[c-nsp] [j-nsp] draft-ietf-mpls-ldp-ipv6-16

[Mark Tinka]

On 21/Feb/15 21:56, Saku Ytti wrote:
> I wish we'd be in market situation where we can choose customers.

It is not an ideal situation, Saku, but this type of request is one
which will make new customers happy and existing customers (very) unhappy.

In our case, considering that we sell more IP than we do l3vpn (although
it's hard to say what the future has in store), it doesn't make sense to
anger a long-standing loyal base.
> In ingress it would be fairly easy to use any ACL/FW key to determine if IP
> TTL is copied to MPLS TTL or if MPLS TTL is set to 255. Maintaining that
> egress PE has same idea as ingress PE might be tricky.
>
> Then of course if there is MPLS loop in core, and your TTL 255 is not
> sufficient in transit, what to do? Then you get in really fuzzy/heuristic
> decision where you won't always be able to do what ideally should be done.
>
> Now that we have two decades of experience, we better know the demands and we
> could do some improvements on MPLS frame format, MPLSv2 if you wish. We could
> address more efficiently and cleanly some of the problems that now require
> imposing two+ labels (1 special stuff coming, 1 special stuff itself).

I'd be open to considering implementing this capability on a per-service
or per-customer basis. Perhaps it's time we re-opened this with the
vendors, as it does have a use-case.

Mark.