Semantic Web Interest Group IRC Chat Logs for 2003-08-06

00:39:08 <FreddieFred> FreddieFred is now known as JibberJim

01:22:18 <GNUPredator-Work> GNUPredator-Work is now known as GNUPredator

05:35:37 <JimJibber> JimJibber is now known as JibberJim

06:10:03 <mdupont_> mdupont_ is now known as GNUPredator

07:47:37 <ChanServ> [#rdfig] This channel is logged and blogged: http://logicerror.com/rdfIRCWelcome

08:52:37 <ChanServ> [#rdfig] This channel is logged and blogged: http://logicerror.com/rdfIRCWelcome

09:29:18 <danbri>http://webservices.xml.com/pub/a/ws/2003/08/05/uml.html

09:29:20 <dc_rdfig> A: http://webservices.xml.com/pub/a/ws/2003/08/05/uml.html from danbri

09:29:30 <danbri> A:|UML for Web Services

09:29:31 <dc_rdfig> Titled item A.

09:29:39 <danbri> A:Interesting for WS description folk...

09:29:39 <dc_rdfig> Added comment A1.

09:53:20 <mortenf> hi dajobe

09:54:13 <mortenf> i assume librdf_storage_contains_statement is intended to only look for statements in the current model?

09:58:11 <dajobe> mortenf: yeah

09:58:15 <dajobe> s/model/storage/

09:58:44 <mortenf> yeah, sometimes i get confused about that, i guess i should remember "storage" as "model storage"...

10:00:09 * danbri had an idea for mozilla

10:00:15 <danbri> 'truth tracking datasource'

10:00:32 <danbri> some datasources just contain triples which may or may not correspond to the state of the 'outside' world

10:00:36 <dajobe> I wish I hadn't used storage, store would have been better but maybe too generic

10:00:50 <danbri> others, like the filesystem or mail ones (if they still use mail/rdf) actually try to keep the world in sync with the RDF view

10:01:07 <danbri> If you delete something in the graph, it zaps your file, if you change its location, it moves your file, etc.

10:01:20 <danbri> It's important to know which datasources have this characteristics

10:01:34 <danbri> mydb.change(danbri,age,29) etc

10:02:14 <swh> heh, we've used model in a confusing way too. in 3store it refers to a subgraph. oh well.

10:03:05 <danbri> and we have 'model and syntax' and 'model theory' confusion too, ho hum

10:03:14 * danbri pities anyone translating this

10:03:55 <swh> i've been sweating blood over the RDFS tests, thier a real pain to run

10:04:15 <danbri> awesome, thanks!

10:04:28 <dajobe> hi steve

10:04:50 <swh> dajobe: hi, thanks for kicking me into doing the tests, found a stupid entailments bug which is good

10:04:51 <dajobe> your 3.2.8 sourceforge release notes said you already did them, maybe I was confused

10:05:04 <dajobe> ah good :)

10:05:22 <swh> dajobe: they did? its not what I meant. we have internal tests, but they turned out to be b0rked

10:05:56 <swh> and we were testing agoings the old WD, not the new one

10:06:11 <swh> I really dont like the bnode->type thing for literals, but I said that on the list

10:06:36 <dajobe> eh?

10:07:06 <swh> that (<f>, <b>, "baz")

10:07:16 <swh> entails (<f>, <b>, _:1)

10:07:24 <dajobe> ?

10:07:28 <swh> (_:1, <rdf:type>, <rdfs:Literal)

10:07:31 <dajobe> _:1 isn't legal syntax

10:07:37 <dajobe> do you mean _:a or rdf:_1

10:07:46 <swh> er, sorry I just meant some bnode

10:07:49 <dajobe> the fomer I guess

10:07:51 <swh> _:a will do

10:08:06 <danbri>http://www.ecademy.com/module.php?mod=club&op=forum&c=628&t=21142

10:08:07 <dc_rdfig> B: http://www.ecademy.com/module.php?mod=club&op=forum&c=628&t=21142 from danbri

10:08:23 <danbri> B:|More FOAF harvester stats

10:08:23 <dc_rdfig> Titled item B.

10:08:51 <danbri> B:"Freddie has found 3,612,159 bits of RDF info. (Properties). These span 179 namespaces."

10:08:51 <dc_rdfig> Added comment B1.

10:08:56 <libby> there's a hell of a lot there, blimey

10:09:36 <danbri> B:That's a *lot* of namespaces. I think we've been too distracted by the triple count, since that has the impact on our databases...

10:09:36 <dc_rdfig> Added comment B2.

10:09:42 <swh> yowza, aparently there are 10k triples in out ontolgy :( I knew it was a bit of a pain

10:10:00 <danbri> in your _ontology_, wow!

10:10:28 <dajobe> which ontology?

10:10:37 <dajobe> ah, portal

10:11:10 <dajobe> "1 http://www.ideaspace.net/users/wkearney/schema/astrology/0.1"

10:11:28 <danbri> oh right, i thought you meant the ontology itself was tubby

10:11:39 <dajobe> the count is properties

10:12:30 <mortenf> heh, am i the only one that uses that ns?

10:12:37 <danbri> B:Even if half of them are crap, or most ill-documented, this still represents a significant amount of work and content.

10:12:38 <dc_rdfig> Added comment B3.

10:12:38 <danbri> ns?

10:12:45 <mortenf> the astrology one

10:12:51 <swh> danbri: yeah, thats the ontolgy, the kb has 8M triples

10:13:20 * danbri wonders if mortenf has looked at swh's SQL storage strategy

10:13:33 <mortenf> at one point i did...

10:13:36 <danbri> (...talking about gluing Redland to SQL stores)

10:16:15 <mortenf> it seems i'm able to maintain a throughput of about 5k triples per minute (when adding), but it's mostly disk time

10:16:39 <mortenf> and with 18M+ triples from spacenamespace:dk, it'll take 3 days to load...

10:16:40 <swh> mortenf: yeah, import throughput is one of a major bugbears

10:16:46 <mortenf> (and take up 6G)

10:17:01 <swh> we can get 1000 a sec when the KB is quite empty, but it drops to about 300/s when we have a few million triples in there :(

10:17:25 <swh> it takes >2 hours to import our whole kb from .xml files

10:17:40 <mortenf> yeah, mine starts at 10k/minute, ends up at 5k/minute, when approaching 100k

10:18:05 <mortenf> 300/s isn't bad!

10:18:17 <swh> Theres a snapshot of whats in the KB at the moment: http://triplestore.aktors.org/manage/snapshot.html

10:18:31 <swh> seems to be full of hundreds of little 4 triple files...

10:18:46 <swh> mortenf: its ok, but I think it can be imporoved

10:19:31 <mortenf> there's always room...

10:19:58 <swh> mortenf: what KB are you using?

10:20:29 <mortenf> kb?

10:20:42 <swh> knowledge base - rdf store / whatever ;)

10:20:46 <mortenf> redland

10:20:49 <swh> whats the canonical word?

10:20:55 <mortenf> i'm not sure...

10:20:55 <swh> k

10:21:11 <mortenf> what hash are you using that fits in 16 bytes?

10:21:26 <swh> top half of an MD% :)

10:21:29 <swh> *MD5

10:21:31 <mortenf> ah...

10:21:48 <mortenf> i'm using the full md5, stored in two bigint's...

10:21:53 <swh> i was going to use crc64, but there are good reasons not to, and MD5 is not significantly slower

10:22:15 <swh> mortenf: it slowwed down the joins too much when it wouldn;t fit in one native SQL field

10:22:22 <swh> and MySQL doesnt have 128 bit ints

10:22:34 <mortenf> yeah, i guess i'll get to that problem at some point.

10:22:47 <swh> mortenf: do you need realtime queries?

10:22:56 <mortenf> but then there's the problem with collisions, have you calculated anything re that?

10:23:01 <mortenf> yep...

10:23:09 <swh> I think most people dont care much, but we need some to drive uis that cant take more than 2ms or so

10:23:32 <dajobe> not haystack?

10:23:43 <swh> haystack?

