RSPF meets the Route table quandary

["Milo S. Medin" (NASA ARC NSI Project Office) <medin@nsipo.nasa.gov>]

Ok, I understand your point.  In OSPF and IS-IS, you can pass around an
ES route too, but using a route with a all 1's subnet mask, bascially making
it a host route, and that allows you to basically float around like you
describe.  I guess this leads me to ask a different question, which is
how many systems really need to float around like this?  If only a few
systems needed this capability, you could still use fixed subnets that
cover most of the ES's, without having to flood their info around.

I guess it would seem to me that if things were this flaky, you would have
to flood ES info around a large area, basically around where that ES
could attach.  If that's the case, you end up flooding ES info over a
large potential area.  At some point you have to draw the line, else you
can't make hierarchical routing work, and you need that given the number
of nodes that float around.  How do you set up where this ES can
attach itself to?  It seems you really need a ES-IS function to make this
work.  ARP doesn't seem like a good fit.  Have you looked at the new Gateway
discovery protocol?  That would seem like a better fit.

I guess I am constantly surprised by what people want to run IP over.  I was
"raised" on ARPANET like networks, and because of that I've never really gotten
interested in the Bell 202 world of packet radio.  Of course, I also work
on designing DS-3 networks these days, so maybe I'm just an oddball.

                                                Thanks,
                                                   Milo