Lists Home |
Date Index |
On Fri, 2002-05-10 at 07:09, Richard Tobin wrote:
> >I'm asking today for a fully-instantiated PSVI, one that gives
> >programmers immediate access to the typed data the PSVI promises
> In fact, the PSVI doesn't provide "immediate access" to typed data.
> There's no property that gives you the typed value of an element or
> attribute. All you get is the [schema normalized value] - which is a
> string - and the type. It would be natural for a PSVI implementation
> to provide the typed value (eg as a Java object of a suitable class),
> but the PSVI itself doesn't address that.
Technically speaking, and limited explicitly to present specifications,
Richard is completely correct.
On the larger landscape of the rhetoric with which schemas and typing
have been sold, particularly the usage which such things are finding in
the Web Services world, Richard is far too modest.
The PSVI does a wonderful job of simultaneously highlighting that we
know a lot about the information we're presenting, and that we've (so
far) done very little to use that information to improve our
representation of that information.
Current PSVI representations are either enormous gouts of XML or
do-it-yourself kits with schemas and documents rather separate. (The
SOAP people have used xsi:type all over in some documents to create
something of an exception that perhaps demonstrates how useful a direct
PSVI representation might be.) Current PSVI APIs - well, we could use
one of those too.
While XML _can_ be used to represent the PSVI, I think it's growing
clearer and clearer that the PSVI could use a firmer base which
explicitly supports and takes advantage of the information in the PSVI.
Ring around the content, a pocket full of brackets
Errors, errors, all fall down!