Semantic Web Interest Group IRC Chat Logs for 2002-10-08

00:11:53 <kmacleod> is there a "best practices" for reading/parsing RDF/XML instances?

00:13:03 <AaronSw> how so? what's wrong with following the spec?

00:14:09 <kmacleod> as in, how to interpret the spec into running code

00:14:20 <kmacleod> techniques, gotchas

00:14:41 <AaronSw> jjc has several useful papers on it, i think dajobe has some notes

00:15:25 <AaronSw> http://www-uk.hpl.hp.com/people/jjc/publications.html

00:16:28 <AaronSw> http://www.redland.opensource.ac.uk/notes/index.html

00:17:05 <tav> tav is now known as tav|sleep

01:26:16 <Skyline_> Skyline_ is now known as Skyline

02:03:04 <kmacleod> . http://bitsko.slc.ut.us/~ken/rss-sxs/RDF::Simple::XMLParser.pm

02:03:23 <kmacleod> can read RSS standalone (with XML::Parser, natch ;)

02:03:47 <kmacleod> but as generic RDF. now working on "real" RDF robustness

02:04:32 <kmacleod> nodes are hashref, keys are predicates, and hash values are arrays of objects

02:05:17 <kmacleod> example usage: http://bitsko.slc.ut.us/~ken/rss-sxs/rsstest.pl

12:55:56 <lilo> [Global Notice] Hi all. We'll need to rehub a good section of Europe. Please bear with us.

14:00:44 <dajobe>http://xfml.org/

14:00:46 <dc_rdfig> A: http://xfml.org/ from dajobe

14:00:54 <dajobe> A:|eXchangable Faceted Metadata Language (XFML)

14:00:55 <dc_rdfig> Titled item A.

14:01:01 <dajobe> A:"an XML language for sharing hierarchical faceted metadata and indexing efforts."

14:01:03 <dc_rdfig> Added comment A1.

14:01:17 <dajobe> A:seems to be related to Topic Maps (or XTM)

14:01:19 <dc_rdfig> Added comment A2.

14:01:52 <larsbot> A:made by same guy who runs [http://easytopicmaps.com|easytopicmaps]

14:01:53 <dc_rdfig> Added comment A3.

14:02:11 <larsbot> A:definitely XTM-inspired, and latest version uses URIs to define identity...

14:02:11 <dc_rdfig> Added comment A4.

14:02:37 <dajobe> A:[http://purl.oclc.org/NET/xfml/core/|XFML Core 1.0] out 2002-10-08

14:02:37 <dc_rdfig> Added comment A5.

14:04:08 <dajobe> A:looks like aiming for an "RSS 1.0" sweet-spot of simplifying things

14:04:09 <dc_rdfig> Added comment A6.

14:04:27 <dajobe> A6:looks like aiming for an "RSS 1.0" sweet-spot of simplifying things for a particular purpose

14:04:27 <dc_rdfig> Replaced comment A6.

14:11:31 <dajobe> yawn, zzz say something I don't know dan

14:16:30 <danb_lap> I mean, please don't bitch about other RDF Interest Group members in this channel and expect the [off] prefix to make it ok. Thanks.

14:17:44 <dajobe> seth has a lovely blog

14:17:55 * danb_lap grins; I heard that too

14:17:55 <dajobe> you can read it here http://radio.weblogs.com/0113759/

14:18:05 <danb_lap> wonderful, thanks ever-so

14:18:14 <danb_lap> don't forget to chump it so others can share your joy

14:18:15 <dajobe> near danny ayers' http://www.citnames.com/blog/

14:18:57 <dajobe> who worked on A:

14:20:51 <dajobe> ann I said was <dajobe> [off] eek, seth has a blog

14:21:02 <dajobe> and that was not any kind of bitch

14:21:07 <dajobe> so don't be so prissy danbri

14:22:22 * danb_lap re-reads the eek more generously, remains confused by the [off]. Whatever.

15:00:11 <q|away> q|away is now known as qmacro

15:22:50 <mhgrove> mhgrove is now known as FujitsuRocks

15:41:00 <FujitsuRocks> FujitsuRocks is now known as Mike

16:02:59 <danja> A: facets are mutually exclusive, so something like disjointWith needed for RDFS version

16:02:59 <dc_rdfig> Added comment A7.

