PE1CHL.010203:

Added AX25-in-UDP encapsulation, compatible with Linux ax25ipd.
Usage: attach encap ax25u <label> <mtu> <call>
or: attach encap ax25u<udpport> <label> <mtu> <call>
When the default route is set to [0.0.0.0] (param <label> default [0.0.0.0])
all traffic to a call not in the route table is sent to all destinations in
the table.  This can be used to distribute multicast traffic (NODES, QST)

PE1CHL.000113:

Got sick of the ongoing trouble with GDBM and switched back to hosts.net
format in Linux version.

PE1CHL.990808:

Added support for attaching Linux network devices.
Usage: attach netdevice <devname> ax25|ether <label> <mtu> [<call>]
This can be used to attach the SCC driver in newer kernels (2.1+) and
the ethernet controller.

PE1CHL.980919:

"smtp reject" command added to allow the sysop of a 24-hour station to
"reject" smtp mail which matches a certain pattern in "source address",
"destination address", or "sending host".  This can be required when the
mail is destined to a gateway to another network (e.g. Internet), and the
sysop does not want to monitor all mail.  Subcommands are:
smtp reject dest <pattern1> [<pattern2>....]
smtp reject host <pattern1> [<pattern2>....]
smtp reject source <pattern1> [<pattern2>....]
set patterns for destination, host and source fields.  (as appearing
in "smtp list" output).	 The patterns are constructed following the
rules defined in the section about "pattern matching" below.
smtp reject mode [y|n]
enable (y) or disable (n) the smtp reject feature.
Rejected mail is refused with a "571 Mail from/via/to <addr> rejected for
policy reasons" message.  Unfortunately, older versions of NET (and
probably xNOS as well) do not handle a 571 reply to a MAIL FROM command,
and will just close the connection and try again later.  When rejection
upgrade to this version, which will correctly return the mail and advance
to the next queued message.

PE1CHL.980201:

New command "netrom bcdest <interface> nottl <destination>" to turnoff the
transmission of NET/ROM TTL values as an extension in NODES broadcasts.
This extension was defined by TheNET and should not cause problems with
nodes not supporting it, but apparently some NET/ROM clones like XNET get
confused by it.
The default is to transmit the TTL values (as before), but it can now be
turned off by "netrom bcdest <interface> nottl NODES" for the default
destination NODES, or by following a "netrom bcdest <interface> add ..."
command by a similar "nottl" variant.  TTL transmission can be re-enabled
by deleting the bcdest and adding it again.

PE1CHL.971025:

Name of tty lockfiles in UNIX/Linux version now stored in environment
variable NETLOCK.  The default is "/usr/spool/uucp/LCK..%s" but this can
be changed using a "setenv" command (before the attach commands).  Some
distributions use "/var/lock/LCK..%s", others use
"/var/spool/locks/LCK..%s" for the lockfiles.

PE1CHL.970808:

Handling of systemnames like linux.pa3edn fixed in AX.25 mailbox (when
forwarding mail to BBSes it uses < PA3EDN instead of < LINUX)

PE1CHL.970622:

Added support for "param <iface> tx y|n" to many types of interfaces.
(before, it was only supported on scc interfaces)

PE1CHL.970602:

Fixed bug that affected only Pentium systems (problem with SALL 32 bits).
This could affect "default" routes and IP filters using "default".

Some other (minor) bugs fixed.

Memory is given back to operating system ("mem r") automatically every
15 minutes in DOS version.

that are not useful in the AX.25 BBS system are deleted when forwarding
messages to a BBS.  The remaining headers are just passed on.  This
when an SMTP mailer is used to handle BBS messages.

PE1CHL.961231:

New command "arp info <hostid>" prints only ARP table entries for
specified host.

Screen swapping in Linux version now also works in xterm.

line separators)

PE1CHL.961030:

Bugfix in FTP permission checking.  When access was allowed to a
directory using a spec like:
guest * c:\public 3
Access would be allowed to all directories whose name started with
"public", e.g. "c:\public.not".  This could be avoided by using the form
"c:\public\" in the ftpusers file, but now it is checked correctly in
the program.

Made SMTP end-of-message handling RFC-821 compliant.  Older NET versions
and all NOS versions I checked are not compliant, but Linux is.  The
result is that messages with lines starting with '.' are sometimes not
transferred transparantly (an extra '.' may be added at the beginning
of such a line, or one may be dropped).  This can cause 'correctable
errors' in 7PLUS file transfers via SMTP.

Screen swapping added to Linux version (only when running on text
console).  Several small bugs fixed in Linux version.

Some more queue length limits added.  Sending very fast pings to the
node should no longer crash it.

Problem with "param <iface> 5 8" in SCC driver (duplex-LAP mode) fixed.

PE1CHL.960424:

Bugfixes in automatic routing / ARP code.  When a user has routes on
more than one node, invalid ARP entries were made.  It is still
important to make sure each user is routed to one node only.

PE1CHL.960324:

Domain Name Server did not echo the question in the response, which
caused problems with the UNIX "resolver".  Interestingly enough, it
worked OK with "named" and "nslookup" (which I used for testing).  This
problem has now been solved, so that direct use of the DNS from
/etc/resolv.conf now works.

Since 960123 version the agressive parameters on NET/ROM interlinks no
longer worked, because of some internal changes that were not completely
implemented.  Fixed.

PE1CHL.960211:

Allocated memory is now released back to the operating system (as far as
possible) when a "shell" command is executed.  This was present for some
time in the Atari ST version, but is now also implemented for MSDOS.
Memory can also be released using the "mem r" command (and the number of
bytes released is printed).  The number of "bytes available from DOS"
will increase.  This feature sometimes allows "shell" commands to be run
after the program has temporarily used a lot of memory.

The command "mem c <corefile>" dumps variables and allocated memory to a
disk file.  This can be used when reporting bugs that result in a memory
leak.

Fixed a bug that sometimes caused unwarranted "don't connect hop-to-hop"
message.

Added more detail to "cannot open file" messages.  The reason a file
cannot be opened is not necessarily that the file does not exist, yet
many users seem to think that.  Therefore, the error code returned from
the operating system is now printed (as text, when possible).  Error
codes vary depending on the operating system, but could include things
like "No such file", "Not enough memory", "Too many open files",
"Permission denied", "Is a directory" etc.

PE1CHL.960123:

New mode for "route trace":
route trace <host> record
Uses the IP option "record route" to find the path to a host using only
a single packet sent to the host.  This requires the destination host to
be within 9 hops, and also requires the "record route" to be correctly
implemented on the systems on the path to the host.  This works right
with NET and recent versions of Linux, but not with most versions of NOS.

New command "route clear [<bits>]"
Clears the routing table.  Optionally a number of bits can be specified,
and only routes with >= this number of bits are deleted.
"source routes.net".

New mode for NET/ROM interface:
netrom interface <iface> <alias> <quality> downlink
This sets up an interface that can be used as a downlink (to make AX.25
connects from the node) but will not accept uplink connections.  This
can be useful on BBS systems that use the BPQ emulation, and want to
forward on some interface, but do not want to provide access to the
NET/ROM network on that interface.

New command "netrom aexclude" allows exclusion of NET/ROM nodes from the
nodelist based on the alias field, similar to existing "netrom exclude"
command (for call).  The match is case-insensitive.  Example:
netrom aexclude "*box"
netrom aexclude "bbs*"
When a callsign is matched by the "netrom include" list, it is accepted
even when the alias appears in the "netrom aexclude" list.
Note that a node excluded this way is not excluded from connects, it
just won't be placed in the nodelist when a nodebroadcast is received.

FTP server can now send short messages at certain places.  When starting
the server, a filename can be passed like this:
start ftp - - c:/net/welcome.ftp
This file will be sent as soon as the FTP server is connected.
When a user has logged in, and a file "message.ftp" exists in his working
directory, this file is shown.
When a "cd" command is done, and a file "desc.ftp" exists in the new
working directory, that file is shown.
Please keep these files small.  They are read in memory at the time they
are being sent.  A few lines should be sufficient, if necessary the user
Note that lines should be 75 characters or less, because a numeric code
is shown in front of each line (adhering to the FTP protocol spec).

FTP server now times out after 3 days of inactivity while in "command"
state (i.e. no transfer in progress, user not typing any command).
This should reduce the number of left-over FTP sessions after a user
has carelessly switched of his system while logged in.

Modified BBS "SID" to include version number and the "M" capability.
(accept message-ID's on any message, not just bulletins)
message, but NET does not do any duplicate checking itself.

"nodeinfo" server now multicasts link information.  It can be displayed
using "node s <hostname>" on a system where the nodeinfo server is
running.  This information is for future use by an automatic IP router.

file remained open.  So, when doing this a number of times, no more
files could be opened.  This is now fixed.

PE1CHL.951026:

Implemented Domain Name Server.  This can be useful to service local
systems that have a client (resolver), like Linux systems and some
versions of NOS.  NET itself cannot use the server because of technical
details internal to the program.
The server implements the A, PTR, MX and HINFO RR's, and uses information
from the "hosts.net" and "alias" files.  To enable it, add the following
command to autoexec.net:
start dns
When asked for an MX record, the server will check if it can deliver
mail to the requested host, and will reply with its own hostname if it
can.  This is useful when local users want to send mail using this
system as a store-and-forward relay.  When they don't want this, they
should not ask for MX info.  (this is usually a settable option)
HINFO is only available when the system's own hostname is requested.

PE1CHL.950819:

TXDELAY measurement added to SCC driver.  For the first packet in a
transmission, the time between "DCD on" and the start of the packet is
measured, and an average value is kept for a number of callsigns on
each interface.
The command "ax25 txdelay <iface> <numtxdelay>" specifies the number of
callsigns to keep for each interface, default is 0.
Example: ax25 txdelay lap48 23
Using "ax25 txdelay <iface>" a list is displayed, specifying the
measured txdelay, the mean deviation on the measurement, and the number
of measurements (samples) that have been made for each callsign.
The command TXDELAY in the NET/ROM also displays this list.
Stations can optimize their TXDELAY setting by aiming for some low value
of measured TXDELAY, e.g. 50 ms.  After a change, a number of
transmissions should be made to allow the average to be updated.  When
the "Mdev" has a low value, the measured TXDELAY value should be OK.
Note that for some modems it may not be possible to get a low value,
because they give DCD before they are actually ready to demodulate data.

IP filtering added.  A table can be built specifying a combination of
source subnet, destination subnet, incoming interface, and IP protocol
range.  Each IP packet is examined using these table entries, and it is
only routed when it does match an entry in the "include" table or does not
match an entry in the "exclude" table.  This is similar to "netrom include"
and "netrom exclude" as in version 940904.
Tables are built using:
ip exclude <source_subnet> <dest_subnet> [d|<iface>|ANY] [prot [prot]]
ip include <source_subnet> <dest_subnet> [d|<iface>|ANY] [prot [prot]]
The subnets are in the usual "[ip.address]/bits" notation as used for
routes.  The interface can be specified or be ANY.  The protocol numbers
specify a range of protocols to match (or a single protocol when only one
number is given).  Examples are:
ICMP	 1
IGMP	 2
IP-in-IP	 4
TCP		 6
UDP		17
RSPF	73
AX25-in-IP	93
IP-in-IP	94
Note that a table entry only matches when ALL fields match.  So to filter
all traffic from and to a single subnet, 2 entries are needed:
ip exclude <subnet> default
ip exclude default <subnet>

When a route is received (via nodeinfo) to a station in one of our own
"node subnets", an "arp publish" entry for that station (AX.25 type) is
automatically added.  The result is that local stations don't need to set
an explicit route to that station, as the node will reply with its address
and will route the traffic automatically.
It can be a problem when two nodes are on the same frequency, one is the
primary node for the subnet and some routes are directed to the other node.
The "node exclude" command has been extended with an "a" flag to exclude
these automatic ARP updates in this case.  The primary node uses:
node exclude secondary_node 0 a
This causes the arp updates from the secondary node to be ignored, while
the routing info (and other node info) is accepted.  To disable all automatic
node exclude default 0 a

The "arp publish" command no longer requires a hardware address.  When
it is omitted, the address of the interface where an ARP REQUEST is
arp publish sys2.pe1chl ax25
This will cause all ARP REQUESTs for "sys2.pe1chl" to be replied with
the address of the interface where they arrive (this is the normal use
of "arp publish")

Settable timeout on "date get" command: a 4th parameter to the
"date get" command is taken as a timeout in seconds.
Example (using default port and tos):
date get pe1chl - - 30

TCP SYN packets are examined when routed, and the MSS option will be
adjusted down to the maximum MSS possible on the incoming and outgoing
interfaces.  Thus, a more optimal end-to-end MSS is chosen, and
fragmentation is avoided (e.g. when running IP over NET/ROM somewhere
inbetween the endpoints)

PE1CHL.950527:

When a critical error occurs, a message is now printed to alert the
operator to the problem.  Critical errors are usually disk read errors,
and they could cause strange problems because NET sets a critical error
handler that just fails the operation (mainly for unattended use), so
there were only indications like "cannot open hosts.net" when in fact
there was a read error somewhere.
The message has an error code, usually 08 (sector not found) or 0C
(general failure).  See a DOS technical manual for full details.
When multiple errors occur during a single command, only the last one is
printed.

MSDOS version: Expert Mouseclock status printed in the "! !" command.

New bug in REJ frame handling (introduced in 950417) fixed.  From now
on, problems in AX.25 handling occurring only with certain software, and
directly related to bugs in that software, will no longer be worked
around but passed on to the author of the defective software.  Sigh.

PE1CHL.950417:

Unfortunate bug in "ax25 lqdump" fixed.  This bug makes dumps made with
older versions almost useless.  (valid data was dumped only on first dump
and on dumps made after 4 days of uptime)

Nasty bug in IP Multicast routing fixed.  (affected only MSDOS version)
Problem with multi-fragment IP Multicast datagrams fixed.

