RDF Core Working Group IRC logs for 2001-08-02

These are the logs from the RDF Core Working Group IRC chat.

Thanks to Dave Beckett for the logger software that produced them.

You are here: Logs Home / 2001-08-02

[00:00:26] dajobe

phayes: publishing a doc that erfers to anon _:bob and really want to do that

[00:00:40] dajobe

... if I can't mean the same thing by the same name, that name is no use of all

[00:00:52] dajobe

danc: nature is that there is no name in rdf/xml and can't refer to it

[00:01:03] dajobe

phayes: then I shouldn't generate _:bob

[00:01:22] dajobe

danc: yes, but this is ntriples and _:bob can't be seen outside

[00:01:41] dajobe

fmanola: then this can't come up?

[00:01:56] dajobe

phayes: are you allowed to use _:bob

[00:02:04] dajobe

danc: no, you use "something"

[00:02:14] dajobe

phayes: then I cannot refer it to you directly

[00:02:23] dajobe

danc: yes, you have to use other mechanism

[00:02:50] dajobe

jang: there are transient docs on the web and fetch them, _:bob etc. are transient identifers you want them to behave lie that

[00:03:10]

* danbri scribes

[00:04:03] danbri

frank: it is one thing to talk about two '_bobs'... It is a very different use case if I, perusing doc1 and getting a genid _bob, returns to source and say "hey, you know that thing you called _bob"...

[00:04:16] danbri

pat: thats the case i had in mind

[00:04:35] danbri

sergey: suggestion... Don't use N3 please. We've barely a grip on RDF/XML.

[00:05:00] danbri

dan: we didn't see these disagreements until we had ntriple to make situation explicit

[00:05:43] danbri

emiller: <shows example with /2001/08/01-ex1 and -ex2.> (@todo: link to doc from em)

[00:05:50] danbri

...

[00:06:03] danbri

jan: theres a mechanism question w.r.t. what pat's sayinh

[00:06:19] danbri

"we have _bob from 2 docs... we want to go back and say more stuff about 'it'"

[00:06:40] danbri

"you can use identifying properties about it

[00:06:58]

* danbri disagrees (quietly)

[00:07:20] danbri

graham: the reason i showed example in ntriples was cos we've decided to use this to describe what parsers do

[00:07:36] danbri

"the unfortunate part of the example was my writing _bob instead of _243234234234324

[00:07:45] danbri

"parser needs to write some kind of labelling for the ntriple for.

[00:08:00] danbri

"now if two docs happen to parse to same ntriple form, incl. genids...

[00:08:19] danbri

"are we going to make parsers responsible for making globally unique genids

[00:08:32] danbri

"or do we couch this in terms of ids relative to a document

[00:09:38] danbri

frank: a query rather than an assertion: "let's pop the stack... seems we've gone in a fairly complicated manner discussing a number of relevant topics. But we started out here with some pretty straightforward questions, ie. firstly whether we want generated identifiers, then whether we want to distinguish them from URIs (either semantically or syntactically)

[00:09:57] danbri

"i don't know that we've done much to answer those questions, or to answer their converse: if we don't like something is the result any better?

[00:10:15] danbri

"if we don't like generated identifiers, we have things that aren't identified, what do we do

[00:10:24] danbri

brian: we've decided that

[00:10:37] danbri

dan: we didn't decide they were URI

[00:10:52] danbri

brian: i thought we had made progress

[00:11:12] danbri

..."that we agreed we can distinguish these things 'in the model'

[00:11:30] danbri

frank: did we decide that they had the characteristics of URIs

[00:11:37] danbri

pat: decisions was...

[00:11:54] danbri

"we agreed there wouldbe a way of distinguishing the 'distinguished nodes' from 'undistinguished ones'

[00:12:00] danbri

dan: i didn't agree to new syntax for this

[00:12:05] danbri

pat: some way...

[00:12:09] danbri

brian: to tell them apart

[00:12:16] danbri

pat: ...syntactically...

[00:12:30] danbri

emiller: we have 3 interpretations on the overhead

[00:13:10] danbri

dan: make the first one a non-URI,

[00:13:26] danbri

emiller: some way of uniquely identifying it, that is or isn't or looks like a uri

[00:13:45] danbri

dan: the wg has ruled out the 3rd situation

[00:13:56] danbri

..."i heard that you could tell the difference in the output

[00:14:15] danbri

"since seeing http://blah in output couldn've been there in the input

[00:14:40] danbri

pat: what's wrong with saying because it contains 'genid' in it

[00:14:44] danbri

dan: not without talking to god

[00:15:03] danbri

..."not in our charter"

[00:15:42] danbri

emiller: (to pat) a lot of us are coming to this from a web architecture p.o.v....

[00:15:50] danbri

brian: dave has the words from previous decision

[00:16:16] danbri

frank: there's a reverse side of it... When you generated this thing that's clearly distinguishable from a URI... Is it therefore _not_ a URI?

[00:16:34] danbri

..."what are its characteristics?

[00:16:38] dajobe

question was: http://ilrt.org/discovery/chatlogs/rdfcore/2001-08-01.html#T22-57-09

[00:16:39] danbri

frank: yes, acc to the URI spec

[00:16:48] danbri

dave reads from logs:

[00:18:02] danbri

emiller: some people are thinking they can peek inside syntactic substructure of uris

[00:18:38] dajobe

answer was: http://ilrt.org/discovery/chatlogs/rdfcore/2001-08-01.html#T23-18-30

[00:19:33] danbri

dan: the words he wrote rule out interpretation 3

[00:19:40] danbri

(someone pls post the 3rd example)

[00:20:32] danbri

pat: similar to one saying, i'll write an axiom in logic with a relation name called 'Not'

[00:20:50] danbri

dan: the text he read says we can tell these apart

[00:20:53]

* danbri now agrees

[00:21:17] danbri

dan: this info is not enough to tell us

[00:21:28] danbri

pat: why can't we say what the rules of the language are?

[00:21:55] danbri

dan: nowhere in rdf 1.0 does it say we can't have http://purl.org/var/.... is not allowed...

[00:22:03] danbri

"now into the opacity argument for not inspecting URI substrings

[00:22:15] danbri

graham: URIs are not our language (ie. IETF spec)

[00:22:17] danbri

---

[00:22:18] danbri

break

[00:22:19] danbri

---

[00:33:31] danbri

dan: some issues are kinda arbitrary, we owe to world to flip coins

[00:33:42] danbri

...i feel this is a non-arbitrary decisions, needs doing properly

[00:34:05] danbri

(discussion of Vegemite-based incentives)

[00:34:13] danbri

dan: i'm happy to defer to another issue

[00:34:23] danbri

emiller: i'd like to end with something we can accomplish

[00:34:32] danbri

...do you think we can do that in 30 mins

[00:34:48] danbri

pat: i voted no and caused a bunch of trouble

[00:35:13] danbri

..."i'd be happy to change and say yes, so long as we ack that we need to introduce notion of scoping, and be crystal clear

[00:35:32] danbri

"it'd be a mistkae to say they have an existential interpretation and have vagueness about their scope

[00:35:48] danbri

sergey: i want to support your (dan's) suggestion by proposing another use case

[00:36:05] danbri

emiller: pat, dan, sergey agreeing...??!

[00:36:23] danbri

sergey: "the case i'm suggesting... trying to factor out all the different concepts we have in our mind...

[00:36:45] danbri

"if we have something we think are anonymous nodes in the document... is there a way to point to this thing in another document?

[00:37:08] danbri

dan: my answer is 'no'

[00:37:25] danbri

sergey: there's this axiom of the web, that rdf allows folk to say anything about it

[00:37:48] danbri

graham: it doesn't, you can talk about the same entity

[00:38:09] danbri

jan: you _can_ providing you're talkikng about a sub-document... ie. make an instantiation of the document... you know what the anonymous resource is...

[00:38:32] danbri

..."database analogy: run a query on a db, you can tell its identity in the database. but i can't publish the private id as a uri

[00:39:04] danbri

pat: a realisation... i think jos said all along... "you can publish a document... if someone can

[00:39:28] danbri

pat: if the original document _is_ enough to pick it out, (jos: by value) then yes you _can_ describe it further

[00:39:48] danbri

brian: my worry, we spent a whole bunch of time on anon nodes... we set out questions...

[00:40:12] danbri

"once a node gets into the model, can i tell it apart, we had that questoin... I want to make sure. Can we confirm we said "yes!"

[00:40:30] danbri

danc: we've only answered it if we rule out 3rd interpretation (@todo: url overhead)

[00:40:57] danbri

brian: <rdf:Description><foo:bar>foobar</foo:bar></rdf:Description> ...->... _ <foo.bar> "foobar"

[00:41:32] danbri

questoin: what goes here< after the '_'

[00:41:32] danbri

dan: ??? (missed)

[00:41:32] danbri

pat: nobody reading any document generated from the xml would ever be able to get hold of the id

[00:41:45] danbri

pat: of course in ntriples it looks public

[00:42:03] danbri

pat: but if in the egs we use 'http:' it sure looks public

[00:42:11] danbri

brian: notion of public doesn't feel quite precise enough

[00:42:20] danbri

..."same question as sergeys 3rd...

