[relax-ng] Minutes for RELAX NG TC Mtg. on 2001-10-25

[Michael Fitzgerald <mike@wyeast.net>]

Minutes for a RELAX NG TC meeting held on October 25, 2001 at 10:30 am EDT
(UTC -04:00)

Attendees

Fabio
James
Josh
Kohsuke
Makoto
Mike

Not attending

David
Norm

Note: No meeting next week or the week after. [For those transitioning from
Daylight Savings Time to Standard Time, the meeting time will remain at
10:30 am EST but will feel like it is an hour *later* not an hour earlier
(hint: if you are north of the equator just remember 'spring ahead, fall
back into bed'!)]

1. Issue 65. Attribute wildcard. The current proposal is that, after
simplification, if an <attribute> has a <nsName> or <anyName> descendant
then it must have a <oneOrMore> ancestor.

James: Are we happy with this?
Makoto: I certainly am.
Kohsuke: Are we sure that this restriction is...
Makoto: I have written a proof...
James: It makes sense from an intuitive point of view...are any opposed to
this?
[No dissent.]

Passed as currently proposed

2. Issue 40. Restriction on interleave.  Any further news from Hosoya-san?
Should we allow <interleave> inside <list>?

[Long]

Kohsuke: I think this is possible with XML Schema using some abstract
syntax...
James: Can you find a reference?
Kohsuke: OK I will.
James: I am not sure we will find a satisfactory solution...
Makoto: The current draft of the spec if fine. Let's go on.
James: OK so we don't allow <text/> in both operands of an <interleave>
which leaves the issue of <interleave> in <list>. Should we allow it?
Makoto: I am leaning towards disallowing it.
Kohsuke: I lean towards allowing it.
Josh: I don't know. Is it allowed in XML Schema?
James: You can do something like it in XML Schema.
Mike: I lean to allow.
Fabio: I lean to disallow...strongly.
James: Where am I? I lean to allow.
Fabio: I am trying to implement <interleave> and am having trouble. When
compared with the small number of use cases and the difficulty of
implementation, I think it falls in the 20 side of the 80/20 rule.
James: Have you implemented <interleave> of elements?
Fabio: No. I have not implemented all of <interleave> yet.
James: I don't think you have to add a single line for allowing <list> in
<interleave>.
Fabio: I guess I'm scared of it.
Makoto: I have heard from another implementor who has complained about this.
His implementation uses back-tracking.
Kohsuke: Did he complain about <list> in <interleave> or about <interleave>
overall?
Makoto: Apart from other restrictions, elements, <text/>, <data> in both
operands, what about the rest of restrictions on <value>?
Kohsuke: They are the same?
Makoto: They are the same but it *looks* complicated. To me, it is only a
subset of a finite set. I am doubtful.
James: Who was it that complained?
Makoto: [Unknown] who did VBRelaxing.
James: Okay, let's see who can live with the opposite...I can live with
disallowing <interleave> in <list>.
Makoto: I can live with allowing it.
Kohsuke: OK, I can live it disallowing it.
Fabio: I can live with allowing it.
Josh: What about 'When in doubt, leave it out'?
Mike: I am wondering about those who have implemented it...was it hard?
James: It took a little while, but that's because I was trying to get the
first part right. After that was done, it wasn't hard...I just did <value>
the same way as the element restriction.
Fabio: I withdraw myself from the question!
Mike: I am still for allowing it.
James: Does anyone have any other uses cases for this?
Kohsuke: It only took me about 100 lines including comments to implement
this.
Mike: It may seem hard but an implementer could look at source code from
Jing or MSV to get some ideas...we need to document this.
Fabio: So the bottom line is ignorance!!
Kohsuke: Somebody needs to write a book...
James: We need to do a cook book for implementers...that will make it
easier.
Josh: I am for not allowing.
Makoto: I am for not allowing.
Fabio: Not allow.
James: Are there other use cases? It is a marginal case, and we can do it
with <choice>...
Kohsuke: That means 8 [instances of <choice>]...
James: Given how rare it is, is that so terrible?
Kohsuke: But it affects data binding...
Makoto: ***After all, schema languages cannot capture everything!***
James: Another argument, if we prohibit <interleave> in <list>, we can
always allow it later in say 1.5 without needing to change the namespace
[which indicates the version number of RELAX NG].
Kohsuke: It's OK. I can live with it both ways.
Mike: If it is easy to add later, then I will go the other way [to disallow
it].
James: I propose that we make this a prohibited path
[list//interleave]...does any one object?
[No objections.]

<interleave> in <list> did not pass.

Other Issues

James: I will need to rev the spec, the tutorial, and the compatibility
spec. I need to implement the compatibility spec...haven't done that yet.
Kohsuke, have you?
Kohsuke: Yes I have.
James: We'd like to have at least two implementations...We need some time to
do some serious proof reading...I will be going to Japan [week after next]
and will meet with Hosoya-san [in part to discuss <interleave>]...not much
sense in having a telcon next week or the week after.
Kohsuke: Do you think there will be any changes or can I submit it [MSV] to
[Sun's] legal for release?
James: I think you can go ahead. I will contact you all for our next meeting
time in mid November.

Mike
=====
Wy'east Communications     http://www.wyeast.net     mailto:mike@wyeast.net