[relax-ng] Minutes for RELAX NG TC Meeting 2002-03-14

[Michael Fitzgerald <mike@wyeast.net>]

Minutes for a RELAX NG Technical Committee meeting held on Thursday, March
14, 2002.

The next meeting of the TC is scheduled for Thursday, 28 March 2002 at 9:00
am EST (UTC -04:00).

Present

Kohsuke
James
John Cowan (observer)
John Evdemon (observer)
Josh
Makoto
Mike
Tom Gaven (observer)

Absent

David
Fabio
Norm

Agenda

1. Datatypes

James: What do people think about us developing our own datatype library?
Makoto: I think we should define our own conformance level for XML Schema
Part 2 (XSD)...
Kohsuke: We could just extend our current guidelines
(http://www.oasis-open.org/committees/relax-ng/xsd-20010907.html).
James: We could recommend our own subset. It would be better rather than
worse.
JohnC: An implementation would just issue warnings [if a datatype were not
in the conformance level]?
James: Kohsuke what did you do about warnings?
Kohsuke: I followed XSV!
James: The problem with XSD is the spec is not clear. There is confusion
between the value and lexical space, such as whether a [unary] + is allowed
to precede 0.
JohnC: So it's self-contradictory!
James: Yes, also things like are decimal points allowed in integers...
JohnC: "If numbers are problematic, then everything is a problem."
JohnC: I have been working on constraining date [types] with regular
expressions. It's just about working.
James: Should we adopt a profile of XSD then?
Makoto: I'd like to adopt a profile. A huge set of types is a problem for
implementers.
John: I think it may be out of scope for the TC to do this.
Makoto: I think we should do it as it will help implementers.
Kohsuke: I am not quite sure.
Josh: Sounds like a good idea to me.
Mike: We have already essentially done it [with XSD guidelines]. I don't
think it's out of scope to do our own profile.
James: I have mixed feelings. A profile is needed, lots of things need it.
JohnC: I'll suggest a compromise. Start the profile, then launch a separate
TC at OASIS to handle it.
James: I don't know if W3C will go for that...
JohnC: Well, XML is just a profile of an ISO standard. [good laugh]
James: Maybe we could send a formal message to the XSD WG and ask them to
create a subset, or else we'll do it.
Kohsuke: Should we just wait another week and then decide?
James: No, let's decide between four options: (1) do nothing, (2)
communicate with the XSD WD, (3) encourage the formation of a new OASIS TC,
or (4) create a RELAX NG spec for the profile. Straw poll:
Makoto: Can't live with 1, can live with 2, prefer 3 or 4.
Kohsuke: I prefer 2 but can live the other options.
Josh: I prefer 3 but can live with any of the options.
John: I prefer 3, but I can't live with 1.
Mike: I prefer 4, but can live with 2 or 3.
JohnE/TomG: We prefer 2.
James: I prefer to do 2, then if no action, do 3, then if that doesn't work
go to 4...perhaps we should try to get OASIS to do 2 for us.

TC agreed to follow the option 2 then option 3 then option 4 approach.

2. DOM Abstract Schemas
(http://www.w3.org/tr/DOM-Level-3-ASLS/abstract-schemas.html

JohnC: They issued a new spec a couple of weeks ago. I've been reading it.
They are doing some things such as universal names, some top-level
declarations, no objects names, load/save. RELAX NG is a reasonable
candidate for it. If we ask for a few things up front, we would hopefully
get them, such as interleave. Will probably provide a fairly good view as
RELAX NG via an API.
Kohsuke: I am not so sure it will be useful.
James: If it's not useful, it's not useful. I don't think it will get used
much.
JohnC: You may be right.
James: Should we take the time to do something?
John: I volunteer to draft a request for some RELAX NG provisions for the
DOM AS people....I believe you can validate with the AS model...
James: Will they define what that validity is then?
JohnC: Well it would rescrutinize as objects rather than as characters.
Makoto: I am not particularly interested.
Kohsuke: It would be nice to be able to do the document editing part...
JohnC: I am interested.
Josh: I don't think we should do it.
Mike: I think we should watch it, but not do anything now. Doing something
early is better than later...
James: Watching it will take TC time. Should we do it or not?
Mike: We don't need to take time now.
JohnE: We agree with Josh, but you could kick it to next week.
James: I don't think it's important.
John: Then I will submit personal comments about adding RELAX NG features to
DOM AS.

3. Type assignment

James: Type assignment is useful in some applications when nonambiguity is
[needed], or you need lookahead.
JohnC: ...when you need some unique types or a list of simple types...
Mike: Doesn't MSV do some sort of type assignment? How does it do it?
Kohsuke: I can't explain it without a lot of jargon? It's not using
context-free grammar.
John: So it's LL(0). [no token lookahead]
Kohsuke: Yes.
James: MSV will not report type assignment.
Kohsuke: Not many are using it.
Makoto: Relaxer does it but oneOrMore of oneOrMore is ambiguous.
Kohsuke: In RelaxNGCC, oneOrMore of oneOrMore is not ambiguous.
James: We can pick up this discussion in a couple weeks.