Steven Pemberton, CWI/W3C
John Boyer, IBM (chair)
Charlie Wiecha, IBM
Uli Lissé, DreamLabs
Leigh Klotz, Xerox (minutes)
Keith Wells, IBM
Nick van den Bleeken, Inventive Designers
Roger Perez, SATEC
Mark Birbeck, x-port.net (remote)
Joern Turner, DreamLabs (IRC, part-day)
Steven Pemberton: Should I forward
them immediately from the mailing list?
John Boyer: As soo as we pick one that
we're going to respond to?
Steven Pemberton: When did the Last
Call start?
John Boyer: February 22, though I
think maybe one thing showed up a little before then, February
14th. I thought we entered a coupl e of these already.
Steven Pemberton: Three. All in
Incoming. One from 05 Dec 2006.
John Boyer: It turns out that this
email is repeated in the CDF feedback and one of them was
mustUnderstand.
Steven Pemberton: I can forward them
all to the database.
John Boyer: We should be making the
spec changes as well.
Steven Pemberton: Oh, how good is
that!
John Boyer: It makes it easy to agree.
So the issues will go into the last-call system and we'll create
changes; should we create the responses as well?
Steven Pemberton: Let's do as much as
we can at the same time.
John Boyer: I have an earlier version of this one. I'm glad it's here.
John Boyer: This one already has an
action assigned to me.
Steven Pemberton: I'm sending it to
the action system. But the March 2nd one (43) is already there and
it's a follow-on.
John Boyer: Did you send any from
February?
Steven Pemberton: John 5th Feb, and
Mark's.
John Boyer: The 5th Feb was Elliot
Rusty Harrold's "getting rid of chameleon namespaces" response and
isn't an issue.
Steven Pemberton: The "where are the
audio video controls" one we have already replied to, and it isn't
phrased as a last-call issue, so we've covered it already.
John Boyer: Do we have an action item to do this? This one will take a couple of hours, so we can't do it here. A lot of the other issues are smaller, but some we will have to take away as an action. Nick?
John Boyer: We had a telcon about
this. The suggestion is that when you say
@serialize=false that the behavior of the relevant and
validate attribute should default to the same value that you set
serialize to, because when you aren't serializing the submission
data, by default you are trying to set up a submission that is just
sending to a URL, and since you aren't sending any data, not
validating it and not pruning the relevant components makes sense.
The proposal is to make that change to validate and relevant and
probably a tweak to the submit processing as well.
Steven Pemberton: OK.
John Boyer: I have spec-ready text
here. When we approve, we send a response; do we record the URL of
the response in the database?
Steven Pemberton: You reply from the
database.
John Boyer: Did we compose the reply
for 3?
Steven Pemberton: If we replied we can
put it in the notes.
John Boyer: I see the reply in there
already. Then the report will have the links to the original and
the reply.
John Boyer: Can someone compose the
reply?
Steven Pemberton: Who has a password?
I'll email Shane.
John Boyer: Can someone look at bullet
point 2 in XForms submit sequence chapter 11 with regard to
serialize=false? The thing we're trying to get working here is when
you say serialize=false that you can resolve a URL even if there is
no submission data. They won't put a ref, and that will resolve to
the instance. Is there a problem with bullet point 2?
Nick van den Bleeken: It says
submission is stopped if there is a non-relevant node.
John Boyer: If you put a ref then you
get what you pay for. If you point to a non-relevant node then why
is submission stopped if relevance pruning is off? It seems like a
fix is needed for step 2?
Leigh Klotz: If serialize=false do we
still do point 2? I guess so because relevant and validate also
would have to be false. Bullet point 2 ought to be split in two,
with relevant processing its own bullet.
John Boyer: Here's the proposed
changes from the email. Added to validate "The default value is the
value of serialize, if present, or true, otherwise."
Same on relevant.
John Boyer: Then on serialize, a
note.
Nick van den Bleeken: Why do we
need relevance pruning? If serialize is false, then relevance
doesn't matter.
Leigh Klotz: If we split the bullet
into two then we put the halt-on-empty to relevance prune. Ah, but
you still want to halt on empty even if relevance is false.
John Boyer: I don't see a big glaring
problem though. If serialize=false you wouldn't set a ref, so if
there is a default element, then the only possible problem is that
the root element of the default nodeset is non-relevant.
Nick van den Bleeken: You set relevant
to false though.
Leigh Klotz: You can serialize nothing
if all child elements are irrelevant.
John Boyer: You get the root
node.