[00:42:40] danbri

brian: there are 3 things that could go in here...

[00:42:49] danbri

"___ <foo.bar> "foobar"

[00:43:07] danbri

brian: dan is asking that we don't allow URIs here

[00:43:31] danbri

strawpolll: can it be a uri

[00:43:33] danbri

most folk: no

[00:43:36] danbri

pat: don't care

[00:43:44] danbri

mike: i'd like it to be a uri and parse substructure

[00:43:56] danbri

sergey: we can have a special namespace

[00:45:13] danbri

danbri: dan persuade me. folk might write bad RDF/XML that used our magic namespace for genids. therefore we can't gurnatee the distinction

[00:45:15] dajobe

danc also said - no, you can't lok inside URI

[00:45:24]

* danbri nods

[00:45:25] dajobe

miked then said - want to parse fragment ids

[00:45:34] danbri

brian: so it can't be a uri

[00:45:39] danbri

Capturing this:

[00:45:49] danbri

we agree it can't be a URI.

[00:46:12] danbri

frank: generated identifiers have a distinguished representation from URIs

[00:46:46] danbri

mike: i'd like to see us say 'we reserve any fragments beginning with an underscore

[00:46:59] danbri

...that way you might know eg what doc it came from

[00:47:02] danbri

...in rdf:ID

[00:47:14] danbri

dan: then you lose expressive power; you lose ability to say 'there exists'

[00:47:24] danbri

marK: maybe you want several types of genids

[00:47:31] danbri

pat: please please don't use a variable as a name

[00:47:52] danbri

jan: this ... doesn't really work. you go to a source, get a doc back; you do again, you get it back. these things are transient in the web.

[00:48:08] danbri

mike: there's nothing to keep the parser from providing an id

[00:48:13] danbri

dan: i suggest thats a bug

[00:48:46] danbri

brian: as i udnerstand the issue with the rdf:ID="_id43454" solution... if i read it in twice, i'm going to geneate the same URI

[00:48:53] danbri

mike: if you use same tool

[00:49:16] danbri

brian:...but you have no way to know that that's about theS same rsource

[00:49:27] danbri

...you're parser is asserting identity when has no right to do so

[00:49:41] danbri

mike: [...] can use daml:equiv...

[00:49:55]

* danbri (didn't capture that)

[00:50:02] danbri

pat: that was what bothered me about _bob

[00:50:07] danbri

...someone else might use it

[00:50:18] danbri

mike: i'd want them to use full URI for doc

[00:50:20] danbri

martin: yes

[00:50:40] danbri

dan: i use '_bob' here, i can't use that _thing_ in any other formula

[00:51:00] danbri

pat: if you were to write that in rdf/xml you'd not see '_bob'

[00:51:14] danbri

dan: in ntriples: an implicit backwards E in front of _:

[00:51:35] danbri

martin: these bobs can't be matched across documents

[00:51:58]

* DanC_ wonders when the meeting is scheduled to adjourn

[00:52:02] danbri

emiller: i'm a little confused about this convesation...

[00:52:45] DanC_

on opacity, is Mike Dean or PatH here on IRC? the opacity axiom is documented, in draft form, at http://www.w3.org/DesignIssues/Axioms.html#opaque

[00:53:00] danbri

..."we've had compelling test cases / experiences in last few years. W.r.t. serey's point. While we say anyone can say anything about anything, here perhaps we can't. Possible, quite probably, that some things may not have names. Those things without may be difficult to further describe

[00:53:23] danbri

"given our current focus, rdfcore, if you give something an ID, you have good chance of merging data...

[00:53:44] danbri

"for those things that don't... not our business to standardise stuff best done in privacy of own computer

[00:54:00] danbri

em: we might eg like sha1 digests, but not our business to specify

[00:54:21] danbri

pat: that's not the issue. Not about knowing 'the' name, but a name.

[00:54:40] danbri

pat: is the act of giving something a name that big a burden that we can't ask them to do it

[00:54:46] danbri

dan: rdf 1.0 did not make that burden

[00:54:58] danbri

"question is, how do we interpret the doc in our 1.0 syntax

[00:55:11] danbri

pat: rdf m+s text is utterly unclear about notion of anon nodes/resources

[00:55:29] danbri

dan: but we DID tell people to write rdf/xml in this form

[00:56:09] danbri

emiller: the 1st intepretation is what we meant first time round; "i feel terrible for setting community back 4 years, not beating the s**t out of editors/WG 1st time round, but that's the situation

[00:56:15] danbri

ora: it's not all your fault!

[00:56:29] danbri

emiller: so, clean up our mess. That's where we're at...

[00:56:34] danbri

dan: problem is that the implementors have done 2nd/3rd thing

[00:56:44] danbri

frank: can someone clarify diff between 1st and 2nd?

[00:56:55]

* danbri requests emiller's doc for the records. URL please!

[00:57:38] danbri

frank: "if point is to interpret the 1st one as an existential qunatifier... what's the difference between my wanting to refer to that something, that ?x, versus referring to some arbitrary named genid. I think there's a difference.

[00:57:42]

* barstow is enjoying this discussion and would love to see it continue but I'm wondering if we have to stop at the top of the hour because the published schedule said the meeting would end at 6:00pm ...

[00:58:00] danbri

dan: pat made this crystal clear. In assertional case same entailment; in query case, nontrivially different

[00:58:09] danbri

dan: we can't go into queries

[00:58:11] danbri

frank: yes...

[00:58:21] danbri

--

[00:58:48] danbri

pat: (attemptign to sum up)

[00:59:07] danbri

"suppose we have existntials, not generated names, there's no real difference logically. what's thefunctional difference?

[00:59:34] danbri

"you lose a little functionality. if there's a handle provided for every existential. if there's no handle, you lose a little functionality.

[00:59:49] danbri

dan: if we want a handle, make it a uri

[00:59:53] danbri

pat: yes, you could take that line

[01:00:30] danbri

ADJOURNED.

[01:00:32] danbri

---

[01:00:48] danbri

chat...

[01:00:50] barstow

barstow has left channel

[01:00:57] danbri

emiller: i feel progress from last few years...

[01:01:03] danbri

pat: issues are becoming clearer...

[01:01:07] danbri

dinner!

[01:01:23] GK-f2f

GK-f2f has left channel

[01:01:31] danbri

danbri has quit

[01:04:29] dajobe

dajobe has left channel

[01:11:09] DanC_

DanC_ has quit

[04:58:16] AaronSw

AaronSw has joined #rdfcore

[14:48:45] jhendler

jhendler has joined #rdfcore

[14:49:26]

* jhendler lurking (Invite as member of W3-SW-CG)

[14:56:31] DanC_

DanC_ has joined #rdfcore

[14:57:46]

* DanC_ wonders if the WG came to any conclusion on anon-resource etc.

[14:58:04] AaronSw

Doesn't seem like it from the logs...

[15:02:11] DanC_

DanC_ has left channel

[16:12:07] barstow

barstow has joined #rdfcore

[16:12:35] barstow

Ora: Pat and I have been thinking and we agree

[16:12:40] barstow

barstow is now known as scribe

[16:12:59] scribe

... we are concenred about the identity of anon nodes

[16:13:10] scribe

... we do know the identity of the annonnodes

[16:13:24] scribe

... the distictness is reserved.

[16:13:40] scribe

Node -a-> 1

[16:13:46] scribe

Node -b->2

[16:14:18] scribe

Yes

[16:14:37] scribe

In the serialization syntax, we give no names to these nodes

[16:15:14] danbri

danbri has joined #rdfcore

[16:15:50] scribe

Pat: the realization that I have, if I do the MT as attached to the graph, then issues like scope of exist quant go away becauset there are no scopes in the graph

[16:16:48] scribe

... ... In N-Trpiles, annonNodes have ttheir own syntax. The annonNode is unique.

[16:17:52] scribe

ACTION: Pat - I'll re-word the MT wrt my new insight.

[16:18:37] scribe

Pat: wrt entailment, if two nodes have same URI, they can be merged; if they do not, they must not be merged.

[16:18:47] jhendler

(err, I mean what school do you attend)

[16:19:19] scribe

Pat: there is no way in [core] RDF to do the graph merging that Eric showed yesterday.

[16:19:37] scribe

EricM: you are correct, it can be done with additional rules - is not part of core RDF

[16:20:20] scribe

Pat: this resolves wether things are public or private [Brian's issue]

[16:20:35] scribe

Pat: this resolves DanC's issue with existen quantifier

[16:20:56] scribe

Pat: the annonNodes do have ID but this has nothing to do with the graph

[16:22:24] scribe

Frank: if you think of the model as being the graph, the nodes in the ggraph have identify; if I merge the graph, the nodes still have identity; the annonNodes just don't have URI.

