See also: IRC log
Henry gives regrets.
Norm summarizes his recent changes wrt xpath-version
No other comments.
Norm summarizes, he'd like to forbid them all.
Mohamed pushes back and suggests that we need generate-id() especially for XPath 1.0 implementatios.
Norm: I'd prefer the simplicity of forbidding them all and come back to it if someone demonstrates a real need for it.
Mohamed: That's ok for me.
Proposal: XProc processors are not required to implement any functions defined only in XSLT.
<scribe> ACTION: Norm to add information about which functions must be implemented and note that XSLT functions do not have to be implemented. [recorded in http://www.w3.org/2008/07/03-xproc-minutes.html#action01]
Henry: It's very clear that the winged hourse allows you to find the type hierarchy. It doesn't let you do things with substitution groups, for example.
Norm: So we couldn't do element construction (which we don't), substitution groups, or the validate expression.
Henry: We could put schema import in, identify it as a feature at risk, and then we could remove it. If we leave it out and ask for feedback, we could be pushed back to Last Call again.
Norm: Yes. I suppose the best thing we can do is add the feature and make it optional. So basic processors can skip it. Schema-aware processors will be able to use types, etc. And identify it as a feature that might get removed if there's strong opposition.
Henry: Seems like a good compromise to me.
Norm: I think we should forbid references to non-builtin schema types if it doesn't support schema import.
Henry: I disagree. I just want to grab an XPath 2 library and use the winged horse.
Norm: In XSLT, you can't refer to a hatsize unless you've imported the schema that defines hatsize.
Mohamed: Where do we use this element?
Norm: At the top of your pipeline.
Mohamed: In XSLT, it's clear that
when we manipulate a sequence of documents, they are related.
In XProc, that's not the case.
... Importing a schema for all the pipeline, even if there will be conflicts.
Norm is confused.
Mohamed: Imagine I have a pipeline that's supposed to handle XHTML 1, XHTML 2, and XHTML 3. They're all in the same namespace, but they have different definitions.
Norm: How will you check?
Mohamed: By looking for an element, html:h, for example
Norm: So you don't need a schema for that
Mohamed: But in each branch, I
might want to be able to do tests with the right types.
... We could add schema-import everywhere that p:namespaces is allowed. Importing may not be sufficient if we want to handle all the cases.
... But if we just want to handle some of the cases, and we want to support substitution groups, then it's probably ok to just put it at the top.
Proposal: We'll add
p:schema-import, allowed at the top of a pipeline, to provide
schema definitions. This is an optional feature.
... We'll note that it may not be sufficient to handle all the possible cases where a pipeline needs to check the types of elements that come from documents using different versions of the namespace.
... We'll leave open for the moment the question of whether or not it's legal to refer to t:hatsize without having imported the schema that defines it.
<ht> I've sent another email -- Norm is right
So scratch the last bit. It is an error if you write an xpath who's type is not in the static context.
<scribe> ACTION: Norm to implement p:schema-import. [recorded in http://www.w3.org/2008/07/03-xproc-minutes.html#action02]
Mohamed: I think we're opening a
new box that may have a lot of side-effects on
... What happens to the context passed to the steps?
Henry: If it's not the static context for xpath evaluation that's effected by schema import, what is?
Norm: It is the static context. But what about the static context for steps.
Mohamed: I think this won't help if we want to go to last call quickly.
Henry: We can add this and make
it an optional feature. We can also mark it as a feature that
is at risk.
... That way we can proceed from last call without it.
... So if no one says, I require this, then we can get rid of it without going back.
<scribe> ACTION: Editor to write it up and see what it looks like. [recorded in http://www.w3.org/2008/07/03-xproc-minutes.html#action03]
Vojtech: It may be difficult to specify the behavior when importing other subpipelines.
Mohamed: Visibility will definitely be an issue.
Henry: The analogy with XSLT is our friend. Whatever they do, we should do.
Out of time.