- From: Graham Klyne <GK@NINEBYNINE.ORG>
- Date: Mon, 29 Jan 2001 17:51:37 +0000
- To: Al Gilman <asgilman@iamdigex.net>
- Cc: "Donald E. Eastlake 3rd" <dee3@torque.pothole.com>, "Weibel,Stu" <weibel@oclc.org>, "'michaelm@netsol.com'" <michaelm@netsol.com>, Aaron Swartz <aswartz@swartzfam.com>, Larry Masinter <masinter@adobe.com>, uri@w3.org, ietf-type@iana.org, dee3@torque.pothole.com, Ted Hardie <hardie@equinix.com>
Hi Al, At 11:08 PM 1/28/01 -0500, Al Gilman wrote: >I never caught up with the requirement being addressed, here. What does CC/PP >actually do with types that it hears about? What does it need to know about a >type? What is the range of types it would come in touch with? What kind of 'type' are you talking about here? A MIME Content-type, or a more abstract math-ish kind of a type? For the former, the currently proposed vocabularies for CC/PP (as opposed to the core structure, which is agnostic on such matters) use a text literal to designate a content-type, per MIME spec. (This may not be how one would do it using RDF from the outset, but that proposal is an import from the CONNEG work.) Assuming the IANA:URN namespace proposal can be established, the plan is that the RDF property that designates the MIME content-type value (and other CONNEG-derived properties) would be named via URN from the CONNEG registry. >Does it ever compare a type and an instance for conformance? The current CC/PP proposals don't attempt that: it is left to the origin server to interpret the RDF property values that represent CC/PP attribute values. >Or does it compare type references for equivalence? I don't think we have any such thing. >What types would one hard-code into CC/PP? Or are the types not protocol >parameters, but rather protocol data in CC/PP? The only types hard-coded in the current proposal are the RDF classes that represent the basic structural elements of CC/PP. There are some recommendations for value types, but they're not hard coded. >It seems to me it would make more sense to create a schema which documents the >structure of the Internet Media Type space. Then reference Internet Media >Types via their place in this schematized index to the existing dictionary of >types. Does CC/PP need to know that text/* is a supertype of text/html and of >text/plain? Does it need to know that application/octet-stream is a supertype >of application/anything-else? If it does, just injecting each type-indication >(header field value) into URI space as an unrelated atom loses a lot of the >value of the Internet Media Type system. In its current proposed form, it knows nothing of these things. But, on a related matter, when we looked at creating the CONNEG feature tag for media types, we took a view that it should focus primarily on the type/subtype elements, and play down the parameters. On review, Ned Freed took a harder line, with which I agreed, which was to completely prohibit Content-type parameter values: many of them just aren't relevant outside of a particular MIME usage (e.g. boundary on multipart) and others are better treated as separate features (e.g. charset on text/*). This was a position taken for media feature descriptions -- other scenarios may have different requirements. #g
Received on Monday, 29 January 2001 13:53:18 UTC