29 Mar 2005

See also: IRC log


Tim Berners-Lee, Dan Connolly, Roy Fielding, Noah Mendelsohn, David Orchard, Vincent Quint, Henry S. Thompson, Norman Walsh
Ed Rice
Vincent Quint
Henry S. Thompson


VQ: Agenda has been revised per DC's suggestions

Next telcon

VQ: 5 April, 1300 EDT

NM: Possible regrets

Roy: regrets for next week

DC: Will scribe next week

VQ: Minutes from last week?

<DanC> (it should say chair: vincent rather than SV_MEETING_CHAIR, but not a big deail)

VQ: HT had action about where to put minutes

HT: Some progress, proposal next week
... This week's will go in W3C date space, as last week

Request from WWW2005 for TAG participation

<DanC> (I plan to *not* be there)

<timbl_> I will be in Chiba

<Roy> not me

VQ: Who will be there

<noah> Noah will not be in Japan

TBL: Will be at WWW2005. . .

<Norm> I, alas, will not be in Japan either

<Vincent> Vincent will not be there

TBL: Doesn't think he will be there for DevDay

VQ: Seems like no-one will be there

TBL: Anyone there for rest of conference?

ACTION: VC to check with DO and ER and, pbly, tell organisers that no-one can be there [recorded in http://www.w3.org/2005/03/29-tagmem-minutes.html#action01]

Reviewing WS-Addressing Core etc.

VQ: Received request from WS-A WG that we review Web Services Addressing 1.0 - Core Editors' copy Date: 2005/03/22 16:22:02 and SOAP Binding (see endPointRefs-47: WS-Addressing SOAP binding & app protocols)
... They have extended the Last Call period through mid-May
... Can we have reviewers for them?

VQ: ER has volunteered -- anyone else?

DC: Has looked at one before . . .
... Not sure I could say anything polite, maybe it's improved since then

NM: We should be watching how identity is being handled in WS-A

NM: I thought there was agreement that ref-prop and ref-param distinction was lost, and agreement that the URI is what determined identity.

<DanC> the "bits go here" text seems to still be there: [[ A reference may contain a number of individual parameters which are associated with the endpoint to facilitate a particular interaction. ]]

NM: But the WG says there is more flexibility, while acknowledging the strong position on identity, allows other practices
... TAG could/should say "This could be misused, you should say that too"

DO: not me, too close to this

<noah> I like Tim's idea: don't just review the document, but instead flag issues we should discuss in TAG meetings/calls

DC: Can we let this hang a few weeks?

VQ: Not finished yet, no rush

DO: They are 99% finished, no reason to wait

VQ: We will return to this in a few weeks when ER is on call

DO: Happy to help privately or on call/www-tag anyone who is reviewing

Reviewing XML Query and XSL WG Last Call documents

VQ: Paul Cotton asks (http://lists.w3.org/Archives/Member/tag/2005Mar/0028.html [member-only]) if we want to review, offers help:

VQ: Are we interested in reviewing all/any of these?

DC: One reason . . . a comment I sent in 2003 (why the special case for % in fn:escape-uri?)
... Two ways to escape URIs, I will check this up in their Last Call

VQ: Will publish on 4 April 2005, Last Call lasts 6 weeks thereafter

DC: Will review Functions and Operators http://www.w3.org/XML/Group/2005/04/WD-xpath-functions-20050404/xpath-functions.html insofar as its a URI standard library

<Roy> They also need to update the references [to URI escaping: http://www.w3.org/TR/xquery-operators/#func-escape-uri]

DC: They are proposing to produce namespace documents, we should look at them, I nominate NW

<Norm> http://www.w3.org/XML/Group/2005/04/WD-xpath-datamodel-20050404/xpath-datatypes.html

NW: I constructed the namespace document, so I'm not the best person to review it

NW: namespace document has an anchor for every name in F&O
... pointers are member-only until 4 April

DC: Please point www-tag to this after it's published

VQ: So we have DC for F&O -- anyone else?
... Paul Cotton was only asking if we wanted to review and if so what, not requiring review of the whole suite

NM: F&O is a good target, language docs not likely to have TAG issues, need Schema and Core to look at data model . . .

ACTION: DanC to review Functions and Operators insofar as its a URI standard library [recorded in http://www.w3.org/2005/03/29-tagmem-minutes.html#action02]

VQ: That's enough

ACTION: VC to reply to Paul that DC is doing F&O, Norm will call attention to namespace document [recorded in http://www.w3.org/2005/03/29-tagmem-minutes.html#action03]

issue URNsAndRegistries-50

<DanC> reviewing 22 Mar minutes... "their deadline is 14 Apr"

VQ: Two actions -- ER to review XRI document, no progress

<timbl_> http://www.w3.org/XML/Group/2005/04/WD-xpath-datamodel-20050404/xpath-datatypes.html -> http://inamidst.com/grddl/demo?uri=http%3A%2F%2Fwww.w3.org%2FXML%2FGroup%2F2005%2F04%2FWD-xpath-datamodel-20050404%2Fxpath-datatypes.html [Broken -- see below]

