Erik Bruchez, Orbeon
Leigh Klotz, Xerox (minutes)
Nick van den Bleeken, Inventive Designers
Steven Pemberton, CWI/W3C (Chair)
John Boyer, IBM
Uli Lissé, DreamLabs
Charlie Wiecha, IBM
Steven Pemberton: Two team members
have been assigned the task of reviewing the charter, and one is
moving this week so the review won't be ready for management review
until at least next Wednesday. This puts us in a crunch for the F2F
at the end of March.
John Boyer: Might they extend they
charter while doing that?
Steven Pemberton: I assume that's a
done deal, though I haven't seen the message. But we'd like to
attract new members to the F2F. I suspect Nick has already
booked.
John Boyer: I'm expecting to request
travel approval as well.
Steven Pemberton: So are we still
going with it?
John Boyer: It seems like the best
idea to me.
Steven Pemberton: As usual, it's
moving slowly. We can then move quickly.
John Boyer: We have one test for
this and there are untested assertions. If you have a submission
resource attribute with a URL including parameters already, we say
that we will combine them with serialized XML data; tacking the "?"
on the end isn't sufficient. You need to decompose the URL to
figure out if it has any URL parameters already and then combine
them? Then you need to combine them.
Leigh Klotz: But you can't re-order
them. That changes semantics.
John Boyer: You can't just look for
the last "?". Mark Birbeck has a blog entry about it.
Steven Pemberton: Treating URIs as
strings considered dangerous
http://webbackplane.com/mark-birbeck/blog/2010/01/treating-uris-as-strings-considered-dangerous
John Boyer: For example, there could
be a #fragment at the end.
Leigh Klotz: Aren't these parts named
in the RFC? Can't we just tell people to add it to the right
part?
John Boyer: The spec doesn't go that
far. It's just the test.
Steven Pemberton: OK, I hear what
you're saying but I don't see how your comments apply to the
test.
John Boyer: I have some more tests
with "?" in the URL parameters.
Steven Pemberton: You are testing the
action URL combining. Got it.
John Boyer: I didn't add a
fragment-identifier test, but there are basic tests for adding
paramters and one for just making sure we add just one "?" as those
are both listed in the spec.
Leigh Klotz: I don't see any problem
adding the fragment test as well, as long as you want to right
it.
John Boyer: 11.9.1.d, as long as we're
in the neighborhood. I know I have an action (nine months ago) to
add tests to section 11. It's still pending. I don't wan to add new
tests
Leigh Klotz: Maybe we can start rev2
test suite and back it with errata when they come out.
John Boyer: Yes, edition2 test suite
now that we have a couple of errata. We can add new tests
there.
Steven Pemberton: Today we're just
OK'ing these tests?
John Boyer: I would not add them the
edition1 test suite.
Leigh Klotz: Can't we just do a thin
test suite?
John Boyer: That's not a bad idea if
we can do it.
Leigh Klotz: I don't see why
not.
John Boyer: Thin-test.
Resolution 2010-01-27.1: We begin to collect XForms 1.1 Edition 2 Thin Test Suite, containing both new or changed tests for existing assertions, and new or changed tests for current and forthcoming errata.
Action 2010-01-27.1: John Boyer to add New tests 11.9.1.b and 11.9.1.c (and .d for #fragment) to XForms 1.1 Edition 2 Thin Test Suite: http://lists.w3.org/Archives/Public/public-forms/2010Jan/0021.html
Nick van: I am producing a grammar
for a formal test suite framework that can easily be transformed to
Selenium or other test frameworks.
Steven Pemberton: Can you send email
to the list explaining how it works?
Nick van: It's presently online.
http://github.com/nvdbleek/com.orbeon.testsuite.w3c
Nick van: I'm working on our internal
tests as well at the moment.
John Boyer: It might be good to see if
the new test 11.9.1.c could be authored differently, as an
example.
