- From: Ian B. Jacobs <ij@w3.org>
- Date: Mon, 11 Mar 2002 19:53:04 -0500
- To: www-tag@w3.org
[This format is a proposed compromise for meeting summaries:
participation, agenda items, action item summary, IRC log.
- Ian]
All present: Tim Berners-Lee (TBL), Tim Bray (TB), Dan Connolly
(DC), Paul Cotton (PC), Roy Fielding (RF), Chris Lilley (CL),
David Orchard (DO), Norm Walsh (NW), Stuart Williams (SW), Ian
Jacobs (IJ)
Agenda items:
* Action item review (below)
* Issue mixedNamespaceMeaning-13:
"What is the meaning of a document composed of content
in mixed namespaces?"
Agenda: http://www.w3.org/2002/03/11-tag.html
IRC log: http://www.w3.org/2002/03/11-tagmem-irc.html
[Quoted as text below]
Previous meeting: 4 March
http://lists.w3.org/Archives/Public/www-tag/2002Mar/0030
Next meeting: 18 March
------------------
Action item review
------------------
Completed:
TBL: Kick off discusiion of mixed namespace interpretation
on www-tag
http://lists.w3.org/Archives/Public/www-tag/2002Feb/0190
PC: Write summary of TAG activity to date
DC/IJ: Send PC summary of TAG work to AC.
PC: Summarize XMl PM workshop
DC: Find XML PM workshop minutes
Open:
RF: Summarize different approaches currently used for mapping
URIs to media types.
IJ: Integrate findings and issues into TAG arch doc toc.
DC: Check with workshop chair, make the XML PM ws minutes available
to the public, per CFP
One pages summaries for arch document, to be completed this
week:
DC, TB: Section 1
NW, SW: Section 2
NW, PC: Section 3
RF, DO: Section 4
CL, PC: Section 5
=========
IRC LOG:
Timestamps are in UTC.
15:16:17 [RRSAgent]
RRSAgent has joined #tagmem
15:16:23 [timdesk]
RRSAgent, pointer?
15:16:23 [timdesk]
See [1]http://www.w3.org/2002/03/11-tagmem-irc#T15-16-23
[1] http://www.w3.org/2002/03/11-tagmem-irc#T15-16-23
15:17:13 [timdesk]
Hmmm... I can't read that, RRSAGent.
15:17:29 [DanC]
tim bray's message about an intro document didn't get to my
inbox. Makes me v. nervous.
15:18:12 [timdesk]
Where was it addressed to?
15:19:26 [DanC]
tag@w3.org
15:19:27 [timdesk]
timdesk has changed the topic to:
[2]http://www.w3.org/2002/03/11-tag.html
[2] http://www.w3.org/2002/03/11-tag.html
15:20:01 [timdesk]
timdesk has changed the topic to: PUBLIC LOG. Agenda:
[3]http://www.w3.org/2002/03/11-tag.html
[3] http://www.w3.org/2002/03/11-tag.html
15:21:54 [timdesk]
The last messaeg I have in my inbox to the tag@ list is 3/5
15:22:34 [timdesk]
Sorrt 3/8 from Chris re(2) agnga ...
15:26:23 [TimBray]
TimBray has joined #tagmem
15:26:39 [TimBray]
hey all, what's the phone#/code again?
15:27:00 [DanC]
Zakim, what's the passcode?
15:27:01 [Zakim]
sorry, DanC, I don't know what conference this is
15:27:06 [DanC]
Zakim, this will be tag
15:27:07 [Zakim]
ok, DanC
15:27:10 [DanC]
Zakim, what's the passcode?
15:27:11 [Zakim]
sorry, DanC, I don't know what conference this is
15:27:16 [TimBray]
:)
15:27:18 [DanC]
phpht
15:27:28 [DanC]
the agenda should say...
15:27:33 [TimBray]
oh right
15:27:35 [Zakim]
TAG_Weekly()10:30AM has now started
15:27:37 [Zakim]
+TimBL
15:27:54 [DanC]
timdesk, [4]http://www.w3.org/2002/03/11-tag.html is not an
agenda; W3C process says the phone number goes in the agenda.
[4] http://www.w3.org/2002/03/11-tag.html
15:28:34 [Norm]
Norm has joined #tagmem
15:28:57 [timdesk]
oops
15:29:08 [DanC]
er... oops; I thought it did. but I can't confirm in
[5]http://www.w3.org/Consortium/Process-20010719/groups.html#Ge
neralMeetings
[5]
http://www.w3.org/Consortium/Process-20010719/groups.html#GeneralMeetings
15:29:09 [timdesk]
Its not schema valid.
15:29:11 [Zakim]
+N.Walsh
15:29:12 [Zakim]
-N.Walsh
15:29:58 [Zakim]
+??P7
15:30:06 [DanC]
Zakim, what's the passcode?
