W3C home > Mailing lists > Public > ietf-http-wg@w3.org > April to June 2014

Re: Alt-Svc header + ALTSVC frame

From: Roberto Peon <grmocg@gmail.com>
Date: Mon, 28 Apr 2014 21:42:27 -0700
Message-ID: <CAP+FsNev-3FR+uYR1RNYJgsePyb-JDunenMYSnGQ9oN2jB3+QA@mail.gmail.com>
To: Mark Nottingham <mnot@mnot.net>
Cc: HTTP Working Group <ietf-http-wg@w3.org>
I'd hope that the immediate GOAWAY wasn't necessary (that would cause a
service disruption for that particular user for a while).
if ALTSVC is advisory only (honestly, I'd prefer it be mandatory, though it
isn't trivial to test for, it can be tested for, like DNS disobeyers can
be), then we probably need a DRAINING frame, which would cause the client
to open up a new connection and make new requests on that new connection
once opened.

On Mon, Apr 28, 2014 at 9:13 PM, Mark Nottingham <mnot@mnot.net> wrote:

> The Alt-Svc header field doesn’t allow advertisement of an alternative on
> another host, to avoid security issues (mostly around someone injecting a
> header that transparently redirects a site to an attacker).
> The ALTSVC frame does allow changing hosts, as long as certain criteria
> have been met (as per the draft).
> If a site wants to upgrade from HTTP/1 to HTTP/2 *and* shift traffic to
> another server, it seems to be it could do so by:
> 1) Serving the HTTP/1 response with an Alt-Svc HTTP header to another port
> on the same host, speaking HTTP/2
> 2) On the HTTP/2 port, upon connection immediately send an ALTSVC frame
> followed by a GOAWAY frame
> That would have the effect of informing the client about the alternative
> on the other host, which would be used once the connection to it was viable.
> Any reason this wouldn’t work / be a good idea? Mostly, I’m interested in
> confirming that everyone has a similar understanding of how Alt-Svc/ALTSVC
> works…
> Cheers,
> --
> Mark Nottingham   http://www.mnot.net/
Received on Tuesday, 29 April 2014 04:42:55 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:14:30 UTC