Re: Proposal to use MIME types to modify interpretation of content

>A W3C group is working on a MIME type proposal for RDF, which is a general
>metadata/knowledge representation format based on XML.

>The suggestion quoted below came up, and I'm slightly uneasy about it.  So
>I thought I'd see if it fits with your view of MIME type usage.

So am I. The intent of a content type is to describe what sort of content
is present. Anything more is really beyond the purview of a content type.

>The proposal is that one content type means that statements in a document
>are asserted to be true, and any other type means that are simply quoted
>(expressed without being asserted).

This usage doesn't exactly parallel any field we've previously defined, but it
isn't all that far away from a content-disposition. Specifically, a
content-disposition could be defined to mean "this is asserted to be true and
you should handle it as such".
Also note that the concept of storing storing other metadata like file names as
part of a content-type (specifically, a content-type parameter) has been
roundly rejected over time. I don't think this sort of metadata is
significantly different. The long standing whine about not being able
to dispatch off of content-type parameters or content-dispositions no
longer holds water either.

>The nearest analogy I can think of is this:  suppose I have two MIME types,
>say, "executable/beans" and "executable/coffee", which both contain
>programs in some programming language.  In the case of "executable/beans",
>the intent may be to take the data and store it in some repository of
>executable code, for use later as and when required.  In the case of
>"executable/coffee", the intent may be that the program is handed to an
>interpreter and the output stored or displayed to a user.

Now this description really does smell like a content-disposition. And the
SIP folks have no problem defining new content-dispositions for this
sort of stuff; the W3C shouldn't either.

>Does this seem like a reasonable use of MIME types to you?

No, it doesn't.

                                 Ned

------------
Graham Klyne
(GK@ACM.ORG)

Received on Monday, 14 May 2001 14:35:30 UTC