Semantic Web Interest Group IRC Chat Logs for 2001-09-24

NOTICE: #rdfig logs are being turned off 2004-12-03. Please switch to the new and shiny #swig channel for Semantic Web Interest Group chat. Change your client to #swig and enjoy the new experience. Or read the latest #swig logs to see what you've been missing :)

01:10:07 <AaronSw>http://www.agfa.com/w3c/n3/rdfsrules.n3

01:10:09 <dc_rdfig> A: http://www.agfa.com/w3c/n3/rdfsrules.n3 from AaronSw

01:10:23 <AaronSw> A:|Rules for RDFS Entailment in N3

01:10:23 <dc_rdfig> titled item A

01:10:27 <AaronSw> A:From Jos de Roo

01:10:27 <dc_rdfig> commented item A

01:10:35 <AaronSw> A:Based on the forthcoming RDF Model Theory

01:10:36 <dc_rdfig> commented item A

01:40:11 <tim> Hmmm.. is A a complete list? Knowing Jos it is. Interesting how he labels the rules.

01:41:37 <AaronSw> A is pretty much a translation of the Model Theory spec's rules into RDF.

01:41:47 <AaronSw> I'm pretty sure it's complete.

01:42:06 <AaronSw> Altho, he just suggested 3 new ones saying that any terms used in a statement are resources.

01:42:45 <sbp> A:q.v. [http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/0323|Re: A proposal for entailment tests]

01:42:46 <dc_rdfig> commented item A

01:46:27 <AaronSw> A: see also [three more suggested rules|http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/0329.html]

01:46:27 <dc_rdfig> commented item A

01:47:00 <sbp> Get rid of that ".html", mister!

01:47:17 <AaronSw> i'm sorry! it's that silly hypermess!

01:47:33 <AaronSw> tim: "interesting how he labels the rules" yeah, i thought similarly... we should get the MT spec to give them URIs.

01:48:03 <sbp> I think "aaa" and "xxx" are confusing variables

01:49:47 * bijan hates the extra space before and after delimiters.

01:50:10 <bijan> { :RULE1 a rdfs:Resource

01:50:17 <bijan> ^ evil!

01:50:26 <sbp> why so?

01:50:41 <bijan> Oh, merely my aesthetic biases.

01:50:49 <bijan> I.e., I find it harder to read.

01:51:09 <bijan> Like that horrid java style thinkg ( putting extra space )

01:51:20 <bijan> As opposed to (putting extra space)

01:51:24 <sbp> I find it easier to read

01:51:36 <bijan> No, you just *think* you do.

01:51:43 <bijan> I assure you that you're completely mistaken! :)

01:52:34 <bijan> (Obviously, I'm joshing, but I will add that I literally do not believe you :))

01:53:54 * AaronSw is certain bijan uses: function(arg1,arg2,arg3) in his Python programs.

01:54:18 <sbp> Yeah!

01:55:03 <bijan> Hmm. No, I try to put a space after the commas.

01:55:07 <bijan> But otherwise, yes.

01:55:24 <bijan> As God herself, in all her blazing glory, specifically told Guido to do.

01:55:51 <bijan> Why he did not write it into the grammar, I shall never know! *Clearly* Satan's work.

01:56:45 <bijan> It's prolly a good think I don't do C, or *surely* I'd have wasted large chunks of my life on Indentation Wars.

02:02:17 <AaronSw> Heh.

02:02:23 <AaronSw> I'm still wondering why:

02:02:26 <AaronSw> function (arg)

02:02:28 <AaronSw> is legal.

02:03:11 <bijan> In prolog it's not!

02:03:16 <bijan> Or rather it means something else.

03:13:24 <sbp> Aaron, would you prefer anonymous nodes outputted by CWM to be identified with some unique ID type namespace, the URN-anon thing I was hacking with, or something else?

03:13:51 <AaronSw> I think that they should probably be treated as plain old RDF anonymous nodes.

03:14:10 <sbp> But what about anonymous predicates?

03:14:20 <tim> So I am left wondering whether to build this log:forAll :x. {} log:implies { :x a rdf:Resource }. into cwm so that I can use Jos's rules directly -- or is the whole idea that you have to turn on each rule - you have control of them. I didn't notice this axiom in his list.

03:14:37 <AaronSw> tim, see the pointer in the chump

03:14:46 <AaronSw> it was an axiom that was accidentally left out.

03:14:46 <tim> Plain old anon nodes in RDF can olny be used for certian cases.

03:14:58 <sbp> exactly, N3 is more expressive

03:15:02 <AaronSw> Well, you simply can't express those other cases in RDF. So why bother trying?

03:15:11 <sbp> I lean towards making a new UUID style namespace for them

03:15:35 <sbp> why bother trying: because they should be expressable, it is useful to have anonymous predicates

03:15:42 <sbp> :Dog [ daml:inverseOf rdf:type ] :Fido .

03:16:09 <AaronSw> I agree it is useful, but you're not going to fix the RDF/XML syntax.

03:16:14 <sbp>http://lists.netsol.com/cgi-bin/wa?A2=ind0109&L=urn-ietf&D=1&T=0&O=D&F=&S=&P=247

03:16:14 <dc_rdfig> B: http://lists.netsol.com/cgi-bin/wa?A2=ind0109&L=urn-ietf&D=1&T=0&O=D&F=&S=&P=247 from sbp

03:16:16 <tim> missing rules: ah i see.

03:16:30 <sbp> B:|URN NID Request (URN-x/anon)

03:16:30 <dc_rdfig> titled item B

03:17:20 <sbp> B:A set of identifiers that are bound to existentially quantified variables

03:17:21 <dc_rdfig> commented item B

03:17:24 <tim> N3 is more expressive but still there are cases (as there must be I suspect in any grammar) when you just can't use the syntax to express the graph without making up anon node names

03:17:59 <sbp> statement and context IDs come to mind... they are nearly always anonymous

03:18:58 <tim> You can have two graphs each of which can be expressed in RDF with anon nodes but when you merge them you need a temp name for the node.

03:19:36 <sbp> avoiding terms used by the author won't work (what if the author invents some new ones), and getting RDF Core to reserve terms is backwards incompatable, and a bit hacky

03:19:43 <tim> An anonymou snode can have 0 or 1 incoming links - it can be the object or 0 or 1 statements.

03:20:07 <tim> So how about just using _:gnnn ?

03:20:27 <sbp> In the XML RDF output? But that's not a URI

03:20:30 <tim> In other wards a "namespace" which in fact does *not* map to any URI ?

03:20:48 <tim> No, its not a URI - but then the node doesn't have a URI.

03:21:10 <tim> (which is sort of where we came in)

03:21:12 <sbp> Well, why not just create a URN namespace for it? That's what I proposed; it seems sesible (i.e. doesn't contravene any Web axioms) to do so

03:21:45 <tim> And explicitly give the existential quantification.

03:22:22 <sbp> yes. You always know that those URI identify existentially quantified resources

03:22:27 <tim> It woul dbe bad to lead people to think that the node could be referenced. A URN might tempt people to point to it.

03:22:48 <sbp> yes, that's a drawback...

03:23:05 <tim> Are you happy with the NTriples situation?

03:23:41 <sbp> Am I happy with "_:x"? Not really. Well, it works, but I'm not jumping up and down about it

03:23:54 <AaronSw> Wy don't you want folks referencing the node?

03:23:57 * tim wonders where spb is on gfthe planet

03:24:01 <AaronSw> s/Wy/Why/

03:24:10 <sbp> I'm in England

03:24:26 <sbp> It's 4:23AM here, so apologies if I'm a little incoherant!

03:24:45 <sbp> You can't reference the node as a node

03:25:02 <sbp> er... viz. you can't say that g1 in one document is the same as g1 in another document

03:25:07 <tim> I wondered why you were coherent [at all] ;-)

03:25:14 <sbp> because they're scoped *to* the document

03:25:18 <sbp> heh :-)

03:25:28 <tim> "A brown haired woman at the meeting said it was raining"

03:26:16 <sbp> <urn:anon:g1> :hairColor "brown"; a :Woman; :attended :Meeting; :said { :CurrentWeather :state :Raining } .

03:26:20 <tim> If there was a URN one would be tempted to refer tothat woman, irresepctive of the fact that there were/may have been 50 brown haired women at hte meeting, 8 of which said it was raining.

03:26:43 <sbp> v. good example...

03:26:52 <AaronSw> Why can't I refer to the woman, though?

03:26:59 <tim> What woman?

03:27:05 <tim> Which woman?

03:27:06 <AaronSw> The one the author was describing.

03:27:20 <sbp> The author said "a", not "the"

03:27:21 <tim> The author saifd such a woman existed. Ther may have been many.

03:27:26 <sbp> exactly

03:28:15 <AaronSw> But was he not describing one woman?

03:28:33 <tim> [ a contact:Person; :a :LittleInCoherent; irc:on #rdfig].

03:28:52 <tim> No, he was saying that that such a woman existed.

