- From: Roy T. Fielding <fielding@ebuilt.com>
- Date: Thu, 17 Jan 2002 17:33:49 -0800
- To: Mark Baker <distobj@acm.org>
- Cc: "Simon St.Laurent" <simonstl@simonstl.com>, www-tag@w3.org, ietf-xml-mime@imc.org
On Thu, Jan 17, 2002 at 08:00:04PM -0500, Mark Baker wrote: > > Therefore, it is not useful to send something as application/xml unless > > the sender specifically wishes it to be processed as generic XML. > > Fair enough. So what if we define "generic XML" to include dispatch > to processors keyed on namespace declarations? Wouldn't that make > it more semantically significant? Namespace declarations define the meaning of tags within a document, not the meaning of the document itself, and certainly not the intentions of the person using the document. Consider what would happen if a person is trying to use a generic XML editor to hand-edit some elements within an SVG file -- the last thing they want is for the editor to try to process the contents automatically as SVG just because it contains a namespace declaration. Let's say you have two resources that both resolve to the same content. GET one and the response is always sent as image/cvg+xml. GET two and the response is always sent as application/xml. Those are separate semantics, and thus separate resources. It doesn't prevent the UA from taking the content and choosing to manipulate it in different ways, but it does define the semantics from the perspective of the person who created the two separate URI used to access those two resources containing the same content. The media type MUST be able to distinguish those two semantics, since nothing else in the protocol will differ between the two messages. That is why HTTP absolutely forbids the UA from guessing the media type based on the content if a Content-Type has been declared in the message. ....Roy
Received on Thursday, 17 January 2002 20:38:29 UTC