This is a personal review of the XPointer last-call draft of 8 January 2001. It reflects the personal views of the author, and these are not necessarily the official position of any group or official body.
First, well done.
The end of para 1 suggests that the extension facilities described here are to be used only with MIME types other than text/xml, application/xml, etc. If it were feasible (I bow to the considered opinion of the WG on this), I think it might be useful to allow extensions even in material shipped with these types. The example I have in mind is XML Schema. XML Schema documents are likely to be shipped as text/xml or application/xml (the XML Schema spec is to be agnostic on this point), but users would, I think, benefit from a scheme allowing the simple identification of arbitrary schema constructs by reference to their nature and local name, without reference to the element structure of the schema document in which they happen to be declared. I don't see any severe problems arising in consequence, but this may reflect naïveté on my part and I am willing to be enlightened.
Section 3.3 seems to imply the opposite, that unknown schemes will be skipped (not treated as errors) in fully conformant applications. That suggests that the wording here might need to be revised.
As one of those who worked on the TEI Guidelines, I thank you for mentioning them here.
For "items comprising minimal conformance" read "items comprised in minimal conformance" or "items included in minimal conformance".
For the reference to IURIs you should probably now substitute a reference to IRIs, as defined by Masinter and Dürst's Internet draft.
For the reference to IURIs you should probably now substitute a reference to IRIs, as defined by Masinter and Duerst's Internet draft.
The inclusion of character ranges is an important feature; don't waver!