To overcome difficulties when editing the NETATTRxxx strings with
certain editors that can't handle the ESC character, these strings can
now include some special sequences like those used in the MSDOS PROMPT
command.  e.g., $E is used for ESC. Note that$ normally indicates a variable in NET, except when the string
is enclosed in single quotes (').  Therefore, to use this method the
setenv commands must be like this:
setenv NETATTRCMD '$E[0;33;1m' Processing of REJ frames in AX.25 (interlink mode) again modified, to work around a problem in some other software. When more than one REJ frame is received from the other side, no more data is sent until T1 elapses. This avoids a permanent loop where the same I frame is sent several times in a single transmission, and several REJ frames are received as a reply, causing the retransmission of the same I frame. NET itself never transmits multiple REJ frames as a response, so this problem only occurred when talking to nodes running other software. PE1CHL.950326: Bug in NETSTATUSLINE fixed (this caused a crash when new connections were made on a session that was previously connected). Multicast tunnels are monitored (using ping) to check if the delay is not too long for proper Multicast protocol operation. If so, they are temporarily disabled until the delay is OK again. The "ip multicast tunnel" command displays the status (Enabled/Disabled) and the reason for disabling: 1=too-long RTT observed recently, 2=average RTT too long, 3=no ping replies received, 4=last ping reply too long ago. This change makes the use of slow AX.25 links for multicast tunnels safer. Note it is still better to use only reasonably fast links for these tunnels, and to avoid the use of tunnels as much as possible. Slow tunnels cause duplicated multicast packets in the network. Received nodeinfo is now checked against NET/ROM "exclude" and "include" lists, to enable lockout of non-network stations. Additionally, "node exclude" and "node include" commands have been added to control the nodes accepted into the nodeinfo list. These commands work like the NET/ROM "exclude" and "include", but they accept an IP subnet and "node scope" as selection pattern. Additionally, a flag specifies if the entry filters all node info (i), or only the route info (r). See also version 950618 for another flag. Examples: node exclude [44.137.0.0]/16 3 all node info from nodes in this subnet, with scope 3 or lower is excluded node include [44.137.40.0]/24 0 i include nodes in this subnet (scope 0 or higher), but only accept the info (not the routes) The filtering can be used to limit the spread of low-scope nodes that are not of local significance. This is done by excluding all nodes below a certain scope, then including all nodes in nearby subnets. Checks for overlapping "subnet" in node info is made. When an overlap is detected, a warning message is printed (each time the offending node info is received). Please coordinate the claims of subnets with other node sysops: each subnet should be claimed by one node only. MSDOS version: support for "Expert Mouseclock" added. When the driver for this clock (a DCF77 receiver connected to a COM port) is detected, the timezone setting is changed automatically depending on the daylight savings time information bits in the received signal. Thus it is no longer necessary to update the timezone setting twice per year. PE1CHL.950316: IGMP implementation completed. When "ip multicast routing on", an IGMP Host Membership Query will be sent with 1800 second interval. Any system wanting to receive Multicast packets will respond with a Report of the address(es) it wants to receive. When another system is heard reporting, the report is canceled. A Multicast address is requested using "ip mcastgroup <iface> <address>", and can be deleted by a "d" after the address. This is all still experimental and mainly useful for my debugging... On Atari ST and MSDOS, the "/o" option of "dir" requests an unsorted directory (as it appears on disk). This can be useful to evaluate the result of "INFO cmdname" and "HELP cmdname" commands in the NET/ROM, as the first file found in the directory that matches the cmdname is shown. Some bugs in the "nodeinfo" server have been fixed. "netrom bcnodes <interface> poll" command sends a modified NODES broadcast that will be recognized by neighboring nodes (running NET or NOS) and instruct them to send their NODES broadcast soon. Using this in the autoexec.net will ensure quick updating of the NET/ROM tables after a restart. "netrom downlink" command now also accepts "circuit" as an interface name, indicating a hardwired connection via a NET/ROM circuit. This is useful when a BBS is attached to a node using a NET/ROM link, and the BBS callsign is different from the callsign of the node associated with the BBS. (which is often the case in the Netherlands) Example: netrom downlink pi8bbs circuit pi1bbs When a NET/ROM user connects PI8BBS, the node will setup a NET/ROM circuit to PI1BBS, and send a "C PI8BBS" there. SCC driver %DCD and %RTS are now kept in 24 separate counters, selected by the local time (hour). The existing "scc b" command still prints the overall percentages, but details are shown when a single channel is reported using "scc b<num>". Example: scc b2 Results gathered by "ax25 lq" can be dumped to a file for later analysis. The commands are: ax25 lqdump <filename> dump link quality data to specified file. when - is specified for the filename and the scheduled dump (see below) is active, the filename given for the scheduled dump is used. ax25 lqdump <filename> <interval> arranges for the link quality data to be dumped every <interval> hours. Example: ax25 lqdump c:\public\linkqual.dat 24 in onexit.net: ax25 lqdump - It is a good idea to dump the link quality data to a file in the public FTP area, so others can download and analyse it. When the file becomes large, rename it to another name, e.g. "c:\public\linkqual.old". The new data will then be written to a newly created file, and the old file can be removed after it has been downloaded. A program to analyse the data still has to be written... the file format is available on request. More screen attribute settings strings introduced, to allow customization of colors (mainly useful in MSDOS version). There is even a simple status line capability. The following environment variables can be set to change screen colors: NETATTRCMD defines attributes for the "net>" prompt NETATTRECHO defines attributes for character echo NETATTREXIT sent before exiting to restore default colors NETATTRMESG defines attributes of asynchronous messages NETATTROUTPUT defines attributes of command screen output NETATTRSES defines attributes of output on session screens NETATTRTRACE defines attributes of trace output NETCLS sent to clear the screen (using current attributes) NETSTATUSLINE defines position and attributes of status line These variables are normally set using "setenv" commands in the configuration file, and should contain ANSI escape sequences that are interpreted by the ANSI.SYS device driver (or equivalent, like NNANSI). Because the values are not yet known at the time NET prints its startup messages, these always appear in the screen colors active at that time. The command screen can be cleared and set to the NETATTROUTPUT colors by using the new command "cls", when NETCLS is set to the sequence "^[[2J". The STATUSLINE has some information about the active sessions. Each session is shown like "1: PI8NOS Con RTU", where 1: is the session number (1* when it is the current session), PI8NOS is the call of the connected station, 'Con' is the status (Connected), R indicates data is in the receive queue, T indicates data is in the transmit queue, and U indicates an "upload" is still in progress. Note that the statusline can interfere with some screen output. It can be turned off by simply not setting the NETSTATUSLINE variable in the configuration file. In the Linux version, a "dbm" version of the hosts.net, called hosts.db by default, is used to lookup the hostnames. This has been introduced in some earlier version, but was mistakenly not documented. The hosts.db file allows a much quicker lookup of hostnames and IP addresses, because index information is present in the file. Thus, NET does not need to do a sequential read of the entire file for each lookup. When the hosts.db file has not yet been created, an error message is displayed each time a hostname has to be looked up. When this happens, use the "host import" command to create a hosts.db, and then exit and restart NET so that it has the file available from the beginning. The hosts.db file is maintained and accessed using a few subcommands of the "host" command (which has no subcommands in the DOS and Atari version). The "host" subcommands are: host closedb close the hosts.db file immediately (it is automatically closed when no lookups have been done for 30 seconds) host comment <hostname>|[<IP_address>] shows the comment info for the specified hostname or IP address, when it is available in hosts.db. e.g.: host comment pe1chl pe1chl: rob host export <filename> convert the data in hosts.db back to hosts.net form, and write it in the specified file (unsorted) host import <filename> [nocomment] loads a file (in the hosts.net format) into hosts.db, creating it when necessary. for each line in the file, a record for translation from hostname to IP address and IP address to hostname is added. furthermore, the comments (after a # character) are stored unless the "nocomment" option is specified (which saves space). host lookup <hostname>|[<IP_address>] lookup a hostname and print its IP address, or lookup an IP address and print the corresponding hostname. this is the same function as the "host" command (without subcommands) performs in the DOS and Atari versions. host reorg performs a reorganization of the hosts.db, e.g. after a lot of changes have been made. this reads the existing hosts.db and writes all records one-by-one to a new file. may take some time. host update <filename> [nocomment] takes a DIFF file between the hosts.net as it was previously loaded using "host import", and the new version. the lines in this DIFF are processed to modify the hosts.db: lines starting with > are treated as additions, lines starting with < are treated as deletions from the file. when the "nocomment" flag is present, comments in the DIFF are not loaded in hosts.db. this command can be used to process the "host update" bulletins that are sent through the network by the address coordinator. PE1CHL.950119: Corrected problem with "record silent". Should now also work with FTP and Finger. Extra checks in packet driver and memory allocator in an attempt to further stabilize it. SCC driver verifies the SCC chip presence. A warning message is printed when some test values cannot be written to and read back from the chip. When a channel fails the test, it can be attached but a messages shows up when "sccstat <channel#>" is used. PE1CHL.950115: Because of the increased codesize and the resulting memory problems in the MSDOS environment, the set of available versions has been modified a bit. A version for MSDOS with both PACSAT and NET/ROM code is no longer available, as this was not very useful anyway. The following versions are now available: Vers Features AtariST DOS86 DOS286 DOS386 Linux ------+----------------------------+-------+-----+------+------+----- AX25 | Only AX.25 protocols, no IP| | ax | | | IP | Only IP (SLIP/ethernet) | | ip | | | | | | | | | User | All except NET/ROM, PACSAT | | ut | u2 | | Sat | Includes PACSAT protocols | ut | | | u3 | net | | | | | | Node | NET/ROM, limited userinterf| nt | | n2 | n3 | Max | NET/ROM, full userinterface| lt | | | l3 | | | | | | | Added AX.25-over-ethernet capability compatible with G8BPQ node. This can be used to operate a node running NET together with one or more machines running applications (BBS, DXcluster) using BPQCODE on the application machines. Setup BPQCODE for ethernet usage using "odidrv" according to the supplied documentation. NET supports only Ethernet_II frame type with PID=08FF, so you have to configure BPQ like that (the default). Install a "packet driver" for the ethernet card in the machine running NET, and attach it using: attach packet <int#> g8bpq <label> <buffers> <mtu> <call> [<enet>] example: attach packet 60 g8bpq bpq 8 256 pe1chl-10 ff:ff:ff:ff:ff:ff This provides you with an AX.25-type interface that sends its packets on the ethernet in a format that G8BPQ will understand. To use it, you will have to register it as a NET/ROM interface in the usual way. The ethernet address is optional, and defaults to the multicast address 01:42:50:51:00:00 like G8BPQ does. You can enter the broadcast address ff:ff:ff:ff:ff:ff here, or the address of another card (when you use only two systems on the network). Note that there can be problems that may force you to use the broadcast address or a specific card address instead of the default address. NET will try to set the receive mode of the packet driver according to the address specified (or defaulted): when it is a multicast address, it will try to select receive mode 5 ("receive all multicast"). This can only be done when no other protocol uses the packet driver yet. Therefore, when you want to attach the ethernetcard for IP use as well, put the "attach packet ... g8bpq" command first. When you want to run other protocols that are started before NET (such as Novell Netware), you may want to use the "pktmode" program supplied with the packet drivers to select mode 5 before loading the other protocol. When this cannot be done and/or the packet driver you use does not support mode 5, you will have to use the broadcast address. This has to be configured in BPQCODE as well. NET does not directly support ODI drivers. When you want to run Netware on the same card, use ODIPKT to simulate a packet driver on top of ODI, or use PDETHER to provide an ODI interface on top of a packet driver. NET/ROM "INFO" (aka "HELP") command can now send information about specified topic. To activate this feature, add a second argument to the "netrom info" command, to specify a directory where information files are stored. It is possible to use the same directory as used by the "finger" server. NET looks for files with the extension .txt in this directory, and displays the first file found which matches the keyword entered by the user. Example: netrom info c:\net\nr_info.txt c:\net\finger c:\net\finger directory contains: connect.txt mheard.txt nodes.txt routes.txt ... etc Now a user at the node enters: H CON The file "connect.txt" is then sent to this user. Any substring of CONNECT (C CO CON CONN CONNE CONNEC CONNECT) will result in this same file being displayed. As the NET/ROM user can only ask for helpfiles by name, and cannot get a listing of available files, it is the responsibility of the sysop to include proper cross-references between the files, starting from the file that is displayed by a single HELP or INFO command (in this example: c:\net\nr_info.txt). This could take the form of directives like "Use HELP <commandname> to get more help" or "Use INFO LINKS to get a description of the available links". Of course, when the directory is shared with the "finger" server, a TCP/IP user can get a listing of files by using "finger @hostname". IP Multicast routing has been implemented. This allows the efficient distribution of datagrams through a network of IP nodes, by sending them to a Multicast IP address (an address in the range 224.0.0.0 to 239.255.255.255). In general, multicast datagrams are only forwarded between network nodes. However, applications running in user stations can request forwarding of certain multicast IP addresses on the local access frequency, using the IGMP protocol. To setup a node for IP multicasting, some study of the links to neighboring nodes and other participating nodes is required. When all nodes in the network run multicasting-capable software and are directly connected, each node only has to set for each interface to a neighboring node: ip multicast interface <name> forward and for each local-access frequency: ip multicast interface <name> all igmp This results in packets being forwarded on the interlinks when they do not arrive on that same interface. On the local access, all packets are forwarded, but subject to requesting via the IGMP protocol. When one single interface is used for a link to more than one neighbor, the command for that interface should be: ip multicast interface <name> all This results in all packets being forwarded, also those that were heard on that same interface, so that neighbor #2 can hear them when they were received from neighbor #1 on that same interface. The setup becomes more complicated when multicast datagrams need to be sent to stations that are not immediate neighbors, e.g. because the immediate neighbor runs software that is not multicast-capable. In this case, a "tunnel" needs to be setup to direct the multicast packets through this gap. Because of the nature of multicasting, the tunnel should be a quickly operating path, preferably not a VC-oriented path like a NET/ROM route. A tunnel is setup using: ip multicast tunnel <destination_IP> <mode> The <mode> is selected from this list: dir - direct transmission to specified IP addr (one hop only) ip - IP-in-IP encapsulation sr - source routing off - delete this entry "dir" is the preferred mode, but it can only be used to help the multicast over a single IP routing hop, using a fixed ARP entry which can contain a number of AX.25 digipeaters. When it cannot be used "sr" can be tried, but it is known to be buggy in many versions of NET and NOS. In that case, use "ip". It is less efficient, but should always work. Example: ip multicast tunnel pi8daz dir Finally, to enable multicast routing, include the command: ip multicast routing on User stations that want to send multicast datagrams but do not need to route them, should not enable multicast routing but should use an entry in the IP routing table: route add multicast <interface> or: route add [224.0.0.0]/4 <interface> A "node information" service has been implemented, to distribute some information about network nodes. Users can request this information on other nodes using "NODES <callsign>". The information is also used internally in the software to help routing. This service makes use of the IP Multicasting protocol, so that has to be configured to make use of this feature. The IP Multicast address 224.0.1.31 is used, with UDP port number 1535. These numbers have been officially registered at IANA. The new command "node" has a number of subcommands to define some information about the node: node ittl <#hops to forward information> node lapinfo "Some information about LAP frequencies and baudrates" node locator XX11YY node neighbors <names of neighbor nodes>... node organization "Name of supporting organization or club" node qth "QTH of node" node rttl <#hops to forward routes> node scope <coverage_area> (0=test 1=local 2=regional 3=interregional 4=national 5=continental 6=worldwide) node subnets local_subnet/bits ... node sysop "Name and/or callsign of sysop(s)" The information strings have a maximum length of 60 characters. They should be kept brief, to avoid very long information packets that have to be fragmented before being sent. When a certain entry (e.g. "node organization") is not applicable, just omit the command. "node ittl" determines the number of hops that the node information is multicast. Setting a reasonably low number here can limit the spread of node information through the network. "node rttl" determines the number of hops that the routing information will travel, and should at least be enough to reach all nodes whose routing decisions are dependent on information in the local routing table. It is not useful to use a higher number than "node ittl". The "node neighbors" should list the IP hostnames or addresses of those neighboring nodes that also route IP (e.g. those running NET or NOS). "node scope" defines the relative importance of the node, and is used to determine the spread of information through the network. Nodes of local importance do not need to be spread all across the continent, and this level can be used to tune that spread. "node subnets" defines the IP subnets that the node routes to local access points, and that are supposed to be routed to this node by other nodes in the network. Only a single node should claim each subnet. (when multiple nodes serve the same subnet, only a single node should list the subnet in this command, the others should only have a set of routes to the stations using that node) Example: node ittl 8 node lapinfo "430.750 MHz, 1200/4800 bps" node locator JO22NF node neighbors pi8ehv pi8esa pi8rni pi8vrz pi8wfl node organization "Scoop Hobbyfonds" node qth "Hilversum" node rttl 4 node scope 4 node subnets PA_NH2net/22 PA_FLnet/22 node sysop "PE1CHL, PE1CRC" To start the node information server, the following command is used near the end of autoexec.net: start nodeinfo The command "node status" shows information about all nodes known to the information server. To get more detail about a single node, use "node status <hostname>". Included in the node information being multicast, is a list of routes to IP stations that have been locally defined to other gateway addresses than the neighbors, and are not in the local subnets. This route information is inserted as a volatile route in the routing tables of the nearby nodes, when necessary to route the traffic for those IP stations. E.g. a station in the PA_GLnet network wants to have his IP traffic routed via the node in the PA_UTnet network, because he has better propagation to that node. He requests the sysop of the PA_UTnet node to add a route to his IP address on one of the local access interfaces. The PA_UTnet node will then send this information as part of the node info, and all surrounding nodes will add a route to that address to go via the PA_UTnet node. No manual intervention is required on other nodes. To get this working: - the metric of the route must be below 8 - the route must not be via a pseudo interface (netrom, encap) - the IP address must not be multicast, loopback, etc. - the route must not point to a neighbor node - the IP address must not be in one of the local subnets - our subnets must not be a subnet of the route Automatic routes are only accepted in a node when it has set the "route obsotimer". Recommended value is: route obsotimer 600 Note that the automatic distribution of route info can make a station unreachable when incorrect routing info is present in the tables. Please check the routing table and make sure no out-of-date or otherwise invalid entries are present. When you have routing table entries that you do not want to be sent to other nodes, use a metric of 8 or more. Example: route add <hostname> <interface> <gateway> 8 route add <hostname> <interface> "" 8 Recording of incoming data in telnet and AX.25 sessions can now be done without displaying all of the data. When the command: record <filename> silent is used, recording will continue but data won't be displayed when the session is not current (on screen). To see if data is still coming in, select the session, and it will be displayed as usual. This feature can be useful to record a large number of messages being read on a BBS, while making other QSO's on different sessions. Similar to the "silent" option of "record", the "upload" command now has a "verbose" option that causes all data to be displayed while it is being uploaded. This can make it easier to notice when uploading is finished. SCC driver parameters (in AX.25 mode) that are expressed in 'ticks' (params 1,3,4 and 7) can now also be specified in milliseconds by adding the 'ms' suffix, as in: param 144 1 300ms This allows more convenient specification of the parameters, and makes them independent from the actual tick rate, which can have several values in the MSDOS environment. The actual value of the parameter, as reported back when using "param <interface>", will be the value in tick time units as before. Note that this function is NOT available for async interfaces (to KISS TNCs), only for the SCC driver. SCC driver now supports up to 14 chips on the PC. (28 channels) "packet driver" buffer allocation scheme has been modified, to pre-allocate enough buffers at attach time, and using these buffers when data is coming in (which is in an interrupt handler). Advantage is that interrupts do not need to be disabled during buffer allocation, which should result in less overruns on SCC channels. Disadvantage is that a buffers*MTU memory area is permanently allocated for this, and the data needs to be copied out of this buffer in the main loop, causing more CPU overhead. The old method can still be selected by specifying a negative <buffers> in the "attach packet" call, e.g. -8 for 8 buffers using the old method. The port number for "rcmd" has now been officially registered at the IANA, and is now 459. Please change all references to port number 333 in your autoexec.net files from 333 to 459. ("start rcmd" command) The "hostname" command can now optionally specify the domain name. When this is done, the lookup of hosts (e.g. when sending mail) is speeded up because the domain name will be stripped off before looking in hosts.net, thus making multiple passes through this file unnecessary. Note that earlier versions just stripped off the domain suffixes one by one, and scanned the hosts.net file each time. This took a lot of time, but it meant that mail with mistyped addresses could often be delivered to the correct host. Now, such mail will bounce back to the sender. Usage of the "hostname" command: hostname <full.domainized.host.name> <domain.name> e.g.: hostname sys2.pe1chl.ampr.org ampr.org hostname$CALLSIGN.ampr.org ampr.org

PE1CHL.940904:

Added some of the TheNET extensions to the NET/ROM code.  The TTL values
in TheNET node broadcasts are now supported, and the uplink node is
shown in the "users" command when it is known (connect from a TheNET
node).  The uplink node is not sent by NET, because this conflicts with
another extension (sending of L4 timeout in connect requests as
specified by G8BPQ).

"netrom exclude" command has been changed to allow "patterns" instead of
only straight callsigns.  The patterns used in this matching are the
same as for the NET/ROM NODES command, see the history entry for version
"PE1CHL.930808".  This allows things like "netrom exclude p[a-e]*" to
exclude any callsign in the PA..PE range from the netrom node list.
A new command "netrom include" allows exceptions from this rule.  A node
is only excluded from the NET/ROM node list when it matches none of the
patterns in the "netrom include" list, AND it matches at least one of
the patterns in the "netrom exclude" list.
The new "netrom exclude" command is not entirely compatible with the
original version, as the patterns must explicitly match the node
callsigns including SSID.  To convert existing "netrom exclude" commands
existing commands.  This will match any SSID in that position.
Try to keep these lists as short as possible, as the processing of these
generic patterns takes considerably more CPU time than the literal
matches that were done in previous versions.  This processing is done

"ROUTES *" command in NET/ROM command interpreter displays AX.25 link
quality of neighboring nodes, when available.

PE1CHL.940720:

Added "route merge" command to merge route table entries. (remove
redundant routes, combine routes to adjacent subnets)
Add a "route merge" at the end of autoexec.net (or at least after the
last "route add" command) to optimize the routing table.

"route" command reports usage count for each entry, so that unused
routes can be found and evaluated.

Txq reported for AX.25 sessions in the "session" command output now is
in bytes, just like the other session types (was in packets).  In the
"ax25 status" command, both bytes and number of packets are reported.

"smtp expiry" now sets shorter timeout when the mail is already on the
system for some time (e.g. when previous attempt ended because of a
system crash).  Now mail even gets returned when the program does not
keep running for "smtp expiry" days.  The minimum timeout is 25 hours.

Added some subcommands to "ftp", like those available in the UNIX ftp
command:
ascii		same as "type a"
binary		same as "type i"
mput		like "tput", but does not send the contents of
subdirectories
reget		same as "get" with "restart" option

Added "SYST", "SIZE" and "MDTM" commands to FTP server (mainly for UNIX
compatability, the "ftp" command does not use them)

NET/ROM interface specification again made more versatile.  It is now
possible to specify an interface as "bbs", which indicates a private
link to a BBS or DXCLUSTER.  Such an interface will only broadcast the
callsign(s) of the node itself, not callsigns learned from other nodes.
This is useful when a BBS or DXCLUSTER is attached using a NET/ROM link
(e.g. BPQCODE) but the attached system should not receive all the
network nodes.
Also, it is now possible to limit the number of uplinks via the same
digipeater (similar to the limitation on links from the same user introduced
in 930808, see below), and the maximum number of digipeaters used to
access the node on each interface.
Syntax:
netrom interface <iface> <alias> <quality>
Example:
netrom interface lap4k8 ALIAS 100 lap 4 2 1
This command defines a NET/ROM interface "lap4k8" which acts as the default
local access point interface, accepts 4 connects from the same user, accepts
2 connects from users coming in via the same digipeater, and accepts only
a single digipeater between the user and the node.

On interfaces where "link quality measurement" (ax25 lq) is running, it is
now possible to switch automatically between "datagram" and "virtual circuit"
mode of IP encapsulation.  This is selected by adding the "auto" keyword
after the mode command:
mode 144 vc auto
mode 430 datagram auto
The automatic selection will select datagram mode when the link quality is
above a certain threshold, which depends on the length of the packet.
Virtual circuit mode is automatically selected when the link quality drops
below this threshold.  The mode explicitly selected in the "mode" command
(vc or datagram) will be used when no link quality figure is available for
the destination of the packet.

IP multicasting support is being implemented.  This is not finished yet.

Numerous small bugfixes and enhanced checks on some parameter values.

PE1CHL.930914:

side to find the percentage of lost frames.
The "ax25 lq [<interface>]" command prints a table of callsigns, and the
statistics gathered.  Each AX.25 link quality frame also contains the
locally stored statistics for up to 10 stations, so that the "ax25 lq"
command can independently report the link quality for both directions of
The first (or only) line for each callsign shows the percentage of
frames received locally, the second line (if present) shows the
statistics received from the other side (i.e. the percentage of our
frames that were received at the other side)
The LastRX field shows the time since the last update of the
information.

Quality shown in NET/ROM "routes" command output corrected.  (was wrong
for a permanent route or a route via digipeaters)

NET/ROM "maxfail" now correctly handled when permanent routes are present.

Gated timer handling made more sophisticated.

PE1CHL.930808:

Added (L)ist command to AX.25 mailbox, allowing users to list their own
messages (a message which has the user's call appearing somewhere in the
output line from "smtp list").  Allows users to check if they have mail
pending at a local SMTP server, and to check if their mail has already
been delivered.

Added command "arp proxy [<iface> ...]" to enable PROXY ARP on one or
more interfaces.  PROXY ARP means an ARP request for a destination is
destination is known and is not on the same interface as where the
request came in.
This feature is mostly useful on an ethernet segment where UNIX machines
are present that can not route packets for the same network address to
different destination machines.  It should probably not be used on AX.25
interfaces (or only with great care)!

New command "route trace <host> [udp|echo [<tos>]]" for tracing IP
routes (like "hop check" in some NOS versions).  Uses UDP datagrams or
ICMP ECHO requests with increasing TTL to find the route to a specified
host.  Returned ICMP "TTL EXCEEDED" messages are used to find the gateways.
The "route trace" commands starts a session to report the output.  When
re-display the list in sorted order, when out-of-sequence info has been
printed.  The following commands can be entered while in the "route
trace" session:
q	quit (close session)
The "route trace" procedure ends when a reply is received from the
specified host, or some "unreachable" message has been returned.  The
state of the session then changes from "Active" to "Complete".  Note
that the final reply may be "Unreachable Port" in UDP mode (when the
host is not running the UDP echo server), or the host may not reply at
all.  Also note that some hosts may be unwilling to return ICMP messages
on ICMP ECHO requests, and others may not allow the routing of packets
for the UDP protocol, so experimentation is required.
When the situation changes during the session, or when gateways internally
modify TTL during ARP resolution, "route trace" gets terribly confused...

New command "smtp post" defines a command to be executed after new mail
has arrived.  Each new mail triggers a 5-minute timer, and the command
is only executed when this timer elapses.  So, more than one mail could
have arrived by then.

