Erik Bruchez, Orbeon
John Boyer, IBM (chair)
Charlie Wiecha, IBM (left early)
Leigh Klotz, Xerox (minutes)
Steven Pemberton CWI/W3C
Steven Pemberton: We should have a
list.
John Boyer: We can list the XForms 1.2
issues. We don't need time slots, right?
Steven Pemberton: Right.
Steven Pemberton: Hours are 9-5.
Leigh Klotz: We can do what we want
for the Xerox-hosted meeting.
John Boyer: I'll put the XForms 1.2
features as the discussion topics, OK?
Steven Pemberton: Sounds perfect.
Resolution 2009-10-28.1: We will discuss XForms 1.2 features at the upcoming FtF.
Steven Pemberton: Shall we meet
Sunday night?
John Boyer: I'm in a remote
hotel.
Charlie Wiecha: I'm not
attending.
John Boyer: Why don't we coordinate on
the list?
John Boyer: What should we do with
non-member testimonials? Put them on our web page?
Leigh Klotz: Can we just put them all
on a wiki page?
Charlie Wiecha: Just on the wiki seems
appropriate.
John Boyer: Yes, we're free to put it
in our own news site.
Charlie Wiecha: [1/2 hour break]
John Boyer: Maybe in the dated group
page.
Steven Pemberton: That works
too.
John Boyer: Steven or I could do
it.
Steven Pemberton: I'm happy to do
it.
John Boyer: I'll send you the text
file.
Steven Pemberton: I won't do it before
Friday.
John Boyer: And update the Overview
page for the news items.
Action 2009-10-28.1: Steven Pemberton to publishe all received testimonials on dated WG page and link from News items.
Erik Bruchez: String normalization
takes place before validation, so when you validate with
normalized-string or token, there's no difference than validating
with a xsd:string. You might be more restricted by using xsd:string
in some scenarios, for example if you use a restriction with an
enumeration. So we should validate the same way with bind/@type and
normalize before validating with normalized-string or token. So we
have a funny behavior, but that's what XML Schema validators do as
well. That's what I understood from my exchange with Eric.
John Boyer: Deep in the schema spec
you can decide which way to go, but they don't say which way to go.
We've taken the default settings because we want to use COTS schema
engines.
John Boyer: So we need to make a small
test change to 5.1.b.
Erik Bruchez: Let's look at it.
John Boyer: I'll put it on the F2F
agenda at the head of the list.
John Boyer:
Action 2009-10-28.2: John Boyer to fix small typo in http://lists.w3.org/Archives/Public/www-forms-editor/2009Oct/0006.html
Leigh Klotz: This post claims we
haven't released the 1.1 test suite.
John Boyer: We did incorporate the
changes. I'm not sure what's up.
Leigh Klotz: Also they ask for a new
version of the XForms Quick Reference. http://www.w3.org/MarkUp/Forms/2006/xforms-qr.html
Steven Pemberton: Good thinking.
Steven Pemberton: Steven Pemberton to update http://www.w3.org/MarkUp/Forms/2006/xforms-qr.html and XForms Guide for HTML Authors for XForms 1.1.
John Boyer: Here's the full test
suite link:
http://www.w3.org/MarkUp/Forms/Test/XForms1.1/Edition1/driverPages/html/
John Boyer: This test gives a
compute-exception and it relies on the fact that nobody defines a
function called "invalid" and should throw an error at start
up.
Leigh Klotz: I responded on his blog:
Thanks for your comments. We will update the XForms guide documents. But we're not sure about your test suite comments. The full test suite is at http://www.w3.org/MarkUp/Forms/Test/XForms1.1/Edition1/driverPages/html/ The implementation report, which shows two implementations of each required feature and one of each optional feature is linked from the XForms 1.1 Recommendation: http://www.w3.org/TR/xforms11 http://www.w3.org/MarkUp/Forms/2008/XForms11ImplReports/results.htm The particular test you mention is a test of error handling for the extensions attribute. The test asks for an error to be signalled if the implementation does not provide an XPath function called "invalid." All implementations passed this test. Please write back, preferably to www-forms-editor, if you have any more comments about the test suite.
John Boyer: We've talked about
various solutions, including xinclude.
John Boyer: Have we talked about
including fragments of multiple layers (binds, UI controls,
instances). For example, a form which collects addresses.
Leigh Klotz: We talked about doing
that with encapsulation.
John Boyer: Right, we talked about
doing it using components. And this might not inject into multiple
layers, because that would be a component, and you'd want to re-use
it multiple times: ship-to, bill-to.
Leigh Klotz: I think it turns
precisely into the components discussion.
John Boyer: OK, so that's out of
bounds as a separate problem.
John Boyer: So we want to discuss this
for a couple of hours.
Steven Pemberton: Sounds good.
John Boyer: We have another issue:
we need a requirements document for XForms 1.2 possibly before
FPWD. Or at least before last call.
Leigh Klotz: I'd prefer after
FPWD.
John Boyer: We need to get started.
Steven?
Steven Pemberton: I don't have much
time before the FtF.
John Boyer: It may be more important
than the testimonials.
Action 2009-10-28.3: Steven Pemberton to provide draft charter document for FtF.
Erik Bruchez: John sent out
examples, in the Wiki page. I proposed some variations.
Erik Bruchez: One scenario was pure
expression. We don't have problems with the processing model there.
There's not much more needed. Variables or a local instance
mechanism can implement complicated functions. You can do that in
XSLT as well, accessing global variables, creating local variables.