10:24:15 <mortenf> my test-case will be a foaf person query, if it can be done in less than a second on 20M triples i'm happy...

10:25:00 <swh> never tried 20M triples, we only have 10 :)

10:25:58 <mortenf> i can give you 18M in N-Triples format if you want them :)

10:26:01 <swh> what query? (?person, <rdf:type>, <foaf:Person), (?person, ?pred, ?obj) ?

10:26:12 <swh> mortenf: maybe later ;)

10:26:19 <mortenf> well, a little more complicated than that, but yep

10:26:29 <swh> If i stuff another 18m triples into our kb i will get shot :)

10:26:34 <mortenf> heh

10:26:54 <swh> sadly my test machine is a laptop and it wont like 20odd million triples much

10:27:27 <swh> - slow disks

10:28:00 <mortenf> btw, my tests are on a 750Mhz Duron, 3+ year old SCSI HD (7200)

10:28:21 <swh> mine are on dual xeon with u320 disks

10:28:26 <mortenf> heh :)

10:28:44 <swh> but before that we were using a PIII800 with crummy old IDE disks - didn't make that much difference though

10:29:03 <mortenf> true, it'll only do one thing at a time anyway

10:29:10 <mortenf> which mysql version?

10:29:35 <swh> yup, the best thing was the import time, faster disks really, help, but I dont find i stress the cpus much

10:29:56 <mortenf> yep, that's what i found as well

10:29:58 <swh> 2 cpus helps for concurrent queries, but it doesn't make individual ones anyfaster ofcourse

10:30:15 <mortenf> perhaps a raid0 on diff channels would help

10:30:16 <dajobe> sounds like a RDB is the wrong hammer, to me

10:30:28 <swh> mortenf: thats what im using

10:30:34 <swh> it does :)

10:30:37 <mortenf> ah

10:30:50 <darobin> dajobe: what else are you thinking of?

10:31:16 <swh> dajobe: actually RDBMS's are quite appropraite - the translation from rdf graphs to relation calc. is pretty simple and they are fast

10:31:25 <dajobe> dunno, it just sounds the overheads are great for triples

10:32:00 <dajobe> swh: in theory, I've seen all the SAIL work too. In practice - slow

10:32:22 <swh> dajobe: slow for what?

10:32:26 <dajobe> maybe if you don't care so much about RDFS or layers above, then it seems like an overhead

10:32:58 <swh> dajobe: yeah, thats likly, but I'd bet that its pretty efficiewnt for just a simple RDF graph match

10:33:15 <swh> we should sort out some benchmarks for RDF stores

10:33:53 <dajobe> those always end up artificial

10:34:11 <swh> yes, but it will give us some idea

10:34:14 <dajobe> some query use cases and big sets of data that use them

10:34:28 <dajobe> s/use them/they use/

10:36:02 <danbri> swh, are you answering Squish/RDQL queries by rewriting into a big self-join SQL query?

10:36:20 <swh> danbri: aproximatly, yes

10:36:43 <danbri> my worry there is that the SQL query optimisers haven't a clue what's going on inside the RDF query, so won't be able to re-order the clauses effectively, doing sensible stuff (functional, inversefunctional) earlier...

10:36:43 <libby> really swh?

10:36:54 <swh> I have a paper if you care

10:37:18 <danbri> maybe oracle have fancy data driven stuff that could figure out the internal structure of the data, but i fear lots of it isn't amenable to simple statistical analysis by the sql engine

10:37:20 <danbri> paper: yes please!

10:37:25 <swh> danbri: they get a lot of clues form the indexes of the tables

10:37:31 <swh> paper: http://eprints.ecs.soton.ac.uk/archive/00007970/

10:37:42 <dajobe>http://eprints.ecs.soton.ac.uk/archive/00007970/

10:37:43 <dc_rdfig> C: http://eprints.ecs.soton.ac.uk/archive/00007970/ from dajobe

10:37:45 <danbri> can i blog it?

10:37:46 <danbri> oh ta

10:37:49 <dajobe> C:|3store: Efficient Bulk RDF Storage

10:37:50 <dc_rdfig> Titled item C.

10:37:51 <swh> with appologies for bad relational calculus, I'm very rusty

10:38:00 <dajobe> C:<swh> with appologies for bad relational calculus, I'm very rust

10:38:00 <dc_rdfig> Added comment C1.

10:38:02 <danbri> do you think they get enough clues?

10:38:10 <danbri> can't read now, but defintely interested

10:38:13 <dajobe> C1:<swh> with appologies for bad relational calculus, I'm very rusty

10:38:14 <dc_rdfig> Replaced comment C1.

10:39:56 <swh> nb. it doesnt do any fancy owl stuff, so the opurtunities for learning fomr the schema are pretty minimal - but it does, eg to decide wether to backchain some entailments

10:42:23 <dajobe> your desc of TAP isn't accurate, they do RDFS and provid emore than GetData

10:42:35 <swh> d'oh

10:42:41 <swh> blame nmg :)

10:43:20 <dajobe> "3.4 RDFLib/Redland" ? wtf?

10:43:30 <dajobe> given that it doesn't even mention rdflib

10:44:38 <swh> ...hold on, opening ps file

10:45:35 <swh> I though rdflib was part of redland?

10:45:44 <dajobe> not at all

10:45:45 <swh> appologies if not, I'l correct it

10:45:54 <swh> ok, whats the distinction?

10:46:14 <dajobe> they are entirely different things

10:46:20 <swh> ah, sorry

10:46:21 <dajobe> eikeon wrote rdflib in python

10:46:24 <dajobe> I wrote redland in C

10:47:35 <mortenf> but redland aka librdf...

10:48:05 <dajobe> that's the library name

10:48:28 <dajobe> redland will have more than just a library

10:48:34 <swh> right, sorry, must have got confused

10:48:50 <swh> a la 3store and librdfsql :/

10:48:57 <swh> ...fixing

10:50:05 <dajobe> I still think MD5 is overkill for this, but we had this argument before

10:50:12 <swh> is the description of Redland otherwise correct? it came from a much older paper and mucht be out of date

10:50:35 <mortenf> dajobe, i'm all open to alternatives

10:50:35 <dajobe> it's mostly your comment rather than description

10:50:52 <swh> dajobe: yes, but MD5 isn't really any slower than crc64 and it makes it hard for people to deliberatly clash keys

10:51:06 <dajobe> I don't see any mention of bnodes there

10:51:14 <dajobe> do you make them illegal URIs?

10:51:38 <swh> dajobe: dunno, bnodes aren't really treated specially

10:51:45 <swh> they should be though

10:51:49 <dajobe> well the parser certainly does

10:52:08 <dajobe> and no literal language either?

10:52:18 <dajobe> you should say what bits of rdf you are removing

10:52:22 <swh> yes, it picks up on that and adds a suffix to try to make them unique the the sub-graph - its hacky though

10:52:38 <swh> dajobe: yes, no datatype support at all - yet

10:52:44 <dajobe> language != datatype

10:52:51 <swh> though it looks like i can implement it without too much pain

10:53:05 <swh> dajobe: language in what sense then?

10:53:15 <dajobe> literal languages

10:53:25 <dajobe> "ou est le language?"-fr

10:53:49 <swh> ok, I thought that was part of datatypes - well, we dont do them either - they weren't in the version of hte RDFS docs I was working form IIRC

10:54:34 <dajobe> glad to see you junked namespaces internally

10:54:45 <dajobe> they might make handy hints for serialization, but not otherwise

10:55:11 <dajobe> hints - which I'm going to add to the rdf/xml output of redland in due course; probably move it into raptor then

10:55:22 <swh> dajobe: yes, thats why we had them, they were a royal pain though

10:56:02 <swh> dajobe: sorry, you lost me, what are you going to move into raptor, rdf/xml serialisation?>

10:56:17 <dajobe> yes, likely

10:56:30 <dajobe> i.e. triples->rdf/xml to join rdf/xml->triples

10:56:35 <swh> that would be handy - we have some code to do it currently, but its not released

10:57:34 <swh> dajobe: are you going to ask for two passes of the tripls, or cache them to build the namespaces?

