Re: [v6ops] Scope of Unique Local IPv6 Unicast Addresses (Fwd: New Version Notification for draft-gont-6man-ipv6-ula-scope-00.txt)

[Ted Lemon <[e-mail]>]

On Jan 7, 2021, at 9:55 AM, Philip Homburg <[e-mail]> wrote:
> I can see a few benefits of Mark's proposal. One is that it is good to
> have a standard representation of information. In particular,
> Mark's proposal would make it possible to have a master zone file that has
> both public and private DNS entries. Then a split-DNS server could serve
> only the public data to the outside world. 

That’s a good point, although it would still be a good point if this were just a feature of the zone file and not of the wire format.

> At the same time, I think it would be great if we can put link-local addresses
> in DNS. 

That sounds like a really heavy lift.

> It may tie in nicely with scope IDs in socket addresses. If a DNS
> record specifies that is valid only on a VPN link, then maybe we can already
> tie the address to that link. No need to change applications, it can be
> hidden in the stub resolver.

Now we need to standardize a way to identify links. This is a Hard Problem. I say this based on experience, not supposition. HNCP tried to do this, not as successfully as I’d hoped. I’ve been working on it for the Thread Border Router work, and haven’t come up with a general solution. Sure, if you have a data center and a managed multi-subnet LAN, and you can just type in configurations, this works, but most networks aren’t like that.  I think the VPN case is probably tractable, but it’s really hard to see a path to broad adoption for this idea.

If there is a path to broad adoption, it probably involves bottom-up work, not top-down design. Most of the ideas I’ve had about this that I think are practicable are very context-dependent. E.g., you can identify that you are on the same link because you received a link-scoped multicast.