[tei-council] the "key" attribute

[Lou Burnard]

Sebastian Rahtz wrote:
> I have raised this before, but it did not not stimulate any response, so
> I am bringing it up again, with the reiteration that it a _mess_.
>
> The key attribute is used in two situations:
>
> 1. on elements  moduleRef, memberOf and specDesc to point to objects 
> in the TEI universe
> which are identifiable by having @ident. It's like the old ID/IDREF 
> pairing.
>
> 2. on elements like persName (all the things to do with naming people 
> and places)
> to point to the canonical object. Unfortunately, our examples use it 
> in two different
> ways
>  2a. as a URI pointer eg <country key="#FR"/>
>  2b as a database lookup somewhere eg  <name type="person" 
> key="ThorJon08">
>
> so that's three distinct data types used on one attribute,
> differering in our examples even on the same element!
>
> What are the choices?
>
> 1. shrug our shoulders and say "bof, compared to the loss of 
> Mourinho's dog it ees a nothing".
>
> 2. regularize all uses of @key to be URIs (screws up <memberOf> royally)
>
> 3. regularize all uses of @key to be database pointers, undefined 
> (loses chance to use external URLs)
>
> 4. replace @key with @target  in all the naming contexts
>
> 5a. allow both @key and @target on all naming objects (5b. making them
> mutually exclusive)
>
> I can live with 4, 5a or 5b. What say the rest of you?
>
> I think it is very important that we decide this before the meeting at 
> month's
> end to finish places, as it is intimately caught up with all that 
> persons/places stuff.
>

In order of decreasing preference:

4

5b

5a

I agree with Sebastian that we *must* do something about this. I think 4 
would be the least painful option, but like him can live with either of 
the 5s. Neither 1 nor 2 is acceptable. 3 would only be an option if we 
were also willing to revise all the <person>, <place> like things to use 
@ident rather than xml:id.