[16:23:21] scribe

Frank: wrt serialization syntax, what characteristics do the annonNodes have?

[16:24:46] scribe

Pat: with N-Triples, annonNode are identified by their unique syntax

[16:25:25] scribe

Frank: if you try to merge multiple N-Triple docs, the app must keep info about where the triples came from

[16:25:57] scribe

Steve: N-Trples therefore is not a good syntax

[16:27:49] ASwartz

ASwartz has joined #rdfcore

[16:27:52] AaronSw

AaronSw has quit

[16:28:00] scribe

Graham: 2 vars are distinct if they have diff tags or if they appear in diff n-triples expressions

[16:28:15] ASwartz

ASwartz is now known as AaronSw

[16:28:43] scribe

... when combining 2 ntriple expressions, all of the tag nodes are assigned arbitrary tags such that distinct nodes always have disnt tags in the resulting expression

[16:28:47] GK-f2f

GK-f2f has joined #rdfcore

[16:28:57] scribe

mike: I'm worried about exposing internal names

[16:29:13]

* GK-f2f I think words to cover this are in: http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jul/0397.html

[16:29:41] scribe

Pa: must differentiate naming the node and giving a name to the thing the node denotes

[16:29:51] scribe

s/Pa:/Pat:/

[16:30:04] scribe

Ora: this is NO change to the M&S spec!

[16:30:11] scribe

Pat: I agree!

[16:30:52] scribe

Mike: in the case where internal names need to be exposed, you will loose the fact that it was anon?

[16:30:55] scribe

Pat: yes.

[16:31:33] scribe

Eric: what does this say about the issue - are these thing distinguishable?

[16:31:54] scribe

Pat: Yes.

[16:32:34] scribe

DaveB: I don't understand the need to add scoping N-Triples.

[16:32:46] scribe

Pat: if you merge on the graphs, there is no problem.

[16:32:56] scribe

DaveB: you merge graphs, not N-Triples.

[16:33:51] scribe

... [not n-triples docs]

[16:34:04] scribe

Jan: we've said scope of N-Triples is the N-Triples doc.

[16:34:04]

* GK-f2f so I have a program that reads two different N-triples documents, and spits out N-triples thayt bresult from merging their graphs -- I'd describe that as merging the N-triples documents

[16:34:22] scribe

Steve: I think we still have to deal with the scoping issue.

[16:34:36] scribe

... Ntriples are forcing us to do that.

[16:35:13] scribe

Arno: we ran into this issue at Adobe.

[16:35:43] scribe

... We have diff docs and compound docs. We solved it by

[16:36:07] scribe

... first thinking not about annonNodes but implicitly named [have a name, wwee dont know what it is]

[16:36:25] scribe

... We have a mechanism to refer to them.

[16:36:54] scribe

... What we've said is consistent with our design.

[16:37:11] scribe

Graham: Ntriples is a syntax.

[16:37:59] scribe

Brian: I think we have some agreement but need to test if we are talking about the same thing.

[16:38:25]

* GK-f2f and a "document" is a character string that matches poroductiions from the "distinguished symbol" of the N-truiples grammar ... i.e. a "sentence" of the Np-triples syntax.

[16:38:36] scribe

... I think Ora and Pat said:

[16:38:48] scribe

... the fundamental model is the GRAPH MODEL!

[16:39:06] scribe

... Ntriples is a syntax for a graph [a serialization foor a graph]

[16:39:30] scribe

... We can have more than one graph.

[16:39:35] scribe

Pat: yes!

[16:40:27] scribe

Brian: when I merge ntriples, the semantics is that I'm merging the graphs.

[16:41:28] scribe

... If we have the two graphs, we can't just concatentate the corresponding n-triples; must first change some names

[16:41:58] scribe

Brian: does the MT theory, Pat?

[16:42:21] scribe

Pat: yes, the MT must be based on the graph, not on N-Triples.

[16:42:40] scribe

... won't need the set of triples in a document.

[16:43:09] scribe

Brian: we have an RDF serialization

[16:43:17] scribe

... we will also have a grammar

[16:43:43] scribe

... we will define semantics by defining a mapping from serialization to n-triples

[16:43:58] scribe

... from n-triples, we have a MT

[16:44:15] scribe

... Why do we have to change that?

[16:45:03] scribe

Pat: the arrow from ntriples to MT must now go through the graph

[16:45:09] scribe

... the graph has a MT

[16:45:25] scribe

... the advant: separates some issues

[16:46:01] scribe

Graham: will we have a MT based on the graph [not the ntriples]

[16:46:52] scribe

Ron: want the graph in the middle; put MT in an arc; put n-triples as an arc, put RDF/XML as a arc

[16:47:32] scribe

Ron: if we split an ntriples doc, how to put it back together?

[16:47:57] scribe

Pat: we can break up graphs.

[16:48:28] scribe

... In the graph, nodes are nodes.

[16:50:14] scribe

Ron: use case is controlled vocabularies

[16:50:33] scribe

... a node may have lots of info

[16:50:41] scribe

... may only want to send some of the info

[16:50:52] scribe

... can send the identity of the node

[16:51:39] scribe

jan: if you need to talk about it, give it a URI!

[16:54:07] scribe

Sergey: I'm not convinced we're all talking about the same thing.

[16:55:37] scribe

... want to explore using annonNodes as existential quantifiers, etc.

[16:56:06] scribe

... By looking at these other approaches, we could get more.

[16:57:23] scribe

danbri: I would like to hear Sergey's view.

[16:57:50] scribe

... I would be willing to give up some RDFS time.

[16:58:32]

* AaronSw thinks you're going around in circles

[16:59:00] scribe

---- Sergey -----

[16:59:46] scribe

[Sergey projects a document that contains his model.]

[17:00:02] scribe

ACTION: Sergey - send this document to the WG mail list

[17:00:27] scribe

Annon nodes as existential variables

[17:00:46] scribe

URI constants: c = {c1,...,cN,...}

[17:00:53] scribe

{ exists, & }

[17:01:04] scribe

Variables: {x1,...,xN,...}

[17:01:26] scribe

graph/document = formula without free variab les (most general consensus?)

[17:01:51] scribe

Applications exchange documents in intermediate format (BLOB), but get formulae (graphs) in the end

[17:02:28] scribe

d1 = t(c1, c2, ce)

[17:02:28] scribe

d2 = exists x1 t(c1, c2, x1)

[17:02:35] scribe

d3 = exists x1 [ t(c1, c2, x1) & t(x1, c3, c4) ]

[17:02:43] scribe

Equivalence:

[17:02:55] scribe

Let -> be entailment

[17:03:09] scribe

d1 = d2 <=> d1 -> d2 and d1 -> d1

[17:04:21] scribe

ad 0): t(c1, c2, c3) -> exists x1 t(c1, c2, x1) [Inference that DanC want]

[17:04:36] scribe

ad 1): d1 = exists x1 t(c1, c2, x1)

[17:04:54] scribe

d2 = exists x2 t(c1, c2, x2)

[17:05:21] scribe

d1 -> d2 and d2 -> d1 => d1 = d1 (fine)

[17:06:03] scribe

ad 2): d1 = exists x [ t(c1, c2, x) & t(x, c3, c4)]

[17:06:16] scribe

How split?

[17:06:43] scribe

[ed note: ... d1=>d1=d2 above]

[17:07:02] scribe

d1' : exists x1 t(c1, c2, x1)

[17:07:26] scribe

d1'' : exists x2 t(x2, c3, c4)

[17:08:03] scribe

Merge: d1''' = exists x1 t(c1, c2, x1) & exists x2 t(x2, c3, c4)

[17:08:34] scribe

d1 -> d1''', but d1''' -/-> d1 => d1 != d1'''

[17:08:58] scribe

irrerversibel change when docs are split and merged (bad?)

[17:09:21] scribe

ad 3: impossible to refer to anon. node in another document withing the model

[17:09:40] scribe

d1 = exists x1 t(c1, c2, x1)

[17:09:59] scribe

d2 = exists x2 t(c1, c2, x2)

[17:10:15] scribe

no way to ask: is x1 in d1 same as x2 in d2?

[17:10:28] scribe

Anonymous nodes as local constants:

[17:10:41] scribe

(Implementaiton perspective)

[17:10:57] scribe

URI constants: C = {c1,...,cn,...}

[17:11:57] scribe

Local constatns: PRG1 = {l1_1,...l1_N,lll}, PRG2 = {l2_1,...,l2_N,...}

[17:12:11] scribe

Rule: prg1 cannot see constants in document

[17:12:41] scribe

ad 0): t(c1,c2,c3) -/->, <-/- t(c1,c2,l1) [fine]

[17:13:08] scribe

ad 1): d1 = t(c1, c2, l1_1), second parse d2 = t(c1,c2,l1_2)

[17:13:18] scribe

d1 != d2 [bad?]

[17:14:13] scribe

...

