To ensure widespread usability across protocols, tokens should follow
   the character set restrictions for XML Names.

This needs a reference. Also, is it really 'should'?
I agree that this makes most sense as a MUST. I can't see any reason to
deviate from the restriction.


It is recognized that the terms used may be ambiguous in the real world.
The aim is to support common sense in the choice of labels, rather than
providing a mathematically consistent algorithm for choosing labels.
Applications using this registry should be designed in the knowledge
that labels may be used imprecisely.