[tei-council] agenda items for Council telecon on 2007-06-15 at 1200 UTC

[Syd Bauman]

First, I just want to verify that the upcoming conference call is
scheduled for this coming Friday 15 June at 12:00 UTC[1].


> As usual I would like to collect things you think that need our
> attention and would like to see on the agenda. Please send them
> either to me or to the Council list. I plan to send out the draft
> agenda next Monday.

I do have one item that, if we had time, I would have wanted to run
past Sebastian first for a feasibility check. However ...

Council has charged me with implementing Schematron rules in the
Guidelines for expressing a variety of constraints that are not
expressed in the RELAX NG schemas produced by the ODDs. A very large
percentage, probably even a large majority, of these rules are going
to be of the form "the X= attribute of <Y> should point to a <Z>".

While at the DH conference at UIUC, I saw a very interesting poster
by Kevin Reiss of CUNY. His major point for our purposes was to
better describe the semantics of XML languages by some extensions to
ODD.

I quickly noticed three major useful suggestions in his poster. One
for describing the inheritance of attribute and one for describing
the semantics of child elements. I think both of these are good
ideas, but I think they require more fleshing out and more work on
the Guidelines than is feasible for inclusion in P5 1.0. However, his
third suggestion is directly apropos to the issue described above,
and I think may well be reasonable to include in P5 1.0.

I have revised his extension a little; here it is.

We add a references= attribute to <attDef> which could be used only if
the datatype of the attribute being defined was data.pointer.[2] The
references= attribute itself has a datatype of data.name, and gives
the name of the element type to which the pointer attribute being
defined should point. The ODD processor would generate the Schematron
needed to validate that instances of the pointer attribute actually
pointed to elements of the type named on references=.[3]

Problem: what would the message in the generated content of the
<sch:assert> or <sch:report> be? I haven't even thought about it yet,
but I doubt it's a show-stopper. This does, however, violate the "no
new additions post Berlin" idea. 


Notes
-----
[1] 06:00 MDT in Calgary,
    07:00 CDT in Chicago,
    08:00 EDT in New York,
    13:00 BST in London,
    14:00 CEST in Paris,
    21:00 JST in Kyoto, and
    Sat 06-16 00:00 NZST in Wellington.

[2] I.e.,
    <rule context="tei:attDef[@references]">
      <assert test="tei:datatype/rng:ref[@name='data.pointer']">The
      @references attribute may only be used if the attribute being
      defined is a pointer.</assert>
    </rule>

[3] I'm not 100% on what the Schematron for this would be, in part
    because I haven't figured out if ISO Schematron is well enough
    implemented we can and should use it (I suspect it is), or if the
    document() function has been implemented properly ala the XSLT
    2.0 specification or not. Shouldn't be too hard to test, just
    haven't gotten to it yet.