Q1. Do you propose to revise RFC 2048 and discourage media types
that are built on top of XML?
Certainly not. People can build media types "that are built
on top of XML", in the same way that XML is a media type that
is built on top of "text/plain". XML inherits many of the
characteristics of text/plain, but it isn't text/plain, because
that's not how it is intended to be processed.
text/plain can be processed with a generic text processor,
but you can't tell the difference between my autoexec.bat
and a web page.
Q2. Do you propose to revise RFC 2376 and discourage such
media types?
I think that the revision of RFC 2376 (necessary anyway)
will need to give some guidelines.
For the most part, the external representations of components
of XML don't need any MIME designation at all, because in
order to process them, they have to be supplied in some context
or wrapped in some kind of packaging or container that doesn't
need to use MIME anyway. (That is, the MIME label might as well be
text/plain or application/octet-stream).
The only cases where you need a MIME label at all is when
the body is being sent as a message body; i.e., in a context
where you might be sending almost anything at all.
MIME is _not_ the only way of labelling data in Internet
protocols.
Even in those cases, applications where the type of message
body can be distinguished by the DOCTYPE within the XML body
need not define a separate MIME type.
In some cases, where the body isn't intended to be processed
by a general XML processor, but a separately coded processor
for some subset of XML that was designed for the application,
or where the bodies don't contain sufficient declarations
internally to distinguish them from other body types,
it might be necessary to register a separate MIME type.
So perhaps the calandaring applications have components
that use MIME but aren't intended to be processed by a general
MIME processor, and so separate MIME types are needed.
It would be useful to note that the generic text/xml or
application/xml designations are only appropriate for those
bodies that contain a DOCTYPE (or whatever new typing
mechanism you're going to add.)
I.e., "XML bodies that are intended to be used in Internet
protocols should always contain a DOCTYPE." You might even
want to restrict the use so that the initial segment is easy
to parse without invoking a general XML parser.
Larry