RE: Cutting the Patrician datatype knot

From: Peter F. Patel-Schneider ([email protected])
Date: 11/30/01


From: [email protected]
Subject: RE: Cutting the Patrician datatype knot
Date: Fri, 30 Nov 2001 19:19:43 +0200

> 
> 
> > -----Original Message-----
> > From: ext Ashok Malhotra [mailto:[email protected]]
> > Sent: 29 November, 2001 08:25
> > To: Stickler Patrick (NRC/Tampere); [email protected]
> > Cc: [email protected]; [email protected]
> > Subject: RE: Cutting the Patrician datatype knot
> > 
> > 
> > > In the above union "data type", the literal "7" maps to
> > > two members of the value space. Therefore, it is not a
> > > valid data type.
> > 
> > If you define a union type then such anomalies can arise.
> > XML Schema specifies how to resolve them.  If a literal can be 
> > interpreted as a valid literal for more than one of the datatypes
> > included in a union, it is assumed to have the datatype of the first
> > datatype specified in the union for which it is a legal literal.
> > 
> > You may not think this is an elegant solution but it solves the 
> > problem. 
> 
> But doesn't that then preclude the use of (hide) a subset
> of the value space? I.e. whatever the intersection is between
> the two (or more) value spaces, only the members of that
> intersection belonging to the first data type are "accessible".

No, again, see the XML Schema specfication.

In short, you can access the hidden subspace in several ways, including
using xsi:type.

> Well, if that's how the union data type is defined to work, then
> it's not technically a problem -- i.e. there really is no actual
> ambiguity in the mapping -- but it would still IMO be a very 
> odd data type ;-)

Sure, you may consider it to be odd, but it is a valid data type, and it
does cause problems for many of the datatype schemes.

> Cheers,
> 
> Patrick

peter


This archive was generated by hypermail 2.1.4 : 04/02/02 EST