Re: [Idr] I-D Action: draft-ietf-idr-error-handling-13.txt

[Jeffrey Haas <[e-mail]>]

On Sat, Jun 14, 2014 at 03:45:22PM +0100, Chris Hall wrote:
> I have a feeling that defining "valid host address" is more trouble
> than it is worth.

I'm also a bit concerned here.

RFC 4271 and earlier went through a lot of wrangling to end up with the idea
syntactically correct vs. semantically correct.

If it's syntactically incorrect, treat as withdraw is quite reasonable.

Semantically incorrect gets heavily tied up into how we're breaking original
rules today in a given spec.  E.g. RFC 5575 zero length.  What happens when
we decide to get clever and end up using a class D/E address in IPv4?  Etc.

For any future Clever Feature, deployment of it will need to be pervasive
enough for the networks using it to make use of it.  This means an island of
semantic correctness for something that didn't used to be that way.

If you're outside of the island, it's semantically incorrect, treat as
withdraw.

It may thus be worthwhile simplifying the section:
"If the next hop field contains a semantically incorrect address within the
context of deployed features and address family, treat as withdraw behavior
should be used."

I'd keep the section on martians.

-- Jeff