15:30:07 [Zakim]
the conference code is 0824, DanC
15:30:44 [Stuart]
Stuart has joined #tagmem
15:31:02 [Roy]
Roy has joined #tagmem
15:31:30 [Zakim]
+DanC
15:31:30 [Zakim]
-DanC
15:31:31 [Zakim]
+DanC
15:31:51 [timdesk]
Zakim, what is the passcode?
15:31:52 [Zakim]
the conference code is 0824, timdesk
15:32:03 [TimBL]
Zakim, whois here?
15:32:03 [Zakim]
I don't understand your question, TimBL.
15:32:07 [TimBL]
Zakim, who is here?
15:32:08 [Zakim]
I see TimBL, ??P7, DanC
15:32:22 [TimBL]
Zakim, ??P7 is Roy
15:32:23 [Zakim]
+Roy; got it
15:32:44 [Zakim]
+??P8
15:32:48 [Zakim]
+TBray
15:33:09 [Zakim]
+??P10
15:33:25 [TimBL]
Zakim, ??P8 is PaulC
15:33:26 [Zakim]
+PaulC; got it
15:33:33 [Stuart]
Zakim, +??P10 is me
15:33:35 [Zakim]
sorry, Stuart, I do not recognize a party named '+??P10'
15:33:44 [Stuart]
Zakim, ??P10 is me
15:33:45 [Zakim]
+Stuart; got it
15:35:01 [DanC]
I have a code in by dose.
15:35:10 [Ian]
Ian has joined #tagmem
15:35:22 [DanC]
dialing in, Ian?
15:36:06 [Zakim]
+Ian
15:36:31 [TimBL]
Zakim, who is here?
15:36:32 [Zakim]
I see TimBL, Roy, DanC, PaulC, TBray, Stuart, Ian
15:36:57 [Zakim]
+ChrisL
15:37:41 [TimBL]
[6]http://lists.w3.org/Archives/Member/tag/2002Mar/0021.html
[6] http://lists.w3.org/Archives/Member/tag/2002Mar/0021.html
15:37:52 [DanC]
[member confidential. :-{]
15:37:54 [TimBL]
Is a message from Tim Bray
15:38:46 [Ian]
NW: One-page summary , section 2. We have different approaches
(more/less technical).
15:38:47 [DanC]
I did read TimBray's pg
[7]http://www.textuality.com/tag/Intro.html . Good for software
folks; doesn't say anything for webmasters. doesn't say
anything about a "space".
[7] http://www.textuality.com/tag/Intro.html
15:38:51 [Ian]
...which way should it go?
15:38:51 [DanC]
s/anything/much/
15:38:56 [Ian]
TB: Should be correct but readable.
15:39:30 [Ian]
NW: SW and I would ilke to unify. There's a spectrum. We
entered at two different points.
15:39:43 [DanC]
I like friendly/chatty.
15:39:51 [Ian]
PC: One-page summary , section 3. Sitting on my desk now.
15:40:08 [Ian]
TBL: Don't hestitate to cc tag@w3.org if you are sending
messages back and forth.
15:40:17 [Ian]
One-page summary , section 4
15:40:26 [Ian]
RF: I'm going to send our current state during this call.
15:40:37 [Ian]
...I received last night and read through; haven't made
comments yet.
15:40:50 [Chris]
Chris has joined #tagmem
15:41:10 [Ian]
One-page summary , section 5
15:41:31 [Ian]
CL: I sent in some notes, but not yet discussed with PC.
15:42:24 [DanC]
if you're setting deadlines, tim, COB Fri doesn't help me at
all. COB Thu is the latest that's useful.
15:42:25 [Ian]
TBL: Goal - end of business Friday. Whatever you've got, please
sent to tag@w3.org, or www-tag if you wish (I'm happy for
www-tag).
15:43:16 [Ian]
DC: Close of biz Friday not good for me.
15:43:36 [Ian]
RF: I'm having trouble getting stuff done on weekend in
preparation of this meeting.
15:44:08 [Ian]
TBL: Please finish drafts by c.o.b. on Weds, then.
15:44:22 [Ian]
Action DC: check with workshop chair, make the XML PM ws
minutes available to the public, per CFP
15:44:24 [Ian]
DC: No progress.
15:45:04 [Zakim]
+DOrchard
15:45:11 [Ian]
Action IJ: Take over DC's action to send summary to AC.
15:45:14 [TimBL]
.Action Ian - take that from Dan - send paul's summary
15:45:18 [Ian]
See
[8]http://lists.w3.org/Archives/Member/tag/2002Mar/0002.html
[8] http://lists.w3.org/Archives/Member/tag/2002Mar/0002.html
15:45:57 [Ian]
Action IJ: Integrate findings into TAG arch doc toc.
