W3C Architecture of The World Wide Semantic Web Task Force Teleconference

26 May 2009


See also: IRC log


jar, DavidB, mhausenblas, HarryH, Stuart, alanr, Timbl




<trackbot> Date: 26 May 2009

<mhausenblas> Previous: 2009-05-12 http://www.w3.org/2009/05/12-awwsw-minutes

<mhausenblas> Scribenick: hhalpin:

<mhausenblas> jar: everything prepared, /me chairing and hhalpin scribing ;)

<mhausenblas> Michael: on ESW Wiki the attachements have been locked

<mhausenblas> ... I'm in contact with Mark Baker to fix that (all figures are gone, for now)

<jar> chair: mhausenblas

<hhalpin> Excellent.

<mhausenblas> yes. shall we wait for you to join?

<jar> didn't you get the announcement?

<hhalpin> I didn't get a formal announcement.

<hhalpin> Just JAR pinging Michael.

<jar> I sent one on the 24th to public-awwsw.

<jar> oh, I see... not formal enough

<hhalpin> It's fine, but people might miss it...

<hhalpin> mhausenblas: let's start with jar's mail about generic resources

Jar's tracing of TimBL's IR notion

<mhausenblas> http://lists.w3.org/Archives/Public/public-awwsw/2009May/0006.html

<hhalpin> jar: main observation was to take an axiomatic approach

<hhalpin> jar: rather than to do it ontologically

<hhalpin> jar: start with generic resource

<hhalpin> jar: then there is a parameter space that helps up index the kinds of resources

<hhalpin> jar: three different ways to set-up parameter space

<hhalpin> jar: timbl's version (time, language, media type)

<hhalpin> jar: distinguish generic resource from the "trace"

<hhalpin> jar: "trace" is a function from a parameter-space to awww:representations

<hhalpin> jar: think of it as an observation/recording/sample of generic resource

<hhalpin> jar: try to make it possible to compare these three different notions

<hhalpin> jar: david booth's "ftrr"

<hhalpin> jar: timbl's "generic resource" idea

<hhalpin> jar: roy's resource model.

<hhalpin> jar: do they make a distinguish between resource and trace?

<mhausenblas> (notes that in http://esw.w3.org/topic/AwwswGenOntDiagrams the diagrams are back in again)

<hhalpin> jar: TimBL does, David doesn't, Roy does in text but formally doesn't.

<alanr> nor I

<hhalpin> jar: the reason the second notion comes up is that we are trying to match this abstract work with what one might observe from the web.

<hhalpin> jar: although as timbl put it, it's ontologically distinct, we need to explain what it means for something to be "on the web"

<hhalpin> alanr: not happy with the idea of "consistent"

<hhalpin> alanr: all we know is that they got back from a 200 response.

<hhalpin> jar: according to timbl a generic resource can have awww:representations without being "on the web"

<hhalpin> jar: moby dick, even if it didn't have representations, would still be a representation.

<hhalpin> alanr: but I thought awww:representations were those in AWWW, which were on the Web, thus back from 200 response.

<hhalpin> jar: they *happen* to be awww:representation.

<alanr> wa-representation = 'representation' in the REST or web architecture

<alanr> >>> (AWWW) sense (NOT in the Xiaoshu or plain-English sense that

<alanr> >>> permits, say, a rock, or a citizen, to be represented)

<hhalpin> I personally not sure if we should talk about "representations" not being on Web.

<hhalpin> Or at least "awww:representations" not being on the Web.

<hhalpin> jar: data that codes resource state.

<hhalpin> jar: not necesarily on the Web.

<jar> http://www.w3.org/TR/webarch/#def-representation

<hhalpin> jar: clear that it can be, since it's not specific to HTTP.

<hhalpin> My personal take is that "awww:representation" can be a more abstract notion, but a subset of them can be retrieved by 200 HTTP response codes.

<hhalpin> alanr: can't have representation without a communication channel.

<hhalpin> jar: but let's do it axiomatic

<hhalpin> alanr: not sure if you can avoid ontological committments

<hhalpin> how does whether or not the "awww:representation" being on the Web or not matter for JAR's axiomatization

<Stuart> afaik the core of alan's dificulty is with S in fact being a wa-representation of G.... because how would one tell.

<mhausenblas> 'A representation is data that encodes information about resource state.'

<alanr> is this an axiom ?

<hhalpin> jar: I am not sure that is a great definition.

<hhalpin> mhausenblas: Well, we should make a decision on whether or not to support this definition.

<hhalpin> jar: it's probably good enough for the moment

<hhalpin> jar: the question is "must a representation be on the Web or not?"

