CLIFF@UCBCMSA.BITNET (Cliff Frost {415} 642-5360) (06/09/89)
In response to John Wobus's complaints about Proteon routing I posted
a suggestion about what might be happening that was filled with careless
errors. One major mistake remains uncorrected, hence this note. (Charles
Hedrick already corrected a couple of them.)
My note implied that Proteon routers on a net will ALL end up sending
their routing tables at the same time. This is a ludicrous statement.
I strongly doubt that this will happen. The type of synching I have
observed with Proteon routers is much more subtle and goes like this:
8 routers +++p80 ring+++ router A ===ethernet=== router B ----net X---
There are 9 routers on the p80 ring. What I have seen is that router A
would receive overlapping updates from router B and another ring router.
When this happened, some packets from router B would be lost. If this
happened enough times in a row, router B would time out the route to net X.
This only happened under the extreme condition of passing approx 350
routes around on the ring. It was also sporadic and infrequent. Under
the circumstances I didn't (and still don't) blame Proteon.
If I keep working at it I may one day get my foot out of my mouth...
Cliff