Leigh Klotz: Not with
application/x-www* serialization; you get nothing.
John Boyer: It's OK to submit
nothing.
John Boyer: So do we remove the
halt?
Nick van den Bleeken: There's a
difference. If serialize is false we don't care.
John Boyer: So we split them up, and
halt if there is no node, then do relevance next.
Leigh Klotz: Point 2 says "the
remainder of the submit process" and then in the very next sentence
modifies it to remove nodes.
John Boyer: Yes, this is the big one
to fix. I've broken 2 into 2 and 3. I've removed the
"non-relevant-node" from 2 and we'll address it in 3. Good so
far?
Uli Lissé: Yes.
Steven Pemberton: [IRC] I have now
forwarded all relevant messages to the DB.
Steven Pemberton: I've created a
bucket for every toplevel section and am moving them into those
sections.
John Boyer: I see what you mean
about the "considered."
Leigh Klotz: We don't say that the
instance data is copied, but we do say "remove."
John Boyer: We can say "ignored." I'd
prefer to say selected above. There's something wrong...we're
trying to transition from the node you selected to the subtree that
it's the root of, all of those nodes. This is hard.
Leigh Klotz: It implies a copy.
John Boyer: Yes, it does. We need a
first sentence to jump from selecting the node to selecting the
subtree. We can just say, are considered. What's a better
word?
Uli Lissé: selected
John Boyer: ...node is
removed...
Leigh Klotz: deselected. It's kind of
a MIP.
John Boyer: OK. Now, if relevant is
true and the root element has been (pruned)...
Leigh Klotz: OK.
John Boyer: Now if you prune the root
element node, we fail the submission? Actually, can you have a
non-relevant node and have any of its children be relevant. So if
the root element is non-relevant then the submission will be
empty?
Leigh Klotz: The behavior as is will
not submit.
John Boyer: So why?
Nick van den Bleeken: We don't want to
submit empty XML serializations.
John Boyer: OK, so we check the root
node after relevance pruning. If any node is selected, then at
least the root node must have been selected.
Leigh Klotz: Can we safely preserve
this invariant by testing after the pruning boundary? Yes, because
if at least one node is selected, then the root node must be
selected, and if the root node was irrelevant then exactly no nodes
will be selected.
Uli Lissé: Selected is
over-used in bullet point two.
John Boyer: [reads]
Leigh Klotz: Which attributes? ref
and?
Nick van den Bleeken: bind
Uli Lissé: Binding
attributes
John Boyer: The binding attribute of
submission indicate a node of instance data...
Keith Wells: Why does it say "for
which it is an ancestor" instead of "descendent?"
John Boyer: Because in XPath the
attribute notes are not descendents but the children are.
John Boyer: Ok, one last time. The
diff marked version shows substantive changes. It's a struggle
here...
Leigh Klotz: Why don't we say "if no
instance nodes are selected"?
John Boyer: To bind more tightly to
the pruning.
Leigh Klotz: Can we move the error
dispatch to one place? I guess not because we need an error on
binding failure. Can we change the binding failure to a binding
exception?
John Boyer: We want to preserve the
two ways that submission can end.
Leigh Klotz: In XForm 1.1 I guess
xforms-submit-error is useful; it's not in 1.0 because there's no
context info.
John Boyer: You can put up an error
message saying "try again or contact tech support."
John Boyer: OK, I'll put this in CVS
for Mark.
http://www.w3.org/MarkUp/Forms/specs/XForms1.1/index-diff.html#submit-submission-element
Leigh Klotz: What's our next
step?
John Boyer: We compose an
issue-tracker reply and Steven needs to send it.
Steven Pemberton: Let me get
there.
Uli Lissé: It's also issue
43.
Steven Pemberton: That's a different
one. Oh, it changed URI.
Charlie Wiecha: Maybe we shouldn't
process them in Incoming.
Steven Pemberton: Right. So we
accept.
John Boyer: We did Modify and Accept;
we had to change section submission processing.
Steven Pemberton: What didn't we do?
Leigh We did more.
John Boyer: Then that's accept, as we
accepted his position.
Resolution 2007-06-13.1: For http://htmlwg.mn.aptest.com/cgi-bin/xforms-issues/Submission?id=1 We accepted the new defaults for relevant and validate and we changed the xforms-submit-event default processing. http://www.w3.org/MarkUp/Forms/specs/XForms1.1/index-diff.html#submit-submission-element and http://www.w3.org/MarkUp/Forms/specs/XForms1.1/index-diff.html#submit-evt-submit bullets 2 and 3.
