[tei-council] issues on Schematron and deprecation

[Syd Bauman]

> >   1) 
> 
> This [#1, / (i.e., root)] has my vote; I can imagine circumstances in
> which you might want to require the presence of a PI or something
> like that.

Yeah, there is something nice about root, here. You can check for PIs
at the root from anywhere else, of course:
      <assert test="/processing-instruction()[3]">
        Not enough PIs for this investigation
      </assert>

I lean towards (1), the root, too. Besides being easy to do, it's
easy to document and explain.

The argument in favor of (3), (4), or (5) is that the expression
generated may be somewhat long and cumbersome, so that default would
save the user a headache. Saving the user "<rule context='/'>" isn't
really much to crow about. That said, I'm hard pressed to think of
constraints where those are particularly useful contexts. (For most
constraints I can think of, it barely matters what we do here: either
anything we generate would be fine because the rule only has to fire
once and it doesn't matter when; or nothing we generate would be
useful, so the user has to write their own context= anyway.)