W3C Forms teleconference August 20, 2008

* Present

John Boyer, IBM (chair)
Kenneth Sklander, Picoforms
Leigh Klotz, Xerox (minutes)
Nick van den Bleeken, Inventive Designers
Paul Butcher, x-port.net
Roger Pérez, SATEC
Steven Pemberton, CWI/W3C
Keith Wells, IBM
Charlie Wiecha, IBM
Erik Bruchez, Orbeon [joined late]

* Agenda

http://lists.w3.org/Archives/Public/public-forms/2008Aug/0058.html

* Previous minutes

* The Forms Newsletter

John Boyer: We haven't had much news to print.
Leigh Klotz: Yes.
John Boyer: Steven, you're working on a story.
Steven Pemberton: Yes, in a couple of days.
John Boyer: So it seems like we should hold off until we get 1.1 out.

* XForms 1.1

John Boyer: This is on the to-do list. Then there's Submission headers, event context, a link issue. We'll discuss those later.

* XForms 1.1 Test Suite

http://lists.w3.org/Archives/Public/public-forms/2008Aug/0054.html

Keith Wells: There are two test cases: 2.3.4 for bind without binding attributes, and some SOAP changes. The other outstanding changes for submission header are for today.
Keith Wells: On the implementation report, we're testing nightlies of FF2 and FF3. I'll run through those results today. There will be updates to the XForms public extension page in early September.

Nick van: I'm hoping to tackle the problem for testing Chiba, but we're in a release here.

John Boyer: We're still looking at doing a report for Ubiquity. It's not complete at this point.
Nick van: Is there any work on annotating the test suite files with attributes for expected values?
Keith Wells: We're looking into the Selenium test suite; it's work in progress and we're working through stumbling blocks right now. I've made the changes in CVS so far.

* XSD 1.1 review

Kenneth Sklander: I've reviewed data types and haven't started structures.
John Boyer: Yes, I misused "XSD." It's XML Schema, not XSD.

* Joint Task Force

John Boyer: I brought the progress issue up in the HCG. The results is that we must coordinate with HTML WG chairs about Forms Joint Task Force.

Erik Bruchez: [joins]

* Ubiquity and Simplification

Charlie Wiecha: We're developing directions to share at the tech plenary.
John Boyer: Also can you let me know when to add the Backplane report to the agenda?
Charlie Wiecha: Maybe 5 minutes next week.
John Boyer: Through our modularization work and the JavaScript processor work, we're communicating that W3C technologies are able to be expressed, implemented, and delivered, without having to involve web browser makers. It's not a showstopper any more. Through modularization efforts, initiated by the XHTML2 WG, we're able to get interoperability on the existing web framework. These are important messages to get through the W3C.
Charlie Wiecha: So I raised the question of connection between simplification and backplane. So John said that if you look at the way Dojo approaches page construction, it's through progressive enhancement: a set of empty divs is scripted into more. We can look at simplification, with additional structures developed under view controls and in the model. It's a recursive, or iterative process, expanding the content, and then the the view. So a more general message about the simplified syntax might be a progressive expansion.
John Boyer: I've been waiting for the forms joint task force to bring up the ARIA argument and ask for xf-calculate and xf-datatype. Those might be more palatable. What do you think?
Charlie Wiecha: We're just starting to think about it for Ubiquity. I had thought of using a namespace, but we might want to revisit that.

* XForms 1.0 Test Suite

Leigh Klotz: So how many tests do I need to check? Do you validate it as chair? Does Kenneth assert it has passed?
Kenneth Sklander: The test suite was from 2005. Did you record what was done then? I have made notes on the changes incorporated in the test suite in the margin. Those have all been incorporated in the test suite.
John Boyer: How many tests are reported? 100 or 400?
Kenneth Sklander: More like 100. It's the entire 1.0 test suite.
John Boyer: So it's 1.0 first edition test suite.
Kenneth Sklander: Yes.
John Boyer: Then we won't run the second or third edition test suite at this point.
Kenneth Sklander: No.
John Boyer: We have the 1.0 first edition reports from others. As long as the implementation report says it is on first edition.
Leigh Klotz: So this is XForms 1.0 Basic Profile First Edition.
John Boyer: Yes.
Kenneth Sklander: Is there a test suite available for the third edition.
Keith Wells: Yes.
Kenneth Sklander: Are there a lot of things to run?
Keith Wells: It's greatly improved, and there is a different way of running the tests.
John Boyer: Numerically speaking, there are over 400 tests.
Keith Wells: No, that's in 1.1.
Kenneth Sklander: We have the 1.1 product now.
John Boyer: Does the XForms 1.0 Basic refer to which 1.0 edition?
Leigh Klotz: So do we want an XForms 1.1 Basic?
John Boyer: Maybe.
Leigh Klotz: There won't be an XForms 1.0 TE Basic implementation report.

