🚀 go-pugleaf

In-Reply-To: harvard!lll-crg.ARPA!hoptoad!tim's message of Mon, 22 Sep 86 12:56:08 PDT
Subject:  InterBridge + FastPath = ?

(sorry if this is duplicate - it bounced back)
 
I think I've come to the bottom of the problem with the fastpath.
You are correct, the RTMP stub in each mac remembers the last bridge
it got an RTMP message from, and all NBP messages are sent to that bridge.

The problem lies in the fact that the fastpath code (so far) does not
recognise RTMP messages from other bridges. Also, it does not maintain
any routining information for the appletalk nets.

To correct this, the fastpath code needs to maintain routine table for
the appletalk, send and recieve valid RTMP messages, and the rebroadcast
the NBP requests based on the routing tables. The Hayes InterBridges seem
to do this fine. However, if the fastpath was the last thing to send an RTMP,
the mac's send NBP requests to it, and it only services the local net.
Since both bridges (the fastpath and the InterBridge) transmit RTMP's
regularly, a "ping-pong" NBP response depending on the last bridge which
broadcast an RTMP occurs.

Anyway, I plan to do a simple RTMP implementation to the latest fastpath
code distributed with CAP2.01. It should be easy and will allow us to
get on with the real work.

If anyone is already doing this, please let me know. (I'll gladly defer to
another).

One the RTMP is going, ZIP should be fairly easy. Anyone interested in this
should first read the official "inside appletalk" and then read the ZIP
document by Alan Oppenheimer. Mine is dated Oct 1, 1985. It describes how
two bridges should interact with NBP.

Sorry for the length of this. And I didn't even flame about the 15 net
hopcount limit. ;-)
(Gosh, didn't Xerox's ethernet have the same limitation?)

Brad Parker
General Computer