17:11:11 <tav|sleep> tav|sleep is now known as tav

17:15:26 <dmiles_and_mea_c> dmiles_and_mea_c is now known as dmiles_laptop

18:23:39 <SethR> yo, danja, what do you think about twisted http://twistedmatrix.com/ ?

18:26:16 <danja> hi Seth

18:27:01 <danja> hey - that looks nice

18:27:41 <danja> add an RDF parser/generator and you're away!

18:29:26 <danja> nice going - http + irc + mail servers + in 0.5MB

18:29:28 <SethR> danja, i got that one already

18:30:00 <SethR> i was thinking of using this to base ~publish~, ~read~ buttons on

18:30:20 <danja> yeah - it'd save a lot of time

18:30:29 <danja> where's the license..?

18:31:44 <SethR> don't know but they say its open source .. there is a #twisted channel here

18:33:14 <danja> busy channel

18:33:38 <SethR> that's good, huh?

18:33:53 <danja> active I 'spose

18:36:18 <danja> heh - they're Extremers

18:36:33 <danja> is Sailor event based?

18:37:01 <SethR> i would say so, but don't know .. where would i look?

18:37:28 <danja> having 'publish' & 'read' buttons would suggest it...

18:37:45 <danja> "if your application is already event based, then integrating Twisted may be only the work of one afternoon"

18:42:42 <SethR> could ideagraph base its i/o on the same classes?

18:43:19 <danja> I guess so, through Jython

18:44:32 <danja> may be a bit inefficient, indirect

18:44:49 <SethR> i think i have sufficient motivation now to download and see if i can get it working

18:45:37 <danja> go for it - I'll have a nosey too

18:46:05 <JibberJim> you're tracing out people...

18:46:15 <SethR> im sure looking forward to the new ideagraph

18:46:43 <danja> JJ - nearly done ;-)

18:47:17 <SethR> will it have spinoff?

18:47:17 <danja> Seth - making quick progress

18:47:23 <danja> yes!

18:47:36 <SethR> :))))))

18:47:46 <danja> the selection stuff is going right in at a low level

18:48:13 <SethR> so you think it is ready to compete with CMap ?

18:48:22 <danja> heh - not yet

18:49:53 <danja> but I'm doing it 'properly', so it soon will be CMap-equiz

18:49:58 <danja> equivalent

18:50:48 <danja> topic maps

18:50:51 <danja> xtm

18:52:06 <SethR> topic maps are great, but there so similary to RDF, why not just use RDF ?

18:54:16 <larsbot> SethR: because many things are easier with topic maps than with RDF

18:56:18 <SethR> easier, smeasier ... if i have to learn a new language, how is that easier for me ... if they are compatable why not just use one to think inside ... i can always relate to TMs as input\output grammars

18:56:29 <danja> larsbot - the TM -> RDFS chumped yesterday(?), it mentions limitations

18:56:37 <danja> what be they?

19:03:58 <larsbot> sorry, lost connection for a while

19:04:42 <larsbot> danja: you mean the "all locators must be URIs" limitation?

19:05:05 <larsbot> or the "perhaps less flexible than one might wish" one?

19:05:50 <larsbot> SethR: this a bit like the relationship java/jython

19:06:07 <larsbot> they are good at different things

19:06:39 <larsbot> to take one example, I'm pretty sure the foaf schema would have been better than it is had it been a topic map application

19:07:45 <larsbot> the difference lies in the way you think about your information, and that difference is significant

19:10:17 <SethR> lars, i think of information in labeled directed graphs. How is the difference significant to me?

19:10:20 <danja> larsbot - sorry, distracted - the "perhaps less flexible" was what I had in mind

19:10:56 <larsbot> danja: it's just that this makes it terribly awkward to work with topic map information in RDF form

19:11:58 <larsbot> you can't say "give me all people born in Lucca", but have to say "give me all topics playing role person in associations of type born-in where topic Lucca plays role place"

19:12:50 <larsbot> born-in should really be a property of the composer rather than a class of resources whose property 'roles' contain the resources that let you connect to what you want

19:13:42 <larsbot> it's as the XML Infoset were to be represented directly in RDF. it would be horrible to work with

19:13:51 <danja> I'm going to have to read that again...slowly...

