W3C Forms teleconference December 3, 2008

* Present

Charlie Wiecha, IBM
John Boyer, IBM (chair)
Leigh Klotz, Xerox (minutes)
Roger Pérez, SATEC
Steven Pemberton, CWI/W3C
Keith Wells, IBM
Nick van den Bleeken, Inventive Designers

* Agenda


* XForms Basic 1.0

Leigh Klotz: I sent Kenneth Sklander's test results to John.
John Boyer: The HTML doesn't look like a report; there's no red or green.
Leigh Klotz: It's the old format. The spreadsheet lists pass, fail, dispute, or inapplicable. We're not going to go back and revise the old test suite for XForms 1.0 that this was against, and we're not likely to re-run it.

* Upcoming Holidays

John Boyer: No call on the 24th or 31st. Anyone else want to nominate th 17th?
Keith Wells: I won't be there on the 17th.
John Boyer: I assume January 7th is OK. Does that seem reasonable to have a call then?
Keith Wells: I think so.
Charlie Wiecha: Yes.
John Boyer: Anyone else away on December 17th?
Charlie Wiecha: We should poll by email for that one.

Resolution 2008-12-3.1: No call December 24th or December 31st.

John Boyer: Email sent. http://lists.w3.org/Archives/Public/public-forms/2008Dec/0008.html

* Upcoming Topics

John Boyer: What will we do on the calls without drafts of modules? SHould we have working calls?
Charlie Wiecha: I can spend some time on the data islands and we can discuss that on the 17th. I'll need some help so we can have it a working session.
John Boyer: Do you want to have that on the 10th as well?

John Boyer: Uli is out the rest of December, and he's currently attached to a number of specs.
Charlie Wiecha: You take a lot of care with your formal spec-writing process and I'm happy to take a crack at the functionality but I'll need help on that side.
John Boyer: Do we need a presentation on SpecXML?
Charlie Wiecha: It's not that level; it's the way to use references and notes and citations. It's no the syntax; it's more the usage.
John Boyer: I hesitate to cancel the calls because we won't do work outside then.

* Next FtF

John Boyer: The upcoming one is at Google. Then the next after that is either in Amsterdam in June or in London arranged with Mark and Charlie. Steven will be unable to travel then. He says he's happy to host it in Amsterdam, but did you have your hopes set on doing it in London.
Charlie Wiecha: That's fine with me too, but I wonder about the travel restrictions? I haven't heard anything yet.
John Boyer: Amsterdam or London will be the same issue through. Hopefully earlier in the year we'll be able to avoid the same situation as this year.
John Boyer: We'll go forward with Amsterdam in June and see what happens.

* Firefox implementation report

John Boyer: How's the report going?
Keith Wells: The HTML page for FF2 and FF3 is on the list.
John Boyer: public-forms?
Keith Wells: www-forms
Leigh Klotz: http://lists.w3.org/Archives/Public/www-forms/2008Nov/0007.html
John Boyer: So this is the new implementation report. I'll publish this in the 2008 space, right?
Keith Wells: Yes.

Action 2008-12-3.1: John Boyer to publish http://lists.w3.org/Archives/Public/www-forms/2008Nov/0007.html in 2008 space.

John Boyer: How did it do? Number of tests passed?
Keith Wells: It's summarized by chapter. I used XSLT to transpose the XML to a tabular format. It would have been nicer to keep it as XForms.
John Boyer: Thank you very much.
Keith Wells: There's also a living implementation report on Ubiquity XForms site for IE7; we're also looking at test results for Firefox 3. It's a living document and follows the codebase and tests. Those should also be able to be used as implementation reports.
John Boyer: Absolutely. We don't have as many tests passed as I'd like and a larger number that could pass with a few technical blocking factors. We need other implementation reports so we can advance XForms 1.1.
Nick van: [joins]

John Boyer: Some of the forms require viewing the submit button. Under firefox the submission won't work because of the cross-domain submission.
Keith Wells: That's a good question. A lot of these tests are XML-instance related. Leigh's echo service is invaluable in showing the instance.
Leigh Klotz: Isn't it allowing it? I put the feature in.
Keith Wells: The plug-in bypasses its. If Ubiquity is a plug-in...
Leigh Klotz: For plus-ins?
Keith Wells: Firefox 3 supports XHR cross-domain submission only for plugins or signed jars.
John Boyer: That's weird.

John Boyer: Mark Birbeck submitted a bug report for XForms 1.1. He said he didn't understand why the data serialization was not included in the event. I explained to him that at the time at least, the working group felt that if someone was going to overload serialization with their own data, serializing the data seemed like extra unnecessary work. Mark's point was that having the serialization available was useful as you could embellish the existing serialization. It makes certain programmatic extensions easier. One is that we could re-do our tests and write a message that captures the xforms-submit serialize. The submission subseuqently failing is irrelevant.

Leigh Klotz: Or I can just fetch the pages by URL and serve them up. It won't help with the local disk which will need upload.
John Boyer: That's fairly interesting. There's the thing they test and then there's the 5 or 6 other things that have to work.
Keith Wells: That's not necessarily a bad thing. It's been that way for a long time, since 1.0.
John Boyer: I guess in the submission case it's proving difficult. We don't have the problem in a practical deployment.
Keith Wells: There's a way to bypass the security through Firefox and that's what Mark and I do for submit. But it's a temporary code change. It's not a burning issue right now.

