IRC log from 2000-09-25

Session Start: Mon Sep 25 10:58:53 2000
[10:58] *** Now talking in #topicmap-rdf
[10:59] *** emiller_ has joined #topicmap-rdf
[10:59] *** emiller_ has left #topicmap-rdf
[11:00] <emiller> hi steve :)
[11:01] <srn> hello
[11:01] <RalphS> present: Eric Freese, Michel Biezunski, Ralph Swick, Steve Newcomb
[11:02] <RalphS> [I intend to put this log publicly accessible at http://www.w3.org/2000/09/25-topicmap-rdf ]
[11:02] *** EricFrees has joined #topicmap-rdf
[11:02] <EricFrees> # Appears as XENO.
[11:03] *** DanC_desk has joined #topicmap-rdf
[11:03] *** EricFrees is now known as EricF
[11:03] *** DanC_desk is now known as DanC
[11:03] <RalphS> previous call: http://lists.w3.org/Archives/Public/www-rdf-interest/2000Sep/0191.html
[11:04] <RalphS> DanC, you intend to be present only via irc?
[11:04] <DanC> er... as of now, yes; sorry... I double-booked myself.
[11:06] <RalphS> 11:58:14 <DanC> ACTION Connolly: distribute pointer to record to www-rdf-interest and xtm-wg
[11:06] <RalphS> done
[11:06] <RalphS> ACTION Michel: send pointer to [???]
[11:07] <DanC> I think Michel did that, in reply to my msg (0191)
[11:07] <RalphS> done: http://lists.w3.org/Archives/Public/www-rdf-interest/2000Sep/0194.html
[11:11] <RalphS> Steve: did folk get a chance to read [the bounded object set stuff]
[11:11] <RalphS> -> Bounded object sets
[11:11] <RalphS> Ralph: oops, no -- sorry
[11:12] <RalphS> Eric: briefly read, seems to refer to some application interpretation
[11:14] <RalphS> Steve: since we're [also] talking about mixing namespaces, it makes sense to me that
        an application will want to know, e.g., about any XLinks and/or RDF statements in other documents.
[11:15] <RalphS> Eric: I prototyped this by extending the RDF seeAlso relation
[11:17] <emiller> <rdf:Property rdf:ID="boundedResource"><rdfs:subPropertyOf rdf:resource="&rdfs;seeAlso" />
[11:17] <emiller> </rdf:Property>
[11:18] *** Michel has joined #topicmap-rdf
[11:18] <RalphS> welcome, Michel
[11:18] <emiller> howdy!
[11:19] <srn> [insert standard greeting here]
[11:19] <Michel> Hello
[11:21] <RalphS> -> RDF seeAlso relationship
[11:22] <RalphS> Eric: BOS seems to be a specialization of rdfs:seeALso
[11:22] <RalphS> Michel: the list of resources inside the bounded object set is like having specialized occurrences
[11:23] <RalphS> ... should this be kept outside of the general topicmap or rdf description or should it be a
        specialized association?
[11:23] <RalphS> Steve: I found I needed to understand how this particular thing worked to understand RDF
[11:24] <RalphS> ... but BOS doesn't need to be on the front-burner; [not necessarily the most important]
[11:24] <emiller> Steve: Interested in knowing what RDF types want to gain from colaboration
[11:26] <emiller> RalphS: RDF community spent a lot of time discussing the model.  Independant of syntax.
[11:27] <emiller> RalphS: seems that RDF and Topicmap may being useing the same model (directed label graph concepts)
[11:27] <emiller> RalpS: Hope is that topicmap group will have concensus about XML/SGML community about a seriealization
        of the directed label graph
[11:28] <emiller> RalphS: Not certain that we won't find differences between RDF and TM
[11:28] <emiller> RalphS: specifically wrt to model (special case vs general case)
[11:30] * DanC is interested to see any topicmap syntax examples
[11:30] <emiller> for the record... eric can;t spell
[11:30] <RalphS> specifically the expectation for the distribution of RDF statements across documents may be different
        from the expectations of the XML topicmap community, but that detail may be revealed later.
