Re: [homenet] Alia Atlas' No Objection on draft-ietf-homenet-hncp-09: (with COMMENT)

[Juliusz Chroboczek <[e-mail]>]

>> While src-dest routing is certainly a solution - and an interesting
>> one - it doesn't seem at all appropriate for an HNCP spec to assert
>> that it is necessary.

> True. However, we were asked to describe the applicability, and
> I consider e.g. tunneling solution inferior so I would rather not
> propose that here.

I was under the impression that there was consensus that tunnelling is
a bad idea, and that source-specific routing is superior.  I may be
mis-reading Alia, but perhaps she means that some deployments might not
require the added complexity of source-specific routing, and might be able
to get away with ordinary next-hop routing.  (Alia, apologies if
I misunderstood.)

Recall that our main concern is to ensure that Homenet routers from
different vendors interoperate.  Source-specific routing, when done right,
interoperates with ordinary next-hop routing under the following
conditions: (1) all edge routers are source-specific, (2) there is
a connected backbone of source-specific routers, and (3) at least one of
the routers in the backbone announces a non-specific default route.  (1)
and (3) are easy enough, but (2) is problematic: it is a global condition,
one that can only be verified with global knowledge of the topology.

So I agree with Markus -- while it might be possible to get away with
a weaker requirement, it is way simpler to make source-specific routing
a MUST.  If there is market demand, we might want to write a different
document that describes a weaker set of requirements for interior-only
(non-gateway-capable) Homenet nodes.

(A more desirable alternative, of course, would be to split the
requirements in HNCP into HNCP requirements and Homenet requirements, but
I gather that there's consensus that it's somewhat late to embark on such
an ambitious project.  I rather agree with that assessment -- we need to
get Homenet out of the door quickly, lest people start deploying IPv6 NAT.)

-- Juliusz