[e2e] Overly Overlay; Peer to peer is commonplace

[Geoff Huston]

Ran Atkinson wrote:
>...which comment thoroughly confirms Jon's point, since the Zebra
>code clearly is not ready (today) for large-scale production networks.
>It might be ready for such in some future world, but it clearly isn't
>today.
>
>         It *would* be good if some/all of the implementation optimisations
>to build truly scalable IP routing protocols were documented.  However,
>most aren't documented, which provides job security for those in the know.

You are possibly working on the assumption that _someone_ knows, but they 
are not telling the rest of us. The alternative explanation is that noone 
knows and there is nothing to tell as yet!

I suspect that the latter is the case - scaleability in routing is (imho) 
all about trade-offs in understanding how pieces of reachability 
information can be discarded without undue loss of functionality of the 
larger system. I'm not aware that this form of tradeoff has passed from a 
imprecisely understood art to that of a definitive set of engineering 
principles so far.

   Geoff