It would be a nice parallel with what XSLT supports already.
Erik Bruchez: If we agreed on the
syntax for function and parameters, what would be missing would be
(possibly) local instances and XForms Actions in functions.
Erik Bruchez: Then iterations. We
might support actions with some exceptions.
John Boyer: The truly painful part
about actions and local instances is that the actions must be
restricted to local instances.
Erik Bruchez: XSLT2 has a "sequence
constructor" which lets you create a variable with nested XML
elements using XSLT (for-each) and other variables. So you have a
whole syntax re-used other places in XSLT which lets you construct
results in subtrees (actually, sequences which may be strings,
elements, subtrees).
John Boyer: So if we went with a
pared-down variable-return syntax we could expand it later.
Erik Bruchez: I think so. In XSLT you
can say 'variable select="3"' or you can use th more complex
syntax.
John Boyer: It would work with XPath
1.0 as well, which is appealing. If I had a function whose params
and return value is defined by as="string" or
as="nodeset" it looks like to me like there's no way
to modify any of the nodes within these functions.
Erik Bruchez: Yes.
John Boyer: That's appealing.
Erik Bruchez: There are no standard
XPath 1.0 or 2.0 functions which modify nodes. Even the node
creation functions don't.
Erik Bruchez: I hope we can re-use
this definition of variables to introduce variables in models and
actions as well.
John Boyer: It would be useful to keep
that as a separate "Category XForms1.2" wiki page so we can track
it the issues there: variable lifetime, etc.
John Boyer: It would be nice to get
this page pared to your functional version of the examples.
Action 2009-10-28.4: Erik Bruchez to change http://www.w3.org/MarkUp/Forms/wiki/Custom_XPath_functions to use params, variables, and returns for FtF review.
John Boyer: One of the issues we
had was xforms-load, which I view as a shorthand for a send and a
submission. But we don't have the attributes from xforms-load
available on submission. We need someone to write up a page on
this. Does this seem like a good idea?
Steven Pemberton: I'm not sure.
John Boyer: We don't have officially
with load is ...
Erik Bruchez: In HTML, it's "_top" or
an ID.
Steven Pemberton: And "_new" for
unnamed new windows.
John Boyer: It seems we should have
the same facility in the submission target attribute. We can do
this in an hour or so at the Ftf.
John Boyer: This may be a 2.0
feature. Obviously when you navigate away it will shut down. We
don't have a way to find out what submissions are in progress or to
stop them. Maybe it's a DOM event.
Steven Pemberton: We're DOM neutral so
we'd need an event.
Leigh Klotz: XHR has progress
events.
Steven Pemberton: We should be close
to XHR if possible.
Leigh Klotz: Anne's document doesn't
have it yet but it's implemented already. Here's Mozilla's
documentation:
https://developer.mozilla.org/En/Using_XMLHttpRequest#Monitoring_progress
John Boyer: Where's the event
defined?
Leigh Klotz: This document is a couple
of years old: http://www.w3.org/TR/2007/WD-progress-events-20071023/
John Boyer: How often is the event
dispatched?
John Boyer: Should we model ours after
this event?
Leigh Klotz: I think we should contact
the XHR and progress-event spec comment lists and ask what the
status is so we can coordinate.
John Boyer: We just want to reference
it and may implement it separate from availability in
browsers.
Leigh Klotz: It would be best if both
XForms and XHR refer to the definition elsewhere and they access it
through onprogress and we access it through XML Events.
John Boyer: Some have asked for
uploads to be done differently. Usually you hit upload and pick a
file and it actually uploads the content at that point.
Leigh Klotz: That's done with a
separate form as the HTML4 upload control does not do that.
John Boyer: Is there any interest in
combining it with submission, so if you pick a resource locally and
your upload is connected to a submission, it would take that file
and upload it using the xforms submission and then some aspect of
the submission, perhaps the returned result with a PUT or POST,
would the URL location on the server where the file was uploaded
to. It seems useful if we could make that upload refer to the
server side resource where it was put by the submission.
Leigh Klotz: I'd say we see if we can
implement this with what we have, and if so, decide whether to put
syntactic sugar on it, and if it can't be then we analyze the
gap.
John Boyer: If you have an anyURI
upload, some local file:url is placed into the data node, so your
submission can be multipart. If I tried to do that today I'd want
to use upload, and when the user finishes using the upload, a URL
appears in the data.
Leigh Klotz: It sounds like we need
insert instead of replace on result of the upload.
John Boyer: If you have the
base64binary it gives you the content in the instance data.
Leigh Klotz: I don't think the spec
says you have to put the data in.
John Boyer: I think it does.
Leigh Klotz: I see. Well, if you have
large data it gets loaded into memory.
John Boyer: If it were easy for the
upload to say compress...
Leigh Klotz: Compression isn't the
issue. If you want programmatic access to files that's a new
requirement.
John Boyer: If you want a 64MB file
you wouldn't bind it to base64binary or hexbinary. But you can't
limit the size. There are some problems.
Leigh Klotz: Do you want to insert
the resulting URL into the other instance?
John Boyer: It's not XML.
Leigh Klotz: If you use it multiple
times you need it.
John Boyer: Yes, dealing with multiple
files also seems important.
John Boyer: I'd like to have a
submission with an upload with local resources sent.
Leigh Klotz: I understand the goal,
but we need to decide if we need syntactic sugar or new
features.
John Boyer: We don't guarantee what
the resulting URL will be.
Action 2009-10-28.5: Leigh Klotz to write example of immediate upload using XForms 1.1 features.