<hhalpin> mhausenblas: ask timbl would help

<hhalpin> jar: Is there such a thing as a resource not being on the Web?

<hhalpin> timbl: resource is a concept in the architecture

<hhalpin> timbl: like an integer, like a data-type

<hhalpin> timbl: that's like asking does a number have to be in a formula

<alanr> ?

<hhalpin> timbl: what would depend on the answer to that question

<Stuart> FWIW I'm still of the opinion that representations are not resources that are on the web, they are part of the mechanisms by which resource state is communicated.

<timbl> What depends on the answer to that question?

<hhalpin> jar: Can a resource have representations that are never returned?

<jar> dbooth, not jar: Can a resource have representations that are never returned?

<alanr> Look, if they have an independent existence then they are resources. But it's been categorically stated that they are not.

<hhalpin> timbl: how would you ever know? that can't be verified. Maybe I could have a secret webserver no-one will ever know about.

<hhalpin> security via obscurity :)

<hhalpin> alanr: I think the question is relevant to jar because can an information resource have characteristics that aren't observable.

<hhalpin> timbl: that's a fundamental issue

<hhalpin> timbl: yes, a resource can have characteristics that aren't available via http

<jar> timbl: A generic resource can have characteristics that are not observable... e.g. IPR

<alanr> harry: resources can, allegedly. what does that have to do with representations?

<hhalpin> timbl: copyright is not "on the net"

<alanr> then a request for copyright should return 303

<alanr> ?

<hhalpin> It's just that you can not, therefore, equate a resource with a set of representations

<mhausenblas> Michael: is this resource metadata? (seems not/more?)

<hhalpin> that in fact, ftrr is false.

<hhalpin> that, therefore, fielding's formalization is a bit wrong as well.

<hhalpin> that's fine.

<jar> authorship is not observable

<hhalpin> alanr: all possible times, all possible requets, all possible language

<jar> (always)

<alanr> harry: Note that we are talking about generic resources in JAR's note. There are defined as those that *can* be communicated.

<alanr> are there information resources that can't be communicated?

<hhalpin> timbl: you *can* convey authorship or not in a representation, but it's often *not* conveyed.

<hhalpin> alan, information of course can be communicated. but it does not necessarily means it *is* communicated.

<hhalpin> alan, also communiation presuppose quite a lot, such as baseline language/ontology alignment between agents, which may not be true.

<timbl> A resource can *have* representations

<hhalpin> jar: ftrr is not sufficient to characterize even an information resource.

<hhalpin> jar, that was my point even about relative URIs.

<Stuart> Fielding and Booth FTTR are 'just' models of the behaviour of a resource.... I don't think either truely takes the view that the sets of accessible representations ARE in fact the resource. ie. what Tim just said.

<hhalpin> jar, a very mechanical point.

<timbl> "ftrr is not sufficient to characterize even an information resource." yes

<mhausenblas> (From "Fielding and Taylor ICSE 2000 "Principled Design of the Modern Web Architecture": resource metadata: information about the resource that is not specific to the supplied representation)

<hhalpin> Stuart, I have heard David argue that it is sufficient quite alot.

<timbl> (deoending on what you mean by characterize)

<hhalpin> jar: we have an agreement here, ftrr is not consistent.

<hhalpin> jar: that's why we call it "trace" rather than resource.

<alanr> harry: "consistent"?

<Stuart> Harry, so the work Mobey Dick is an FTTR!

<jar> hey.. ftrr is just *different* from what Tim's talking about (GR)

<hhalpin> jar: want to think about that definition.

<hhalpin> ftrr=trace

<hhalpin> ftrr != resource

<hhalpin> in some cases, possibly ftrr CAN = resource

<hhalpin> but not necessarily and often unlikely.

<hhalpin> alan, not sure what is meant via consistent by jar quite yet.

<hhalpin> I'm just going to say that we need to STOP arguing about whether or not ftrr=information resource

<hhalpin> it's unproductive and incorrect.

<timbl> Why?

<hhalpin> The simple relative URI argument shows that it is incorrect without processing.

<jar> I wanted to say what it meant for G to be served by a server S at a URI U

<timbl> I think the idea that "trr=information resource" is unproductive and incorrect.

<timbl> Ok, so you mean we stop arguring because we agree.

<hhalpin> Timbl, because if a resource = ftrr, then a resource can be defined purely in terms of returned representations.

<hhalpin> yes.

<timbl> Ok, glad we agree.

<hhalpin> we agree, minus possibly david booth, whose not here.

<hhalpin> but i have heard this argument for about a year, and I don't think it's useful and is trivially shown wrong by things like relative URIs.