03:29:41 <AaronSw> And why should that prevent me from saying more about this existing woman?

03:29:52 <sbp> but of course, such URNs are still persistent and properly transcribable identifiers... the urn urn:anon:woman always gives the node "woman" as an existentially quantified variable, scoped to the bounds of the document in which it is used. But people will see it as being an actual label for one woman, in the same way in which you confused "the" for "a"

03:30:12 <AaronSw> Well, I meant "the" "a" woman.

03:30:20 <sbp> and they will probably ignore the bounds of the quantification, too

03:30:32 <tim> I'm leaning toward _:gnnn as it avoids the question, he need fro an arbitrary choice here. I'm also headed for bed as unlike some I don't work though the night well.

03:30:39 <tim> tim is now known as a_tim

03:30:41 <AaronSw> Like, "I'm thinking of a woman with brown hair." "Does she have blue eyes?"

03:30:48 <AaronSw> a_tim: ;-)

03:30:50 <sbp> lol @ a_tim!

03:31:16 <sbp> [ a :Bed ]: I should probably lean towards the same

03:31:39 * a_tim night

03:31:52 <AaronSw> G'nite.

03:32:02 <sbp> 'night. I'll compose a better email in the morning...

03:34:01 <AaronSw> Bye.

03:36:04 <myipaqdon> hello

03:43:31 <sbp> Gotta run

06:48:04 * chaals waves

09:13:14 <danbri>http://hans.breuer.org/ports/

09:13:14 <dc_rdfig> C: http://hans.breuer.org/ports/ from danbri

09:13:35 <danbri> C:|Win32 port of GTK Python extensions

09:13:35 <dc_rdfig> titled item C

09:13:45 <danbri> C:Did I need any more incentives to learn Python?

09:13:45 <dc_rdfig> commented item C

09:30:59 <dajobe> lovely; abiword loads Psion documents natively...

10:21:32 * danbri waves

10:21:44 <dmiles> good morning :>

10:21:47 <dajobe> hey dan

10:22:04 <danbri> hello

11:00:48 <sandro> 'morning....

11:01:29 <danbri> hiya

11:17:00 * chaals waves

11:17:21 <libby> chaals, can I aska quick question? dunno if this is your thing....

11:17:32 <chaals> yep...

11:18:14 <libby> dyou know if it's possible to guarantee that a xml schema name plus an element name will give you a unique key?

11:18:38 <chaals> The namespace plus name is supposed to give you a unique thing.

11:18:45 <chaals> (that's why it has URI syntax).

11:18:46 <libby> like, you can definitely say that there aren't two titles within a document?

11:18:57 <libby> I was wonderign that but couldn't find anything :(

11:18:58 <chaals> Errr...

11:19:12 <chaals> It is an error to have two things with the same ID in a document

11:19:26 <libby> so within a schema the namespce plus name is unique

11:19:29 <libby> ok

11:19:30 <chaals> (i.e. two elements who have an attribute of type ID, where the value is the same)

11:19:44 <chaals> yes. But that doesn't mean you can't have two titles.

11:19:54 <danbri> an occurance of an element, not the element conceived of in that abstract, right?

11:20:13 <chaals> html:title (where the html is expanded to the namespace) can only occur once in a document or it is not valid.

11:20:24 <chaals> but svg:title can occur all over the place.

11:20:31 <libby> in an instance document title could appear more than once but in a schema doc it couldn't?

11:20:42 <libby> ok, unless the schema doc used more than one namespace

11:20:42 <chaals> Oh, I see.

11:20:54 <chaals> I don't know, you would have to check the schema spec.

11:21:13 <chaals> I suspect it might be an error, because it would be painful to deal with multiple appearances.

11:21:25 <libby> bah, hoping not to ave to do that! thanks chaals

11:21:33 * chaals could go have a look....

11:21:53 * libby shoudl stop being so lazy and read the spec

11:21:58 <danbri> but this isn't really about XML Schema; its about instance data...

11:22:15 <danbri> ...which might be defined using a different XML schema language, or non at all, ie. just be wellformed.

11:22:17 <danbri> I thought.

11:22:19 <danbri> lib?

11:22:31 <libby> I don't know any more.... :(

11:22:45 <libby> I think so.

11:25:28 <libby> basically, I want to be able to create an RDF relationship between a title (say) in an XML document and a concept defined in RDF

11:26:07 <libby> so I want to be able to refer to (and point at using a URI) that title element using RDF

11:26:30 <danbri> bit like an... annotation?

11:26:38 <libby> er, yeah, I guess

11:26:58 <libby> the reason I asked you chaals is because you talked about doing somethign similar with EARL I think once

11:28:13 * danbri considers registering http://www.omnigroup.com/applications/omnigraffle/ for RDF Schema spec image editing

11:29:03 <chaals> Oh.

11:29:09 <chaals> So for example:

11:29:11 <chaals> .....

11:29:14 <libby> uhuh

11:29:25 <chaals> <title id="foo">the title!</title>

11:29:27 <chaals> ....

11:29:33 * libby finds this stuff ver difficult to explainfor some reason

11:29:33 <chaals> somehwere in an XML doc.

11:29:38 <libby> yep

11:30:05 <libby> and then the uri of the element would be xmldoc.xml#foo?

11:30:15 <danbri> confusing: it's the terminology. "element" sometimes gets used to mean _the_ thing html:title, sometimes _an_ instance of html:title in a doc.

11:30:16 <chaals> And you want to say <x:about resource="http://example.new/exam#foo">....

11:30:35 <chaals> but if you have

11:30:45 <chaals> <html xmlns="blah">

11:30:48 <chaals> <head>

11:30:49 <libby> I think I want to say stuff about all titles in the document, which was why I was tyalking about referencing it within an schema document

11:31:06 <chaals> <title>You didn't update this, you drongo</title>

11:31:09 <chaals> </head>

11:31:13 <chaals> <body>

11:31:14 <chaals> ...

11:31:30 <chaals> then there is only one title, and you could happily refer to it with an Xpath or Xpointer

11:31:56 <libby> i.e. in your example?

11:32:16 <chaals> example 2 (above)

11:32:42 <chaals> so example 3:

11:32:57 <chaals> an SVG doc that has an outline of each of libby and chaals and dan pasted over a background jpg.

11:33:03 <chaals> and each of those has a title already.

11:33:16 <chaals> and you want to say something about all the titles in the document at once.

11:33:21 <libby> danbri, yeah the_ thing html:title i what I want

11:33:24 <libby> yeah

11:33:37 <libby> I think so. I'm rtying to say somethigna bout the meaning of title

11:33:50 <chaals> example 4:

11:34:24 <chaals> for all HTML documents, the title is <foo/>

11:34:34 <chaals> (where foo is some whole statement)

11:35:13 <libby> yeah, that's it

11:36:09 <chaals> Aaah. So what you want to do is annotate the schema for HTML (or whatever)

11:36:21 <libby> yeah, that's it

11:36:42 <libby> (sorry it was so painful to arrive at what I wanted!)

11:37:07 <deltab> hmm, that's what I thought you wanted: sorry

11:37:59 <chaals> well, if I was smarter I would have figured it out quicker.

11:38:28 <chaals> Anyway, I looked for that in the schema spec - is there only one html:title definition?

11:38:43 <chaals> Actually there is a thing where you can redefine or extend something.

11:39:08 <chaals> It might be that you can add an annotation by <redefine>ing the element.

11:39:18 <danbri> Ah, right: I thought initially you wanted instance level qualification

11:39:21 <libby> ok...

11:39:22 <danbri> This is easier

11:39:26 <libby> is it?

11:39:30 <danbri> Yep.

11:39:49 <danbri> Take a look at the Meaning Definitional Language spec I have printed out around here someplace. It's like a fancied up schematron.

11:40:04 <libby> started out thinking it was simpler then changed my mind

11:40:58 <libby> I do want to be able to get from an instance document element to the identifier for the schema element, and the similer it is to do that the better

11:41:08 <libby> s/similer/simpler

11:42:33 <libby> ok, will do

11:42:46 <danbri> yeah, that's the tricky bit

11:42:58 <chaals> so: start at the eelement, dig up its schema, make an annotation on the definition of the element. Only breaks if the element declaration is distributed through the schema, or repeated?

11:43:26 <libby> I think so...

11:43:37 <libby> I don't want to make an annotation though, but search annotations

11:44:59 <libby> it would be better if I could actually give a uri to the definition of the element in the schema, but it doesn't really need one if I can construct a unique identiofer for it using the element name and the schema uri

11:45:39 <libby> I guess treating an xml schema a bit like an RDF schema

11:46:54 * libby starving...needs food now

11:47:05 <libby> thanks for all your help guys :))

11:47:22 <libby> very grateful

11:47:23 <libby> :)

11:47:37 <chaals> aaah, I think it is easy again then.

11:47:42 <chaals> go eat and talk after...

