W3C Forms teleconference September 3, 2008

* Present

Charlie Wiecha, IBM
John Boyer, IBM (chair)
Mark Birbeck, x-port.net
Paul Butcher, x-port.net
Steven Pemberton, CWI/W3C (left early)
Keith Wells, IBM
Leigh Klotz, Xerox (minutes)
Nick van den Bleeken, Inventive Designers
Erik Bruchez, Orbeon (joined late)

* Agenda


* Previous Minutes

* Administration

John Boyer: Steven will be chairing the next two weeks. Maybe it's time for blue-sky sessions and reducing call times.
Steven Pemberton: I've always been fine with 60-minute calls.
Keith Wells: If we don't have any thing to talk about we could finish early.
Steven Pemberton: Sure.
John Boyer: That sounds like a good idea.

* Rich Web Application Backplane

Charlie Wiecha: Take a look at the streamlined syntax in Ubiquity. There's no repeat or XPath syntax yet. There's some connection between backplane and an early proposal multi-modal using XHTML+Voice using XML Events. Before streamlined syntax, there was no implied data model though. So we may be able to revisit XHTML+Voice and use the implied model and go back to MM at the Tech Plenary. I'll put it up on the wiki.

* News item from Steven

* XML Schema 1.1 review

John Boyer: It would be good to get this.
Steven Pemberton: Hopefully Kenneth should get it to us by Wednesday.
John Boyer: I'll send him mail.

* Coordinating with HTML WG chairs about Forms Joint Task Force

John Boyer: I'm coordinating for Tech Plenary and have some responses.

* Sentence about XHTML Modularization

Steven Pemberton: I sent it today.

* 1.0 Basic impl report

Leigh Klotz: I haven't started this yet. I've been working on submission headers.

* Are xforms-insert and xforms-delete notification events?


John Boyer: In modularization we're separating this out, and in repeat it says that certain behaviors behave as if they were in response to xforms-insert and xforms-delete. So technically it's a notification event as there is no default handler, but it's not a pure notification event from the author's standpoint any more. Should we attach a note?
Steven Pemberton: Can we also say that the XForms processor is also notified?
John Boyer: We don't rigorously define "notification event." Erik has interpreted it to mean author event only, but if you look closely at http://www.w3.org/TR/xforms11/#evt-insert it is what comes after default action. So we could remove "notification event" and say "default action: none."
Steven Pemberton: If we have no normative meaning for notification event, then that's fine.
John Boyer: Perhaps we can just delete them throughout the spec and just leave "default action."
Leigh Klotz: So it won't have a default action but repeat will listen for it?
John Boyer: Yes. http://www.w3.org/TR/xforms11/#ui-repeat-processing
Steven Pemberton: [leaves]
John Boyer: The repeat module uses the event context information to find out if the inserted or deleted nodes have anything to do with it, and if so, update. This gives a more precise definition of when the repeat updates happen.
John Boyer: So the action is for me to remove "notification event only" completely from the XForms 1.1 spec. Is everybody happy with that?

Resolution 2008-09-3.1: We remove "notification event only" from XForms 1.1 as it is not normatively defined and is causing confusion.

Action 2008-09-3.1: John Boyer remove "notification event only" from XForms 1.1 as it is not normatively defined and is causing confusion.