<hhalpin> alanr: there are parts of the formalism I agree with and dont'.

<hhalpin> jar: I'm trying to just do formal because I want to remain ontology/interpretation free as long as possible.

<hhalpin> jar: thus to prevent arguments about representations/resource definitions

<hhalpin> jar: I use term "consistent"

<hhalpin> would like to clarify this notion consistent

<hhalpin> ah.

<hhalpin> timbl: the server can send back something, a representation?

<alanr> tim: We are still figuring out what "that" means

<Stuart> FWIW jar: I think the mean is clear to me... but the difficulty is in th determination that the wa-representation is in fact a wa-representation G (when S is asked the question using U)

<hhalpin> timbl: if a server sends something back that isn't a representation

<hhalpin> timbl: is it not on the Web?

<hhalpin> jar: can we get representations that aren't representations of a resource?

<hhalpin> timbl: there is a URI, that denotes a resource

<hhalpin> jar: take an arbitary URI, ask the queston, whether some particular server, is it giving you a reprsentation of a generic resource?

<hhalpin> timbl: is a arbitrary resource identified? how?

<hhalpin> jar: moby dick example

<hhalpin> jar: Am I getting representations of moby dick?

<hhalpin> stuart: I think AlanR's question is how to determine if a reprsentation *really is of* moby dick?

<hhalpin> jar: you have to assume you can do that

<jar> Given a URI U, am I getting a representation of Moby Dick from some server given URI U?

<Stuart> Tim: here's the bit of JAR's email that we are discussing (I think)

<Stuart> OK, now suppose that S is an HTTP server, and G is a generic-resource,

<Stuart> and U is a URI. Define "S is consistent with G at U" as follows:

<Stuart> if whenever S receives an HTTP GET request with request-URI U and

<Stuart> responds with a 200 response is received,

<Stuart> the RFC2616-entity in the 200 response is a wa-representation of G,

<Stuart> then S is consistent with G at U.

<jar> Denotation isn't empirical.

<hhalpin> maybe say "the RFC2616-entity in the 200 response is information about the state of G"

<hhalpin> notes that at least we agree ftrr != resource

<hhalpin> jar: we need to figure out if we descriptive/pre-scriptive

<hhalpin> jar: how do we then understand what it means for something to be "on the Web" or not.

<mhausenblas> Michael thinks it is all about naming things. We can use whatever we want

<jar> it's a bait and switch

<mhausenblas> addressing, labelling, denoting, referring to, etc.

<hhalpin> that's why I wanted to subclass Information/Generic Resource with a class WebResource that has Representations that are available via HTTP 200.

<mhausenblas> or simply: identify as in UR *I*

<hhalpin> Also note that I think, in general, formal systems parasite off of natural language.

<jar> Are the representations we see coming from S (from requests on U) consistent with S thinking that U names resource G?

<hhalpin> it's hard to not do that, and systems that do not do it, formal systems, are very ambiguous actually, rather than unambigous, unless their domain is carefully construed (i.e. real numbers, etc.)

<hhalpin> jar: how do we know if a URI identifies a resource with that name?

<hhalpin> jar: maybe the URI names a different resource?

<mhausenblas> so, I read, a URI {addressing|labelling|denoting|referring to|identifies|names|whatever-jar-wants-to-call-it} a resource

<hhalpin> timbl: how can you tell? maybe you can look at the diff?

<jar> "on the web"

<hhalpin> jar: you need to relate this abstract realm to the real realm, the realm "on the Web"

<jar> G is "on the web" at U

<jar> (according to some S)

<hhalpin> jar: "on the web" is a relationship

<timbl> G is is_on_the_web_at "http:/..."

<hhalpin> jar: moby dick is "on the web" at some particular URI

<alanr> x on the web =def x exists(u) x on the web at u

<hhalpin> jar: this becomes a falsiable statement

<jar> G is "on the web" at U (according to S)

<timbl> <http://www.gutenberg.org/etext/2701> dc:title "Moby Dick".

<Stuart> jar: can you explan how it is falsifiable?

<jar> you can do a GET and look at the result, and check to see if it's a rep of G

<alanr> some said we agree on the semantics of "on the web at U". No we are trying to understand what that could mean.

<Stuart> ok.... but that goes out of the realms of the 'formal' system and relies on a human inspection and a human judgement.

<timbl> <http://www.gutenberg.org/files/2701/2701.zip> dc:title "Moby Dick".

