- 13:02
- *** Log started on 2000-10-10 UTC-0400 ***
- 13:02 [RRSAgent]
- *** RRSAgent (swick@seahorse.w3.org) has joined channel #topicmap-rdf
- 13:04 [RalphS]
- (log may be some minutes behind from some mirror sites)
- 13:04 [DanC]
- I'm getting "we're sorry, your access code is not recognized" using the code from Eric's message of
- 13:04 [DanC]
- Tue, 10 Oct 2000 08:53:39 -0400
- 13:05 [DanC]
- operator suggests changing 1st digit to 2
- 13:06 [DanC]
- Attendance: Connolly, Swick, Dave Beckett
- 13:06 [DanC]
- Attendance: Steve Newcomb present
- 13:07 [emiller]
- *** emiller (emiller@krash.dev.oclc.org) has joined channel #topicmap-rdf
- 13:07 [RalphS]
- *** RalphS (swick@24-6-128.wireless.lcs.mit.edu) has joined channel #topicmap-rdf
- 13:07 [RalphS]
- +Eric Miller
- 13:08 [DanC]
- SteveN claims to be in a channel with the same names, but we can't see him.
- 13:09 [DanC]
- expected: brickley
- 13:09 [DanC]
- expected: freese
- 13:10 [DanC]
- attendance: regrets in advance: EMiller probably needs to leave soon
- 13:10 [emiller]
- hi steve?
- 13:11 [RalphS]
- log is http://www.w3.org/2000/10/10-topicmap-rdf
- 13:11 [DanC]
- expected: michel
- 13:12 [RalphS]
- -> http://www.w3.org/2000/09/25-topicmap-rdf previous discussion
- 13:12 [DanC]
- agenda ideas, anyone?
- 13:13 [RalphS]
- +Dan Brickley
- 13:13 [DanC]
- old business: Connolly to invite danbri; DONE.
- 13:13 [RalphS]
- +Art Barstow
- 13:14 [DanC]
- [more old business, not reviewed]
- 13:14 [DanC]
- ===== upcoming topicmap stuff
- 13:14 [DanC]
- Steve: we've set a deadline of [?] so we can announce at XML 2000
- 13:15 [DanC]
- SN: I've been asked to chair
- 13:15 [DanC]
- SN: not sure I will, though
- 13:16 [DanC]
- old business: I think this is done: emiller: send announcement to www-rdf-interest
- 13:16 [DanC]
- old business: I think this ACTION is still pending: SteveF: grab some dublin core examples [and do something... not sure what.]
- 13:17 [DanC]
- XML 2000... is that this? [[[
- 13:17 [DanC]
- "Don't miss your chance to join us at XML DevCon
- 13:17 [DanC]
- Fall 2000, November 12-15, 2000, San Jose
- 13:17 [DanC]
- Doubletree Hotel, San Jose, CA. For more
- 13:17 [DanC]
- details, contact Dale Sherman at 212.251.0006
- 13:17 [DanC]
- ext. 12 or at dale@camelot-com.com. "
- 13:17 [DanC]
- "Our New York event sold out."
- 13:17 [DanC]
- "You'll be joined by 4,300+ XML attendees"
- 13:17 [DanC]
- http://www.xmldevcon2000.com/
- 13:17 [DanC]
- ]]]
- 13:18 [emiller]
- http://www.gca.org/attend/2000_conferences/XML_2000/default.htm xml 2000 conference?
- 13:18 [RalphS]
- +Michel
- 13:18 [danbri]
- *** danbri (danbri@w3cdhcp8.w3.org) has joined channel #topicmap-rdf
- 13:20 [DanC]
- ah... emiller is right; Washington DC [date?]
- 13:20 [DanC]
- I note that the word "washington" doesn't occur on ....XML_2000/default.htm
- 13:20 [DanC]
- ah... it does, but as graphics, not text.
- 13:20 [emiller]
- xml 2000: Dec 3-8 Washington DC
- 13:23 [DanC]
- grumble... GCA conference page isn't WAI-happy; no alt text for the most important info on the page: date, city
- 13:24 [RalphS]
- +Eric Freese
- 13:24 [DanC]
- connolly claims keywords/description are widely deployed; emiller disputes this claim.
- 13:25 [danbri]
- emiller accepts action to circulate evidence pro/con this claim to some public archived list.
- 13:25 [danbri]
- emiller?
- 13:25 [DanC]
- old business:
- 13:25 [DanC]
- [12:58] <DanC> ACTION SteveF: grab some dublin core examples [and do something... not sure what.]
