- From: Julian Reschke <julian.reschke@gmx.de>
- Date: Thu, 20 Mar 2014 19:15:58 +0100
- To: Mike Bishop <Michael.Bishop@microsoft.com>, Rob Trace <Rob.Trace@microsoft.com>, Mark Nottingham <mnot@mnot.net>
- CC: "William Chan (ιζΊζ)" <willchan@chromium.org>, HTTP Working Group <ietf-http-wg@w3.org>
On 2014-03-20 19:03, Mike Bishop wrote: > While I wasn't in London, it sounds like the consensus there was "don't block the spec." However, there are a couple approaches to "don't block" that are being used by different people. > > Some seem to be advocating a normative reference, which makes HTTP/2 dependent on the Alt-Svc draft, and achieves "don't block" by getting the draft done so quickly that it doesn't (further) delay the HTTP/2 RFC. When we're already unable to finish HTTP/2 on schedule, this makes me worry that the draft would be rushed. Can we give it a try instead of having this meta discussion? :-) > Others are advocating an informative reference, where Alt-Svc would be dependent on the HTTP/2 spec and could finish afterward -- that corresponds better to my understanding of non-blocking. The problem with that approach is that with no extensibility mechanism in HTTP/2, if Alt-Svc needs *anything* from the base protocol it has to be known before HTTP/2 goes to RFC. That makes the obvious interpretation of not blocking rather problematic, now that Alt-Svc isn't simply using headers. An informative ref is not going to work if we put the alt-svc framework into a separate document. > What is the most contained set of changes we could make, either to HTTP/2 or Alt-Svc, to decouple them? There's no way to decouple them. The alt-svc framework needs to live in either (a) a separate spec or (b) in HTTP/2. In case of (a) (which is what IMHO the current authors of these specs prefer), HTTP/2 will normatively reference the alt-svc draft. In case of (b), all "users" of the alt-cvs framework will need a normative reference to the HTTP/2 spec, which looks weird to me.# At the end of the day, the number of documents really is not relevant for the time we need to finish them. Best regards, Julian
Received on Thursday, 20 March 2014 18:16:30 UTC