[11:31] <RalphS> Steve: [to Michel] perhaps RDF is a template?
[11:31] <RalphS> ... a set of relationship types
[11:32] <RalphS> Michel: if RDF defines hierarchical relationships, that is easily represented in topicmap
[11:32] <RalphS> ... but maybe something more fundamental going on, having to do with the fact that RDF connects nodes with arcs
[11:32] <RalphS> ... this may represent the underlying model of topicmaps
[11:32] <emiller> http://www.isogen.com/demos/RDF/rdfarch.html (not found)
[11:33] <emiller> [August 12, 1997] A response to the incipient Resource Description Framework (RDF) proposal, in terms of the
[11:33] <emiller>       AFDR architecture mechanism (HyTime Second Edition): "Using the RDF Data Model as an SGML Architecture,"
[11:33] <emiller>       by W. Eliot Kimber. The document summary: "Describes the author's attempt to use the Resource Description
[11:33] <emiller>       Framework (RDF) data model as an SGML document architecture conforming to the Architectural Forms
[11:33] <emiller>       Declaration Requirments annex of ISO/IEC 10744:1997 (HyTime Second Edition). Includes an explanation of RDF
[11:33] <emiller>       base architecture as well as examples of using that architecture derived from the examples used in the RDF working
[11:33] <emiller>       draft."
[11:34] <RalphS> Steve: that reference is to Architectural Forms
[11:34] <RalphS> Steve: the AF issue is not so much about modelling, mostly about syntax
[11:36] <RalphS> Steve: re directed graphs versus hyperlinks...
[11:36] <RalphS> ... we learned an important lesson: people don't "get" nodes and arcs
[11:36] <RalphS> ... when we did a nodes-and-arcs syntax for topicmaps it was a non-starter
[11:36] <RalphS> ... this may be one of the issues RDF is experiencing
[11:38] <RalphS> Michel: for computers to process efficiently, nodes and arcs are silly
[11:38] <RalphS> ... for representing human knowledge, we have tried to group fundamental nodes and arcs together within an appropriate construct
[11:39] <RalphS> Steve: the result of processing some set of topicmap documents is a single directed graph
[11:39] <RalphS> ... but the problem is trying to make the syntax represent this directly
[11:40] <RalphS> Eric: as someone who has been trying to convey nodes and arcs for some time, the people who do 'get it'
        get it immediately.  Others seem to never get it.
[11:40] <RalphS> ... have to convey [nodes and arcs] in a context that people are familiar with
[11:41] <RalphS> ... SGML community seems to have the hardest time, since they've become so focussed on syntax [to the exclusion of modelling]
[11:44] <emiller> RalphS: Discussing reifincation
[11:44] <emiller> RalphS: Is there a similar concept in Topicmaps
[11:45] <RalphS> [is there a concept in topicmap like reification; making a statement into a first-class object?]
[11:45] <emiller> EricF: based on reading od ISO doc there isn't
[11:45] <RalphS> EricF: I [and ???] have been pushing for giving IDs to associations
[11:45] <RalphS> ... there is a proposal before the ISO committee to do thi
[11:46] <RalphS> Steve: the ISO standard makes first class objects out of association _types_
[11:46] <RalphS> ... everything that is a first class object _is_ a topic (or uses a topic as a proxy)
[11:47] <RalphS> ... the proposal before ISO is that a topic can be a proxy for an association
[11:47] <RalphS> Michel: associations can have IDs; this is a common attribute that can be placed anywhere
[11:48] <RalphS> Steve: the debate at the ISO WG meeting was that associations could not participate in associations
[11:48] <RalphS> ... thus the need for a proxy topic
[11:49] <RalphS> ... a proxy topic is "just a topic"; one whose subject identity is an association
[11:49] <RalphS> Steve: associations express n-ary relationships between topics
[11:51] <RalphS> Michel: I prefer to represent things with the least possible number of constructs
[11:53] <RalphS> EricM: I am hoping for as seamless an integration as possible of the two sets of building blocks
[11:55] <RalphS> ... realize that we have different names for very similar concepts
[11:56] <RalphS> Steve: I'd like to characterize RDF nodes-and-arcs as putting the arcs at the center
[11:56] <RalphS> ... topicmaps puts the nodes at the center; more people seem to get this
[11:59] <RalphS> Michel: about the notion of topic itself...
