usenet@t... <usenet@t...> wrote
in
<8f2742d7-79d0-4c33-bb39-3dcfb92d3362@e...>:
> On 1 Feb, 07:49, Pavel Lepin <p.le...@ctncorp.com> wrote:
>> mnicke...@gmail.com <mnicke...@gmail.com> wrote:
>> > <Record>
>> > <foo id="a">... </foo>
>> > <foo id="b">....</foo>
>> > <foo id="c">.....</foo>
>> > </Record>
>>
>> > Is it possible to create an xsd file that enforces that
>> > within the <Record> block there exists exactly three
>> > foo tags each containing a different id value (from the
>> > enum a, b, and c )?
>
>> <xs:element name="foo">
>> <xs:complexType>
>> <xs:simpleContent>
>> <xs:extension base="xs:string">
>> <xs:attribute ref="id" use="required"/>
>> </xs:extension>
>> </xs:simpleContent>
>> </xs:complexType>
>> </xs:element>
>> <xs:attribute name="id">
>> <xs:simpleType>
>> <xs:restriction base="xs:string">
>> <xs:pattern value="(a|b|c)"/>
>> </xs:restriction>
>> </xs:simpleType>
>> </xs:attribute>
>> </xs:schema>
> 
> I don't know what other peoples' opinions are, but in
> general I think I would avoid doing a ref to an attribute. 
> My reasoninig is that in XML documents attributes are
> generally unqualified (although there are numerous
> exceptions to this such as xml:lang etc.).  If an XML
> namespace is present, then globally defined attributes 
> are namespace qualified.  Agreed, this schema does not
> define a targetNamespace, so the issue doesn't arise; at
> the moment.

Ah well, I don't really have an opinion. I simply don't
tinker all that much with schemata, so I'm a bit in the
dark as to what the corresponding best practices are. In
this case I defined the attribute separately without even
thinking twice about this (or the consequences).

> However, if as the schema is evolved it's decided to give
> it a target namespace, then the attribute will become
> qualified. This could potentially make it difficult to
> migrate any existing XML instances.

Makes sense to me. However, I believe that all newly
developed schemata should be namespaced from the start. In
that case, the issue would have been immediately obvious
anyway.

> Just my thought - and I'd interested to hear others!

Having thought about it for some five-ten minutes I tend to
agree in general. Defining attributes separately and
referring to them in element definitions should probably be
avoided, unless there's a very good reason for this (like,
some sort of meta-information present on many different
elements: but in that case it makes perfect sense to
namespace the attribute as well, so there).

-- 
<>There is no phenotype</>