[HECnet] TCP based bridge / router, what's the status?

[Paul_Koning at Dell.com]

On Aug 29, 2012, at 5:24 AM, Peter Lothberg wrote:

...
DECnet over Ethernet, makes the assumption that unless there is a collition, the packet makes it to the
receiver.

Not quite.   The assumptions are that it's a datagram service with a fairly low loss rate.   It assumes order, and at-most-once delivery, though if that's violated I don't think things are likely to fail badly.   It is actually tolerant of some pretty weird errors, for example a non-transitive network (A can see B and B can see C but A can't see C).   That's rare in Ethernet bus networks and even less likely in modern switched networks, unless you misconfigure a firewall.   And many other network architectures get seriously confused if this happens, but DECnet explicitly covers this case.

DECnet over P-T-P links assumes there is underlaying reliable transport, eg, DDCMP.


Running a ether bridge serving multiple nodes over TCP will cause problems like what Jonny outlines, basically
it'sa head-of-line problem.

Are you talking about congestion or head of line blocking?   I'm not sure how you get from there to the problem that was seen.   Multiple correlated packet losses?   That might do it.


But if the Bridge makes a mesh of TCP connections (one per destination) it will work....

-P


	paul