- From: Willy Tarreau <w@1wt.eu>
- Date: Tue, 1 Jul 2014 22:55:26 +0200
- To: Johnny Graettinger <jgraettinger@chromium.org>
- Cc: Jeff Pinner <jpinner@twitter.com>, "William Chan (?????????)" <willchan@chromium.org>, Jason Greene <jason.greene@redhat.com>, HTTP Working Group <ietf-http-wg@w3.org>
On Tue, Jul 01, 2014 at 04:46:15PM -0400, Johnny Graettinger wrote: > > > > As I said above, all that dangerously complicated stuff for a dead > > > connection anyway is not worth it. Better kill the connection and restart > > with something sane if needed. > > > If HAProxy is speaking HTTP/2 to backend application servers, and I can get > an application server to send a response with more than 16K of headers, > doesn't that give me an easy way to tear-down any other active stream to > that server? In 1.1 it would return 502 bad gateway. In 2.0, assuming I manage to implement it to the server, I don't expect it to be able to multiplex multiple streams over a single connection before we've finished HTTP/3. I'm serious. 2.0 is a major architecture change for low-level intermediaries. So I'd say that for now, with the elements I have, a protocol error on 2.0 will be handled the same way as in 1.1 : the connection will be torn down and the effects will be exactly the same : only the request using that connection will be impacted. I'm not saying nothing better could be done, of course. But at least for me right now it's not obvious :-/ Willy
Received on Tuesday, 1 July 2014 20:55:53 UTC