11:57:25 <libby> back

11:57:36 <libby> if you'rte not too busy

12:04:56 <dmiles> i was thinking of a use for the prolog in webbrowser.. i was disconnected from the net using a cached version of the jdk/html and wanted to "seach" for something

12:05:11 <dmiles> err "search"

12:05:25 <chaals> me, busy?

12:09:59 <libby> you're never busy chaals ;)

12:10:19 <libby> so what's the answer then? "<chaals> aaah, I think it is easy again then"

12:13:19 <sbp> [[[

12:13:19 <sbp> 11:25:28 <libby> basically, I want to be able to create an RDF relationship between a title (say) in an XML document and a concept defined in RDF

12:13:20 <sbp> 11:26:07 <libby> so I want to be able to refer to (and point at using a URI) that title element using RDF

12:13:21 <sbp> ]]]

12:13:30 <sbp> [sound of me falling on the floor crying]

12:14:11 <libby> why? :(

12:14:19 <libby> don't cry sbp

12:14:51 <sbp> the problem is, namespaces are partitioned, so it's difficult, if not impossible to identify XML QNames in RDF. Unless you create a model for namespace partitions, and because languages each use their own version of namespace partitioning...

12:15:06 <libby> ah...

12:15:16 <sbp> For example, there are *two* titles in XHTML

12:15:34 <sbp> <title>My title</title>[...]

12:15:41 <libby> yep

12:15:56 <sbp> The element title and the attribute title don't nececssarily have the same semantics

12:16:11 <libby> no, I get you

12:16:18 <sbp> But, when you use the XHTML namespace in RDF, you get the same URI for both. Ugh

12:16:25 <sbp> So you have to model it...

12:16:43 <chaals> this is why you need to dig into the schema, where oe is declared as an element andd one is declared as an attribute.

12:16:50 <libby> right, you've tried to explain this to me before, soz for being slow

12:16:55 <chaals> (and into the group, so they don't do that again....)

12:16:55 <sbp> [ :namespace <http://www.w3.org/1999/xhtml>; :att "title" ] .

12:17:17 <sbp> yeah!

12:17:46 <sbp> well, it's not their fault, and it's not even really up to them to define URIs for the terms. Although they could have, and arguably should have done

12:18:27 * sbp isn't sure that he has been over this for a while at least

12:18:59 <libby> so, could I use id on the elements defining the elements I'm interested in withing the schema/

12:19:05 <libby> ?

12:19:52 <sbp> Um... you can point into the XML Schema defining the language that you're working with, and then invent a property to say "the element/attribute defined by"

12:20:06 <sbp> Using id: that's another cool approach

12:20:30 <sbp> but it only works if the XML Schema in question has IDs defined :-)

12:20:35 <chaals> right.

12:20:37 <libby> right. what's the best way of pointing into the schema? is id ok, or do I have to use xpath?

12:20:40 <libby> ok, right

12:20:45 <chaals> id is great if it is there.

12:20:48 <sbp> if it has IDs, use them...

12:20:49 <libby> I have control over the schemas I hope

12:20:50 <sbp> yeah

12:20:52 <libby> cool

12:21:02 <chaals> or you can build a schema where you redefine each element in turn, by adding an ID i there isn't one.

12:21:10 <chaals> You coul probably do that with an XSLT hack.

12:21:13 * sbp wonders if xsd:id is an XML ID... should be

12:21:19 <chaals> (even I managed to figure out some XSLT to do that ;-)

12:21:23 <sbp> yeah, cool idea!

12:21:27 <libby> I totally owe you guys beers/britvic55

12:21:30 <danbri> yes, I was going to suggest that too; transform the schema into a good one that's similar

12:21:38 <sbp> Ooh, britvic55 :-)

12:21:41 <chaals> libby, are you going to be in brighton?

12:21:48 <sandro> britvic55?

12:21:53 * chaals promises to drink Sean's beer ;-)

12:21:55 <libby> that's the plan, though not done anything aqbout it yet

12:21:56 <sbp> :-)

12:21:59 <libby> heh

12:22:26 <libby> britvic55=organge and lemonade I think

12:22:40 <libby> yeah, it should e cool

12:22:48 <sandro> Maybe I wont chump http://www.britvic.co.uk/home.asp

12:23:00 <sbp> Yep. It confuses bartenders, since they haven't made/sold the drink for ages

12:23:33 <libby> heh

12:23:37 <danbri> s**t, taxi at door

12:23:42 * danbri rushes off

12:23:52 <libby> sbp, they want us to make changes to anntations paper by wed!

12:23:52 <sbp> c'ya DanBri

12:23:55 <sbp> aha:-

12:23:55 <libby> see ya

12:23:56 <sbp> [[[

12:23:56 <sbp> Britvic 55

12:23:57 <sbp> A beautiful drink, and amazingly cheap in certain parts of Cambridge. Although strictly not a beer, it has a head to warrant an ale-like ranking on this page. 55% OJ gives it its name, of course, the remaining 45% coming from lemonade. Gorgeous.

12:24:01 <sbp> ]]] - http://www-stu.christs.cam.ac.uk/~grb26/breviews.htm

12:24:08 <sbp> by Wednesday! Aaargh!

12:24:20 <libby> I think we eed to ask for moere timr - phil's away at the moment

12:24:27 <sbp> BTW, congrats on that...

12:24:39 * sbp isn't sure if he emailed or not... gmgpgmgh

12:24:45 * chaals waves goodbye to Dan's departing back

12:25:37 <libby> no probs, congrats yourself.

12:27:03 <sbp> then again, who could refuse a paper from us? eh?

12:27:55 * sbp wonders where the Web would be without Google

12:29:23 <sbp> Wow...

12:29:29 <sbp> chaals, you still there?

12:29:29 <libby> h'm welll plenty of places refuse my papers, so it must be yours and phil's good influence :)

12:29:37 <sbp> :-)

12:29:38 <sbp> [[[

12:29:40 <sbp> NEW! Anne Pemberton declares the Web AccessiBlog not a useful site. That means its just the ticket for everybody else!

12:29:47 <sbp> ]]] - http://www.joeclark.org/accessiblog/

12:29:57 <sbp> Bizzare

12:30:01 <libby> weird

12:31:38 <sbp> they have a somewhat long running feud...

12:32:38 * chaals is hiding

12:32:59 <chaals> well, if either of them would grow up a bit life would be better for the rest of us.

12:33:21 <chaals> (Anne more so, since she's ben there longer, and Joe more so, since he claims to be an expert)

12:33:29 * chaals goes back to hiding

12:33:38 <sbp> I fully agree

12:33:46 <chaals> what, that I should hide??

12:33:53 * chaals goes to have a cry

12:33:58 <sbp> Don't worry chaals, no one can prove that it's you on IRC :-)

12:33:59 <chaals> s/cry/coffee/;-)

12:34:08 <sbp> s/coffee/beer

12:34:10 <sbp> :-)

12:37:51 <chaals> no, I am in France.

12:39:29 <sbp> Does that make a difference? "un bier, c'il vous plait"

12:42:12 <sbp> Hmm... ERT call soon

12:42:43 <libby> sbp, have you got your allthe time internet access now?

12:43:11 <sbp> nope

12:45:05 * chaals is back.

12:45:10 <sbp> wb chaals

12:45:14 <libby> whoo!

12:45:17 <chaals> Yep, the difference is that the biere isn't that good here.

12:45:33 <chaals> Hey Sean, did you go back to the community centre and claim your 2 quid?

12:45:42 <sbp> Oh crap

12:45:46 * em_lurk wonders if danbri is still with chaals?

12:45:53 <chaals> nope, we sent him home.

12:45:54 <em_lurk> em_lurk is now known as em

12:46:02 <sbp> Hey em

12:46:14 <sbp> How's the Primer coming? Finished yet? :-)

12:46:15 :) whether he liked it or not, eh?

12:46:30 * em waves and pokes sbp at the same time

12:46:35 <sbp> lol

12:47:12 <chaals> well, he seemed to like it.

12:47:18 outline and set of relavant documents to be sent by tomorrow

12:47:22 <chaals> He's coming back in a minute or shortly

12:47:29 <chaals> well, a couple of weeks anyway.

12:47:59 from there, i'm planning a teleconferece of the participants to make sure we're on same page, and solicit ownership on pieces/parts of the primer

12:48:31 chaals, do you know early oct dates he

12:48:36 * sbp wonders what the projected timescale for this is

12:48:37 's planning on being back there?

12:50:17 <chaals> 3-5

12:50:29 <chaals> well, early morning of the 4th - 5 I think.

12:50:40 <chaals> Although he might be shifting that back a day to be 3-5

12:50:52 * chaals will be bacck here on the 4th

12:50:52 <sbp> Ooh, understatement of the week:-

12:50:55 <sbp> [[[

12:50:56 <sbp> The February 2001 Recommendation date for XHTML 2.0 may seem optimistic, but it certainly sets a target.

