- From: Martin Thomson <martin.thomson@gmail.com>
- Date: Mon, 1 Jul 2013 10:44:43 -0700
- To: James M Snell <jasnell@gmail.com>
- Cc: Amos Jeffries <squid3@treenet.co.nz>, HTTP Working Group <ietf-http-wg@w3.org>
On 29 June 2013 07:27, James M Snell <jasnell@gmail.com> wrote: > On Jun 29, 2013 1:03 AM, "Amos Jeffries" <squid3@treenet.co.nz> wrote: >> The cache will store the pushed response under: >> hash(URL)+hash(Vary:"image/jpeg") >> >> When looking up future HIT the cache for this client on explicit-GET it >> will look up >> hash(URL)+hash(Vary:"text/html, image/jpeg, image/gif") >> >> I assume you can spot the difference. > > I didn't miss that. Pushed resources and content negotiation used together > have rather, um, "interesting" complications. However, those don't change > the request headers in the push promise and response headers in the HEADERS > question which is what I was addressing. I think that we've reached a conclusion. One very different to what is described in SPDY/3 (the version we inherited). A client without a cache entry will not care about what the cache key was. A client that has already received the document is likely to get a cache miss due to the above effect, but at least the behavior is "correct". Subsequent pushes, assuming that the server is consistent with itself, should produce a cache hit. I can see a future for UA-sniffing here. that makes me a little sad, but optimizations tend to have that effect. BTW, the idea of RST_STREAM+If-None-Match is appealing. I've seen lots of approaches to this problem and that is somewhat novel. I haven't thought it through fully, but it could add something, depending in a number of conditions. Like any optimization, it does need to be tested out, and it's probably early days for that sort of tweak.
Received on Monday, 1 July 2013 17:45:10 UTC