> I think we can't really avoid using QNames,
I agree. Given that we need to support enumerations of type QName, it seems
unavoidable that a RELAX NG validator will need to use information about
in-scope namespaces.
> but perhaps we can provide an
> alternative for those who want to avoid them, e.g., something like the ns
> attribute suggested earlier. Yes, all that typing is error prone, but
people
> like to do things their own way. I think we could let them do things both
> ways.
TREX already does this. You can always use an explicit ns attribute instead
of a prefix on a QName (except for a QName enumeration where use of QName is
unavoidable).
James