10:57:49 <swh> s/are you/do you/

10:57:54 <dajobe> I guess my first thought is your scheme wouldn't work for me since it drops some rdf bits. And I didn't even get to contexts

10:58:23 <swh> yes, our requirements are pretty minimal

10:58:25 <swh> contexts?

10:58:45 <dajobe> I've not written any code yet, so the ideas are these so far: 1) a streaming one which just does what it can 2) a "pretty" one which will suck everything into memory and try to emit things grouped

10:58:46 <swh> I'm considering adding storage of the lang/datatype stuff though, it wont hurt much and it might help some people

10:58:58 <dajobe> fixing bnodes is critical

10:59:14 <swh> dajobe: we currently take two runs through the tripls, its a bit memory hungly otherwise

10:59:16 <dajobe> a quick hack won't do it. You've got to make sure they work ok

10:59:42 <dajobe> I've not got to where you explain 'model' yet, so maybe they are redland-contexts

10:59:53 <swh> agreed - are you concerned about roundtripping bnodes? its something we rejected, but maybe its neccesary

11:00:31 <dajobe> I'm concerned about: <swh> yes, it picks up on that and adds a suffix to try to make them unique the the sub-graph - its hacky though

11:00:39 <swh> nb. our current rdf/xml export does put hte bnodes back, but you could trip it up by creating a uri with the _: namespace :(

11:00:41 <dajobe> which doesn't seem like you guarantee they work

11:01:37 <swh> dajobe: potentially thier could be collisions with bnodes between graphs, but its unlikly, making it guananteed would be easy though

11:01:56 <swh> its just not that high up the list of features

11:01:56 <dajobe> I've just added some new support to raptor that might help there

11:02:05 <swh> ?

11:02:08 <dajobe> you can either tweak the default bnode id gen algorithm

11:02:22 <dajobe> or register a handler that'll let you do your own thing

11:02:28 <swh> that would certainly make the code cleaner

11:02:33 <dajobe> look here:

11:02:39 <swh> currently I intercept bnodes and mangle the uris

11:02:47 <dajobe> no need for that

11:02:51 <dajobe>http://www.redland.opensource.ac.uk/raptor/libraptor.html

11:02:53 <dc_rdfig> D: http://www.redland.opensource.ac.uk/raptor/libraptor.html from dajobe

11:03:00 <dajobe> D:|Raptor API

11:03:01 <dc_rdfig> Titled item D.

11:03:21 <dajobe> D:new in 0.9.11 "Added raptor_set_default_generate_id_parameters and raptor_set_generate_id_handler to control the default generation of IDs, allow full customisation."

11:03:21 <dc_rdfig> Added comment D1.

11:04:02 <swh> great

11:04:25 <dajobe> D:so you can either tweak the "genid1" function such as starting as "genid<bigint>" or you can do your own thing, such as output anything you like

11:04:26 <dc_rdfig> Added comment D2.

11:04:46 * mortenf wonders if he'll have to do something storagewise re bnodes in redland, if parsing two context into same model...

11:05:03 <swh> dajobe: I'd just put an autoincrement field in the DB and bump it every time I do an import

11:05:09 <dajobe> that's the idea

11:05:26 <dajobe> mortenf: I'm going to do similar for redland

11:05:35 <mortenf> k.

11:05:38 <dajobe> probably gettimeofday() at redland start plus some prefix

11:05:51 <swh> currently I use the has of the model uri, time and pid or something like that

11:06:06 <swh> *hash

11:06:10 <dajobe> so maybe "_genid"+timeofday_at_redland_start+(inde++)

11:06:29 <dajobe> time's expensive ot fetch sometimes, I'd just take it once

11:06:37 <swh> yup, i do

11:06:48 <dajobe> and some systems don't have a pid

11:07:08 <dajobe> (non-POSIX)

11:07:16 <swh> 3store is only portable to POSIX

11:07:25 <swh> *shrug*

11:07:34 <dajobe> threads

11:07:49 <swh> good point

11:07:52 <swh> hm....

11:08:30 <swh> should be pid+thread_number I guess, but the import code is not threaded anyway

11:08:54 <dajobe> you might be running in a threaded environment without knowing it

11:09:04 <dajobe> for example, libxml2 can be built threaded

11:09:14 * mortenf hasn't checked threadsafety, but should be able to run concurrently otherwise

11:09:20 <dajobe> or if 3store is linked into a GNOME/gtk app...

11:09:23 <swh> no, because the import code is in a comandline tool :/

11:09:26 <swh> that will change though

11:09:33 <dajobe> links with raptor, which may use libxml2

11:09:48 <swh> librdfsql doesnt have any parsing methods - there external

11:10:00 <swh> and 3store is just a bunch of tools that link against librdfsql

11:10:03 <dajobe> try ldd on your binary anyway

11:10:11 <dajobe> it might be a threaded mysql

11:10:15 <dajobe> but I'm labouring the point!

11:10:50 <swh> anyway - I think just having an autincrement field in the DB would work fine :)

11:10:59 <dajobe> yes

11:11:27 <dajobe> I'm probably going to have some hidden triples at some point in redland's stores for such things

11:11:33 <dajobe> hidden in that they won't be visible from user apis

11:12:08 <swh> currently all of 3stores can be queried (things like import history and so on) via the 4th arg to RDQL patterns

11:12:22 <swh> there is some stuff that isnt stored as triples though (like cache state)

11:13:21 <swh> eg (<a>, <b>, <c>, ?model), (?model, <foo:lastUpdated>, ?when) will give you the last updated time the the abc triple

11:13:25 <dajobe> your hybrid entailment approach looks nice

11:13:41 <swh> dajobe: its pretty much neccesary

11:13:43 <dajobe> you should link to this paper in your impl report

11:13:50 <dajobe> *cough* *cough*

11:13:55 <swh> heh ;)

11:14:47 <dajobe> C:a nice hybrid approach to rdfs entailment - forward chaining & backward (at query time)

11:14:49 <dc_rdfig> Added comment C2.

11:16:17 <swh> dajobe: is a uri that starts _: always a bnode, even if its explitly stated in the graph?

11:16:36 <dajobe> it's not an url, deliberatly _: is an illegal url prefix

11:16:42 <dajobe> but _:foo are always bnodes

11:16:42 <swh> ah, ok

11:17:27 <swh> in that case, bar the dubious uniqing technique 3store should be ok for bnodes

11:17:44 <swh> the export code mangles them back into bnodes with ID's IIRC.

11:17:54 <dajobe> rdf:nodeID

11:17:58 <swh> yup

11:18:05 <swh> is that legal?

11:18:17 <dajobe> not with rdf:ID

11:18:35 <dajobe> but preserving the actual bnode tokens in the original rdf/xml is not required

11:18:40 <swh> sorry, I meant nodeID - I think - the code is nmg's

11:19:22 <swh> dajobe: right, we dont, which means we cant roundtrip arbitrary graphs, but it would be hard to guarantee anyway and subgraphs can grow

11:19:33 <swh> s/and/as/

11:20:27 <dajobe> jena also doesn't guarantee that but I think it always generates fresh bnodeids on output mapping from (intenral bnodeid, new bnodeid) rather than just emitting the internal bnodeid

11:20:45 <swh> ot: whats the right way to check that the positive entailment tests are satisfied?

11:21:16 <dajobe> your output graph matches the expected one

11:21:44 <dajobe> writing the entailed grpah to ntriples and graph comparing with the expected .nt would work

11:22:06 <swh> the conclusions docuemnt is incomplete AFAICT

11:22:40 <swh> so should the conclusions be a subgraph of your entailments, or should premise+entailments = premise+conclusions

11:23:02 * dajobe looks for jang

11:23:44 <dajobe> conclusions should be isomorphic with premise+entailments I think

11:23:50 <dajobe> er, conclusions

11:23:54 <dajobe> argh, ignore me

11:24:23 <dajobe> you take (input triples) and entail some more - you never remove triple - and that gives you (output triples)

11:24:49 <swh> yes, but how does that relate to the conclusions doc?

