[tei-council] specDesc spec vs. reality

[James Cummings]

On 09/12/13 12:29, Lou Burnard wrote:
> It *used* to be "all those available" before we invented
> attribute classes. Then we argued quite a lot about whether it
> should include all those available by inheritance or only all
> those locally defined. Then you implemented "only those
> explicitly stated"

I would expect that if I said:

<specDesc key="foreign"/>

that the default would be that I would get all attributes, local 
or inherited. Possibly might have some indication of their source.

If I wanted only some I would specify them as:

<specDesc key="foreign" atts="rend corresp ana facs change"/>

and I would expect that they would come with their descriptions 
in the order specified by @atts. I'd also expect that valLists 
were not displayed unless I added an attribute asking for it. 
(whatever that attribute might be.)

>> all I can think of is to say <specDesc
>> display=“{full|medium|minimal}”>
> That just punts on deciding what "full" or "medium" means, or
> leaves it to implementors to decide

I don't think that is necessarily bad, it means that all sorts of 
display features could be categorised into full/medium/minimal 
but prefer your idea below for withValList.

> You could  say  <specDesc key="valListkey"/> if valLists were
> identifiable, but as they're not something more precise is
> needed: how about
>
> <specDesc atts="a b c" withValList="a b">
>
> defaulting to "no valList display".

My only worry about that is what if you always want to show all 
valLists. You don't want to update the <specDesc/> in every page 
when we add a valList to an attribute. (i.e. the attribute didn't 
have a valList before, so we didn't include it in @withValList, 
but now it has so we have to go find all those places and update it.)

> While we're looking at this, I have long wanted some way of
> grabbing <egXML>s from the specs for display in the
> descriptive prose (or indeed vice versa)

I agree. At various times I've suggested re-examination of how 
our corpus of examples works. (I believe I suggested separating 
them from specs and prose into a separate corpus and then pulling 
them in wherever required -- but I'm open to other improvements.)

-James

-- 
Dr James Cummings, James.Cummings at it.ox.ac.uk
Academic IT Services, University of Oxford