Added IP-in-IP encapsulation.  Encapsulation means that an entire IP
datagram is (again) sent as an IP datagram, i.e.  another IP header is
added in front of it, and it is sent to some other IP host for further
routing.  This allows (temporary) fixing of routing problems, and/or
transport of datagrams through networks that do not route net-44
traffic.
Example of a case where IP-in-IP is useful:
A --->- B --->- C --->- D
Assume, A wants to route IP traffic to D through the gateways B and C.
Normally, that means that each gateway must have set up an entry in its
route table to forward traffic to D to the next system.  Now, assume
that for some reason B does not know that traffic to D has to be routed
via C, but it does correctly route traffic to system C, and C does route
traffic to D correctly.  It may be that B is a system which does not
understand the subnetting used in net-44, and can therefore not route
traffic for different net-44 hosts via different gateways. Encapsulation
can be used in this case to send datagrams for D encapsulated in
datagrams for system C.  System B will route the datagrams as if they
are for C, and C will see that it is an encapsulated IP datagram and
will route the original content to D.
To set this up, the following commands are necessary at A:
attach encap ip ipip 512
This will attach an encapsulation 'pseudo-interface' called "ipip" (for
IP-in-IP encapsulation), and will route all traffic for D to this
interface with C specified as a gateway.  NET will encapsulate all
traffic for D in IP datagrams sent to C.  C will route the original
datagram to the original destination D according to its local routing
table.  A single encapsulation pseudo-interface can be used for all
encapsulation routes you may want to define (unless you want to define
different values for MTU, in which case you can attach multiple
encapsulation interfaces) Take care that encapsulation does not create
routing loops, as these can cause a lot of traffic bouncing back and
forth, and are difficult to break.

Added AX.25-in-IP encapsulation.  Like with the IP-in-IP encapsulation
described above, an entire AX.25 packet is sent as an IP datagram.  The
AX.25-in-IP variant allows AX.25 traffic over local area networks for
which no AX.25 drivers are available, e.g. SLIP or ethernet.
To use this encapsulation, an encapsulation 'pseudo-interface' is
attached which will serve as the interface for all AX.25 operations.
Any AX.25 packet sent to this interface will be sent to an IP host
determined by checking a special routing table.  The following commands
are used:
attach encap ax25 axip 256 <callsign>
param axip <callsign> d			(drop a route)
param axip default <ip-address>		(set default route)
After attach, only a default route to [255.255.255.255] is in the table.
Using only this route should be adequate for use on a shack ethernet (at
least when packets to [255.255.255.255] are routed to this network).
As the routing table entries are SSID-dependant, you may have to enter
several routes for the same station when different SSIDs have to be
reachable via the encapsulation interface.

MSDOS version: Added support for "block transfer" functions in the
FOSSIL and MBBIOS interface routines for "attach com".  This should be
more efficient when using high speed on async ports.  Because of a bug
in MBBIOS this is currently only used for receive with MBBIOS.

Async speeds "57600" and "115200" now supported in Linux version.
Line locking fixed so that it deletes stale lockfiles.

Improved efficiency of "broadcast" protocol hole filling, by randomizing
the request packets sent.  Before (and in PB) the requests were always
sent in byte-offset order, so that everyone is always filling the file
from the lowest address.  After a while you receive a lot of duplicate
Now, the requests are sent in several different orders, selected by a
random generator.

New command "broadcast post [<command>]" runs a command whenever the
dirfile.dbp file is closed (after two minutes without receiving broadcast
packets).  This command can start a program to compact the dirfile.dbp

New command "broadcast request <server> <file_id> p" requests only the
PFH of the file, e.g. when you see a large file in the status listing
and want to check what it is.  Because only a few bytes are requested
you will always be put at the front of the queue, and get your PFH
updated quickly.

shown in the "ftl0 status" command, and the final figure is printed and
logged when the upload completes.  Note that this figure only reflects
the performance during the last connect to the server, when multiple
connects were necessary to upload the entire file.

G8BPQ hostmode emulation improved.  New (930808) emulator has to be used
with this version!
- downlinks from the NET/ROM to the G8BPQ emulator are now properly made
as a Host(# call) link, instead of via a local loopback AX.25 connect.
so, the Hostmode application can now get the correct paclen value for the
connection to the remote station.  when an F6FBB BBS is integrated with
a NET/ROM node, the paclen can be set to 0 in PORT.SYS, and the BBS will
- finally figured how DEDHOST tells BPQCODE that keep-alive packets have
to be sent (AUTOTIMER flag).  This is undocumented in the G8BPQ Hostmode
interface specification...
now, NET sends keepalives (2 NUL characters) when an incoming connection
to a DX-cluster is via NET/ROM, and disables the idle-disconnect timer
when it is an AX.25 connection.

NET/ROM accepts command "*** LINKED to <callsign>", but only when issued
as the first command on a connection from the G8BPQ Hostmode emulator.
This is used by G8BPQ Hostmode applications that want to setup outbound
connects using a different callsign than the default (e.g. F6FBB gateway)
Other commands starting with *** are ignored by the NET/ROM command
interpreter. (this seems to be a convention for passing status
information, F6FBB also ignores it)

NET/ROM "NODES" command can now selectively display nodes using a pattern
match.  The match is only applied to the node callsigns, not the aliases.
E.g:	nodes pi8*	show only nodes in the Netherlands
nodes *hwb	show only nodes whose call ends in "hwb"
The following constructs are valid in this pattern:
*	matches any sequence of characters
?	matches any character
[abc]	matches a, b or c
[a-l]	matches any character in range a..l
[^abc]	matches any character except a, b or c
[^a-l]	matches any character except those in range a..l
(this is NOT the same pattern match syntax as used in some other commands!)

NET/ROM "NODES <nodecall>" command (and NET "netrom route info") command
show smoothed round trip time and mean deviation for the specified node,
when this info is available.  (only when connect and/or info have been
exchanged with the specified node after it has been added to the nodelist)
The round trip time information is also used to calculate the NET/ROM
transport timeout when setting up new circuits.

NET/ROM "CONNECT" command now handles digipeaters the same way as a fast
connect on the uplink is handled.  So, a user can connect a station on
another node using:
CONNECT <user> VIA <node>
Before, this resulted in an AX.25 downlink connect of <user>, with <node>
as a digipeater.
Also, it is now possible to specify more than 2 digipeaters in both a
fast connect and a CONNECT command to the node.  The remaining digipeaters
will be sent in the CONNECT command to the next node, so it should be possible
to connect through multiple nodes (when they run this version of the software,
or other software that understands "C <call> VIA <digi>...").  example:
CONNECT <user> VIA <entrynode> <intermediatenode> <exitnode>

NET/ROM now detects hop-by-hop connecters, and issues a warning message
when a user has connected from a neighbor node and tries to setup a
circuit to another neighbor node.  It then reduces to NET/ROM window to
decrease the troughput on this circuit.  Users should be told to connect
from uplink to downlink node in one step (or as few steps as possible),
as this enables more efficient routing and reduces processing in each node.

Again improved NET/ROM routing by optimizing broadcast packets sent to
specific destinations, so that more alternative routes are stored.
NET/ROM broadcast is sent soon after one is received from a neighbor
which was not in the tables (to speedup recovery after a crash).  Also,
updating of alias fields more carefully checked to prevent errors when
nodes running other software have locked node entries with wrong alias
names.  (alias names cannot be locked in NET, not even for fixed entries)

NET/ROM can limit uplinks by a single user, to avoid over-use by some
people.  The default limit is 3, it can be changed by an aditional
parameter to "netrom interface":
Of course this has only meaning when "lap" or "uplink" is present.  It
is suggested to set a low value (e.g. 2) on 1200 bps LAP, and a higher
value on a highspeed LAP interface.
Please note that you should NOT include "uplink" or "lap" on an interface
which is to be used for a NODE-to-NODE interlink!  The efficiency of the

MSDOS and Atari version: new command "grep <pattern> <file>..." to scan
for a pattern in one or more files.  See entry for PE1CHL.920701 for
description of pattern syntax.  It is NOT the same as for the NODES command!

MSDOS version compiled for 386/486 now available.  This saves memory and
should run more efficient by using 32-bit registers when appropriate.
This version identifies as "BC.L386" and will immediately terminate with
a message "386+ Required" when run on a 8088 or 80286.  So be careful when
remotely updating a node with such a processor: load the right version!

PE1CHL.930430:

This is a minor 'bugfix' release.  It fixes problems in the packet
driver interface, the NET/ROM command interpreter, and "arp publish".
The memory management in the Atari ST version has been fixed to avoid
problems with TOS 2.06.
Also, a new G8BPQ Hostmode emulator is included with this release.  It
no longer supports the sending of *** CONNECTED messages, because it was
unclear when these messages had to be sent.  (apparently I don't
understand the G8BPQ Hostmode interface specs...)

PE1CHL.930411:

AX.25 queue length for "virtual circuit" mode is now limited.  IP Packets
beyond a certain limit are discarded, and a "Source Quench" is sent
back.  This should prevent allocation of too much memory for a bad link.
because these make matters worse when the link is bad in one direction!
(ping replies pile up at the remote end, and normal traffic is slowed
down even more than it already was...)

G8BPQ emulation now also includes UI frames.  This means MAIL beacons
can now be sent, and the application can monitor UI frames as well.
The "g8bpq" command now has 3 subcommands:
g8bpq status: like the old "g8bpq" command, shows the status of all
streams.  A range of streams can be shown by specifying
first (and optionally last) stream number.
g8bpq active: like "g8bpq status", but shows only those streams for
which an application is or has been active.
g8bpq ports:  setup the mapping between G8BPQ radio port number, and
the interface names normally used by NET.  this is
necessary because UI frames are normally sent to
"radio ports" by G8BPQ applications, where these ports
are known by number.  The NET interface names supplied
to the "g8bpq ports" command get the BPQ port numbers
1,2,3 etc.  E.g.: "g8bpq ports 144 430" will setup
BPQ radio port #1 = NET interface "144", and
BPQ radio port #2 = NET interface "430".

PC version now compiled with Borland C 3.1.  This results in more
efficient, but slightly larger code.

Memory allocation routines revised.  This should decrease the amount of
memory grabbed when the program is left running for a long time.
New command "mem s" displays some allocation statistics that can be
useful when searching for memory leaks.  Please include a "mem s" output
when you think you have discovered a leak.

AX.25 timers T1 and T2 values are now randomized +/- 10% from their
preset values, to reduce the risk that two or more stations are
constantly colliding when their timers run out at exactly the same time.
It was reported that this often occurred on point-to-point links, where
both stations were sending RR(P) frames at the same time, and continued
doing so until the retry limit was exceeded and the link failed.

"dirfile.dbp" located in the directory used for broadcast files.  No
attempt has yet been made to detect duplicate entries, or to request
hole-fills when parts of the directory are missing.  After the pass of
the satellite, one should clean this file using "pfh -c dirfile.dbp" and
then convert it to a printable directory using "pfh -d dirfile.dbp >dirlist".
When the file grows too large to maintain it this way, simply delete it
and a new file will be created at the next satellite pass.  Note that
the "dirfile.dbp" file remains open during the pass, and gets closed
only when the 1-minute broadcast idle timer runs out, or when the

Server "tp" has been renamed "time" to be consistent with the list of
standard protocols in RFC-1340.  Please change "start tp" to
"start time" in autoexec.net.

IEEE 802.3 packets with sap=x06 are now correctly recognized as IP packets.
Tracing of IEEE 802.3 packets corrected.

A nasty problem with FTP bytes/s measurement, that sometimes caused a
crash when a file was sent in 0 seconds, has been eliminated.

Slow response (sometimes no response at all...) for internal AX.25
connections investigated, and hopefully eliminated.  This problem would
occur when a local application was connected from the NET/ROM, or when
setting up a connect to the own AX.25 BBS or NET/ROM.

When using SCC card, the transmission time for a single AX.25 connection
is now limited to 4 seconds, by varying the effective MAXFRAME value
depending on the length of the packets being sent. (no more than
4*bitrate bits are sent in a single transmission)

In the MSDOS version, the bus timing for the SCC driver has been
modified.  The driver now uses an "IN AL,0E4H" instruction to introduce
the proper amount of delay after a write to the SCC (5 PCLK cycles, or
about 1uS).  Tests have shown that this IN instruction provides a
reasonably constant delay on a wide range of machines.  Earlier versions
used a delay routine that introduced a fixed number of clock cycles of
delay, which varied widely between an old XT and a fast '486 system.

A version of the program for the Linux operating system has been
created.  It is compiled using GNU C.  Features are like the earlier
XENIX versions.  Some details are handled different:
-   The "telunix" server provides remote logins to the system using
"telnet" to the port where it was started (default 23).  This
requires superuser capabilities, so the program should be installed
with mode and owner like this:
-rwsr-xr-x   1 root     users      293897 Mar 18 20:49 net
Normally, NET will run with the privilege of the user that started
it, but during short periods when logging "telunix" servers on and
off, it switches to "root" privilege.  The "telunix" server starts a
/etc/getty process on the tty corresponding to the free pty it has
allocated, so this should not be done in /etc/inittab.
(as was required in the XENIX version)
When the file /etc/default/getty.telunix exists, it is used to
provide the getty settings.  If the file does not exist, the normal
settings file /etc/default/getty is used.  See "man getty" for
details.
-   The select() system call is used as much as possible, to wakeup
the program only when something has to be handled.  This saves CPU
time.  In XENIX, select() has bugs and is not supported on all
types of devices.  Linux handles select properly, and therefore it
is used by default (when variable SELECT is not defined).
-   The "termcap" entry of the controlling terminal is used to get
keyboard sequences for the "special keys" used in line editing.

All filenames used in the program are now "cleaned" before they are used
to open a file.  Previously, this was done only in certain cases.  The
actions perfomed by the "cleanup" depend on the operating system, but
generally they include removal of control characters, and conversion of
'/' and '\' to the pathname separator character normally used on the
system.  E.g.: now it is possible to say "record /tmp/file" on the
Atari, where this would not have worked previosly (TOS requires a '\' as
the pathname separator).

Handling of TELNET options has been significantly extended and debugged.
This does not really apply to the use of TELNET in the amateur radio
TCP/IP network, but is rather important when using TELNET to log in to a
host on a local area network.
The following options are handled by the telnet client:
Transmit_binary, Suppress Go Ahead, Status, End of Record,
Echo, Terminal Speed, Terminal Type, Window Size, X Display location.
These are handled according to the respective RFCs.  Some details:
- Terminal Type option is accepted only when an environment variable
"TERM" exists.  When the connected host asks for the terminal
type, the contents of this variable are sent.  So, to get the type
"ansisys" sent to the host, use: setenv TERM ansisys
- X Display location option is accepted when the environment
variable "DISPLAY" exists.  Like with TERM, the value of this
variable is sent when the host asks for the X Display location.
When the value of "DISPLAY" starts with a colon, the value sent is
the local hostname followed by the DISPLAY variable.
- The value sent for Terminal Speed is the speed of the interface where
the output for the connected host is routed.  This may not reflect
the speed of the entire path to the host...

Commands to control TELNET operation (the former "echo" and "eol"
commands) have been moved into a new command "tnoptions". Current
subcommands of the "tnoptions" command are:

debug	sets option negotiation debug mode.
debug 0: no debugging
debug 1: show option negotiation
display	when the current session is a "Telnet" type session,
the input mode and negotiated options are shown.
echo	is the former "echo" command.
echo accept: accept a "will echo" command from the remote
side, setting the input mode to TELNET.
echo refuse: refuse a "will echo" command from the remote
side, setting the input mode to LINE.
eol		is the former "eol" command.  determines the sequence
sent when the ENTER or RETURN key is hit.
eol standard:	CR+LF
eol null:	CR+NUL
eol unix:	LF
mode	when the current session is a "Telnet" type session,
sets the input mode.  valid modes are:
mode line:	local echo, full line of text is sent when
RETURN is pressed.  Lines saved in history.
mode raw:	separate characters are sent with no
translation.
mode telnet:	separate characters sent, special keys are
translated to TELNET-defined sequences:
Ctrl-B	-> Break
Ctrl-C	-> IP (Interrupt Process) + Synch
Ctrl-O	-> Flush Output
Ctrl-P	-> Synch
Ctrl-T	-> AYT (Are You There?)
Ctrl-_	-> EOR (End Of Record)
mode unix:	like mode telnet, but the following
Ctrl-D	-> EOF (End of File)
Ctrl-Z	-> Suspend Process
Ctrl-\	-> Abort Process
send	send a telnet option or suboption.  intended primarily
for debugging (in association with "tnoptions debug 1").
the option to be sent must be given as a decimal number,

SMTP handling of "." has been fixed.  The transmission no longer breaks
when a line consisting of only a "." is present in the message body.

SMTP sequence number handling fixed.  Sequence number no longer increments
by 2 when a message is forwarded.

The broadcasts of NET/ROM routes now include all routes that are above
the "minimum quality to be broadcast" and "minimum obsolescence to be
directed to.  This reduced the number of entries in the route table, but
it also limited the possibilities to recover from link failures.  With
the new memory management, the extra routes shouldn't be a problem.

Maximum number of simultaneous NET/ROM circuits has been increased from
32 to 64.  Please report when this is still insufficient.

NET/ROM transport layer packets received out-of-sequence are now held on
re-sent by the source after the missing packet was transferred.  I
believed that this would not be an issue, because the network links are
AX.25 virtual circuits.  In practice, maintaining the resequencing queue
seems to be worthwile.  (after receiving the NAK reply from the
destination node, the source node sends only a single packet, and the
next acknowledgement will indicate the correct reception of the packets
on the resequencing queue)

When "exit" command is used to terminate NET, all outstanding data on
transmit queues is sent, and one second is allowed for interrupt-driven
buffers to clear.  This solves problem with "param <interface> 255" in
onexit.net, which is supposed to let a TNC leave KISS mode.  (that
didn't always work).

Screensize determination in MSDOS version has been enhanced.  Now the
"screen swap" option will also work on super-VGA modes, e.g. 80*50 or
100*40 mode.  When the screen swapping seems to assume the wrong screen
size, it can be overridden by putting NETCONF=s80*25 on the command line
or in the DOS environment.

The "ax25 persist" command can only work on AX.25 interfaces where the
driver allows setting of the persistence.  This is true for KISS TNC's
attached to a COM port, and for SCC cards.  It is not possible to set
the persistence with drivers attached via "attach packet", so "ax25
persist" cannot be used with those.  NET now detects this condition and
removes the "ax25 persist" setting, instead of printing an "Invalid
parameter" message every time the "ax25 persist" routine wants to change
the persistence.

File name used by "rcmd" command modified: periods in the hostname given
to the "rcmd" command are now first replaced by underscores.  Then, NET
tries to open the file named: RCMDDIR/hostname.portnum
where RCMDDIR is the rcmd directory (contents of the RCMDDIR environment
variable), hostname is the hostname passed to the "rcmd" command, and
"portnum" is the port number given to the "rcmd" command (or the default
459).  When this file cannot be opened, the file RCMDDIR/hostname is
tried.  This change allows different rcmd passwords for the same hostname
(but for different ports), and avoids trouble with systems which can
handle only a single period in a filename.

PE1CHL.921225:

Many bug reports I have received lately turned out to be related to
configuration parameters.  Users unaware of subtle side-effects modified
critical parameters like "buffers" and "AX.25 timing", and later
complained about strange effects or memory problems.  Because the system
works best when everyone uses the same parameters anyway, most
configuration commands have now been removed, and the parameters have
internally been set to values that have proven to be fine during years
of experience.  This has the added benefit of reducing the program size,
and making it easier for newcomers to get the program running. The
affected parameters and their values are:

buffers 8*nchan		(8 buffers per channel of SCC interface)
smtp maxclients 10
smtp maxsessions 1
smtp timer 4000 500
tcp irtt 15000
tcp mss 512			(1K on an ethernet interface)
tcp timeout 7200
tcp ubound 3600
ax25 maxframe 4 16
ax25 paclen 128		(paclen depends on speed on SCC interface)
ax25 pthresh 64
ax25 retry 10
ax25 t1 15000 5000
ax25 t2 2000 500
ax25 t3 600000
ax25 t4 900000
ax25 window 2048
netrom param 3 50 6 5 20 3 24 3
netrom param 4 120 3 10 180 10 16 1800
netrom param 7 1 0 0
flow on

This change has made the file "config.net" mostly redundant.  The few
remaining parameters have been moved to the autoexec.net files in the
examples.

Internal buffer handling has been modified.  Should no longer see TCP
segments with "CHECKSUM ERROR" when using an SCC card.

DOS version: Fixed bug in memory allocator that caused strange behaviour
when running out of memory.

Fixed bug in the SCC driver that could cause a system crash at high baudrates.

A kick of a TCP session with no outstanding bytes will now re-send the
last transmitted ACK.  This can sometimes wakeup the distant station.
Also, when a RESET comes back it is known that the other station has
restarted the system and it is no longer necessary to wait for data.
A manual kick no longer increases the backoff.

Variable-maxframe computations in AX.25 now also use the interface speed
and length of packets when determining an optimal value for maxframe.

A bug in the AX.25 conference bridge, that could cause a crash, has been fixed.

