RE: XPath Data Model proposal

Jeni already brought most of the reasons why I don't think your model
would work and actually provide a type safe system that would make the
XPath data model independent of the validation process.

Jonathan already presented you with a way to get most of what you want
using the current framework.

Let me address some technical issues below and potentially in some other
mails (I have not read all of them yet) that you may have missed.

Best regards
Michael

> -----Original Message-----
> From: public-qt-comments-request@w3.org [mailto:public-qt-comments-
> request@w3.org] On Behalf Of John Cowan
> Sent: Thursday, June 05, 2003 9:53 AM
> To: Jeni Tennison
> Cc: Norman Walsh; Jonathan Robie; Mike Champion;
public-qt-comments@w3.org
> Subject: Re: XPath Data Model proposal
> 
> 
> Jeni Tennison scripsit:
> 
> > Just to fill in the gaps here, presumably you're only suggesting
that
> > an XPath 2.0 processor be able to do anything with xsi:type
attributes
> > holding the names of types that it recognises, and that other type
> > annotations should be ignored?
> 
> In effect, I think that is achieved by the penultimate paragraph in
> 3.6, which says that unknown types are to be promoted to xs:anyType or
> xs:anySimpleType, as the case may be.

[Michael Rys] Promoting unknown types to xs:anyType and xs:anySimpleType
does not work if you want to be able to have different semantics for
typed data (strong typing) and untyped data (weak typing) and want to
make a distinction between the most specific type being untyped and the
most specific statically given type being the most general type.

xs:anyType/xs:anySimpleTypes are very general types that serve as the
base types of the type hierarchy. Saying that untyped data becomes the
most general type is not the right approach (although if you just do
dynamic typing you may get away with it). 

You need to distinguish between something that you know to be untyped
and something that you do not quite now what type it is during type
inference (mostly if you perform static type analysis and static
typing). 

Since XML Schema unfortunately mixes these two concepts (xs:anyType in
the schema definition language indicates the most general type, and on
the PSVI instance annotation the untyped case), XQuery/XPath needed to
split them. And since it is easier to change the type given in the data
model than the type name used in the schema definition language, XQuery
introduced xdt:untypedAtomic (and will also need a companion for untyped
Complex/Any types).

Thus, any XPath/XQuery type system that is using XML Schema based type
names should NOT use xs:anySimpleType/xs:anyType to denote untyped data.

[Note: there would be a slightly more complex alternative, using the
notion of the most strict type that then would have the notion of
"strict xs:anySimpleType" being the same as xdt:untypedAtomic]
 
> > Also, I assume that you're suggesting that the XPath 2.0 processor
> > validate the content of the annotated element, such that, for
example
> > <foo xsi:type="xs:integer">rubbish</foo> isn't labelled as being of
> > type xs:integer. Or are you suggesting that elements be labelled
with
> > types without necessarily being valid against those types?
> 
> Hmm, that's a point I hadn't considered.  It's formally messy, because
> there is no formal definition of type validity in the DM -- it can
> delegate to the XML Schema documentation for that.  OTOH, asking an
XPath
> processor to cope with examples like that is more than a bit awkward.
> 
> > (The same answers will apply to xdt:attributeTypes, I imagine.)
> 
> Indeed.

[Michael Rys] And this is one of the main reasons why the current system
of requiring some form of validation equivalent before populating the
data model is important.

That the PSVI and data model is preserving xsi:nil/xsi:type as
attributes in addition to adding information about their values to the
element/attribute nodes is already bad enough. But having to recognize
them even if you do not provide a validation step before is bad. And
requiring the data model now to perform its own validation flies into
the face of composable specs that so many people want from the W3C.


Best regards
Michael

> --
> John Cowan  jcowan@reutershealth.com  www.reutershealth.com
> www.ccil.org/~cowan
> I am he that buries his friends alive and drowns them and draws them
> alive again from the water. I came from the end of a bag, but no bag
> went over me.  I am the friend of bears and the guest of eagles. I am
> Ringwinner and Luckwearer; and I am Barrel-rider.  --Bilbo to Smaug
>

Received on Thursday, 5 June 2003 15:41:30 UTC