12:51:02 <sbp> ]]] - http://www.xmlhack.com/read.php?item=282

12:51:03 sbp, timescale is fairly quick... i'd like to see something back to wg by mid-late oct... my concern is the larger community might think we're overcomplexifying something that was already considered overly complex, i'd like to have the primer availiable to help address these concerns

12:51:40 <sbp> Yes. The new MT is going to be very dense for people to wade through

12:51:43 3-5, ok... thats what I was looking for him for...

12:51:48 <sbp> mid-late oct: that's cool

12:52:20 sbp, time frame something you'd be willing to consider contributing to?

12:52:37 sorry... given the time frame, is this something you'd be willing to consider contributing to?

12:52:49 * em can't type quick enough

12:53:04 <sbp> Yeah, I'll help as much as I can/you want

12:53:30 <sbp> Hmm...

12:53:31 <sbp> [[[

12:53:32 <sbp> The value of XHTML 2.0 will be the change to XLink from HTML's

12:53:32 <sbp> linking syntax. Still modular, probably with some more complex

12:53:32 <sbp> events and forms.

12:53:37 <sbp> ]]] - http://lists.w3.org/Archives/Public/www-html/2000Feb/0125

12:54:00 thanks sbp

12:54:12 <bijan> For example, almost all beginners to RDF go through a sort of "identity crisis" phase, where they confuse people with their names, and documents with their titles. For example, it is common to see statements such as:-

12:54:13 <bijan> <http://example.org/> dc:creator "Bob" .

12:54:24 <sbp> you may as well just stick swintro/semanticWeb-long together with a bit of blu-tack and be done with it :-)

12:54:26 <bijan> I have a retort to that ;)

12:54:45 <sbp> Yeah! That annoys me quite a bit

12:54:49 <bijan> First, if you're going to make this criticism, then you should *slam* the RSS 1.0 DC module.

12:54:56 <bijan> Which explicitly mandates this behavior.

12:55:05 <sbp> ah, and DC actually says to use the persons name

12:55:20 <bijan> However, Bob is just a literal string, so how can a literal string write a document?

12:55:20 <jang> there is no reason to slam the behaviour...

12:55:22 <sbp> but it doesn't state that you explicitly have to, which I think is an error with DC

12:55:27 <bijan> But *that's* wrong.

12:55:31 <jang> ...providing your definition of dc:author is explicit

12:55:37 <jang> ie, "it doesn't do what you think it does"

12:55:40 <sbp> the definition is not explicit

12:55:41 <bijan> you're making the use/mention confusion.

12:55:58 <bijan> I hope.

12:56:05 <bijan> So, here's how I get round that problem.

12:56:07 <jang> no...

12:56:11 <sbp> Go read the Dublin Core schema, it says that the property is usually used to indicate the name of the author. That sucks

12:56:18 <sbp> [period]

12:56:35 <jang> go on.

12:56:36 <sbp> people are not names, names are not people

12:56:51 <sandro> It's simple and to the point. Make another schema for object-based document metadata. The rules linking the two may even be somewhat clear.

12:56:51 <jang> you invent a dc:author' that does what it ought to

12:56:52 <bijan> First, i don't think it's a *confusion* per se...we often use a variety of ways to refer to objects, and a variety of indirect ways to attribute things to them.

12:57:02 <jang> and a person's-fullname that does what it ought to

12:57:03 <sbp> if you're going to define a property that says "this points to the *name* of the creator", then you should do that

12:57:07 <jang> and compose them

12:57:40 <sbp> it's O.K. for creating sub properties, but I don't think that it has any value as a property as-is

12:57:42 <bijan> Granted, in a regimented language, we might want things more explicit, but there's a big tradeoff with such langauges...they often are very hard for pepole to use.

12:58:05 <bijan> Ok, so here's my desparate attempt to save string literals...

12:58:17 <sandro> {$doc dc:author $name} <=> { $doc sbp:author $author; $author sbp:name $name}

12:58:19 <bijan> First, observe that my *name* isn't a string literal either.

12:58:43 <bijan> or, at least, not necessarily. It can be *recorded* or *encoded* as a string literal.

12:58:46 <sbp> { { :x dc:creator :y . :y a rdfs:Literal } log:implies { :x :writtenBy [ foaf:name :y ] } } a log:Truth; log:forAll :x , :y .

12:58:50 <bijan> Or perhaps, *named* by a string literal.

12:59:24 <bijan> This suggests that perhaps we should record string literals as an 'escape' mechanism.

12:59:38 <sbp> you have to add the conditional that it only applies when y is a literal... because people can still use other resources as the ranges of dc:creator

12:59:48 <sbp> s/ranges/objects

12:59:51 <jang> bijan: thank goodness!

13:00:11 <sbp> viz., dc:creator rdfs:range rdfs:Resource . AFAICT

13:00:15 <jang> your name isn't a string literal... at last, some sense!

13:00:23 <sandro> Ah yes -- predicates based on the LOGICAL SYMBOL instead of the object. (ie whether it's a literal or not -- that the symbol, not the object.)

13:00:26 <jang> there's nothing wrong with having

13:00:35 <jang> <som object> xxx:name "bijan"

13:00:43 <sbp> A name is a string literal

13:00:44 <bijan> Thus, how to interpret a string literal is application dependant.

13:00:57 <jang> providing you understand that "bijan" is just a rendering into text form of a more abstract literal notion

13:01:22 <sbp> A name make a very good string literal. Trying to derive any information from that in context is like trying to process natural language (well... it actually is), pointless, and not what the SW is about

13:01:37 <sbp> yeah, it's the lexical value of bijan's name

13:01:53 <bijan> right, encode it as a string literal, of a sort which says "don't sub parse this"

13:02:14 <jang> the fact that our serialisations are unable to capture literals explicitly (we have to use serialised forms of them) is just a fact pof life

13:02:16 <bijan> but whether it's to be used as the textual representation...or as my *name* is app dependant.

13:02:20 <jang> but one that needs making explicit

13:02:27 <jang> but only if you want to confuse people :-/

13:02:41 <sbp> hmz

13:03:21 <jang> excuse random typing... fingers falling off from the cold at the moment

13:03:24 <sbp> RDF Core have yet to really define what the structure of a Literal is

13:03:44 <jang> we could just remain agnostic on that

13:03:53 <jang> (from a MT point of view, at least)

13:03:54 <sbp> I don'

13:03:58 <sbp> t think you should

13:04:07 <bijan> So, take a less, or perhaps more, contentious example.

13:04:16 <sbp> you have to let people know, for example, how to give the natural language of a string in context

13:04:31 <bijan> <http://foo.com/> dc:date "9/25/2001".

13:04:37 <sbp> [ xml:lang "en"; rdf:value "word" ] . # perhaps

13:04:43 <bijan> On your reading this is saying that a date is a string literal.

13:04:51 <jang> no!!!

13:04:51 <bijan> But *clearly* that's not what it's saying.

13:05:01 <sbp> DC should define a datatype for dc:date...

13:05:07 <bijan> And equally clearly it's not implying it.

13:05:56 <jang> I wrote something to that effect recently : http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/0212.html

13:06:03 <jang> perhaps not brilliantly articulate

13:06:09 <bijan> And an application profile could say "object of a dc:date must follow spec yadda yadda"

13:06:19 <bijan> Which, in fact, the RSS DC module *does* do.

13:06:47 <sbp> Pff, that information should be in the schema for DC. DAML makes that easier

13:07:06 <bijan> Sean: if I understood you, i don't see how that helps your case.

13:07:11 <bijan> What do you mean by "that information"?

13:07:29 <jang> it's handy for a serialisation to specify a format for representing a literal value

13:07:39 <jang> but the representation _isn't_ the value itself

13:07:47 <sbp> [[[

13:07:48 <sbp> In other words, I tend to draw a big

13:07:48 <sbp> box around the

13:07:48 <sbp> _:a <rdf:type> <xsd:Date> .

13:07:48 <sbp> _:a <rdf:value> "2001-09" .

13:07:48 <sbp> and call all of that a literal.

13:07:53 <sbp> ]]]

13:07:56 <sbp> good call

13:08:22 <sbp> that information: that the range of dc:date is an appropriate RDF defined datatype

13:08:38 <sbp> RDF defined: having information about it's definition in RDF form

13:09:03 <bijan> That's an approach...and one that coudl be recovered from the <> dc:date "2001-09" version.

13:09:15 <bijan> For what *that's* worth.

13:09:51 <sbp> Here's how we did it for EARL: http://www.w3.org/2001/03/earl/0.95datatypes

13:10:06 <bijan> Well, either you have to elide the issue or you have to go nuts "fixing" this.

13:10:28 <sbp> the processor would have to have built in knowledge of the XSD datatypes, and the URIs used to refer to them...

13:11:05 <bijan> I tend to draw a big

13:11:05 <bijan> box around the

13:11:05 <bijan> _:a <rdf:type> <xsd:Date> .

13:11:05 <bijan> _:a <rdf:value> "2001-09" .

13:11:06 <bijan> and call all of that a literal.

13:11:11 <sbp> why go nuts? just say: "x is how it will be" and that's it

13:11:26 <sbp> yeah, you could also have a language in there too

13:11:42 <sbp> _:b xml:lang "en" .

