Re: [DNSOP] Pity (was Re: Expiration impending: <draft-jabley-dnssec-trust-anchor-11.txt>)

[Suzanne Woolf <[e-mail]>]

(no hats)

On Oct 9, 2015, at 3:25 PM, Paul Vixie <[e-mail]> wrote:

> no longer. he needed client-subnet for his
> business, that year, not this year. i didn't and still don't like that
> protocol and i fear its long term implications, but i understand why he
> did it the way he did it.

I think client-subnet was actually an interesting example of several things.

When it was originally brought to the IETF, people didn't like it, so it didn't get through the WG (DNSEXT?). The Options registry policy at the time was, I think, Standards Action, so they couldn't get an IANA-registered codepoint, either.

As Paul points out, they wrote their technology and started using it anyway.

The registry policy was changed, partly because of cases like this, to Expert Review, and a codepoint was requested, without an RFC. It was granted, so the technology could be extended by the use of a registered codepoint.

The original authors of the draft came back to the DNSOP chairs and expressed an interest in documenting the protocol anyway. We encouraged them to do that-- we figured that since there's a properly registered codepoint, and the technology is in use even though DNSEXT five years ago didn't like it, documented is better than not.

The current client-subnet document describes the technology currently in use, and suggests that the protocol could be revised, in light of experience, to work better, including making it easier for clients to opt out.

I would hope that when that protocol revision is undertaken that the authors will be willing to document it in the IETF, and that they get the review they need to help them make a better protocol.

If we succeed, there will be a better protocol, and review in the IETF will be part of why it's better, and it will be documented in an RFC.

If we fail, there may or may not be a better protocol, but the IETF will have nothing to do with it, and it may never be documented in an RFC. 

I'd like to see the first option happen. I think it's entirely possible, and will do what I can to enable it. 


Suzanne