* Submission header problem

Leigh Klotz: I'm trying to get agreement with Philip Fennel and Aaron Reed. There are two new things: combine looks valueable, though error or exception may be not be (and what error) and then we need to allow multiple value elements to allow the submission method to combine the values.
John Boyer: xforms-submit-error
Leigh Klotz: Does this happen at submission time or serialization time?
John Boyer: It could be validation, prior to serialization. We have an xforms-submit-serialize event but I don't know if we allow that to quit or indicate errors.
John Boyer: The values could be done with concat.
Leigh Klotz: Then you need to know to quote and use commas, which is really HTTP's responsibility, unless we pull back and say all headers are MIME headers and ask teh user to do the quoting and ordering and so on.
John Boyer: http://www.w3.org/TR/xforms11/#submit-evt-submit We calculate submission headers after validation. See number 5. There might be an xforms-submit-error there.
Leigh Klotz: We'd need an error type.
John Boyer: Yes.
Leigh Klotz: I'll propose one.
John Boyer: What's the use case for the condition?
Leigh Klotz: It's the opposite of replace. It gives you an error if you attempt to set a header that is already set, most usefully from an existing browser header.
John Boyer: Or if you have nodeset/header with calculations.
John Boyer: It sounds like the headers are mixed in. Can you calculate the headers before initiating the URI request?
Leigh Klotz: I assume it's a loop unrolling that happens to create a "display list" of names, values, and combinators, which are then serialized into MIME in the HTTP method handler (or whatever).
John Boyer: So the problem is that maybe it's not possible in step 5 to ask the browser what steps are possible.
Leigh Klotz: Step 5 calculates this; step 9 puts it into use.
John Boyer: You'll have to say that you aren't doing the combining in step 5.
Leigh Klotz: Actual determination of header serialization is done in the submission serialization.
John Boyer: What if there is not a combine attribute?
Leigh Klotz: That's what I'd like combine="error" for. It lets you set the header once, as long as it doesn't conflict.
John Boyer: What do we currently say? "The header element can be used to contribute information to the preamble of a submission in a manner appropriate to the protocol." So there's no previous behavior.
John Boyer: So what about the default for combine for Accept being what people most want?
Leigh Klotz: That would not satisfy the use case since if you default to append, it doesn't do anything for content negotiation.
John Boyer: Append is implementable.
Nick van: [irc] I think we can use append as default, because if you have multiples in your document you want this default.
Leigh Klotz: So if we make append the default, then we don't need error.
John Boyer: That sounds like the best thing. It also makes the existing behavior the same. Adding multiple value elements is the same; existing content still works.
Leigh Klotz: Is the best copy of the schema the one in the spec?
John Boyer: Yes. It is linked.
Leigh Klotz: I'll produce a change packet.
John Boyer: I'll check it in for you.

* Review of XPath Functions Module


Erik Bruchez: [joins]
John Boyer: I did the new organization of this module and then will discuss that before cutting it into five-odd specs. Section 2 (XPath Function Library) is broken into four specs if we do it this way.

John Boyer: There are data integrity functions, date and tiem functions, conditional (predicate, decision) functions, and math functions. There is some question about whether random belongs in math or data integrity. I bought Erik's argument that it belonged in the math package.
John Boyer: Any comments?
Leigh Klotz: What's left in here once you pull these out, or do you replicate it?
John Boyer: More or less replicate, but there's some slight tweaking.
Leigh Klotz: So four, not five.
John Boyer: The introduction would look more like the introduction to section 2. I'm not sure that we need to refer to EXSLT any more; that's for the consuming profile.
John Boyer: Is everybody happy with this division of functions? OK, then I'll divide these into four specs and get them ready for publication.
Nick van: [irc] yes, only the name 'conditional functions' is a bit vague
John Boyer: One of the things that I saw recently in another WG in Interaction (HGC) is that they published a first public WD and Last Call WD. So there is precedent to take some of these to last call. We should also do instance and binding pretty soon.
Leigh Klotz: The theory being that it's content that we already have anyway.
John Boyer: Yes, it's division points. We're publishing these as individual modules and all have XForms 1.1 modules. There's a slight deviation with instance, but the CR period is a time for implementations, but do we need test suites and implementations separate from XForms 1.1?
Leigh Klotz: Is there a separate test module for each module in 1.2?
Nick van: [irc] creating test suites for the modules is a lot of work but for adoption outside XForms you probably want separate test suites
John Boyer: Functions will be easy, but binding and instance data will require tweaks: binding has context (not just insert and delete). So we can't use XForms 1.1 implementations for binding. So we can go to CR at least.

John Boyer: Nick, you're on vacation so we can't start on bind.
Nick van: It's been a while since I've looked at it.

* XForms 1.1

John Boyer: Nick, have you heard from Joern on XForms 1.1?
Nick van: I'm still working on resources; vacations are a problem right now.

John Boyer: Keith, how about Ubiquity testing?
Keith Wells: The XHTML+XML content type support is lacking. We can rename the tests to get past that point but I don't want to do that right now.
John Boyer: Is there a browser where the test suite works with Ubiquity?
Keith Wells: It doesn't yet handle .xhtml files and I have changes which will make it review. In IE it's working with minor problems.
John Boyer: Do the test suite files work in IE with Ubiquity?
Keith Wells: Not without these changes.
John Boyer: It turns out we published CR at the end of November last year so it would be good to go to PR in under a year. There is also a publication moratorium around the Tech Plenary, so it will be around November at best. We can report Ubiquity under Firefox, but testing under IE is necessary. Since the Firefox plugin is available, that's valuable.
Keith Wells: We are re-doing the Firefox implementation report at your request; I'm finishing that up in day or two.
John Boyer: Zip files are a problem for the W3C list servers so rename it to .data.

* IRC Minutes


* Meeting Ends