.4 progress

[dan%cs.wisc.edu%ucsd.edu@Sdsc.BITnet]

   Yes, Virginia, there is going to be a .33 alpha w9nk.4 release of
the increasingly creaky and overburdened pre-nos NET.  Not only will
this one be equipped with the usual .33 alpha features such as
the mysterious vanishing upcall and the AX.25 connection with a mind
of its own, but .4 will have (ta da!) NET/ROM transport.
   With NET/ROM transport support, AX.25 stations which try to connect
to you via NET/ROM will get your mailbox (if it's turned on) instead
of a busy signal.  Also, you will be able to initiate and accept chat
sessions via NET/ROM.  For example, if you want to use a BBS that is
reachable via the W9XYZ NET/ROM, you would type:

        netrom connect w9xyz

and wait for the "Connected" state message.  Then you'd type:

        c w9wi-1

or whatever the name of your BBS was.  So, you don't have to connect
to your local NET/ROM to use the NET/ROM system; that is taken care
of automatically.
   If an AX.25 user connects to you via NET/ROM, then the mailbox and/or
chat session will give that user's callsign as the originator, not the
callsign of the NET/ROM node he is using.
   The actual benefit you get from this new feature is, I've got to
tell you, all out of proportion to the work it's taken to develop it.
I mainly did it for the following reasons:

        1) I'm tired of explaining to people why their friends have
           to use digipeaters to connect to them.

        2) I thought it might help advance the cause of TCP/IP to
           have all these funny NET/ROM nodes that, when you connect
           to them, say, "Welcome to the TCP/IP mailbox!"

        3) I wanted to demystify NET/ROM by showing that someone
           could implement it in a couple weeks.  Take that,
           Software 2000!

        4) I still have hopes, when the NOS version comes out, of
           expanding the mailbox into a more full-featured mail
           and communications system.  NET/ROM transport will
           give easy regional access to it.

        5) It was there.

   The implementation is based on protocol 6 from Tanenbaum's book
on computer networks, not on TheNet.  It's all completely original,
from-scratch code.  The source code will accompany the release in
case anyone wants to check me on this.
   The bad news is that it's not quite finished.  I still have to
add a bunch of user commands, put in flow control (CHOKE), which is
kind of nasty, and convince myself that the protocol implementation
really works.  Then I'm going to think real hard about whether to
release it without doing adaptive retry timers.  Native NET/ROM
transport uses a fixed transport timeout interval; this is one of
the reasons why NET/ROM networks congest so easily, and why NET/ROMs
mysteriously disconnect from you during busy periods.  I'd like to
do adaptive round-trip timing like TCP, but I'm not sure how to do
it with a sequenced packet protocol where every frame in the send
window has its own timer.  Anyone have any advice?
   .4 will also have a minor improvement to the mailbox.  The
X-BBS-Msgtype: line will be added to SMTP mail headers when the
'S' command has an extra letter (e.g. 'P', 'B', 'T') and will
contain that letter in upper case.  This was done in response to
complaints from those demanding Israelis (just kidding, guys) who
need to be able to reconstruct Bulletin-IDs from the Message-ID
line.  This way your program can tell if the message came in as
a bulletin, and pull the appropriate information back out.

   73, Dan