Good stuff Troy.
As xml is self describing it might indeed be a good for the reasons I briefly touched on (point adornment etc.). While, at this point, it may be overkill, I see nothing wrong with having the capacity to store what the current identified needs are (wait have they been identified?), rather, the problem is when a container or spec cannot host future unknown requirements.
While one might guess I'd now say "yeah, has my vote" no, I say -- "this worthy serious consideration, put it on the list of proposed formats and let us go now and identify what actually needs to be stored now as an absolute minimum, and what possibly be required for the future" -- which should be done first.
Rhetorical and not so rhetorial questions: What are the deliverables? What is the workflow? What minimum data do we need to start with?
Insomnia -- it's a lot more than lack of sleep.