> James Clark wrote:
> >
> >If the spec is going to say "Implementations may issue error messages for
> >duplicate attributes", then the spec has to define exactly what
constitutes
> >a duplicate attribute. How should it do this?
>
> I have GNF normalization in mind. But I am not sure if we should put
> that in the spec.
GNF normalization seems like a complicated way to explain duplicate
attributes (especially since it can be very expensive to implement). I'm
still not clear what constitutes a duplicate attribute in GNF when you take
name classes into account.
> >I am also concerned that this will create interoperability problems (like
> >the ambiguity constraint in XML). Either it should be an error and
> >implementations should be required to detect it, or it should not be an
> >error and implementations should be required to accept it.
>
> I don't think so. Implementations may issue warning messages, but they
> are required to continue. Thus, the only difference is warning messages.
> Different XML parsers already show different warning messages. What is
> wrong about that?
I wrote that before I saw your correction that changed from proposing error
messages to warning messages.
James