John Boyer: So we should go forward with XForms 1.0 with its reference to basic profile. We should consider XForms 1.1 if someone wants to do the tests.
Leigh Klotz: And for XForms 1.2 there is no need because it's all modularized.
John Boyer: Right. So we can publish the implementation report in our dated space.
Leigh Klotz: And I'll make sure that the reference from XForms 1.0 Basic is to the dated specs.

* bind

Nick van: No progress yet.

* Bug fix issues

John Boyer: CR is the implementation period, but do we need to go back to last call? No alarms are up for me yet but if you see any issues, please let me know.

* Common event info not exposed in event() function

http://lists.w3.org/Archives/Public/public-forms/2008Aug/0048.html

John Boyer: We expose the context but we don't expose the the target. It's fairly easy to add in the event function. It's the regular XML Events pieces of data. It's going to mean more test suite work and more implementation work. Does anyone else feel that the event function is broken as a result of not providing the event information? Is there some reason to prefer that it doesn't provide that information?
Erik Bruchez: I want to expose it but I don't have an opinion for 1.1 or not. In our implementation, we export target and event name. We expose custom information as well: bubbles, cancelable, etc. We also have a facility to add custom context information. The most useful one is the target, and the event name can be useful.
John Boyer: This isn't as big an oversight as with submission, but it is an oversight. It's localized to the event function.

Leigh Klotz: Let's see if we can capture the questions so we can repeat them next time, because eventually we'll have to say no. Is anybody opposed to the feature in general? Is it important to have? Is it important to have now, or can we wait? Does it cause last call? If so, does it cause limited last call?
Erik Bruchez: That's a good analysis. I think there are a dozen or so things of this level and we should push them out.
John Boyer: Good point. Thanks for the analysis Leigh. We'll push this forward.

* Submission header problem

Leigh Klotz: It worries me that the appending and the accept-type corner cases are the ones raised by implementers.
John Boyer: Then do the rest later.
Paul Butcher: Appending can be handled with concat.
Leigh Klotz: I don't think you can access the current header to concat. I don't think we can legislate what HTTP serialization does with Accept-Type because we don't have a good interoperable and useful suggestion yet.
John Boyer: So can you write the note?

Keith Wells: We need to be lax on how it can appear?
Leigh Klotz: For generic headers, test and accept either serialization in HTTP but for accept type, it's vague.
Keith Wells: Here are some other issues with multiple xforms:headers each with multiple xforms:value and multiple nodes that match.

Leigh Klotz: For headers of the same name, I'd say we go with document order for the xf:headers, followed by instance order of the nodeset, followed by document order of the xf:value within the header. For headers of different names, there is no need for order testing.
John Boyer: Why is uniformly appending not implementable?
Leigh Klotz: Because it needs to be moved to the HTTP serialization section, because the protocol decides what append means. We can restrict to just one header with multiple values, but I'm not sure what it buys us.
John Boyer: It doesn't buy us much.
John Boyer: Can you proceed?
Leigh Klotz: I can write the text but I'm worried that nobody would implement it. It lets us set the headers that are not set, but there's no way to say what to do with pre-existing headers, which cuts out the ability to do content negotiation for REST services.
John Boyer: So a reasonable implementation could make use of a replace-vs-append flag. And without it we don't achieve our goal.
Leigh Klotz: We need something, but I don't know if it's the attribute. We need to say what happens if you have multiple headers with conflicting values of the attribute, for example.
John Boyer: Ok. So can you write it up?
Leigh Klotz: That's the question: do we do the conservative thing that is implementable but not useful, or do we go back to the drawing board. So you and I agree, butr do we have consensus?
John Boyer: Any objections?
Steven Pemberton: [irc] I think the use case is essential.
John Boyer: OK. So if it doesn't meet the needs we need to remove it from CR anyway if nobody implements it. So let's go forward.

Action 2008-08-20.1: Leigh Klotz to investigate replace attribute on header and discuss test cases with Keith Wells and consult with potential implementers and users and provide spec-ready text for XForms 1.1.

* The xforms-link-exception

http://lists.w3.org/Archives/Public/www-forms/2008Aug/0001.html

John Boyer: So can he do a setvalue inside the link exception handler to set a value from the inline instance? We said the processor would be going down anyway.
Paul Butcher: By the looks of it, it looks like it would do the setvalue, but the default action would stop processing. But outputs and messages wouldn't happen.
John Boyer: If the instance in the source attribute were the first, would it matter?
Paul Butcher: That's a difficult question.
Charlie Wiecha: We might just call this error in my stuff.
John Boyer: We can't make any change for 1.1. For 1.2, we'll use this additional xfor-link-error as a way to get out for the problem. The question is still open about when he can write to the instance. So we'll take corrective action in a future version.

* IRC Log

http://www.w3.org/2008/08/20-forms-minutes.html

* Meeting Ends