John Boyer: OK next item.
Nick van den Bleeken: Can we go to the
next item in the issue tracking system?
Steven Pemberton: I haven't finished
categorizing them.
John Boyer: We should process them by
date.
Steven Pemberton: I have this one in
twice.
John Boyer: It's a duplicate.
Steven Pemberton: I'll make a box
called duplicate.
John Boyer: How do we handle
duplicate?
Steven Pemberton: It's a
classification for our purposes. We still have to give it a state.
That was issue 2 in model.
John Boyer: We probably added his 05
Dec email. It was on a telecon and I wanted to make sure it showed
up during last call.
Steven Pemberton: I'm just going to
put it in trash.
Leigh Klotz: Same issue, same person,
trash. Same issue, different person, requires two responses because
they might disagree.
Steven Pemberton: Right.
John Boyer: Can you ask for approved
but not implemented? So it's approved, we're going to get rid of
it. Do we have to implement it right now?
Leigh Klotz: There are two
references, plus section 3.4.
John Boyer: I will remove them.
Uli Lissé: Plus schema.
Action 2007-06-13.1: Mark Birbeck to remove mustUnderstand from XForms 1.1 Schema.
John Boyer: The mustUnderstand
section will still remain in the TOC for the non-diff-marked
version and then we remove it after last call.
Leigh Klotz: You can just diff mark
the content instead of the section and do a diff add that says
"this section removed".
John Boyer: The XML spec DTD doesn't
allow nested divs.
Action 2007-06-13.2: John Boyer to manually remove empty mustUnderstand section for XForms CR 1.1 after automatically stripping diff markup.
Resolution 2007-06-13.2: We implemented http://lists.w3.org/Archives/Public/www-forms-editor/2007Mar/0002.html by removing mustUnderstand module and references to it. XML Schema changes are pending.
John Boyer: David says that you
shouldn't be allowed to set readonly=false.
Leigh Klotz: I believe it was Mark who
had an objection to this.
Mark Birbeck: I think I do but I
accept I might not quite understand David's objection. ... It seems
to me fairly safe no matter when the calculation is
performed.
John Boyer: I've used much the same
logic myself. Can we follow this into the user experience? When the
form first comes up, the data node will be blank and set to some
nonempty value. When the user types, deletes the text, and hits tab
to commit the value, the UI binding pushes an empty string into the
data node; does the calculate run again?
Mark Birbeck: Good question; no it
doesn't.
Uli Lissé: Why?
John Boyer: The node gets on the
change list for recalculate dependencies but it doesn't get
recalculated.
Mark Birbeck: Right, it doesn't.
John Boyer: The note at the bottom
http://www.w3.org/TR/xforms11/#evt-recalculate
says that when a recalculate event is...must calculate computes
nodes in l and nodes dependent on l. So the
calculate does run.
Mark Birbeck: Good.
John Boyer: So that's good or not
depending on your point of view.
Charlie Wiecha: It replaces it unless
you clear it. It's like taxcut; if you clear it, it goes back to
default; otherwise stays.
Leigh Klotz: That's not this example
from David which is calculate=1.
Charlie Wiecha: No, it's Mark's
example.
Charlie Wiecha: There was also some
cross-instance issue from David.
John Boyer: I'm having trouble seeing
the problem with this case. We run all calculates on startup.
That's why that part of the spec says l and things
dependent on l. On rebuild we throw everything on
l. I don't think that's a problem. If the only dependency
is itself, then it's only going to run whenever there is a rebuild
and when somebody modifies the node to which the calculate is
bound.
Leigh Klotz: He's implying that an
insert does a rebuild of the whole model.
John Boyer: Insert does a rebuild of
the whole model.
Leigh Klotz: Then an insert in another
instance will cause the calculate to re-run.
Charlie Wiecha: That may be
fine.
John Boyer: I don't think you can do
that kind of isolation.
John Boyer: We should say explicitly
that it's ok to change readonly. This is a modify and accept. We
could agree with David that the spec needs a note, but we think it
actually operates the other way; you are allowed to set
readonly=false() on a calculated node and we can say a
few things about it.
Mark Birbeck: In 1.0 spec 4.3.6, we
have a note that says that authors should not refer to the current
node in calculate expressions.
John Boyer: I don't have a problem,
because it says "should" instead of "must." (RFC2119) I think there
is some concern that other than Appendix D issues, such as
side-effects .=.+1. We wouldn't be able to get interoperable
behavior. "should" means you should have a decent reason; setting a
default value is a decent reason. Maybe that's where we put that
in.