13:12:04 <bijan> Jan: You didn't mean that one had to represent what is intened by something like <> dc:date "2001-09" in that manner?

13:12:51 <bijan> sbp: start with rss1.0

13:12:59 <bijan> WHen you get the spec changed, let me know.

13:13:21 <sbp> { dc:date rdfs:range xsd:Date . <> dc:date "2001-09" } log:implies { [ a xsd:Date; rdf:value "2001-09" ] } .

13:13:56 <bijan> Yeeeeeeeeeeeeeeeees?

13:14:11 <bijan> Are you agreeing or disagreeing?

13:14:13 <bijan> And with what?

13:14:14 <bijan> :_)

13:14:38 <sbp> rss1.0: I tend to avoid things that use terms as both properties and classes

13:15:14 <sbp> considering that "rdf:Property daml:disjointWith rdfs:Class ." is a pretty sensible axiom to have

13:15:19 <jang> bijan: that's not what I meant

13:15:26 <jang> I meant two things:

13:15:57 <jang> which my memory has halved; anyway...

13:16:03 <sbp> lol

13:16:08 <jang> if you see <> dc:date "2001-09" you accept that

13:16:13 <jang> oops, shaky keyboard

13:16:22 <jang> you accept that "2001-09" is the representation of a date

13:16:27 <bijan> Yes.

13:16:31 <jang> the fact that it looks like a unicode literal is an accident

13:16:51 <jang> and the value you plug into the model you're building at the sharp end of "dc:date" _is_ a date

13:17:01 <jang> you can tell this...

13:17:03 <bijan> yes.

13:17:09 <jang> either by reference to an rdf schema

13:17:20 <jang> or by using some hackery in the serialisation

13:17:21 <lilo> [GlobalNotice] Hi all. Services will be back in a moment, with the usual splits accompanying it. Please bear with us.

13:17:22 <bijan> Although, I don't see why the model couldn't use an unicode representation if it wanted to :)

13:17:36 <jang> well, it doesn't have to

13:17:44 <jang> my point of view is completely backwards compatible

13:17:48 <bijan> No, it doesn't have to.

13:17:54 <jang> I'm talking about a mindset thing here...

13:18:04 <bijan> But it doesn't have *not* to :)

13:18:06 <jang> in other words, your system might actually use a unicode representation

13:18:14 <bijan> Right, but it's a date.

13:18:18 <jang> but it should "consider" it to be a date

13:18:22 <jang> yep

13:18:37 <ChanServ> ChanServ has changed the topic to: Semantic Web Chat - http://www.w3.org/RDF/Interest/#irc

13:18:37 <ChanServ> This channel is logged and blogged: http://logicerror.com/rdfIRCWelcome

13:18:49 <sbp> so the internal representation will have to be tagged in some way so that it can follow that literal around...

13:19:12 <jang> sbp: what's rdf:Property

13:19:15 <jang> ?

13:19:19 <bijan> sean: well, perhpas "tagged" in the sense of "is the object of a dc:date" :)

13:19:22 <jang> that's the idea sbp

13:19:31 <jang> that's more like what I had in mind

13:19:31 <sbp> what do you mean, "what's rdf:Property"?

13:19:42 <jang> earlier... "rdf:Property daml:disjointWith rdfs:Class ."

13:19:53 <sbp> yeah, what's wrong with that?

13:20:56 <jang> I dunno, I'm just having a bout of attention defecit disorder

13:21:32 <sbp> tagged: I mean that in an internal store, both the lexical unicode representations of the literals in :x :y "2001-09" . and <> dc:date "2001-09" . will be the same, but that the literals will be of different types, so it will hvae to assign labels to the literals to follow them around

13:21:48 <bijan> I didn't quite get what it had to do with "fixing" rss.

13:22:04 <sbp> and RDF Core needs to start shouting out about this, before people start implementing broken things...

13:22:18 <bijan> sbp: well, as I said, one way it coudl be "tagged" is by noting it's the object of a dc:date :)

13:22:20 <jang> actually, there's a problem with this

13:22:26 <sbp> RSS: I was just ranting... that classes are disjoint from properties, and yet RSS has a classproperty. Yuck

13:22:26 <dc_rdfig> Label RSS not found.

13:22:42 <jang> there neds to be a distinction drawn between what people generally think of as "objects" and literal values...

13:22:45 <bijan> Worse, or better, you could just parse it willy nilly and hope for the best!

13:22:59 <jang> also, a literal may have multiple equivalent representations

13:23:02 <sbp> "objects": in the RDF sense? I guess not

13:24:48 <sbp> better: probably"

13:24:51 <sbp> s/"/!

13:28:33 <sbp> Aaron, are you there?

13:29:39 <sbp> when you through doing whatever it is you're doing, could you possibly explain why rss:image is a property and a class, and what's being done to rectify the problem

13:29:47 <sbp> if anything

13:29:48 <AaronSw> Feb 2001?!

13:29:53 <sbp> yeah!

13:30:11 <sbp> That really was the target date

13:30:19 * sbp looks at the calendar

13:30:28 <sbp> they seem to be on target...

13:31:07 <jang> heh, at the rdfcore f2f...

13:31:31 <jang> I almost mistyped our expected date of delivery as "WWW20002"

13:31:41 <sbp> lol!

13:32:43 <sbp> You'd still me the target

13:32:46 * sbp ducks

13:32:51 <AaronSw> still me?

13:33:03 * sbp is just kidding

13:33:14 <sbp> no, just anyone in general. Just ignore me

13:33:25 <sandro> "still miss the target" I assume

13:33:46 <AaronSw> rss:image: Because Edd said it should be, and I wasn't well enough RDF-versed to catch it at the time and danbri was lounging on some tropical island.

13:34:01 <AaronSw> (No offense to Edd and danbri.)

13:34:05 <sbp> Ah, a mistake. So what are ye all doing ta fix it?

13:34:26 <AaronSw> Umm, well, you see, it's sorta like this.... umm... nothing.

13:35:17 <sbp> Gflmergl

13:35:31 <AaronSw> Well, what can we do?

13:35:39 <sbp> the only thing that I can fix of is of using the wonderful powers of inference to fix all of the current RSS

13:36:55 <sbp> reify, then convert any statements with the predicate "rss:image" into similar statements with the predicate "rss:hasImage"

13:37:13 <AaronSw> "the only thing that I can fix of is" Wow, what a sentence. You make more sense late at night.

13:37:15 <sbp> and do a similar thing to class uses, converting "rss:image" to "rss:Image"

13:37:20 <sbp> lol

13:37:39 * sbp is doing too many things at once

13:38:19 <AaronSw> Aww, shoot! Jeremy is objecting to stating that Literals are Resources.

13:38:29 <sbp> Where?

13:38:39 <AaronSw> on w3c-rdfcore, re: the model theory.

13:38:49 <sbp> Where?

13:38:54 <AaronSw> on w3c-rdfcore, re: the model theory.

13:39:07 <sbp> He said that they should be anonymous

13:39:07 <dajobe> you can't slip in such a thing at the last minute

13:39:27 <bijan> The object of a statement (i.e., the property value) can be another resource or it can be a literal; i.e., a resource (specified by a URI) or a simple string or other primitive datatype defined by XML.

13:39:36 <AaronSw> Well, I think it's rather obvious.

13:39:46 <bijan> Hmm. Given W3C XMl schemas....

13:39:58 <sbp> Literals are resources. Every addressable object in the universe is a resource!

13:40:06 <jang> a literal is not an object

13:40:13 <jang> a literal is a self-describing datum

13:40:21 <sbp> it's still a resource

13:40:22 <jang> sorry, knee-jerk reacion

13:40:54 <bijan> sean: but is it a resource (treated as a resource) *in this context*

13:41:01 <sbp> a self-describing datum can still be addressed, either in a model, or in an internal RDF store. Hence, it is a resource, by definition, and it is silly to suggest otherwise

13:41:18 <sbp> in this context: er...

13:41:29 <sbp> yes

13:41:52 <jang> y point really is...

13:42:04 <sbp> according to the "[ rdf:value "2001-09" ] ." view of things

13:42:04 <jang> what would people use RDF to say about literals?

13:42:13 <jang> either: things that are already patently true, or lies.

13:42:44 <bijan> sean: if it's sensible to mention them rather than use them in this context, that's the line you should be arguing.

13:42:51 <jang> although, "is the literal representation of the name of..." is a reasonable property

13:42:53 <bijan> Not "but they're *objects* in the world"!

13:43:03 <jang> at least, danbri would argue so to annoy me :-)

13:43:29 <sbp> it doesn't matter whther or not they are mentioned, the fact is that they are *mentionable*, and are therefore defined as resources

13:43:55 <sbp> resonable property: sure

13:43:57 <bijan> No no no.

13:44:21 <bijan> Look, let me define a quicky language. It has two words A and B.

13:44:42 <sbp> Ooh, a binary language

13:44:48 <bijan> A denotes a resource and can be use to attribute things to it.

