[
Lists Home |
Date Index |
Thread Index
]
> From: Gavin Thomas Nicol [mailto:gtn@rbii.com]
<snip/>
> I think the hey,value pairs have shown themsevles to be insufficient.
> Look at MIME media types for a prime example, or the Accept-*
> headers.
> Many headers have a built-in ad-hoc syntax for representing structure.
Quite apart from that, the notion that every message type has to have a
unique MIME media type runs against the grain of a common and perfectly
valid approach to XML processing that insists "application/xml" (or perhaps
something a bit more specific, such as "application/soap+xml") tells the
protocol all it needs to know about the document "type", and that the notion
of a single "type" for the document, rather than a modularized mix of
fragments from different namespaces, is not useful.
If REST insists that the only type of compound document that can be sent
over HTTP is one that makes all of its consistuent parts visible at the
protocol layer (such as MIME multipart formats), then that certainly
constrains uses of XML over HTTP. It seems to me to even be in conflict with
the direction of many XML standards.
Maybe Simon St. Laurent's xmlns media-features proposal can help, here,
although I'll confess I'm still a bit confused as to what media-features are
(and I am even more confused as to why they use an ad-hoc syntax for
representing structure that I thought was reserved by MIME for comments).
|