Leigh Klotz: Should we make that
edit?
John Boyer: After the break.
John Boyer: The "formal objection"
to Last Call isn't part of the process. The earliest time for that
is CR.
Steven Pemberton: I agree. The
objection was based on the fact that we had not replied to earlier
issues; we have now done that.
John Boyer: We fished out the issues
at the bottom of the referenced emails and addressed them.
Leigh Klotz: The HTTP authentication
one was addressed here:
http://lists.w3.org/Archives/Public/www-forms-editor/2007Apr/0050.html
John Boyer: Jan also replied to one of
them.
Nick van den Bleeken: Test suite.
http://lists.w3.org/Archives/Public/www-forms-editor/2007Apr/0056.html
John Boyer: Here is my reply on QName,
which we fixed in XForms 1.0 Errata, and 1.1 as well.
http://lists.w3.org/Archives/Public/www-forms-editor/2007Apr/0055.html
Resolution 2007-06-13.3: We have closed all outstanding technical issues from Bjoern Hermann. We received no response on our final messages.
John Boyer: Back to this, we should
say the effect is not defined "in all cases."
Uli Lissé: Then we have to
define the cases.
Steven Pemberton: ...
John Boyer: It's not going to
crash.
Uli Lissé: What condition
John Boyer: calculate=if(sometest(),
., .+1). Appendix D doesn't say because it's not clear. There are
suboptimal algorithms that don't give the same effect. The feeling
at the time is that we wanted people to do alternate
algorithms.
Leigh Klotz: By algorithm do you mean
topological sort or additional stuff?
John Boyer: The topological
sort.
Leigh Klotz: I think there is no
implementation that does not do topological and yet achieves the
same result. Mark's implementation and Orbeon don't do topological
sort and obtain different answers depending on document
order.
Mark Birbeck: One of ours does, and
one doesn't use topological sort. The one that doesn't gets
different answers. You were talking about the recalculation
sequence algorithm. But the prior step says that implementations
must reduce the number of vertices, and I read that to mean that we
wouldn't recalculate this node.
John Boyer: If you look at the total
shape of the MDG, there are nodes not reachable from the set of
nodes l, but that original set of nodes is part of the
pertinent subgraph. So take the set of nodes fed to the
recalculation engine and use those as the roots of directed
subgraphs and work out the closure of those nodes on the graph;
that will often be smaller than the whole graph.
Mark Birbeck: I agree, but if you use
the node as the origin that has a calculation based on itself, then
I don't think that would give the subgraph. The pruning step would
remove the recalculation of that node. So you asked if that node
would be recalculated if the user changes it to space (empty). You
said yes, I said no. It looks to me now like it wouldn't.
John Boyer: We may have to change
words to make it clearer: the other part of the spec says what
happens; this part of the spec is subject to 4.3.6. We have to
change this to make it do what 4.3.6 says.
Mark Birbeck: OK.
Leigh Klotz: Is this the section that
says you remove nodes that depend on themselves?
John Boyer: You remove the edges that
point to themselves, not the node itself. If the appendix does say
what you described, then it needs work. The usual transitive
closure definition; a node is reachable from itself by having an
edge that goes from that node to itself.
Mark Birbeck: How do you get to the
compute node?
John Boyer: The vertex that represents
the value of a node is the same one as the one that the calculate
is associated with; there aren't two separate vertices, which has
been a problem in the past. The calculate MIP is not different from
the value of the node in the MDG. The vertex may represent a
readonly, a constraint, or the value of a node, where that value
could be calculated.
Mark Birbeck: I don't see that.
John Boyer: The last paragrph of
section D1 (1.0) C1 (1.1) says..."a vertex exists for each instance
node to represent the expression in the context of the node." So,
we are trying to say that we create vertices; if you have a
calculate that binds to 20 nodes you'll get 20 vertices.
Mark Birbeck: Yeah.
John Boyer: Those 20 vertices are
representative of the text content and are associated with the
expression as well. Maybe we need to take another look at the
appendix to make sure it says the correct things to say what
section 4.3.6 says; if there's some problem. But we did a lot of
work on section 4.3.6 to make sure it includes in the recalculation
the nodes that are in the list l.
Mark Birbeck: If you look at C4, in
the diagram, the only things that can be in the last are A and B. C
is recalculated because of a changed A or B. So you are saying that
if C were to appear in the list it would be recalculated. But how
can it get in the list?
John Boyer: If you set readonly false
and change it.
Mark Birbeck: So if v=a*c I don't see
how we get an edge.
