TAG Weekly meeting

16 May 2006

See also: IRC log



T.V. Raman, Ed Rice, Vincent Q, Norm W, Dan C, Noah, TimBL


            David O, Henry T


Vincent Q


T.V. Raman




<raman> Scribe: Raman

<noah> +1 to cancelling next week. I almost surely can't make it. I.e. at risk -> regrets

Resolution: No Call next week

<DanC> I seem to be available 30 May

<EdR> I'm available next week.

No call May 23, next call will be May 30

Ed will be scribe for May 30 call

<DanC> close enough for me http://www.w3.org/2006/05/09-tagmem-minutes.html

Meeting minutes from last week approved

<DanC> (it's not "just giving a name" to the issue; it's deciding to add something to our issues list, which is non-trivial.)

<timbl> http://2006.xmlconference.org/

Q: Should tag have a meeting in Jan in Cambridge?

Dan: Yes but not strongly

Ed: No

Noah: Not instead of December meeting and not unless there is sinergy

Norm: TAG owes it to the community to be available to meet

Raman: No

Tim: Will go with anything happening in Cambridge, no strong feelings either way

Vincent: Not necessarily interested in the Jan meeting.

Vincent to report back to Steve that TAG is not strongly interested in a shared Jan meeting

F2F Agenda

Security -- Possibly on the second day.

<DanC> draft June agenda


Multiple content types for the same URI

Possibly continue discussion of Noah's MetaData finding

Aspects of SemWeb Architecture

<DanC> (indeed, "finishing a document" and "starting a document" are patterns)

<DanC> (yes, I like targeting the boxes; it tends to be worth nailing those down as a group; sometimes the finding has to come first, and sometimes it can come after.)

<Zakim> DanC, you wanted to try provoking a bit

<noah> I wonder whether the versioning discussion is important to Dave at the F2F?

<DanC> (I think the connection from metadataInURI-31 to semweb arch is pretty arbitrary too.)

<timbl> (Norm, when it comes to NamespaceDocument-8 , it is worth saying in lots of places that any program should basically look namespace documents up one, maybe on installation, or maybe every few months, and keep it in a persistent cache. This is so the servers (at w3.org for example) don't get drowned in just re-requests for namespace documents. w3.org has a problem wit hDTD lookups no. )

<Norm> (Hmmm. I see.)

<DanC> (I am profoundly uninspired when it comes to security. It seems important, but darned if I can say anything specific about it.)


<noah> Dated draft: http://www.w3.org/2001/tag/doc/metaDataInURI-31-20060511.html

<DanC> (were any reviewers assigned? I read the whole thing pretty closely. I'm tempted to say we should go thru the boxes one by one.)

Noah/MetaData Finding

Noah: looking for the high-order bit answer : Is this in the right direction

Raman: suggest publishing after the F2F to get community feedback

<EdR> Ed: Noah, I think at the 30,000ft level.. I like the format and structure and find it very clear.

<DanC> (hmm... I think if I had started with the 2nd box, it might have done better. "Guess information from URIs only when the consequences of an incorrect guess are acceptable.")

Reviewers for Noah's document: Raman, Ed

<scribe> ACTION: Vincent to send reminders [recorded in http://www.w3.org/2006/05/16-tagmem-minutes.html#action01]

<DanC> (that one is short, suite, and compelling.)

Tim might also review.

<DanC> (s/suite/sweet/)

<noah> (Interesting: I originally had that first, and thought: gee, everyone's been so concerned that the main thought is "don't infer metadata", that I figured I better lead with that.)

<noah> (I'm very sympathetic to trying to find tighter wording for that constraint. Will work on it.)

<DanC> (yes, I think that one needs to go first, but it's not worded as well; it doesn't provoke immediate "yes, I agree" nor "no, I disagree" responses.)

<DanC> # Issue abstractComponentRefs-37

Abstract Component Refs Issue 37

<noah> (sounds like a plan, I'll try and tighten it)

<EdR> http://www.w3.org/2001/tag/issues.html#abstractComponentRefs-37

<DanC> [[

<DanC> Can you confirm that this URI...

