I think you've got it right, but I can't explain the reasoning.

It seems reasonable that derivation by extension results 
in unioning of the wildcard constraints, but I don't know
why attribute groups have their wildcard attributes
unioned (except for processContents, in which the
first attribute wildcard wins).  I can only guess that
the WG realized the wildcard component was insufficient to
represent the "proper" combination of the wildcards and
decided to make a somewhat arbitrary approximation.

xan

-----Original Message-----
From: xmlschema-dev-request@w... [mailto:xmlschema-dev-request@w...] =
On Behalf Of Michael Kay
Sent: Thursday, April 15, 2004 2:16 PM
To: xmlschema-dev@w...
Subject: Attribute Wildcards (another try)


I submitted the following question a couple of weeks ago and got no
response. Perhaps the people who know were on holiday, so I'll try =
again...

Normally, if a complex type references two attribute groups, then the
attributes permitted on an element are the union of those permitted by =
the
two attribute groups.

But if the two attribute groups both contain attribute wildcards, the
attributes permitted  are the intersection of what's allowed by the two
wildcards (i.e. an attribute has to match both).

However, if one complex type containing an attribute wildcard extends
another complex type also containing an attribute wildcard, then the
attributes permitted are the union of what's allowed by the two =
wildcards.

Have I got it right?

And if so, can someone please explain the reasoning?

Michael Kay


From janet@w...  Fri Apr 16 10:49:01 2004
Return-Path: <janet@w...>
X-O