15:46:00 [Ian]
TBL: I thought I had done that.
15:46:10 [Ian]
TB: Should be integrating issues as well as findings.
15:46:13 [Ian]
IJ: No progress.
15:46:13 [DanC]
TimBL, you're doing well to review the actions in the maximally
confusing order. ;-)
15:47:06 [Ian]
Action RF: Summarize different approaches currently used for
mapping URIs to media types.
15:47:08 [Ian]
RF: No progress.
15:47:18 [Ian]
--------------------------------
15:47:38 [Ian]
mixedNamespaceMeaning-13 : What is the meaning of a document
composed of content in mixed namespaces?
15:47:42 [Ian]
[9]http://www.w3.org/2001/tag/ilist#mixedNamespaceMeaning-13
[9] http://www.w3.org/2001/tag/ilist#mixedNamespaceMeaning-13
15:47:52 [Ian]
"The Interpretation of XML documents
15:47:52 [Ian]
"
15:47:55 [Ian]
[10]http://www.w3.org/DesignIssues/XML.html
[10] http://www.w3.org/DesignIssues/XML.html
15:47:56 [DanC]
the agenda includes "3/18" referring to a date; pls don't do
that. write 18 Mar or 2002-03-18 .
15:48:47 [Ian]
TBL summarizing: There is a set of things like embedded xslt,
xinclude, xquery, etc. These things elaborate in a particular
way how they are to be replaced.
15:49:01 [Ian]
...the meaning of the xml document should be that after
elaboration.
15:49:22 [Ian]
...processors should expect to see this type of content
anywhere in an xml document and be prepared to deal with it.
15:49:49 [Ian]
...there are certain architectural principles such as piece X
can replace itself but can't touch other parts of the XML tree.
15:50:54 [Ian]
PC: Please elucidate the encryption problem early in XML.html.
15:51:03 [Ian]
TB: DO has a rant on this as well.
15:51:28 [Ian]
TBL: We know how to say what an xhtml document is (spec +
schema).
15:51:41 [Chris]
This model does not seem to describe XSL-T processing very well
15:52:25 [Ian]
TBL: For XML Encryption, people have been running an entire
document through a processor. But when several of these
functional pieces are together, there is a big-ending type
problem.
15:52:33 [Ian]
TBL: When you treat as functions, there are no problems.
15:52:39 [Ian]
CL: Not "no problems" but "different problems".
15:52:48 [DanC]
I wonder if the "functions" view is consistent with the "stack
of SAX filters" implementation technique.
15:53:38 [Ian]
TBL: Suppose part of a document is encrypted. E.g., xhtml
document with a meaty bit in xml encryption.
15:54:19 [TimBray]
I assume there are use cases...this notion of partial
encryption seems like very poor security practice, but what do
I know?
15:54:21 [Ian]
...This is not a valid xhtml document. So what is it? The
proposed answer is: if you support xml encryption, then you
expect it anywhere, and should be able to decrypt in the right
setting (e.g., right keys). The "meaning" is after that
encryption.
15:54:31 [Ian]
s/encryption/decryption/
15:54:38 [DanC]
q+
15:54:47 [Chris]
I wonder how this functional approach affecxts things like
validation.
15:54:51 [Ian]
TBL: If you have an encrypted thing that contains an xinclude,
you must decrypt first before getting xinclude.
15:55:02 [Dave]
Dave has joined #tagmem
15:55:02 [Ian]
PC: When does schema validation get done.
15:55:17 [Dave]
+q
15:55:19 [TimBray]
q+
15:55:21 [Norm]
The functional way doesn't work for me either, just because
XInclude is inside xsl:template doesn't mean I want the
template fired before the xinclude
15:55:21 [Chris]
q+
15:55:23 [Ian]
TBL: Schema validation doesn't make sense until after the
functional expansion has been done.
15:55:23 [TimBray]
+q
15:55:24 [Dave]
q+
15:56:09 [Ian]
DC: How is this Web architecture? [DC groks expanding xml
encryption and finding xml includes in their place.]
15:56:17 [Ian]
TBL: Architecture based on sending information to each other.
15:56:50 [Ian]
PC to DC: We started talking about MIME types. TimBL started in
by saying that the namespace of the root element was very
important. Some of us argued against that (citing xslt
example).
15:57:00 [Ian]
....when you look at it you can't tell what type of document it
is.
15:57:05 [DanC]
"root element NS is important" ... important for what?
15:57:28 [Ian]
PC: So, this is "Web architecture" following logic that got us
to this question.
15:57:49 [TimBL]
q+ PC
15:57:54 [Ian]
TBL: The problem is when recipient goes through different
process than author expected.
15:58:03 [Ian]
DC: I still don't get why this is important.