John Boyer: I need to read this; I
don't know why I used C and D; it's been a long time. I see; v and
w are representing the calculated value and the validity property
of an instance node c. If you bind an input control to node c, and
you set readonly=false on node c, then you should be able to cause
the vertex labeled v to enter the changelist that is input to the
recalculate, which is exactly the issue here.
Mark Birbeck: I see. I hadn't spotted
that in our implementation. That would be a vertex for the
calculate and a separate vertext for c if it appeared anywhere
else.
John Boyer: V is representative of the
...
Mark Birbeck: There is a path; it's
being used in w. I think you're right. I don't think we've got that
extra step. V represents C, essentially. We've got a possibility
where C could be in the list independently of V.
John Boyer: Yes, David had that a
couple of years ago with separate vertices for the text content of
the node and the calculated value of the node.
Mark Birbeck: Yes.
John Boyer: That caused a certain
amount of grief.
Leigh Klotz: What's the basis of the
objection now?
Nick van den Bleeken: It says they are
explicitly ignored.
John Boyer: When you are forming a
dependency list for a node; it never lists itself in its
dependencies. You could create one by doing <bind nodeset="c"
calculate=".+1"/> We remove those. The rebuild sequence starts
with a list l that we push down to recalculate. The
recalculate algorithm picks up the list l and runs them
all, including the self-reference calculates which technically have
to dependencies.
Leigh Klotz: So the root nodes
l are included in the result even if they have no entries
in the MDG.
John Boyer: The closure includes the
node you started with. It's bullet point 2, in C1.
Leigh Klotz: But there are no nodes in
the MDG for them?
John Boyer: Yes, there are, because
they have values, particularly when you put a calculate on
them.
John Boyer: Back to the note in 4.3.6:
do we want to take it out?
Leigh Klotz: Should we say don't do
.=.+1
John Boyer: Or don't have
side-effects, but that's not well defined.
Leigh Klotz: It's an informative note
so let's take out the prescriptive stuff and just put in an example
of what might be counterintuitive and say that <bind nodeset="c"
calculate=".+1"/> will have effects that may vary by
implementation.
John Boyer: OK [types]. That's half of
it, then.
node's value in a calculate on the node, as in the following example, may have effects that vary by implementation: <bind nodeset="x" calculate=".+1"/> Model item properties other than calculate, such as required or readonly are well-defined in the presence of self-references.
Roger Perez: So it's not a
loop?
Leigh Klotz: No, it's never a loop.
But some implementations may do calculations more frequently, so
this expression will increment more often in some
implementations.
John Boyer: We have to look at
6.1.2, the readonly property. It says the default value is
specified. David said it wasn't clear. So a note that says you can
explicitly set the value to false.
Nick van den Bleeken: David thought
you couldn't change it.
John Boyer: Why does he think
that?
Nick van den Bleeken: Becuase of the
problems with calculate he believes.
John Boyer: We're trying to acknowldge
that the spec could be clearer. We're going to say that we don't
agree that you can't set readonly=false on a calculate node. There
are cases where if you use a self-referential value and the
processor uses an algorithm other than topological sort that you
can get other effects, but we haven't seen any conformant
implementations not using topological sort. The ones that don't use
it now don't achieve the same effects in the other cases, so they
are non-conformant.
Leigh Klotz: I think that Orbeon is
moving to using topological sort.
John Boyer: So now we put in a note
showing readonly=false on calculate.
Leigh Klotz: How about an
example?
John Boyer: The default value
example.
Leigh Klotz: And in XForms 1.2 we can
put in a default attribute.
John Boyer: Yes. [adds example to
readonly section]
Resolution 2007-06-13.4: For http://htmlwg.mn.aptest.com/cgi-bin/xforms-issues/MIPs?id=45 we accept with modification; we find that calculate merely defaults readonly to true, and that it can be set to false, and that there are use cases, namely default value. We tested the use case and found it works. We changed the note in 4.3.6 [ http://www.w3.org/MarkUp/Forms/specs/XForms1.1/index-diff.html#evt-recalculate] and put an example in MIP for readonly [ http://www.w3.org/MarkUp/Forms/specs/XForms1.1/index-diff.html#model-prop-readOnly]
Nick van den Bleeken: Here is my
response:
http://lists.w3.org/Archives/Public/public-forms/2007Jun/0000.html
Nick van den Bleeken: Here is the link
to the minutes:
http://lists.w3.org/Archives/Member/w3c-forms/2007JanMar/att-0107/20070214.html#resolution1
Nick van den Bleeken: Here's the
proposal: http://www.w3.org/MarkUp/Forms/wiki/Sorting_data
Leigh Klotz: I thought Kenneth said
that collation was problematic because of the number of codepoints
for small devices.
