[rrg] [lisp] Economic issues of long-path / stretched routing

[Dino Farinacci]

> You peer with your upstream ALT routers based on prefixes they are
> responsible for, and where the ALT routers are doesn't (necessarily)
> follow physical topology.  For any prefix there will probably be more
> than one upstream ALT router.  Yes, you announce your prefix to all
> upstream ALT routers responsible for the shorter including prefix.
> Table size should be minimal, depending on how things are configured
> in your neighborhood.  It could just be default from each.  At most it
> would be a set of highest-level prefixes, maybe some special ones, and
> some longer ones for local efficiency.

And your neighborhood really doesn't need back-doors so we can keep  
the ALT hierarchical by shaping it as an acyclic tree and not a mesh  
of trees or a mess of links connecting up and down at *different*  
hierarchical levels like the underlying Internet is configured today.

Dino