HamCall (October 1991)/HamCall (Whitehall Publishing)(1991).bin/amateur/g8bpq/changes.bpq

Text File | 1990-10-14 | 11KB | 293 lines

    Recent changes to G8BPQ switching system (TheNode)


Version 3.20 29/5/89


Displays FRMR counts in stats (should have been in 3.17, but wasn't).

Obeys NOMODE setting following a C SWITCH. Mainly so RLI's Netwalker will work.

Allows a 'via' string in UNPROTO line of config file.

Tidies monitored data by removing surplus newline at end.

Includes new TNC command MMASK, which controls selective monitoring of
radio ports. A port will only be monitored if the corresponding bit in MMASK
is set, ie to monitor only port 1, set MMASK to 1, to monitor only port 3,
set MMASK to 4, to monitor ports 1 & 3, set MMASK to 5. Default is FF hex.
Setting MMASK=0 is equivalent to setting NOMON ON, so NOMON has been withdrawn.


Disconnects user if he gets more than 6 'Invalid Command' responses in
succession. This is primarily to catch the situation where a BBS somehow gets
attached to a Node, and swaps 'invalid command' messages for ever.

Allows the ROUTES and NODES lists to be saved whilst running, so that they can
be restored on reload. See separate file SAVENODE.DOC for details.

Improves timeout of 'hung' sessions. This may get over the problems where
ports become unusable after disconnection in certain circumstances. It does,
however, mean that the session timeout (normally 15 mins) is applied to host
sessions.

Addition of a system to allow multiple applications (servers) to run above
the code - see file APPLS.DOC for details.

Support for digi-ed routes, and locked nodes, has been added, but not tested -
partly because I dont really understand how NET/ROM copes with having both
direct and digi-ed routes to the same neighbour. If anyone wants to use the
facility, send me a message (or a 'phone call).


3.21 2/7/89 (Released for beta test only)



Fix letter 'J' in CWID (was sent as dah dah dad dit).

Fix possible overflow of ROUTES list.

Fix loading of BPQNODES when over 4k, or with calls with 2-char SSIDs.

Fix C BBSCALL/BBSALIAS from node.

Fix error in L4 NAK/RETRY logic, which could cause session to hang.

Allows commas (as well as spaces) between calls in a 'via' string.

Allows a separate callsign/alias pair for each radio port for level 2
access only.

Fixes problem with putting 'via' in lower case in CONFIG file UNPROTO.

Removes 'RTS RAISED' message from TNC emulator.

Removes  redundant spaces from end of 'connected to' msgs.

I've made a start at long last to improve the documentation - the first
stage is a definition of the node command set (COMMANDS.DOC). I would
appreciate any feedback on it, and what you would like to see documented
next!


Version 3.22 10/7/89


Fixed initialisation of PORT ALIAS - was rubbish in 3.21.

Fixed BPQNODES to handle calls more than 7 bytes long.

Fixed problem which could cause loss of the initial BBS 'signon' message,
and possible FRMR conditions, when receiving a repeated SABM (due to UA
being missed, or channel conjestion).

Relax callsign validation rules to allow '/' chars in callsigns.

Drops the CTS bit in the emulated modem status register when too many
messages queued for transmission.

There was an error in APPLS.DOC - application names have a maximum length
of 12 chars (not 2).


UFQBBS param is replaced by HOSTMODE - see BPQCFG.TXT for details.

Attempts to support AA4RE. I've run it, but am not familiar enough with
it to be certain it's doing all it should. Please let me know how you get
on if you try it. See AA4RE.DOC for configuration info.


Version 3.23 15/7/89

Version 3.22 introduced a couple of bugs (User applications and UFQ BBS
stopped working) - hence 3.23!



Version 3.50 23/9/89


Improved format of PORT definition in CONFIG file. See PORTS.DOC

Supports Sharing of hardware interrupts (with suitable cards) and
Quadram Quadport.

Allows data to be loaded into EMS memory. This saves up to 64k below
the 640k limit, but prevents the use of EMS for anything else. This is
mainly to help AA4RE users who are very short of store.

Acts upon the max connects parameter passed from AA4RE.

Attempts to prevent overrun errors on AA4RE monitored data.

Supports KISS params (TXD, SLOTTIME, PERSIST, FULLDUP).

Accepts 's' qualifier on CONNECT command, which causes the user to (S)tay
connected to the node from which the connect is issued, if the circuit is
closed by the far station.

Accepts a set of additional 'COMBIOS' commands, primarily for use with
user written application programs. See INT14.DOC for details.

Allows you to conect to ALIAS-SSID, thus forcing an L2 connect, even when
ALIAS is 6 chars long.

Ignores connect request (SABM) from a node if it is locked into the ROUTES
list with a quality of zero.

Allows you to limit incoming connects on a particular port to a list of
up to 8 calls, specified with the VALIDCALLS parameter of the PORTS
configuration. This is mainly for use for HF BBS ports, but could be used
to control access to a dedicated truck channel.

Connects to PORTCALL or PORTALIAS will now go to the BBS if NODE=0



Version 3.51 7/10/89

Sorted NODES List (at last!)

New versions of BPQNODES and BPQDUMP

TX Underrun counter should now work.