[17:14:35] scribe

[ed. note: I give up - assume Sergey will post this his file to the list ...]

[17:16:06] scribe

...

[17:16:17] scribe

Observations:

[17:17:38] scribe

o A does not caont URI (disjoing) If A and C overlap, we cannot distinguish anon. nodes from the others. But: since the same procedure for assigingin constants from A, this is irrelevant. A can be viewed as subset of C that is extremely unlikely to be used

[17:18:25] scribe

o Application that neeed not communicate may not local IDs. If communicate using syntax that contains "holes", fine. No global autogeneration algorithm required.

[17:19:01] scribe

o If no standard assignment algorithm is required, ad 1) is still violated (parsing twice)

[17:20:46] scribe

....

[17:21:13] scribe

Sergey: there can be a formal mechanism to help base arguments about anon. nodes.

[17:21:41] scribe

... point 2: there are multiple options for implementing anon. nodes [they all have advantages and disadvantages]

[17:22:45] scribe

... want to ground the decisoion. This also helps define the application semantics.

[17:22:55] scribe

Jan: this is very useful. However, your existance proof is false.

[17:23:28] scribe

... You have no way of knowing where things come from.

[17:23:56] scribe

... The algorithm doesn't reflect that a URI may return the same thing through time.

[17:25:14] scribe

Pat: the question is what is the semantics. Is is temporary [the doc]?

[17:25:26] scribe

... M&S is talking about graphs.

[17:27:09] scribe

Sergey: I think explicit genids would be useful.

[17:27:34] jhendler

jhendler has quit

[17:28:03] scribe

Brian: if you parse the same doc, should an anon description have the same identity?

[17:30:25] scribe

Ron: if everyone generates the same identifier for an anon node, it could be useful but it also could be dangerous.

[17:30:37] scribe

... That's the choice: useful vs dangerous.

[17:30:54] scribe

danbri: it is very dangerous.

[17:31:22] scribe

Brian: can you do more things this way?

[17:32:07] scribe

Ron: you can do more things because then you can hang additional stuff off of it.

[17:32:19] scribe

Jan: there are better ways to do this.

[17:32:39] scribe

Frank: what exactly is the question?

[17:33:27] scribe

Brian: do you need everyone to use the same algorithm?

[17:33:49] scribe

Ron: if folks agree on an algorithm, you can do additional stuff.

[17:35:49] scribe

Pat: I don't think these examples are helpful. They introduces more confustion. It doesn't talk about the graph.

[17:36:21] scribe

... We don't need to introduce the variables.

[17:37:32] scribe

Eric: we need to put a stake in the ground and move on. We need to focus on the graph to agree on stuff.

[17:38:38] scribe

Pat: I'll update the MT based on the graph within a week.

[17:58:11] scribe

======== Break Over =========

[18:00:20] GK-f2f

RDF Syntax -- Dave Beckett Leads

[18:01:51] GK-f2f

http://www.ilrt.bris.ac.uk/people/cmdjb/talks/rdfcore-sebastopol/

[18:02:57] GK-f2f

slide 2

[18:03:18] GK-f2f

slide 3

[18:03:24] GK-f2f

(ntriplesreview)

[18:03:27] GK-f2f

GK-f2f has quit

[18:03:31] jang

jang has joined #rdfcore

[18:04:34] jang

(talks about benefits of ntrlpes as simple serialisation)

[18:04:41] jang

slide 4

[18:04:50] DanCon

DanCon has joined #rdfcore

[18:05:21] jang

dave'd been looking at reexpressing the grammer in terms of the infoset, rather than '<' etc

[18:05:39] jang

grammar looks simpler, smaller

[18:06:23] jang

we o xml syntax -> ntriples -> graph -> MT

[18:06:27] jang

s/o/go

[18:06:39] jang

points at:

[18:06:57] jang

http://ilrt.org/discovery/2001/07/rdf-syntax-grammar/

[18:07:16] jang

dave points out, as an example, the 6.3 entry in the table

[18:07:40] jang

dve's had to invent a syntax for expressing XML infoset

[18:07:57] jang

slide 5

[18:08:31] jang

change to slide: <=> "the graph" in each case

[18:08:50] jang

dave mentions other formal proposals for syntax notations

[18:08:55] jang

slide 6

[18:09:09] jang

answer to "is ntriples sufficient?" DB: yes

[18:09:15] jang

(db = dave beckeytt)

[18:09:36] jang

still open on question 2: what formalisms should be used to express grammar

[18:09:45] jang

=> slide 8

[18:09:47] jang

and done

[18:10:01] jang

wy is BNF so bad?

[18:10:27] jang

we can use BNF, but sould be expressed in terms of infoset, not <, & etc

[18:11:04] jang

PH: ntriples syntax may bem misleading

[18:11:14] jang

URIRef and anonnode should be changed

[18:11:19] jang

DB: I've already changed that

[18:11:52] jang

brian:

[18:11:57] jang

there are two questions

[18:12:05] jang

1. how we repsent the grammar formally and precisely

[18:12:12] GK-f2f

GK-f2f has joined #rdfcore

[18:12:17] jang

2. how do we define the transform from rdf/xml into core representation

[18:12:27] jang

(and is it mechanically executable)

[18:12:46] jang

closest bri has are attribute grammars

[18:12:54] jang

danbri: schematron is the closest I've seen

[18:13:11]

* AaronSw (informally) proposes XSLT

[18:13:14] jang

db: uneasy about it's completeness

[18:13:33] jang

restating question about transformation:

[18:13:44] jang

rdfxml must be translated by a parser into some representation of the graph

[18:14:02] jang

is there a way of describing this transformation mechanically and executably?

[18:14:13] jang

SM: there's a new parser that uses javacc

[18:14:25] jang

javacc ives you the grammar definition

[18:14:35] jang

you can introduce bits of code into the grammar it uses

[18:14:51] jang

bri: that's basically an attribute grammar with java as the attributions

[18:15:05] jang

danbri: we sould note xslt has been used for this

[18:15:21] jang

bri: tried it, it was very large, not a good way of descibing the transformation to an implementor

[18:15:30] jang

danc has also got an xslt parser, danbri knows of another

[18:15:35] jang

(can't remember who by)

[18:15:52] jang

bri: wants a compact gramamr that can be transformed into xslt, for instance

[18:15:57] jang

danbri: that's what schematron does.

[18:16:06] jang

bri: talks about jeremy's parser

[18:16:25] jang

he had the problems due to M+S talking about characters.

[18:16:26] danbri

the other rdf xslt parser was by jason diamond

[18:16:33] jang

so he did javacc with a grammar in terms of SAX events

[18:16:48] jang

this is pretty handy

[18:17:05] jang

dave's looking for suggestions; he's stil inthe investigation phase

[18:17:15] danbri

xslt parsers: see http://www.xmlhack.com/read.php?item=757

[18:17:34] jang

art: is the WG constrained to using w3c or other standard?

[18:17:44] jang

are we bound to standardised mechanisms?

[18:18:12] jang

em: i don't think so. We must be able to represent the grammar in something that's as familiar as possible to the other xml techs

[18:18:33] jang

em: likes XDuce; but if it's one of equals he'd prefer sometng else

[18:18:39] jang

dan: relaxng looks promising too

[18:19:03] jang

danri: if we want others to have a good look at our spec, we should extend them the same courtesy

[18:19:23] jang

em: only a few people from, eg, DC will be interested in reading the spec?

[18:19:33] danbri

(re politeness: specifically within the w3c xml family of specs)

[18:19:35] jang

daveb: we want a single normative mechanism

[18:19:46] jang

em: ebnf, for example.

[18:19:54] jang

daveb: it's basically in terms of characters

[18:20:06] jang

sm: can xslt produce non-xml docuemnts?

[18:20:08] jang

all: yes

[18:20:35] jang

sm: asks for smple syntax

[18:20:45] jang

dave: we've got ntriples, that's what we've been using

[18:21:03] jang

sm: what about dealing with reification, literls, etc.? it's going to keep growing

[18:21:23] jang

brian: is ntriple broken? do we anticipate it stopping working?

[18:21:34] jang

ph: the only possible problem is scoping, i think we've resolved that

[18:21:46] jang

bri: then we stick with ntriple until it's demonstrated that it's broken

[18:22:12] jang

jang volounteers to help dave with the investigation

[18:22:19] jang

dave asks: can we include jeremy?

[18:22:25] jang

art: also interested

[18:22:32] jang

brian: AP! ask jeremy about this

[18:22:42] jang

dan: suggests schematron to invesigate

[18:22:59] jang

AP: ang, dave, art to investigate and come back with a reccommendation

[18:23:16] jang

Graham: let's be absolutely clear wat we consider the primary audience?

[18:23:40] jang

graham: i ask because XSLT exists and may be very good, but it is probably not very good at expressing the concepts to a human reader

[18:23:48] jang

is the human developer the primay audience?

[18:23:50] jang

bri: yes

[18:24:02] jang

sm: s ntriple going to be xmlised?