Program has been made more compliant with RFC-1122 and RFC-1123.  This has
affected:

arp:    no longer sends a burst of requests when an ARP entry times out while
an FTP or SMTP session is running.  only a single ARP request is
sent and up to 10 packets are buffered waiting for a reply.

telnet: end-of-line sequence has been slightly modified.  now, 3 different
sequences can be selected using the "eol" command:
eol standard:	CR+LF
eol null:	CR+NUL
eol unix:	LF
everyone should try to use "eol standard", and should only switch to
a different setting when this is really necessary when talking to
some nonstandard implementation.

after connection to the telnet service, the server will now send the
telnet option "WILL SUPPRESS GA", as required per RFC-1123.  3 bytes
are sent in both directions after session setup for this option.

smtp:	SMTP server constructs a return path when mail is sent through
multiple systems. this should make it easier to reply to such mail,
and helps the server when it has to return a message to the sender.

PE1CHL.921114:

Interface MTU in attach commands is set to 128 when a smaller value is
specified (which would lead to a crash when IP frames had to be fragmented
to very small parts)

Improved operation of "netrom filter" and AX.25 BBS with F6FBB BBSes.

When using ESCC chips, the RTS signal can be made inactive immediately
after the closing flag, to improve turnaround time at higher speeds.
Set "param xx 4 0" to enable this.  Warning: some modems and/or transceivers
will not transmit long enough to send the last packet in a series, when using
this feature.  In that case, use "param xx 4 2" to introduce the smallest
possible time delay at the end of each packet.

PE1CHL.921018:

This is a major release, with many new features and enhancements.
Please report any bugs you find in there...
Make sure you update your configuration files, as there are some important
changes.  It is best to modify one of the example files to suit your
configuration.

Atari .TOS version: font can be switched between the existing font at startup,
and an IBM-compatible font.  This is useful when accessing a BBS that uses the
IBM special graphic characters, which display as hebrew characters in the
original Atari ST font.	 Switching is done using:
screen font a		(Atari font)
or:	screen font i		(IBM font)
The EGA font can only be used on monochrome monitors.  Note that the
"screen font" command sets the fontsize in the GEM version of the program.
Fonts can't be switched in that version (although you could load an alternative
font before starting NET)

MSDOS version: added the G8BPQ hostmode emulator.  See separate G8BPQ.DOC file
for details.  This emulator uses AX.25 ports 11..18, and is started using
"ax25 start g8bpq [<intno>]".  (<intno> defaults to 127)

Some revisions in the memory allocation routines, attempting to reduce memory
fragmentation.

SMTP server does not put a Received: header above mail when it is talking
to itself (e.g. after receiving AX.25 mail, or when expanding an alias)

FTP server indicates the transfer type (ASCII or IMAGE) when opening a data
connection, to hint certain common mistakes (e.g. getting a .ARC file in
ASCII mode)

AX.25 packet tracing now includes the packet length, to make it easier to find
packets that exceed certain length limits.

Enhanced tracing of Ethernet packets.  Can now also display IEEE 802.3
packets, and knows some names for non-IP Ethernet types.

Packet driver: It is now possible to specifiy a packet driver class when
attaching, instead of getting the default class from the driver and using
that.  When a packet driver supports multiple classes (e.g. Ethernet and
IEEE 802.3) it is now possible to select the alternative class.	 Specify 0
when the default class is to be used. The <class> values for packet drivers are:
ETHERNET	1
PRONET_10	2
IEEE8025	3
OMNINET		4
APPLETALK	5
SERIAL_LINE	6
STARLAN		7
ARCNET		8
AX25		9
KISS		10
IEEE8023	11
FDDI		12
INTERNET_X25	13
LANSTAR		14
SLFP		15
NETROM		16
PPP		17
Syntax of the attach packet command:
attach packet <int#> <class>|0 <label> <buffers> <mtu> [<call>|anytype|<types>]

When a packet driver has class "Ethernet" or "IEEE 802.3", the extra keyword
"anytype" at the end of the attach packet statement will allow all types of
packet to be received.	This can be useful when tracing the traffic on an
ethernet.  It is also possible to specify up to 4 ethernet types in HEX.
Examples:
attach packet 60 0 enet 8 1500 anytype
attach packet 60 0 enet 8 1500 800 806 8137

MSDOS version: The SCC driver now requires an alternative timing source when
it is running under DESQview.  A DRSI card has a Z8536 which will automatically
be programmed for this purpose (see below).  Other SCC cards must either
dedicate one channel for timing purposes, or they must use a TSR which will
deliver accurate timing interrupts on some other interrupt number than the
normal 08 or 1C vectors.  DESQview treats those interrupts in a special way,
which results in horrible SCC channel timing when another task is running.

The keyboard input routine can send a sequence before and after all keyboard
input characters echoed to the screen.	This can be used to set a different
attribute for text entered at the keyboard, and text displayed by NET.
In a session screen, this will make it easier to differentiate between the
text one has typed, and the text received from the other station.
To enable this feature, use "setenv" to store the desired escape sequence
in the following variables:
NETATTRECHO		sent before echoed characters
NETATTROUTPUT	sent before program output (after echoed characters)
NETATTRCMD		sent when returning to command mode (from session)
Example (for use on MSDOS systems with ANSI.SYS loaded):
setenv NETATTRECHO	 "<ESC>[0;1m"
setenv NETATTROUTPUT "<ESC>[0m"
setenv NETATTRCMD	 "<ESC>[0m"
Note that you have to include a line like this in your CONFIG.SYS:
device=c:\dos\ansi.sys

MSDOS version now recognizes DPMI, Windows 3.x on DOS 5.0 or higher, and OS/2.
(in addition to the existing support for DoubleDos and DESQview)
The proper "end timeslice" call will be made to release the processor when
there is no work to be done.

New command "broadcast blockftype" will block all files of specified filetypes.
This can be used to block "Satgate" traffic, CCD camera files, etc.
Examples:
broadcast blockftype		will list currently blocked types
broadcast blockftype none	all blocked types will be cleared
broadcast blockftype 02 d3	type 02 and d3 (hex) are blocked

PE1CHL.920701:

In the "ftl0 dir <server> a <date+time>" command it is now possible to add a
list of filetypes that are not to be included in the directory.	 This allows
to get a directory without all the automated-gateway files.  Example:
ftl0 dir uo22 a 920503160410 02

It is now possible to cancel an FTL0 request even while it is active.  In
this case the server will be disconnected to cancel the execution of the
request.  This allows the cancellation of a directory request when the link
has failed and the server had been immediately re-connected.  You can then
display the directory and re-issue the request with an updated "hightime".

In "netrom status" command, the timers now show the number of remaining
seconds before the timer expires, instead of the elapsed time. (there is no
room for the expiry time in this layout)

When someone connects to the NET/ROM from another NET/ROM node (i.e. incoming
on a NET/ROM circuit), and asks for a NODES or NODES * list, those nodes that
can be reached via the same neighbor that is used to reach him will be omitted
from the list.	A message (# NODES routed via XXXX not shown) is sent at the
end of the list to explain this.  The reason is that users connecting through
the network are only interested in nodes further down the path.
When a connect is made to another node, the same check is made to avoid that
nodes are connected that are reachable only via the same neighbor that the
incoming circuit also uses.  This avoids that a circuit is set up for which
all info packets have to be sent from one node to another, and then back to
the originating node.
A message "alias:call} No backward connects along incoming ROUTE" is shown when
such a circuit connect is attempted.
The second and third parameter of "netrom param 7" can be set to "1" to
restore the old behaviour (displaying all nodes, allowing all circuits).

A new command, "netrom filter" has been added to allow selective filtering
of certain patterns from NET/ROM uplink or downlink traffic.  This can be
used to inhibit bulletin forwarding on busy local channels during certain
times of the day (in combination with the "at" command).
Subcomands are:
netrom filter call [<call> <iface>|d]
enters a callsign in the table of stations whose traffic is to be
to be monitored.  normally you would enter the callsigns of all
local BBS stations in this table.  The SSID is not used in the
matching.  Specification of "d" instead of the interface will delete
the callsign from the table.  The "netrom filter call" command without
further parameters will display the callsign table.
netrom filter mode [y|n]
shows the status of the filter, or sets it ON (y) or OFF (n).
would normally appear in an "at" command, e.g.:
at 18:00 netrom filter mode y
at 00:00 netrom filter mode n
netrom filter pattern <pattern1> [<pattern2>....]
sets one or more patterns to be matched.  when a packet matches one
of these patterns, and the call of the connected station appears in
the callsign table, and the mode is "y", the packet is deleted.
this way, one can effectively obstruct certain types of traffic, e.g.
bulletin forwarding.  for a description of patterns see the pattern
matching description below.  Examples of useful patterns:
'^SB [A-Z0-9]* @ [!-_]* < [A-Z0-9]* \$[!-_]*$'		(RLI/MBL boxes)
'^F[AB] B [A-Z0-9]* [!-_]* [A-Z0-9]* [!-_]* [0-9]*$' (FBB boxes) Added "DCD-gated timing" for AX.25. While a signal is received or the transmitter is sending a packet, on a half-duplex channel, the AX.25 T1 and T2 timers will be halted. This causes less unnecessary re-transmissions when the channel is so heavily loaded that the transmitter cannot be keyed. Because a KISS TNC does not report this information to the computer, this feature is only available when using the SCC driver, with a plugin board like the PA0HZP OptoPcScc, DRSI PC*PA etc. The commands "ax25 t1" and "ax25 t2" now accept two values, the first of which is used for KISS interfaces, and the second is used for an SCC interface. The second value can be much smaller, as only the channel-idle time will be counted. Suggested values are: ax25 t1 15000 5000 ax25 t2 2000 500 Further improvements in the dynamic adjustment of MAXFRAME on AX.25 connections. NET will always start at MAXFRAME 1 when a connection is made, and will gradually increase the MAXFRAME up to the specified value when the connection is error-free. When a REJ or RNR frame is received the effective MAXFRAME steps back to 1. When a re-try is necessary to get a frame acknowledged, the MAXFRAME is halved. This algorithm should improve the channel efficiency by avoiding large MAXFRAME values on poor links, but still allows a large MAXFRAME when it causes no problems. Yet another kludge in the AX.25 handling, that should eliminate the double RR frames sent by the NET/ROM uplink. A user-defined sequence of characters may be inserted between trace output lines. The sequence will be inserted just before the date/time header. The default is "*** ", it can be changed using the command "trace separator <string>". In the string, "\n" can be used to insert a newline character. Example: trace separator "\n**** " Tracing of AX.25 packets now shows the state of the two "reserved" bits in the SSID byte. Normally these two bits are set to '1', and nothing will be shown. When one or both of the bits are '0', the indication [d], [e] or [de] is shown after the SSID, like in: PE1CHL-7[e]->PI8RNI-1[d]. The [d] corresponds to bit 5 in the SSID byte being clear, and this is in use to indicate a DAMA master. The [e] corresponds to bit 6, and indicates the use of extended sequence numbering when present in the source call field. Extended sequence number (modulo-128) option added to AX.25 coding. This allows the use of a high maxframe value on highspeed half-duplex links, especially when they are point-to-point. Most useful for links between NET/ROM nodes. The command "ax25 extended <call> [<iface>]" enables the modulo-128 mode for the specified station. A connection will be opened using SABME instead of SABM. When the other station replies with a DM or FRMR, a re-try is done using SABM, and the connection is set up in standard modulo-8 sequence number mode. Be careful! Not all AX.25 implementations reply to an SABME! The "ax25 extended" command should only be used when it is certain that the other station can handle the modulo-128 option, and ONLY on point-to-point links. The command "ax25 maxframe" now accepts two values for MAXFRAME. The first is used on modulo-8 connections as before, and can be 1-7. The second is used on modulo-128 connections, and can be 1-127. Because modulo-128 mode is only established during connection setup, and packets sent on this connection have a 2-byte control value, bit number 6 in the SSID field of the SOURCE call is cleared to indicate that the packet belongs to a modulo-128 connection. This is used to display the correct format when tracing. New command "sccstat b" shows the percentage of time that an AX.25 interface was busy. Separate columns are shown for the DCD (carrier detect) and RTS (request to send) signals. Note, however, that some modems will assert the DCD line when RTS is activated, so it may be that the percentage DCD also includes the percentage RTS. The command "sccstat b c" can be used to clear the counters and start a new measurement. MSDOS version: when SCC driver hardware type 08 (DRSI PC*PA) is selected, the Z8536 on the DRSI card is programmed to generate 10ms timer interrupts. This yields more accurate timing, just as described under "PE1CHL.920311", without the use of one of the SCC channels on the card. When "external clocking" is selected in the "attach scc" command, NRZ encoding instead of NRZI can be selected by specifying "ext/nrz" in the baudrate field. NRZ encoding is used by the DF9IC 9600 baud modem. To calm down certain people, "tcp kick" commands are now only acted upon when they are spaced at least 10 seconds. New command "ttylink", same as "telnet" but defaults to TCP port number 87 instead of 23. New command "rcmd", same as "telnet" but defaults to TCP port number 459, and performs automatic login to an rcmd server. To use this feature, a file must be present with the systemname as filename, in the directory "\net\rcmd". This file should contain the password of the system to login to. Example: rcmd pi8rni (the password should be in file "\net\rcmd\pi8rni") NET will wait for the login challenge (3 groups of 5 numbers), and will randomly select one of them. It will then send the corresponding password characters as defined in the password file \net\rcmd\pi8rni. When the password is correct, the remote system will respond with its prompt. The directory for the password files can be changed by setting the environment variable "NETRCMD" to a different value. New command "unproto" sets up an UNPROTO session. Any data typed on such a session is sent as UI packets (line by line), and any UI packets received on the interface specified during setup will be displayed. This can be useful for "chat nets" on a quiet frequency, and it is also useful as a general activity watch. When used on a PACSAT it will display the general information that these satellites send as beacon frames. Syntax: unproto <interface> <callsign> [via] [<digipeaters>] The specified callsign will be used as destination for the UI packets. It does not influence the display of received packets. MSDOS version: support for sending IP datagrams over NetBIOS, as specified in RFC1088. Attach a NetBIOS-supported interface using: attach netbios <lana_num> <label> <buffers> <mtu> e.g.: attach netbios 0 nb0 8 512 <lana_num> is the adapter number as used by NetBIOS. Try 0 when you don't know. <buffers> limits the number of receive buffers used by the driver, when packets are coming in and NET is not (yet) handling them. <mtu> is the max packet size. The IP datagrams will be sent as NetBIOS datagrams, using the destination name "IP.XX.XX.XX.XX", where XX.XX.XX.XX is the HEX representation of the destination IP address. Broadcast packets will be sent to "IP.FF.FF.FF.FF". Your IP address must be defined before the NetBIOS interface can be attached, because the name to be entered in the NetBIOS name table depends on this value. Thefore, put the "ip address" command before the attach command. This has been changed in the example configuration files. New command "version" re-displays the program version and copyright, that is normally displayed only at startup. Please mention the version number (date) when you report any bugs. Added a new command "fkey" to define function keys. It has always been possible to do this using the capabilities of ANSI.SYS or another external program, but as this was not widely understood amongst NET users, it has now been added as an internal command to NET. This also allows for a more convenient setup of the programmed keystrings, e.g. from CONFIG.NET The "fkey" command can program any sequence of characters for the function keys F1..F10. Examples: fkey f1 "ax25 status" 13 fkey f2 "connect 144 " fkey f9 -2 24 "session t" 13 A string will be stored literally, and has to be enclosed in quotes when it contains a space. A numeric value can be used to define a control character or other special key. Values for special keys are: -20: HOME -19: UP -18: PgUP -17: LEFT -16: RIGHT -15: END -14: DOWN -13: PgDN -12: INS -11: F1 -10: F2 -9: F3 -8: F4 -7: F5 -6: F6 -5: F7 -4: F8 -3: F9 -2: F10 When only "fkey f1" is entered, without any values, the programming for that key is erased. The command "fkey" by itself will print a list of programmed keys, with unprintable characters represented by [decimal_value]. The "screen" command, previously existing only in the Atari ST version, has been extended and now also is available in the MSDOS version. Subcommands: "screen swap [y|n]" enables/disables screen swapping. When this is enabled, separate screens are created for "command", "trace" and each "session" in the range 0-9. When switching between sessions, the existing screen is restored. This often makes it easier to run multiple qso's, as the previous text typed in each session can still be seen after typing many other lines and/or commands. However, the price to pay for this is *memory*. Each screen costs about 4Kbyte for the standard 25*80 mode, so that about 48Kbyte is required for all screens. Therefore, it is not available in the middle-model MSDOS version. Also, the screen swapping is done using direct access to the videoram, and may therefore cause problems in multitasking environments. On MSDOS systems, the access can be done via BIOS, by setting the environment variable "NETCONF=v0" before starting NET. However, access via BIOS is so slow that it isn't really practical to use this method. On the Atari ST, setting "screen swap on" automatically switches to direct videoram access ("screen direct"), so if you are using a nonstandard video adapter it won't work. New subcommands for "session" command: "session #" displays the maximum number of sessions. This is a configuration constant, that can be modified by setting the environment variable NETSESSIONS to a numeric value *before NET is started*, e.g. in a .BAT file. "session trace" switches to "trace" screen (when screen swapping enabled). Return to command screen by pressing a key. This key will also be the first command character, i.e. one switches back to command mode simply by starting to type a new command. FTP transfers can now be restarted, when the session had to be reset and the destination file was correctly closed. Make sure that you have set up the same environment as during the failed FTP (i.e. the same directory and transfer "TYPE"), and then issue one of the following command versions: get <remotefile> <localfile> restart put <localfile> <remotefile> <restart_offset> You have to specify both filenames, even when they are the same. To restart a "get" command, you only need to add the keyword "restart". To restart a "put", determine the number of bytes that have been transferred during the previous attempt, using a "dir" command. Then, pass that number as a third argument to the "put" function. You cannot restart a "put" when you don't have "overwrite" access at the remote system. Always make sure that the previous connection has been reset, before restarting a transfer! When a "dir" displays "0" for the filesize, or you are uncertain if the remote system has reset it's connection (e.g. because your machine crashed and the remote does not know that) *DO NOT* restart the transfer! First ask the operator of the remote system to reset the FTP session! Failure to observe this rule will result in a damaged file at the receiving end, because it is opened for write twice, and conflicts occur. The restarting is done by sending a REST command plus a byte offset, immediately before the RETR or STOR command that starts the transfer. Although the FTP spec specifies this use of the REST command, it does so only for the "block" and "compressed" modes of transfer, not for the "stream" mode that NET supports. Therefore, restarting a transfer is not compatible with other FTP implementations, and should only be used between machines running NET. When the remote system refuses the REST command, "abort" the transfer and start all over. New command "ftpstat" shows the status of all active FTP sessions, both client and server. More detail about one session can be shown using "ftpstat <ftp>", where <ftp> is the control block address obtained from the first listing. SMTP message_id allocation fixed. Now, the message ID in the header matches the actual ID in the queue, and it is incremented by 1 for each message processed. (previously it incremented by 2) Messages that are on the SMTP queue for a long time can now be automatically returned to the sender. Use "smtp expiry <days>" to define the number of days message are allowed to wait for delivery. "smtp expiry 0" disables this feature. The filetime of the .wrk file of the message is taken as the reference time of the message (normally it is equal to the time at which the message arrived at the station). Each time an outgoing SMTP session gets RESET, each message on the queue for that session is examined. When the time it is on the queue (current time - .wrk filetime) exceeds the number of days specified in the "smtp expiry" command, the message is returned to the sender with a short header explaining the problem. Reasons why an SMTP session gets RESET include: - ICMP "Unreachable" message received in SYN Sent state - timeout in SYN Sent state (the timeout is set to "smtp expiry" days) - "exit !" from NET - manual reset by operator (tcp reset <tcb>) "smtp hold" command added to allow the sysop of a 24-hour station to "hold" smtp mail which matches a certain pattern in "source address", "destination address", or "destination host". This can be required when the mail is destined to a gateway to another network (e.g. Internet), and the sysop is required (by license conditions) to monitor all mail going through that gateway. Subcommands are: smtp hold dest <pattern1> [<pattern2>....] smtp hold host <pattern1> [<pattern2>....] smtp hold source <pattern1> [<pattern2>....] set patterns for destination, host and source fields. (as appearing in "smtp list" output). The patterns are constructed following the rules defined in the section about "pattern matching" below. smtp hold mode [y|n] enable (y) or disable (n) the smtp hold feature. smtp release <job_id> release the specified smtp job, when it was in hold status. When "smtp hold" is in use, the sysop has to check using "smtp list" if jobs are being held, and if he wants to release them. The held jobs are indicated with a character D, H or S in the first column of the "smtp list" output. After they are released, the first character will be R. Note that exit/restart of NET, or a RESET on the SMTP connection with another host, may reset the status of a message to 'blank', and thereafter again to D, H or S. A general "pattern matcher" has been introduced, which is now used in the following places: AX.25 mail forwarder, "r" and "w" script lines in .bbs files. FTL0 "Open" packet detection smtp hold netrom filter Please note that you may have to modify some of your ".bbs" files, when they contained "r" or "w" lines with some of the special characters defined below. The pattern matcher uses the same syntax as UNIX programs like "grep", "sed", "vi" etc. Note that this is different from DOS filename wildcards "*" and "?". The following special characaters are recognized in patterns: ^ when it appears at the beginning of a pattern, it indicates that the pattern should match only at the beginning of the string. e.g.: "^abc" matches "abcde", but not "xyabc".$	when it appears at the end of a pattern, it indicates that the
pattern should match only at the end of the string.
.	matches any (single) character
[abc]	matches (single) characters "a", "b" and "c".
[a-z]	matches any (single) lowercase character
[^0-9]	matches any character that is NOT a digit
*	indicates the character or special that immediately precedes
it, can be present any number of times (0..infinity)
e.g.:	a*	matches any number of "a" characters
.*	matches any number of any characters
[0-9]*	matches any number of digits
\	generally indicates that the following character is to be
treated without special meaning.  e.g.: \. matches a ".".
$$opens a "memory", which is closed by "$$".  the contents of
the memory can be referred to anytime later in the pattern,
using \num, where num is the sequence number of the memory.
e.g.: $$[0-9]$$\1\1\1 matches "0000", "1111" etc.
\{	used with "\}" for specification of a limited number of matches
of the preceding character or special.
a\{1,5}		matches "a", "aa", "aaa", "aaaa" or "aaaaa".
b\{3,\}		matches strings of 3 or more "b" characters.

