1. use the internal subset to declare IDs.
Wouldn't this have even deeper implications?
It shouldn't, for conforming implementations of XML.
By this I mean that a
validating parser won't validate a document that has no DOCTYPE
references (treats it simply as well-formed), but *will* try to validate
a document in an all-or-nothing fashion if there is.
The XML spec doesn't say any such thing. In fact it's quite clearly
possible to validate documents with no <!DOCTYPE ...> declaration
(and get lots of errors). The conformance section of the XML spec
describes validating parsers, and non-validating parsers, but not this
"sometimes validates" behavior you describe.
You may be confusing the XML standard with what Microsoft does
in some of their implementations.
- Dave
p.s. why is the IETF mime list CC'd?