Tag F2f 3 March 2006

3 Mar 2006

See previous session, agenda

See also: IRC log


Tim Berners-Lee, Dan Connolly, Noah Mendelsohn, David Orchard, Vincent Quint, T. V. Raman, Ed Rice, Norm Walsh
Henry S. Thompson



Vincent: I had a discussion with Mark Baker, and confirmed that we would be happy for him to join a teleconf to help us address the issue he has over End Point References. let us invite him March 21 or 28.

<Ed> Scribe: Timbl

<scribe> ACTION: Vincent to Negotiate date of his attendance with Make Baker [recorded in http://www.w3.org/2006/03/03-tagmem-minutes.html#action01]


David: I thought I would step back a bit.

<DanC_lap> (pointer to "part one"?)

David: We are looking at 2 documents: part1, which tlaks about versioning XML, talking about design decisions. Terminology section, motivation, stragegy.
... Part 2 has XMLSchema sepcific stuff.

<DanC_lap> Ext/Vers terminology with generic/xml split

David: This has ben out there for a year or so. After discussion wih Noah etc, and thing abiout how these thinga are actually done, iot seems there is a more interesting story to be told about language versioning.
... Talking about different parts of the system which may operate on XML; and also the system above XML.

<noah> Do we have links to Dave's most current drafts on XML Versioning? We should probably include those in the minutes. I think he prepared some just before the Cambridge December 2004 F2F.

David: I was using hReview over the weekend. Found that these people have names of things like FN for full name and I wondered how they are going to deal with versioning.
... Are we going to have several layers on the onion? I would like 2 more l;ayers:

<noah> Ah...I think Dave's latest drafts are at http://lists.w3.org/Archives/Public/www-tag/2004Nov/0071.html , unless I've missed a newer one.

<DanC_lap> the issues list agrees with you, noah.,

David: 1) The notion of systems being built up where each component has a different understanding of a language .... Noah calls it partuial understadning ... see my bl;og entry @@@ about java going to XML going to SQl. I talked about the notion of different flavors of system accessing the same XML. Like T having [missed]

2) Talk about language evolution in general. We have a story about Markup exten'y, but I'd like also to talk about for example µFormats. I don't know whether to do CSS eveolution and URI evolutoin as well... maybe too far togo.

<DanC_lap> (css versioning is a permathread, internally, in discussion of validator.w3.org )

scribe: i'd like to also think about how when you have XML>Java>SQL you end up with constraints from each stage on the syntax of names etc. The initial idea of "ignore unkowns" is not well defined in these systems. What does it mean to 'ignore'?

@@@ Link to Daves's blg, someone?

<Zakim> DanC_lap, you wanted to noodle on the idea that the end-game for ext+vers is Web Arch Vol 2: Language Evolution and the Self-Describing Web: extensibility, versioning, composition

scribe: So let us discuss whether this is a workable method of looking at this.

DanC: What is the endgame? For many things, it is WebArch Volume 2 ... though who onws how we slice up the picees in the meantime.
... Re CSS versioning, we have a CSS versioning, there is a problem that there is no version indication in CSS and the validator can't work out what it should be.

Raman: One of the things which has happened in W3C is that every slightly new version of a document gets a new namespace, even for a new namespace. I think tyhis is a mistake.

If you have to namespaces, you have ns1:section and ns2:section the spec of section has not changed at all, but it has a totally new name.

scribe: I conclude this is too chaotic.
... (... Multimodal working group for example? Maybe voice browser?)
... In XForms, we had an experimental namespace but we are thinking of graduating the working bits t the original namespace.

<Norm> The XSL and XML Query WGs used this technique for evolving versions of several XPath 2/XQuery namespaces

scribe: Look at SVG. SVG 1.0, 1.1, 1.2 and tiny each have different namespaces!
... Graceful failover is really important.

Dave: I have been trying to get the design decisions clearer in the document.

<Zakim> noah, you wanted to ask whether I'm looking at the latest draft

Raman: Make the same mistakes consistently over W3C at least ;-)

<DanC_lap> (whee! life without CVS)

Noah: Where is the latest draft? My link above is to the latest version I found.

<noah> http://lists.w3.org/Archives/Public/www-tag/2004Nov/0071.html

VQ: That is linked from the findings page.

[discussion about the fact that the latest version is not the one in CVS and linked from the Findings page, Agenda, etc.]

<DanC_lap> Ext/Vers terminology with generic/xml split

<Zakim> DanC_lap, you wanted to ask VQ to take an action to write up "lack of CSS versioning syntax is trouble" or "on versioning in CSS" and to relay a concern about SMIL versioning and

DanC: I'm not persuaded that XHTML 2 merits a separate REC-level namespace from XHTML 1 and to argue that people making new languages shouldn't be obliged to give their

<scribe> ACTION: Vincent to Write to www-tag about CSS versioning being a problem "levels" [recorded in http://www.w3.org/2006/03/03-tagmem-minutes.html#action02]

DanC: Someone told me that the 3 smil profiles each have their own namespaces and this will be a disaster. Does the finding say that is a problem:

Dave: I didn't come up with a hard rule for making a decision.