13:44:53 <sbp> How are you today? A B B B A B A A B

13:44:54 <bijan> B does not.

13:45:16 <bijan> So, I can say "A is a resource"

13:45:20 <sbp> Er... you've lost me already. A denotes what now?

13:45:33 <bijan> Any resource, pick one, it doesn't matter.

13:45:37 <sbp> O.K.

13:46:05 <bijan> Now, from one point of view...the meta-AB-linguistinc, A and B are objects.

13:46:15 <bijan> Words to be precise.

13:46:18 <bijan> I can name them.

13:46:45 <bijan> So, it makes sense to say "B" is a resource.

13:46:53 <sbp> yes...

13:47:15 <bijan> But that doesn't mean that in my specification of the language, I want to think of B as a resource.

13:47:17 <bijan> It's not.

13:47:25 <sbp> but it is!

13:47:27 <bijan> A, now *A*, that's a resource.

13:47:31 <sbp> how can "B" not denote a resource, in the RDFS sense of "resource"? A resource is "that which can be denoted", and if B is denoting something, then that thing is a resource. It is a logical impossiblity to denote something which is not a resource

13:47:32 <bijan> No, "B" is a resource.

13:48:08 <bijan> Well, "B" can denote nothing at all.

13:48:14 <sbp> if it exists, and especially if it exists within the realms of a computer system, then it is addressable, and it is a resource

13:48:14 <bijan> Or it can *express* but not denote.

13:48:36 <bijan> If it's a free variable, it won't denote anything.

13:48:39 <sbp> well then, it is perfectly useless, and not a name

13:48:50 <bijan> I didn't say it *was* I name.

13:48:53 <bijan> It's a *word*.

13:48:56 <sbp> free variable: universally quantified variables are still resources

13:49:06 <bijan> See, you're confused.

13:49:10 <sbp> if it's "a" anything, then it is a resource

13:49:24 <bijan> metalinguistically, yes. In teh object language, no.

13:49:52 <bijan> "B" is a word. It "is" also, lets say, a variable.

13:49:59 <jang> sbp: "how can B not denote anything?"

13:50:06 <jang> answer: because D(B) is undefined

13:50:33 <sbp> fine. It is not a resource. Now what?

13:50:44 * sbp has a teleconference quite soon

13:51:37 <bijan> Just pointing out that literals may be resources, but you need a better argument than "the language items Literals are themselves resources"

13:52:18 <bijan> Here's another example, URI's aren't resources.

13:52:24 <sbp> It still seems to me that there is no sound logical basis for that statement, according to what I understand as being the definition of a resource

13:52:25 <bijan> I.e., in rdf statemetns.

13:52:50 <bijan> If they were, then we could only talk about URIs

13:52:52 <sbp> URIs are resources... they can have URIs: data:,http://www.w3.org/

13:52:59 <bijan> See, that's the confusion.

13:53:00 <sbp> oh, I see what you mean...

13:53:10 <bijan> Yes. that's the use/mention difference.

13:53:14 <sbp> cool

13:53:51 <sbp> but it is incorrect to say that literals are not a subClassOf rdfs:Resource

13:54:00 <bijan> BTW, it's only the most common mistake that everyone...from babes to Betrand Russell.

13:54:09 <bijan> makes. ;)

13:54:12 <sbp> :-)

13:54:54 * sbp has to get ready for the call

13:55:09 <bijan> ta

13:55:11 <sbp> bijan, jan, many thanks for the discussion :-)

13:56:04 * dajobe thinks we need a clear, small uri/resource, use/mention document

13:56:53 <sandro> Should it point to all the bad examples in widespread use?

13:57:04 <dajobe> I said small :)

13:57:13 <jang> only if it says more than "don't do this"

13:57:23 <jang> instead, "don't d this, do XXX instead"

13:58:45 <sandro> Would it define a correct answer to whether <http://w3.org> necessarily denotes a web page?

13:59:11 <jang> nothing "necessarily" denotes anything

13:59:13 <AaronSw> What's a web page?

13:59:19 * dajobe groans

13:59:37 <AaronSw> No, I'm just curious.

13:59:40 <jang> there's not one universal denotational function

14:01:12 <sandro> So jang thinks the use/mention document would not need to open that can of worms..... That sounds like RDF theory, without helping people on RDF practice, but maybe that's right.

14:04:31 <jang> well, the question poens up many issues...

14:04:41 <jang> such as, what happens if I mirror a web site?

14:04:49 <jang> what happens if I put multiple language versions online?

14:05:04 <jang> actually, those are principally the two big issues

14:13:41 <bijan> Hmm. I think http://w3.org, as standardly read, denotes a location! :)

14:14:23 * bijan is just stirring up trouble...being a troll...don't fall for it!

14:15:13 * lasDesk is amused to discover, after being without an internet connection for 4 days, that the discussion is *exactly* where she left it.

14:15:34 <bijan> What goes around comes around and around and around.

14:18:17 <dajobe> so we are in what is generally called "pre-consensus" on this topic

14:18:23 <bijan> No.

14:18:29 <bijan> Pre-pre consenus.

14:18:59 <dajobe> that is meaningless

14:19:23 * lasDesk wonders at the idea of pre-consensus....we agree that we don't agree?

14:19:26 * bijan points to his last /me statement.

14:19:53 <jang> protoconsensus

15:02:12 * dajobe notes googling for various webby things often ends up with hits on these chatlogs

15:12:29 <sbp>http://www.w3.org/TR/uri-clarification

15:12:30 <dc_rdfig> D: http://www.w3.org/TR/uri-clarification from sbp

15:12:41 <sbp> D:|URIs, URLs, and URNs: Clarifications and Recommendations 1.0

15:12:41 <dc_rdfig> titled item D

15:13:04 <sbp> D:Report from the joint W3C/IETF URI Planning Interest Group, W3C Note 21 September 2001

15:13:04 <dc_rdfig> commented item D

15:14:58 <sbp> D:The classical view? Yuck. Go contemporary!

15:14:58 <dc_rdfig> commented item D

15:16:39 <sbp> D:This is a good note, because it covers the first in a huge list of very important questions that need to be addressed

15:16:39 <dc_rdfig> commented item D

15:18:21 <sbp> "Some have an RFC, but are not included in the IANA list.

15:18:22 <sbp> "

15:18:42 <sbp> and hence the IANA list is out of date... who maintains it, anyway?

15:19:00 <sbp> Pff: (last updated 2001 August 20)

15:19:22 <sbp> But the bit that really cracks me up is: "In the Uniform Resource Locators (URI) definition "...

15:20:55 <sbp> Hmm... "alternative trees". Can you imagine a "prs." tree?

15:28:31 <sbp> s/prs./prs-

15:29:54 <Morbus> unterminated regular expression at line 1.

15:31:13 <sbp> pardon?

15:31:21 <sbp> s/prs./prs-/

15:32:16 <sbp> the trailing slash is implied, if that's what you meant... :-)

15:32:30 <Morbus> :)

15:34:33 <sbp> Gotta run

16:37:29 http://www.adobe.com/aboutadobe/pressroom/pressreleases/200109/20010924xmp.html

16:37:29 <dc_rdfig> E: http://www.adobe.com/aboutadobe/pressroom/pressreleases/200109/20010924xmp.html from em

16:37:46 E:|XML (Extensible Metadata Platform) from Adobe

16:37:46 <dc_rdfig> titled item E

16:37:57 E:XMP (Extensible Metadata Platform), this framework enhances workflows so

16:37:57 <dc_rdfig> commented item E

16:37:57 that content can be applied seamlessly between print, Web, eBooks and other

16:37:57 media. XMP provides Adobe applications and partners with a common metadata

16:37:57 framework that standardizes the creation, processing and interchange of

16:37:57 document metadata across publishing workflows. XMP will be incorporated

16:37:58 into all Adobe products eventually and is available for developers via a

16:38:00 software development kit (SDK).

16:38:08 E:XMP incorporates many World Wide Web Consortium (W3C) standards. W3C

16:38:09 <dc_rdfig> commented item E

16:38:09 produces the technologies that serve as the foundations for Web

16:38:09 architecture, including XML itself and Resource Description Framework

16:38:09 (RDF), the foundation for Metadata on the Web, and Semantic Web developments.

16:52:49 * danbri catches up

16:55:29 * em notes to fix XMP chump item as comments didn't make it

17:55:13 * AaronSw waves

18:38:31 E:XMP (Extensible Metadata Platform), this framework enhances workflows so that content can be applied seamlessly between print, Web, eBooks and other media. XMP provides Adobe applications and partners with a common metadata framework that standardizes the creation, processing and interchange of document metadata across publishing workflows. XMP will be incorporated into all Adobe products eventually and is available for developers via a softwar

18:38:31 <dc_rdfig> commented item E

18:38:31 e development kit (SDK).

18:38:42 E:XMP incorporates many World Wide Web Consortium (W3C) standards. W3C produces the technologies that serve as the foundations for Web architecture, including XML itself and Resource Description Framework (RDF), the foundation for Metadata on the Web, and Semantic Web developments.

