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

Re: Alternative Service Indication

From: (wrong string) 陈智昌 <willchan@chromium.org>
Date: Thu, 1 May 2014 18:40:31 -0700
Message-ID: <CAA4WUYiMMjidi6GmKE_0a_uxU4JfoHEiM10aEOYzxeCBe+9Zhw@mail.gmail.com>
To: Erik Nygren <erik@nygren.org>
Cc: Martin Thomson <martin.thomson@gmail.com>, HTTP Working Group <ietf-http-wg@w3.org>
I somehow missed the previous discussion thread on it. Sorry about that. I
have some comments now though :)

* I'd like to hear more people discussing it. The only participants on the
last thread that I saw were Martin, Mark, and Erik. Can others chime in?
* It's interesting to me that this appears to be mandatory for clients. I
need to think about this more deeply.
* I need to think more deeply about how necessary this is. My default as a
client is not to provide something unless there's a clear reason to do so.
Are there any examples of things *breaking* without this? AFAICT from
previous discussions, there are some server suboptimalities, and this would
be nice to provide. I'm trying to evaluate how "nice" this is vs mandatory.
I consider the Host header to be pretty important to the web as it operates
today, but am unclear on how this compares.
* Does sending an ALTSVC frame play nice with connection sharing? I haven't
thought about it too much.

Cheers.


On Thu, May 1, 2014 at 6:27 PM, Erik Nygren <erik@nygren.org> wrote:

> This looks good to me.
>
>    Erik
>
>
>
> On Thu, May 1, 2014 at 5:21 PM, Martin Thomson <martin.thomson@gmail.com>wrote:
>
>> The discussion we've been having on this topic seems to have
>> converged, or at least cooled a little.
>>
>> I've put together a pull request that combines the following changes:
>>
>> 1. Clients can send an ALTSVC frame to indicate the service that is in
>> use.  This is HTTP/2 only, obviously.  The main advantage of this is
>> that you can send a single frame.
>>
>> 2. Clients can send an Alt-Svc header field to indicate the service
>> that is in use everywhere else.  This has to be on every request
>> (i.e., no change).
>>
>> 3. Remove the Service header field in favour of the above.
>>
>> 4. Add some security considerations around the use of these indicators
>> for tracking clients.  I was a little concerned about this at first,
>> but now that I've done the thinking, the concern seems manageable.
>>
>> See the text here:
>> https://github.com/http2/http2-spec/pull/474
>>
>> --Martin
>>
>>
>
Received on Friday, 2 May 2014 01:40:59 UTC

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