Code added to recover from 'lost interrupt' problems.

Change to USERS display to identify circuits shutting down.

TNC Command Handler no longer returns 'Eh' to blank input lines.

NEWMODE Now correctly spelt (but still ignored!).

Correctly handles SENDPAC characters above 1F Hex.

Accepts '*** LINKED to CALLSIGN' command to node to change callsign used
for connections from the node. This is to allow the real user's callsign
to appear in connects from Gateway programs running above the switch.
The normal rules for SSID processing apply - ie if you enter a call of say
G8BPQ-2, the call used in downlink connects will be G8BPQ-13. Connects
to other nodes will contain the call as entered.

TNC Command handler now accepts Y/N in as equivalents of ON/OFF


See file WARNING! for details of problems with AA4RE.



Version 3.53 Beta - 9 Dec 89

(Several subversions of 3.52 were issued for test only)

BYE (Disconnect) command added to Node.

ROUTES display includes indication of active neighbours.

NODES XXXX command returns 'Not Found' rather than the full NODES list
if XXXX is not a known Node Callsign or Alias.

ROUTES display has option which gives additional statistics (R *) -
the fields currently displayed are I frames sent, I frames re-sent,
Route MAXFRAME, Route FRACK, Time NODES last heard. An option to set
MAXFRAME and FRACK independently for each neighbour has been added to
the ROUTES section on BPQCFG. This is the first stage of an attempt to
optimise L2 parameters automatically.

Stack Switch in INT 14 handler reinstated, in an attempt to track down
problems with AA4RE. It was removed between 3.23 and 3.50.

Fixed problem with Max Connects being set to zero if running UFQ

Control Chars removed from monitored data - they could cause problems with
AA4RE.

The 'All ports in use' message now includes the requested application.

NODES broadcasts are now fragmented, with a 10 sec delay between each
packet. This is designed to reduce QRM, as well as load on buffer pool
(and TX power supplies).

Fixes TXDELAY and SLOTTIME problem with KISS TNCs on QUADPORT cards.

FLOW command is now implemented in the TNC emulator. The code is rather
crude - in particular Echoing is suppressed, so if you want to use FLOW
you should set ECHO off, and use local echoing.

Ignores SSID's when checking calls in VALIDCALLS

STATS display now shows the percentage busy of each port, averaged over
60 secs. Two figures are shown, the first is time transmitting, the
second is the total busy time (including transmitting). This feature
only works with HDLC cards (not async links).

Includes code to help debug problem of lost buffers. If your system
loses buffers whilst running, I would appreciate a BPQDUMP taken when a
significant number are 'missing', and maybe I'll be able to track down
the cause.

Includes a facility to reduce the broadcast quality of a node on a
particular port, if its best Neighbour is on the same port. This is intended
to assist in arranging that, with a multiport system, packets will tend to
be send out on a different port to that on which they arrive, thus reducing
the chances of a collision of a frame with the ack for a previous one. See
enclosed XXXX.msg for more information. Parameter is QUALADJUST, and is the
percentage REDUCTION in quality, so zero has no effect, 100 would reduce
transmitted quality to zero.

Attempts to detect and recover from 'hung' async ports. These result in
loss of the buffer pool, usually with the KISS TNC 'STA' led on. Recovery
attempts are counted in the 'FRAMES DISCARDED' stats (which were only used
for HDLC ports). There is a finite (though small) chance of chars being
lost from, or duplicated in, the frame being sent, possibly leading to
corruption, so if you get a significant number of recovery attempts (say
more than 1 per day) please let me know.

Supports CTEXT. This is sent to stations who connect to the NODE using
the NODE Alias at level 2. Connects to the NODE callsign are not affected,
to minimise the impact on BBS forwarding systems.

CR char added to front of '*** DISCONNECTED' message, to try to reduce
hangups, especially on RLI systems.

Stops sending an L4 disconnect with zero session ID if a connect is aborted.

Extends FRACK when using digis by the normal (2*n+1) multiplier (used to be
n+1)


Version 3.53 23 Dec 89.

The following changes were made after the beta test version was issued, so
havnt had quite so much testing!

Correctly responds to I frames with the 'P' bit set - it used to send
RR, now sends RR(F).

Sets the 'P' bit on I frames if at the L2 window, or if no more to send.
This speeds things up considerably on good links, as it avoids the RESPTIME
delay.

L4 code now sends an immediate ACK when it has L4 window of frames unacked.
This again speeds things up on good/fast links, as it avoids the L4 RESPTIME
delay. The improvement on hardwired 9600 baud links is spectacular!

First release of Remote Sysop facilities - see SYSOP.DOC


Stop Press!

It has just been suggested it me that running my system on a 286 machine
using a disk Cashe program in Extended Memory can cause the occasional
'hang'. As the 286 has to be reset to get back from protected mode when
accessing extended memory, it is just possible that interrupts could be lost.
If you are running such as system, and have a reliability problem, it would
be worth trying without the cache for a while. Please let me know the results
of any such tests. (Thanks to G6FCI for this info).


73, de John Wiseman, G8BPQ @ GB7DAD

ps. My postcode is wrong in some documentation - it should be NG3 6AE.