Nick van: For every test HTML file I
have another XML file that describes how to test the form. There
are elements that say to fill in the first input field, an output
control present with a value, must be invalid (assertions), etc.
The tests don't need to change.
John Boyer: So it's a gramar for
describing the automated test.
Nick van: I've looked at Ubiquity and
Chiba tests using Selenium, but they don't work with each other
beacuse the tests are implementation-specific. So I've created a
Selenium "User Extension" to query the XForms state, but it was too
verbose, and I've created a grammar for writing the tests.
Leigh Klotz: This is of course
valuable for testing XForms used in production, where the test is
of the application and not the XForms implementation.
Nick van: It could also work for
Picoforms.
Nick van: It makes it easier to have
tests done for multiple implementations because XForms implementors
must just write a test adapter for my grammar. I'm not sure if the
WG feels about this, but at least we could link it from the W3C
site. It gives a more formal way to test.
John Boyer: I think that would be
valuable for all current tests. That is the essence of "Improving
the XForms 1.1 test suite" http://lists.w3.org/Archives/Public/www-forms/2009Jun/0017.html
agenda item. It sounds like the way forward.
Action 2010-01-27.2: Nick van den Bleeken to send description of test framework to public-forms
John Boyer: It would be good to
have the automation added to Edition1, in a parallel
directory.
Leigh Klotz: Or just put it in the
existing test suite, since it doesn't affect conformance.
John Boyer: Right, addition of XML
files for tests.
Nick van: I hope it will be completed
for all tests before the F2F.
John Boyer: Fabulous. This has been a
huge problem.
Steven Pemberton: Great.
Nick van: The XPath 2.0 module made me
think of this, as a test report is a prerequisite.
John Boyer: Does it prohibit a
second submission while the processor is still workingo n the
first? The test claims you should not see an xforms-submit-error.
In XForms 1.0 we had some implementations with synchronous
submissions from xforms-submit-event to xforms-submit-error as a
blocking process. So dispatching two in the same submission would
be successive. The test is correct there because the submissions do
not overlap. Now in XForms 1.1 submissions are asynchronous, it's a
race condition.
Leigh Klotz: I think the intersection
between submit-once and asynchronous submission isn't done right. I
can see situations where you want to allow multiple outstanding
asynchronous submissions.
John Boyer: The submit-once is for
convenience.
Leigh Klotz: Yes, I proposed it.
Steven Pemberton: The submit button is
supposed to be inactivated.
John Boyer: It's bound to the
submission via id. We should have something that takes a long
time.
Nick van: Do you know that this
works?
Leigh Klotz: I'll give you a sleep
test; that's not hard. Do you want a 200 back or do you need it to
fail?
John Boyer: So we need something like
echo that conditionally fails.
Action 2010-01-27.3: Leigh Klotz to provide sleep and fail version of echo for xformstest.org
Steven Pemberton: Our intention was
to make the submit unavailable; not that the user should get an
error message.
John Boyer: The user gets an error
message if the form author hooks xforms-submit-error.
Steven Pemberton: So the trigger looks
like a normal trigger but it doesn't do anything.
John Boyer: Right. The form author
could provide an error message. We have an error type in the
context information so they can check error type.
Leigh Klotz: I wouldn't be surprised
if some implementations don't just gray out xf:submit.
Steven Pemberton: That's what you
really want. So that you can't click it again.
Leigh Klotz: I think it's hard to do
with xf:trigger but it should be possible for implementations to do
with xf:submit.
John Boyer: Usually people to
setvalue, setvalue, send.
Steven Pemberton: That's too bad.
You'd like to wait for all actions.
Erik Bruchez: We have an extension
attribute on trigger/@modal=true. The whole screen
becomes unavailable until actions are done.
John Boyer: Does that include
asynchronous submissions?
Erik Bruchez: No. The sync/async isn't
it. In our implementations we don't default to asynchronous. We
have partial asynchronous submissions. Forgetting that question, it
works quite well with an attribute to control availability. Our use
case was that we didn't want the user to do anything. You could
imagine the trigger being model or the whole UI being modal.