15:58:08 [Chris]
ack DanC
15:58:12 [TimBL]
ack DanC
15:58:13 [Ian]
TBL: Do you think that defining what an xml document means is
important?
15:58:17 [Ian]
DC: Not in the general case.
15:58:30 [Ian]
TB: DC may be right. Does this problem go away after processing
model work starts?
15:58:52 [DanC]
q+
15:58:55 [Ian]
TB: On when to do schema validation: I could right a schema for
partially encrypted documents.
15:58:57 [Norm]
What TimBray said. We went around in circles over this at the
procmodel workshop: there is no right answer.
15:58:59 [DanC]
note to self: about XInclude and Schema
15:59:06 [Ian]
TBL to TB: You could, yes, but there are unanswered questions.
15:59:18 [Ian]
TB: This problem goes away, IMO, once we have a processing
model.
15:59:30 [Ian]
TBL: This is my contribution to the processing model debate:
[11]http://www.w3.org/DesignIssues/XML.html
[11] http://www.w3.org/DesignIssues/XML.html
15:59:32 [DanC]
Norm, TimBray, there's no
15:59:39 [DanC]
"processing mode" working group chartered presently.
15:59:40 [TimBL]
ack TimBray
15:59:52 [Ian]
TB: See pipeline proposal:
16:00:12 [TimBL]
ack Chris
16:00:16 [Ian]
TBL: Pipeline is half of processing model.
16:00:17 [DanC]
XML Pipeline Definition Language Version 1.0
16:00:17 [DanC]
28 February 2002, Norman Walsh, Eve Maler (XML Pipeline
Definition Language Submission)
16:00:20 [Norm]
q+
16:00:25 [DanC]
[12]http://www.w3.org/TR/xml-pipeline/
[12] http://www.w3.org/TR/xml-pipeline/
16:00:33 [Ian]
TB: TBL, you're hosed if you just get an xml document and don't
know all the namespaces.
16:00:52 [Ian]
CL: My problem with TBL's writing doesn't describe schema
processing or xslt processing.
16:01:38 [Ian]
TBL: My document doesn't tell you the order of processing.
16:01:46 [Ian]
CL: It does give one order, that's not universally applicable.
16:02:03 [DanC]
TimBL, when PC asked about order, you said "yes, you have to
decrypt before xinclude". I'm confused.
16:02:37 [Ian]
TBL: You can choose to elaborate functions or not. Depending on
who you are. But the meaning of the document is what you would
get if you expanded them all.
16:02:49 [TimBL]
ack Dave
16:02:55 [Ian]
TBL continuing: But it doesn't matter to me who does what, for
example.
16:03:18 [Ian]
DO: On the XML encryption issue: One of the tenets (or
invariants) of the Web (IMO) is that documents are
self-describing.
16:04:01 [Ian]
DO: You can encrypt portion of a message (e.g., SOAP) and XML
Encryption doesn't say what to do in this case. I suggested to
them that, anyone who encrypts should make some sort of change
to keep a document self-describing.
16:04:13 [Ian]
TBL: If you encrypt an entire document, there will be a mime
type for this.
16:04:27 [Ian]
...if you are talking about the case of subtrees....
16:04:52 [Ian]
TBL: What do you mean by self-describing? Do you mean that by
looking at mime headers you should know what to do?
16:04:53 [TimBray]
in DO's argument s/encryption/xinclude/ or s/encryption/xquery/
- point still holds
16:05:17 [Ian]
DO: If you "just look at a[n encrypted] document" and think
it's xml, the content type lies about what's really there.
16:05:37 [Ian]
...the first thing that has to happen is that there needs to be
a clue that somethign has been encrypted.
16:05:53 [Chris]
I note that this model means that you can never schema validate
anything but the fully-expanded result
16:05:57 [DanC]
q?
16:06:14 [Chris]
which means that any xml language for a function becomes
non-validatable
16:06:20 [Ian]
TBL: When you say the first thing that needs to be done is
decryption, that may not be the case.
16:06:21 [Chris]
q+
16:06:26 [DanC]
q-
16:06:40 [Ian]
TBL: Suppose that a browser does incremental rendering until
the viewport has been filled.
16:06:58 [Ian]
...the browser could choose to not retrieve images or decrypt
content unless absolutely necessary (lazy processing).
16:07:15 [Ian]
PC: I now understand DO's rant.
16:07:43 [Ian]
PC: DO is concerned about case when mime type says "foo/bar"
but you have to decrypt part of a document to get "foo/bar"
content.
16:08:03 [Chris]
The MIME type becomes an assertion of what will result after an
arbitrary amount of processing
16:08:11 [Dave]
q+
16:08:23 [TimBL]
q+
16:08:36 [Ian]
PC: I thought that the XML Processing workshop came to a
conclusion - that there is no generic processing model. There
are always counter examples.