VQ: Deadline is 14 April, so we need to move
... HT to produce initial draft with DO

<Norm> I get an EOFError, timbl_

HT: Barest of beginnings http://www.w3.org/2001/tag/doc/URNsAndRegistries-50.xml -- next step is to fold in material from DO

<Zakim> DanC, you wanted to suggest VQ let the XRI folks know we're interested but would like an extension beyond 14 Apr to review

DC: 14 April is hard to meet, but they should know we're about to comment
... Could you (VQ) ask them for an extension

ACTION: VQ to send note to OASIS TC requesting extension to end of month [recorded in http://www.w3.org/2005/03/29-tagmem-minutes.html#action04]

Issue httpRange-14

VQ: We ran out of time two weeks ago, picking up on this now that TBL is here

<Roy> http://www.w3.org/2001/tag/issues.html#httpRange-14

<DanC> What do HTTP URIs Identify?

<DanC> (bummer Ed's not here)

VQ: Let's try to develop this in to a finding on httpRange-14

TBL: SWBPG has said they'd like to see resolution that any URI can be used for anything

<Roy> http://lists.w3.org/Archives/Public/www-tag/2005Mar/0101.html

TBL: We need understanding amongst TAG not just about httpRange-14, but also an improved glossary -- resources, function of http
... Looking for formal description of web in SemWeb terms

<DanC> (some work on formalizing webarch terms http://www.w3.org/2001/tag/fdesc54/ )

TBL: Compromise between my position that if retrieving a URL gets a picture [subsequent retrievals will always get some kind of picture, i.e. the picture is what the URL is for]
... Contra position [if the URL is really for what it's a picture of, then subsequent retrievals might give you an HTML page describing that same thing] e.g. Dublin Core are using vanilla URL as a property
... TBL replies OK, but you can't reply 200 [OK], because you can't give back the relation itself
... Maybe we should come up with a new return code for this case

<Zakim> noah, you wanted to ask for clarification of Tim's statement: you should expect a picture

TBL: Note that you can't reasonably return the whole WordNet DB

NM: Clarification of picture example
... Do a get, what comes back is a picture, entitled to expectation that subsequent GETs will also be some kind of picture, although its encoding might change
... NOT that the URI was really identifying Noah, so almost anything might come back again
... Narrow range is the range that returns 200

TBL: Life Sciences identifiers -- instead of complicated scheme, convert to URI, no # there now, not obvious where to put it
... Namespaces where # doesn't make sense, where / is really sensible
... Not as architecturally clean as saying all http space is info resources

<Zakim> ht, you wanted to discuss another compromise

HT: I hear two positions:
... (1) http URIs can be used for both things you normally expect to retriev from and things you don't normally expect to retrieve from
... (2) while maybe you could, it reduces confusion if we had some syntactic way to distinguish those two cases
... Topic Maps have a special syntax around the *use* of a URI.
... The topicmap architecture works as (2)... it's not just business-as-usual when you want to point outside the web
... there are proposals like tdb: from Masinter and wpn: by myself et al
... some of the folks that are opposed to the "use #" position are opposed because they don't like that particular mechanism, although they agree some distinction is necessary, so maybe we should state the goal in a higher-level way that doesn't commit to a particular mechanism ...

<Zakim> Roy, you wanted to consider that SW applications could accept URIs that indirectly identify concepts through information resources if there is an N:1 relationship between the

<timbl_> sweb: http:

TBL: To look at way of relieving peoples pain, where we keep seeing new schemes as new applications arrive, maybe a new scheme for asking about semantics

<DanC> (there's also HTTP for setting up telephone calls, called SIP. but that's another story)

HT: yes, that's v. close to the wpn: story

RF: Want to go back to the question of why we want to make the distinction
... What do we gain by saying "All the http: identifiers have this constraint"?
... As opposed to saying "We have one http resource which gives information about another http resource"?

TBL: Use cases?

RF: Technology examples -- we could change the language this way, or the usage that way, pros and cons

TBL: Dublin Core is a popular test case
... dc:Title is not a relation, its a conceptual thing, if you dereference it you get a page which says: "This is the abstract 'title' property"
... We could say -- just put a # there, or use wpn:, or . . .

RF: Those are just 'solutions' -- how does it make anything better?

<DanC> (found RF's comments that seem 180 degrees away from timbl's position http://lists.w3.org/Archives/Public/www-tag/2003Jan/0148.html )

<timbl_> Here:

<timbl_> http://www.w3.org/2005/moby/dick

<timbl_> { <http://www.w3.org/2005/moby/dick> dc:creator ?x } => { ?x a :Answer }.

HT: Consider the difference in ordinary language between "Moby Dick wasn't a real whale, but was based on a real whale that sank a ship off the coast of Peru in the 18th century" versus "Moby Dick isn't the Great American Novel, but it will do until the real thing comes along." Smart human beings can understand the different referents by clever use of context, but we use typographic conventions to make the difference clear when we can. Machines aren't as clever, we should make the distiction evident whenever we can.

