I suspect the reason for this is that <language> itself allows text content, so that's where the description belongs. However, <language> requires @ident, which must be a value according to BCP 47, so perhaps there is an argument for descriptive content in <langUsage> for cases where nothing whatsoever is known about the languages used. Cheers, Martin On 15-08-10 01:24 AM, Lou Burnard wrote:
The process of TEI content models proceeds apace and (as might be expected) is turning up a few oddities in some unfrequented corners.
For example, the content model of <langUsage> is currently (in the source)
<content> <oneOrMore xmlns="http://relaxng.org/ns/structure/1.0"> <choice> <!-- <rng:ref name="p"/>--> <ref name="language"/> </choice> </oneOrMore> </content>
Now, this is clearly nonsense : you can't have an alternation with just one option. But did we really want to remove the possibility for having a purely documentary <langUsage> such as <p>Contains various unrecognised Polynesian languages</p>?
Can anyone remember why we made this change ? I would have expected to see something like
<choice> <oneOrMore> <rng:ref name="model.pLike"/> </oneOrMore> <oneOrMore> <rng:ref name="language"/> </oneOrMore>
myself.