16:09:17 [Ian]
PC: XSLT 2 will allow stuff after PSV
16:09:30 [Ian]
PC: I don't think the workshop was a failure; it showed lots of
counter-examples.
16:10:17 [Ian]
PC: People at workshop felt that XML Core WG would be a good
forum for this work.
16:10:51 [Chris]
q-
16:11:19 [DanC]
ack PC
16:11:22 [DanC]
ack Norm
16:11:23 [Ian]
NW: When I first read TBL's paper, I thought it specified a
processing order. I don't think it's possible to do in an
arbitrary order.
16:11:35 [Ian]
NW: I think that unless you have a private agreement, there's
no right answer in the general case.
16:11:40 [Ian]
+q
16:11:44 [Ian]
q+
16:11:57 [DanC]
ack Dave
16:12:02 [Chris]
it specifies an order by the order of the immediate children of
the root
16:12:04 [Ian]
DO: I suggested in a paper at the workshop that this work be
done in XML Core WG.
16:12:25 [Chris]
I agree with Tim that this means the porder can be altered,as
long as you can rewrite the instance
16:12:31 [Ian]
DO: I think there should be an explicit processing model for
how to process XML documents, and a language that describes it.
16:13:01 [Ian]
DO: For some applications, there may be an explicit processing
model. A minimalist suggestion is that we can say "For
document-oriented XML Dcouments, use the following order...."
16:13:22 [Ian]
...perhaps in machine-oriented realm, this may not be the right
way to do things.
16:13:48 [TimBray]
q+
16:14:38 [DanC]
per the agenda, we're on issue mixedNamespaceMeaning-13. lemme
swap that in...
[13]http://www.w3.org/2001/tag/ilist#mixedNamespaceMeaning-13
[13] http://www.w3.org/2001/tag/ilist#mixedNamespaceMeaning-13
16:14:54 [Dave]
lol
16:15:14 [Ian]
NW: TBL, you seem to be asserting that if you elaborate b
before a or in the other order, the results will be the same.
16:15:28 [Ian]
TBL: Yes, I am within the constraint that the elaboration has
no impact on the rest of the document.
16:16:26 [Ian]
NW: Suppose I build a TOC before or after processing XIncludes.
I might get 3 or 4 chapters depending on when I expect
XIncludes.
16:16:52 [TimBray]
you'll be happy to hear that I plan to try for some concluding
noises
16:17:01 [Ian]
q-
16:17:20 [TimBL]
There is a draft by Simon StLaurent aboyt putting the sum of
the nsamepsaces in the mime type - this addresses DavidO's
concern. This is a concern.
16:17:33 [Norm]
When people put X in their document, they *do* know what they
mean. But if you had me your document, *I* don't know what you
meant. You have to tell me.
16:17:40 [Chris]
Not clear that a list of namespaces helps too much
16:17:43 [Ian]
TB: I think that if you look at what came out of processing
model workshop, and our discussion here, it's hard to make
statements about processing that are wide in their
applicability.
16:18:08 [Ian]
TB: I think it's a good idea to define a processing model.
However, at the moment we don't have a formal way to write this
down.
16:18:10 [Chris]
q+
16:18:21 [DanC]
ack timbl
16:18:23 [DanC]
ack timbray
16:18:36 [DanC]
after chris
16:18:41 [Ian]
TB: Given that we only have human-readable prose. We might want
to start a processing model group that provides language
designers a way to write down their processing model.
16:19:34 [Ian]
TB: You can ignore my processing order, but I have the "right"
to express it.
16:19:42 [DanC]
q+
16:19:51 [Chris]
TimBL has just mnade my point
16:19:59 [Chris]
well, part of it
16:20:00 [DanC]
take you off the queue, then?
16:20:05 [Chris]
no
16:20:08 [DanC]
ok
16:20:09 [Ian]
TBL: In XML you sometimes get this "battle" between precedence
at the attribute level.
16:20:30 [DanC]
not to self: who made a promise that all XML namespaces can be
arbitrarily mixed?
16:20:58 [Dave]
q+
16:21:08 [DanC]
note to self: on CR exit criteria for schema/XINclude/etc.
16:21:18 [Ian]
TBL: We're shipping xslt and xinclude. People think they know
how to interpret these things. We will either get a de facto
clash (e.g., you can't use both at the same time) or there is a
simple processing model and the xml processing model is on too
high a plane.
16:21:42 [Ian]
CL: Subgroups are agreeing among themselves on processing
models, for a limited set of processing.
16:21:48 [Ian]
...localized architecture.
16:21:49 [TimBray]
TBray said: a processing model facility is good even if people
can write conflicting demands with it
16:21:54 [Norm]
q?
16:21:59 [DanC]
ack Chris
16:22:12 [Ian]
CL: It's not clear that we can generalize to all combinations.
