[tei-council] limited phrase, take 2

[Lou Burnard]

I think I prefer the second of your proposals (i.e. make 
model.limited.phrase a superset of existing low level model classes), for 
the following reasons
a) it's what we originally decided to do
b) it makes it a little easier for users to decide for themselves what 
does into model.limited.phrase: because it works with the existing model 
definitions, it's a bit easier to say "I want that kind of thing in my 
header too" (or not)

c) it avoids having to rethink all the existing class assignations

Reasons (a) and (c) are purely pragmatic of course and largely conditioned 
by the desire to get P5 out sooner rather than later.

Another (third) option, if we do want to reopen this can of worms though, 
might be to define a new macro "anorexicPhraseSequence".

Let's remind ourselves of a use case for this. I am (currently) rather 
obsessed by the BNC, so forgive me for harping on about that. I want to 
be  able to say that the <p>s in my corpus headers have a content model 
which  is a specific subset of the <p>s in my corpus texts (only PCDATA, 
not PCDATA or <s>, for example). How would each of these proposals help me 
with that problem?