- 13:26 [DanC]
- SteveF: ... and see if they could be mapped into topicmap-ese
- 13:26 [DanC]
- SteveF: no progress
- 13:26 [ArtB]
- *** ArtB (barstow@24-6-195.wireless.lcs.mit.edu) has joined channel #topicmap-rdf
- 13:26 [DanC]
- Michel: agenda request: general RDF/topicmap model issues
- 13:27 [DanC]
- ====== general RDF/topicmap model issues
- 13:27 [Anonymous]
- *** Anonymous (Anonymous@dns.datachannel.com) has joined channel #topicmap-rdf
- 13:27 [Anonymous]
- # Appears as XENO.
- 13:27 [DanC]
- Michele: (1) What do we call metadata? seems that for topicmap, there is no distinction. it seems that in RDF there is.
- 13:28 [danbri]
- [[[The distinction
- 13:28 [danbri]
- between "data" and "metadata" is not an absolute one; it is a distinction created primarily by a particular application, and many
- 13:28 [danbri]
- times the same resource will be interpreted in both ways simultaneously.
- 13:28 [DanC]
- Swick: we didn't mean to draw any sharp distinction between metadata and data in the RDF specs. Did we?
- 13:28 [danbri]
- ]], from rdf model and syntax intro
- 13:28 [RalphS]
- Ralph: when RDF refers to 'metadata', we mean it in the sense that metadata applications are one use of RDF
- 13:29 [DanC]
- Michel: so (1) is not much of an issue.
- 13:29 [DanC]
- Michel: (2) topicmaps are described as objects, where we describe the data but not the methods. so... nothing about processing.
- 13:31 [DanC]
- connolly: we expect RDF to be processed, but we don't mean to specify any particular processing. This is the same as topicmaps, no?
- 13:31 [DanC]
- Michel: ok... so (2) doesn't seem to be an issue either.
- 13:31 [danbri]
- more context for (2), from rdf schema section 3: [[For example, while an RDF schema can assert that an
- 13:31 [danbri]
- author property is used to indicate resources that are members of the class Person, it does not say whether
- 13:31 [danbri]
- or how an application should act in processing that class information. We expect that different applications will
- 13:31 [danbri]
- use these constraints in different ways
- 13:31 [danbri]
- ]]
- 13:31 [DanC]
- Michel: (3) RDF schema is specified as relationships between resources; topicmaps have another layer...
- 13:32 [danbri]
- src for that quote, http://www.w3.org/TR/2000/CR-rdf-schema-20000327/#constraints
- 13:32 [DanC]
- -emiller is excused.
- 13:34 [danbri]
- [discussion of whether 'arcs' and/or 'topics' are --in some sense-- neutral]
- 13:34 [RalphS]
- Steve: "a topic is the thing that gives form to a platonic idea"
- 13:35 [RalphS]
- Michel: can't write a relationship between 2 documents...
- 13:35 [RalphS]
- ... must say "document A is represented by topic 1, document B is represented by topic 2, topics 1 and 2 are related."
- 13:36 [RalphS]
- Michel: e.g. in the RDF examples, "author" would be a topic
- 13:36 [RalphS]
- ... "subject" would be in the occurrence role and the role would be "description"
- 13:38 [RalphS]
- DanC: earlier it was said that every topicmap could be written as a directed labeled graph...
- 13:38 [RalphS]
- ... is the reverse true; every DLG can be interpreted as a topicmap?
- 13:38 [RalphS]
- Steve: yes
- 13:42 [RalphS]
- DanC: the smallest RDF model is empty, but the next-smallest has 2 nodes and 1 arc. What is the smallest topicmap?
- 13:42 [RalphS]
- Eric: 1 topic
- 13:44 [EFreese]
- *** EFreese (EFreese@dns.datachannel.com) has joined channel #topicmap-rdf
- 13:44 [EFreese]
- # Appears as XENO.
- 13:45 [RalphS]
- DanC: smallest is 1 topic with a label?
- 13:49 [RalphS]
- DanC: from RDF's point of view, for example, every GIF image on the Web is a resource
- 13:49 [RalphS]
- Michel, Steve: topicmap wants these to be deliberately stated as 'stand-ins' for topics
- 13:49 [RalphS]
- ... may be that there is an extra layer in topicmap that is "notion identification machinery"
- 13:51 [RalphS]
- DanC: in general, [we] think of a URI as playing the role of a noun-phrase in natural language
- 13:53 [RalphS]
- Michel: "degenerate topics" are topics without names [and without semantics]
- 13:54 [RalphS]
- Steve: suppose we took the position that association links, if they don't point to topics...