We should still write down the localized parts.
16:22:52 [Ian]
TB to CL: Are you asserting that we should look with favor on
two groups getting together and saying "Always do A before B?"
16:23:09 [Ian]
CL: No, we should recognize that they are already doing this
and either help them or tell them not to.
16:23:38 [Ian]
DC: I asked at CR discussion which went first xinclude or
schema; people not interested in the answer at the time.
16:23:43 [Norm]
q+
16:23:48 [Norm]
ack danc
16:23:51 [Ian]
DC: ...the cat is out of the bag as far as I'm concerned.
16:24:11 [Ian]
DC: If we say now that you can mix our specs in any order and
get a consistent result, that's like returning 2 years in the
past. May not be easy.
16:24:16 [Ian]
q?
16:24:36 [Ian]
DO: The processing model comes up when you want to mix content.
16:24:49 [Ian]
....software is written to understand particular namespaces.
16:25:17 [Ian]
DO: ...the scale of things we could do is to define an explicit
processing model that could be contained within each particular
document; to give the author a way to say what they think the
proper order is.
16:25:23 [Ian]
DO: Another extreme is to do nothing.
16:25:37 [Chris]
I remember I wanted to praise the value of namespaces that do
not alter the infoset
16:25:39 [Ian]
DO: Slightly less than that is to write down instructions for
commonly used processing.
16:26:21 [DanC]
q?
16:26:26 [DanC]
ack Dave
16:26:53 [Chris]
The longer we leave it, the more existing practice we have to
break to enforce a single global architecture
16:26:54 [Ian]
NW: TBL said there were two possibilities: clash or simple
model. There's a third possibility - different communities will
get used to different processing models.
16:27:31 [Ian]
NW: ..and we will have to have ways to describe which model we
are using in which case.
16:27:57 [Chris]
q+
16:28:08 [Ian]
TBL: Can we make a default processing?
16:28:23 [TimBray]
q+
16:28:31 [Ian]
NW: Maybe xhtml browsers need to have a fixed processing order.
Not the same as saying that all applications need the same
processing order.
16:28:40 [Ian]
TB: Maybe there's an 80/20 point we can hit.
16:29:08 [Ian]
TB: Getting 80% right would be a huge boost to
interoperability.
16:29:15 [Norm]
q?
16:29:25 [Ian]
IJ: Is TB's document good enough as a default order?
16:29:36 [Chris]
q+ Paul
16:29:51 [Ian]
PC: I'd like to probe TBL's arch principle - functions that
have no side effects elsewhere in the tree.
16:30:24 [Ian]
PC: I hear TBL saying if you have this type of XML, then a
browser could choose which functions to elaborate at any given
time.
16:31:04 [Ian]
TBL: I was deliberately being independent of who is doing the
processing. But you are right, by not allowing side effects,
you allow independence.
16:31:23 [Ian]
PC: This means that this model works only for functions that
don't have side effects.
16:31:34 [Ian]
...which of our functions today don't have side effects?
16:31:49 [Ian]
PC: Norm gave a counter-example that adjusts the document by
putting in a table of contents.
16:32:05 [Ian]
TBL: The table of contents expands in place. But because it can
take input from outside the tree....
16:32:17 [Ian]
PC: XML Schema has side effects when there are referential
integrity constraints.
16:32:17 [DanC]
q?
16:32:32 [Ian]
TBL: I don't see XML Schema as a function.
16:32:36 [Chris]
q+
16:32:46 [Ian]
TBL: I see it as a thing that only applies once all functions
elaborated.
16:33:01 [Ian]
TB: I don't buy that. Some people will want to validate
documents with some encrypted parts.
16:33:24 [Ian]
PC: E.g., encrypted header in a SOAP message.
16:34:21 [Norm]
On the subject of schema ref-integrity constraints: I believe
they're tree-local. So they would actually work on a
tree-by-tree basis.
16:34:25 [Ian]
CL: On whether schemas should be applied at the end [@@scribe
missed comment@@]
16:34:34 [Dave]
q+
16:34:55 [Chris]
model implies you can never schema validate a function writen
in an xml syntax
16:35:10 [Chris]
because it dissapears before the schema processor ever sees it
16:35:33 [Ian]
TBL: I think we need to design a system where people can use
these functions anywhere in a document. But yes, you might want
to use schema techniques to detect that some particular piece
has been encrypted.
16:35:34 [DanC]
q+ Paul
16:35:36 [TimBray]
q+
16:35:48 [Ian]
DO: I could see in this processing model that you would want to
do two passes of schema validation.