Atari version: fixed bug in allocator that caused "2-bomb" exit when a malloc
failure occurred.

PE1CHL.920420:

New command "sccstat f <chan#>" sends 30 seconds of 'flags' on AX.25-type
interfaces.  This is useful when evaluating the performance of certain
modems (especially the HAPN 4800 baud modem) on a scope.

PE1CHL.920329:

New "date" subcommand:
date get <host> [<port> [<tos>]]
requests the time from an RFC868 Time Protocol server, which is
normally located on UDP port 37.  Prints the difference with the
15 seconds, sets the local system clock to date/time received.
This command can be used to keep the system clock in sync with the
clock on another machine.  You can define a "date get" command to be
executed at some time each day, e.g.: at 03:00 date get pe1chl

PE1CHL.920311:

PC version: SCC driver option added to use one of the SCC channels as an SCC
timer tick generator.  It turns out that multitaskers like Desqview treat the
55ms system timer interrupt in such a way that a running task does not receive
all timer ticks.  This results in random timing errors when a program is
running in another window. (you can observe this as unusually long flag leaders
and trailers when listening to your transmissions)
When a spare SCC channel is available, it can now be set up as a 10ms timer, to
be used instead of the system timer: add the option "t<n>" at the end of the
"attach scc init" command, where <n> is the channel number to be used for
timing.	 When no hardware dependent options ar used, two zeroes must be
inserted between the <frequency> parameter and the option.  Example for use
with the PA0HZP OptoPcScc card:
attach scc 2 init 150 4 2 0 1 168 3 p4915200 0 0 t3
This defines channel number 3 (the last channel on the board) as a timer tick
generator.
The timer tick generator defined this way generates 10ms ticks.	 When you have
defined your own values for SCC channel parameters 1,3,4,7 or 11: remember to
re-calculate the values of these parameters based on this tickrate.

PE1CHL.920221:

