I am with Fred in questioning whether QName is the right
construct here anyway. If we don't intend to use the muws:category mechanism
then let's use URIs. And if we don't intend to let people define their own
values then forget URIs and let's just make this an enumeration of possible
strings.
William
I'm just saying that putting QNames which are not defined
by schema or WSDL is not a good practice. Do a URI, string or define an XML
element in the schema with the QName you need.
So, you'll get
<muws-xs:OperationalStatus><muws-xs:Available/></muws-xs:OperationalStatus>
instead
of
<muws-xs:OperationalStatus>muws-xs:Available</muws-xs:OperationalStatus>
This
is just proper XML practices. I think this is very
simple...
-- Igor
Sedukhin ..
(igor.sedukhin@ca.com)
-- (631) 342-4325 ..
1 CA Plaza, Islandia, NY 11749
Sedukhin, Igor S wrote:
It should
be
<OperationalStatus>{any}</OperationalStatus>
instead of <OperationalStatus>xs:QName</OperationalStatus>
QNames cannot just hang out
there. They have to be declared in schema, WSDL, policy or some other
document. Declaring them merely in the text is unheard of so far. I suggest
we don't do it.
Therefore there should
be XML element declarations which QNames correspond to the
intended status values. e.g.
element with muws-xs:Available
QName corresponds to available status.
-- Igor Sedukhin ..
(igor.sedukhin@ca.com)
-- (631) 342-4325 ..
1 CA Plaza, Islandia, NY 11749
I guess I'm missing something. I
thought the issue was [primarily] that if QNames, properly schema-defined or
not, appear as text, then XML rewriters may muck'em up by changing namespace
prefixes, etc. I don't quite see how their proper schema def'n solves
that particularly, unless we assume that all xml rewriters are schema
savvy.
Since the type is properly defined, it seems legal. It
does, though, for the reasons alluded to above, seem to be a less-than-good
way to do it. A URI would work, as would a new type definition, a
restriction on string, which enumerated the possibilities ("Available",
"Partiallyavailable", "Unavailable" and "Unknown").
Failing that,
'twould seem that making it an element carrying information content would be
the most reasonable way to go (as we do with categories).
I'm with
William in not quite understanding what "declaring them merely in text is
unheard of" means. Surely they need to be defined if they are
QNames. But having QNames in text, albeit now recognized as less than
the best of ideas, does happen. Isn't that used for the "extends"
attribute (list of QNames) in WSDL 2 (or did it
change)?