[18:24:05] jang

bri: no plans yet

[18:24:12] jang

sm: then we can't use xslt?

[18:24:21] jang

al: no, it can produce anything including text files

[18:24:30] jang

em: it can do tree transforms to text

[18:24:40] jang

steveP: what is the role of the ntriples syntax?

[18:24:46] jang

normative or for testing?

[18:24:54] jang

I'd be opposed (I think) to it being normative

[18:25:13] jang

bri: we need a way to represent the graph. we have to be able to write down the graph transformation

[18:25:20] jang

bri: in mymind, ntriples is for that

[18:25:24] jang

steve: but it's not a graph

[18:25:37] jang

ph: we could draw pictures in the spec

[18:25:54] jang

we need ntriples for testing, not for the defintiion

[18:26:11] jang

brian: I need some way of writing down my test-cases

[18:26:25] jang

I'd rather use one representation of a gaph

[18:26:30] jang

graph, even

[18:26:56] jang

dave: we've got a mixture of text and formalisms at the moment

[18:27:04] jang

ph: software exists to construct and transmit graphs.

[18:27:23] jang

ph: wy don't we make the exposition in the definitive document conform to the graph directly?

[18:27:40] jang

every time a graph is pictures, we can give the ntriples representation

[18:28:02] jang

danbri: i aree largely, but I'd be concerned if we say all sorts of wooly non-normative thngs about ntriples

[18:28:08] jang

it's as normative as the rest of the spec

[18:28:28] jang

ph: I was responding to brian's desire that ntriples be the way graphs are described

[18:29:19] jang

fm: one role ntriple could play in the exposition is to illustrate some of th epotential misunderstandings they may experience

[18:29:47] jang

people have been sending ntriple-ish stuff back and forward for disambiguation by email

[18:29:57] jang

so ntriples could be used as an example of a serialisation

[18:30:02]

* danbri agrees

[18:30:08] jang

to make the point that the graph model is the central issue

[18:30:32] jang

brian: when I think of ntriple, i think it behaves exatly like a graph

[18:30:46] jang

ph: all the issues of name scoping have not been properly articulated

[18:31:09] jang

graham: the advantage of using graphs directly: it'll prevent opthers from falling into the same mental trap

[18:31:32] jang

em: we've been trynig to do this for 4 years, unsuccessfully: saying "it's the graph stupid"

[18:31:46] jang

people understand the serialisation more than the abstract notion

[18:31:53] DanCon

DanCon has left channel

[18:31:54] jang

ora: I don't think that's true.

[18:32:19] jang

people see the serialisation and don't understand it represents a graph

[18:32:28] jang

em: eg,xml people see it as an xml document

[18:32:50] jang

graham: we should do everything twice in the document: once as a graph and once as ntriples

[18:33:17] jang

ora: in some sense, choosing xml was a mistake. people see xml and consider it to be just xml, not a graph

[18:33:38] jang

every time i speak about the graph, people get it though. I've stopped talking about xml and people just get it

[18:33:46] jang

em: I've seen exacty the opposite

[18:34:03] jang

people ask, "but what does it look like?" meaning, where are the angle-brackets

[18:34:41] jang

em: people are deploying a lot of apps that just happen to be rdf-friendly, eg, rss - most users just consider it to be an xml document

[18:35:12] jang

dan: what are we trying to achieve? we're not trying to write the rdf tutorial or do modelling

[18:35:57] jang

we're not trying to write the tutorial here - in that context, does anyone have anything else to add?

[18:36:20] jang

mike d: we've produced another serialisation for rdf. if we play it up, won't people start using it?

[18:36:36] jang

how important is it to emphasise that the xml serialisation is the preferred syntax

[18:36:45] jang

bri: we're not chartered to develop a new synta

[18:37:00] jang

mike: it's becoming bigger. it's for test cases basically

[18:37:24] jang

ph: it sends a good messge - there are at least two maybe more serialisations of rdf

[18:37:35] jang

M+S doesn't hammer this home sufficiently

[18:37:54] jang

arno: this can create some confusion. eg, DC has multiple representations

[18:38:14] jang

documents that have different forms tend to be interpreted differently

[18:38:36] jang

em: let's not go there yet. it is non-trivial to convince people to deploy multiple syntaxes

[18:38:58] jang

ph: we either say, there is one preferred syntax, and not mention any others, or we shoud say

[18:39:06] jang

rdf is about graphs and there may be muktiple syntaxes

[18:39:24] jang

danbri: this has been very important to exaplin to people.

[18:39:50] jang

em: I want that, yes: we're building on the first M+S

[18:40:16] jang

this diagram (referring to the graph -> MT, ntriples, serialisation diagram) is really importnat

[18:40:35] jang

e: priority should be to clarify the model (graph) and focus on the rdf xml serialisation

[18:41:12] jang

ora: says "S-expressions" and gets lynched

[18:41:12] jang

brian: moving on to schema

[18:41:12] jang

minute break, back in a tick...

[18:55:51] jang

back: rdf schema issues

[18:56:00] jang

danbri to lead, brian to timekeep + chair

[18:56:15] jang

eric has noes on laptop

[18:56:21] AaronSw

http://lists.w3.org/Archives/Public/www-archive/2001Aug/0003.html

[18:56:30] AaronSw

- notes

[18:56:32] jang

ap: eric/danbri to ensure dan's document goes online

[18:56:41] scribe

======== DanBri - RDFSchema ========

[18:57:57] jang

w3c think rdfsch is more or less done

[18:58:06] jang

we've had feedback... particularly from daml+oil

[18:58:36] jang

rdfschema work stopped waiting for xml schema - that's now done

[18:58:49] jang

we need to take the useful bits ( datatypes) into rdf schema

[18:59:08] shellac

shellac has joined #rdfcore

[19:00:01] jang

WG sucessor (web-ontology) is planned

[19:00:55] jang

dan points out: what we decide/do next doesn' have to be writen in stone, so we make pragmatic ecisions on what the next WD looks like

[19:01:55] jang

domain + range is an open issue; dan proposes we skip over this because there's a good answer

[19:02:19] jang

this is a no-brainer

[19:02:47] jang

ora gives a bit of background to rdfschema; we're after an OO extensible type system to rdf

[19:03:02] jang

we're after very little.

[19:03:54] jang

ora: te properties of properties are global - no class-specific constraints

[19:03:59] jang

we fixed this in daml

[19:04:17] jang

domain + range: this is the open issue

[19:05:12] jang

dan talks about daml work getting pushed into schema/web-ont - we don't know or care yet what's going to ahappen about this

[19:05:23] jang

dan: ora - class-contextualised constraints may come later

[19:05:50] jang

dan: any dissent to conjunctive interpretation of range+domain?

[19:07:04] jang

art: is there any evidence that people are using current semantics?

[19:07:07] jang

jan: I've seen some

[19:07:14] jang

AP: jan to write up the fix/workaround for this

[19:07:36] jang

ron: possible to change the namespace to not break stuff for people?

[19:07:55] jang

dan: yes, it's possible, my preferred take:

[19:08:15] jang

there is a thing called rdf:domain which the rdfschema people have previously made an erroneous statement about

[19:08:36] jang

ph: introducing a new namespace isn't always the most painful thing

[19:08:53] jang

AP: rdf schema editor to fold conjunctive decision into the raph

[19:09:53] jang

APPROVED: multiple domains, ranges, with conjunctive semantics

[19:10:01] jang

pretty much carried unanimously

[19:11:21] jang

approved by ora, brian, art, jos, dave b, martin, ph, ron d, frank m, sergei, kwon, em, arno, stephen p, jan, raham,

[19:11:31] jang

we record no objections: ron daniel abstained

[19:11:44] jang

(danbri also voted in favour)

[19:12:26] jang

rdfs:domain & rang constraingts or rdfs:domain were missing from the schema - this is just a typo

[19:12:33] jang

proposal to fix this

[19:13:17] jang

ron: was the editorial oversight due to non-discussion/ non-decision or was a decision recorded

[19:13:24] jang

but didn't make it to the doc?

[19:13:37] jang

dan: not certain; but the pictures we had show these values

[19:13:46] jang

proposal: editor to fold these into the next WD

[19:14:36] jang

all in favour, no abstentions, no against

[19:14:40] jang

APPROVED

[19:15:15] jang

subclassing containers...

[19:15:36] jang

dan: a compelling case for this was not allowed

[19:15:41] jang

s/allowed/made

[19:15:47] jang

for the next wd, we say: future work

[19:16:18] jang

AP: jan to provide explanation of how we'd add this

[19:17:35] jang

proposal: no change on this issue in next draft of rdfs

[19:17:50] jang

we take as resolved on the issues list

[19:18:13] jang

(recording accurately the nature of the resolution)

[19:19:50] jang

all in favour: abstain frank, no against

[19:19:55] jang

APPROVED

[19:20:34] jang

datatyping....

