TCP-group 1992
[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
A question about RSPF
- To: bill@platypus.uofs.edu
- Subject: A question about RSPF
- From: "k1io, FN42jk 04-Feb-1992 1447" <goldstein@carafe.enet.dec.com>
- Date: Tue, 4 Feb 92 14:55:36 EST
RSPF is certainly intended to cope with non-reciprocal routes!
I'm assuming that in your example (A, B, and C, where C cannot hear B
but all of the other five possible one-way routes work), all three nodes
are running RPSF. What should happen is that C broadcasts that it hears
A, B broadcasts that it hears C and A, and A broadcasts that it hears C
and B. Then the paths are built.
The current AX.25 implementation is crippled by NOS and ARP. Since B
hears C, it tests out the adjacency. It does this by ping, which
begets ARP. Ping can go to C via A, and C can theoretically reply
directly to B, but C can't really send the response to B because C can't
ARP B. Ping to C's AX.25 address fails because C can't hear it, so that
adjacency _should_ not show up in "rspf links".
Now for B to ping C, it has to put C into its Route table. Once it's
there, even if RSPF marks it as "bad", it remains there... that's a
problem in the current 2.1 implemention. It's not purging bad routes
from IP Route.
I'm not sure exactly what's happening in your case. Anyway, we hope
Mike N1BEE will be able to fix this during his RSPF 2.2 implementation
work.