See also: IRC log
No regrets heard.
Paul: Are we still planning to try to address that?
Norm: I think I'd like us to take a look at it
Richard: Is the TAG responsible for overlapping things?
Norm: I think the TAG is looking at some related issues, but they probably hope we'll provide some guidance.
Richard: The description in the charter is pretty vague, perhaps we could get more specifics?
Norm: Yes. Indeed.
<scribe> ACTION: Norm/Henry to attempt to provide a more crisp description of what's needed as a first step towards getting to this work. [recorded in http://www.w3.org/2009/04/09-xproc-minutes.html#action01]
Richard: Use cases would be a good place to start. I've long imagined that one such use case is to answer the question "what does a web browser do with an XML document"
Norm: I'll work this into the agenda more regularly so that we can track our progress.
Norm attempts to summarize.
Norm: I think the answer, whether the prose is clear or not, is that steps can't see their own inputs and outputs. The question of 5.11 is an attempt, I think to address the special case of p:output on a compound step.
Richard: The outputs of a compound step are surely in the same state as the other step children of a subpipeline?
Vojtech: I thought that in 5.11
the phrase "In all cases except the p:output of a compound
step" was redundant.
... When I read it, I went looking to see what was so special, but in fact I think it's covered by the other definitions. It isn't special.
Norm: Fair enough, I'm happy to remove the phrase if it causes more confusion than clarity.
Vojtech: Unless I missed something, I wasn't sure.
Further discussion of 2.5
Norm: In 100, the magic phrase from 2.5 is "the container's contained steps". Compound steps see their siblings, but not themselves.
Vojtech: I have a compound step.
It sees the output ports of its contained steps. Suppose one of
the contained steps is a compound step.
... Now inside that compound step, this step inherites the visibility of the output ports from its parent, which means that it sees it's output port.
Richard: No, it sees the output ports of its siblings, not its parent.
Norm: I think that second bullet
in 2.5 needs to clarify that it doesn't apply to the contained
... To fix issue 100, we need to say "The union of all the declared outputs of all of the containers's contained steps *except this step* are added to the readable ports." But in better English.
Norm: Coming back to 101, I now think that prose is correct. It wouldn't be allowed according to the rules and rather than rewrite the rules to allow it, we're simply stating an exception.
<scribe> ACTION: Norm to fix the rules in 2.5 to satisfy CR #100. CR #101 can be closed without action. [recorded in http://www.w3.org/2009/04/09-xproc-minutes.html#action02]
Vojtech: If you have the validate with XML Schema step and you pass multiple schemas, what does that mean.
Norm attempts to describe the schema validation rules of XSD.
Vojtech: And what about xs:include and xs:import.
Norm: I propose we wait for Henry's input.
Norm: It boils down to whether the 3rd or 4th bullet in 7.2.9 applies. I don't think it much matters.
Vojtech: I think we should say that it's c:data without a content-type or with a content-type that specifies a text content type...something like that.
Norm: That works for me.
... Proposed: make the change that Vojtech outlines.
Norm: My proposal is that "you lose" if you get mixtures of slashes and you turn on fix-slashes.
Vojtech: So you lose if you need a mixture fo forward and backward slashes in the filename
Norm: Only if you turn on the
... It seems like we have two choices, leave it as is or invent a new escaping mechanism.
Richard: We could use a private use character or allow the fix-slashes option to specify which character to replace with the platform-specific slash.
Norm: We've made pretty significant changes to p:exec already.
Richard: I think I'd say that no-translation is applied unless you specify the fixup and then that fixup is applied everywhere.
Vojtech: It sounds good to me.
Norm: How about I write up a proposal that does this and we see if we like it.
<scribe> ACTION: Norm to write this up as a proposal. [recorded in http://www.w3.org/2009/04/09-xproc-minutes.html#action03]
Vojtech: What do multiple source documents mean? Is it only to allow no documents?
Norm: Yes, I think it probably was.
Vojtech: So what happens if you pass two, is it an error?
Norm: I think we should either
say that its an error or say that its
... Do you have any command-line tools that accept a sequence of documents on stdin?
Richard: No, I don't think so.
Norm: I propose we make it an error in V1 to pass a sequence of more than one document.