Lists Home |
Date Index |
At 05:08 PM 5/13/2002 -0400, Mike Champion wrote:
>5/13/2002 3:14:45 PM, "Danny Ayers" <firstname.lastname@example.org> wrote:
> >Hmm - not sure I like this direction at all. float, double, dateTime - all
> >pretty ugly, I'd leave the definition of things like this to separate
> >namespaces/schema, and leave them out of these specs altogether. In fact,
> >I'd probably drop everything but booleans & integers.
>All I really care about here is that there be a conformance level that is
>not a big
>conceptual stretch for XPath 1.0 users and not a burden for implementers.
I agree that this is needed.
In general, an XQuery user can ignore the presence of datatypes and the
right thing happens, unless they make a mistake. Implementors, on the other
hand, have to implement all these data access functions, which is a royal pain.
>[FWIW, I'm not persuaded by the "it's just a bunch of 1-line functions,
>no big deal for implementers" argument. Each of those has to be put
>in a test suite, tested, regression tested regularly, documented, the
>translated, the help desk trained upon, the XQUERY FOR DUMMIES readers who
>can't understand the subtleties supported, ad infinitum.]
Yes, though I frankly wonder if most of these types are important enough to
put in the XQuery for Dummies book. But I can't say that, or you'll rub it
in my face....
The XML Schema Part 2 Datatypes tend to be treated as a component that is
dropped in. People use none of them, or all of them (though one might argue
for just the primitive types, or just the primitive types plus integer.) 
So I agree, fewer types would be better. On the other hand, for the user,
too many types is better than none at all, since most of these types can be
safely ignored. For the implementor, that does not hold.
 And if you really want to get me in a bad mood, ask me to defend the
decision that integer should not be primitive. But if you do that, I simply
won't respond to your email.....