[relax-ng] Minutes of RELAX NG TC 2001-10-11

[Michael Fitzgerald <mike@wyeast.net>]

Minutes for the RELAX NG TC meeting held on Thursday, October 11, 2001, 10:30 am EDT (UTC -04:00)

Attendees

Fabio
James
Kohsuke
Makoto
Mike
Norm - momentarily, with regrets

Not attending

David
Josh

1. Errata in current spec

(a) data/except//empty

Yes, we will disallow empty as a descendant of data/except.

(b) start//empty

Yes, disallow empty as a descendant of start.

(c) disallowing relative URIs for ns and datatypeLibrary should not disallow
empty string

Yes. 

(d) unreachable elements

Yes. Reword restriction. Fixes bug in 4.18. See http://lists.oasis-open.org/archives/relax-ng/200110/msg00007.html.

(e) attributes named xmlns. What if there's also a non-empty ns attribute?

Yes, this should be an error if ns="".

(f) Disallow attributes with namespace "http://www.w3.org/2000/xmlns" (the
namespace URI that the infoset binds to the xmlns prefix)?

Kohsuke: Just say nothing. It is just an unusual pattern.
James: I will think about it and propose a solution [and post to list].

(g) <list> with multiple child elements

Makoto: This is obvious.
Kohsuke: Yes.

Done. See http://lists.oasis-open.org/archives/relax-ng/200109/msg00076.html.

2. Issue 57 (whitespace normalization)

James: Whitespace is tricky. Let's take another week to think about. [Makoto had not considered the issue fully.]

3. Issue 55: external grammar reference mechanism

Kohsuke: I agree to close this with no action.
Makoto: I am happy to close this. The door is open for future extensions.
James: This is not a nasty problem.

Closed with no action.

4. Associating RELAX NG schema with a document.

Makoto: What about the catalog-based solution?
James: Perhaps inventing a catalog-based system is not too complex but I think it is too complex for RELAX NG. I think it fits more along the lines of general XML processing.
Makoto: A PI is out of scope for RELAX NG?
Kohsuke: A PI is not a complete solution in the short term. We need a language neutral solution. There are going to be JAXP processing properties for this.
James: It's going to be tricky.
Fabio: I'd like a PI solution.
Makoto: I don't want a PI.
Kohsuke: Abstain.
Mike: Not critical to have a PI now. An implementation could meet their own needs with a foreign attribute.
James: I will make this a FAQ entry essentially saying "Not for now." We need a general solution that is not RELAX NG specific.

A solution for associating schemas will not happen for 1.0 of RELAX NG.

5. Martin Bryan's comments based on DSDL solutions paper.

James: Most of Martin's comments in his document have to do with instance documents and are not really part of schema processing. For example you could do RELAX NG processing after Xinclude processing.
Makoto: What is the status of this document?
James: I think it is part of a new work proposal that is part of SC 34 [?].
Makoto: I don't think it is.
James: I think so.
Makoto: You could simply request more input.
James: I think that is a good way to go...but we need to provide a proper answer for the minimum order question (minOccurs/maxOccurs question).
Kohsuke: There is a shorthand for that.

James will reply that we will not do minOccurs/maxOccurs for 1.0 and request clarification or expansion on other issues.

6. Issue 40: restriction on interleave

Kohsuke: A cached branch must have a disjoint element name! Same as attribute processing.
James: That's a natural restriction. Would Jing not explode [combinatorial explosion] if this restriction were in place?
Makoto: Yes.
James: But this is not sufficient for subsumption of other schemas. We don't want derivative-based implementations to blow up.... We don't seem ready to agree on a restriction on text.
Makoto: It is important to discuss the restriction of the wildcard when a wildcard is a subordinate of oneOrMore. I will talk to Hosoya-san.
Kohsuke: What about oneOrMore and zeroOrMore of attributes? It's not complete.
Makoto: I think it is complete.
James: Kohsuke, can you open an issue on this?

Action Item: Makoto to discuss current status or progress with Hosoya-san.

7. Schedule for releasing RELAX NG 1.0

James: We need at least two 100% compliant implementations before we release 1.0, along with all the specs. [tutorial, lang spec, compatibility spec, data type spec]. I'd like us to be finished by XML-1 (Dec. 9-14, 2001). [Orlando]
Makoto: What about November?
James: Yes, November I think. Kohsuke, how close are you to finishing MSV?
Kohsuke: It is almost fully compliant. The changes are very simple but the release process [at Sun] takes about a month.
James: We need to discuss the remaining restrictions so we can do the implementations.
Makoto: [By the way] I am doing two chapters for a book on schemas by Addison-Wesley. I will be discussing RELAX NG.
James: I won't be discussing RELAX NG at my keynote at XML-1, but I will in my presentation.

Mike