John Boyer: Maybe an item of
dicsussion could be to look at this in future features? We want to
fix it but we want better features.
John Boyer: I like not being able to
DOMActivate the trigger as long as something is happening.
Leigh Klotz: At the same time I'd like
to consider adding multiple activations for asynchronous
submissions.
John Boyer: The stability might be
compromised. The data will show up in the same place.
Leigh Klotz: I have some experience
with this. As long as you put in a request ID and your system gives
it back you can track things. I'd like to see it considered.
John Boyer: OK.
John Boyer: So back to the tests.
Do you want a second test after the succesful and unsuccesful case
that the flag gets set back?
Leigh Klotz: Not necessarily; I just
wanted to provide the server side in case we want to do it.
John Boyer: I don't think it's a
problem to write.
Leigh Klotz: If you want to when
you're doing the other.
Charlie Wiecha: I'm happy to write
it.
Action 2010-01-27.4: Charlie Wiecha to fix 11.2.a http://lists.w3.org/Archives/Public/public-forms/2010Jan/0022.html http://lists.w3.org/Archives/Public/public-forms/2010Jan/0022.html http://lists.w3.org/Archives/Public/public-forms/2010Jan/0022.html
John Boyer: That can go back to
Ubiquity.
Charlie Wiecha: Yep.
John Boyer: A future agenda item.
Action 2010-01-27.5: Leigh Klotz to add Submit and Trigger Availability during (multiple) submissions to XForms Future Features
John Boyer: I fixed this.
Leigh Klotz: OK.
Steven Pemberton: Some of these
examples don't test statements in the specification.
John Boyer: The one we just discussed
has a backing statement originally.
Steven Pemberton: How about correct
construction of URLs.
John Boyer: Section 11.9.1 talks about
the correct separator and URL parameters.
Leigh Klotz: The correctness
definition is in an IETF document, but we still need to test
it.
Steven Pemberton: We don't test
redirection, and we don't have any spec for it.
John Boyer: It would be reasonable to
test some of that stuff that we reference. Nobody has made that
level of analysis. I'm not sure. The test suite has grown a
lot.
Steven Pemberton: There will be discussion of DOMActivate at this week's HCG call.
Erik Bruchez: I read that XPath 2.0
Rec is now three years old this week. It was stable before that.
It's time to implement them.
John Boyer: You wouldn't happen to
have a Javascript implementation of that. Don't you need an XML
Schema engine?
Erik Bruchez: You don't
actually.
John Boyer: I thought it knew
types.
Erik Bruchez: That's XSLT 2. XPath 2.0
has type info access but only requires the type libraries. It
doesn't assume a schema behind it.
John Boyer: So we decide whether it
can consume schemas?
Erik Bruchez: You don't need any XML
Schema Part 2, only the Part 1 built-in types.
Erik Bruchez: There's a write up on
our Wiki about how the xforms bind (and schema) types influence
expressions. User defined types is a different question. If you had
bind/@type=xsd:date what happens when an XPath
expression reaches that node?
John Boyer: We have union types for
example; we don't want those to fail jyust because they're not the
primitive types.
Erik Bruchez: If you use typed nodes
in expressions in XSLT 2.0, once the XPath expression runs, it has
a guarantee that the value is a date. In XForms we have a different
situation. An xs:date on an element is either empty or contains a
non-date, you're still allowed to run an XPath expression on that
node. So what are you allowed to do?
Steven Pemberton: Who should we invite
to dicuss this?
Erik Bruchez: I think once we are
ready to tackle it, I would ask Mike Kay.
Steven Pemberton: Would you be willing
to do that.
Action 2010-01-27.6: Erik Bruchez to prepare for contact with Michael Kay on XPath 2.0.
John Boyer: We haven't addressed
the original issue.
Steven Pemberton: We'll take that up
next week.