<hhalpin> mhausenblas: representatioin is *current* state of resource

<jar> X is a rep of G is shorthand for X is a rep of G in its current state (or some state)

<hhalpin> timbl: the sense of the statements may be different, and when I give you that URI, I *think* you are getting back a reprsentation of what I wanted

<timbl> <http://www.gutenberg.org/files/2701/2701.txt> dc:title "Moby Dick, or, the whale by Herman Melville".

<timbl> 1.2MB

<timbl> thank you

<jar> "X is a rep of state of G" not the same as "X is a rep of G in its current state" since latter allows the rep to contain stuff that's part of G but not part of its state

<jar> mhausenblas, next steps?

<hhalpin> timbl: so we can put representations on the web, then we can talk about how we can use it, but if we stop and say "what if someone hacked into server?"

<hhalpin> timbl: about the error case

<hhalpin> timbl: then the models becomes much messier

<hhalpin> timbl: we have to figure out who owns domain, and all of these things.

<mhausenblas> +1

<hhalpin> timbl: but let's focus ontology about what happens when someone does it right

<jar> +1

<hhalpin> so, we can assume that the representations are jar:consistent, i.e. that the representations describe the resource.

<hhalpin> ok, here's an example.

<jar> except, "denotes" and "identifies" are troublesome

<hhalpin> if the resource U = "Moby Dick" and HTTPRequest(U) = "picture of blue cheese" therefore, we are in an error state.

<hhalpin> To attempt to remove human agency in some attempt to formally define "representation" is doomed.

<hhalpin> We have to assume humans can communicate about the world using languages.

<hhalpin> jar: we have to assume the premise that people can both identify and communicate about Moby Dick

<hhalpin> timbl: they can look at it and say "yes, it's moby dick"

<hhalpin> note that in philosophy of language, this is called defining a sense or content of a representation via assertoric content.

<hhalpin> see Michael Dummett

<hhalpin> alanr: so we need two people

<hhalpin> jar: we don't need people, want to define consistent

<hhalpin> alanr: we are trying to use reverse reasoning about what resource could be.

<hhalpin> jar: it doesn't constrain them.

<Stuart> jar: would a formulation that involved two URI be clearer - are the wa-representations available using U1 consistent with the wa-representations available at U2 - ducking the problem of any out of system way or referring to the resource - eg. you were using the name Mobey Dick as a non-URI way of referring to the resource.

<hhalpin> that JUMP is justified.

<hhalpin> Sorry, that's how language works.

<hhalpin> we have to assume people can identify and use language to denote things to accomplish things.

<timbl> No it is 60 mins

<hhalpin> if we can't make that jump, then we are stuck.

<hhalpin> Please see Dummett on Frege or almost any work in philosophy of language after 1950.

<hhalpin> mhausenblas: 9th of June for next telecon

<alanr> I'll be abroad. Can't figure out what time it will be

<hhalpin> yes

<hhalpin> i'd like to review IRW ontology

<alanr> harry - make sure all terms have definitions and that they are not circular

<hhalpin> alanr - all definitions are circular to some extent

<alanr> cop out

<hhalpin> that's Searle

<hhalpin> No, it's true.

<hhalpin> Definitions come in groups of related concepts

<hhalpin> that's not a cop-out.

<alanr> excuse for not being clear

<hhalpin> you can't understand a concept without other.

<hhalpin> that's the recursive dictionary problem.

<jar> harry +1

<alanr> there's something called the world

<hhalpin> otherwise you try to ground out in "elementary sense data"

<mhausenblas> [adjourned]

<hhalpin> which is an argument that has been proven incorrect again and again.

<hhalpin> see recent work in cognitive science or even just Wittgenstein's duck/rabbit argument.

<alanr> in any case a definition that refers to its own term isn't building on anything

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2009/05/26 14:06:21 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.135  of Date: 2009/03/02 03:52:20  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/"Moby Dick"/"Moby Dick, or, the whale by Herman Melville"/
Succeeded: s/NO/No/
Found ScribeNick: hhalpin:
WARNING: No scribe lines found matching ScribeNick pattern: <hhalpin\:> ...
Inferring Scribes: hhalpin:

WARNING: 0 scribe lines found (out of 323 total lines.)
Are you sure you specified a correct ScribeNick?

Default Present: jar, DavidB, mhausenblas, HarryH, Stuart, alanr, Timbl
Present: jar DavidB mhausenblas HarryH Stuart alanr Timbl
Agenda: http://lists.w3.org/Archives/Public/public-awwsw/2009May/0008.html
Found Date: 26 May 2009
Guessing minutes URL: http://www.w3.org/2009/05/26-awwsw-minutes.html
People with action items: 

[End of scribe.perl diagnostic output]