figDesc in p4 allowed <hi> in its content. In p5 it does not. In p4 hi is a member of hgphrase, along with "distinct emph foreign gloss mentioned soCalled term title" In p5 hi is a member of model.hiLike, while all the others are members of model.emphLike, and model.hiLike is not included in the content of figDesc, which seems like an oversight.
I was going to argue that <figDesc> is for occasions when markup is not appropriate, eg for constructing an HTML @alt or @title attribute on <img>, or for rendering in speech.
But then I looked at what it does allow...
The mistake, I _think_, is that model.hiLike should be a member of model.limitedContent, not model.emphLike
This is a reclassification proposal then, and quite a big one. Allow model.hiLike in model.limitedContent. All in favour?
There is macro.limitedContent and model.limitedPhrase. Which is intended here?
Sorry, my fault back in may. model.limitedPhrase. Add hilike to this, and it becomes parallel to emphLike (as it oughta)
At rev 9756 added model.hiLike to model.limitedPhrase. Nothing seems to have broken, and <hi>is now available in <figDesc>