<DanC> http://www.w3.org/2005/08/sparql-protocol-query/#wsdl.interface(SparqlQuery

<DanC> ]]

<DanC> http://www.w3.org/2005/08/sparql-protocol-query/#wsdl.interface(SparqlQuery

<DanC> http://www.w3.org/2005/08/sparql-protocol-query/#wsdl.interface(SparqlQuery)

<Zakim> timbl, you wanted to ask whether RDF/A would help

by the way I have a hard stop in 4 minutes; could someone else scribe for the final 30 minutes?

<timbl> <#wsdl.interface(SparqlQuery> rdfs:describedBy <http://www.w3.org/2005/08/sparql-protocol-query/#wsdl.interface(SparqlQuery>.

<DanC> do you mean rdfs:definedBy ?

On the other hand, RDF/A is the closest we have to something working in xhtml, so let's not knock it.

<timbl> <#wsdl.interface(SparqlQuery)> rdfs:describedBy <http://www.w3.org/2005/08/sparql-foo.wsdl>.

<timbl> http://www.w3.org/TR/2005/WD-rdf-sparql-protocol-20050914/sparql-protocol-query.wsdl

<DanC> huh? I have tons of stuff working with GRDDL that doesn't use RDFa, raman. RDFa is much less close.

need to leave

<noah> Are we losing our scribe?

<timbl> rdfs:seeAlso

<Norm> ScribeNick: Norm

Norm becomes scribe

Some discussion of relative merits of "a href" and "rdfs:seeAlso"

timbl: For a machine that's "seeAlso aware", the namespace document is useful

<noah> TBL: I'm hoping the TAG will eventually set out guidance on which things, like rdfs:see also, a machine should follow

danc: What sort of machine are we talking about? The consumer of .wsdl is a web services toolkits.
... They could be taught to follow see also links, but I'm not sure what the value would be.

<Zakim> noah, you wanted to ask about conneg on the representations

noah: We could start with a RDDL document and later add RDF with conneg.
... That led to a discussion of whether or not conneg should be used to serve alternate formats.
... But these are secondary resources and I'm not sure we have a good story for talking about links with fragids in this context.
... Does the link with fragid represent the same thing if RDF and HTML are conneg'd?
... I'm not sure we ever decided that.

DanC: Yes, that's the state of play, and we have the same problem in other areas, like QT functions and operators.

noah: I think I'd like the answer to be "yes" for better or worse.
... We've established '#'s for some things and they're out in the wild so they better work.
... I think if I want a prose explanation I should be able to get that and if I want an RDF explanation, I should be able to get that too.

timbl: What's the relationship between the pieces

<DanC> (I'd like a name for the analog of 404 in #-space... when you get a representation and it has nothing matching that fragment. Anybody got a suggestion? unbound fragment ref?)

noah: For the Schema data types, you can have <baseuri>#integer, <baseuri>#double, etc.
... I don't think we have complete closure about what's identified by <baseuri>
... I take it as an abstraction for the namespace. The things that seem like documents are representations of that resource, I think.
... I believe timbl's position is that the <baseuri> refers to the document-ey thing I get back.

<DanC> (I think the 2 positions noah are describing aren't observably distinguishable.)

timbl: There's a pun in URIs used for two things; syntactically it's used as the prefix. But by itself it identifies the namespace document.
... I think information resources always identify documents.

DanC: I didn't think that's where we landed.

timbl: Representations are the actual bitstreams. If the resource is a list of things, I'm happy to have the list in different orders if they're unordered.

<DanC> what timbl actually said was "... I use information resource only for things that have a beginning, middle and end"

<DanC> and I meant to ask "really? that doesn't sound like things that can be posted to."

