Re: router-router NHRP

[Robert Coltun <[e-mail]>]

Hi,
	The refreshing of the routed path by the source (or some
intermediate NHS) seems to be the scaling problem here. Perhaps
one way to deal with this is to have the border NHS routers keep state
of the source network or host (and perhaps the last-hop NHS)
and rather than wait for a request to revalidate the resolution,
have the NHS verify that the reverse path is correct for all
entries that relate to a changed route - if it isn't a purge is sent.
There would additionally need to be keep-alives
along the routed (NHS) path so that if a border node fails (or someone
in the NBMA routed path fails) the resolution is also purged.

I belive this is a variation on something proposed in the framework
document (to send a poisoned reverse) as you actually want to 
poison the reverse path.

Comments?

---rob