<scribe> ACTION: DanC to Look at the document and see if it is good for informing on this SMIL problem of multiple namespaces. [recorded in http://www.w3.org/2006/03/03-tagmem-minutes.html#action03]

DanC: I am not sure whether versioning tehcnology in the technology can be done without the xperience of having done veriosn 1 and 2 in version3

David: You actually make decisions about versioning in your langauge whgether you like it or not

Raman: the other thing which i would like to say is like when xHTML 1 did 1 namespace. What people often do is to append to a namespace a little string to indicate a profile, and it seems that is acceptable. Maybe we could in hindsight say that that is wrong

<Zakim> Norm, you wanted to observe the XSLT case

Norm: Xslt2.0 uses the same NS as XSLT1.0, and so does docbook
... There is a whole matrix of what happens with version n document being processed by version m agent.

DanC: Please do CR test cases for that

Norm: [nods with a smile]

Raman: Can we do examples?

<DanC_lap> 8.1 Version Strategy: all components in new namespace(s) for each version (#1)

<DanC_lap> (at home, I don't ask "where does this spatula belong?" becuase it might not have an established home yet. I ask "where would you look for this?" which pretty reliably produces an answer. ;-)

<noah> TBL: We could now look at where people have made different namespaces.

<noah> TBL: We could see whether this has caused problems for software.

<noah> TBL: We could also look(?) whether there is generic metadata to express that one namespace is "compatible" with Edinburgh sense with each other?

<noah> DC: I've done some of that/

<noah> TBL: In RDF?

<noah> DC: Yes, for 3 standard W3C ???? (types of compatibility?)

<noah> DO: We should incorporate in finding?

Tim: [missed]

<DanC_lap> tv+ tvraman

<Zakim> noah, you wanted to discuss namespaces a bit

Tim: Should we look at generic dispatch code which can tranform an XML doc from one new namepsace into an old one using metadat, and then feed it into an old version client?

Noah: It is easy where there is a clear mapping between betwen different mappings. Not sure there is an easy 89/20 situation.

Tim: I was only talking about total back-compatibility.

<DanC_lap> re RDF vocabulary for "backward compatible", see http://www.w3.org/2000/10/swap/util/changePolicy which defines #stableSpecification etc. that formalize the 3 options in http://www.w3.org/1999/10/nsuri

Noah: Namespaces are for preventing name clashes.
... in XML Schema, we debated how many namespaces, and felt one namespace per document might make sense. Or should it be one namespace per language?
... When you start doing one namespace per version, you end up renaming things which have not changed.
... When people write an XPath for all the paras in a document, and the document may have one of many namespaces for paragraph, the XPath becomes a total pain.

<tvraman> I have an xslt in emacspeak that converts rss to xhtml and one for atom to xhtml. The poor stylesheet has to carefully handle the same element in three different RSS namespaces and ends up duplicating the rules

Noah: Sometimes you decide something really changes its semantics, so you want to search only a new one.
... You can image a certain XPAth processor which reads teh metadata, and is smart about that.

<Zakim> tvraman, you wanted to add that today browsers kick off an impl based on namespace, and this was especially true of IE in switching between html and xhtml. But that shouldn't be

Dave: UBL does that: change the NS all the time, but have XPaths which are huge and recommended to do anything with UBL.

Raman: Yu can use the namespace to disptach a different code, but the two versions may in fact share code. Maybe the sofwatre instalation problem should look at the namespace, and also the has-feature stuff from DOM3.

<Zakim> dorchard, you wanted to respond that schema did make some decisions about multi-ns wrt extensibility..

DanC: I hate DOM3 has-feature with a purple passion

<DanC_lap> (it could be that has-feature was fixed to use URIs. anybody know? but even so, I think it's the same anti-pattern as has_attr in python.)

<DanC_lap> . 8.2 Version Strategy: all new components in new namespace(s) for each compatible version (#2)

<DanC_lap> in http://lists.w3.org/Archives/Public/www-tag/2004Nov/att-0071/versioning-part1.html

Dave: Noah talked about Schema and namsepsaces. i think one of the biggest issues in the XML community is that multiple namnespace languages -- see pattern 8.3 -- for versioning eg for WSDL They have abstract ionterface, soap binding, and HTTP binding. The idea was to be able to rev the namespaces differently.
... The issue is that you can't write a Schema for this.
... You wantto be able to say that the binding must occur here, whatever its version.
... Couldn
... 't use substitution groups
... There is flexibility in eth Schema design, including adding names to a namespace and new versions is really important design center.
... Taking one extreme, you could have every name in each namespace! What would a verison of the language be?
... i worry about going to far that way.

Vincent: You wanted to get back to the issue of terminology....