16:36:14 [TimBray]
Here's an 80/20 point strawman:
16:36:22 [TimBray]
1. Decrypt the parts you are equipped to
16:36:26 [TimBray]
2. Do any Xinclude
16:36:34 [TimBray]
3. Run XSLT if required
16:36:47 [TimBray]
</ol>
16:37:12 [Ian]
[Time check]
16:37:37 [DanC]
summary (a) TimBL wants a "default -- what a document means"
(b) folks want stuff like a pipieline language (c) Chris would
be happy with a browser-only solution.
16:37:45 [Ian]
DC: We have an IETF teleconf today.
16:38:32 [Dave]
Ian, you missed my comment on trying for an 80/20 point that is
simple and generally human-oriented as a first step. And you
missed the same point on my earlier discussion
16:38:41 [Chris]
Chris would be happy with defined pipelines of localised scope
(for example, anm xml user agent scope, a SOAP scope...)
16:38:46 [Ian]
Thanks DO.
16:39:10 [Ian]
PC: Much of this is about the infoset. Even schema PSV doesn't
touch the infoset.
16:39:18 [TimBray]
4. Validate
16:39:26 [Ian]
DO: I will note that XML Schema gave extensive comments to
XInclude about what they wanted in infoset.
16:39:40 [Ian]
PC: I thought encryption and signing ran across infost.
16:39:50 [Ian]
DO: No, they are designed to use XPath.
16:39:58 [Dave]
xpath 1.0
16:40:03 [Ian]
TBL: Canonicalization model.
16:40:22 [Ian]
PC: When we want to sign things that depend on schema
attributes, then we'll have to pull away from xpath 1.0 data
model.
16:40:43 [Ian]
TBL: No, when you sign something, you sign the xpath model; pre
schema validation.
16:40:54 [TimBL]
I am very surprised that PaulC suggests that XIncldue does not
change the infoset.
16:41:01 [Norm]
Me too
16:41:20 [Ian]
PC: I find it interesting that we have a series of specs that
act like functions, and another series that doesn't behave like
this. Is this an architectural problem?
16:41:24 [TimBray]
yep
16:41:42 [Dave]
q+
16:42:25 [Ian]
TBL: The case of encryption and dsig - not defined on surface
syntax; defined on xpath, which those specs consider an
infoset. They are defined on an abstract syntax.
16:42:52 [DanC]
I think PC's point is about functions on infoset (i.e. abstract
syntax) vs. functions on surface syntax
16:43:20 [Ian]
TB parenthetically: Whatever model we take, I think we should
not limit modifications to infoset; you'll need to be able to
modify the syntax as well.
16:43:30 [DanC]
s/modify/exchange/
16:43:31 [Ian]
TB: If there's an 80/20 point, where is it? Is it as simple as
saying:
16:43:53 [Ian]
a) In the absence of indications to the contrary, first
decrypt, then process includes, then run xslt, then validate.
16:43:56 [TimBL]
q+
16:43:57 [Ian]
TB: Is this worth doing?
16:44:10 [DanC]
I like the idea of a user-agent processing model.
16:44:17 [Ian]
s/first decrypt/first decrypt what you can/
16:44:17 [DanC]
sortof.
16:44:59 [Ian]
CL to TB: This means that you can assert validity of some
subtrees, but must leave what you can't decrypt.
16:45:18 [Ian]
DO: I like TB's idea. I think from UA perspective, most people
will do validation before (something else).
16:45:27 [Ian]
PC: This won't work in XSLT 2.0.
16:46:02 [Ian]
DO: I hear PC would be more comfortable if schema validation
were done before xslt.
16:46:15 [Ian]
PC: I think that "it depends". I think that the 80/20 cut might
change with XSLT 2.0.
16:46:33 [Ian]
PC: ....templates will look for data types, not just element
and attribute names.
16:46:41 [Ian]
PC: ....if 80/20 cut changes, that makes me nervous.
16:46:43 [Dave]
q?
16:47:02 [TimBray]
q=
16:47:04 [Zakim]
TimBray, if you meant to query the queue, please say 'q?'; if
you meant to replace the queue, please say 'queue= ...'
16:47:09 [TimBray]
q+
16:47:13 [Ian]
TBL: When you run a C program, should you calculate strings
first, or math, or built-ins?
16:47:36 [Ian]
TBL: When you look at them as functions, deciding which to do
first is a ridiculous question.
16:48:14 [Ian]
TBL: Schema is a more interesting question. When you are
looking for a piece of a document, even if you've not
elaborated a piece, you may find a particular element instance.
16:48:30 [Ian]
TBL: It could be that schema processing could be done as a
local elaboration.
16:48:37 [Ian]
q?
16:48:46 [Ian]
PC: This is too simplistic a model for how schema runs.
16:48:46 [DanC]
Ian, you're available to scribe next week, 18Mar, yes?
16:49:18 [Ian]
PC: TBL's description is correct for simple cases.
