On Mon, 20 Mar 2000 10:48:40 PST, ned(_dot_)freed(_at_)innosoft(_dot_)com
said:
I note in passing that we've already registered some XML-based types
with an "XML" suffix, and no problems have been reported with this
usage.
Has there been any registration of a 'foobar-xml' that duplicates
an existing 'foobar', but with an XML wrapping?
AFAIK, no. I don't see a reason to change an existing registration in any
case -- when we've made changes to the media types system in the past
we've grandfathered old registrations that don't "fit" the new rules.
If so, have any
user agents been known to have been modified to use the -xml as the
sort of trigger we have been discussing?
I always try to code stuff like this before making assertions as to its
ease of implementation. In this case I did so with the user agent I
use (PMDF MAIL). The changes for -xml handling were very simple and seemed
to work just fine.
I also attempted to code the changes for dispatching off of a global parameter
and adding a global parameter when sending. The former turns out to be doable
but kinda ugly; the latter isn't doable in my UA short of a new product release
because the format of a critical data file has to change to accomodate it. (I
don't have a complicated GUI to contend with either.)
FWIW, this also got me started looking at implementation of content-features
support in our software. This is, well, interesting. The facility is well
specified as far as I can tell, but part of it involves implementation of a
expression canonicalization routine. Now, I used to work on the symbolic
algebra system in a package called MATHLIB, so this is known territory for me,
but I wonder how well someone who's never done anything like this would handle
it. It certainly is way more complex than anything else we've discussed,
although its addition doesn't present a problem from an interface or file
compatibility standpoint.
And there's also the problem of generating feature expressions describing stuff
that varies from document to document -- as I see it this can be done either by
a generic analysis step or else by adding some communications channels
composition programs can use. Both of these amount to pushing the problem off
to some other component, and I have to wonder whether or not composition agents
will actually evolve to the point of providing such information or analysis
programs of this sort will be written.
Ned