Noah: We could tell a story which goes beyond XML.
... There is a story which is independent of XML, and it becomes useful in the XML level. When people do unmarked up content inside the XML have versioning issues. too.

DanC: Encourage you to write down the generic story.

Noah: Dave thought that the XML people would not keep reading stuff which was generic.

Dave: I don't want many views of the same thing from many angled, and too much abstraction.

DanC: Makes sense for bits about text strings to be there but not the first bit and ignored by the XML person whi an XML problem.

Noah: people are asking me to do this -- or I can do metadataInURI

<Zakim> timbl, you wanted to suggest limiting the effort we put into the XML layer in that there are limited things on can do at that level, and suggest that future parts of it deal with

<DanC_lap> (managing one's todo list for the TAG is something I have certainly not solved to my own satisfaction.)

<DanC_lap> (I will repeat that I like the way dave and I have found a machine-readable formal artifact, i.e. the violet uml diagrams.)

<noah> TBL: Maybe it will fall out of the terminology, but I think telling a general story has value.

<noah> TBL: Dave, are you thinking at focussing on XML-specific or not?

<DanC_lap> Ext/Vers terminology with generic/xml split

<noah> DO: Noah didn't characterize my latest story as well as he should. I've moved toward including more general stuff.

<noah> DO: But be careful, the 2004 version was almost all XML.

<dorchard> http://lists.w3.org/Archives/Public/public-xml-versioning/2006Feb/att-0012/ext-vers-generic-uml.png

Dave: We should do several but not too many case studies of differenet languages.

Vincent: CSS is diff from XMl -- not a metalangauge, and under the control of one working group. It ios comparable with HTML.

Noah: They have a versionming strategy

<DanC_lap> (DO, do you distinguish betwee "ignore the whole element" vs "ignore the tags" ? css has an interesting design there.)

<noah> TBL: When you use XML for something specific like CDF, or HTML, you can tell stories about things like must-ignore.

<noah> TBL: You can and should describe that whole branch of the tree.

<noah> TBL: They are all in one form or another presentations languages, and you can say a lot about them.

<noah> TBL: RDF is another branch you can say a lot about.

<noah> DC: You've said that before, why are you saying it now?

<noah> TBL: Because we're discussing the form of the document. How much general stuff, early, etc.

<noah> TBL: I'm suggesting the document start totally generic, then move on to XML, then two branches, one for presentation languages, and one for RDF.

<noah> TBL: Now having said that's a logical order, I note that some people have said that will be hard to read, and maybe we need a different organization.

<noah> TBL: You have examples where you talk about names "consisting of" pieces. That's not an appropriate explanation in the case of RDF. RDF doesn't have a fundamental notion of "consists of". It's not tree-like.

<noah> DO: I say the name language has the constraints that a name "consists of" the piece parts.

<noah> TBL: right, that handles any context-free grammar

<noah> DO: I think consists of can be mapped to things like RDF

Tim: But when you extend RDF systems you don't change the CFG.

<Zakim> noah, you wanted to say, the story about the terminology needs to be general

<Zakim> tvraman, you wanted to add: ass CSS starts getting used to say interesting things on the right hand side of the equal sign, versioning becomes an issue --- microformats being a

<DanC_lap> (timbl, I don't think you convinced DO to add a CDF section and a semweb section. I wonder if that was a stalemate or if you've accepted that we don't need one or if you're likely to write one.)

<dorchard> (I wasn't sure whether the suggestion was to have a CDF/SemWeb/someotherthing/someotherotherthing terminology sections..)

Raman: When you talking of versioning non-XML, especially with th..... we all know how to format books. How are you goig to version something like what Hakon did the other day? If I have written my book this year, but the style sheet language changes, then some sections of the book may not com out tomorrow?
... What is the true content? The source or the presentation?


Decision to move the final session in a high granularity silicon-based environment.

Tim: What happened to the math in the whiteboard in Edinbugh?

Back compatability being the interpretation of a document being a subset of the]intent of the author

or something.

<DanC_lap> (fyi, timbl, http://dig.csail.mit.edu/breadcrumbs/blog/2 )

<DanC_lap> er...

<DanC_lap> (rather: http://dig.csail.mit.edu/breadcrumbs/node/87 Using RDF and OWL to model language evolution )

Summary of Action Items

[NEW] ACTION: DanC to Look at the document and see if it is good for informing on this SMIL problem of multiple namespaces. [recorded in http://www.w3.org/2006/03/03-tagmem-minutes.html#action03]
[NEW] ACTION: Vincent to Negotiate date of his attendance with Make Baker [recorded in http://www.w3.org/2006/03/03-tagmem-minutes.html#action01]
[NEW] ACTION: Vincent to Write to www-tag about CSS versioning being a problem "levels" [recorded in http://www.w3.org/2006/03/03-tagmem-minutes.html#action02]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.127 (CVS log)
$Date: 2006/03/22 13:40:55 $