[19:21:56] jang

ron: originally we discussed this and decided to wait for xml:schema

[19:22:12] jang

proposal is to take in what DAML+OIL did, throw it in and then argue abot ti later

[19:22:19]

* GK-f2f INFO: CC/PP uses XML schema datatypes - http://www.w3.org/TR/CCPP-struct-vocab/

[19:22:41] jang

this sin't for the next WD, just as the next step

[19:23:11] jang

proposal (refined)

[19:23:25] jang

we expect to work in this area, informed by the daml+oil work

[19:25:41] jang

AP: graham to send to working group how CCPP does datatypes

[19:26:01] jang

AP: ajn to do the same with the EASEL approach

[19:26:20] jang

ora: the daml+oil approach is clever because if you don't care, you don't get hurt

[19:26:49] jang

proposal: to go away and investigate and report back to the group

[19:27:00] jang

dan: taskforce to consider the adoption of...

[19:27:43] jang

adopt daml+oil/xml datatypes as initial foray into the issue

[19:27:55] jang

we don't consider closure on this issue a must-have for the next WD

[19:28:18] jang

drop the "adopt"

[19:28:30] jang

final proposal should come from EM's document, he's editing it now

[19:29:22] jang

volounteers: danbri graham, martin, jan

[19:30:15] jang

all in favour of the taskgroup

[19:32:10] jang

brian leaves to order pizza

[19:33:48] jang

rdfs-primitive-properties

[19:36:28] jang

AP: pat to go into some more detail on why the know-tying at the top of the hierarchy in rdfs is not a set-theoretical hole

[19:37:14] shellac

shellac has quit

[19:37:20] jang

s/know-tying/knot-tying

[19:38:04] jang

proposal: we don't think this is a problem

[19:38:32] jang

so we close the issue, with the expositional urden associated

[19:39:52]

* GK-f2f INFO: Horrocks, et al paper is at http://www.semanticweb.org/SWWS/program/full/paper40.pdf

[19:40:20] jang

dan explains how we go around answering this (process issues)

[19:40:30] jang

we're obliged to respond to feedback

[19:40:57]

* GK-f2f Last URI was wrong one ... still looking

[19:41:42] jang

all in favour. no abstain, no against (brian absent)

[19:41:59] jang

cycles in subclassof

[19:42:02]

* GK-f2f I think this is the right one: http://www.semanticweb.org/SWWS/program/full/paper11.pdf

[19:42:36] jang

daml+oil didn't like this

[19:42:58] jang

dan: personal bias towards what we've got is a bad usability problem explaining it to people

[19:43:25] jang

graham: one reason to change is that this is one of those things that can't be described within the schema framework

[19:43:27] jang

thus we drop it

[19:43:46] jang

ora: doesn'tcare

[19:44:05] jang

dave: biggest implementation in this area doing this already? will we break something?

[19:44:17] jang

dan: opens floodgates

[19:44:34] jang

ron: recalls experience was that people wouldn't want cycles in te subgraph relationship

[19:45:00] jang

this will break a lot of implementations if we use this. in particular, stuff out there won't be doing cycle detection

[19:45:12] jang

danbri: the system is gullible if it's not checking this restriction

[19:45:33] jang

ron: no, that's not fair - if the spec tells you there are no cycles, then there is a performance optimisationto not bother checking

[19:46:01] jang

frank: that argument can be ade for any syntactic description. this is absurd if taken to its extreme conclusion

[19:46:20] jang

dan: I was going to ask someone from DAML how strongly they care

[19:46:36] jang

graham: we need to make the change ASAP if we're going to hurt as few people as possible

[19:47:11] jang

dave: rather than do nothing, I want to know now what we're doing, are we gogin

[19:47:17] jang

to change it

[19:47:51] jang

PH recalls why daml people wanted this - it was fought hard over

[19:48:01] jang

ora: frank + ian do class equality doing this

[19:48:01] jhendler

jhendler has joined #rdfcore

[19:48:10] jang

which frees them from having another relationship

[19:48:16] jang

and that this is DL accepted practice

[19:48:58] jang

frank: I did circulate a paper sumarising the major arguments for this change from DAML

[19:49:11] jang

people are going to write these; what do we want to happen?

[19:49:56] jang

do we barf, notice this, flag it up as a possible problem, etc?

[19:50:11] jang

frank: there are a number of large-scale ontologies with cycles.

[19:50:52] jang

PH: the critical case for DAML+OIL thinking was that subsetting relationship might be made by multiple people

[19:51:12] jang

the logical conclusion is that the two classes are co-extensive

[19:51:47] jang

PH: the issue is, is subclassof lessthan or lessthanorequal

[19:52:00] jang

one has cycles, one doesn't; we really need both

[19:52:34] jang

EM: when merging large ontologies, we can't prohibit cycles from happening

[19:52:40] jang

the issue is, what does it mean? (PH)

[19:53:09] jang

this is the only place where two ontologies could contradict each oter (in rdf + rdfs)

[19:53:20] jang

dan tries to close this

[19:53:37] jang

can we resolve that people who care about this go away and come back with some advice?

[19:53:56] jang

ron would like a vote

[19:54:14] jang

brian returns at this point.

[19:55:09] jang

em: options: taskforce, or discuss now (useful)

[19:55:18] jang

ron: third option: strawpoll?

[19:57:23] jang

dave: programmers from OO languages dont like this

[19:58:27] jang

non-binding strawpoll

[19:58:45] jang

insufficient consensus on this

[19:58:54] jang

dan stresses we're only talking about the next WD

[19:59:52] jang

sergei says why he's against (because of dave's point)

[20:00:58] jang

ron: suggests we record this that we insert a question into the next WD

[20:01:04] jang

asking for feedback

[20:01:10] jang

this now becomes the proposal....

[20:01:23] jang

ORA: AP - talk to ian +frank t get the background on this

[20:02:55] jang

proposal: we stick something in the WD saying "we're looking for feedback - we're going to pull this - how badly does it hurt?"

[20:03:11] jang

PH: daml will invent daml:subclassof f you don't take this out

[20:03:44] jang

frank: the daml+OIL people gave us explicit feedback, which strongly mentioned this

[20:04:27] jang

frank: also want explicitly recognised that frank sent the feedback t the WG list

[20:04:41] jang

this shouldn't be news to us we HAVE feedback already!)

[20:05:36] jang

ron: by nserting this in the document then this becomes a resonse to the DAML+OIL people

[20:05:44] jang

PH: that sounds perfectly fine

[20:06:09] jang

jos: we're discussing subpropertyof too

[20:06:23] jang

danbri: yes, we take this to be the case

[20:06:45] jang

can frank modify his document into something to put in the next WD?

[20:07:30] jang

em: propose flaed in the next draft

[20:07:48] jang

also: someone (frank) to go back to the DAML+OIL people and ask for moe convincing arguments

[20:08:21] jang

AP: pat to take this back to the DAML people at the next telecon and bring the feedback to us

[20:09:02] jang

summary: OO programmers are confused, people are trying to code-generate classes (java) for this

[20:09:06] jang

thus we have to go back

[20:09:39] jang

proposal: open the issue, take the stuff to DAML (PH) and continue the discussion

[20:10:03]

* AaronSw (informally) notes java doesn't even have multiple inheritance, so it's not really a good example

[20:10:08] jang

frank to own this issue.

[20:10:22] jang

aaron: it has multiple inheritance of interfaces

[20:10:30] jang

(after a particular fashion)

[20:10:38] jang

not of implementations

[20:10:56]

* AaronSw scrunches his face up...well, yeah

[20:12:21]

* GK-f2f Java no MI of classes, true, but it does have MI iknterfaces

[20:12:22]

* AaronSw notes (informally) that python allows inheritance cycles

[20:12:30] jang

propose: open issue (frank owns ) plus PH, ora to take back to DAML any feedback from this

[20:12:35] jhendler

jhendler has left channel

[20:21:49] jimH-lurk

jimH-lurk has joined #rdfcore

[20:42:20] scribe

scribe has quit

[20:44:10] barstow_

barstow_ has joined #rdfcore

[20:50:25] jang

back after lunch

[20:50:47] jang

plan: finish schema, open mike

[20:50:51] jang

danbri...

[20:51:25] jang

spo semantics (inheritance)

[20:51:35] jang

inheritance of range+domanin

[20:51:48] jang

seem to have fixed a lot of this with range & domain

[20:53:16] jang

jan: subproperties should inherit conjunctively the range+domain of their superproperties

[20:53:24] jang

ron: is there a clarification that's been asked for?

[20:54:48] jang

general agreement with jan's statement of this

[20:55:24] jang

AP: dan clarify prose to reflect this position accurately

[20:56:03] jang

then issue closes

[20:56:30] jang

subclass of a subproperty

[20:56:40] jang

(previous issue RESOLVED)

[20:57:30] jang

are rdfs:Class and rdfs:Property disjoint?

