- From: Poul-Henning Kamp <phk@phk.freebsd.dk>
- Date: Wed, 02 Jul 2014 10:37:29 +0000
- To: Mike Belshe <mike@belshe.com>
- cc: James M Snell <jasnell@gmail.com>, ChanWilliam(ιζΊζ) <willchan@chromium.org>, HTTP Working Group <ietf-http-wg@w3.org>
In message <CABaLYCsvLCZPcDObg2A2yP8yG+aunsT1ZPyy_Ha9hGBCY15Heg@mail.gmail.com>, Mike Belshe writes: >> We have two (or more ?) major-ish browsers nixing HTTP/1 upgrade. > >protocol upgrade has always been optional at the client discretion. this >is not new with http/2. Absolutely, just like my decision to implement HTTP/2 or not implement HTTP/2 is entirely discretionary. But such discretionary decisions people will make really big footprints on the future success of HTTP/2 in the wider web, so these "political statements" should weigh heavy in the judgement of the drafts future prospects. Of course if the only thing cares about are a few $BigWWW sites, then that doesn't matter, but if the question is if HTTP/2.0 is worthy of its name, it matters a LOT. >> We have a number of proxies (involved in about 30-50% of all HTTP1 >> content delivery) nixing CONTINUATION and we have a lot of webmasters >> who have yet to see any evidence that adding HTTP/2 support would >> ever be worth their while. >> > continuation is not a "throw the whole thing out" issue. Actually it is. It's a clear indication of technical ineptitude when it comes to good and clean protocol design. I'm 100% certain that you'd get flunked by Andy Tanembaum for a END_STREAM flag which doesn't, and if that doesn't do it, a framesize with a maximum payload of (2^n-1) certainly will. >So now you're saying we should nix http/2, because you started talking >about http/3. I'm saying that if we rush into Last Call with the draft we have now, we have no choice but to start talking about HTTP/3 right away, because it is simply not good enough. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@FreeBSD.ORG | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence.
Received on Wednesday, 2 July 2014 10:37:52 UTC