John Boyer, IBM (chair)
Charlie Wiecha, IBM (scribe)
Paul Butcher, WebBackplane
Leigh Klotz, Xerox (scribe)
Nick van den Bleeken, Inventive Designers
Keith Wells, IBM
Steven Pemberton, IBM
Uli Lissé, DreamLabs
http://lists.w3.org/Archives/Public/public-forms/2008Dec/0028.html
John Boyer: Who can chair in two
weeks?
Steven Pemberton: I can chair January
21st.
John Boyer: We're meeting at
Google. The Wednesday telecon is Feb 4. The virtual day is Feb 5.
Travel is Feb 6 or 7. The meeting is Feb 9, 10, 11.
Charlie Wiecha: I'm not going to get
approval.
Steven Pemberton: I anticipate being
ok, as it's the one meetign I'm travelling to.
John Boyer: I haven't asked yet.
Nick van: I booked my flights already,
so I'll be there.
Steven Pemberton: Did anybody know
that Raman got married.
John Boyer: Have you done the
public list response needed?
Leigh Klotz: I should start it.
John Boyer: Maybe you could correspond
with Eric to double-team it as he did some work on this topic in
the past when we last visite the technical details.
John Boyer: I need to look at this
and make it available alongisde the Firefox one. They said it's
better than 90% conformance.
Steven Pemberton: And it's great that
it comes from EMC.
Action 2009-01-7.1: John Boyer to review/publish EMC implementation report http://lists.w3.org/Archives/Public/www-forms/2008Dec/0020.html
Steven Pemberton: Can we invite
them to join the working group?
John Boyer: Sure.
Steven Pemberton: Who did they send it
to?
John Boyer: www list.
Steven Pemberton: I'll invite them if
you don't mind.
Action 2009-01-7.2: Steven Pemberton to invite EMC to join Forms WG.
Paul Butcher: We're not planning to
do a report for formsplayer, at least for the next couple of
months, because we're working on Ubiquity.
John Boyer: It's a very highly
conformant implementation so it would be nice to have its coverage.
Maybe we can do a triage and find out which tests we don't have
covered and get a constrained report.
Keith Wells: I can do that.
Action 2009-01-7.3: Keith Wells to examine Firefox XPI, and EMC implementation reports and find out which test cases are not covered by both tests.
John Boyer: It was surprising that
the power() function wasn't there in Firefox.
Keith Wells: I'll take a look at
it.
Action 2009-01-7.4: Keith Wells to confirm power() function in Firefox and request its implementation if necessary.
John Boyer: That and the question mark at the end of serialization=none URIs. There's a pending action item on me to fix the spec to say that when serialization=none you don't get the question mark. There is someone who cares (using it for ATOM work).
John Boyer: Any other implementation report issues or test issues?
John Boyer: Wasn't there an issue
with empty binds.
Paul Butcher: The test was that the
events weren't coming from the correct elements.
John Boyer: The issue was <bind ...
/> vs. <bind> ... </bind>?
Paul Butcher: Yes.
John Boyer: We don't have tests that
look at some of the challenges that look at XML content in an HTML
parser; this is one of the issues that we hit.
Leigh Klotz: There was a similar
problem with XHTML 1.0 and the space in <br >
Steven Pemberton: We said if you think
this is going to be served in HTML browsers you should put the
space in before the slash; modern HTML browsers don't require the
space any more.
Leigh Klotz: So John's asking about
how legislate around a similar issue involving short-form close of
bind.
John Boyer: Is it OK to change the
text of a test to pass in this case?
Steven Pemberton: It's the same in XML
and you could argue that it's required for certain browsers, so I
think it's pragmatic and nobody would argue.
John Boyer: It works in IE. The binds
aren't being interpreted correctly but the UI controls receive the
events. It's the HTML parser that's removing parts of the form,
before Ubiquity gets it. We're still trying to make sure we have
more accounting for how XForms gets used on the web, and if we an
implementation that says, subject to certain limitations, you can
use XForms in your HTML browser, it might be reasonable
implementation.
Leigh Klotz: But didn't the issue go
away with br?
Steven Pemberton: In HTML browsers,
the slash just goes away. It doesn't know about the slash close. So
it interprets it as an unclosed element and waits for the close.
All they did was allow the absence of the space.
John Boyer: So we'll consider these XML issues separately from conformance.
Nick van: I added a MIP element
with name and value and cascade.
John Boyer: You need the inherit and
cascade for more flexibility to casecade with non-boolean
ancestors.
Leigh Klotz: I did an analysis of a
bunch of them a while back. Or, And, first-wins, last-wins, all
must match, etc.
Uli Lissé: [irc] wouldn't it be
easier to just have a cascade attribute, which specifies an
and or an or combinator?
John Boyer: And how can you ask the
MIP value of a data node?
Nick van: I was planning to do
that.
John Boyer: That opens a huge can of
worms, if it can ask the value of a constraint or readonly. The
only think we can create dependencies on node values, so then we'd
create dependencies on node properties.
Charlie Wiecha: [irc] i don't recall
whether we decided to bite off custom mips for this iteration of
the bind module?
John Boyer: It may not be so bad if we
leave off the ... like id() returns a variable nodeset. We could
say the MIP function is like that.
Leigh Klotz: Aren't they the same as
calculates with values on a separate instance? So they are not
circular references but instead calculates.
John Boyer: Yes. But right now if the
function returns a nodeset, if it references are a nodeset then
those are dependencies. So there's one generic location outside
XPath functions where we make tests about what's being referenced.
The MIP function would return a number, a boolean, or a string, but
unless the internal implementation of that function is able to
access some API to say "I am setting up new dependencies."
Leigh Klotz: What new
dependency?
John Boyer: It would say it's
referring to this property of this node.
Leigh Klotz: What if the mip element
was moved out to be defined statically, and then the mip-value
function took a node and the mip name? That's a lot like calculate
and the compute system should handle it.
John Boyer: The containing expression
is now dependent on the MIP named X on node Y. No other function
has had to be able to declare out to the calculation engine that
it's dependent on the MIP value, not the node value.
Leigh Klotz: OK. But as Charlie said,
is this for combining old MIP types or is it for adding custom MIP
types.
Nick van: It's for our MIPs and
cascading them, just for our own MIPs, but it also allows creating
custom MIPs.
Leigh Klotz: So solve both problems
with one mechanism.
Nick van: Maybe we don't need cascade
and combine and maybe not even type. But the same API would need
those.
Leigh Klotz: If we move the custom
definition out into a separate element and refer to it, and drop
the unnecessary attributes now, then we have what's in 1.1.
Charlie Wiecha: [irc] i would prefer
to keep this module at the level of current, 1.1, function. do we
need to talk about cascade at all then? separating def and use
would be great.
Nick van: So it would be a child
attribute on bind. So there should be an element named mip and
attribute calculate.
Leigh Klotz: It's hard to schema
validate if you introduce attribute names by definitions inside the
document.
Nick van: In the bind I'd like to use
readonly as an attribute on bind, for example. Can't we allow every
attribute on bind?
John Boyer: There are two issues: for
schema validation we can't have declarations within the form
dynamically adding new valid attributes.
Charlie Wiecha: [scribe via IRC]