Re: [mif] Domain name issue in draft-stenberg-mif-mpvd-dns

[Markus Stenberg <[e-mail]>]

> On 15.3.2016, at 12.26, Ray Bellis <[e-mail]> wrote:
> Looking at section 3 of this draft, I have a problem with the
> specification where it says that for e.g. a /64 the DNS lookup should be
> for _pvd.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.x....ip6.arpa, and likewise the
> /80 example (with 12 leading zeros).
> 
> Taking the /80 example, it would be impossible given this specification
> to distinguish between a /80 where the bits between 64 and 80 were all
> zeros and the corresponding /64 - they would both end up at the same
> point in the ip6.arpa hierarchy.
> 
> In my opinion (and I do have some prior art in this space, see RFC 7216)
> the "_pvd" label must be at the point in the ip6.arpa that presents the
> same level as the prefix being queried for.  Any leading zeros below
> that point (i.e. to the left in the domain name) MUST be omitted.
> 
> Hence in the /80 case, it would be _pvd.<20 labels>.ip6.arpa and in the
> /64 case it would be _pvd.<16 labels>.ip6.arpa

Actually we considered it a feature and not a bug, i.e. subdelegation would still result in same PVD available. I am not sure your scheme would work with subdelegation though.

Given broad example, ISP has /40 for home users, but ultimately the prefix in use may be e.g. /64 (or even something more narrow). Having the _pvd.* in 'all' prefix lengths would be obviously available, or performing (up to 128) lookups with subsequently shorter/longer prefix lengths.

I am not sure how client would know to start lookups from /40, or from /56 assigned to a home, if the link has /64.

Cheers,

-Markus