11:25:07 <dajobe> they are same as the output

11:25:24 <swh> for the praser tests its easy as the output do is always complete, but not all of the entailments tests are

11:25:45 <swh> I dont think so

11:26:04 <dajobe> it's not my area, and I can't confirm

11:26:23 <dajobe> nobody else has reported this problem

11:26:36 <dajobe> except with rdf:_1 test something you mentioned I think

11:27:12 <swh> take eg rdfs-no-cycles-in-subPropertyOf/test001.(nt|rdf)

11:27:28 <swh> the conclusions file (.nt) is only a subset of the required entailments

11:27:55 <swh> the rdf:_1 thing is the opposite

11:28:15 <dajobe> I can't help, your asking the wrong person.

11:28:36 <swh> ok, i'l do what I think is intended - it sounds like its what the jena people did anyway

11:29:45 <swh> ok, off to lunch now

11:29:52 <swh> swh is now known as swh_lunch

12:04:29 <danbri> dajobe, syntax question:

12:04:31 <danbri> 9: <foaf:topic>

12:04:31 <danbri> 10: <foaf:Person>

12:04:31 <danbri> 11: <foaf:name>Alice</foaf:name>

12:04:31 <danbri> 12: <foaf:knows xyz:foo="1000!">

12:04:31 <danbri> 13: <foaf:Person>

12:04:32 <danbri> 14: <foaf:name>Bob</foaf:name>

12:04:34 <danbri> 15: </foaf:Person>

12:04:36 <danbri> 16: </foaf:knows>

12:04:38 <danbri> 17: </foaf:Person>

12:04:40 <danbri> 18: </foaf:topic>

12:04:44 <danbri> is line 12: illegal?

12:04:49 <danbri> (assuming 9 is a property, etc)

12:04:59 <dajobe> yes

12:05:11 <dajobe> um, maybe?

12:05:12 <dajobe> no

12:05:13 <dajobe> ?

12:05:17 <danbri> bummer, it spoils an argument i was making.

12:05:17 <danbri> heheheh

12:05:20 <danbri> rdf syntax--

12:05:32 <dajobe> don't blame me, I didn't make it!

12:05:38 <danbri> rdf validator doesn't like it

12:05:50 <dajobe> yes, it is illegal

12:05:53 <dajobe> I was confused briefly

12:05:58 <dajobe> on a property *element*

12:06:00 <danbri> Oh, I don't blame you in the slightest. Your spec is way better. It's the syntax that's the issue.

12:06:07 <dajobe> you can only have property attributes if the properyt element is empty.

12:06:19 <dajobe> on a *node element*, that doesn't apply

12:06:19 <danbri> If I yes exactly

12:06:41 <danbri> I knew <foaf:knows mensa:iq="200"/> would work

12:06:51 <dajobe> this is one more abbrv I'd like to kill, but it seems that XPackage uses it....

12:07:00 <dajobe> yeah it is seductive

12:07:29 <dajobe> you can make things that look xml-attribute style but don't quite give the triples you might expect

12:11:58 <danbri>http://rdfweb.org/topic/EggheadExample

12:11:59 <dc_rdfig> E: http://rdfweb.org/topic/EggheadExample from danbri

12:12:19 <danbri> E:|Snippet of RDF, showing a syntax variant

12:12:19 <dc_rdfig> Titled item E.

12:12:32 <danbri> E:Wanted this page to cite from weblog discussions re rss2

12:12:32 <dc_rdfig> Added comment E1.

12:13:29 <dajobe> gosh you used an untyped node, I didn't think you did that :)

12:13:55 <danbri> the type is implied by the rdfs:range on foaf:knows :)

12:14:12 <danbri> I don't use rdf:Description if i can help it

12:14:17 <dajobe> foaf:topic you mean?

12:14:56 <sbp> rdfs:domain you mean?

12:15:07 <danbri> that too, but no.

12:15:20 <danbri> the knows points to a node we only describe with 'iq=200' property

12:24:22 <danbri> E:RSS2/RDF discussion (a bit one-sided) [http://blogs.law.harvard.edu/tech/comments?link=http%3A%2F%2Fblogs.law.harvard.edu%2Ftech%2F2003%2F08%2F04%23a167&p=167&u=tech|continues...].

12:24:22 <dc_rdfig> Added comment E2.

13:02:12 <swh_lunch> swh_lunch is now known as swh

13:31:12 <DanC-AIM> ZZ:

13:31:13 <dc_rdfig> Label ZZ not found.

13:31:25 <DanC-AIM> ZZ:

13:31:25 <dc_rdfig> Label ZZ not found.

13:32:08 <DanC-AIM> byebye

14:38:57 <DanC-AIM> BLURB: Extreme Markup notes

14:38:59 <dc_rdfig> F: Extreme Markup notes from DanC-AIM

14:39:36 <DanC-AIM> 1ogger_1, pointer?

14:54:17 <dajobe> logger_1, pointer?

14:54:17 <dajobe> See http://ilrt.org/discovery/chatlogs/rdfig/2003-08-06#T14-54-17

15:08:29 <DanC-AIM> F: see [http://ilrt.org/discovery/chatlogs/rdfig/2003-08-06#T14-54-17 notes]

15:08:31 <dc_rdfig> Added comment F1.

15:08:55 <DanC-AIM> (fix please? I can't make a pipe on this gizmo)

15:09:31 <DanC-AIM> I haven't seen liam yet, but I'll try to ask him about fragments, danbri

15:09:40 <danbri> thanks DanC, appreciated.

15:09:45 <libby> F1: see [http://ilrt.org/discovery/chatlogs/rdfig/2003-08-06#T14-54-17|notes]

15:09:45 <dc_rdfig> Replaced comment F1.

15:09:58 <deltab> what gizmo?

15:10:46 <DanC-AIM> -- Nishikawa "a metadata network for bridging people and places"

15:11:08 <DanC-AIM> Using oasis geolang PSIs

15:12:16 <DanC-AIM> Hmm a resume is used as a 'subject indicator' for a person. That confirms my intuition that tm:subjectIndicator makes sense as an rdf property

15:12:29 <DanC-AIM> Kinda like foaf:topic

15:13:15 <libby> deltab, I think it's a T-mobile sidekick

15:13:18 <tav> tav is now known as tav|offline

15:13:19 <DanC-AIM> ... Speaker drops EricM's name, mentions damloil:unambiguousProperty [sic]

15:13:41 <DanC-AIM> ... Fly by of dublin core stuff

15:13:59 <DanC-AIM> .google WearableGizmo

15:14:01 <datum> WearableGizmo: http://dm93.org/z2001/WearableGizmo

15:14:05 <libby> who's the speaker?

15:14:20 <libby> ah, mary right?

15:14:26 <DanC-AIM> ... Mention of freenode, libby miller...

15:14:34 <DanC-AIM> Yes, Mary.

15:14:45 <libby> yeah I gave her a hand with some detail re foaf stuff

15:15:30 * danbri remembers chatting w/ Mary about Topic Maps, RDF and FOAF a few times, if this is same Mary

15:15:41 <DanC-AIM> ... Discussion of indirect identification of people...

15:15:48 <libby> yeah it is, she was at xmleurope. she's very nice

15:15:52 <DanC-AIM> "changeability of URLs"

15:16:09 <DanC-AIM> Yes, very nice

15:16:25 <libby> on the left: http://swordfish.rdfweb.org/photos/2003/05/06/2003-05-06-Pages/Image4.html

15:16:30 <libby> I shoudl catalogue taht one

15:16:50 <danbri> good idea

15:17:04 <DanC-AIM> Kaled used uml diagrams for presenting topicmap patterns. Diagrams++

15:17:13 <libby> I forgotten who second from left is...

15:17:39 <DanC-AIM> /me wishes he had submitted his travel tools stuff for this conference... Considers a nocturn

15:17:47 <wkearney99> wkearney99 is now known as bill|rebooting

15:18:39 <DanC-AIM> (somebody wanna find Mary's paper in the exml website?)

15:19:22 <DanC-AIM> Hmmm... 3-ary relation.

