[tei-council] num
Lou Burnard
lou.burnard at oucs.ox.ac.uk
Wed May 27 18:24:54 EDT 2009
Please forgive me for bringing this up again, but I am not convinced
that the decision we appear to have agreed in Lyon re.# 2673045 is the
best one and I am uncertain how to proceed.
If you look at the ticket
(https://sourceforge.net/tracker/?func=detail&aid=2673045&group_id=106328&atid=644065)
you'll see that what Gabriel wants is the ability to represent the value
of the @value attribute on num as a fraction e.g. 34/56
In Lyon we agreed that the way to do this was to define another
attribute @valueType with possible values decimal, float, and ratio. I
can do that easily enough, but it won't help Gabriel's original
requirement unless I also change the datatype of the @value attribute
to permit values such as "34/56" (presumably by defining some sort of
pattern). At present @value is data.numeric which is mapped to a choice
of XSD schema types "double" and "decimal", neither of which would
support fractions.
In Lyon we did not discuss this aspect of the problem (or if we did I've
forgotten it) and I can't see any discussion of it in the ticket
either. Was the intention to (a) just forget about validating the
@value attribute or to (b) extend its possibilities to include a pattern ?
Thinking about this problem on the train home , I felt that this was
really not the right way of approaching the issue. I think what Gabriel
asked for in the first place -- a different value-bearing attribute --
was the right answer. Having an attribute that tells you whether
another attribute is
decimal, float, or ratio seems redundant in a world where you already
have datatype checking.
Extending the range of an existing constrained datatype to include
something which many processors will regard as just plain broken seems
a really dangerous idea. And I don't even want to think about the
problems of how you determine what combinations of values for these two
attributes are mutually consistent.
My recommendation is that we don't introduce a @valueType attribute.
Instead we introduce a @ratio attribute (better names welcomed), the
value of which is constrained by a pattern something like \d+\/\d+ and
which is not allowed to coexist with the @value attribute (that is
something I think we can already represwent in ODD, tho I suppose a
schematron rule would also be a good idea)
Apologies again for digging this up again, and for the delay in
implementing decisions we thought we'd agreed to!
More information about the tei-council
mailing list