[tei-council] datatypes
Syd Bauman
Syd_Bauman at Brown.edu
Sun Sep 18 01:16:51 EDT 2005
Not surprisingly, my disagreements with this proposed list of
datatypes fall mostly in the places where it differs from those
recommended in EDW90 (as I said I was going to modify it on the
conference call, but haven't bothered since Lou in essence moved the
declarations to DTYPES).
* tei.data.probability: Hmmm... this might work. It is a bit
confusing. The problem is that, because the recommendation in
DTYPES has removed the percent sign, there is no way to distinguish
"1" meaning 100% from "1" meaning 1%. However, we could write a
prose rule that required users who want to use "1" to mean 100%
write it as "1.0" (which, because the percentages are declared as
integers, and the check is done on the lexical space, not the value
space, could not be a percentage). This seems somewhat fragile to
me, so I still think the original idea (requiring a percent sign
for percentages) is better, but this one is not implausible.
Note also that the EDW90 pattern permits fractions of a percentage,
whereas the DTYPES pattern does not. I'm perfectly happy either
way. (There is no current application of this datatype where
fractional percentages would make any sense at all; one can imagine
that it might be used for something else for which more precision
would be appropriate, though.) In either case, having now done the
calculation to figure out how tiny a number can fit into an
xsd:float (~1.4e-45), I am convinced we should change xsd:double
to xsd:float.
* tei.data.numeric: the change removes support for a constituency
that we already now about: those who need to enter floating point
numbers. Furthermore, I still claim it makes sense to permit
percentages. (One could argue, though, that the percentages should
be limited to 8 characters maximum (effectively limiting them to 3
or 4 decimal places of precision), so that any tei.data.numeric
value could fit into 64 bits.)
Well, I've gotten through group 1, numeric. I hope to tackle the
others tomorrow.
More information about the tei-council
mailing list