18:38:42 <dc_rdfig> commented item E

19:25:23 <timbl-lap> Re RDF schema for infoset -- I have looked at it before. It uses "." in localnames which doesn't convert to N3.

19:25:25 <timbl-lap>http://www.w3.org/2001/04/infoset.

19:25:25 <dc_rdfig> F: http://www.w3.org/2001/04/infoset. from timbl-lap

19:26:59 <timbl-lap> It contains (when N3-ized) multiple classes as domains, and I supect means there union instead of their intersection.

19:27:00 <timbl-lap> :parent a rdfs:Property;

19:27:00 <timbl-lap> rdfs:domain :Character,

19:27:00 <timbl-lap> :Comment,

19:27:00 <timbl-lap> :DocumentTypeDeclaration,

19:27:01 <timbl-lap> :Element,

19:27:03 <timbl-lap> :ProcessingInstruction,

19:27:05 <timbl-lap> :UnexpandedEntityReference;

19:27:07 <timbl-lap> rdfs:range :InfoItem .

19:56:20 <sbp> Hmm... multiple ranges/domains are taken to be conjunctions, usually

19:56:35 <sbp> Gotta run

20:00:16 <theran> You could fix n3, I guess. Is there a problem with #\. in XML?

20:36:55 <jrn_work> jrn_work is now known as jrn

20:49:59 <AaronSw> Hi Dan

20:50:07 <AaronSw> Nice work on uri-clarification -- just reading it now.

20:56:56 <AaronSw> I can't even find any mistakes to show I read it carefully.

21:02:47 <theran>http://lists.w3.org/Archives/Public/www-rdf-comments/2001JulSep/0288.html

21:02:47 <dc_rdfig> G: http://lists.w3.org/Archives/Public/www-rdf-comments/2001JulSep/0288.html from theran

21:03:25 <theran> G| My comments on the ntriples syntax for test cases.

21:04:24 <theran> G: |My comments on the ntriples syntax for test cases.

21:04:24 <dc_rdfig> commented item G

21:04:52 <dajobe> theran: URI syntax will change, didn't you read my reply on www-rdf-comments ?

21:05:30 <AaronSw> theran: what files would you be byte-comparing?

21:05:31 <dajobe> cmp isn't appropriate to compare models where some nodes have no names

21:05:47 <AaronSw> iow, ntriples files with bNodes

21:05:48 <dajobe> i.e. graphs

21:09:28 <theran> why should cmp not be appropriate?

21:10:02 <AaronSw> Because you can name the bNodes different things in N-Triples.

21:10:04 <dajobe> you can't compare graphs where some of the nodes are given local names (bNodes) that are not URIs

21:10:05 <theran> if you are defining an output format for compatibility testing, then the oracle should be simple, even if that means restrictions on what may be emitted.

21:10:41 <AaronSw> Well it would be pretty useless as a testing format if we removed a key part of the language from it.

21:10:46 <dajobe> graph compare algorithms are easy to find in textbooks

21:10:59 <dajobe> and these are test cases, so the graphs ar small

21:11:21 <AaronSw> a number of ntriples comparison tools have already been written.

21:11:32 <theran> formal ways of defining orderings on sets are easy to find in textbooks.

21:12:35 <AaronSw> theran, do you know a way to generate a canonical string from arbitrary RDF graphs? if so, please do tell the working group

21:13:56 <theran> are you saying that two ntriples graphs may be the same if the node and edge sets are not isomorphic?

21:16:43 <AaronSw> isomorphic?

21:18:51 <dajobe> AaronSw: does the draft MT specify graph comparison equality? I can't rememeber

21:19:36 <theran> there is a bijective function between the sets.

21:23:45 <theran> ``AaronSw: a number of ntriples comparison tools have already been written.'' ok.

21:24:10 <AaronSw> model theory: Hmm, I think it just said that if two graphs were the same then they were equal. I don't think anything more is needed in the new graph syntax.

21:24:22 <AaronSw> theran: not in common lisp, though ;-)

21:24:29 <AaronSw> but i can't imagine its too hard.

21:24:38 <theran> so I take it these use some specific definition of sameness not written down in the draft.

21:25:12 <AaronSw> sameness? no, they just do the conversion from bNodes to empty circles in "tidy" graphs

21:25:31 <AaronSw> AFAIK, there's no way to serialize a graph like that in a fully canonical way.

21:26:05 <AaronSw> i.e.:

21:26:13 <AaronSw> <a> _:c .

21:26:19 <AaronSw> is equivalent to

21:26:24 <AaronSw> <a> _:foo .

21:26:27 <AaronSw> etc.

21:27:57 * DanC plays around with RDF family trees...

21:28:27 <AaronSw> Ooh, cool. Perhaps a GEDCOM importer? ;-)

21:28:43 * AaronSw did a bunch of family tree stuff two years ago or so for a school project.

21:28:55 <AaronSw> kept trying to get the GEDXML guy to use RDF but he was on to other things.

21:34:08 <theran> correct me if I am wrong, but: (1) lexographic comparison is an absolute order on the node and edge sets. (2) a bNode will be introduced by a triple of the form ``<a> _:x''

21:36:56 <AaronSw> what do you mean by introducing a bNode?

21:38:01 <theran> If I have an anon node, when I go to serialize it, this will be as part of a triple with two other (possibly anon) nodes.

21:38:11 <AaronSw> yes

21:39:53 <theran> My point is that since we expect the node and edge sets to be isomorphic, and we have at least a partial order on both, then we can specify a way of generating the serialized names s.t. the same graphs will have the same names generated. Again, correct me if I am wrong.

21:41:21 <AaronSw> I thought the same thing but I was never able to find an algorithm that worked.

21:42:01 <AaronSw> I

21:42:09 <AaronSw> I'd be happy to be disproved, tho.

21:52:15 <theran> Try this one out (there's no particularly deep reasoning here---it's quite possibly all wrong)... We have an ordering on the vertex and edge sets, which are finite. So we can number them. For example (t_n,t_m,t_l). Except for bnodes, which have no names. Name bnodes after the lowest numbered node,arc pair that points to them. Break ties based on the lowest numbered arc leaving the bnode.

21:53:19 <theran> I must admit, this is not all that satisfying.

21:53:23 <AaronSw> Yeah, that was the one I came up with...

21:53:28 <dajobe> that is all fine (if it works - not sure) but n-triples is meant to be easy for people to write; the machines can do the not-very-difficult work to compare graphs, and this has already been implemented a few times already

21:54:15 <theran> I want to be able to trust that two different oracles are in fact the same.

21:54:21 <dajobe> oracles?

21:54:44 <theran> So either you need a cannonical representation, or you need a formal description of sameness.

21:55:24 <AaronSw> The definition in the model theory is not formal enough for you?

21:56:16 <theran> I want a definition in terms of ntriples documents. Not in terms of abstract structures, since we are not testing those.

21:56:30 <theran> Two ntriples documents are the same iff ...

21:56:34 <dajobe> n-triples has no meaning apart from defining a graph for th emodel theory to interpret

21:56:58 <AaronSw> The model-theory describes how to convert N-Triples documents into a graph, no?

21:56:59 <dajobe> but we could put the graph equality thing in n-triples since it is useful there with the test cases nearby

21:58:13 * DanC doesn't think the model theory tells how to convert N-triples to a graph.

21:58:17 <theran> Also, if the M&S document was considered precise enough, we would not be doing this exercise.

21:59:14 * theran notes that since & usually commutes, we could call it the S&M document too. :)

21:59:16 <AaronSw> DanC, the paragraph starting "A graph can be viewed as a set of triples corresponding to its edges; the correspondence between an ntripleDoc and an RDF graph is that the graph has one node for each uriref, bNode or literal identifier in the document, and one edge for each triple, directed from the node corresponding to the subject to the node corresponding to the object." apears to.

21:59:21 <DanC> er... isn't the graph equality thing in the test case spec?

21:59:53 <AaronSw> Doesn't appear so.

21:59:59 <DanC> odd

22:00:29 <DanC> how do folks use the test cases? they have to know how to do graph comparison, to use the .nt files,no?

22:00:49 <AaronSw> Yes, I'd think so.

22:01:06 <dajobe> unless you have no bNodes

22:01:10 <AaronSw> Perhaps the draft should include pointers to jena.rdfcompare and its associate paper.

22:01:11 <DanC> so it seems some reshuffling is in order.

22:01:20 <theran> ``Each test case is defined in RDF/XML. The expected result is defined using the [N-Ttriples] syntax. [N-Triples]is a line-based, plain text format for representing a RDF triple. Each RDF/XML file that contains valid RDF has an associated [N-Triples] file. The contents of the [N-Triples] files have been reviewed by the RDF Core Working Group and are considered authoritative (subject to the warnings provided in Status of this Document).

22:01:22 <theran> ''

22:01:29 <AaronSw> jena: to show that there is working code behind this

22:01:38 <theran> That's the text from the draft.