16:49:38 [Ian]
TBL: I have 2 problems with TB's proposal as the 80/20 cut.
16:50:00 [DanC]
I'm slightly at risk 18Mar; travelling. plan to be available,
but things could go bad.
16:50:08 [Ian]
a) For one thing, it's over-constraining: if in one part of a
document there is encryption and another is xinclude, I may
want to do one at one moment and another on the plane.
16:50:28 [Dave]
q+
16:50:40 [Ian]
TBL: Making an ordered list will knock out software that does
things in a different order when the order is relevant.
16:50:53 [Ian]
TBL: I like the summary:
16:51:00 [Ian]
- Order doesn't matter unless,
16:51:13 [Ian]
- Your function applies or interacts with another part of the
document.
16:51:19 [Norm]
I don't think it works for subtrees either.
16:51:28 [Ian]
TBL: There are a lot of cases that TB's order doesn't solve.
16:51:56 [Norm]
I might have a bit of XSLT that counts things in my subtree.
The processing-or-not of xinclude inside my subtree will effect
the count.
16:52:00 [Ian]
TBL: In my model, siblings are processed in an arbitrary order.
16:52:12 [DanC]
TimBL, I suggest you take TBray's comment as feedback that your
proposal isn't understood. [as he said]
16:52:20 [Ian]
TB: If this order doesn't break anything substantial, then
TBL's 80/20 may be acceptable. I need to reread.
16:52:42 [Ian]
PC: If we attach a name to TBL's default, then we need to allow
people to say "Use the default or don't."
16:52:55 [Norm]
q+
16:53:15 [Norm]
q-
16:53:15 [Ian]
DO: I don't think we'll get to an 80/20 position on this. I
think the TAG should punt on this issue. Maybe give advice to
another group to discuss this.
16:53:18 [Ian]
q+
16:53:30 [DanC]
TimBL, over to you
16:53:51 [Ian]
DO: If I can specify an attribute to say "use default
processing", then I can go with that.
16:53:52 [TimBL]
q=
16:53:53 [Zakim]
TimBL, if you meant to query the queue, please say 'q?'; if you
meant to replace the queue, please say 'queue= ...'
16:54:38 [Dave]
Wonder if the rathole map will be a rathole
16:54:46 [DanC]
Zakim, please put your help messages in /me actions or in
-notice- thingies, not in the log.
16:54:47 [Zakim]
I don't understand 'please put your help messages in /me
actions or in -notice- thingies, not in the log.', DanC. Try
/msg Zakim help
16:55:02 [Stuart]
+q
16:55:05 [Ian]
PC: I agree that, if we could define components that met TBL's
constraints (local subtrees only), then having any kind of
processing model would be easier.
16:55:36 [DanC]
Chris, I'd like to see you write something up about an
xml-user-agent way of doing things.
16:55:57 [Ian]
------------------
16:55:58 [Dave]
Roy, did you get my text?
16:56:06 [Ian]
TBL: Please recall to send 1-page summaries by Weds.
16:56:19 [Zakim]
-TBray
16:56:19 [Zakim]
-Stuart
16:56:20 [Ian]
Adjourned.
16:56:29 [Zakim]
-Roy
16:56:30 [Zakim]
-ChrisL
16:56:31 [Zakim]
-Ian
16:56:38 [Zakim]
-DanC
16:57:02 [Stuart]
I think it's not just about confining an effect to a sub-tree.
I think the sources that an XML function draws on are also
significant
16:57:20 [Zakim]
-TimBL
16:59:41 [DanC]
Ian, do you think the sort of summary I sent about last week's
meeting is sufficiently easy to do that it doesn't impose a
burden?
16:59:52 [Ian]
I haven't seen it.
17:00:23 [DanC]
summary has (a) attendance (b) list of agenda items, with
actions and decisions under each.
17:00:46 [Ian]
URL?
17:00:48 [Stuart]
I think this is the case with Norm's TOC example which draws
it's influences broadly from a document.
17:01:24 [DanC]
* notes from 4 Mar TAG telcon [was: IRC log?] Dan Connolly
(Fri, Mar 08 2002)
[14]http://lists.w3.org/Archives/Public/www-tag/2002Mar/0030.ht
ml
[14] http://lists.w3.org/Archives/Public/www-tag/2002Mar/0030.html
17:02:43 [Ian]
Yes, that's do-able.
17:03:05 [Stuart]
bye
17:03:08 [Stuart]
Stuart has left #tagmem
17:04:04 [TimBL]
Ian, I will be around. i want grab some lunch while the q is
short.
17:04:16 [TimBL]
RRSAgent, bye
--
Ian Jacobs (ij@w3.org) http://www.w3.org/People/Jacobs
Tel: +1 718 260-9447
Received on Monday, 11 March 2002 20:45:37 UTC