[11:59] <RalphS> ... topics are like hubs; they concentrate many connections
[12:00] <RalphS> ... we have a constraint that forbids a link between any two nodes unless a topic first exists
[12:00] <RalphS> Ralph: in RDF you can't write an arc without writing its name
[12:02] <RalphS> Steve: it feels to me right now that the reification of a statement is a topic
[12:03] <RalphS> ... e.g. an index term in a document index is an expression of the concept that links two pages together
[12:05] <RalphS> Michel: who are we targetting? developers or information holders?
[12:05] <RalphS> Ralph: information holders -- many more of them than there are developers, I hope.
[12:12] <RalphS> EricM: next steps?
[12:13] <RalphS> EricF: the XTM modelling group speaks "directed labeled graphs" quite a bit
[12:13] <RalphS> ... the nodes have unique ids, whether the association has a unique id is up for discussion
[12:14] <RalphS> Steve: the topic has an identity; two occurrences and the association between them
[12:14] <RalphS> EricF: the XTM Use Case group has RDF on its agenda to look at
[12:14] <RalphS> ... the Modelling group hasn't decided whether or not RDF will be a primary focus for November
[12:15] <RalphS> Michel: the topicmap community has projects on query languages, inference rule descriptions, constraint of topicmap structures, etc.
[12:16] <RalphS> ... these are in the background; we're concentrating first on the core syntax
[12:16] <RalphS> ... RDF seems to have some similar projects
[12:16] <RalphS> ... beneficial to exchange points of view on these other areas
[12:18] <RalphS> Michel: I'm reluctant to say that automatic processing should heavily use inference rules
[12:19] <RalphS> Steve: I have a feeling that there is nothing in the current topicmap discussions that could not be expressed in RDF
[12:19] <RalphS> EricF: agree
[12:20] <RalphS> EricF: I am experimenting with defining "public topics" for the RDF built-in concepts of Bag, etc.
[12:21] <RalphS> ... given data expressed in RDF or in topicmap we should be able to extract the same graph
[12:21] <RalphS> Steve: it may be an effort to produce a deterministic transformation
[12:22] <RalphS> EricM: then why do both?
[12:22] <RalphS> EricF: document versus data point of view
[12:24] <DanC> ok... my other call is done.
[12:24] <RalphS> EricM: separate sets of tools would be a waste
[12:24] <DanC> are we ~30min into this, or ~90min?
[12:24] <RalphS> [DanC: 90 mins done]
[12:25] <RalphS> EricM: my impression is that ISO topicmaps are finished and that XTM is trying to figure out how to write them on
        the Web.  My suggestion is that RDF be the way this is done.
[12:25] <RalphS> +DanC on call
[12:25] * DanC joins, with apologies
[12:27] <RalphS> Steve: all data is subject to a variety of transformations.  What is the 'true' lingua franca
        -- what is it that is really the same between all applications?
[12:27] <RalphS> EricM: the RDF data model
[12:28] <RalphS> ... a better serialization [coming from] XTM would be an improvement
[12:30] <RalphS> EricF: the communities need to get down enough into the details to convince themselves that there are no differences in the models -- or that those differences are unimportant
[12:31] <RalphS> Steve: we're facing a self-imposed deadline of December for an XTM 1.0 encoding.  It's unlikely we'll have an RDF data model by then.