15:20:57 <dajobe>http://www.mulberrytech.com/Extreme/Proceedings/html/2003/Nishikawa01/EML2003Nishikawa01.html

15:20:57 <dc_rdfig> G: http://www.mulberrytech.com/Extreme/Proceedings/html/2003/Nishikawa01/EML2003Nishikawa01.html from dajobe

15:21:12 <dajobe> G:|A metadata network for bridging people and places, Mary Nishikawa

15:21:13 <dc_rdfig> Titled item G.

15:21:16 <dajobe> G:all images are 404

15:21:17 <dc_rdfig> Added comment G1.

15:23:08 <DanC-AIM> ... 1/2 joke about finding non-midnight times for international telcons. [calendaring is everywhere!]

15:25:09 <DanC-AIM> Er... Kal Ahmed, rather.

15:25:53 <danbri> BLURB:

15:25:53 <danbri> hm

15:25:53 <dc_rdfig> H: from danbri

15:27:11 <DanC-AIM> /me wishes for badwidth with which to polish up esw.w3.org/topic a bit in anticipation of visitors from exml

15:27:36 <danbri> H:"We are *not* talking hacks, here, except in as much as the whole WWW is We are *not* talking hacks, here, except in as much as the whole WWW is a great big hack." (TimBL, [http://lists.w3.org/Archives/Public/www-tag/2003Aug/0039.html|www-tag], 2003-08-06).

15:27:37 <dc_rdfig> Added comment H1.

15:27:43 <DanC-AIM> Hmm.. Geolang using cia factbook and a few other interesting sources.

15:28:02 <danbri> H1:"We are *not* talking hacks, here, except in as much as the whole WWW is a great big hack." (TimBL, [http://lists.w3.org/Archives/Public/www-tag/2003Aug/0039.html|www-tag], 2003-08-06).

15:28:02 <dc_rdfig> Replaced comment H1.

15:29:17 <DanC-AIM> Demo: ltm sources... Omnigator...

15:29:23 <libby> kal was at that calendaring workshop in bristol, DanC-AIM, don;t know if you remember

15:29:38 <DanC-AIM> Had to change the PSIs for offline use. [sigh!]

15:29:45 <danbri> H:hacking++

15:29:45 <dc_rdfig> Added comment H2.

15:30:39 <danbri> co-depiction and path-cross are natural fit... as in 'where did i meet this person before...?'

15:31:07 <DanC-AIM> "... There is a lot of discussion of how to use URIs as identifiers." <- understatement of the day!

15:32:16 <danbri> I so wish URIs were just another property/value pair in the RDF graph, it'd free up RDF from all the URI navelgazing, and make for a clean way to have things have multiple URI names.

15:32:35 <DanC-AIM> Doubt it.

15:32:37 <danbri> I should shut up griping and see if its implementable...

15:32:53 <danbri> OK, RDF will never be free of navelgazing ;)

15:33:33 <DanC-AIM> You might ask sandro. He persued that for a while. Found it thoretically doable, but practically too tedious.

15:33:52 * danbri nods, supposes it varies across applications

15:33:55 <DanC-AIM> I'm pertty sure you'd end up with the same problems.

15:34:23 <Mutiny> i'm curious, is it some kind of faux paus to have the URI you're using to identify a real-world object be the rdf description of that object?

15:34:28 <danbri> most of my concerns are for apps where you need a ton of identity reasoning to get the data de-fragmented enough to use (foaf harvesting and general rdf harvesting on top of that)

15:34:49 <danbri> it's a faux pas to even mention the idea ;)

15:35:04 <danbri> not at all. It is just one of those rathole topic areas where discussions swirl around and around.

15:35:39 <danbri> Some people do it that way, others get all flustered cos they see it as conflating things with their descriptions. Others care greatly whether your URI contains a '#' or not.

15:35:42 <DanC-AIM> Foaf is a great situation: it exploits the agreement we *do* have about uris (e.g. My homepage is http://...) and bootstraps a network of new info.

15:37:17 <danbri> Yeah, the homepage thing is interesting. A purist model would do 'person1-workPlace->org2--homepage->...'; FOAF shortcuts that with workplaceHomepage, bypassing things that we don't have widespread uris for.

15:37:23 <Mutiny> is it bad form to have a rdf:about for a foaf:person?

15:37:46 <danbri> we should probably add in the other property though (workplace) so the rule can be written expressing the design pattern

15:37:48 <DanC-AIM> People love this omnigator thing. I guess a standard library of navigational stuff and some tools to browse it would go a long way in getting rdf adopted.

15:37:53 <danbri> same w foaf:schoolHomepage

15:38:38 <danbri> mutiny, again, some people do it that way. I don't find much value in claiming some string to be 'the', or even 'a' URI for a person; I stick with other strings that relate to the person in named ways but which uniquely pick them out.

15:39:29 <DanC-AIM> Nothing wrong with using workplaceHomepage. Do define it as a composition of workplace and homepage in the schema.

15:39:54 <danbri> ...need to add "workplace" first, but yes I think well worth doing that.

15:40:03 <Mutiny> ah, in this case it relates to my earlier question of if having a subject uri be a rdf description of a thing

15:40:06 <danbri> Can I express that pattern in OWL? I had thought not...

15:40:23 <danbri> mutiny, yup, same issue

15:40:27 <DanC-AIM> Indeed, not...

15:41:15 <danbri> I've added all the OWL i can think of (without reading the spec again) to FOAF, mostly class disjointness and inverse-functionals, plus the Ontology tag.

15:41:20 <DanC-AIM> I defined a compsition gizmo somewhere.

15:41:26 * danbri thinks that's enough to count as being an ontology

15:41:38 <danbri> danc, i'd be interested see that. its a pattern i value and re-use a bunch.

15:41:46 <DanC-AIM> Disjointness? Interesting... For example?

15:42:25 <danbri> xmlns.com/foaf/0.1/index.rdf

15:42:27 <danbri> er

15:42:29 <DanC-AIM> Indeed, composition is a pattern worth writing up. Does it have a wiki topic yet?

15:42:34 <danbri>http://xmlns.com/foaf/0.1/index.rdf

15:42:34 <dc_rdfig> I: http://xmlns.com/foaf/0.1/index.rdf from danbri

15:42:41 <danbri> wiki, not to my knowledge

15:42:45 <danbri> I:|FOAF in RDFS/OWL

15:42:46 <dc_rdfig> Titled item I.

15:43:36 <danbri> I:Some classes are disjoint. Eg. Person and Document.

15:43:37 <dc_rdfig> Added comment I1.

15:43:58 <DanC-AIM> Disjointness in foaf... So you can write clearly wrong (I.e. Inconsistent) foaf thingies.

15:44:16 <danbri> I:...which was useful in explaining possibility for [http://rdfweb.org/topic/FoafContradictions|FoafContradictions].

15:44:16 <dc_rdfig> Added comment I2.

15:46:14 <bitsko> danbri: re. rdf in rss, no, you don't need to repeat yourself for echo ;) at least not so strongly. note the discussion on http://intertwingly.net/wiki/pie/RdfAndEcho

15:46:44 * danbri bbiab

15:51:19 <danbri> did someone link the discussion on dave's blog?

15:51:44 * danbri glad ppl are following it... I wonder who'll be first, RSS2 or Echo, to specify a nice RDF-friendly embedding mechanism.

15:52:17 <DanC-AIM> For the bandwidth-challenged, bitsko could you redice that discussion to a number between 0 and 1, i.e. Odds that echo will use rdf/xml syntax?

15:52:27 <DanC-AIM> -- Freese

15:52:45 <tav|offline> tav|offline is now known as tav

15:53:01 <danbri> I like 'rdfn:rdfxml="on"' or somesuch to signal 'this sub-tree of XML is in RDF/XML syntax, and applies to the thing the current element is a property of'.

15:53:13 <DanC-AIM> Lexus/nexus: 32000 sources. 4.1billion docs +1milion/day. Stuff that's 300years old. ?? Terabytes.

15:53:21 <danbri> I believe currently unlikely that Echo will use RDF/XML syntax. Nobody is actively pushing that.