19:14:06 <larsbot> SethR: labeled directed graphs are nice, but not good design tools. everything essentially becomes node soup

19:14:20 <danja> nodles?

19:14:24 <larsbot> yeah :)

19:14:32 <larsbot> topic maps force you to lift prominent nodes out of the soup relative to the rest of it

19:14:38 <larsbot> helps you think more clearly

19:14:57 <larsbot> to take one example: nobody would make an occurrence type "place of work home page" as in foaf

19:15:19 <larsbot> they'd make an "organization" class and associate the person with that organization, then have a "home page" occurrence for organizations

19:15:42 <SethR> the semantics of the noodles are in the semantics of the arc labels .... that semantics is so diverse its hard to standardize on it quite yet ... but we can standardize on the fact that RDF and TM are both labeled directed graphs.

19:15:54 <larsbot> sure

19:16:19 <larsbot> the difference is not earthshattering. I guess it's similar to the java/jython difference

19:16:25 <larsbot> certain things are best done with the one, others with the other

19:16:43 <larsbot> some people never realize that and do all their work using just one of them

19:16:58 <SethR> perhaps its a small step twards interoperable programming methods, but its a step none the less

19:17:31 <danja> thanks - I've periodically nosed around the docs, and Kal's API, but never really got a handle

19:18:23 <larsbot> danja: lots of people seem to have it that way, just like lots of people seem to have that problem with RDF

19:18:27 <larsbot> I'm not sure why this is

19:22:25 <SethR> i have a theory, if people share graphs (all people actually working on the same graph), then they naturally evolve to share the semantics of the arc labels .. whether that vocabulary is in TM or RDF makes no difference.... just a conjecture ... comments?

19:24:35 <larsbot> I think if we prepare the RDF/TM ground correctly that will happen

19:24:49 <larsbot> certainly that is what I hope and am trying to work towards

19:25:04 <SethR> lars, wanna share a graph?

19:25:09 <larsbot> what we need is a way to express (in RDF) how an RDF graph maps to a topic map

19:25:20 <larsbot> an a way to express (in topic maps) how a topic map maps to RDF

19:25:25 <larsbot> SethR: sure :)

19:25:41 * danja Mentograph churns

19:26:55 <danja> larsbot - is it (theoretically at least) possible to describe a universal TM <-> RDF mapping?

19:27:57 <larsbot> danja: you mean a generic schema-indpendent one? I don't think so

19:28:22 <danja> where might it fall over, do you reckon?

19:29:07 <larsbot> let's say we have rdf triple (a, b, c)

19:29:22 <larsbot> I know "a" refers to a resource, and that a resource is the same as a subject

19:29:28 <larsbot> so I can think of "a" as a topic

19:29:50 <larsbot> but "a" is actually a URI, so I need to know: is this the URI of the resource, or the URI of something that describes the resource?

19:30:25 <SethR> you got to define that using an arc label. then it would be clear.

19:30:44 <larsbot> yep, but then you've got a specific mapping, not a generic one

19:31:14 <larsbot> "b" is easier. a property URI must be conceptual, so I can assume it refers to a resource describing the property

19:31:30 <SethR> its generic if you agree to use the same label to specify it

19:31:49 <larsbot> for your sense of "generic", but not for mine :-)

19:32:14 <larsbot> what I meant was something that would let you just convert any RDF graph to a topic map, and vice versa, with no added information at all

19:32:24 <larsbot> sadly, I don't think that's doable

19:32:28 <danja> hmm - "a" I glaze over when that stuff comes up on rdf-interest, xml-dev etc

19:32:37 <larsbot> sorry, danja

19:32:54 <larsbot> if I use a meaningful example it becomes less clear, but I can do it anyway

19:32:57 <danja> are you saying that TMs care more about the nature of the URI

19:33:00 <larsbot> yes, I am

19:33:02 <danja> ?

19:33:05 <danja> oops

19:33:46 <danja> so the problem there is RDF -> TM

19:33:51 <SethR> so convert the TM to a labeled directed graph that distinguishes between the URI of something and it's Resource

19:34:03 <larsbot> danja: yes. going the other way is much easier

19:34:30 <SethR> when the graph goes back to RDF, then you need to define an arc label in RDF for that purpose

19:34:37 <larsbot> SethR: yeah, you can do that.