<noah> But Henry, has anyone proposed the >same< URI for the whale and the book text? I don't think so. The question is, I think, whether the possibility of a slimy slippery non-information whale can be eliminated either by inspection of the scheme and/or by rc=200

HT:Noah, see Tim's example!

RF: So you dereference, and see notes about reading Moby Dick the book

TBL: If we don't have some convention, what's to stop me using the above URL for the book

which makes ... dc:creator <timbl> false

TBL: So the URI clearly identifies what you get when you retrieve, i.e. the web page

RF: Can you extrapolate this to all examples of http: I don't think so

<Zakim> noah, you wanted to ask a question about role of media types in the use of #

NM: Mechanistic view of the # position:
... 1) WebArch says interp of fragid is determined by media type of thing retrieved;
... 2) # proponents say if I want to talk about the whale I'll use a #

<timbl_> racine [is the word I use to refer to the part of a URI before the #]

NM: 3) LHS is a URI, I do a retrieval, something comes back, with a media type

<DanC> my bumper sticker explanation is: foo#bar refers to what bar refers to in foo

NM: What do we need to know about this media type wrt the # convention?
... Since I know I can't put the puppy in the wire, I still need something there which the #puppy connects to
... What's the connection between the media type and what kind of media types can I deploy that makes this work
... Surely I can't use text/html, because that media type says # resolves to fragments

DO: ref Abstract Components finding, advice to media type definers

NM: Suppose we said we had a class of resources that always returned a 0-byte document, with a media type which tells you you get abstract referents from fragids

TBL: That's twisted
... See DC's foo#bar above, media type tells you how to interpret what you get as foo, and how to find bar in it
... RDF says bar refers into the universe of discourse

NM: I want to refer to people, I design a media type that allows this, can I do that w/o having 4billion parts, in some explicit way?

DC: You could do the 0-byte story, but it's unlikely to work very well

NM: You're telling me to ground it in the document, how?

DC: WebArch doesn't guarantee to allow you to solve your problem your way

TBL: Possible way, in the language you say after the # identifies something in this space, could be an algorithm, not a navigation story
... So the spec. could say to formulate a SPARQL query

NM: So racine always retrieves the same thing, why not move it in to the media type?

TBL: No, media type is ...rdf, [missed the rest]

NM: Document is rdf or sparql query, #Noah is parameter

DC: Reduces to 0-byte case

<DanC> (I hope we get to talking about what to write soonish)

TBL, NM: [scribe missed an exchange here, sorry]

TBL: Reading document and understanding it is the way forward, but navigation within it is not necessarily what the fragment will do

<Zakim> DanC, you wanted to ask RF about XML signature algorithm identifiers

<DanC> use of fragments as names is irresponsible

<DanC> . http://www.w3.org/2000/09/xmldsig-more#...

DC: XML Sig algorithm identifiers, with # in them, I recall Roy saying use / there

<Zakim> Roy, you wanted to invert the moby dick example by taking the original text of Moby Dick and decorate it with the original author, publisher, and date metadata and place it on an

RF: If using defining examples, use plausible and thorough ones
... Stipulate we produce a full electronic reconstruction of a particular edition of Moby Dick?

TBL: In that case there's no argument

RF: There should be some -- it's still not a book, it's a web page

TBL: But it's the Work of Art

<Zakim> ht, you wanted to worry about weakness of media type

HT: Henry says: depending on the media type makes me nervous, because sometimes I can't dereference

<timbl_> Another time Henry (offline)

VQ: Making progress -- can we work from existing doc't (DesignIssues/HTTP_URI), or start from scratch?

DO: Ponder more

TBL: Discuss more
... We're not going to reach conclusion in 5 minutes

VQ: More discussion another time

DC: One part of goal was to bring new members on board, so we go again when we have ER with us
... It helped to have two weeks notice -- next time?

[all]: Yes, next week

VQ: Remaining agenda items postponed until next week

<timbl_> Therom 1. A group with n people takes O (n^2) to come to a conclusion.

<Roy> units?

<timbl_> Therom 1a. A group with n people takes O m* (n^2) to come to a conclusion. with n-m people at each meeting

Summary of Action Items

[NEW] ACTION: DanC to review Functions and Operators insofar as its a URI standard library [recorded in http://www.w3.org/2005/03/29-tagmem-minutes.html#action02]
[NEW] ACTION: VC to check with DO and ER and, pbly, tell organisers that no-one can be there [recorded in http://www.w3.org/2005/03/29-tagmem-minutes.html#action01]
[NEW] ACTION: VC to reply to Paul that DC is doing F&O, Norm will call attention to namespace document [recorded in http://www.w3.org/2005/03/29-tagmem-minutes.html#action03]
[NEW] ACTION: VQ to send note to OASIS TC requesting extension to end of month [recorded in http://www.w3.org/2005/03/29-tagmem-minutes.html#action04]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.119 (CVS log)
$Date: 2005/04/01 11:48:52 $