Nick van den Bleeken: It's not
mandatory. XQuery says the number of collations is implementation
dependent.
John Boyer: XForms Basic could
implement only Unicode code points [http://www.w3.org/2005/xpath-functions/collation/codepoint].
Leigh Klotz: Do we need to specify the
profile?
John Boyer: Let's read the cited
http://www.w3.org/TR/xquery-operators/#collations
Leigh Klotz: We need to say which ones
are supported.
John Boyer: One sentence on Unicode
then.
Leigh Klotz: Who defines these
collations? The relative namespace URI (really a URN) that isn't
dereferenced is odd.
John Boyer: That's their issue.
Leigh Klotz: We might not want to buy
into it.
John Boyer: We can throw a compute
exception if it's not defined.
Leigh Klotz: What's their error.
John Boyer: [reads] "an error is
raised"
Leigh Klotz: What's this about the
"default collation for the static context?"
John Boyer: We don't have a way to do
that.
Leigh Klotz: So we should say it's the
Unicode point and drop the third argument since it drags in a lot
of questions from XPath 2.0 which we don't have. If you want the
third argument, it's underspecified and we can't define
interoperability for it in XForms 1.1. Implementations can still
take the 3rd arg, but it will be an extension out-of-spec, just as
some implementations are currently using XPath 2.0 already
anyway.
John Boyer: OK. Let's try that. Do we
mention the collation URI? Can we just describe the behavior?
Leigh Klotz: We either reference the
XPath 2.0 spec normatively and profile it by collation URI or
describe it and mention XPath 2.0 upwards compatibility in a Note
with an informative reference to http://www.unicode.org/unicode/reports/tr10t
.
John Boyer: We don't need a note,
though. We've just done the work of the compatibility.
Resolution 2007-06-13.5: We implement http://htmlwg.mn.aptest.com/cgi-bin/xforms-issues/XPath?id=44 by a substantive change, adding a sort function that uses Unicode collation http://htmlwg.mn.aptest.com/cgi-bin/xforms-issues/XPath?id=44
Action 2007-06-13.3: Nick van den Bleeken to read http://www.unicode.org/unicode/reports/tr10t and assure that we are compatible.
Uli Lissé: Is this going to
www-forms-editor automatically?
Steven Pemberton: It looks like
not.
John Boyer: Is there a re-send?
Nick van den Bleeken: Yes, it sends
immediately without confirmation.
Uli Lissé: I cc'd www-forms and
www-forms-editor.
John Boyer: They're all coming from
www-forms-editor so replies ought to go there without you having to
CC them.
Leigh Klotz: We don't have the replies
yet at all so we don't know.
Steven Pemberton: I suspect the
replies are queued up somewhere. Ah, they've been spam-blocked.
Steven Pemberton: This was entered
in as one issue.
John Boyer: There are six.
John Boyer: Issue 1 is a
misunderstanding; it applies to XForms 1.0 and was corrected in
XForms 1.0 TE. We mention it in the id() function. Is it worth
pointing out that xml:id works for us? We mention it only in the
id() function.
Steven Pemberton: I think we should
allow id or xml:id, but of course not both at the same time.
John Boyer: We can use either as long
as they are type id. We could make note of the fact that xml:id is
allowed or you can use any attribute of type ID.
Leigh Klotz: The questions is where do
we put it in the text?
John Boyer: Right after id; move id
after anyAttribute for sorting anyway.
Resolution 2007-06-13.6: We add a note to Common Attributes: "Elements can be identified using any attribute of type ID (such as xml:id), not just the id attribute defined above."
Leigh Klotz: It's probably left
over from thin-spec.
John Boyer: I guess they mean a
citation.
John Boyer: We refer to it from tbe
abstract, and it's in the Normative references section; they want
informative references.
Resolution 2007-06-13.7: We move XForms 1.0 to an informative reference.
John Boyer: Right now we have these
two datatypes that we have defined in XForms 1.0 to allow empty
content. They already allow empty content in XQuery 1.0 and XSLT
2.0 but we don't have access to them.
Leigh Klotz: How did they put them in
the XML Schema namespace?
Resolution 2007-06-13.8: We cannot move xforms:dayTimeDuration and xforms:yearMonthDuration to the xs: types from XQuery 1.0 / XSLT 2.0 because XForms 1.1 is based on XML Schema 1.0 and XPath 1.0.
