Re: complexType : extension of a sequence by a choice ?

Henry Thompson writes:

> I'm still a little nervous about a conditional
> invariant -- that is, on your proposal the V1
> application would have to be _much_ smarter -- instead
> of depending on one invariant across the board, without
> appeal to type information, it would have to know in
> some detail what the types of the nodes it was
> processing were.

True, although I would have thought that any application doing anything 
with this sort of polymorphism would need to be built with practial 
knowledge of the type.  Whether it gets that knowledge formally from the 
schema or from some other spec isn't the point.  If nothing else, it needs 
to know what we capture as the value of block={extension}, and if 
extension is allowed, it should in general know the content model of the 
prefix covered by the base type.  Once you know that, is it such a leap to 
say: "well actually, the content model is <xsd:all> and doesn't have a 
prefix defined in the type at all...instead, it sets a lower bound on the 
elements that must appear, regardless of order."  I'm not convinced it's 
as big a leap as what you're implying. 

Anyway, I think this has been a useful discussion.  Let's see whether 
there is a groundswell of interest from users claiming that extended "all" 
makes an 80/20 cut.  Then we can worry about whether it would unduly 
complicate applications.  Thanks!

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------

Received on Friday, 13 February 2004 10:40:48 UTC