- From: <paul.downey@bt.com>
- Date: Fri, 21 Apr 2006 15:43:19 +0100
- To: <petexmldev@tech-know-ware.com>, <edday@obj-sys.com>, <Paul.V.Biron@kp.org>
- Cc: <gcowe@origoservices.com>, <public-xsd-databinding@w3.org>, <public-xsd-databinding-request@w3.org>
> I agree with Ed and Paul here. Support for xs:choice seems fundamental. > You could offer some advice along the lines of "You may wish to check that > your tool supports xs:choice. If it does not, you may wish to choose > another tool!" That's OK for someone taking a schema to build a client or a service who has control over the toolkit employed, but not at all helpful for the author of a schema or WSDL wanting to reach a wide marketplace of tools selected by potential customers. > However, I wouldn't advice against using xs:choice in schemas. I propose we consider: 0) no pattern for xs:choice 1) include 'choice' as a Basic pattern with a design consideration - we can treat desgin considerations as "warnings" rather than errors. 2) offer an a alternative pattern for choice: <choice> <element name="optionA" ../> <element name="optionB" ../> <element name="optionC" ../> </choice> using: <sequence> <element name="optionA" minOccurs="0" ../> <element name="optionB" minOccurs="0" ../> <element name="optionC" minOccurs="0" ../> </sequence> Some empirical testing with toolkits might help here, If it is only one old version of a toolkit which bails/ignores choice, then we can provide a pattern for choice without (1), otherwise I'd suggest a combination of (1) and (2), or (2) alone. Paul
Received on Friday, 21 April 2006 14:43:28 UTC