John Boyer: "xsd suggests the XML
Schema datatypes namespace." In XML Schema 1.0 there's no XML
Schema datatypes namespace. If you look in the Schema 1.0 Par 0
primer and Schema Part 2 Datatypes, both use xsd: as the prefix for
the XML Schema namespace. So both documents have done the same
thing we do. Only in XML Schema Part 1 do they use xs. Pretty much
all the time we're using xsd.
Uli Lissé: xsd has its own
namespace. http://www.w3.org/TR/xmlschema-2/#namespaces
John Boyer: In the po.xsd example in
XML Schema part 0 they use
xmlns:xsd="http://www.w3.org/2001/XMLSchema". How does it
work?
Leigh Klotz: Your XML Schema processor
is supposed to deal with it.
John Boyer: So I didn't know.
Leigh Klotz: It's almost always xsd we
want so it would be wrong to change the prefix throughout.
John Boyer: Nowhere in Part 2 does it
ever define xmlns:xsd.
Leigh Klotz: So where we say
xsd:schema that's wrong, I guess. But I see plenty of examples both
in XML Schema Part 0 and elsewhere that use xsd: to refer to the
XMLSchema namespace, not the XMLSchema-datatypes namespace, so I
think we should keep xsd: as is, but add xs: and use that for
xs:schema and its child elements. In the reponse we can
explain.
John Boyer: And we add a definition of
xs: at the beginning.
Resolution 2007-06-13.9: We change xsd:schema and children to xs:schema and children, and add a definition of the xs: namespace, but leave occurrences the definition xmlns:xsd="http://www.w3.org/2001/XMLSchema" and the uses on datatypes, because that's what XML Schema Part 0 does. Implementors will already note that XML Schema requires handling of type definitions in the xmlns:xsd="http://www.w3.org/2001/XMLSchema-datatypes" namespace without special instruction from XForms.
ACTION:Mark Birbeck to update XForms Schema to move xsd:schema to xs:schema (prefix change)Steven Pemberton: Could we replace
this with a reference?
John Boyer: Sure, what
reference.
Leigh Klotz: Should we ask Martin Dürst? He
gave us the list.
John Boyer: They are taken from
version 3.2 of the Unicode standard. Someone said we should be
using Unicode 5.0.
Steven Pemberton: Martin says
5.00.
John Boyer: They've edited their spec
in place so we just continue to refer to it from
http://www.w3.org/MarkUp/Forms/specs/XForms1.1/index-diff.html#ref-unicode-scripts
and get Unicode 5.
Steven Pemberton: So we can take out
our list and use http://www.unicode.org/Public/5.0.0/ucd/PropertyValueAliases.txt
John Boyer: So we lose Braille and a
couple of other ones.
Steven Pemberton: No, it's there; it's
Brai.
John Boyer: We have
braille. And we have canadianAboriginal.
Steven Pemberton: Tokens are
case-sensitive.
Steven Pemberton: I'll just ask Martin
for suggested replacement text.
John Boyer: That sounds good.
Leigh Klotz: What about if? We could
introduce namespaced versions of all functions.
John Boyer: That would be cool. Then
people using XPath 2.0 could use those namespaces and it would work
in XForms 1.1 processors.
Nick van den Bleeken: Can't we just
say that they are the default namespace?
Leigh Klotz: XPath doesn't have
default namespaces.
John Boyer: XPath 1.0 doesn't have
namespaces on functions at all; people just did them. If we said
the default namespace is XForms then the XPath functions don't
exist anymore.
Resolution 2007-06-13.10: We agree that XForms-specific procedures should be harmonized with XPath 2.0 functions library where possible.
John Boyer: It seems we need
someone else to write the text for the Luhn function and make the
ISO/IEC 7912-1 an informative reference. I don't want to use the
Wikipedia reference they suggest as normative as it's
volatile.
Leigh Klotz: The Wikipedia article
mentions the expired patent: "U.S. PATENT 2,950,048 Computer for
Verifying Numbers.", H.P. Luhn Let's just use the reference to the
now-expired patent as the normative reference and call it "luhn
algorithm" in the spec, so that web searches will take you to the
Wikipedia entry.
Steven Pemberton: There's a later
issue that says let's not call it Luhn.
Leigh Klotz: It's not a function
either, in the patent.
