producing the draft of XML Datatype spec

At the request of Dan Connolly, this message is an attempt to describe how
we produced the first working group draft of the XML Dataytpe spec (part of
the XML Schema spec) [1].

The short answer is...

The draft itself was authored in XML using the XML Spec DTD written by Eve
Maler [2].  The HTML version of the draft was produced by an XSL stylesheet
originally written by Eduardo Gutentag.

The longer answer is...

The editors have found Eve's DTD very easy to author with.  We were up to
speed very quickly, due largely to the "intuitiveness" of most of the
content models (at least, they are intutive to me).   We have only had to
refer to the "report" (or the DTD itself) infrequently to know what was or
was not legal in some context.

So far I've only found one minor thing that I'd change and a few small pet
peeves.  As to the pet peeves, I think the content model of things like list
items should be mixed and the DTD requires paragraph-type block structures,
i.e., the DTD requires


	<ulist>
		<item><p>this is an item</p></item>
		<item><p>this item has <emph>emphasized</emph>
text</p></item>
	</ulist>

whereas I'd rather do

	<ulist>
		<item>this is an item</item>
		<item>this item has <emph>emphasized</emph> text</item>
	</ulist>

Its not really a big deal, but it is a personal pet peeve (see below).

The thing I would change, however, is the content model of the editorial
note element (actually, the edtext sub-element of ednote).  Currently, it
allows only #PCDATA, with the rationale

	The content of edtext need not be more complicated than #PCDATA
because the note doesn't need to contribute to the "real" content of the
document.

We've found that it is useful to include sub-elemenets in edtext such as
various reference or link types, lists, etc.

Also, we've introduced a "usage convention" for the optional date and name
sub-elements of ednote which goes against the stated description of ednote
("The ednote element identifies commentary passed between editors and
authors of a document.").  In particular, we've used "signed" ednotes for
communication between and "unsigned" ednotes for communication between the
editors and the WG.  We'll see if the distinction is useful.

The XSL stylesheet produced by Eduardo is great!  Makes it so easy to
produce html output.  One thing that isn't noted (in the report or in the
stylesheet itself) is that it uses 2 constructs *not* in the current WD of
XSL, but which are implemented as "extensions" in the 19990307 version of XT
[3].

I've made a few changes to the XSL here and there, tho nothing major.  For
instance, I had to add templates for a few element types such as some of the
reference types (xnt|xspecref|xtermref) and some pieces of the table model.
Also, I've redefined several templates to produce output more to my liking
such as removing "unnecessary" paragraphs in list items (re: my pet peeve
above), sorting the references, etc.  Lastly, I've introduced a mechanism
which gathers all of the *issues* together in an appendix for easy
references, which hopefully will be useful to WG members.

We intend to share all mods with both Eve and Eduardo (and this list, if
anyone is interested) when we are more sure that our initial changes should
stick.

References
[1] http://www.w3.org/XML/Group/1999/04/WD-xdtl-19990326.html
[2] http://lists.w3.org/Archives/Public/spec-prod/1999JanMar/0003.html
[3] http://www.jclark.com/xml/xt.html

Paul V. Biron
co-editor, XML Datatype Language (XML Schema WG)
Kaiser Permanente, for Health Level Seven

Received on Thursday, 1 April 1999 21:07:25 UTC