- 13:55 [RalphS]
- ... are regarded as pointing to the "subject identity" of a topic
- 13:55 [RalphS]
- ... where "subject identity" is some hunk of information that is regarded by the author of the topicmap as the best possible unambiguous expression of what notion he is thinking about
- 13:55 [RalphS]
- ... "subject identities" are the binding points in topicmap land
- 13:56 [RalphS]
- Steve: if multiple topic links point at the same resource saying it is their subject identity, then all those topics are regarded as being one and the same
- 13:56 [RalphS]
- Steve: could relax the [linking] constraint such that everything pointed to is either a topic or is a subject identity
- 13:57 [RalphS]
- DanC: sounds right
- 13:57 [RalphS]
- Michel: could get confusing to implementations to have such choices between two interpretations
- 13:57 [DanC]
- hmm... while that solves some things, it may introduce items from the RDF issues list: "are we pointin got the XML element, or the RDF resource it denotes?"
- 13:58 [RalphS]
- Michel: I understand "topics" as the connecting hub
- 13:59 [RalphS]
- Steve: let's say that an association link is the equivalent of an RDF arc...
- 13:59 [RalphS]
- ... if we say that association links are not required to point at topics,
- 13:59 [DanC]
- something I'm noticing... topicmaps seem, like XLink, to be n-ary at heart, whilc RDF is binary at heart. RDF's binariness is kinda painful, to me, lately, so I have an open mind here.
- 14:00 [RalphS]
- Steve: ... but if they do point at topics they work just as before,
- 14:00 [RalphS]
- ... however if they don't point at topics then consider what they point at as being subject identities
- 14:00 [EFreese]
- TM are n-ary
- 14:00 [RalphS]
- Michel: doesn't seem implementable
- 14:00 [danbri]
- danc: does the DAML work address the binary/n-ary concern? eg. common strategy for doing n-ary over edge-labelled graphs...?
- 14:00 [danbri]
- (um... that was a question to danc not an attribution!)
- 14:01 [DanC]
- I've run into the binary/n-ary issue in DAML work. I came up with an approach (see restrictedBy etc.). I'd hardly call that a "common strategy" but it's there.
- 14:01 [RalphS]
- [Ralph worries about two meetings here!]
- 14:04 [RalphS]
- Steve: [is it the case that] if you point to a resource that isn't a topic, then that resource doesn't have any occurrences, right?
- 14:08 [RalphS]
- DanC: in RDF identifiers are only necessary to be sure you are talking about the same thing in two different documents
- 14:08 [RalphS]
- ... different URIs in different documents *might* be talking about the same thing but we need more smarts to deduce that
- 14:08 [RalphS]
- ... the same URI in two different documents is always talking about the same thing
- 14:09 [RalphS]
- DanC: RDF syntax does have a way to talk about things without having a URI for them
- 14:11 [DanC]
- ACTION SteveN: propose to relax that contraint in windam(sp?) and report back. (Michel's reservation noted)
- 14:11 [DanC]
- ... back to Michel's list of model differences.
- 14:12 [RalphS]
- Michel: notion of something like RDF Schema ...
- 14:12 [DanC]
- Michel: (4) about RDF schemas... in some ways there's an equivalent, but in some ways there is...
- 14:12 [RalphS]
- ... no equivalent idea in topicmap
- 14:13 [RalphS]
- Michel: can define some [similar things] in DTD. see definitions of "housekeeping topics"
- 14:14 [RalphS]
- Michel: topicmap has "topic type", "topic name", ??
- 14:14 [RalphS]
- ... value of a link is unconstrained
- 14:16 [ArtB]
- [I couldn't understand the third item in Michel's list but I thought he said "inference rule" ...]
- 14:18 [RalphS]
- EricF: "public subjects" are something like the shared concepts that are defined in an RDF schema
- 14:20 [RalphS]
- Steve: consider "scope"
- 14:21 [RalphS]
- ... when we assert an arc; when we say that two things have a relationship
- 14:21 [RalphS]
- ... the scope can be "unconstrained"; there is no situation we know of in which the arc is not valid
- 14:21 [RalphS]
- DanC: e.g. "the successor of 2 is 3"
- 14:21 [RalphS]
- Steve: right, though this might be an expression of someone's opinion
- 14:22 [RalphS]
- ... however, could be narrow scopes
- 14:22 [RalphS]
- ... e.g. "sensible teeth" vs "sensitive teeth". In the French language, only the context distinguishes which meaning of "sensible" is meant.