Atari version: made MIDI port fully interrupt buffered.	 To use this feature
transmit), use the following form of attach command:
attach asy 3 1 ax25 430 2048 256 9600 $CALLSIGN-7 The old method remains available, using "attach asy 3 0 ..." This change greatly improves the performance of the MIDI port, also when used as a SLIP connection between two machines. Atari GEM version: fontsize is now selectable, using "screen font <fontsize>". Values 8,9 and 10 work best on monochrome screens, but larger values are also possible. The default on startup is 10. (8*16 system font) Fixed problem in SLIP/KISS receive routines that could cause the receiver to go deaf when a specific (erroneous) sequence of characters was received. new server: RFC868 Time Protcol. Start it using "start tp". It will return the current date and time in the format specified by RFC868, when an empty datagram is sent to its port (UDP port 37). The "date" command now has a number of subcommands: date request <host> [<port> [<tos>]] requests the time from an RFC868 Time Protocol server, which is normally located on UDP port 37. Prints the difference with the local clock, when a reply is received. date started: shows program uptime and starttime date set mmddhhmmss[yy]: sets the date and time date tzset <zone>: sets timezone, consisting of name and offset westwards from UT, e.g. "GMT0", "MET-1". you *must* use this form instead of the older "setenv TZ MET-1", to get meaningful responses from the Time Protocol server. PE1CHL.920212: Funny behaviour of SCC driver on some systems seems to be corrected now. If you still see a large number of interrupts (Exints or Spints) please report it with details of your configuration. Fixed bug in new segmenter. Also made some changes to LAPB module to get more reasonable behaviour in case RNR frames are received and/or sent. Keeping some of the "freeing garbage" messages in memory so that they can be re-displayed (using "mem g"). This also allows checking of freeing garbage situations when accessing a remote system via RCMD. The displayed info is still of limited use except when a symbol table of the running version of the program is available. PE1CHL.920209: Fixed a nasty bug that caused NET/ROM links to remain connected when the user requested a disconnect. Also this one may have caused some "freeing garbage". Added "BYE" and "QUIT" command to NET/ROM to disconnect (not shown in menu). Better args checking in NET/ROM and BBS code to reduce chance of loops when a BBS is talking to command interpreter (BBS: "I don't understand that"; NODE sent Info file; BBS: many times "I don't understand that"; etc etc) PE1CHL.920107: Included new segmenter as present in NOS. This should end the incompatibility problems between NOS and NET.PE1CHL when using "mode vc". Of course, it will introduce the same incompatibility between old and new versions of NET. Avoid the use of "mode vc". Fixed overflow problem that could cause the TCP retransmission timer to go haywire, when "backoff" and/or "SRTT" had high values. Result was a rapid transmission of packets at a 1-second interval. Changed memory allocation routines in Atari version (Turbo C). Should prevent problems with TOS versions before 1.4, that sometimes could not open files when a lot of memory was already allocated. Dropped support for EMS memory in PC version. It added extra overhead to the SCC interrupt handling, and caused nothing but problems anyway. Added possibility to hard-wire downlink ports for the NET/ROM. When a certain AX.25 station is known to be present on some port only, a "connect" request to that station can be automatically directed to that port. This makes the use of a dedicated port for a BBS less confusing for the users, and can also enforce the use of a highspeed downlink when available. Usage: netrom downlink <call> <port> [<digipeaters>] E.g.: netrom downlink pi8utr lap.48 PE1CHL.911222: TNC2 emulator added to UNIX/XENIX version. Use is like DOS version (see entry for version 890515, amongst others) Instead of the "comnumber", a full pathname for the UNIX device to be used to access the TNC2 emulator must be given. This can be a TTY or a PTY device. Examples: ax25 start tnc2 /dev/tty10=430/3 (KISS mode on a TTY) ax25 start tnc2 /dev/ttyp9=144 (cmd: mode on a PTY) When a TTY is used, the emulated TNC22 "appears" on the RS232 port controlled by that device. It will always run at 9600 baud. When a PTY is used, the name of the "slave" side of the PTY must be used in the start command, and a program running in the machine can access the TNC2 using the "master" side of the PTY (i.e. /dev/pty9p in the example above) To test the functioning of the TNC2, use "cu" or another communications program on the pty, or on another tty using an external loopback cable. Fixed pty location problem in telunix server. It should no longer silently close a connection when an active pty is still available. Some extra checks built into the SCC driver for MSDOS should avoid the stray transmit interrupts that occurred on some machines when using "vector" mode. Input packets on the SCC driver and the SLIP and KISS protocols on async ports are checked for a maximal length. Packets longer than the MTU of an interface plus the header overhead for AX.25 are discarded. Sometimes very long packets were received when noise was present on the input, and this lead to excessive use of memory by the drivers. This was probably the explanation for the use of large amounts of memory by NET, which was later freed to the heap but never returned to the operating system. Calculation of tcp MSS option value in SYN packets changed, to reflect above change. The MSS value relects the MTU of the interface on which the outgoing packets are routed. Using different MTU values on interfaces of the same kind (e.g. AX.25) could cause problems when the outgoing and incoming packets are not passing through the same interface. To avoid trouble, use only a widely-accepted MTU: 256 for AX.25, 1500 for ethernet. Calculation of local time for "at" command changed. Now, it should no longer be so sensitive to the setting of the TZ variable. Corrected bits/second value printed for very large FTP transfers. Fixed obscure timer problem that caused sudden trigger of *all* timers in the system under heavy load circumstances, resulting in the loss of all connections for which some timeout is active. PE1CHL.911126: "ax25 start tnc2" command now accepts optional /<kissrx> suffix to each argument. The result is that the TNC2 will be started in KISS mode, with the KISSRX parameter set to the specified value. This can be useful in systems where an MSYS or BPQ program is connected to a node running NET, using the KISS protocol on a serial link. Example: ax25 start tnc2 1=lap/3 Improved checks on duplicate use of a callsign-ssid combination. Now, duplicate use causes an error message, instead of subtle problems like "ARP not working" or "No ACK messages on broadcast request". SCC driver: enhanced handling of DCD. Before, a when DCD dropped while a packet is received, the packet was always discarded. Now, two alternative modes of operation are possible: param <iface> dcd n DCD-changes are ignored, the receiver is permanently enabled. In this mode, a packet is never discarded because of DCD loss. However, when noise is present on RxD it will cause a permanent interrupt load on the system. Normal operation restored by: param <iface> dcd y param <iface> 11 <n> Whenever DCD falls, the receiver remains enabled for <n> ticks. (the duration of a tick is displayed at startup, normally 10ms on the Atari ST, 55ms on the PC). Setting this parameter to a suitable value will eliminate the packet loss caused by a flickering DCD, as can sometimes be observed with the G3RUH modem. Setting parameter 11 to 0 will restore normal operation. These DCD handling options do not affect the "channel busy" detection that is performed in half-duplex mode (CSMA). The P-persistent channel access algorithm takes the unsmoothed DCD as input. NET/ROM now allows a fast connect to another station using the forms: connect <station> via <entrynode> <exitnode> connect <station> via <node> This is similar to the facility offered by Flexnet. You will get a connect to the local node, and the link to the specified station will be automatically set up. When the local node detects a problem (e.g. <exitnode> not in nodetable), it returns a DM (Busy). When a problem at the remote node occurs, the normal error message will be returned on the connection. The old step-by-step connection method remains available. NET/ROM downlinks are now made with a digistring containing both the source node and the local node. This allows easy reversal of the path to connect back to the calling station. "netrom param 7 0" can be used to disable this feature. AX.25 T1 time calculation based on smoothed round trip time. Enabled only for NET/ROM links on interfaces without uplink capability (pure interlinks), and for FTL0 links. This should improve performance on fast dedicated links, especially when they are fullduplex. PBLIST format from DRAFT spec (Nov 4, 1991) implemented, because OSCAR 22 has started to use it, and this caused an infinite loop with "broadcast download". PE1CHL.911020: Increased initial TCP local port number from 1001 to 1024, because HP-UX (and possibly other systems) refuse FTP data channels with a port number below 1024. Extra option to "broadcast download" allows insertion at the front of the download queue, for higher priority when selecting a file to be downloaded. Usage: broadcast download <sat> <file> prio Use this when you want to download a personal message while there is already a number of other files in the queue. Fixed error that caused crash when an OK reply on a "broadcast request" command was received while no "broadcast download" was in progress. Don't use "b r" command in 911005/911006 version! PE1CHL.911005: "Broadcast download" command added, that allows unattended, automatic downloading of files from a Pacsat Broadcast server (satellite). A "broadcast server" has to be defined (usually this is done in autoexec.net) "broadcast download <servername> <file_id>" adds a file to the list of files to be downloaded. As soon as a PBLIST packet is received from the specified server, NET automatically sends "broadcast request" packets to get the specified file(s). "broadcast download <servername>" without file_id prints the list of files queued for download. A "*" indicates the file currently being requested. Files can be canceled from the queue using "broadcast cancel <rqst_CB>", where the <rqst_CB> can be found in the list. New command: "broadcast kick", causes current broadcast file to be closed, and complete files to be written to the bcst directory. Same thing as happens after 1 minute without incoming broadcast packets. FTL0 downloads and broadcast requests or downloads of already complete files now prints an error message, instead of transmitting a request which results in nothing useful. New command "smtp maxsession" defines the maximum number of incoming SMTPs from one remote host. Default is 1. When an incoming SMTP session would cause the maximum to be exceeded, the oldest connection is closed. This feature was introduced in an attempt to defend against other software that resets connections after a small number of retries, thereby leaving open SMTP sessions that will never complete. These SMTP sessions occupy resources (memory, open files) on the system and may eventually lead to problems. Note that resetting a connection on which data is still outstanding is against the SMTP and TCP specs, so it is really the remote system which is at fault. PC "large" version: setting the environment variable "NETEMS" to any value, before starting NET, causes some EMS memory to be added to the heap. An EMS 3.2 or 4.0 compliant EMM must have been loaded in CONFIG.SYS. A message about EMS will be displayed before the usual startup banner, when the EMM is recognized. Using EMS memory can leave more DOS memory available to run a SHELL from NET. However, when using the SCC driver at high baudrates, you may get overruns because the EMS memory has to be banked-in at interrupt time. EMS memory cannot be used when "attach packet" is in use. PE1CHL.910828: New command: "ftl0 post <server> <command>" specifies a command (and parameters) to be executed whenever all queued requests for that server have been successfully executed. The command is started at the moment the FTL0 server is disconnected. It can be used to process the results of the FTL0 session in an automated system. Example: ftl0 post uo14 source c:\net\uo14post.net Use "ftl0 post <server> #" to clear a previously specified command. Adding the keyword "delete" to an "ftl0 upload" command now causes the file to be deleted after it has been upload successfully. This is indicated in the "ftl0 status" output by the indication "[del]". Adding the keyword "perm" to the "bcst abort <bcst_CB>" command now aborts the file AND marks it as "received", so that no further attempts will be made to assemble the file. (e.g. to prevent reception of a big file you have upload yourself, or you are otherwise not interested in receiving) The command "netrom logmask" has been added to allow selection of the events to be logged. The logmask is a hex number, being the OR of the following values: 0x0001 user connected NET/ROM (AX.25) 0x0002 user connected NET/ROM (NR4 circ) 0x0010 downlink okay 0x0020 downlink fail 0x0040 circuit okay 0x0080 circuit fail 0x0100 link failed PE1CHL.910818: This version includes support for the NEC PC98xx, written by Kohjin JR1EDE. To accomodate this (and probably other MSDOS machines that are not PC clones), the environment variable NETCONF is checked during startup. When it is defined, it is searched for the following strings: k0 use MSDOS call 06H & assume PC keyboard codes (default) k98 use NEC PC98xx BIOS call for keyboard input t0 use PC BIOS 55ms ticks at 40:6C (default) t1 use MSDOS get_time call t98 use 8253 timer in NEC PC98xx The NETCONF variable can be defined using a SET command in a .BAT file, or by passing it as a parameter when starting NET. Example: NET NETCONF=k98t98 You can put this command in a .BAT file together with the COMBIOS startup command to make startup of NET more convenient. On non-PC-clones, you will have to set the "ftl0 trace" to 00ef instead of 00ff, to avoid a hangup during an FTL0 session. PE1CHL.910808: "Broadcast servers" can be defined (like FTL0 servers), using: broadcast server <servername> <iface> <call> <blocksize> <reqwait> [<minw> <maxw> <UIdest>] defines a Broadcast server (satellite). a user-chosen name can be assigned to each server, and the remaining parameters are defined as required for the different satellites available. examples: broadcast server uo14 9k6 uosat3-11 244 6 broadcast server uo22 9k6 uosat5-11 244 6 1 3 pblist-1 parameters can be modified by re-entering the command broadcast server shows a list of servers defined using the above command [the last four parameters of the command are not yet used - experimenting...] A new form of "broadcast request" can now use the information stored in this way. Usage: broadcast request <servername> <file_id> [hole|stop] This saves the typing of interface name, callsign and blocksize for each request, as this information can now be referred to using the server name. The old form of the "broadcast request" command still works, and can be used when an alternative blocksize is to be specified. New command "broadcast trace [<hexval>]" controls the unsolicited messages printed by the broadcast receiver (like "ftl0 trace" does for the FTL0 messages). Parameter <hexval> is the (hexadecimal) OR of the following values: 0001: print PBLIST packets sent by server 0002: show which commands are sent to server 0004: print ACK packets received from server 0008: print "file 100%" messages 0010: print completed file messages Added audible FTL0 feedback for PC too. It does not sound as nice, and it ties up the CPU for 5 milliseconds, but well... there is no sound-generator chip, so... Unless you turn it off using "ftl0 trace", a high beep is sounded when the server acks an upload packet, a lower-pitched beep sounds for each received packet. Added mail distribution optimizer as suggested by PA2AGA. When the mail is to be sent to multiple users at the same host (e.g. via an alias entry that is used as a mailing list), the message is sent to that host only once, with multiple recepients. This is done by scanning the recepient list for equal hostnames. Aliases for the same host are not recognized as such. Program can now be compiled using Borland C++ version 2.0 PE1CHL.910802: Enhanced disk-full checking when receiving broadcast (those UoSAT-22 pictures, you know...) Also changed fill character to 0x00 when incomplete files are saved. This is useful when receiving raw pictures. (filler used to be '?') The first fragment of a file is now always saved when received during broadcast. Useful when changing files are broadcast. UNIX version: ihave.bc and frags.bc now put in broadcast directory properly. Please delete any file xxxx\*.bc you may have on disk (xxxx = broadcast directory) PE1CHL.910720: FTL0 has been enhanced with a p-persistant automatic connect. Now, when the "Open" packet is received, the connection attempt will be made after a random delay between "minw" and "maxw" seconds. These values *must* be specified in the ftl0 server command, just before the bbstat. PG uses minw=1 and maxw=3. Example: ftl0 server uo14 9k6 uosat3-12 130 2 2 1 3 bbstat "Open" MAKE SURE YOU UPDATE YOUR AUTOEXEC.NET when installing this version! Some minor changes in the NET/ROM handling should again improve it's operation, and the appearance of the NODES list (less unreachable nodes). Fixed bug in tty driver that caused trouble with raw-mode telnet. Fixed CR/CR/LF sequence in Atari rcmd (Turbo C only) PE1CHL.910622: "ftl0 dir" commands with "f" (file_id) and "t" (to_addr) specification now accept an optional extra parameter, the HIGHTIME to be used for the selection. This allows more efficient selections to be made with the current satellite software implementation, that optimizes it's search when a HIGHTIME is given. Example: To get "all messages sent to me and all messages sent to ALL after" 910622 01:05:00", use the follwing two commands: ftl0 dir uo14 t *all* 910622010500 ftl0 dir uo14 t pe1chl 910622010500 Then run PFH with the -c option to compact and re-sort DIRFILE.DL, and run PFH with the -d option to display the directory. The next time, use the HIGHTIME indicated by PFH on the last line of the -d display. Using this technique, one can keep track of all personal messages and bulletins without having to download all of the satellite's directory. NET/ROM "users" command now differentiates between established connections and connections in setup phase using the symbols "<-->" and "<..>". PE1CHL.910609: The alias for NET/ROM routes added automatically when a packet is received from a node for which no route exists is changed to "######". Previously, such routes got no alias at all. Failures of a NET/ROM link are now logged to the logfile. When a file received with broadcast is 100% complete, a message is printed. At that time, you could enter a new broadcast request. (the current software on the satellites allows only a single broadcast request per user) The program still waits for 1 minute without received broadcast packets before it writes all the received data to the permanent files in the broadcast directory. New command: "broadcast cleanup [<percentage>]" will abort all files that have less data than a specified percentage (default 10%). This provides a quick way to get rid of those files of which only a few fragments were received. When "broadcast stop" is executed, the fragments of files received will be saved in a file "frags.bc" in the broadcast directory, that will be reloaded when broadcast is started again. When the "broadcast stop" command is in the file "onexit.net", this allows exit from NET and a later re-start, while retaining the data for partially received files. PE1CHL.910603: "netrom nodefilter mode" and "netrom nodetimer" now can be set per interface. Specify the interface name as an extra parameter. When name is omitted, the value is set for all interfaces, as before. Because these commands now only act on actually initialized interfaces, these commands have to be moved from their original location in CONFIG.NET to a line below the "netrom interface" commands in AUTOEXEC.NET. new mode "smtp mode forward". works like "smtp mode route", except that it won't write new .txt files in the mail directory. useful for unattended nodes that can forward mail (using "alias" file or the user%host1@host2 construct), but do not want to accept local mail. PE1CHL.910525: Fixed problem in Broadcast Protocol Hole-list request: the requested list did not include the hole between the last received fragment and EOF. When this was the only hole, a request for the entire file was sent. Idle-timers for FTL0 connections have been changed to accomodate for long processing delays aboard the satellite. Now, instead of disconnecting after timeout the program first tries to poll using RR(P) and simply resets the connection when there is no reaction. Hopefully this prevents the looping behaviour that was often seen when the link failed. When a DM is received on an open AX.25 connection, the connection is no longer automatically re-established. It goes to "disconnected" state after the message *** LINK RESET from .... Fixed problem in mail destination check: the SMTP server refused mail for dest@call.bbs when there is no route to [255.255.255.255], even when the proper forwarding file for call.bbs existed. This is because "call.bbs" destinations are internally mapped to this IP address. In 910330, this problem can be worked around by defining the route. (e.g.: route add [255.255.255.255] 144) Added support for the Z85130/Z85230 ESCC chip to the SCC driver. This is enabled by using "attach escc" instead of "attach scc" to attach the channels. The "init" attach line can use either "scc" or "escc", with no difference. The total set of chips driven by the SCC driver can be a mix of Z8530/Z85C30 (SCC) and Z85130/Z85230 (ESCC). The ESCC can run fullduplex without the external divide-by-32. When no divider is fitted, the "d" option in the baudrate field should be omitted (as always). The main advantage of the ESCC is the larger TX FIFO (4 bytes instead of 1) and RX FIFO (8 bytes instead of 3), allowing a longer interrupt latency without causing overrun/underrun errors. Therefore, the chip can be used at the high edge of the speed range supportable using interrupt-driven I/O (19200 and possibly 38400 baud). It also solves some bugs in the original SCC, and creates a few of its own. Unfortunately, it is quite expensive, so for now the best choice is likely to remain the SCC. PE1CHL.910508: Fix in AX.25 mailbox: mail can now be terminated with ^Z anywhere on a line, not only on first position. The part of the line before the ^Z will be written to the mail, but a CR/LF will allways be appended. Necessary when F6FBB BBS forwards mail to it, because FBB leaves partial lines ending in ^Z in mails, and hangs the forwarding. Other boxes do the same as NET.PE1CHL now does, avoiding the problem... PE1CHL.910428: Changed NET/ROM: The output of the command-interpreter is now sent as separate packets instead of a single message (with MORE bit set on all packets but the last). This was done to prevent crashes in old NET/ROM nodes that occur when long messages are sent, e.g. the output of "N * *". The consequences of this change are: 1. the output is now sent in 236-byte packets (instead of 256) 2. when the link fails you may receive an incomplete response Fixed check in NET/ROM command interpreter to make sure only the NET/ROM call valid for the interface the user is connecting on can be used, not those of other interfaces. (e.g. PI1NOS vs PI8NOS) Added some extra output flushes in Broadcast and FTL0 (some messages appeared only when a key was hit) PE1CHL.910330: Fixed problem in "netrom bcdest". When more than one destination was given, some garbage was transmitted in the first netrom broadcast to the second (and further) destinations. Removed limit of 10 on "smtp maxcli". You can set any limit now, but understand that the program can open only 20 files and that failure to open a file may occur if too many sessions are active. This does not include any sessions in "SYN sent" state. SMTP server now checks mail destination more throughly. When the destination system is not reachable the mail is refused. PE1CHL.910317: Fixed problem that caused crash when "exit" was done on PC and SCC driver was in use. (caused by timer tick interrupt number fix) When connected by a forwarding/reverse-forwarding BBS, the AX.25 mailbox now disconnects after a single command error (after sending "Huh?"). This avoids looping when the two systems get out of sync and send error messages to eachother, something that could happen with an F6FBB BBS. Added printout of "ack" message sent by satellite on "broadcast request". Added "broadcast request" to stop a broadcast, like: broadcast request 9k6 uosat3-11 af7 244 stop Added command to grab UI frames sent to specified (or all) destinations. This can be used to get satellite telemetry, and other data sent as UI frames. The grabbed frames are appended to a file, further processing is required to make them readable. This is supposed to be done in an external program. Only those frames not locally processed by other protocols can be grabbed. Commands: ax25 uigrab ax25 uigrab <filename> [<dest>...] ax25 uigrab off The first form displays the current state and destinations. The second form enables the grabbing, specifying an output file and 0 or more destinations. When no destinations are given, ALL non-handled UI frames are grabbed. One or more destinations (like QST QST-1 MAIL TLM) can be specified to limit the amount of data grabbed, especially when TCP/IP traffic is present. The third form of the command stops the grabbing. The data is dumped in the output file as a sequence of headers and packets. The headers are formatted like this: struct axui_header { unsigned long magic; /* magic number for re-sync */ #define AXUI_MAGIC 0x41585549L long rxtime; /* timestamp */ char ifname[6]; /* interface name */ short int len; /* length of following data */ }; The rxtime is the number of seconds since 1970 at the time the packet was processed. The ifname field contains the first 6 characters of the interface name, \0-terminated when it is shorter. len specifies the number of bytes in the received packet. The received packet immediately follows the header. It includes the AX.25 header, consisting of shifted-ASCII addresses, ctl byte (03 or 13) and PID. PE1CHL.910228: NET/ROM routing broadcasts are made configurable. The "netrom bcdest" command can add or drop destinations to/from a list for each NET/ROM interface: netrom bcdest <iface> add <dest> [<digi1> [<digi2>]] netrom bcdest <iface> drop <dest> [<digi1> [<digi2>]] The existing destinations can be displayed using "netrom bcdest <iface>". By default, only the destination NODES exists. It can be dropped using "netrom bcdest <iface> drop nodes". One or more destinations can be added to the list. Keep in mind that the node broadcast can be very lengthy when you add multiple destinations, and some temporary memory is needed at that time to queue all the packets. Limit the number of destinations to the minimum. Node broadcasts can now be completely disabled on a single interface while still having NET/ROM capability on that interface, by dropping all destinations. The NET/ROM now also accepts node broadcasts via up to two digipeaters. PC: the interrupt number to be used for SCC timer ticks can now be specified as an optional parameter to the "attach scc init" command. The default is 08, i.e. the hardware timer interrupt. Possible other candidates are 1C and 70. The driver assumes that the timing hardware has been initialized before and a handler is in place that resets the pending interrupt. It only chains it's handler on the specified vector. The tickrate should be somewhere between 18 and 200 ticks per second. Example: attach scc 2 init 150 4 2 0 1 168 3 p4915200 0 0 1c Note that the 2 hardware-dependent parameters after the clockfrequency should be present in this case. Use zeroes when these parameters are not needed. Also note that the 1C vector is not suitable when multi-tasking DOS replacements like Desqview and Doubledos are in use. Vector 70 can only be used in an AT, and needs a TSR routine to setup and handle the RTC timer interrupt. PE1CHL.910218: Fixed bug in FTL0 downloading on PC: It did not properly send the continue offset when a download was resumed after a link failure. This resulted in corrupted files when the download could not be completed in one go. FTL0 download: files are now named *.pdl during download, and renamed to *.dl when download succesfully completes. "broadcast request" command: the blocksize is now optional, and defaults to 244. so, you can now request the broadcast of a specific file using a command like: b r 9k6 uosat3-11 af7 the blocksize *must* be given when you want to send a hole-list: b r 9k6 uosat3-11 af7 244 hole note that this is still not supported in the onboard satellite software. SCC driver: the sccstat command now accepts a channel number to display info about a single channel. Added "calibrate" facility to SCC driver: sccstat c <chan#> sends 30 seconds of zeroes (NRZI pattern 101010...) on AX.25-type interfaces. This is intended for tune-up of TCM3105 modems and for link tests. When 30 seconds isn't enough, simply re-issue the command (we're still waiting for that SCC card test program) PE1CHL.910127: Corrected bug in "bcst abort" and "bcst complete" that caused either a "not a valid control block" or a crash when these commands were used. Added "ftl0 cancel <Req_CB>" to cancel a queued request. <req_cb> can be found in the "ftl0 status" output. Active requests cannot be canceled. Added "ftl0 directory <server> s <selection> intended for use by an external selection compiler. <selection> is a string of hexvalues, forming a valid FTL0 SELECTION. Note that the form "ftl0 directory <server> a <yymmddhhmmss>" is most efficiently processed by the satellite software. All other forms can take a long time to execute onboard the satellite, with the danger of losing the connection in the meantime. PE1CHL.910123: Added "ftl0 directory <server> a <yymmddhhmmss> gets a directory like described below. selection is on upload_time later than the specified date/time. part of the date/time spec can be omitted, defaulting to zeroes. Added "ftl0 trace [<hexval>]" sets the trace level for FTL0 transactions. the hex value is the OR of the following values: 0001: print BBSTAT packets sent by server (only when jobs are queued) 0002: show which commands are sent to server 0004: show state changes on AX.25 link (connected, busy etc) 0008: print alert message when link disconnected permaturely 0010: Atari ST only: audible feedback of ack'ed and received packets a high click is sounded when the server acks an upload packet, a lower-pitched click sounds for each received packet. from 910807: also implemented in PC version PE1CHL.910119: Automatic connection of FTL0 server built in. When there is work queued, and the server is heard sending a specific UI frame, the connection is automatically set up. For this purpose, the "ftl0 server" command has been extended with optional <UIdest> and <UIstring> parameters specifying the destination address and a string to be matched in the UI frame. For UOSAT3 and PACSAT these parameters would be specified as "BBSTAT" and "Open", e.g.: ftl0 server uo14 9k6 uosat3-12 130 2 2 bbstat "Open" When a packet to <UIdest> is received while work is queued, it is printed. !!!! see 910720 for a change in the ftl0 server command !!!! FTL0 client now doesn't connect the server anymore when an upload request is queued. An explicit "ftl0 kick" is required to connect the server, or the automatic connect feature must be used. New FTL0 commands: ftl0 directory <server> f <file_id> ftl0 directory <server> t <dest_address> gets a directory from the specified server, and stores it in the file "dirfile.dl" in the directory specified using "ftl0 homedir". (see below). The first form gets a directory of all files with a file number above the specified value, the second form searches for files with the specified destination address. The wildcard characters "?" and "*" can be used in the destination address. example: ftl0 dir uo14 t *pe1chl* ftl0 download <server> <file_id> [<register_dest> [<lock_dest>]] queues a file for download from an FTL0 server. <file_id> is the id of the file to download (on the FTL0 server) <register_dest> and <lock_dest> are sent as specified in the FTL0 spec. when omitted these default to 0, as required for ordinary file downloads. ftl0 homedir <directory-name> sets the directory used by FTL0 download and directory commands to store the received files. A subdirectory will be created, named using the callsign part of the source address (e.g. UOSAT3 or PACSAT). PE1CHL.910106: Limited FTL0 client support built in, to upload files to PACSAT-type satellites. The following commands are recognized: ftl0 server <servername> <iface> <call> <paclen> <maxframe> <frack> defines an FTL0 server (satellite). a user-chosen name can be assigned to each server, and the remaining parameters are defined as required for the different satellites available. examples: ftl0 server uo14 9k6 uosat3-12 130 2 4 ftl0 server po16 psk pacsat-12 130 1 6 parameters can be modified by re-entering the command ftl0 server shows a list of servers defined using the above command ftl0 status <servername> shows the status of that server (including any requests queued) ftl0 upload <server> <filename> [<file_id>] queues a file for upload to the specified server. the filename should be a full pathname to avoid problems when a "cd" is done. of course the file must have a valid Pacsat File Header! when the <file_id> is specified the upload is treated as a continuation of an earlier attempt. when no link exists to the server, the "ftl0 upload" command tries to establish it. ftl0 kick <server> try to establish a link to the specified server, to attempt to execute all queued requests for that server. must be issued when a BUSY was returned at the "upload" command, or when the link failed for some reason. When the client has a link to a server, it attempts to execute all queued jobs for that server. When this is finished, it disconnects the server immediately. Re-connection takes place when new jobs are submitted, or when the "ftl0 kick" command is issued. When the link fails, it is not necessary to re-queue any of the jobs. These are held as long as the program is running. Should it be necessary to exit from NET and lateron resume an upload, the upload must be re-queued using the file_id received at the first attempt. This file_id can be found in the output of the "ftl0 status" command, and is also logged in the logfile. PE1CHL.901226: New broadcast subcommands: broadcast abort <bcst_CB>: this terminates the specified broadcast file reception and throws away all received data (unlike "broadcast complete" which saves it, even when the file is incomplete) broadcast request <iface> <call> <file_id> [<blocksize> [hole]]: sends a broadcast request to the specified call, requesting transmission of file_id. the blocksize specifies the packet length and can be up to 244. When "hole" is specified, the hole-list (list of missing packets) is sent with the request. (note: this is not yet supported by the satellite...) The "session" command can now send a line of text to a session. Usage: session 1 "one line of text" This sends the argument (if more than one word is has to be quoted) to session 1. When "session c" is specified it is sent to the current session. This can be used in "source" files and/or with the "at" command to start FTP transfers at a convenient time, or to login to an FTP session. One could put the following in a file and execute it using "source": ftp pe1chl ses c "user guest" ses c "pass rob" ses c "type i" Some enhancements to the TNC2 emulator make it more compatible with a real TNC2. Now, the "paclen", "maxframe", "frack", "resptime" and "check" parameters are really used for TNC2 connections. Of course the defaults are the same unreasonable values as used by the TNC2 firmware, so these should be set before a connection is made (especially "frack"). PE1CHL.901215: When the AX.25 forwarder receives a NO reply when sending a message, it is returned to the sender and the forwarder skips past the 'b' command. This allows bulletin forwarding to boxes that allow abbreviated forwarding. The SID string has to be sent explicitly. The new command 'f' in the forwarder script allows reverse forwarding. A string to request reverse forwarding can follow the 'f', the default is "F>". Example of a script to forward to MBL/RLI boxes using abbreviated forwarding: c 430 pi8utr connect the mailbox w> await the > prompt s[NET-H$]			send SID
r>				next line must be > prompt
l				loopback point
mS%s %s@%s < %.6s $%.12s send a mail or bulletin rOK response must be OK u%s send subject b send body w> await the > prompt n send next mail (if any) from loopback point f request reverse forward PE1CHL.901212: Shortened the version string. This version has so little left from the original 871225.33 base that this seems to be warranted. Added Pacsat Broadcast Protocol reception capabilities. The system can receive files transmitted using this protocol. Currently OSCAR14 and OSCAR16 use it, and OSCAR19 will follow shortly. The commands currently defined are: broadcast start "pathname": This will enable the storage of received broadcast frames. The pathname refers to a directory where the completed files fill be stored. Example: broadcast start "c:\bcst". broadcast stop: Disable the broadcast receiver. All completed files will be moved to the directory specified in broadcast start, and all fragments of incomplete files will be discarded. broadcast status: Shows the status of the broadcast receiver, including some statistics and the status of each incomplete file. The fragment list of an incomplete file can be shown using "broadcast status <bcst_cb>". The file currently (or last) being received is marked with a '*'. broadcast ihave: Shows a list of all files that have been stored away in the directory specified in the "broadcast start" command. This list is kept to prevent duplicate reception of files. broadcast complete <bcst_CB>: Forces the storage of a received file. When this command is given for a file that is not yet complete, the gaps in the file will be filled with question marks '?'. Any broadcast packets received for files "under construction" are first stored in a temporary file (the name of this file is shown in the "broadcast status" output). Only when the file is complete it is moved to it's destination. The directory for temporary files can be modified by setting the TMP environment variable (or TMPDIR on the Atari ST). This can be done using "setenv" (see below). The broadcast receiver kicks a 60-second timer each time a broadcast packet is received. When this timer elapses (i.e. after 60 seconds without incoming broadcast packets), the list of incoming files is examined and all complete files are moved to their destination directory. The same thing is done when a "broadcast stop" is executed. This implementation allows full CPU attention during the highspeed transmissions from OSCAR14, assuming that the timer only elapses after the satellite is out of reach. When you want to exit NET before the timer has elapsed, execute a "broadcast stop" to update and close all files. The directory specified in the "broadcast start" command holds a file "ihave.bc", which is updated when a file is stored away. A copy of this file is kept in memory for fast lookup when a broadcast packet is received. Any packets belonging to files in the list are ignored. To save memory, it is best to remove the old entries from this file when it is certain that they are no longer broadcast. This can be done using a text editor. It is also possible to delete the file. The broadcast receiver will store incoming files in subdirectories of this same directory, named after the sourcecall used in the broadcast packets. e.g.: files received from OSCAR14 will be stored in the directory UOSAT3, because this satellite sends broadcast data as UOSAT3-11->QST-1. The individual files will be named iiiiiiii.tt, where iiiiiiii is the 8-digit HEX file ID (as assigned by the satellite) and tt is the 2-digit HEX filetype (as assigned by the uploader). The handling of broadcast packets at 9600 baud (as sent by OSCAR14) imposes a considerable load on the processor and disk. As it has been observed that some slower systems already have trouble with 9600 baud AX.25, it can probably be stated that an AT system is needed to successfully receive broadcast at this speed. Tests on the Atari ST (using the SCC interface, TOS 1.4 and a RAMdisk for temporary files) have shown no difficulties. 871225.33.PE1CHL.901129: Implementation of IP over X.25, currently only for the EiconCard adapter. This allows routing of IP packets via an X.25 network, thus allowing the construction of a bridge between different local area networks using NET. The implementation follows RFC877, transmission of IP datagrams over Public Data Networks. Unfortunately, the cost of a connection to the public X.25 network is way out of reach for amateur use... This feature is composed of the following subcommands: attach eicon <port> <label> <mtu> Attaches an EiconCard, previously installed and configured using the configuration tools delivered with it. Only the X.25 Network-level software is required (ACCESS/X25). <port> will normally be 255 when a single card is installed. Subsequent ports are numbered 254, 253 etc (as defined by Eicon). <label> is the name of the interface. <mtu> should be set to a value of 576 or less to comply with RFC877. When an IP datagram is larger than an X.25 packet, the X.25 MORE bit is used to fragment it. When this is unwanted, the MTU should be set to a value equal to the maximal X.25 packet size (128 or so). arp add <hostname> x25 DTE-address Used to map the IP address corresponding to <hostname> to a DTE address valid in the X.25 network. This situation is similar to the method used when routing via NET/ROM. A DTE address is a string of digits, up to 15 in length. x25 inactime [<time>] Sets the time an X.25 VC is allowed to be idle. after this time the circuit is disconnected, and set-up again when a packet has to be transmitted. The default is 120 seconds. x25 status Shows the status of the X.25 Virtual Circuits. x25 disconnect <circ_cb> Disconnects an X.25 Virtual Circuit before the inactime has elapsed. Experimental version compiled using Turbo C on the Atari ST. Most problems seem to be solved. Please report bugs found in this version, that were not present in the earlier MWC versions. 871225.33.PE1CHL.901122: Introduced "tcp ubound <time>" command to set an upper bound to the Retransmission Timer used by TCP. This will clip the timeout to a certain level when the backoff increases, so that there is a better chance that the session stays alive. <time> is specified in seconds, with a minimum of 1800. A value of 0 will turn off this feature (and restore operation as it always was). Fixed "mheard" so that "mheard <iface> 1" no longer hangs the system. Temporary files in the PC Turbo C version are now created in a fixed place, instead of in the current directory. This should also eliminate the failure to remove the tempfiles, that occurred when the current directory was changed between creation and deletion of a tempfile (a bug in the Turbo C library). When the environment variable TMP has been set, it's value is used as the directory name for tempfiles. If not, they are created in the root directory. Put something like "setenv TMP C:\tmp" in autoexec.net to define TMP, and make sure the directory exists. Added trace code for the "broadcast" protocol used on microsats (K8KA/NK6K). 871225.33.PE1CHL.901108: Some source cleanup using prototypes. Possibly the memory leak problem has been solved (a suspicious pointer conversion was tracked down and removed). Implemented ICMP "Host Unreachable" and "Source Quench" message generation by the IP router, when the link protocol detects such problems. This is currently true for NET/ROM, when it has no route to the requested destination or it's output queue is becoming too long. Handling of these messages has always been present in the package. The result of this modification is that you may get the message "Closed (ICMP: Host unreachable)" when a TCP connection is attempted using a route via the NET/ROM network. Added "ax25 txq <axcb> [a|h|l]" command to display the contents of an AX.25 transmit queue. Primarily intended for evaluation of the effects of above mod. When an outgoing NET/ROM link is set up, the AX.25 connection now ignores incoming packets with pid=Text. This has become necessary because more and more NET/ROM clones are sending a connect text to everything issuing a connect, causing the local side of the connection to start a NET/ROM level 7 session. Because of this change, some programs will no longer be able to simultaneously have a NET/ROM- and a normal TNC session with a station running NET. 871225.33.PE1CHL.901014: Fixed minor bugs. Added "memstat dump <file>" to Atari ST version, to dump the heap to a diskfile. This can be used to debug the program, but is otherwise of little use to the end-user. When you see the number of "bytes used" in the normal memstat output steadily increasing, please issue a "memstat dump a:memdump" and send the resulting floppy to me including your configuration files and a short description of the situation at the time this happened. 871225.33.PE1CHL.901007: PA0HZP fixed the packet driver interface in PC Turbo C version Fixed bug in TCP header routine causing garbage printout when TCP headers returned by ICMP are traced. An earlier fix caused complete failure of the returned TCP header handling... Fixed double-linefeed problem in PC (Turbo C) type-a FTP Problem with long lines in AX.25 forwarder (introduced in 900807) fixed. It crashed the system when the message contained lines longer than PACLEN. The forwarder still wraps very long lines (it inserts a CR after 255 chars). Remember that mail is NOT transported transparently. When you want to send files without any modification, it is best to ARC and then UUENCODE them. The "ax25 framesammler" has been removed. This option caused mis-sequenced and damaged frames in AX.25 connections. The command "ax25 maxsammler" no longer exists, so it should be deleted from the "config.net" file. AX.25 BBS SID changed from [NET-$] to [NET-H$]. The BBS gurus say that is more appropriate. 871225.33.PE1CHL.900813: Fixed minor bugs in Turbo C version for the PC. SCC driver now allows complete specification of the clocking mode, so that all modems requiring- or providing external clocks can now be controlled. The value to be written to WR11 of the Z8530 can now be specified as a hexadecimal value after the existing baudrate specification, separated by a colon (e.g. d1200:66). This is not necessary for the standard AFSK modems and the HAPN 4800 baud modem, as these do not require external clocks. Refer to a Z8530 technical manual for more information about the value to be written to WR11. 871225.33.PE1CHL.900807: Fixed a problem in the AX.25 forwarder: when no "r" or "w" line was present in the script after "m" "u" or "b", the forwarder did not proceed. Therefore it was impossible to forward to mailboxes that do not prompt for the subject and/or the message, or don't send a CR after such prompts. This is now fixed, and the following script works: # forward test to personal mailbox c 430 pe1chl-15 w> # now we are connected, send each mail l mS%.0s %s u%s b rOK w> n This sends the S <call> command, the subject, and the entire message in one go, and then checks for a response containing "OK". For more detail about forwarding scripts see the explanation further down this file. The forwarder now concatenates multiple lines from the message into a single packet, whenever the total length of these lines is below paclen, except when the effective window (maxframe) is less than 2. The answer to a often-asked question about NET/ROM: You can use different calls for the NET/ROM on different interfaces (e.g. a PI1 call on 23cm). Specify the calls you want to use in the "attach" commands for the different interfaces. Somewhere after these attach commands the autoexec.net file contains the line "attach netrom$CALLSIGN-9".  When this
line is changed to "attach netrom" (without the callsign), the NET/ROM will use
the interface callsigns on each band.  Make sure that you use a "hidden" alias
(starting with a # sign) for the interlink nodes of the NET/ROM!  The alias can
be separately specified for each interface in the "netrom interface" lines.