15:53:53 <DanC-AIM> So less than 1% odds, I gather.

15:53:55 <bitsko> I would venture to say that it's already concensus that it will not.

15:53:56 <danbri> I tried sketching an rdf mapping, but it was ugly because they want to ascribe properties to blobs of encoded XML, and that made for lots of extra wrapper nodes in the graph.

15:54:18 <danbri> well, I also wouldn't be suprised if Echo fails to settle down by end of summer.

15:54:32 <danbri> It is too biased to weblogging use case at expense of things like Job-info syndication which are gaining interest

15:54:38 <bitsko> I think the timeline is a little longer than end of summer, yes

15:55:00 <bitsko> can you expand on "ascribe properties to blobs of encoded XML"?

15:55:16 <danbri> see http://www.w3.org/2001/sw/Europe/200307/echopie/ for my example...

15:55:28 <dajobe> encoding format, language

15:55:33 <danbri> I took an echo feed, dunno if still fashionably formatted, and rdfized it.

15:55:53 <danbri> -

15:55:53 <danbri> <content>

15:55:53 <danbri> -

15:55:53 <danbri> <rdf:Description ne:mode="escaped">

15:55:53 <danbri> <dc:format>text/html</dc:format>

15:55:54 <danbri> -

15:55:56 <danbri> <data xml:lang="en-gb" rdf:parseType="Literal">

15:55:58 <danbri> etc

15:56:17 <danbri> dave notes that one bit of info could be snuk in as a datatype of the literal

15:56:40 * danbri wishes they weren't encouraging escaped content anyway... is that a fixture, bitsko?

15:57:48 <danbri>http://www.ecademy.com/node.php?id=9875

15:57:48 <dc_rdfig> J: http://www.ecademy.com/node.php?id=9875 from danbri

15:57:58 <danbri> J:|"RDF is the new sex"

15:57:59 <dc_rdfig> Titled item J.

15:58:03 <danbri> J:lol

15:58:04 <dc_rdfig> Added comment J1.

15:58:07 <DanC-AIM> The echo project is designing its process as well as its technology, right?

15:58:13 <bitsko> escaped content is principally for use for already extent weblog archives and for transition of current weblog software that does not tidy user input. it is not recommended for future development use.

15:58:16 <danbri> yup. in a wiki.

15:58:40 <danbri> one concern i had was issue churn: the wiki wasn't leaving a clear enough papertrail for stuff that'd already been decided

15:58:49 <bitsko> more clearly, it is recommended against for future development use.

15:58:52 <danbri> ok thats good, bitsko :)

15:59:29 <danbri> even if not escaped, we'd still need to parseType="Literal" it in RDF/XML syntax, and as a literal it can't directly have properties, so any properties would need to be of something else.

15:59:39 <DanC-AIM> Has anybody raised the nasty copyright/trademark/patent questions? Or how to manage churn? How to decide when they're done? Or schedules?

15:59:50 <bitsko> yes. we have a real papertrail problem. it needs a good, simple solution that isn't obvious, nor have I really seen a good example elsewhere, except in versioned documents or codebases

16:00:19 * danbri feels on the edge of saying 'we' rather than 'they' with Echo, but not until rss-dev wg decides to throw in its lot with Echo

16:02:02 <bitsko> the specs will be published most likely with a current organization (IETF, W3C, or OASIS, whoever steps up as the easiest to accept the documents). "done" will be the milestone of publishing the first versions of the specs.

16:02:16 <DanC-AIM> But so far it's a high-trust process, I gather.

16:02:32 <danbri> yes

16:02:39 <bitsko> yes

16:02:43 <Mutiny> why was everybody hatin' on RSS 1.0 again? because it was called 'RSS?'

16:03:03 <bitsko> Mutiny: some

16:03:04 <danbri> yes, big fights over name, control etc. tired everyone out.

16:03:17 <danbri> also some didn't like the rdf. and others used that as a stick in the naming/control fight.

16:03:23 <Mutiny> rename RSS 1.0, call it a day? ;)

16:03:24 <danbri> ugly exhausting mess.

16:03:50 <bitsko> Copyright of the specs will be given over to the org they're submitted to, most likely

16:03:52 <danbri> "Echo 2.0"

16:03:54 * danbri ducks

16:03:54 <DanC-AIM> Freese: tm4j worked for initial prototype, but I ran into scaling issues. Out of memory.

16:04:46 <bitsko> volunteers have stepped forward to clear the project name itself, as soon as some realistic candidates come up

16:04:57 <dajobe> lol

16:06:05 <DanC-AIM> Freese: ran into topic name clashes... Had to switch to display names.

16:07:07 <Mutiny> is anybody jaded with the "echo" thing already, too?

16:07:23 <DanC-AIM> Freese: I need to use paragraph and list markup in my topicmaps. Topicmap DTD doesn't allow that. Hope the WG fixes that. [good luck! RDF has spent several man-years on it and isn't done!]

16:07:30 <bitsko> the schedule is like open source, it's done when it's done. the move to post snapshots, six for the API and two for the format now, are really good

16:08:09 <danbri> I'm intrigued by Echo. Both technology wise (API stuff interesting) and process, reminds me of our experiments here in lighter-than-trad-WG collaboration (query, calendar; wiki, irc, ...).

16:08:15 <DanC-AIM> Good large open source projects have schedules.

16:08:43 <danbri> Does Debian have a schedule now?

16:08:49 <dajobe> the API stuff looks good, because it's written up in a doc that not everyone is editing

16:09:45 <DanC-AIM> Debian establishes schedules when a release looks near...

16:10:05 * danbri nods

16:10:10 <bitsko> yes, it's intentional that discussion is solidified into edited specifications.

16:10:24 <DanC-AIM> Folks agitate for more schedules in the debian project, but consensus hasn't emerged.

16:11:23 <DanC-AIM> Frese mentioned tolog... He's the 3rd or 4th to do so. Do rdf query and tolog cross-pollenate?

16:11:45 <libby> not sure about tolog in particular

16:12:10 <libby> talking to Ann Wrightson, she's keen for some TMQL thing to interoperate w RDF query

16:12:12 <danbri> I'm also interested in Echo process because FOAF has gone crazy in last few months, so am looking for lightweight process ideas to test out there.

16:12:39 <libby> as far as I know topic maps qls are at the usecases/feature gathering stage, although there are various implementations

16:16:04 <DanC-AIM> /me noodles on the cost of stable publishing using dns and http... On the possibility of using freenet or at least bittorrent so that the consumers fund the publication naturally

16:17:43 <bitsko> as far as the wiki "process", I'm a long time wiki user and believer in its ability to coalesce discussion among other long time wiki users. it's new users, particularly a swell of new users, that mucks things up for a while.

16:20:10 <DanC-AIM> Yes, wiki is a good facilitator, when there's a community that does agree on some stuff. But it's not magic; if people don't agree, I wonder how to tell.

16:20:37 <bijan> Can anyone tell my why rdf test case: statement-entailment/test002 isn't bogus

16:20:58 <bijan> Given this langauge from RDF Semantics:

16:21:02 <bijan> """A reification of a triple does not entail the triple, and is not entailed by it. (The reason for first is clear, since the reification only asserts that the triple token exists, not that it is true. The second non-entailment is a consequence of the fact that asserting a triple does not automatically assert that any triple tokens exist in the universe being described by the triple. For example, the triple might be part of an ontology describing a

16:21:20 <bijan> Premises: http://www.w3.org/2000/10/rdf-tests/rdfcore/statement-entailment/test002a.nt

16:21:39 <bijan> Conclusion: http://www.w3.org/2000/10/rdf-tests/rdfcore/statement-entailment/test002b.nt

16:21:51 <bitsko> if two long-time wiki users disagree yet are not malicious, then the idea is that they document their disagreement and continue to pick away at until it's as clear and concise as possible

16:22:43 <bitsko> "thread mode", by comparison, is evil

16:22:50 <bijan> This case seems to require that a reification is entailed by any triple

16:22:51 <bijan> heya ken

16:23:20 <bitsko> hey bijan

16:23:20 <bijan> D'oh!

