classless nets and home ASs

Mon Apr 25 17:37:39 UTC 1994

Elise,
	My feeling is that there is some level of atomic aggregation,
below which, you will be using igp routes. This goes back to the
original definition of Autonomous System. I would like to see the home
AS represent that transition from igp routing to egp routing. That is,
if you inject a route into the global routing table, then you have
some level of responsibility for that route and this should be documented
somewhere.

	Don't think of it as no more A,B,C networks, but rather as
30 different sizes.

Erik

P.S. Would routing within a AS be sub-atomic routing? :-)

> During this tranisition from the classful world to the classless,
> home ASs still have meaning.  It does not seem clear that home ASs
> will be relevant when we treat all routes as classless entities,
> because most assignments can be further subdivided. There will be
> no base unit such as class A, B or C on which to hang the home AS
> tag. We may want to move from "one net/one home AS" to "prefix and length/
> announcing AS" - since an AS downstream or upstream may aggregate
> a subset or superset.
> 
> I'm not commenting directly on your suggestion on how to express
> meaningful information about the specific nets that AS 701 has
> withdrawn from its announcment to AS 690, but trying to explore
> how we should look at keeping information about classless entities.
> These lists (regional-techs and bgpd) seem like the right place to
> discuss what information the operators need in a classless routing 
> world. 
>               --Elise
> 