[20:57:45] jang

ora has seen an instanc of this

[20:57:58] jang

didn't see any reason why this shouldn't be permitted

[20:58:28] jang

dan talks about rss:image

[20:58:47] jang

PH: the real question is "what does this mean?"

[20:59:15] jang

em: finds it very confusing

[20:59:41] jang

dan: default thing is to do nothing; the spec's silent on this

[20:59:56] jang

dan's candidate meaning is "coincidence" - ph's MT does this

[21:00:44] jang

proposal: can we restate this as "are Property and Class disjoint?"

[21:01:07] jang

the proposal is to record this issue in this style

[21:01:54] jang

and to do nothing in the next WD

[21:03:56] jang

we move on

[21:04:23]

* barstow_ notes to GK that MIT [thus W3C] is experiencing network problems ...

[21:04:37] jang

(this is our resolution)

[21:04:52] jang

onlie char encoding

[21:05:12] jang

proposal: editor to fix this

[21:05:16] jang

RESOLVED

[21:05:25] jang

(we don't want to rathole on this nobrainer)

[21:05:30] jang

versioning:

[21:05:36] jang

known and had problem

[21:05:40] jang

s/had/hard

[21:05:59] jang

this is very very difficult. Nobody really appears to know how to do this. Open research issue

[21:06:17]

* AaronSw doesn't think so

[21:06:41] jang

Proposal: note this is very hard, close the issue.

[21:07:30] jang

PH: proposal "wepropose to not answer this!"

[21:07:57] jang

in other words, we leave the spec as it is

[21:08:11]

* AaronSw is vehemently opposed to that proposal

[21:08:15] jang

moving on

[21:08:22] jang

you're not here; take it to email

[21:08:49] jang

transitive subproperty

[21:09:35] GK-f2f

Jan: Counter-example sisterOf subproperty of siblingOf

[21:09:50] jang

proposal: the answer is "no" it's not transitive

[21:10:13] jang

AP on anyone who cares: find an explanatory piece of prose on this

[21:10:17] jang

AP on JAN to do this

[21:10:41] jang

AP on editor: chase jan on this

[21:12:32] jang

movin on

[21:12:46] jang

(ron notes that he's not convinced in this case)

[21:12:56] jang

we make no changes to the next draft; this issue remains open

[21:13:33] jang

we do that. frank, steve P abstain

[21:13:37] jang

next

[21:13:44] jang

subclassof and instance clarification

[21:16:03] jang

frank: we must ensure that we consider the original email messages, not a summary of the issues

[21:17:08] jang

that and the next two issues (isDefinedBy semantics and editorial)

[21:17:28] jang

no action on tehse before the next WD

[21:17:44] jang

carried; a couple of abstentions (jan, steve P)

[21:19:08] jang

we leave these until later

[21:22:44] jang

new WD in one month

[21:22:44] jang

new WD of rdfs due on september sixth

[21:24:16] jang

jan notes he has acounterexample to the transitive subproperty of subproperty question

[21:29:58] jang

we go on to "where next"?

[21:30:05] jang

schema new WD by sept 6

[21:30:17] jang

syntax: we have a taskforce

[21:30:27] jang

model: pat has an action on him

[21:30:45] jang

also: second half - sergei's mechanisms for implication analysis

[21:31:12] danbri

danbri has quit

[21:32:01] jang

ron especialy points out that splitting is not a requirement, merely something to consider

[21:33:04] jang

steve p asks : can we actually get a proposal out of this?

[21:34:01] jang

PH; two pieces of rdf are identical iff they map to the same graph

[21:35:23] jang

s/identical/equivalent

[21:36:02] jang

two rdf documents are equivalent iff they map to teh same RDF graph

[21:37:41] jang

two rdf graphs are the same when :

[21:37:51] jang

1. they are graph isomorphic

[21:38:03] jang

2. no two nodes are labelled with the same URI

[21:42:18]

* AaronSw doesn't get point 2 at all

[21:43:43]

* GK-f2f stuff on graph theory at: http://www.math.fau.edu/locke/graphthe.htm

[21:44:14] jang

we can't specify this precisely

[21:44:14] jang

so we agree that this needs more thinking about

[21:44:14] jang

we HAVE agreed that the graph is the central idea to RDF

[21:45:42]

* GK-f2f see in particular 1st para of http://www.math.fau.edu/locke/graphmat.htm

[21:45:46] jang

ron: the graph is the central concept for RDF, there are multiple graphs

[21:46:15] jang

ron reads out a whole bunch of statements that indicate we need to think

[21:46:22] jang

AP: ron to send this to the list

[21:47:01] jang

frank: as a matter of exposition, the graph model is central and the other representations are to be interpreted n that light

[21:47:06] shellac

shellac has joined #rdfcore

[21:47:06]

* AaronSw knows what isomorphism is... not sure why RDF needs a special requrement though

[21:47:11] jang

the current text doesn't really make this clear throughout

[21:47:53] jang

in the course of making these points, we have to be careful that the message is carried throughout the whole document

[21:48:26] jang

brian: agrees; we're lookig fora rewrite, not an editing job.

[21:50:00] jang

we look at te schedule

[21:50:34] jang

we're running a lttle behind :-)

[21:51:10] jang

are there better notions of what revised dates we should commit to?

[21:51:30] jang

em: we should discuss what our delivrables are

[21:51:38] jang

we know one:rdfs, we have a date

[21:52:08] jang

re: pat's attempts aove: jang greed they had the same logical entailment, but that that was not where teh anon node issue lied

[21:52:14] jang

s/lied/lay

[21:52:25] danbri

danbri has joined #rdfcore

[21:53:57] jang

we ask how many people would be interested in focussing on a document

[21:54:23]

* AaronSw signals agreement

[21:55:05] jang

if we had to pick to each of:

[21:55:11] jang

primer, model, df/xml, schema

[21:55:20] jang

which would they be?

[21:55:41] jang

we add "test case repository" as a deliverable

[21:56:48] jang

interested in primer: 5

[21:56:57] jang

model: 8

[21:57:07] jang

syntaxL 4

[21:57:13] jang

schema: 4

[21:57:17]

* AaronSw volunteers for primer

[21:57:22] jang

test cases: 2

[21:57:45] danbri

aaron/primer: :)

[21:58:32] jang

AP; (repeated) action item to get rdfs done

[21:58:43] jang

some of these depend on pats revised model

[22:00:19] jang

danri: we can get the telecon bridge available at other times too

[22:01:59] jang

ora: are we issuing a version of the existing spec or a new spec?

[22:02:37] jang

as comparison, there is a new XML spec.

[22:03:24] jang

ora notes that we tried to originally eparate model and syntax, and it was too hard

[22:03:49] jang

pat: is the document primarily definitive or understandable?

[22:04:57] jang

em/ora: why we smushed the documents together originally

[22:05:18] jang

we were looking for primer and spec and al sort of things

[22:05:32] jang

dave: document format is to be left ntil much later, let's produce the pieces first

[22:07:13] jang

ow many people are interested in being the editor/document layout person

[22:07:14] jang

graham is

[22:07:38] danbri

danbri is

[22:07:51] jang

graham: i sense there's significant support for the idea that model and syntax be separated

[22:07:56]

* AaronSw is interested in nitpicking

[22:08:06] danbri

(danbri is...interested in being on any group working on document partitioning)

[22:08:21] jang

good, but people are talking - I'm not going to butt in with this one (to email, you'll not be left out)

[22:08:56] jang

dave:do we need coordination?

[22:09:13] jang

em: yes, really. I'm looking for where this can take place/be centered

[22:09:23] jang

s/centered/centred

[22:09:58] jang

ron: proposal to identify a team leader for each of the items, including overall documen structure

[22:10:06] jang

this is har work, but I think that's what we need

[22:10:57] GK-f2f

brian M suggests pick a leader for the overall breakdown, and defer selecting others

[22:11:40] jang

brian proposes to take the document leader job - it's the chair's jo

[22:11:44] jang

hear, hears

[22:12:02] jang

that is document structure ONLY

[22:12:21] jang

AP: brian to take the list of sections and come back with something more cocrete to look for volounteers

[22:12:55] jang

frank: could we consider structuring theseas web things instead of PODs?

[22:13:06] jang

there, we close.

[22:13:27]

* GK-f2f I think the docs should be printable as PODs if required

[22:13:42] jang

reopen: schedule rearrangement

[22:13:42]

* GK-f2f (POD = plain old document?)

[22:13:45] jang

yep

[22:14:13] jang

brian thinks that www11 would be a good place to annonce rec

[22:14:14]

* AaronSw thinks web things == goodness

[22:14:35] jang

hard narrative stuff and hard to print out to read onthe plane :-)

[22:14:58] jang

www11 is in may 2002

[22:15:25]

* AaronSw thinks web things != no print version

[22:16:15]

* GK-f2f yes, but I woukld want it to be a single printable doc, not lots of separate web "pages"

[22:16:41]

* AaronSw thinks that this is what XML is for -- one XML document can be distributed in multiple versions

[22:16:58]

* AaronSw also likes one-document specs, FWIW

[22:18:30] jang

AGREED: to announce REC at www2002

[22:18:35] jang

(or aim for that)

[22:19:23] jang

kwon's presentations

[22:19:23] jang

AP: kwon to get this on the web

[22:22:30] danbri

(hmm... agreed: We would really really like to announce REC at www2002...)