* xforms-submit-serialize receives empty submission-body value


John Boyer: This is relevant to what we're saying. I'd like to get some opinions. So what is provided is used in lieu of the data submission that is normally constructed. If you look further down, Mark's original issue is that the submission body should contain the real data; whether someone has modified it or not is not important. If they have captured it, they can use the data as a starting point for constructing their own alternate serialization. The explanation of why is that in a lot of cases, someone wants a radically different serialization gives. They can ask for the instance document and create a serialization anyway. His examples is addresstype=home and they could replace it with sub-elements. They're going to have to parse the serialization anyway. So it doesn't give you much.
Leigh Klotz: It includes relevance pruning.
John Boyer: True. We don't expose IDL for that but a reasonable implementation of an XForms processor would allow you to find those MIPs out anyway. But you're right, we don't expose that in the processor. It just seems to be a nice feature, but at the end of the day, perhaps in some future version we could put in a switch to say whether you want the data serialization to be created for submit-serialize or a serialize-with-relevance function to be available. It seemed more like an additional feature. I'm not getting a lot of consensus.
Nick van: [irc] is it for performance that we don't want to always provide it?
Charlie Wiecha: If we were doing it from scratch I'd do it the way Mark says, but there is the momentum of it being that way already. It might be appropriate to do it in the modules.
John Boyer: IDL for serialiation-with-pruning.
Leigh Klotz: Not just pruning; what if we add other features to serialiation?
Charlie Wiecha: Right, the whole serialization.
John Boyer: Or we could generically have a style-like attribute with parameters saying what you want.
Charlie Wiecha: And Nick's IRC question?
John Boyer: Yes, since there is an alternative serialization.
Nick van: If it's mostly for performance then detect when someone needs it and then serialize.
John Boyer: How?
Nick van: By analyzing the XPath expressions.
John Boyer: Which ones?
Nick van: You can only access it inside the document. So if you don't use it anywhere in the document then you are safe not to serialize it. If you want a smarter solution it may also be possible.
Charlie Wiecha: [IRC] TMM -- Too Much Magic
John Boyer: It would be more direct to have a switch.
Leigh Klotz: If there's no way to tell whether it's produced then I don't see it.
Charlie Wiecha: You're only looking at XPath, not scripting.
Nick van: You can process the scripts. I don't like asking the author. You have to remember to switch it off or on.
John Boyer: If we made the change, it would seem to be better to say that it has it, not that we say how to optimize its production.
Nick van: It's the same as events.
Charlie Wiecha: That's easier to determine with event registrations.
John Boyer: I can't see people parsing the script code. Once you get to the point where you dispatch the event...
Nick van: Can you do it in the implementation of the event section? So if nobody asks for it, it's not available, and a smart implementation can generate it when someone reads it. Then you don't have to parse your XPath expressions.
Leigh Klotz: So I think we should remove the performance issue from the discussion of whether we want to do this. We can hear from implementors later if they can't optimize it.
John Boyer: So that node can be set to empty string or the initial data serialization.
Nick van: So why not provide it if the reason is only performance.
John Boyer: We're a year past candidate rec. It's a last-call comment. Right now we're having trouble getting implementation reports. It's one more cool thing we can do. It's useful for testing but what else? Are we solving a big problem here?

Nick van: Or maybe we can change it in a future version. Do we now check to see the value is empty?
John Boyer: Good question. Maybe someone adds text and we see text. It makes sense to see what we're testing. It looks like the test doesn't test for empty-string in the beginning. We'd have to add a test for it. It would be 11.3.c.
Charlie Wiecha: My vote is for IDL in a module.
John Boyer: How about you, Leigh?
Leigh Klotz: Ask Paul to provide (1) use case and (2) implementation proof. We can't deal with your concern of it being too late.
Keith Wells: I vote for not doing it.
Charlie Wiecha: We know they can do it, so why decide later?
Leigh Klotz: The test case doesn't test it not being there so let's let it slide if they want to implement it. They if they do it and perhaps Firefox does it, we can put it in the next spec. That's how things ought to work anyway -- we standardize implementations, rather than the other way around.

* Javascript implementation of replace all submissions


John Boyer: We need a public response for this from Paul Butcher.

* Problem with xf:copy and xf:delete


* Submission headers fix


John Boyer: Until we get implementation reports finishing this merge into the spec isn't high prioity but it's bubbling to the top.

* Problem with test suite test


Keith Wells: This problem has been here since 1.0. I find it worrisome.
Leigh Klotz: This issue appears to be in the PicoForms test report as well.
Keith Wells: I've received feedback from Aaron and Merle.
John Boyer: The first test I opened had a problem. The others are good...
Keith Wells: I'll fix the test.
John Boyer: It's supposed to test bind overriding nodeset. The test form seems different. The outcome it describes would only be possible by having the bind override the ref on the select1 itself.
Keith Wells: That's messed up. And I can take out the submission.

Action 2008-12-3.3: Keith Wells to update test http://lists.w3.org/Archives/Public/public-forms/2008Oct/0038.html to fix binding and remove submission if necessary.

* XForms for HTML

John Boyer: We still need the shortname request. Is Steven Pemberton here? I'll put it in the HCG request.

* Last call on XSD Component Designators


John Boyer: This sounds like the thing we heard before, of the common patterns.
Leigh Klotz: I found it: it's an XPointer scheme for innards of XSD: http://www.w3.org/TR/2008/WD-xmlschema-ref-20081117/#key-scd

Syntactically, the first part is a URI without a fragment identifier, and the second part is an XPointer fragment identifier. An absolute schema component designator therefore is a URI reference. For example: http://example.org/schemas/po.xsd#xscd(/type::purchaseOrderType)

Leigh Klotz: It seems like an XPointer scheme for indexing into an XSD document at a slightly higher level than the XML elements that comprise the document.
John Boyer: Can you ask them if it's relevant to XForms?
Leigh Klotz: Sure.

Action 2008-12-3.4: Leigh Klotz to ask WG if http://www.w3.org/TR/2008/WD-xmlschema-ref-20081117/ is relevant to XForms.

* IRC Log


* Meeting Ends