Most peculiar. The pure odd content model generated on dec 4 just needs a <sequence> round it. I cannot remember (or understand) why I added constraint into the alternation instead. Must have been something I ate. Have fixed it now in dev. On 16/03/16 20:58, Syd Bauman wrote:
I really think this can't go out like this. The problem crept in when Lou went to fix a problem with DTD generation from PureODD. Apparently only this file was affected.[1]
Lou -- can you describe what the DTD problem was, in the hopes that we can come up with another solution?
If we can't come up with a PureODD solution, we could add a Schematron constraint.
Notes ----- [1] https://github.com/TEIC/TEI/commit/bfc5e4e55dbec131f25a87090e24c52d86f28c06
The current released content model of <constraintSpec> is
( model.glossLike | model.descLike )*, constraint?
The content model of <constraintSpec> in the dev branch is
( model.glossLike | model.descLike | constraint )*
Thus the new content allows multiple <constraint>s. As much as I was the one arguing for this in the past,[1] until we're settled on exactly what it *means* to have 2 or more <constraint>s inside a <constraintSpec>, I think we should stick to the old content model.