See also: IRC log
No regrets given
Alex: Essentially, we want to get
it right without describing a particular algorithm
... I took the text from the XML spec and tweaked it a bit
... The algorithm in the proposed Appendix G is based on Henry's proposal.
Norm: I'm happy with the changes to section 5, I'll need to study the appendix more
Alex: The appendix is non-normative, right?
<ht> HST agrees
Richard: About the retreival
URIs, if you did it naively, you might think that you had to go
and fetch the document every time to see if its retreival URI
winds up being the same as one you've already got.
... But in practice, if its literal URI is the same, you can assume its actual URI is the same too.
Alex: Do we have language elsewhere about consistency of resource?
Richard: Even if it did change, there's no gaurantee that you'll see it.
Norm: So we should say that in section 5
Richard: Either that, or some generral statement along the lines that Alex suggested
Norm: I think a general statement is probably the way to go
<MoZ> I think we cannot avoid caching neither
Alex: Maybe we could do a special case for import. Shouldn't we say that however you get a URI, the declarations will be consistent.
<scribe> ACTION: Alex to tweak the text of of p:import [recorded in http://www.w3.org/2008/01/24-xproc-minutes.html#action01]
Norm: I thought we were clear on 95, but yes we should be
<scribe> ACTION: Norm to check the clarity of the spec [recorded in http://www.w3.org/2008/01/24-xproc-minutes.html#action02]
Norm: Static error 27.
Richard: Namespaces don't appear in the attributes property in the infoset
Alex: I think we should allow
users to set xmlns: attributes.
... Namespace fixup will correct user errors.
Richard: I don't think so.
... Namespaces aren't attributes as far as the infoset is concerned.
Alex: I think p:add-attribute could add namespaces.
Richard: Bad. Bad. Bad.
Henry: Here's the question: can you use xsl:attribute to add a namespace decl?
Henry: Right, then add-attribute shouldn't either.
Alex: But you could still do it with a transform
Henry: I think my analysis holds
Richard: The rule should be that you can't manipulate namespaces by manipulating namespace attributes
Henry: The only circumstance you need it for is when you want to create a prefix for a QName that you're going to use in a value
Alex: I think add-attribute needs a modification and set-attributes needs a clarification.
<scribe> ACTION: Alex to consider the attribute steps and make appropriate changes [recorded in http://www.w3.org/2008/01/24-xproc-minutes.html#action03]
Richard: In a sense this is already implicit in the infoset conformance section.
Mohamed: I'm not sure I understand the path. The original question was do the namespace attributes get copied. The answer is 'no'
Alex: Right. We're going to clarify that set-attributes doesn't copy them and clean up the other steps too.
Alex: Event/consequence ordering is determined by the connections.
Norm: What I recall is that if
you mandated document order, then you'd have to make forward
reference an error. But the impl can fix the order, so why make
theuser move teh steps around?
... It seems an unnecessary burden on the user.
Henry: I still don't know what he means by "do things"
Some discussion of what ordering actually means
Richard: I think there are
straightforward cases where it's natural to write in the
pipeline using forward references.
... Consider a straightforward pipeline with an XSLT step in the middle where the stylesheet is generated. It's natural to put the stylesheet generation at the end of the pipeline.
... Is it always possible to write pipelines without forward references?
Norm: I think so.
Richard: Another way to say this is that the XML syntax is just one representation of the pipelien. It could also be represented with a box-and-arrow diagram.
Alex: We're all in agreement,
... Maybe a note?
<scribe> ACTION: Henry to reply and close the issue [recorded in http://www.w3.org/2008/01/24-xproc-minutes.html#action04]
Alex: We have comments that
continue to straggle in.
... What is our plan?
Norm: We're going to do another
last call, as soon as that's ready we'll push it out
... Until then, I don't see any harm in looking at the comments that come in.
rrsange, set logs world-visible