19:34:43 <larsbot> exactly

19:34:52 <danja> pull another URI from the sky?

19:35:06 <larsbot> nah, just define an xtm:subjectIndicator property

19:35:11 <danja> ok

19:35:29 <SethR> so TM is more precise where it comes to map\territory problems .. very fine ... but RDF can express those too, we just need to defien a schema for that purpose

19:35:29 <larsbot> then you can say (bnode, xtm:subjectIndicator, http://psi.ontopia.net/rdf/#Topic)

19:35:46 <larsbot> then you know that the bnode represents something described in the http resource

19:35:54 <danja> that makes sense

19:36:13 <larsbot> SethR: precisely. I've been nagging the RDF folks about this for a long time, and they have seen the problem, but there's no solution yet

19:36:25 <danja> is it on the issues list?

19:36:48 <larsbot> danja: I think so. I'm not familiar with the RDF issues list

19:37:15 <dajobe> I'm not familiar with that problem

19:37:20 * dajobe distracted now elsewhere

19:37:21 <danja> like I said, I always glaze over when the nature of URIs comes up...

19:37:27 <larsbot> danja: the problem is when this starts out on the RDF side. then you've got (http://psi.ontopia.net/rdf/#Topic, rdf:type, infoset:InformationItemType)

19:37:51 <larsbot> dajobe: other RDF people are. I discussed it with Eric Prud'hommeaux only yesterday

19:38:36 <larsbot> now you don't know whether the RDF is talking about the web page or the thing that corresponds to a topic

19:38:54 <larsbot> danja: glaze over, yeah, lots of people have that

19:38:56 <larsbot> :-)

19:40:37 <danja> and this difference is well-defined in TMs?

19:40:40 <larsbot> dajobe: he called it his biggest problem right now

19:40:45 <larsbot> danja, yes it is

19:41:00 <danja> might RDF borrow the definition?

19:41:24 <larsbot> yeah, certainly. there's no magic involved, and it can be solved easily in RDF

19:41:27 <larsbot> let me try to explain. I've failed with RDF people before, but I may succeed now :)

19:41:29 <danja> (give it back on Tuesday)

19:41:36 <larsbot> sure :)

19:42:24 <danja> fire away larsbot

19:42:34 <larsbot> was garbage-collecting :)

19:42:55 <SethR> ill try to mentograph as you talk

19:42:59 <larsbot> let's say you want to say "the XML spec was written by Tim Bray", and "XML is a meta-language"

19:43:20 <larsbot> a plausible way to say it in RDF would be

19:43:42 <larsbot> (http://www.w3.org/TR/REC-xml, dc:creator, bnode-for-tim)

19:44:03 <larsbot> (http://www.w3.org/TR/REC-xml, rdf:type, foo:meta-language)

19:44:13 <larsbot> well, semi-plausible, but still

19:44:40 <larsbot> the problem is that this doesn't distinguish between the resource (the spec) and the language (the abstract notion)

19:44:54 <SethR> timBl would say that http://www.w3.org/TR/REC-xml is a document ... but that http://www.w3.org/TR/REC-xml#spec is the spec resource you are describing

19:45:08 <larsbot> SethR: exactly, and that solution does IMHO not work

19:46:01 <larsbot> what would work would be to use a bnode for http://...#spec, and then use a special property to say "the resource is described at http://www.w3.org/TR/REC-xml"

19:46:20 <larsbot> this solution has been proposed within the RDF community, and it would work perfectly with topic maps, as well as with RDF

19:46:32 <larsbot> argument against, I'm told, is that it's a significant change to RDF

19:46:51 <larsbot> (when talking about the RDF community I'm paraphrasing what other people tell me, though)

19:47:18 <larsbot> if you always used bnodes for the things TimBL calls documents you would be safe, basically

19:48:31 <danja> hang on, REC-xml and REC-xml#spec are two different URIs

19:48:40 <larsbot> yep, they are

19:48:52 <SethR> ok, but i dont get that this is any change to RDF at all ... rather it is just good data modeling in RDF ... can you document where the RDF community has rejected the use of bNodes as you have described?

