[Prev][Next][Index][Thread]

Re: text/xml volunteers needed



In message <199702010419.UAA28885@boethius.eng.sun.com> bosak@atlantic-83.Eng.Sun.COM (Jon Bosak) writes:
> The ERB decided not long after the November XML rollout that it was
> time to register the text/xml MIME type.  Since then, I've been

Jon, 
	This is very dear to my heart - I can't volunteer (see below).

[...]

Henry Rzepa and I spent > 1 year proposing the value of a new top-level
mime type (chemical/*) - in the end it faded into nowhere.  It raised 
strong feelings (some supportive) but my .sig for a MIME type application
may not be the best start at present. [In 
_practice_ the whole of our community now use chemical/* and so
CML will be transmitted as the (ugly and IANA-incorrect) chemical/x-cml.]

One suggestion made was that we should use text/sgml for CML, which though
_technically_ correct is useless _in practice_.  This is a problem which
is going to reoccur as we develop non-textual applications.  If I send
someone a molecule/integrated_circuit/etc. and they read it in a current 
'SGML browser' it won't
do them much good at present.  IFF the browser recognised the DTD _and_ 
had access for routing the whole problem to an appropriate system, fine - 
but that is asking a lot.

A second possibility is application/xml.  This would tell the client that
a text/sgml processing system was inappropriate for the contents of
the message.  Indeed the first priority (assuming that there were no MIME
subsubtypes) would be to process the DOCTYPE and retrieve the DTD.  Then 
the system would almost certainly have to retrieve the processing software
for that application.  With CML/XML this would be something like:

client receives application/xml document
client fires up non-textual XML processor (e.g. a parser + generic TOC tool)
client reads DOCTYPE
client resolves FPI (or URL) in doctype.
client retrieves DTD and parses the document (including validation).
client asks for processing software (stylesheets and/or Java), presumably
	from server or gets indirection from server.
client downloads software and processes document.  If document contains
	XML-LINKS to an application/xml document (possibly with a different
	DTD) it (may) repeat the whole sequence of operations for the 
	linked document.

(You will also see why I consider it important for XML-LINK to give as 
much information as possible about the type of the object before actually
reading it.  MIME types seem the best way. )
> 
> Here's your chance to cover yourself with glory and become one of

It is probably premature to submit application/xml at the same time
(see the complexity of the above) but I think it would be a useful
placeholder.  (And, in general, if people are developing XML applications
which don't fit into text/ , image/ , audio/ then they should either
be considered for application/xml or ... :-).  I cannot remember where
the official process has got to on subsubtypes, but we could surely
benefit from (say) application/xml.cml or similar


	P.
-- 
Peter Murray-Rust, (domestic net connection)
Virtual School of Molecular Sciences, Nottingham University, UK
http://www.ccc.nottingham.ac.uk/~pazpmr/