- From: Clinton Wong <clintdw@netcom.com>
- Date: Mon, 5 May 1997 16:42:34 -0700 (PDT)
- To: www-talk@w3.org
In future versions of HTTP, would it be a nice feature to have a version number as a parameter to the content-type and accept header in HTTP? Let me explain... To specify that a document is HTML 3.2 compliant, one might want to upgrade HTTP to include things like: Content-type: text/html (v3.2) or to specify components of HTML, how about extra parameters to specify what capabilities are expected of the client? For example, if frames are used, the header might look like: Content-type: text/html (v3.2/frames) Likewise, the browsers might want to send something like: Accept: text/html (v3.2) And the server could provide HTML 3.2 instead of some future version of HTML. (Perhaps run a filter program to downgrade newer html tags/parameters to be 3.2 compliant) Here's the motivation for printing out content-type versions and/or components: - Browsers know what the content is without downloading it. For example, if a browser can handle GIF87 but not GIF89, it could use HEAD to determine the GIF version number and skip GIF89 images. Likewise for HTML and other media types. - Web site designers no longer need to do a "timed redirect" with the default page saying "if your browser doesn't switch pages in a few seconds, click here." I suspect that they do this with the assumption that "newer" browsers will understand the redirection and older browsers won't. From an architecture standpoint, this won't work well in the future. There will be a lot more browsers in the future that support redirections but vary by HTML/applet/plugin capability. - One can keep statistics about the capabilities of web browsers and servers. It might be possible to say things like "84% of our hits are from browsers that understand Cascading Style Sheets." Support of media-type version information could lead to: - HTTP media-type content negotiation, so a browser can get content that makes use of its features. (And not receive content that it can't support) - With persistent connections, the content and communication between the client and server could become specialized as more transactions are performed. Historically, Mosaic would print out all the MIME types it supports. Today, browsers only print a few MIME types in its Accept header, like GIF, JPEG, xbitmap, or */*. In future versions of HTTP, a typical connection might have the following logic: - Client requests a URI from a web server. - There are two copies of the URI at the server. One is HTML 3.2 compliant and the other is HTML 4.0 compliant. The server tells this to the client and asks the client which version it would like. - The client replies to the server with a preference of HTML 3.2. The server remembers this preference for the rest of the (persistent) connection until told otherwise. - The server responds with a HTML 3.2 version of the URI. These are just a few ideas... maybe they're good ideas, maybe not. If you have comments, feel free to email me. Regards, Clinton
Received on Monday, 5 May 1997 19:42:42 UTC