- 14:24 [RalphS]
- Michel: a topic can have many names
- 14:24 [danbri]
- [aside: (real world) eg. http://xmlns.com/wordnet/1.6/Tooth is one notion of tooth that has a unique web name]
- 14:24 [RalphS]
- ... but we know when we are talking about the same topic
- 14:24 [RalphS]
- Michel: to know what thing the name applies to, we filter it using the scope
- 14:25 [RalphS]
- ... when we give a name for an association or for a role, we really give a set of names that are represented by a topic
- 14:25 [RalphS]
- Steve: this is separate from addressing
- 14:26 [RalphS]
- Steve: topics have names, which are not their addresses, and they have occurrences
- 14:26 [RalphS]
- ... topics play roles with respect to other topics; these are their 3 characteristics
- 14:26 [RalphS]
- ... all 3 characteristics have scope
- 14:27 [RalphS]
- ... scope is expressed as a set of topics (not topic names)
- 14:28 [RalphS]
- DanC: consider a typical RDF sort of example; "Bill Clinton" and "United States" are related by "is President of"
- 14:28 [RalphS]
- ... this will not always be true, but we want to represent the fact that it was true in 1999
- 14:29 [danbri]
- more, er, context for the record. Graham Klyne posted this proposal for representing contexts to the RDF Interest Group: http://lists.w3.org/Archives/Public/www-rdf-interest/2000Sep/0028.html
- 14:31 [RalphS]
- DanC: RDF uses reificiation to represent statements then any statement can have additional arcs. There is no restriction to the arcs that can be added to a statement.
- 14:32 [RalphS]
- Michel: this representational idea might come out of the "conceptual model" subgroup
- 14:32 [DanC]
- my point is that there's a relationship topicmap:scoped-by(?association, ?set-of-topics), and I wonder if topicmap:scoped-by can have subproperties, and whether topicmaps have a slot for them.
- 14:36 [RalphS]
- DanC: different example; "red book is good". Bob believes this statement, Fred refutes it.
- 14:37 [RalphS]
- ... in topicmap I expect I can put Bob and Fred into the context of the statement but I don't know how to say Bob believes it and Fred doesn't.
- 14:38 [RalphS]
- DanC: I want to structure this such that it is clear that Bob and Fred disagree.
- 14:42 [RalphS]
- DanBri: can you ascribe probabilities to associations in topicmap?
- 14:42 [RalphS]
- ... for example, that Bob gives probability 90% to the statement "book is red"
- 14:42 [RalphS]
- ... does this make sense in topicmaps?
- 14:42 [RalphS]
- Steve: people make new topics, e.g. for chronology
- 14:43 [RalphS]
- ... could do the same with probability levels
- 14:46 [RalphS]
- Steve: wrt "topicmap only permits positive statements" ...
- 14:46 [RalphS]
- ... if you are interested in the fact that an association is absent from a list, this is a 'query'
- 14:47 [RalphS]
- ... so if you want to know 'Joe does not believe', you look for the absence in a query
- 14:47 [DanC]
- I think that if I look at topicmap:scoped-by as a subproperty of rdf:type, I can be happy.
- 14:48 [DanC]
- i.e. "scoped by" is a sort of one-place predicate mechanism in topicmaps, and we use rdf:type for one-place predicates.
- 14:50 [danbri]
- ralph: "is there an explicit notion of unknown? can I distinguish between not knowing and false?"
- 14:53 [RalphS]
- Steve: not in general, you need to know something about the semantics of the specific association
- 14:54 [RalphS]
- ... in order to know how you might differentiate between "unkown" and explicitly absent
- 14:57 [danbri]
- more context: interesting work bridging the two worlds... http://www.cse.ogi.edu/~shawn/rpe/bowersrpe.pdf
- 15:03 [DanC]
- michel's message http://lists.w3.org/Archives/Public/www-rdf-interest/2000Sep/0194.html
- 15:05 [DanC]
- Michel: we've been able to exploit XLink implementations to deal with one topicmap dataset
- 15:09 [DanC]
- proposed: Mon, 23 Oct [what time?]
- 15:09 [DanC]
- 1500Z (11amET)
- 15:09 [DanC]
- ok by me.
- 15:09 [DanC]
- RESOLVED.
- 15:09 [ArtB]
- *** ArtB (barstow@24-6-195.wireless.lcs.mit.edu) has left channel #topicmap-rdf
- 15:10 [DanC]
- ADJOURN.
- 15:10
- *** Log ended on 2000-10-10 ***