See also: IRC log
Henry gives regrets for 8 Oct
Vojtech: I wonder if err:XS0005 is clear in the case of compound steps.
Norm: Unless there's a specific suggestion for how to improve the description of err:XS0005, I'm inclined to leave it.
No dissent heard.
Norm: I propose that that closes issue 163
Norm explains what he did wrt replacing "binding" with "connection".
Vojtech: I think it's an improvement.
Henry: I'm in favor of the change
without having looked at it.
... It makes sense and I trust you to have gotten it nearly right, so I'm not fussed.
Norm: Is there anyone who wants more time?
Norm: I propose that we adopt the alternate wording as the new status quo.
Vojtech: With this change, I think we should update the schemas and DTDs.
Henry: I'll fix the schemas
Norm: I'll fix the RELAX NG.
Henry: How likely is it that
people are going to try to do this?
... If it's something that's going to happen a lot is going to be compromised to allow something that isn't going to happen a lot then I'm opposed.
... It seems to me that most of the benefit of what you just said would be to simply change the error from static to dynamic.
Norm: The wrinkle is you only get one declaration.
Vojtech: How would this influence the V1 pipelines. Suppose I have a V1 step and I simply use p:identity and I intentionally refer to an output port called foo.
Norm: That would still be an error.
Vojtech: If we decide to introduce some changes, this is something we can test.
Alex: Your saying it would be a
dynamic error to evaluate a step that contains unknown children
... But your saying unknown inputs/outputs have a well-defined behavior.
Norm: Explains his rationale again, about the fact that there's only one declaratin.
Alex: That's a value judgement about how important the output is. In that case, it would be nice to know that the output was non-critical.
Henry: Add into the mix the fact
that we have a story about what you're allowed to do without
changing the name of the step.
... I think Alex's observation is correct that someone may refactor the structure of a step so that the fundamental output is split across two ports.
... The correct answer there is that we have to extend our notion of what amounts to a category of backwards compatible changes.
... At the moment it only contains optional options, but we could extend it to include collateral inputs and outputs.
... So if you're going to make a non-benign change, you must use a new step name.
Norm: Yes, I think that would help.
Henry: I think it's the usual story, Norm you should write up an editor's draft for us to review.
Norm: I'm tempted to delay publication of our interim CR draft by at least a week so that we can resolve this.
Alex: I'd prefer to resolve this first.
<alexmilowski> (sorry, I have to go. I have an appointment at 9am.)
Vojtech: There's the related question of what is the default version.
Norm: I think if you don't import a library, you get what the processor does.
More discussion about what to do with input/output ports
Norm: I'm happy to make it a dynamic error to attempt to read from the port in the V1 case, it just looks like more work.
<scribe> ACTION: Norm to promote the alternate text draft to the status quo. [recorded in http://www.w3.org/2009/10/08-xproc-minutes.html#action01]
<scribe> ACTION: Norm to produce an alternate draft with the proposed versioning changes. [recorded in http://www.w3.org/2009/10/08-xproc-minutes.html#action02]
Norm: I will be going to
... Welcome back Vojtech!