16:23:23 <bijan> These are negative entailment tests

16:23:23 <bijan> Duh, that' clears it all up

16:23:23 <bitsko> the pie wiki is overwhelmed with thread-mode discussion

16:23:23 <bijan> Thank you bijan

16:23:27 <bijan> you're welcome bijan

16:23:37 <DanC-AIM> Well, bijan, it's perhaps counter-intuitive, but it limits the extent to which rdf stuff has to know anything about reification. Only parsers need know.

16:23:39 * bitsko bijan: :)

16:23:45 <bijan> No, really, your reply was fast, courtrous and effetive

16:24:07 <bijan> No dan

16:24:13 <bijan> I though it was a pos entailment test

16:24:21 <bijan> Which would contradict the Semantics docuemnt

16:24:24 <bijan> But i's a negative one :)

16:24:31 <bijan> Which tests the assertion of the semantics document

16:27:36 <bijan> Sorry if I derailed a converstation. I wonder if there's a easy format tweak to make this fact more obvious

16:32:22 <DanC-AIM> Not at all.

16:32:31 <DanC-AIM> /me -> lunch.

16:33:44 <danbri> danc, your /me's are showing up as: <DanC-AIM> /me -> lunch.

16:33:46 <danbri> (intended?)

16:35:34 <deltab> could be due to the "gizmo" he's using

16:36:33 * danbri expected so

17:15:53 <DanC-AIM> byebye

17:50:17 <danja> danbri - re. owl:disjointWith in FOAF schema

17:50:32 <danja> shouldn't the range & domain be owl:Class rather than rdfs:Class

17:50:51 <danja> ?

17:51:14 <dmwaters> {global notice} Hi all, we seem to be having problems with one of our main rotation servers, the server is being pulled from rotation now. any further messages will be given in wallops

17:52:47 <danja> maybe I missed something somewhere - doesn't sound feasible reengineering all the classes

17:53:36 <danja> foaf:Person rdf:type owl:Class

17:53:44 <danja> etc

17:54:27 <danbri> yeah, thats a pain.

17:54:35 <danbri> I will stick with rdfs:Class

17:54:50 <danja> is the disjoint still valid though?

17:54:53 <danbri> JimH noticed I'd used owl:Class in some examples

17:55:34 <danbri> yeah, I don't believe OWL WG claim that rdfs:Class and owl:Class are disjoint. They couldn't use owl:disjointWith to express that anyways, without a contradiction :)

17:56:19 <danja> heh, twisty

18:01:14 <danja> must have been the same issue with daml:Class I guess

18:01:58 * GNUPredator lurks

18:02:10 <GNUPredator> Hey dajobe, its me mdupont

18:02:20 * bijan not sure what the daml:Class isue is :)

18:02:45 <bijan> daml:Class and rdfs:Class had more different semantics than rdfs:Class and owl:Class, I think.

18:03:59 <danja> but what about saying things like sameClassAs when talking about rdfs classes?

18:04:52 * bijan not been recently thinking about these nuances, but...

18:06:11 <GNUPredator> GNUPredator is now known as GNUPredator-afk

18:06:53 <bijan> I think...

18:06:56 <bijan> Probably

18:06:59 <bijan> That that's legal owl full

18:07:14 <bijan> Given 5.1 of S&AS

18:07:51 <danbri> that using rdfs:Class is OK, so long as OWL Full? cool, had assumed so.

18:07:58 <bijan> Not sure

18:08:09 <bijan> Since there are possible syntactic restrictions as well.

18:08:12 <bijan> But that's my guess

18:08:12 <danbri> DL folk can strip out the stuff they don't like and make a translation, shouldn't be hard...

18:08:31 <bijan> owl:Class's extension is always a subset of the rdfs:Class extension

18:08:38 <bijan> And in owl full, they are identical

18:12:47 * danja remembers he's forgotten the need for owl:Class anyway

18:13:29 <bijan> In DAML and rdfs:Class of the time, rdfs:Class forbad subClass cycles

18:13:44 <danja> ok...

18:13:49 <bijan> Whereas they are extremely convenient, especially for defining sameClassAS

18:14:16 <danja> (except in non-tabling Prolog)

18:14:36 <danbri> we fixed that 2 yrs ago though, are there other reasons for still having 2? I guess so...

18:14:41 <bijan> In OWL

18:14:43 <bijan> Yes

18:14:45 <bijan> there is

18:15:47 <bijan> And it's not a problem in non-tabling Prolog if you're careful :)

18:15:55 <bijan> have to do loop detection no matter what

18:15:58 <bijan> In OWL

18:15:58 <sandro> Is it that owl:Class does not include any owl:Classes, but rdfs:Class does?

18:16:20 <danja> bijan - remind me to ask you about prolog later

18:16:45 <sandro> right, sorry.

18:16:52 <bijan> owl:Class is not always identifical to rdfs:Class (in extention)

18:16:54 <bijan> Sure

18:17:29 <bijan> I'm not sure the name separation strongly helps

18:17:59 * danja pondering identifical in extention

18:18:00 <bijan> If RDF core had embedded the distinction in RDFS, the separation might never have happened

18:18:11 <bijan> See S&AS 5.1

18:18:22 <bijan> In DL the extension of owl:Class is a subset of rdfs:Class

18:18:26 <danja> ta

18:18:36 <bijan> In OWL full, indeed, part of what makes it full, is that they're identifical

18:22:14 <bijan> There was a nice presentation by pat hayes at the tech plenary in march

18:25:38 <bijan> Slides: http://lists.w3.org/Archives/Public/www-archive/2003Mar/0033.html

18:25:38 * danja only groks AS&S on full moons, leap centuries...

18:25:51 <danja> ooh - thankss

18:26:03 <bijan> Lbase: http://lists.w3.org/Archives/Public/www-archive/2003Mar/att-0018/LBASE-new.html

18:26:21 <bijan> PFPS's persentation on layering: http://www-db-out.bell-labs.com/user/pfps/talks/semantic-layering/Overview.html

18:28:58 <bijan> Some early discussion on this: http://lists.w3.org/Archives/Public/www-webont-wg/2002Aug/0139.html

18:30:14 <bijan> Hmm. This thread is leading me to think that having rdfs:Class in a disjoint etc is forbidden

18:31:34 <bijan> Aha!

18:31:39 <danja> ?

18:31:57 <bijan> """The basic distinction is that DAML classes (i.e., members of daml:Class) can only have resources as their instances (i.e., not literals)."

18:32:13 <bijan> Though, it seems, that they don't, now, mind have a restricted and extended owl:Class

18:32:22 <bijan> See: http://lists.w3.org/Archives/Public/www-webont-wg/2002Aug/0116.html

18:33:14 * danja admires bijan's search prowess

18:33:31 * bijan bows

18:34:19 <bijan> IRC discussion: http://www.w3.org/2003/06/05-webont-irc

18:34:55 <bijan> If owl:Class were removed...

18:34:57 <bijan> """PatH - then OWL DL willl have theorems that are false in RDF "

18:35:05 <DanC-AIM> -- Bill Kent: "the unsolvable identity problem"

18:36:02 <danja> ah

18:36:12 <bijan> danja: danc would be able to speak more specifically

18:36:21 <bijan> Having actually participated in the discussions

18:36:59 <sandro> Is DanC just reporting on the interesting talks, or are they all like that?

18:37:12 <bijan> More recent debate: http://lists.w3.org/Archives/Public/www-webont-wg/2003Jun/0151.html

18:37:34 <bijan> Indeed, jim is raising precisely the thing bugging me at the moment.

18:37:45 <bijan> If I were *actually* diligent, I'd have chumped all this.

18:37:49 <bijan> logger, pointer

18:37:56 <DanC-AIM> All like what?