[22:25:46] jang

kwon's questions....

[22:25:55] jang

(from last side)

[22:26:10] jang

PH: rdf useful within machines for storing metadata

[22:26:21] jang

is this an "in" for getting RDF involved?

[22:26:54] jang

kwon's wg chair wants to go with rdf

[22:27:00] jang

but they're suffering from tool availablility

[22:27:25] jang

metametadata storage is still currently hard, esp. with rdfs in its current state

[22:27:39] jang

em: lots of people are squeamish because rdfs is not a rec

[22:28:05] jang

em: thus I'd really like to get rdfs out the door asap

[22:28:28] jang

ph: it's startling that ean entire country waiting for us to make up our minds

[22:28:49] jang

em: there are now 6 countries that have mandated DC metadata in xml/rdf in all govt produced documents

[22:29:12] jang

so the frivolous question of pat's is actually very accurate

[22:29:29] jang

dan: people see us getting interested in AI/KR issues, theyr'e worried by this

[22:29:44] jang

ron: "are we supporting the austrailian DC standard?"

[22:29:50] jang

we get these issues all the time

[22:30:06] jang

em: yes, a lot of tese people are simply waiting on a REC

[22:30:52] jang

(now dajobe scribe)

[22:31:03] jang

arnot - adobe

[22:31:14] jang

... toolkit and specification now available

[22:31:21] jang

... invite anyone interested to join program

[22:31:45] jang

... c++ and source available under an open license, probably open source

[22:32:05] jang

bwm: to kwan

[22:32:12] jang

s/kwan/kwon/

[22:32:42] jang

... toolkits - redland, raptor by daveb, rdf api - sergey, jena - bwm

[22:32:52] jang

... cslisp - ora, kinda-perl - dan

[22:33:00] jang

... help available, please ask

[22:33:13] jang

jang is now known as dajobe

[22:33:39] dajobe

rond: demo

[22:33:56] dajobe

... presentation to time

[22:34:47] danbri

aside, danbri's perl rdf stuff: http://ilrt.org/discovery/2001/06/rdfperl/

[22:34:50] dajobe

... visual maps

[22:35:01] dajobe

... tim bray's antarti.ca

[22:35:27] dajobe

... demo contains interwoven-generated time info

[22:35:37] shellac

shellac has left channel

[22:35:50] II

II has joined #rdfcore

[22:36:02] dajobe

... stories visualised in map form

[22:36:17] dajobe

... number of articles is area

[22:36:27] dajobe

... 'importance' by how titles are visualised

[22:36:44] dajobe

.... separate view on content using SIC codes

[22:37:27] dajobe

... example of rdf 'stuff' sent off to a different company, made into a demo

[22:37:46] dajobe

em: can you make this public?

[22:37:48] dajobe

rond: have to see

[22:39:08] dajobe

dajobe: maps.net taken rdf from dmoz too

[22:39:17] dajobe

em: short turnaround, fantastic story

[22:39:41] dajobe

---

[22:40:13] dajobe

em: until I got tools that knew daml+oil, did it dawn to me what daml+oil was up to

[22:40:27] dajobe

... workflow for w3c was really interesting with model and merging, equilvanentTo

[22:40:29] II

II has left channel

[22:41:04] dajobe

phayes: simple stuff in daml+oil has biggest bang-for-buck

[22:41:18] dajobe

... which is what we find. Nobody much uses the advanced stuff

[22:42:05] dajobe

emiller giving w3 demo

[22:42:45] dajobe

wg chair visualising

[22:43:00] dajobe

... object of type 'chair', make it a square ...

[22:43:26] dajobe

.. chairs really don't know the unique ID of WG and don't care

[22:43:41] dajobe

... but know name and its email address etc

[22:44:07] dajobe

... tere is no unique ID for wg

[22:44:18] dajobe

... some may use homepage, email address or charter (danbri)

[22:44:23] dajobe

... and all of those are OK

[22:44:49] dajobe

... don't want to impose new requirements, but let them describe as they see them

[22:44:56] dajobe

... and ground in what they know

[22:45:19] dajobe

http://www.w3.org/Talks/2001/07/30-swws/slide36-1.html

[22:45:30] dajobe

(url typed by hand)

[22:46:04] dajobe

.. people waned to know announcements by activity eg.. everyting by XML activity

[22:46:04] dajobe

... and the WG chairs don't need to add this

[22:46:19] dajobe

... people who describe activity structure have different anmes than the chairs do for wgS

[22:46:27] dajobe

... they can make their descriptions in a different way

[22:46:48] dajobe

... so long as they agree on the id for the entity, they can merge (e.g. mail addr)

[22:47:08] dajobe

... so without the notion of the contact:mailbox as daml:equilalent we couldn't merge

[22:47:20] dajobe

... so need daml peroperties to do this

[22:47:38] dajobe

... interesting to see how processing this info wtih different levels of tools became a powerful thing

[22:47:55] dajobe

... and these things can be incremenetly layered. Int his case I needed damil:equiv

[22:48:00] dajobe

... but in other forms, I didn't

[22:48:07] dajobe

(slide 39-1.html)

[22:48:35] dajobe

... some get merged because of unique ids, some from daml:equivalent too

[22:49:07] dajobe

... we can do this by graph merging mostly and sometimes need daml

[22:49:21] dajobe

... incrementaly layering functionality

[22:49:30] dajobe

... greate experience to get hands on the tools for this

[22:49:43] dajobe

... and sometimes we realise we can weave into the workflow assigning unique ids for these

[22:50:00] dajobe

... lwo hanging fruit for daml is uniqueproperty, damlequivalent, ... (lost 3rd)

[22:50:13] dajobe

... very powerful

[22:50:34] dajobe

3rd was daml:unambiguous

[22:50:56] dajobe

--

[22:51:12] dajobe

ora: was mandated in daml program for all participants to use daml on their pages

[22:51:31] dajobe

... if you looked at the feature usage, most people just used rdf schema, very few daml bits

[22:51:52] dajobe

phayes: if you looka t daml+oil working at daml reseacher level

[22:51:53] dajobe

...

[22:52:06] dajobe

... they are running into limitations of daml+oil

[22:52:17] dajobe

... and hence has divergent pulls to simplicity, complexity

[22:52:26] dajobe

danbri: I've run into those concernts, more of a spectrum

[22:53:01] dajobe

... data merging is critical, before daml I had something monoproperty. daml properties don't license all the merging done in em's demo

[22:53:10] dajobe

monoproperty was 'same for all time'

[22:53:18] dajobe

em: rdf notion of layering

[22:53:29] dajobe

... daml may require more layers, but if done in this way, remains useful and power

[22:53:38] dajobe

danbri: sw-cg job is to get thesecharters layered

[22:53:48] dajobe

s/charters/components/

[22:54:10] dajobe

phayes: what this community needs is a combination of things from KR ...

[22:54:20] dajobe

... GOFK(???) ...

[22:54:30] dajobe

... some features that are pathetical easy from 1956 or something

[22:54:39] dajobe

... and some things so hard we put them off ...

[22:54:49] dajobe

... "full temporal sensitivity in changing worlds" ...

[22:54:58] dajobe

... exciting ...

[22:55:07] dajobe

... redirecting our attention to problems we put off

[22:55:21] dajobe

... and can't put off to the next millenium. Must do now, or yesterday.

[22:55:29] dajobe

bwm: pat is excited!

[22:55:34] dajobe

... wrap up

[22:55:38] dajobe

... thanks to everyone

[22:55:45] dajobe

thanks to brian

[22:55:47]

* dajobe claps

[22:55:54] dajobe

thanks scribes

[22:56:03] dajobe

more free gifts...

[22:56:11]

* AaronSw claps, grabs free gifts

[22:56:14] dajobe

XML schema f- the guide to w3c xml schema

[22:56:20] dajobe

MEETING CLOSED

[22:56:25] dajobe

DONM - months away :-)

[22:56:30] jimH-lurk

jimH-lurk has left channel

[22:56:50] dajobe

AaronSw: will try to grab one for you. mostly xml.com articles in a book

[22:56:54] GK-f2f

Yes, done!!!!

[22:57:04] AaronSw

thanks, dajobe

[22:57:10] AaronSw

Good work everyone!

[22:57:35] dajobe

logger here will be closing shortly... over to #rdfig

[22:57:39] GK-f2f

I think the acronym above was GOFAI

[22:57:54] GK-f2f

(Good Old Fashioned AI)

[22:58:00] AaronSw

GOFKR ;-)

[22:58:13] barstow_

barstow_ has left channel

Provided by Dave Beckett, Institute for Learning and Research Technology, University of Bristol