[tei-council] Notes from today;s conference call

[Syd Bauman]

> The easy way to do this is the following:
>   element automobile {
>      (
>        attribute engineType { "int_combustion" | "hybrid" | "electric" }
>      |
>        element engineType { text }
>      ),
>      text
>   }

Right, although I think this model has the further restriction that
<engineType> must precede any character data inside <autombile>.


> For IDREFS, a model like this would work kind of nicely.

Right, but you misnamed the element. Also, we'd want separate
children elements, not 1 child with lots of URIs (which would be just
as hard to parse as if they were in attribute :-). So we'd want
something like

  element ptr {
     (
       element target { xsd:anyURI }+
     |
       attribute target { xsd:anyURI }
     )
  }

And, of course, everything would have to use patterns for
indirection; and we actually use the XML syntax in the ODDs. Thus the
problem META will have to address (whether now, in "6 months", or
never), is how to resolve the <attDef> below so that the second
<rng:ref> in the <content> points to it, and it is not instantiated
elsewhere.

  <content>
    <rng:choice xmlns:rng="http://relaxng.org/ns/structure/1.0">
      <rng:oneOrMore>
        <rng:ref name="target"/> <!-- a new element declared elsewhere -->
      </rng:oneOrMore>
      <rng:ref name="tei.ptr.attributes.target"/>
    </rng:choice>
  </content>
  <attList>
    <attDef ident="target" usage="???">
      <equiv/>
      <desc>specifies the destination of the pointer by supplying a
        single URI Reference. If there is more than one target, no
        <att>target</att> attribute should be specified, and instead
        a series of two or more <gi>target</gi> child elements should
        be used.</desc>
      <datatype>
        <rng:data xmlns:rng="http://relaxng.org/ns/structure/1.0" type="anyURI"/>
      </datatype>
      <valDesc>A single URI Reference; may use the TEI XPointer schemes.</valDesc>
    </attDef>
  </attList>

Mind you I have no idea whether or not this implementable.
Furthermore, I have not thought about what we'd do with usage= of
<attDef>.

BTW, I'd be perfectly happy if for every "normal" TEI element X
(i.e., those for which we do not want to take advantage of
co-occurence constraints) we needed to specify
   <rng:ref name="tei.X.attributes"/>
somewhere inside its <content> if it made this workable.


I'm begining to think I should have posted this to the META list
instead ... in fact, I'll cross-post it there, and if we want to
continue this technical discussion I suggest we move it there.