Re: nhrp-05 - Purge packets

[Andrew Smith <[e-mail]>]

> From [e-mail] Thu Nov  2 18:14:12 1995
> To: [e-mail] (Andrew Smith)
> Cc: [e-mail], [e-mail]
> Subject: Re: nhrp-05 - Purge packets 
> In-Reply-To: Your message of "Thu, 02 Nov 1995 12:02:30 PST."
>              <[e-mail]> 
> Date: Thu, 02 Nov 1995 17:56:28 -0800
> From: Bruce Cole <[e-mail]>

Bruce,

> > Doesn't the receiver also have to match on some other things e.g.
> > 
> > - that the Purge comes from the same guy that supplied the cached 
> >   information: match on Source IP Address or Source NBMA Address
> 
> Possibly; may be a problem for multi-homed stations.

Are you saying that the protocol does not guarantee
that the purge will come from the same place as the original info? Or
at least that it does not know whether the purge came from the same place
because it cannot trust the Source IP Address?

I think that it would be nice if at least the Source IP Address were the same
even if the Source NBMA address had no guarantees. There are some very
ugly failover/transient cases that we discussed in MPOA where two seperate
NHSs give conflicting Purges to a client (Yakov - do you remember exactly
how we reached the conclusion that a client needed to get purges from the
same place that it got its original cached info?). In fact the whole
issue of multiple NHSs and failover cases is somewhat under-specified 
currently in NHRP.

> > - that the QoS matches what was cached.
> 
> I don't think so; of course, use of QOS by NHRP is still under specified.

I think you are assuming that QoS-based routing will not be used with NHRP
around. It would be a shame to make that assumption.


Andrew


********************************************************************************
Andrew Smith					TEL:	[phone]
Technology Synergy 				FAX:	[phone]
Bay Networks, Inc.				E-m:	[e-mail]
Santa Clara, CA
********************************************************************************