Re: [Bug 3062] Need to write examples in the spec as valid XML

Sebastian Rahtz wrote:
> 
> Felix Sasaki wrote:
> 
>> that's true as well. I think we have two choices:
>>
>> - 1 Rely on the prose description, which says (or should say) what is
>> mandatory. It says even more than a schema, e.g. that the value of
>> its:selector is an XPath expression.
> 
> fair point.
> 
> if we have prose which is normative, you might as well
> go the whole way and write BNF productions by hand (as I
> assume XSLT does), and the schema(s) can be part of implementation
> testing.
> 
>> - 2 Rely on the ODD definitions, and say that these are normative. (no
>> matter if they are written as "production rules" or visualized in a
>> style like http://www.tei-c.org/release/doc/tei-p5-doc/html/TD.html ).
> 
> in that case parts of the schemas *do* become normative.

mm ... can't we say that the names of elements and attributes, their
occurrence constraints and - as for attributes - are normative?

> 
> I am sorry to have raised this so late. I had been assuming
> until relatively recently that the schemas were normative
> and I hadnt thought through the implications of them not being so.

How about a survey in the group - hands up:

1 Who would use the schemas "as is"?
2 Who would create his own ITS schema and use it every time he
integrates ITS into a new or existing schema?
3 Who would create a new ITS schema each time he combines ITS with a new
or existing schema?

Looking at the wide range of schema designs, I thought that 3 would be
likely, but I'm not sure anymore ...

- Felix