Cart
[Home] [By Thread] [By Date] [Recent Entries]
Andrew Welch wrote:
>> So in order to get your constraint you have three choices:
>>
>> 1) Adopt RELAX NG instead of XSD.
>> 2) Adopt Schematron as well as (or instead of XSD)
>> 3) Adopt XSD 1.1 (if you can find an implementation) and use their
>> assertions.
>>
>
> there's a 4th I think:
>
> 4) annotate the category elements with an xsi:type:
>
> <category xsi:type="privacy" scheme='http://dig.com/privacy' term='private'/>
> <category xsi:type="fishingType" scheme='http://dig.com/fishingType'
> term='coarse'/>
>
> ...then just define the types "privacy" and "fishingType" in your
> schema.
I suspect that will not work, but it certainly may be worth a try.
To make this work, don't you have to define the <entry> element to use
wildcards? For example (loosely)
<element name="entry" >
<sequence>
<element name="category"> <!-- open category -->
<xs:any namespace="##targetNamespace"/>
And if you wanted to make the first optional, alarm bells go off that many XSD processors
are likely to disallow it (due to quasi UPA implementation, regardless of the standard).
So you lose the ability to say "the second element must be a category" while gaining the
ability to override the schema in the instance. The trouble with this use of xsi:type is
that it really doesn't give you anything in terms of stricter validation or information
modeling: it lets you get a validation result in spite of the modeling!
Cheers
Rick Jelliffe
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] |