[12:31] <RalphS> EricF: because of the discussions from Extreme, people will ask about the status
[12:32] <RalphS> ... XTM working group meeting is scheduled for October
[12:33] <RalphS> ... if we identify some differences that can be brought to the modelling group by then, ...
[12:33] <RalphS> Steve: October 12-14, Swindon
[12:36] <DanC> I propose: if the XTM WG will (a) adopt a data model that is essentially the same as the RDF data model, and (b) provide a serialization that is interchangeable
        with the RDF serialization (via XSLT, say), then the RDF IG will investigate W3C endorsement of this other serialization
[12:38] <DanC> retry: if the XTM WG will (a) propose a data model that is essentially the same as the RDF data model, and (b) propose a serialization that is interchangeable
        with the RDF serialization (via XSLT, say), then the RDF IG will investigate W3C endorsement of this other serialization
[12:41] <RalphS> Steve: it is not necessarily true that the way to serialize a data model is also the best API to that model.
[12:43] <RalphS> DanC: XSLT is turing-complete, so any transformation should be writable in it
[12:43] <RalphS> Steve: it could be monstrously inconvenient to do it in XSLT
[12:44] <RalphS> Ralph: what are the canonical representations you expect for typical topicmap applications?
[12:45] <RalphS> Steve: if two expressions are the same they are merged
[12:45] <RalphS> ... not trivial to do in XSLT
[12:45] <RalphS> DanC: it is straightforward in rdf to have anonymous things
[12:45] <RalphS> ... and to later discover that these two anonymous things are the same instance
[12:46] <RalphS> Steve: even if they are in different documents?
[12:46] <RalphS> Eric, Dan: yes
[12:48] <RalphS> EricM: syntax transform provides a transition path for current RDF apps
[12:51] <DanC> Steve: for next meeting, agenda proposal: hash thru exactly what we're going to say to our constiuencies.
[12:53] <DanC> proposed: 1-3pEST 10 Oct
[12:53] <DanC> Tuesday
[12:54] <srn> OK
[12:54] <RalphS> swick agrees
[12:54] <emiller> eric yes
[12:54] <DanC> dan ok
[12:54] <EricF> 12-2 CST - agreed
[12:54] <DanC> michel OK
[12:54] <RalphS> == 1700-1900 UTC
[12:54] <DanC> ACTION Connolly to invite danbri
[12:55] <DanC> re verbiage: start with DanC's proposal above
[12:56] <DanC> NB: DC 8 is next week (8th dublin core workshop; URI?)
[12:56] <DanC> s/NB/N.B./ i.e. Note Well
[12:56] <RalphS> Michel: Oct 3 and 4 in Chicago I'm teaching a topicmap tutorial
[12:56] <DanC> http://www.infoloom.com/workshops.htm
[12:56] <emiller> DC8: (http://www.ifla.org/udt/dc8/call.htm)
[12:58] <DanC> ACTION SteveF: grab some dublin core examples [and do something... not sure what.]
[12:59] <DanC> Michel: feel free to send questions by email [about the Paris topicmap example, I assume]
[13:01] <DanC> ACTION Michele: send pointers to topicmap serialization examples
[13:02] <DanC> ralph to publish this log
[13:02] <DanC> ACTION emiller: send announcement to www-rdf-interest
[13:02] <DanC> (and to xtm-wg, I suggest/assume)
[13:03] <RalphS> I will close and publish the log as soon as Michel adds the URIs
[13:03] *** EricF has left #topicmap-rdf
[13:03] <Michel> http://www.infoloom.com/tmsample/xmle99mp.htm
[13:03] <Michel> This is the URL for the serialized topic map.
[13:03] <Michel> Bye
[13:04] <RalphS> bye, thanks Michel
[13:04] <DanC> ah! thanks, Michel