Resolution 2007-06-13.11: We change the Luhn reference from ISO/IEC 7912-1 to expired patent "U.S. PATENT 2,950,048 Computer for Verifying Numbers.", H.P. Luhn. We change terminology from "Luhn formula" to "Luhn algorithm" to make finding the Wikipedia entry easy. We retain "Luhn function" for now.
Uli Lissé: So accept or
modify and accept?
Leigh Klotz: We did the schema thing
slightly differently from the way they wanted.
John Boyer: No substantive
changes.
Nick van den Bleeken: Not even the
schema namespace change?
Leigh Klotz: No. It's just a prefix
spelling.
John Boyer: We can break up into
groups and work on April's issues
Leigh Klotz: Pairs, one each with
write access.
Charlie Wiecha: I haven't found any
that work out as easy.
Keith Wells: If we break out into
pairs we will come up with "trumpable" decisions. I feel that we
had a lot of progress and we're about to get into easier
ones.
Leigh Klotz: I could go either way;
I've been busy so it's not an issue for me.
Keith Wells: I feel we won't make as
much progress. If anybody feels they can go tackle an issue, feel
free I guess.
Resolution 2007-06-13.12: We do not break into groups (yet).
John Boyer: We got through March;
maybe we can break through April tomorrow.
Keith Wells: Like Aaron's.
John Boyer: Someone could break out
and find that most of those are editorial.
Charlie Wiecha: I'll do that.
Action 2007-06-13.4: Charlie Wiecha to read Aaron's three last-call submissions and pull out the points that are not merely editorial and require group attention during the F2F.
John Boyer: We don't refer to the
requirements at all. We could just delete it.
Steven Pemberton: Or put it
normative.
John Boyer: Or remove it.
Resolution 2007-06-13.13: We remove the XForms 1.0 requirements and do not add XForms 1.1.
Steven Pemberton:
Internationalization Tag Set. It's a tagset for marking up
documents about how they are to be translated.
John Boyer: Do translate label, don't
translate item. It's somewhat hopeless if you can use ref on a
label.
Leigh Klotz: The instance itself has
its own idea, and it's the problem of whoever made the
instance.
John Boyer: And if the instance is
inline? This isn't normative, right?
Steven Pemberton: We might be happy to
do this after last call.
Leigh Klotz: You mean 1.2? After last
call?
Steven Pemberton: It's a host
language. Oh wait, they want a rules file.
John Boyer: It's like a Schema.
Leigh Klotz: They want a Note?
John Boyer: Yes, it would have to be
afterwards. Maybe they could make an example.
Leigh Klotz: By the way to tlink from
http://www.w3.org/205/11/its
links to their CR version. They are at REC now.
Leigh Klotz: I'm happy to work with
them on this.
Uli Lissé: What is the
state?
Steven Pemberton: Defer; Closed. We're
not saying no; we're willing to discuss it.
Resolution 2007-06-13.14: An ITS rule file is not a normative part of the recommendation and we would be happy to provide some meeting time to help I18N WG create it, but we do not have the resources to create it.
John Boyer: The XML example file is
corrupted.
Leigh Klotz: Use an entity reference
so it doesn't get corrupted again. http://lists.w3.org/Archives/Public/www-xsl-fo/2001Oct/0017.html
says á
John Boyer: 225 is wrong; that's an
a.
Steven Pemberton: It's 233. But
actually, since it's Italian, not French, it's "latte" with no
accent.
Resolution 2007-06-13.15: We accept the correction to XForms and SVG Example by removing the accent mark, as the word "latte" is Italian, not French, so it's "latte" with no accent.
John Boyer: They don't like
compound documents?
Leigh Klotz: I wrote this example. He
says it's not a good authoring practice, but this might not be
authored; it could have been created by a program using ITS.
Steven Pemberton: The reader is being
confronted with extra detail. Also, it's using an old XHTML2
URI.
Keith Wells: I agree with Steven; I
believe the language complicates the example.
John Boyer: The choices is an issue as
well.
Steven Pemberton: Then we need to
reply saying we'll re-write it.
Leigh Klotz: Or remove it. If we
simplify it it may still have problems.
Steven Pemberton: Why not take an
existing XHTML example.
John Boyer: We lose the Schema
then.
Steven Pemberton: So we will replace
it with a simpler example that does not use multi-lingual
text.
Leigh Klotz: I still don't agree that
with the point that we shouldn't create XForms documents with
multiple languages in them. They might not be authored that way but
it should be OK to present them.
Steven Pemberton: It's not necessary
for the response.
Resolution 2007-06-13.16: We re-write and simplify G.1 XForms in XHTML to be in one language, and move to XHTML1.