Thanks for the correction, DanC

noah: The resource is the potentially infinite collection.

timbl: (reference to information theory) when you look this thing up, you're going to be more informed. An information resource to me is that information, not the subject of the information.

noah: would it be reasonable for me to define a resource which is all the square roots of all the integers.
... blah-blah-blah#144 refers to the number 12.
... Or "/", I'm just talking about the infiniteness of the set.
... One representation si a java program that computs the squre roots

timbl: For me, a representation is a string of bits and some metadata. What you get in http.
... Those bits, in the given language convey the information that was the information resource.

noah: If the table wasn't infinite; if it was the square roots of the first 100 integers. I could then just give you an HTML page that conveyed it as a table.

timbl: No. The representation of the set must have a different URI. An information resource isn't a set of numbers.
... The statement that the set contains these numbers is an information resource, but that's distinct from the set.

noah: I would have thought they could be conveyed as information.

timbl: We played with the words a lot

<DanC> (where timbl says "it's improtant to distinguish between the set of numbers and the description of it", I'm not yet convinced. I agree that you _can_ distinguish, but I don't know why it's important to.)

timbl: It's not coherent not to distinguish between them.

noah: I want to distinguish them, but I think they're both information resources.
... What I hear timbl saying is that the only one I'm happy to call an information resource are the ones that are documenty

timbl: It's really important because the web is about communication and when I give you a URI I expect you to be able to get information with that URI.

noah: If you ask people what a namespace is, I don't think they'll say "document". It's more set like.
... Once we say "I've got that" now at some level, by the time we get to representations, everyone agrees that what we get is a document.
... The problem is that given a namespace in my left hand, there are lots of different kinds of documents that I might like to write; in RDF, in HTML, in English, in French, etc.
... But that leaves us in the position of asking what is the fundamental document that the namespace URI names (because I have to pick one). But then we trip over how one is a representation of the other.

<DanC> (I find timbl's position mildly more appealing, but the argument seems to be by assertion. It's maybe good enough to convince me, but it's not at all good enough for me to take and convince other people.)

noah: What's really fundamental is the set; how can we use webarch to say that that is on the web?

timbl: we could make it clearer by having a 303 response.
... As a result, the only thing that's identified by the URI is some collection of documents.
... It's not neat and tidy, but none of the processes that get the URI really need the abstraction.

noah: I could say that I control that namespace, yes?

timbl: Yes, you can talk about the document, but they all use the DC namespace to talk about how they're managed.

Scribe isn't sure he captured that

timbl: We don't have a way in rdf of saying that this property is in a namespace; we don't have the concept of a namespace.
... The namespace concept is only used in common parlance.

Norm isn't sure that the fact that RDF doesn't need the concept means the concept isn't useful.

<noah> Hmm. So if the URI with no sharp sign is for "the document(s) about the namespace", as opposed to the set of names, then maybe to talk about the set you need: http://example.org/uriOfNamespaceDoc#namespaceThisIsAbout

Vincent: DanC did you have a goal in mind?

DanC: Yes, that one of the documents would say that the SPARQL example is good or not.

<noah> Well, I guess what I'm hung up on is that in practical terms, the namespace is an important thing. People talk about namespaces all the time. If we don't have a simple, first class way to give a Web name (I.e. URI) to the namespace itself, it seems we've lost something important.

vincent: I don't think we can go further today.

<noah> Fodder for the F2F?

Noah: We don't have a clean simple story about what a namespace URI identifies that avoids a 20 minute discussion

Norm: I share Noah's concerns about the practicality

vincent: I'll plan to schedule discussion about abstractComponentRefs again when Dave is present.
... Adjourned

<timbl> I note that DanC's point was different form Noah's. Dan C seemed to be asking whether something could be a set of names and alos a document. Noah and Tim seemed to agree that the set of terms and the document were distinct things, and just differ in which they were suggestsing was names by the NSURI.

<DanC> I'm not interested to persue that point any more, fyi.

<DanC> " whether something could be a set of names and alos a document"

<DanC> hmm.

<DanC> well, maybe. I don't advocate it, in any case.

<DanC> There's a clear-and-present question in the semweb best practices WG: can a wordnet word (synset) be an information resource?

Summary of Action Items

[NEW] ACTION: Vincent to send reminders [recorded in http://www.w3.org/2006/05/16-tagmem-minutes.html#action01] to review Noah’s MetaData document to T.V., Ed and TimBl.
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.127 (CVS log)
$Date: 2006/05/16 18:35:01 $