[relax-ng] Minutes RELAX NG TC telecon 2002-04-25

[Michael Fitzgerald <mike@wyeast.net>]

Minutes for a RELAX NG TC teleconference held on Thursday, 25 April 2002 at
9:00 a.m. EDT (UTC -04:00)

Our next telecon will be on Thursday, 9 May 2002 at 9:00 a.m. EDT.

Attendees

David
James
Josh
Kohsuke
Makoto
Mike
Norm

Not Attending

Fabio
John

Agenda

1. Escaping characters with \x{n}

James: Should we allow character escapes anywhere, resolving them at the
preprocess level?
Makoto: Are they allowed in tag names?
James: Yes
Kohsuke: Do we do normalization before processing escapes?
James: Yes. We do normalization first, then escapes, then other processing
Makoto: Exactly the same as XML?
James: Yes, except we allow escaped characters anywhere...XML does not allow
character references anywhere
Makoto: You can escape any UTF-16 character?
James: Yes, any that are valid in XML
Makoto: What about C0/C1 control characters?
James: Same as XML.
Makoto: Do we want to wait until XML 1.1?
James: No, let's go ahead. We will do XML 1.0 and deal with XML 1.1 when it
comes along.

Resolved: we will allow characters escaped with \x{n} anywhere.

2. Long string literals

James: We will use multiple adjacent strings. No more quote doubling [that
is] you have to break the strings into pieces in order to embed quotes
Mike: In Java you can embed quotes with \" \' -- we can't do that...
James: Backslashing escapes in regexes is hard ... you would need four
backslashes to get a single backslash, also in annotations.

Resolved: we will allow multiple adjacent strings to be concatenated.

3. Literal newlines in strings

James: Do we terminate strings with \x{} or use literal newlines?

Resolved: we will allow literal newlines.

4. div { ... }

James: We had a bare braces proposal, but it makes the grammar ambiguous and
there were problems with ##. I'd prefer not to have any special grammar
hacks.
Kohsuke: I think a keyword is better than bare braces []
James: Bare braces were too clever for their own good.

Resolved: we will use the div name as a keyword with braces and drop the
bare braces proposal.

5. ## syntax for <a:documentation>

James: ## seems to work well...
Mike: What happens if you don't declare a namespace/prefix for the
documentation element?
James: The processor will put one in for you [or] you can declare it
yourself. If the prefix a: is already used, [Trang] will use aa: instead of
a:, or aaa:, etc.
Norm: I prefer it as a special token...this is a complete win.
James: Does anybody have any problems with this?

Resolved: we will use ## for a:documentation

6. An encoding declaration

James: We don't want an encoding declaration now?
Makoto: I want to avoid special hacks; they annoy users.
Norm: [reviewed proposal]
James: You get UTF-8 or if the UTF-16 BOM is present, you get UTF-16
Makoto: [made a comment here I didn't capture]
James: If there is no charset parameter, you get ASCII
Makoto: Can't step 3 of you proposal be implementation dependent?
James: Yes
Kohsuke: So the Japanese can't use something like [ISO-2022-JP]...
James: We can't stop a processor from providing a command line option [that
could do this]...we'll do UTF-8/UTF-16 only
David: I have do depart now...I will post a proposal about improving RELAX
NG adoption
Kohsuke: I just thought it would be nice to have [some provision for other
encodings]
Makoto: We should encourage the use of a separate program like native2ascii
[to handle some cases]
James: So we will provide no magic encoding declaration then...
Makoto: I am not happy with this but I can live with it
James: I guess that's the closest one gets to happiness. <-- quote of the
week

7. Annotating the grammar element

Norm: There is an undiscussed issue: annotation of the grammar element
James: Yes, we can make the grammar element explicit...we can then place
annotation attributes on it more easily...I want to avoid using QNames
[without square brackets] because this may cause problems later if we want
to use QNames in tokens...what about the last point in the message 00075
[http://lists.oasis-open.org/archives/relax-ng/200204/msg00075.html]?
Mike: I am a little confused...I have the message in front of me now...I
prefer the square brackets example best [that is

[xml:lang = "en"] grammar {
  start = ...
  ...
}

]

James: That's what Norm and I like best too.
Makoto: Are we going to punt and decide this later?
James: We are not punting exactly...let's just decide...any not in favor of
this?

Resolved that we will have an explicit grammar element that we can annotate
with attributes in square brackets

8. Relative order of ## and annotation attributes [also in message 00075]

James: do people like a over b? I think b is more natural [## first,
[a:defaultValue="0"] second]...Java uses the formal comment first...any
opposed?

Resolved that we will use the relative order of ## first, [] second

9. Mixed schema

James: Makoto asked a good question about whether or not we should allows
mixed schema [RNG with RNC and vice versa]...with includes...we could
autodetect them...would require a MIME type
Norm: Let's not
Kohsuke: I think this would be hard to implement...comparing the difference
between include and included...
James: Implementations could experiment with this if they want...we won't
prohibit that
Makoto: I am happy with this
James: Kohsuke, have you tried this?
Kohsuke: No

Resolved [I think] that we won't require implementation to support both RNG
and RNC schemas but we also will not prohibit implementations from
supporting this if they want to

Makoto: Has anyone tried my online validator for RELAX NG?
[See http://www.asahi-net.or.jp/~eb2m-mrt/onLineValidation.html]
Mike: Yes, it's very nice. I particularly like the pull-down menus [that
populate] the text boxes...it is the best online validator I have seen
Kohsuke: Will you make the source available
Makoto: [Chuckle] Yes, but it is [sort of a mess]
James: How are you doing it?
Makoto: With Java and JSP...I will disclose the source...