See also: IRC log
Rui gives regrets for 4 Sep
Norm gives regrets for 11 Sep; Henry will chair if there's a call.
Norm: Henry is working to get our charter extended through December to align with the other XML WGs.
Norm: I think there's only one
item, making the step naming rules non-normative.
... I think we made them normative on purpose and we're unlikely to want to change them.
Henry: I think that's the right thing to do. It gives us a basis for describing steps in URIs if anyone ever wants to.
Proposal: Make no change.
Norm: Also from Jim, a request
for namespace aliasing instead of p:namespace-rename.
... I don't want to make this change.
Alex: He's asking for something that's related, but not the same as namespace-rename
Henry: The only possible virtue I
can see is that it might save you some typing if you happened
to have the prefixes in scope.
... I don't think his proposal is substantially different, so why make the change?
Norm: Ok, I'll see if that satisfies him.
Norm: This is about RDF and there are some suggested step types here.
Henry: I think we should respond as positively as possible and suggest that absolutely, the Semantic Web Deployment WG should publish a pipeline library with these steps defined as soon as possible.
Norm: I agree, I think the best thing we can do is suggest that the folks with the relevant expertise publish the library.
Henry: They'll get all the interop they need by doing it that way, using a W3C URI for the step namespace.
Norm: I think we might get some pushback that they would like this to be in p: and in V1, but I'm not sure we can practically satisfy that request.
Proposal: Suggest that this is something that folks with the relevant experience undertake asap, but that we won't plan to include it in p: in V1.
Norm: This is from Vojtech, who
gave regrets, but I think I can describe it.
... Vojtech observes that p:load requires support for http: and file: URIs and wonders if p:data should do the same.
Mohamed: There are now three locations where we need to be clear on what schemes are supported, p:load, p:http-request, and p:data
Norm: And p:document, as it
... It seems odd to call out http: and file:, but maybe it's useful for interoperability
Alex: I think it's useful.
Norm: Does it make sense to rephrase this as general note along the lines: implentations are encouraged to support all practical schems for loading resources, and in particular they SHOULD support file: and http(s): URIs.
Mohamed: That's fine for me.
Richard: I can imagine an implementation that doesn't have any files
Norm: Sure, but then that's a good reason not to obey the SHOULD.
Proposal: Make a general note as
... and adjust the spec accordingly.
Norm: I'm inclined to agree, but I'm not sure what the available APIs actually provide.
Henry: Well, that's critically important. I'm less inclined to do this if there's no way to implement it.
Norm: Ok, I'll investigate.
<scribe> ACTION: Norm to investigate RELAX NG APIs for switches to control DTD Compatibility conformance. [recorded in http://www.w3.org/2008/08/28-xproc-minutes.html#action01]
Norm: I just wanted to confirm
that the WG agrees with me that an XSLT 2.0 step is free to
produce PSVI-rich results even if @psvi-required appears no
... Even if you said psvi-required=false, a step can still produce PSVI-rich infosets.
... Extra PSVI stuff shouldn't cause any problems.
Henry: Seems right. Do we need to change anything?
Norm: Maybe just add some editorial clarity.
Mohamed: We can put psvi-required on declaration, library, and pipeline. But not on individual steps.
Norm: Interesting, that's true,
these are declaration-level attributes, not instance
... But maybe that's ok, there's no value in having a single step in the middle of a pipeline require PSVI support. I don't think.
... So I guess the question remains, if you say psvi-required=false, is it an error to produce a PSVI?
Henry: I don't think so, in fact a PSVI capable processor might just ignore psvi-required.
Proposal: Add some editorial clarification.
Norm: I think the salient point is that we should say that how options and parameters are bound outside the pipeline is implementation-defined.
Proposal: Say so.
Mohamed: The example uses p:namespaces where it isn't needed.
Norm: Let's let the editor reconsider the example
Mohamed: I think the example predates the default rule, and that's the problem.
<scribe> ACTION: Norm to reconsider and fix this example [recorded in http://www.w3.org/2008/08/28-xproc-minutes.html#action02]
Mohamed: Vojtech observes that it's not clear when we use the short format for the option.
Norm: If you use the short form then you can't use p:namespaces, so you better not need it.
Norm: This encrypt/decrypt and
I'm trying to work with the Security WG to resolve this
... We do have a use case in our requirements document that requires it.
Mohamed: That's why I've proposed to not normalize too much, maybe we should just say that's what parameters are for, to say implementation-defined.
Norm: Yeah, but having a standard
step with no interoperability is a little funky.
... If the Security WG can help us get these crisp, then we'll prbably need to include them, otherwise, we'll have to do something else, maybe what we're suggesting for the RDF steps.
Henry: Maybe at TPAC08 we can hand this off to Frederick if we promise to help.
Norm: That's not a bad idea
... I think we should just leave this open a little longer.
Mohamed: I found two or three things that weren't in sync. For example, for p:insert, I proposed to accept different kinds of nodes depending on the position option.
Norm: I see what you mean, but there are no constraints on p:insert so we can just let the invalid document catch it, right?
Mohamed: What about matching a PI before the document element
Norm: Oh, right. I see. I guess that should really be anything except attribute nodes.
Proposal: Fix p:insert
Mohamed: And p:replace should be the same.
Richard: It should only allow you to match things that can appear on the child axis.
Norm: But matching a document node and inserting a comment as the first-child is a reasonable thing, right?
Richard: It doesn't make any sense to insert something before or after the *document node*
Norm: I'll have to expand the prose a bit.
<scribe> ACTION: Norm to take another crack at getting this right for all the steps. [recorded in http://www.w3.org/2008/08/28-xproc-minutes.html#action03]
<ht> I like Richard's formulation: p:insert allows matching of anything which may have children if where is first or last, and anything that _is_ a child if where is before or after
Norm: We need a normative reference for UUIDs, anyone know what it is?
Alex: There's an ISO
... It's in the references for the RFC 4122
Norm: Ok, I just looked right past that when I scanned the references.
Proposal: Add the reference
Henry/Alex: Maybe we can refer to both, because the RFC is easier to read?
<MoZ>  ISO/IEC 9834-8:2004 Information Technology, "Procedures for the
<MoZ> operation of OSI Registration Authorities: Generation and
<MoZ> registration of Universally Unique Identifiers (UUIDs) and their
<MoZ> use as ASN.1 Object Identifier components" ITU-T Rec. X.667,
Norm: I don't mind having the RFC in the non-normative references.
Proposal: Say they're hex
Henry: We've been glibly
addressing these issues without classify the changes. I think
the way we've proceeded has been perfectly fine.
... At some point we'll have to decide.
Norm: I think we've been making clarifications and small changes.
Henry: Strictly speaking, anything that invalidates a review is out of bounds.
Norm: I don't want to make any of
... If I think anything I change in the spec isn't a clarification, I'll bring it back to the WG.
Mohamed: Probably you should make an announcement that there's an implementation.
Norm: Okie dokie.
Henry: Have we invited review of the spec?
<scribe> ACTION: Norm to make sure we've solicited review from the relevant WGs [recorded in http://www.w3.org/2008/08/28-xproc-minutes.html#action04]