22:01:43 <dajobe> well, I did write it

22:01:53 <dajobe> no, that was Art actually

22:02:32 <theran> The draft goes on to say that ntriples is a line based description of an RDF graph.

22:02:52 <theran> So we have reduced the problem of comparing RDF to comparing ntriples.

22:03:24 <AaronSw> Right, and the MT reduces the problem to comparing graphs, no?

22:03:25 <dajobe> which is my point, we could usefully put one simple graph-compare-in-ntriples algortihm in that doc

22:03:45 <AaronSw> I agree. You could even steal Jeremy's from the jena.rdfcompare paper.

22:03:47 <theran> But we could just use RDF!

22:03:53 <dajobe> we are

22:04:02 <theran> RDF is a representation of an RDF graph too. :)

22:04:04 <dajobe> steal: yup

22:04:14 <DanC> just use RDF? huh? how? ah... you're looking for a canonical linear representation of a graph. no can do.

22:04:34 <DanC> perhaps we should say that (no can do) in the test cases doc.

22:04:56 <DanC> fun with RDF for family trees: http://www.w3.org/2001/02pd/dm93fam.png

22:05:55 <AaronSw> family trees: cool!

22:06:18 <theran> Dan: can you explain? This is obviously true for general graphs. But we have a lot more structure here.

22:06:40 <AaronSw> Hmm, interesting: http://orl01.drc.com/daml/HW/Assignment3/GedDemo/Converter.htm

22:06:47 <AaronSw> Does your stuff use the same RDF format, DanC?

22:08:30 <DanC> doubt it, Aaron. I just made my format up.

22:08:53 <AaronSw> thought so

22:08:56 * DanC sees a blank page at .../Converter.htm

22:09:03 <AaronSw> You do?

22:09:18 <DanC> well, it says "GedML 2 DAML Converter" at the top, with a logo, and that's it.

22:09:25 <AaronSw> Perhaps turn JS off?

22:09:37 <DanC> <meta name="GENERATOR" content="Microsoft FrontPage 4.0">

22:09:43 <AaronSw> :-)

22:10:02 <Morbus> heheheh

22:12:12 <AaronSw> That should get them an F for this assignment.

22:12:34 * DanC tries opera, gets a little further..

22:12:37 <DanC> lots of 404s

22:12:44 <DanC> when looking for their ontology, for example

22:12:49 <AaronSw> More useful stuff at http://orl01.drc.com/daml/HW/Assignment3/GedDemo/ perhaps?

22:13:02 * DanC tries daml ontology library...

22:14:56 <AaronSw> Well, there's google cache: http://www.google.com/search?q=cache:f-dNVqDaoe4:orl01.drc.com/DAML/Ontology/Genealogy/current/+geneology+rdf&hl=en

22:15:05 <DanC> aha... http://www.daml.org/2001/01/gedcom/gedcom.daml

22:15:17 <DanC> hyper-damlized: http://www.daml.org/cgi-bin/hyperdaml?http://www.daml.org/2001/01/gedcom/gedcom.daml

22:15:53 <dajobe> must tell you the story of the UK govt 'interoperabiloty' web site that required me to use lynx to navigate the frames, to submit the form I couldn't do in mozilla to download the Word file describing the web interop. policy :)

22:16:45 <AaronSw> DanC, some rules for that ontology: http://www.w3.org/2000/10/swap/test/gedcom/gedcom-relations.n3

22:18:11 <DanC> yeah...odd stuff in there: ... log:implies <data:,>

22:19:00 <DanC> BLURB: geneology fun with RDF

22:19:00 <dc_rdfig> H: geneology fun with RDF from DanC

22:20:24 <DanC> H:a bit of [family tree|http://www.w3.org/2001/02pd/dm93fam.png], [in RDF/n3|http://www.w3.org/2001/02pd/dm93fam.n3]

22:20:25 <dc_rdfig> commented item H

22:20:40 <DanC> H:some [gedcom-relations|http://www.w3.org/2000/10/swap/test/gedcom/gedcom-relations.n3]

22:20:40 <dc_rdfig> commented item H

22:23:08 <DanC> H:a [gedcom daml ontology|http://www.daml.org/2001/01/gedcom/gedcom.daml], [hyperdaml-ized|http://www.daml.org/cgi-bin/hyperdaml?http://www.daml.org/2001/01/gedcom/gedcom.daml]

22:23:08 <dc_rdfig> commented item H

22:23:57 <DanC> H:a [converter|http://orl01.drc.com/daml/HW/Assignment3/LargeScaleContent-B.htm] from some XML gedcom thing to DAML

22:23:58 <dc_rdfig> commented item H

22:25:39 <DanC> H:a [ged2xml converter|http://www.topxml.com/downloads/default.asp?id=ged2xml]

22:25:39 <dc_rdfig> commented item H

22:31:31 <DanC> hmm... didn't know gedcom was from The Church of Jesus Christ of Latter-day Saints

22:31:47 <DanC> H:a [gedcom spec|http://www.gendex.com/gedcom55/55gctoc.htm]

22:31:47 <dc_rdfig> commented item H

22:32:28 <AaronSw> Oh, you didn't? They're very big into geneaology.

22:33:37 <DanC> it make sense; I just didn't make the connection 'till now

22:34:43 <sbp> Do DAML have a persistence policy?

22:35:23 <sbp> Google suggests not: http://www.google.com/search?hl=en&q=site%3Adaml.org+persistence

22:35:25 * DanC wanders off...

22:38:05 * sbp wonders if Dan's wandering is related to my wondering

22:38:23 * AaronSw wonders if DanC has a persistence policy

22:38:53 * sbp wonders if DanC really *has* wandered off

22:39:34 * sbp concludes so

22:48:46 <AaronSw> Sheesh, jos, you're quick at reading these logs!

22:49:11 <AaronSw> Heh: "where some of those aunt's are actually M(ale) which is a bit strange..."

22:49:47 <AaronSw> Jos, have you tried using http://impressive.net/people/gerald/2000/11/irc/cgi-irc/irc.cgi ?

22:49:55 <AaronSw> It should work from work.

22:55:34 <AaronSw> Server: irc.openprojects.net

22:55:38 <AaronSw> Port: 6667

22:55:48 <AaronSw> Channel: #rdfig

23:09:43 <dajobe>http://www.slashdot.org/articles.rss

23:09:43 <dc_rdfig> I: http://www.slashdot.org/articles.rss from dajobe

23:09:55 <dajobe> I:|Slashdot RSS 1.0 (i.e. RDF/XML) articles news feed

23:09:55 <dc_rdfig> titled item I

23:10:12 <dajobe> I:with multiple namespaces including Dublin Core

23:10:12 <dc_rdfig> commented item I

23:10:34 <dajobe> I:happened a while back when the codebase switched to Slashcode 2.0

23:10:34 <dc_rdfig> commented item I

23:20:53 <AaronSw> Hi jos!

23:20:56 <AaronSw> I see that it worked!

23:21:20 <JosD> Aaron, thanks for that hint! How are you today?

23:21:33 <AaronSw> No problem. I'm doing quite well, and you?

23:22:40 <JosD> Very good, although I have to go to my dentist tomorrow, and I'm so scared

23:23:03 <sbp> any teeth need whipping out?

23:23:07 <AaronSw> Ouch.

23:24:31 <JosD> not really (I hope), just some small broken part...

23:24:44 <sbp> Well, my motto is that the dentist is always there to help... (that's how I get through it)

23:25:33 <JosD> Aaron, I also think that LV is a subset of IR

23:25:48 <AaronSw> Cool. :-)

23:26:13 <sbp> Hooray

23:26:57 <JosD> I just don't see how to fix the N3 rules if they weren't

23:27:25 <sbp> what does the MT currently say about that?

23:27:33 <JosD> Hi Sean!!!

23:27:42 <dajobe> hi there Jos

23:27:53 <sbp> Hi. Sorry, I have an awful habit of just butting in...

23:28:01 <AaronSw> Well, it's sure good to have you here, Jos...

23:28:07 <AaronSw> Now you can scribe at telecons. ;-)

23:29:17 <JosD> Sean, agnostic

23:29:47 <sbp> ah, I found your message about it... http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Sep/0339

23:30:13 <JosD> Well Aaron, my typing speed is a fraction of yours...

23:32:43 <JosD> Sean, that's indeed a message about that point

23:32:46 <AaronSw> I can help with the typing, it's producing the minutes I don't like.

23:33:23 <sbp> I'm glad someone's sticking up for the LV subset of IR viewpoint! :-)

23:33:31 <JosD> Oh, Hi Dave, what a pleasure to meet you!

23:33:57 <dajobe> yeah, hope you can be around on friday; at lteast reading it

23:35:41 <JosD> Thanks Aaron!

23:49:25 <AaronSw> hello

23:49:31 <Iron_SpermWhale> Hello

The IRC chat here was automatically logged without editing and contains content written by the chat participants identified by their IRC nick. No other identity is recorded.