See also: IRC log
Norm notes that he added a link for 2.1 and added a 2.6
We have continuing regrets from Richard and Henry. Also Rui for next week.
Norm summarizes the proposal.
Norm suggests that we no longer allow the context to be empty.
Some discussion of what to do in the case where there's no primary input port to default to and the expression is known not to need a context.
<MoZ> NOrm, what are you proposing for evaluating "$foo='2'"
Alessandro suggests binding to an empty document node.
Norm: I suggest we direct the editor to give that a try.
moz, do you want to call in to talk about p:map?
We'll skip 2.3 today until Mohamed can be present.
Alex: Do we have a parking lot for future things?
Norm: Not at the moment.
... I suggest we setup a wiki page for that, at e.g., wiki.xproc.org.
<MoZ> +1 for wiki parking page
Norm wonders if the supporters are content for V2.
They appear to be.
We'll reconsider this for V2, no change for V1.
<alexmilowski> + 1 to new-name
Alex: There's authentication, encoding stuff, and serialization options. Is anything other than serialization controversial?
Alex: p:http-request and p:store
... XSLT2/XQuery1 Serialization describes how to control serialization.
... many are similar to XSLT 1
s/step we/spec we/
Norm: Do you think the complex serialization stuff is needed on p:http-request?
Alex: I have had to control XML serialization; given that, I can see how I might need to.
Norm: Don't I also then need it on escape-markup?
Alex: There are a few things that don't apply, but maybe.
Norm: I worry about the implementation burden, but there is a stand-alone spec we can point to.
Alex: We could say that only XML was required.
Norm: I guess that would be ok.
Norm decides not to persue the "p:store" vs. "p:serialize" steps.
Alessandro: A slightly different
look at it is that, at least in my mind, implementors will want
to reuse existing code.
... If they have an XSLT 2 engine, they'll be OK.
... For implementors with XSLT 1 engines, what are they going to do? Can we limit the required set to just what's supported in XSLT 1?
Alex: I'd have to reread the serialization spec to find out.
Norm: Alex, can you look at that?
<scribe> ACTION: Alex to review the serialization spec to see if we can define a set of parameter/value pairs that we require that happens to be supported by XSLT 1 and XSLT 2. [recorded in http://www.w3.org/2007/07/12-xproc-minutes.html#action01]
Norm: Have we accepted these
ideas in principle?
... I guess so.
Alex: I want to be able to reuse
options on several steps.
... I propose a pipeline-level p:serialization element (like xsl:output) to handle this case.
... Once you have a named p:serialization element, you can refer to it from the p:outputs where you want it.
Norm: This is only for p:output elements on the p:pipeline, right?
Norm: Maybe we could turn it around and make it like p:log where p:serialization points to the port it wants to be applied to.
Any further discussion?
Norm: What about character maps?
Alex: I left it out.
... That requires even more declarations in our pipeline.
Norm: I bet it'll come back, but
I'm fine with leaving them out now.
... Please add an editorial note about it.
Norm: I saw the proposal for http-get, did you also change the http-request?
Alex: I only added the
serialization and authentication options and some words about
... The two issues I see are (1) is there some simplification for simple GETs and (2) should this be required?
Norm: I thought we agreed to make it required.
Norm: I have real reservations about p:http-get.
Alex: I'm not sure I follow.
Norm attempts to summarize Henry's desire for a simple http-request option that just returns the response directly with no c:http-response/c:body wrappers.
Alex: If we want an href option then I think we should look at expanding what "p:load" does.
Scribe fails to capture some discussion
Norm: I guess we're splitting
Henry's request in half: a simple result vs. a simple
... I hear Alex in favor an attribute on c:http-request for providing a simple response, but not on anything for providing a simple request.
Norm: So what distinguish a simple get from load is the authorization stuff.
Alex: And http-request can handle non-XML stuff.
Norm: Ok, I think I'm happy with that. If you just want to do a simple get, use p:load; if you want authorization or fancy encodings, get out the big guns.
Alex: That sounds good to me.
Proposal: An attribute for a simple response from p:http-request, but http-request does the whole deal.
Alex: And we clarify that p:load is expected to be able to do HTTP...
Norm: I want to know what stands
between us and last call.
... Please be prepared to enumerate the issues you know of next week.