- From: Al Gilman <asgilman@access.digex.net>
- Date: Thu, 17 Jul 1997 13:31:58 -0400 (EDT)
- To: w3c-wai-wg@w3.org (WAI Working Group)
to follow up on what Daniel Dardailler said: > [I had said]> > > The requirements for HTTP should come from the Styles work area. > > HTTP and Styles should mutually work out how much information > > about the media capabilities of the client session can/should go > > from client to server in support of autogeneration of pages or > > selection of styles. They will have to figure out whether it > > will work to put stylesheet subtypes in the existing Accept > > header or whether some new header field possibly named Target_Media > > would work better. > > Now you've confused me. Why would HTTP be extended here since the > model is as follow: > > - the server sends *one* HTML document with multiple pointer > to various SS in it, categorized by media target (see example in > http://www.w3.org/TR/WD-html40-970708/present/styles.html#adef-media) > - the user-agent decides which one it wants, and fetch it. > > In which circumstances would the user-agent need to negociate further > with the server (e.g. send it a preference list). > Probably just my confusion. I was thinking there was a scenario which would work like the server-side decision to send .gif or .jpg. I wonder if the single-HTML-document scenario isn't a little over-confident. There is a possible scenario where adaptations are done in a page generator engine at the server based on capabilities sent from the client. This approach has more ultimate capability, that is it can be taken further if you push it, than what one can do with styles. I might not want to rule it out. My preferred strategy is to work the flexible presentation problem in a way that treats the WWW as a black box, i.e. don't assume what is client vs. server functionality. Then make sure we have enough communication capability (and cooperation rules) so that a range of solutions are supported. On the other hand, I think that the characterization of media targets is more important, and is _our_ problem, not _their_ problem. Somehow, we need to come up with a way to wean server operators off the use of the User Agent string to discriminate service decisions. And the right short list of media targets could really help. -- Al Gilman
Received on Thursday, 17 July 1997 13:32:08 UTC