Syd Bauman wrote:
>> ok, probably true, but either way, <num>, <measureGrp>, <geo> all
>> belong in model.measureLike, which is allowed inside paragraphs. So
>> is <geo> more anomalous than <measure>?
> But my point is that <geo> doesn't belong in model.measureLike. Yes,
> it is semantically a kind of measurement, but no, it doesn't make
> sense to permit it in many of the places in which <meausure> makes
> sense. 
> I don't think either is anomalous, but I think that <geo> has a very
> particular purpose, syntax, and semantics, whereas <measure> is
> pretty general-purpose.

I agree with Syd that it seems weird to me to have <geo> in (from model.measureLike) and thus available in all 
sorts of weird places.  In my head it has always been an abstracted 
meta-data like element.

So <placeName ref="#foo"> points to a <place xml:id="foo"> in a 
<listPlace> in the header.  So what happens if I'm transcribing a 
geographical gazetteer... well then <listPlace> is allowed anywhere a 
<list> is allowed.

But then I wondered what I'd do if I honestly came across a lat/long 
just randomly mentioned in a transcribed text as two decimal digits 
according to WGS84.  I have to say the chances are pretty unlikely in 
the kind of source texts I work with.  But say in a digital-born text it 
might appear as something like "Verona is somewhere around: 4526'N 
1059'E." (which isn't WGS84 if those degree signs come out in your mail 
client) rather than "Verona is somewhere around: 45.433333 10.983333" 
(which is). It is the requirement for it to be WGS84 (or for you to 
indicate what other format it is in), which leads me to think of <geo> 
as metadata.

Thinking about this makes me feel that geographical references like this 
in a source text (that is not a geographical gazetteer of places) are 
the same kind of thing that names are with respect to <person> elements. 
  That is, they are tokens referring to a thing, rather than the thing 
itself. Thus, if I were so minded, I would mark up any references to 
lat/long whatever their format with some phrase level element and point 
back to the header where I had transcribed abstract information about 
that location.  At the moment, I am unable to envision a reasonable 
circumstance when true lat/long information would not refer to a 
location, since that is the raison d'etre of any such lat/long information.


Dr James Cummings, Research Technologies Service, University of Oxford
James dot Cummings at oucs dot ox dot ac dot uk