On Thursday, June 16, 2005, 1:35:09 AM, MURATA wrote:
Chris>> With respect, you are, since I pointed one out in this thread. You may
Chris>> disagree with it, but you are not unaware of it. I have already
Chris>> mentioned content negotiation of schemas and the problems caused by
Chris>> serving both RNG and XSD as application/xml.
MM> Yes, of course. You pointed out the problem.
MM> Chris, Mark, and Martin think that an attribute indicating media types
MM> is very useful for standalone schemas and are inclined to forget
MM> embedded schemas for now.
Or at least, to say that reliably identifying a standalone schema does
not necessarily constrain any future solution for embedded ones.
MM> I think that an attribute specifying primary
MM> namespaces, for example, is more useful and it works for standalone
MM> schemas as well as embedded schemas.
Possibly; a mere list is not really quite sufficient, as the xslt
template shows. Nesting order is also important.
MM> I also think that the use of media
MM> types by the referrer should be avoided, since they may be different
MM> from what the server says.
Which has been looked at by the TAG and found to be consistent with Web
Architecture as long as it is treated as a fetching hint and not defined
to be authoritative over what the server sends, in the very few cases
where that turns out to be different.
--
Chris Lilley mailto:chris(_at_)w3(_dot_)org
Chair, W3C SVG Working Group
W3C Graphics Activity Lead