19:49:50 <larsbot> SethR: I don't understand that this is a change either. I don't know that the bnode proposal has been rejected, I just know there's resistance to it. AFAIK the jury is still out

19:50:45 <larsbot> danja: take http://www.w3.org/TR/REC-xml#dt-wellformed, then

19:50:52 <danja> the bnode points to the doc URI and the concept URI

19:51:03 <larsbot> let's say I want to talk about that para, *and* the concept of wellformedness...

19:51:06 <larsbot> same problem

19:51:18 <SethR> thing is the use of bNodes here does not break any RDF spec that I am aware of, so this supposed "resistance" is just your subjective groking of what people say in a non-normative manner

19:51:27 <larsbot> danja: hmmmm. can you give example triples?

19:51:37 <larsbot> SethR: absolutely

19:51:58 <danja> bnode spec REC-xml#spec

19:52:10 <danja> bnode concept REC-xml

19:52:27 <danja> (ugh) or vice versa...

19:52:32 <larsbot> danja: yeah, that works.

19:52:55 <larsbot> with that solution the URIs always refer to the resources they dereference to

19:53:02 <larsbot> that's the key issue

19:53:04 <danja> and how would that map to TMs

19:53:19 <larsbot> in RDF, when you say http://... you don't always mean the thing I get if I follow that URI

19:54:00 <larsbot> mapping probably by turning bnode into topic

19:54:04 <larsbot> spec is then occurrence of topic

19:54:14 <larsbot> REC-xml then becomes subject indicator of topic

19:54:18 <danja> but topic needs a URI?

19:54:23 <larsbot> no

19:54:27 <larsbot> a topic is a bnode, really

19:54:34 <danja> ah, right

19:56:21 <dajobe>http://www.imsproject.org/metadata/imsmdv1p2p1/imsmd_bindv1p2p1.html#1208705

19:56:22 <dc_rdfig> B: http://www.imsproject.org/metadata/imsmdv1p2p1/imsmd_bindv1p2p1.html#1208705 from dajobe

19:56:28 <dajobe> B:|IMS Meta-Data RDF Binding

19:56:28 <dc_rdfig> Titled item B.

19:57:19 <dajobe> B:their use of modelling of language is interesting, using interpretation properties after TimBL - see [http://www.imsproject.org/metadata/imsmdv1p2p1/imsmd_bindv1p2p1.html#1208815|section A.4.3 -Language-specific strings]

19:57:19 <dc_rdfig> Added comment B1.

19:58:18 <dajobe> B:avoiding xml:lang entirely - recommending against it in fact

19:58:18 <dc_rdfig> Added comment B2.

19:59:00 <larsbot> B:extremely interesting. they're a hair's breadth away from published subjects

19:59:01 <dc_rdfig> Added comment B3.

19:59:56 <dajobe> B:thought you'd like it :)

19:59:56 <dc_rdfig> Added comment B4.

20:10:32 <larsbot> danja/SethR: thanks for this discussion. helped me clarify my thoughts quite a bit :)

20:12:49 <SethR> dialogue was good ... the graph is better ! ... but takes a bit longer

20:13:58 <larsbot> would be cool to see it when it's ready

20:16:28 <danja> thanks larsbot - cleared up stuff for me nicely

20:17:24 <larsbot> glad to hear it :)

20:19:12 <danja> now got 'AI' to watch - quality of the evening descends... ;-)

20:19:20 <larsbot> AI?

20:19:31 <danja> movie

20:19:38 <dajobe> barely

20:19:41 <larsbot> oh, I see :-)

20:19:49 <danja> heh

20:19:53 * larsbot is very eurocentric

20:19:59 <dajobe> ignore it; watch Minority Report instead

20:20:25 <larsbot> heard good things about it, but think I will ignore that, too

20:20:45 * danja is centro-eurip

20:21:18 <danja> thanks for rec. but the DVD goes back to hire shop tomorrow :-(

20:37:54 <dmiles_laptop> dmiles_laptop is now known as mea_culpa_now

21:42:52 <mea_culpa_now> mea_culpa_now is now known as mea_culpa_for_re

21:42:59 <mea_culpa_for_re> mea_culpa_for_re is now known as mea_culpa_real

23:11:44 <mea_culpa_real> mea_culpa_real is now known as mea_culpa