W3C home > Mailing lists > Public > ietf-http-wg@w3.org > January to March 2008

Re: Content negotiation for request bodies

From: Adrien de Croy <adrien@qbik.com>
Date: Sun, 24 Feb 2008 19:02:19 +1300
Message-ID: <47C1086B.1080201@qbik.com>
To: David Morris <dwm@xpasc.com>
CC: ietf-http-wg@w3.org

David Morris wrote:
> On Fri, 15 Feb 2008, Adrien de Croy wrote:
>> Simply because the first agent to send a message is the client (rather
>> than the server), the server can know advertised client capabilities
>> (i.e. Accept- headers) before choosing to send a resource that
>> complies.  The client however cannot know what the server supports,
>> because the server has not had an opportunity to advertise any
>> capability.  There's no server welcome message.  So clients send message
>> bodies blindly, whereas servers send message bodies with knowledge.
> Well it would be rare for a client to send a significantly sized request
> body without a previous interaction with the server or one if its peers
> in the web based application. It is not uncommon for a web page to direct
> an upload or post transaction to another server, but the html page source
> should be aware of target's capabilities. A browser should be able to
> infer server capabilities from previous responses. If needed, we can
> perhaps add some kind of hint of server preferences along with
> implementation 'advice' which suggests that clients remember prior
> experience with a server for a modest interval.
> This should reduce error case to a few stupid servers which don't provide
> proper clues.  Tring what is likely to succeed based on history with the
> server or its application peers will actually have a very low error rate.
I can't even begin to imagine the administrative nightmare of trying to 
provide proper hints in a redirect about capabilities of some other server.

Why put all that burden (support in software for the feature, admin 
burden to configure it, and ongoing maintenance burden) onto millions of 
people when you can solve it in the protocol?

As a gross generalisation, heuristic solutions to problems when a 
deterministic solution exists are generally more trouble and effort than 
they are worth.


> Even non-browser clients can be expected to have or acquire knowledge
> as to a server's capabilities.
> It would also help if HTML Forms, etc. had the ability to express server
> preferences.
> Dave Morris

Adrien de Croy - WinGate Proxy Server - http://www.wingate.com
Received on Sunday, 24 February 2008 06:01:23 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 11:10:44 UTC