871225.33.PE1CHL.900710:

On "exit", the program now waits for the interface queues to drain.  This
should solve problems with param commands placed in "onexit.net".  This often
did not function reliably because the interfaces were reset *before* the
characters were actually transmitted.

Atari version now supports up to 5 async channels.  The 2 params after
"attach asy" (<addr> and <vec>) can now have the following values:

1	 0	ST RS232 port using Bcon* calls (BIOS)
3	 0	MIDI port
6	 0	ST-compatible RS232 port on TT030
7	 0	MODEM2 port on TT030
8	 0	SERIAL1 port on TT030
9	 0	SERIAL2 port on TT030

The old <addr> values 0 and 2 remain valid (for RS232 and MIDI).
Ports 6 and 7 could also be used on an ST with TOS 1.2 or higher, and a user-
written device driver (loaded from the AUTO folder) that supports the Bcon*
routines on BIOS devices 6 and 7.  However, on an ST it is not possible to set
the mode and baudrate for these channels so your driver must set the port to
8bit, no parity, 1 stopbit, no handshake and the correct baudrate.
The driver must also supply a FIFO buffer for transmit and receive.

Unix/XENIX release.  Only supports async I/O for now... (SLIP/KISS/NRS)
To attach a serial line, use an attach command like this:

attach asy /dev/tty1a 0 ax25 ax0 512 256 9600 pe1chl

- the tty device (/dev/tty1a) must be "disable"d, and have read/write permission
for the user running NET.
- the "0" value can be replaced by "r", "c" or "rc" to set the RTSFLOW and
CTSFLOW options for the device.
- the next two fields are the type (ax25) and name (ax0)
- the "512" is the size of the read() buffer used for receive
- the "256" is the MTU (paclen)
- the next fields are the speed (bits per second) and callsign

The program is available in 2 versions, net286 and net386.  The '386 version
can use the select() function internally to improve efficiency.	 This can be
enabled by starting the program using: net386 select=kt
The 'k' and 't' enable the use of select() for keyboard(stdin) and tty devices.
It is also possible to set an environment variable SELECT to this value.
Testing using SCO XENIX386 release 2.3.2 has shown that the select() function
somehow clobbers the tty input and output buffers, so the use of this option is
not yet recommended.

The "dos" commands CD, COPY, TYPE, RENAME etc. are not implemented.  Use a
subshell or a different window to perform these functions (except CD).
Incoming mail is stored in the /usr/spool/mail directory (or to the directory
defined by the NETMAILS).  This directory must be writable by the user running
NET to be able to receive mail.	 The standard mailer can read the incoming mail,
but I do not yet know how to prepare outgoing mail.  Mail stored in the
directory /usr/spool/mqueue (NETMAILQ), using the same format as in the DOS
version of NET, is transmitted by SMTP.	 This directory *must* exist and be
writable for the user running NET.
All configuration files are located in the "current directory" as effective when
the program is started.	 It is recommended to create a user "net", and to put
the configuration files and the program in it's home directory.  NET can then
be started from the .profile, so that logging in as "net" is sufficient.
The FINGER files are stored in a subdirectory "finger", *without* the .txt
extension used by the DOS version of NET.

871225.33.PE1CHL.900623:

Fixed problem that caused crash when an excluded station was connected from
the NET/ROM.

Many users complain about problems with IP and/or selected AX.25 servers.
Please note, that in this software each AX.25 server (defined using
"ax25 port") *must* have a *different* SSID, and that each interface (defined
using "attach") *must* have an SSID *different* from all SSID's used by AX.25
servers.  Different interfaces may use the same SSID, but cross-band
digipeating will not be possible in that case.
The autoexec.net file provided as an example follows these rules, but many
people have modified the SSID's and subsequently had strange problems...
When you use the -0 SSID for an interface, the TNC server *must* have some
other SSID, or IP will not work.  To avoid problems, everybody is hereby
encouraged to use the following SSID scheme (proposed by PE1LGT):

SSID  Interf	Function
-0    All	TNC (normal ax.25 connect)
-1    All	MBOX
-2    144	2m interface
-3    144	mheard/NetDigi 2m
-4    4k8	2m/4800 baud interface
-5    4k8	mheard/NetDigi 2m/4800 baud
-6    All	Conference Bridge
-7    430	70cm interface
-8    430	mheard/NetDigi 70cm
-9    All	NET/ROM
-10   aaaa	interface
-11   aaaa	mheard/NetDigi
-12   23cm	23cm interface
-13   23cm	mheard/NetDigi 23cm
-14   bbbb	interface
-15   bbbb	mheard/NetDigi

aaaa and bbbb usable for other bands, e.g. 28 or 50 MHz

avoid using the -14 and -15 SSID's on bands where a local NET/ROM
is operating.

871225.33.PE1CHL.900611:

Atari version: moved disable/restore handler to trap #7, to allow the use
of trap #5 as a micro_rtx system call.	This is only possible for MWC release
3.0.0 and above.  No testing done with micro_rtx.

Fixed protocol problem that caused AX.25 failures when connected to G8BPQ
nodes.

Some changes made to improve portability.  Attempt made to run under XENIX.
(it compiles okay, but the async port handling still causes headaches)

871225.33.PE1CHL.900530:

TZ or TIMEZONE (depending on machine+compiler) environment variable now gets
a default value GMT0 when it has not been set by the user.
This prevents funny behaviour when "at" is used (default was Pacific Standard
Time).	You can still set your own preferred timezone.

Automatic ticks/second and "at" command removed from Aztec version. The library
function time() returns an unusable value, and the Aztec version reported an
incorrect value of 36 ticks/s because of this.

871225.33.PE1CHL.900522:

Fixed problem in SMTP that sometimes caused the program to crash (when mails
containing very long lines were received)

Also, built extra test in watchdog so that "watchdog ?" no longer causes the
PC to crash.  Now, specification of 0 or a non-numerical port number also
disables the watchdog (instead of clobbering port 0, the DMA controller).
Of course, this still leaves plenty of opportunities to scramble the system
through specification of weird port numbers.  Be careful!

871225.33.PE1CHL.900520:

Fixed problem in NET/ROM level 4.  Sometimes the transport-level timeout timer
did not run while frames were outstanding, causing connection failures under
certain circumstances.

871225.33.PE1CHL.900519:

"watchdog" command now also installed in PC version.  It can only drive an
external watchdog timer.  The syntax of the watchdog command is:
watchdog [<port> <value>]
When specified, the <port> and <value> select the output port to be used for
the watchdog.  The value is OR-ed to the value read (input) from the port, and
this value is output to the port.  Some time later the value is complemented and
AND-ed to the port value.  The result should be that one or more output bits
toggle at each pass through the main loop.
The watchdog triggering can be stopped by specification of a zero <value>.

871225.33.PE1CHL.900513:

New command "tcp active" shows only those TCP sessions that are not in the
"Listen (S)" state.  This gives a shorter list, especially when many servers
are started.

Fixed "from" address in AX.25 mailbox (to provide proper returnpath for errors)

Small changes in SMTPCLI to make it more RFC821-compatible

871225.33.PE1CHL.900506:

Fixed problem in NET/ROM: when multiple NET/ROM level 4 packets were received,
some with the MORE bit set, and the total length of info was over 256, the AX.25
link transmitted packets over 256 bytes in length.  Now these are split in
multiple packets each up to 256 bytes in length.

871225.33.PE1CHL.900503:

Improved NET/ROM handling of failing routes.  When a NET/ROM broadcast is
sent, nodes with only 0-quality routes are broadcast too (with that 0 quality).

Corrected TNC2 emulator message at disconnect.	Must not print callsign.

Improved the FRAMESAMMLER.  Now, the functionality is the same as the Digicom
and RMNC implementations.  It should now be safe to set "ax25 maxsammler 7",
independent of the other station's maxframe.

New command: netrom users
Displays the local NET/ROM users, just like the "users" command when connected
to the NET/ROM.

Atari version only: added "log proc" command, to dump the processor exception
area to the logfile, when it contains a valid exception (bombs) dump.
When this command is inserted after the "log <filename>" command in autoexec.net,
information about a preceeding crash is put in the logfile.
The format is:
Sun Apr 29 15:54:00 1990 exception dump (BP=0005F126)
00000000 00000020 000023ED 00000020 00000013 00000000 00000001 00000001
0004E71B 00020284 00000000 0004B5D8 00FDD1F4 00000000 00001804 000017BC
04FC0B12 000061A0 23080000 180800FC 9DA20000 000A0000 5F360000 03100000
00460000 000000FE
The BP value is the current basepage address.  This could be the same as
the basepage address at the time of the exception, but this is uncertain.
The following lines are a dump of the area at 0x384, see an Atari internals
manual for more detail.
This command is only intended for software debugging and/or bug reporting,
when you don't understand it's purpose, you do not need to use it!

Adapted some source modules for compilation using Turbo C 2.0 on the PC.

Solved problem in NET/ROM causing crash when user aborts pending connection.

netrom level 4 configurable.  just set the "netrom param 4" parameters
to 0 to disable net/rom level 4.  then, the node will always return "busy" when
connected from another net/rom node, and node connects from the net/rom always
fail.

871225.33.PE1CHL.900415:

NET/ROM (level 4 & 7) support completed.
It is now possible to connect other nodes, and to list the connected users.

New NET/ROM commands:

INFO (replaces IDENT)

MHEARD [<interface>]
Show the MHEARD list

PORTS [<interface>]
List available NET/ROM ports (interfaces) and optionally selects
the interface to be used for downlink connects.

When a CONNECT to an AX.25 station is made, an interface is selected:
- For AX.25 connected stations: the interface they are connected on
- For NET/ROM connected stations: the default (LAP) interface
The interface can be changed by a "port <interface>" command before the
CONNECT.

New ("net>" prompt) commands are:

netrom info <filename>
sets the name of the file to be sent after the normal response to the
"INFO" NET/ROM cmd. It can be used to send a station description.
Keep this file small!!	The length should not exceed 1 kilobyte!

netrom interface <iface> <alias> <quality> [lap|uplink]
the extra option "lap" must be given on only ONE of the "netrom interface"
lines to specify the default interface for downlink AX.25 connects.
(Local Access Point).
the option "uplink" can be given on other lines, to specify that
user-uplinks are allowed on that particular interface.	when neither
"uplink" nor "lap" are given, the netrom will not process commands when
connected.

871225.33.PE1CHL.900325:

Again changed the behaviour when REJ is received - probably a conversion to
the newer AX.25 code would be a better idea...

Added an interrupt-latency measurement routine to Atari version, to evaluate
the effects of some efficiency improvements.  This is not intended to be a
feature useful to the end-user.	 The "latency" command is used to switch this
feature on- and off, and to display the results.

871225.33.PE1CHL.900314:

Added capability to write (or append) incoming mail to a specified file.
This can be used to keep a file in the public area uptodate using a
mailing list.  (e.g. hostsfiles, tcp-group mail)
When the first line of a "mailbox" file (mail\xxxx.txt) is of the form:
write c:\net\public\hosts.137
or:
append c:\net\public\tcpgroup
the incoming mail will be written (or appended) to the specified file,
not to the mailbox file.
When a suitable "alias" file entry is made, it is possible to write the
mail both to a file and to a mailbox.

Incoming mail is now marked with the source (when "new mail arrived" is
printed on the console)

871225.33.PE1CHL.900304:

Added a second parameter to "smtp timer": the time to be used when a
mail needs to be forwarded.
Whenever a mail arrives that has to be forwarded the "smtp timer" will be
set to this value and re-started.  This allows a long time between polls
of the mailqueue, while still forwarding mail within reasonable time.
This parameter should not be set too low, to allow more than one mail to
be queued and the SMTP session to be terminated before the forwarding is
started.  Recommended value is 500 seconds.
The main timer can now be set to slightly more than one hour, to set the
time between attempts of the AX.25 BBS forwarding.  Exactly one hour is
not recommended because this could synchronize the attempts with normal
inter-BBS forwarding.  4000 seconds is a good value. (smtp timer 4000 500)

871225.33.PE1CHL.900303:

HOME and END (Atari: Ctrl-Home) keys now move the cursor to the beginning
and end of the input line, respectively.

Fixed SCC speed-setting using "param" for Atari ST (it crashed the program)

Corrected TCP windowsize for ascii-mode transfer of FTP, and SMTP.

871225.33.PE1CHL.900211:

Added forwarding from NET to RLI/MBL-style mailboxes and personal maildrops.
This is kind of a hack, as several conversions have to be made on SMTP
The feature operates as follows:  When mail must be sent to a host, a
check is made if the hostname is of the form <call>.bbs, and if a file
exists in the MQUEUE directory with this name.
If both of these conditions are met, the mail is forwarded using a script
in the MQUEUE\<call>.bbs file.
The script contains lines, of which the first character is an opcode:

#	this line is a comment
c	only as the first line: specifies the initial connect to be made.
this can be a connect to the bbs, or to a NET/ROM node that is
used to reach the bbs.	syntax is like the net> connect command.
s	send string to remote BBS or node.  can be used to send strings
containing connect commands to NET/ROM nodes.
w	wait for specified string to be received.  lines not containing
this string are ignored.  useful during connect and disconnect
phase, to eat banners sent by the BBS.
r	checks for a specified string to be received.
when a line not containing this string is received in the message
transfer phase, it is mailed back to the message sender (it
probably is an error)
while in the connect phase, the box is simply disconnected.
l	marks the point in the script where the connect ends, and the
messages are transferred.  ("loop")
m	defines the line to be sent to send a message. line is used as an
argument to sprintf, with the following string args:
1. message type ("P")
2. destination call
3. destination bbs
4. source call
5. message id
u	send subject of message. line is an argument to sprintf, with
the message subject as an argument.
b	send the body of the message. any text after the "b" is sent as
the end-of-message marker.  the default is ^Z.
n	defines the end of the message transfer lines and the start of
the disconnect phase.  control will branch back to "l" when
more messages are to be sent ("next")

Examples: to directly connect to a BBS (AA4RE-type):

# forward to PA3APN
c 650 pa3apn
w==>
# now we are connected, send each mail
l
mS%s %s@%s < %.6s $%.12s rBetreft u%s rBericht b ropgeslagen w==> n # all done, send "bye" command sB w73 To connect to a BBS via NET/ROM: # forward to PI8EAE via PI8RNI-7 NET/ROM c 430 pi8rni-7 sc pi8eae rConnected to w} # now we are connected, send each mail l mS%s %s @ %s < %.6s$%.12s
rSubject
u%s
rmessage
b
r}
n

Another NET/ROM example:

# forward to PI8ABT via PI8RNI-7/PI8ABT-7 NET/ROM
c 430 pi8rni-7
sc zbl70
rConnected to
sc pi8abt
rConnected to
w>
# now we are connected, send each mail
l
mS%s %s@%s < %.6s $%.12s rOmschrijving u%s rBericht b ropgeslagen w> n As can be seen, the scripts must be tuned to match the exact responses sent by the local BBS, such as the "subject" and "message" prompts, and the final character of the BBS prompt (">" or "}"). It is possible to write scripts that ignore the prompts (just omit the "r" lines and put only a "w>" or "w}" after the body) but this will kill the error checking capabilities of the forwarder. Remember that NET/ROM will cancel a connection when you send something without waiting. Always include an "rConnected" line after a NET/ROM connect command. Added support for "Type of Service" as offered by TCP. TOS is composed of characters 0-7,D,T,R for precedence, low delay, high throughput, high reliability. Low-delay will force UI frames to be used, High-reliability will force connected mode, both overriding the selected default mode (using "mode" command). Other TOS selections may be effective for particular IP implementations and/or subnetworks, but are currently not acted upon by NET. With "telnet" and "ftp" and "start", specify TOS using a 3rd/4th parameter. Otherwise set/change the TOS using the "tcp tos <tcb> <tos>" command. You cannot change the precedence of an existing connection. 871225.33.PE1CHL.900123: Atari ST: Now, tempfiles are deleted immediately after use. netrom exclude <nodecall> [d]: excludes certain nodes from the nodelist. The "d" option removes a call from the exclude list. This can be used when it is known that a node has it's parameters set up in such a way that a connect will not be possible because a retour route can not be established. (so-called "Belgian parameters") 871225.33.PE1CHL.900113: Swapped Rcv-Q and Snd-Q in ax25 status output to make it consistent with other status listings (session, tcp status) Fixed an error in sccvec.asm, that caused malfunction in large model on PCs. Atari ST: Now compiled it using MWC 3.0.9 instead of 2.1.7. A savings of 1.9% in code size! 871225.33.PE1CHL.900109: Fixed problem in FTP client causing a crash when the user typed "type l". This caused a BUS ERROR because "type l" needs an extra argument and a NULL pointer was dereferenced when this argument wasn't supplied (attd by SM0IES) 871225.33.PE1CHL.900107: Added support for the DRSI PC*Packet adapter in the SCC driver. (hardwaretype = 08) Added support for external transmit/receive clocks in the SCC driver. Use "ext" instead of the baudrate to get RTxC=RX clock, TRxC=TX clock. With the SCC driver in AX.25 mode, a new speed can now be set using "param <iface> speed <speed>". This only works when internal clocking was specified during init, and it cannot switch between internal and external clocking. In AX.25 mode, the transmitter can be disabled using "param <iface> tx n". This is useful for unattended stations (when interference reports have been received by the control operator he can issue this command via rcmd). ARP entries added to the table as a result of ARP packets received will now be marked with the interface of their origin. It is no longer assumed that an address received on some interface is also valid on another. Interfaces may be specified in arp add, drop and publish commands using the syntax "<type>:<interface>". e.g. arp add pe1chl ax25:144 pe1chl-2 This was primarily done to solve the problems that occur when 4800 baud and 1200 baud modems are used in parallel on the same frequency. Now it is no longer necessary to enter fixed ARP entries for 4800 baud. The selected method of specification of the interface keeps the "arp" command compatible with earlier releases of the program. Added the W9NK TCP/IP mailbox. This is an AX.25 server that allows AX.25 stations to send SMTP mail to the owner of the box and to other TCP/IP stations to which the box owner can forward SMTP mail. The mailbox accepts forwarding from the de-facto standard BBS systems. To use the mailbox, initialize AX.25 port #6, and start the ax25 mbox server: ax25 port 6 conn pe1chl-1 ax25 start mbox \net\finger\rob.txt The (optional) filename specified after the start command specifies the "info" file for your box, sent to the connected user when he types an "i". Incoming mail is only routed (to the local user or to others) when the SMTP timer ticks. You'll have to set your "smtp timer" to a reasonable time to receive your mail without long delays (say, up to 7200 seconds). Added support for multi-channel KISS TNCs. These boxes encode the radio channel number in the upper nibble of the KISS type byte. To use this feature: Attach the physical device using "attach asy" or "attach com", as before. The channel number zero will be accessed using the defined parameters. To access channels 1 to 15, use: attach kiss <parent> <channel> <label> <mtu> <call> The <parent> is the physical interface attached before. <channel> is the channel number used by the TNC (1..15). e.g.: attach kiss ax0 1 ax1 256 pe1chl-7 This will make the second channel of the KISS TNC available as "ax1", with a maximum IP packet length of 256 and a callsign "pe1chl-7". Modified "copy" to allow device specs in Atari version. Hope it does not break down on other versions. 871225.33.PE1CHL.891225: Added "x" option to "tput" FTP subcommand to update attributes of a file after it has been sent. This ensures the file will have the same date/time and "read only" status on the destination after a transfer. This is implemented using an "XATR" FTP command that is sent after each "STOR", containing this information in an MSDOS specific format. For now, this is only implemented in the Atari ST version. Also made response on "MKD" according to RFC959. "at" command corrected, now it always uses the local time. Added "netrom route drop * <iface> <neighbor>", to drop all NET/ROM routes via specified neighbor (e.g. when you know it has been taken out of service). This only affects non-permanent routes. Added NET/ROM level 3 parameter "maxqueue", specifying the maximum number of AX.25 packets allowed on an outgoing AX.25 queue. Any packets that have to be queued above this limit are dropped, and an alternative route to this destination is chosen, if possible. Added NET/ROM level 3 parameter "maxfail". When more than this number of link failures occurs on an AX.25 link to a neigbor, all non-permanent routes via this neighbor are dropped (until we next hear a broadcast from him). This should limit the forever-retrying on dead links (dead nodes). Set "maxfail" to zero to disable this feature. 871225.33.PE1CHL.891217: Added "tput" FTP subcommand to put a (sub)tree. Before using it, the target directory must be set using "cd", and the file type (usually I) to be used for all the transfers should be selected. Then, a "tput <dirname>" or a "tput <wildcard spec>" will start automatic creation of subdirectories and transfer of files. The session should remain selected to allow display of messages and execution of further commands. Use "abort" to terminate the process. DON'T change the local "current directory" when an incomplete pathname was specified for the "tput". Solved some problems with pathnames specifying a drivename and/or the "." notation for current directory. Implemented "netrom tcpip interlink" to let the NET/ROM pass only visible nodes and #TCPIP. (#others are not broadcast, but appear in the nodelist) This is similar to the "TheNet I-version", but without the disadvantage of inhibiting TCP/IP traffic. When "netrom nodefilter exclusive" mode is selected, this provides very good control over the number of nodes in the broadcast and the NET/ROM traffic through the node. Added extra delays in the SCC driver (MSDOS version only) to allow it to run on very fast machines. 871225.33.PE1CHL.891209: NET/ROM nodefilter mode "exclusive" added. Works like "accept", but additionally no NET/ROM transport packets are accepted from stations not listed. Only stations in the "netrom nodefilter" list are accepted for NET/ROM traffic. A NET/ROM retour route is now also created when only 0-quality routes exist to the sender (this was only done when no routes existed). Experimental NET/ROM level 4 code added. Accepts connects from other NET/ROM stations but cannot connect to others. "netrom status" lists the current connections. Changed "isdir" to fix a problem with the MSDOS/MSC version (it said that a device was a directory, so "copy file prn" would fail) Made configuration depending on memory model and compiler, to ease maintenance and prevent distribution of unworkable versions (like MEDIUM MODEL with lots of memory-demanding functions) 871225.33.PE1CHL.891031: Atari version: memstat <number> mallocs and frees the specified number of bytes, to get these from the OS in one big chunk. Old versions of the system have trouble with programs making a lot of allocation requests, and the Mark Williams C runtime makes requests in small (2K) units. Putting "memstat 30000" or so at the beginning of autoexec.net will acquire some memory to start with. The <number> is a long integer. Fixed problem in netrom route handling that caused inconsistency (and hangup) when memory could not be allocated to store received routes (see above). Fixed broken "dir" command (wouldn't list directories after "improvement") 871225.33.PE1CHL.891001: route info <ip addr> command will print the IP route to another host improved handling of directories in DIRUTIL.C (because of TurboC problems) 871225.33.PE1CHL.890924: Greatly improved error handling in SMTP server (disk full, out of mem) Also introduced timeout in server (waiting on commands) Introduced "at" command to run NET command at certain time-of-day. Syntax is: at hh:mm command [params] example: at 18:00 stop ftp "at" command without params prints a list of scheduled commands. Only one command can be started at each time. Use "source" command to start a sequence of commands. (stored in a file) stopping FTP/SMTP/RCMD servers does not produce a log message anymore improved handling of AX.25 excluded calls. these cannot be connected anymore, also from bridge, netdigi etc. 871225.33.PE1CHL.890918: Minor change to make NET/ROM "param" command output compatible with NET/ROM. (some programs interpret this output for a better display) Fixed problem with timeouts in SMTP client (it failed when a mail took more than "tcp timeout" time to transfer) Log now flushed every time, log without params doesn't flush anymore Added watchdog timer (called from timer interrupt, reset from main loop). The "watchdog" command sets timeout values (in seconds) for 2 cases: the first value is used while in NET, the second when executing a SHELL. When a 3rd arg is given, it's (HEX) value will be put out on bit 0 of the printer port every time the watchdog is reset, followed by a zero value. This can be used to trigger an external (hardware) watchdog timer. This should not affect printer operation, as no strobe pulse is sent, only a change on the datalines. (for now, watchdog functions are only present in the Atari version) Solved problems with "*" and "?" in filenames in ftpserver 871225.33.PE1CHL.890910: Made a lot of small changes to compile the package with Turbo C on the Atari. (Mainly suspicious constructs warned by Turbo C, and MW C dependencies) It does not yet work correctly when compiled with Turbo C. Introduced "onexit.net" file, executed during "exit", just after all connections are closed. Can be used to send param 255's or copy files. Added an extra check in the Atari SCC driver to detect a permanently-low IRQ line from the SCCs. 871225.33.PE1CHL.890825: Yet another dup_p replaced by copy_p: the SCC driver transmitted garbage when an AX.25 packet was acknowledged when a retry was already sitting in the SCC transmit queue... (this could also happen with KISS, but it is less likely). Now, in lapb_output() a packet is copied to a new buffer instead of dup'd. (same in recover()) 871225.33.PE1CHL.890814: Fixed the last (?) flow-control problem in the NetDigi (and possibly other AX25 services driven by the transmit upcall). It now only asks for as many bytes as will fit in the maxframe window. This is the "final PC release" distributed in Aztec, MSC Middle and MSC Large. 871225.33.PE1CHL.890812: Corrected a problem introduced in the router in 890809 (only /32 and default work), and made the router cache more effective. PC: Fixed bugs in the SCC interrupt handler. This also affected the handler for the 3c500 in the Aztec version of 890812. "date" command now can set the date and time in MSC-compiled versions for the PC. 871225.33.PE1CHL.890809: Finally tracked down a problem with ctime(), which only appeared using MSC. It loses it's terminating \n under some circumstances, which caused funny results in smtp. Added yet more checks for NULL pointers (from malloc). Now prints "No Space!!" when no buffer is available for a ping (with length >0). This can cause mesages when a repeated ping is set up. Now I know how TCP sessions could stay in CLOSED state. Fixed FTP state change upcall handler, so that it should not happen anymore. MSC-compiled version (without PACKET and 3C500 drivers) can now be compiled in LARGE model on the PC. This should solve the problems of running out of heap space, at least when you have enough free memory in the machine (or section or window). This version is about 30% bigger and slower than the standard MIDDLE model version. 871225.33.PE1CHL.890731: Set a timeout on CLOSED state (TCP). A session was left in this state after an FTP (directory) with all timers stopped, but I don't know how it happened. Changed appearance of NET/ROM "param" command output to be compatible with NET/ROM 1.2 871225.33.PE1CHL.890725: Converted all .asm files to pmacros.h usage. Should be usable with MSC, TC and Aztec. NETROM4 define introduced to omit all NET/ROM level 4 support. Changes to pc.c, dirutil.c to make it run under MSC 5.1 Fixed trouble with receiving from Fossil (again) 871225.33.PE1CHL.890722: Fixed problem with hostname translation in 870720. It was only printed in the incoming session message, not in the session display. Also scanned all sources for occurrence of inet_ntoa and psocket, and fixed declarations. Fixed silly bug introduced in 890720 (resolve did not work on PC) PC version: Fixed problem with TNC2 emulation on COM port (device number off-by-one) PC version: Fixed hardware handshake problem for KISS tnc's (again...) TNC2 emulator can now XON/XOFF when FOSSIL driver is used Fixed problem in FTP login sequence (attd by dg2kk) 871225.33.PE1CHL.890720: All versions: Included an ip_address-to-hostname translator supplied by pe1jlj (after some modification). Incoming telnet sessions are now labeled with the hostname instead of the IP address. All versions: Added -f flag to "delete", "rmdir" and "mkdir" commands, to make these silent in case errors occur. Useful to clearout temp directories and lockfiles from AUTOEXEC.NET, without getting complaints about the non-existence of these files. Use it like "del -f${NETMAILQ}*.lck".

SCC driver: When DCD is active during initialization of the driver, the receiver
is enabled immediately.	 This used to happen only on inactive-to-active
transitions of DCD.  (Added after request from pa0hzp who was testing his board)

NET/ROM: When a NODES broadcast comes in which contains routes to a certain
destination with a better quality than the currently active route, that better
route is again selected, even when it was dropped before because of a link
failure.  This is done to prevent infinite re-routing of NET/ROM links in case
of a temporary link problem (there used to be no mechanism to re-select a
nodes regularly, these nodes should be filtered out to prevent repeated tries
of the direct links to these nodes.

PC version: Added "date" command. Prints or sets the system date and time.
date setting does not work yet.

PC version: Included the possibility to emulate a TNC2 on an external COM port.
Algorithm is now as follows: when COM1-COM6 is specified in the
"ax25 start tnc2" the code checks for the presence of an INT14 handler for
that port.  If it finds one, it will start the emulator on that external port.
MBBIOS or FOSSIL should be loaded for the selected port.  (the standard BIOS
handler will lose big time!!! no interrupt handling in the ROM BIOS...)
The baudrate of the external port is fixed to 9600 (8 bits, no parity, 1 stop).
When a port above 6 is used, or there is no INT14 handler present, the code
continues to use the "virtual COM port" accessible from the other doubledos
section to emulate the TNC.
NOTE: All this only works in versions compiled to support COMBIOS/MBBIOS/FOSSIL
instead of the built-in "asy" device.

PC version: timeout on transmit character emulation routine (virtual COM port)
increased from 5 to 25 seconds, to allow for very bad AX.25 links to a BBS
running on TNC2 emulation.  The WA7MBL bbs does not test for timeouts, and
therefore drops characters when one occurs.

Attempted to make slightly more heap space available to those poor PC users, by
putting more common text strings in variables (so that they appear in the 64K
datasegment only once).	 Now 31552 bytes of heap, hooray.
There is still more than 16K of fixed text in the data segment...

871225.33.PE1CHL.890622:

ping length [<value>]
default <value> is 0
defines the number of extra bytes to be send with each ICMP echo reply.
useful for testing path and modem quality.
maximum value <value> is 4095

ax25 maxsamml [<no# frames>]
default <no# frames> is 0
A framesammler algorithm has been implemented. Non-sequential receiving of
frames on AX.25 connections is possible now.
This option is disabled when <no# frames> is 0
100 % data integrity is only guaranteed when using this option with a
<no# frames> value wich 4 or lower and if the other side has a Maxframe
of 4 or lower.

Atari version: added "screen" command, to select NET's method of screen output:
screen bios	use BIOS call Bconout
screen direct	use direct writes to screen memory (fastest)
screen tos	use standard output (default)
"screen direct" only works on standard resolutions, not on "big screens".

871225.33.PE1CHL.890515:

new ax25 server: tnc2 emulator
This server allows you to emulate a TNC2 running TAPR 1.1.6 code.  This
TNC2 can then be used by client programs running in the other section when
using doubledos. (from the 890720 version it is also possible to connect the
TNC2 emulator to an external COM port)
This allows you to run a BBS concurrently with NET, using the same hardware
(KISS tnc, SCC board or whatever) to interface to the radio(s).	 Also, other
packet programs can be run this way, as long as they interface with the TNC
using INT 14H (as is the case with MBBIOS, COMBIOS etc).

The emulator tries to mimick the functional aspects of the TNC2 software as
well as possible.  Features not supported are:

- setting of operational parameters like timing, async port settings etc
- monitoring and it's associated parameters
- calibration and testing
- multiple streams (users)
- battery backed-up RAM

Most other features are supported, including CONVERSE and TRANS mode,
input line editing, and even KISS.

There is one extra command, not found in TNC2 software, that is introduced
to compromise the efficiency of CPU usage with the requirements of the client
program:

KISSRX	<number>

This command selects which frames will be "received" by the emulator when it
is running KISS mode (KISS ON, RESTART).  This can be set using param 6 when
KISS mode is already selected.	The meaning of the number is:

1 receive frames addressed to "MYCALL" only (the call of ax25 port 5)
and also to QST and NODES.  This is the default value.

To use the tnc2 emulator:

Run NET under DoubleDOS.  NET needs about 230K.	 Find the exact requirement
by increasing the section size in small increments, until the free space printed
by the "memstat" command immediately after startup does not increase any more.

- define ax25 port 5
- start ax25 server "tnc" with parameter list.	each parameter is of
the form comnumber=interface.	 example: ax25 start tnc2 3=144 4=430

After starting NET, run "NETCOM" in the other partition, specifying the TNC
port numbers (from above command).  You can also specify the /b flag, which
enables a buffer between the client program and NET.  Some BBS programs don't
like this, experiment with it.	Example:
netcom /b 3 4

Therafter, you can use "YAPPB", "WA7MBL BBS" etc in that partition, configured
for COM3, COM4 etc.   Do NOT use any of the real com ports (COM1, COM2).

If you have more than 2 COM ports in the machine, use higher numbers.
NET allows up to COM255, but the packages you want to run may be more
limited.  For first experiments, use YAPPB <com#>.

Variables now supported in NET commands. see environ.txt for details

FTP server has an extra feature: a lister for the contents of .ARC files.
usage: (ftp subcommand!)
dir <filename>.arc
This also works with the dir command at the net>_ prompt.

changes for PC version:

Async interface for kiss and slip devices is now under control of an external
device driver, an interrupt driven int 14 handler (COMBIOS). This can be either
MBBIOS or FOSSIL, well known with w0rli & wa7mbl sysops, and FOSSIL is from the
world of fido / opus phone bbs systems.

mbbios and fossil.

to attach a COM port use:

attach com <number> slip|ax25|nrs <label> <mtu> <baudrate> <handshake> [<call>]

<number> is the COM port number (1, 2, ...)
slip, ax25 of nrs specificies the protocol to be used (ax25=KISS)
<label> is the name of interface (144, 430 etc)
<mtu> is the maximal transmission unit (paclen) (256 for ax25)
<baudrate> can be selected from these values:
110, 150, 300, 600, 1200, 2400, 4800, 9600, 14400, 19200,
28800, 38400, 57600, 115200, 330400.
The high baudrates (> 9600) mean "asking for problems".
A standard XT can reliably run up to 4800, a fast AT can
probably manage 9600 baud.
<handshake> is a letter "n" of "h", and specifies the use of hardware
handshaking (RTS/CTS) on the COM port.	This would normally not
be used, but it is rumored that AEA TNCs which exhibit the
wellknown KISS bug would run more reliable when hardware
handshake is in effect.
<call> is needed for ax25 and nrs interfaces only. It is the IP call
used for this interface.

Examples:

attach com 1 ax25 144 256 4800 n pe1chl-2
attach com 2 ax25 430 256 9600 h pe1chl-7
attach com 3 slip sl0 1024 9600 n

Remember that the translation of COM port number to hardware address is now
performed by the MBBIOS or FOSSIL. These programs must be properly configured

starting this release, smaller versions of NET.EXE are available for those
stations who don't need all the features of the big NET.EXE.
these versions do not include the NET/ROM code, the rcmd server and some of
the more exotic device drivers.	 they are suitable for packet-radio-only users
that do not want to leave their system on unattended.

871225.33.PE1CHL.890206:

netrom supports route [<call>] command.

scc driver included. see sccdoc.arc file on this disk.

buffers [<no. buffers>] command included.
Specifies the number of receive buffers for the scc interrupt handler.
Belongs to the scc driver.

shell can now start a dos program. (instead of an interactive shell)
syntax:
shell /c <program name> [<program options>]

In fact, "shell" starts the program identified by the environment variable
COMSPEC, and passes it the parameters you specify. See your DOS manual for
an explanation of the /c option (which is interpreted by COMMAND.COM).

On the Atari ST, the shell program to be run is specified using the special form:
shell =programname

871225.33.PE1CHL.890115:

==  Known bugs solved in this version:

Scanning of HOSTS.NET file improved.
usage of tab and space characters is allowed now.

Record <filename>
now also works with ftp and finger sessions.

Control characters are echoed to the screen as ^<character>.
They are however still sent as control characters.

With ^V it is possible to insert ANY control character.
Type ^V, followed by the control character you want to insert.
(ref. tnc2)

SMTP server is now capable of automatic multi-hop forwarding:

When the smtp timer counts to zero, or a smtp kick cmd is given, the mail
is first sent to pa0zzz (the local station).  The mailer chops off the
rightmost call and @, replaces the rightmost % by an @, and puts the mail
in \spool\mqueue.  This process repeats itself (when everybody is working
with this version of NET) until john@la4zzz is reached.	 Of course there
has to be a path of known stations running NET 24h/day.

871225.33.PE1CHL.881202:

== known bugs solved in this version:
AX25 connects were lost at midnight (00.00: *** LINK FAILURE)
(only pc version)
The control block adresses are now 4 hex digits again (no ffffedca any more)
Temporary files are erased again. (no "eaaaaa22.591" files in various dir's)

== new features:
Log command now also flushes the buffer contents to the logfile.
Type command has an extra parameter: the number of lines to type
example:	cd \spool  ;go to spool dir
log	   ;flush buffer to file
type net.log -10  ; list last 10 lines of net.log
to see who is doing what.
Memstat command gives the # of malloc errors not at the beginning, but at
the end of the list

New server: rcmd (remote command server)

format:
start rcmd [<port> [<key string>]]

defaults: <port>=459, <key string> = null
e.g.: start rcmd 459 qwert2yuiop5asd8fgh6jkl1zxcvbnmqwe5uiashjkzxcbn1234567890

(to be inserted in autoexec.net file, together with the

accessible for remote stations with:

telnet <hostid> <port>

e.g.: telnet pe1zzz 459	  (from net>_ command line)

the rcmd server then returns three series of random numbers, five each. One of
these series must be returned with the corresponding characters from the
key string.

e.g.:suppose that the server was started with this line:

start rcmd 459 gqweGrerDFtDyuWioGpRaYsdYJfJhIKjLzxYcfRGwevbFEnSWmlcfkHtldrDC

then, a session could look like this:

telnet pe1zzz 459
rcmd 871225.33.PE1CHL.881202 pe1zzz.ampr
58 24 3 18 32, 59 13 26 12 25, 71 23 12 8 44
ryJDY				   /* we choose the second series of numbers*/
pe1zzz.ampr net>_

At this point, the user is in complete control of pe1zzz's NET system !
This can be very dangerous, because there are dos commands like delete and
chdir, and your complete system is within reach.  It is important to make
the key string as long as is practical, for example 50 or 60 characters,
all for the sake of safety. In theory it is possible that someone
monitoring the channel can, after a while, reconstruct the key string.
Therefore it's a good practice to change the string after a while.  This
server is specially meant for automatic stations without operator. Don't
start the server just to impress other people !. And whatever you do:

CHANGE THE DEFAULT KEY STRING !!!!

871225.33.PE1CHL:

Long forgotten history....