18:38:09 <bijan> logger_1, pointer

18:38:09 <bijan> See http://ilrt.org/discovery/chatlogs/rdfig/2003-08-06#T18-38-09

18:38:17 <bijan>http://ilrt.org/discovery/chatlogs/rdfig/2003-08-06#T18-38-09

18:38:18 <dc_rdfig> K: http://ilrt.org/discovery/chatlogs/rdfig/2003-08-06#T18-38-09 from bijan

18:38:24 <DanC-AIM> There's a spectrum. I remark on what occurs to me as remarkable

18:38:36 <bijan> K:|Buncha pointers plus discussion about rdfs:Class vs. owl:Class

18:38:37 <dc_rdfig> Titled item K.

18:38:49 <bijan> K:This is me exercising not quite due dilligence.

18:38:49 <dc_rdfig> Added comment K1.

18:39:28 <DanC-AIM> "some talks were informative. Others reinforced my sense of confusion" - Kent

18:39:50 <DanC-AIM> K:ooh... Nifty.

18:39:51 <dc_rdfig> Added comment K2.

18:40:28 <bijan> Oops, that was Jonathan, not jim

18:40:37 <DanC-AIM> /me wishes for a similar item on the literal foo... Esw:TextValues

18:43:10 <DanC-AIM> "one thing I like to do, as a kind of hobby, is talk to various tech support people"

18:43:12 <bijan> danja: that last thread is very insight breeding, i think

18:43:43 * bijan thinks that DanC-AIM has been possessed by a beat poet

18:48:36 <danja> plenty of reading - many thanks #rdfig-collective

18:51:33 <DanC-AIM> This Bill Kent guy that I'm quoting could pass as a beat poet, I suppose, though he introduced himself as something of a standards wonk, conceptual modelling guy, etc.

18:51:54 <DanC-AIM> He has long hair, though. :)

18:52:07 <danja> where did you find him?

18:53:13 * danja remembers wife due home, expecting to be fed...

18:53:31 <DanC-AIM> Here in Montreal... At extreme markup.

18:59:01 <danbri>http://weblog.infoworld.com/udell/2003/08/05.html#a716

18:59:02 <dc_rdfig> L: http://weblog.infoworld.com/udell/2003/08/05.html#a716 from danbri

18:59:24 <danbri> L:|Using RSS 2.0 and RDF together, Jon Udell

18:59:25 <dc_rdfig> Titled item L.

18:59:52 <danbri> L:Follows up the discussion I started in the RSS2 site.

18:59:52 <dc_rdfig> Added comment L1.

18:59:53 * danbri reads

19:02:07 <DanC-AIM> Ooh... Rdf is on udell's radar?

19:02:22 <danbri> yup :)

19:02:44 <danbri> he's one of the committee of three looking after Winer's RSS2 for Harvard...

19:04:09 <DanC-AIM> Committe of 3? Whence comes that tidbit?

19:04:32 <danbri> there was a press release

19:04:36 * danbri digs it out

19:05:15 <danbri>http://blogs.law.harvard.edu/tech/announceRss2

19:05:15 <dc_rdfig> M: http://blogs.law.harvard.edu/tech/announceRss2 from danbri

19:05:23 <danbri> M:|RSS 2.0 Specification moves to Berkman

19:05:24 <dc_rdfig> Titled item M.

19:05:34 <DanC-AIM> Thx

19:06:30 <danbri> M:The "[http://blogs.law.harvard.edu/tech/advisoryBoard|RSS advisory board]" is currently Dave Winer, Jon Udell, Brent Simmons.

19:06:30 <dc_rdfig> Added comment M1.

19:06:59 <danbri> excellent, Jon buys the multi-namespace story: I'd like to surface those issues, in the spirit of better understanding how to achieve the "Web of complementary namespaces" that Dan rightly envisions

19:26:26 <ChanServ> [#rdfig] This channel is logged and blogged: http://logicerror.com/rdfIRCWelcome

19:27:00 <lilo> [Global Notice] Hi all. We just experienced a very interesting problem with services and an EU hub. We're investigating, and suspect it won't recur anytime soon.

19:33:36 <lilo> [Global Notice] Hi all. Please bear with us, we *may* have a major dehubbing and rehubbing in just a sec.

19:34:33 <lilo> [Global Notice] Looks like not. Thanks for your patience.

19:38:09 <bill|rebooting> bill|rebooting is now known as wkearney99

19:46:19 <wkearney99> any calendar rdf folks about here?

19:47:03 <DanC-AIM> Yo

19:47:25 <wkearney99> is there a vocabulary you know of that handles 'types' of events?

19:48:04 <wkearney99> as in being able to rdf:resource="" for various sorts of event happenings. in this case for political campaigns.

19:48:41 <wkearney99> as in 'appearance', 'fundraiser', 'backroomdealswithcorporatemasters', etc...

19:51:24 <danbri> I don't know of a good event taxonomy.

19:51:58 <danbri> I have a fantasy of making one from a list of all the verbs, since verb "occurances" are happenings...

19:52:10 <danbri> probably not quite what you're after though

19:52:41 <wkearney99> I'm thinking of using this as a 'force the issue' example as to why using RDF (and triple retrieval) would be a good idea on newsfeeds.

19:53:28 <wkearney99> as in, have a core vocab that covers a reasonable range of events, backed up by a decent schema that explains the relation of the elements to those in other namespaces.

19:58:08 <DanC-AIM> I'd look at the wordnet/sumo mapping stuff. And/or cyc.

19:58:49 <DanC-AIM> But no, there isn't a slam dunk ready-to-go namespace of event types including fundraiser.

19:59:14 <wkearney99> any others? what about one for public relations?

19:59:57 <DanC-AIM> Maybe prism has some relevant terms?

20:00:23 <DanC-AIM> Look in the daml library; you never know.

20:02:17 * sandro vaguely remembers something called X-Talks, which was a DAML thing about publicizing public events.

20:07:48 <sandro> X-Talks maybe == ITTalks. http://daml.umbc.edu/ontologies/talk-ont.daml http://daml.umbc.edu/

20:14:48 * wkearney99 looks it over...

20:16:04 <wkearney99> any idea who was coordinating that vocab?

20:16:47 <DanC-AIM> Tim Finin

20:16:59 <wkearney99> ah, at UMBC...

20:17:17 <DanC-AIM> Yup

20:17:59 <DanC-AIM> W3C xml schema has evidently grown a TLA: WXS.

20:18:12 <sandro> Huh!

20:18:27 <danbri> That's xml.com/ORA's name for it

20:19:06 <danbri> maybe it serves us right for picking the most generic name when we knew there'd be competition, i dunno

20:19:46 <DanC-AIM> Sandro, who has the tuesday chairing ball? Is it in the air?

20:20:07 <DanC-AIM> I'm quite happy that webont picked OWL

20:21:15 <DanC-AIM> Naming is a terribly important part of the process that's totally chaotic at w3c

20:21:47 <sandro> I don't know about tuesday, Dan.

20:22:44 <DanC-AIM> -- Fuchs on an XML Schema algorithm

20:25:51 <DanC-AIM> Fuchs defines, in addition to the traditional first(A) and follow(A), a new confusion(A)

20:27:10 <DanC-AIM> I wonder how long our telcon reservation lasts, sandro

20:27:26 <sandro> you mean like how many years, Dan?

20:28:34 <DanC-AIM> No, I think telcon reservations last about 3 months

20:28:45 <DanC-AIM> I mean in the database

20:29:38 <DanC-AIM> Em is the reservation contact.

20:30:02 <sandro> huh. *shrug*

20:32:14 <sbp-> sbp- is now known as sbp

20:36:25 <DanC-AIM> I see your msg, but I can't reply outside the channel

20:38:06 <DanC-AIM> RFE for chatbridge: private message support

20:39:14 <sandro> Ah.

20:39:50 <DanC-AIM> Hmm what news from the july swad-eu meeting, I wonder...

20:40:59 <DanC-AIM> /me surfs...

20:42:15 <GNUPredator-afk> GNUPredator-afk is now known as GNUPredator

20:49:21 <DanC-AIM> I lose. Checked esw list wiki, meeting agenda page. Can't find any records of what happened. Bummer.

23:17:08 <_joshua> DanC-AIM: The AIM gateway I gave you can do personal messages

23:36:20 <dje> dje is now known as lackadaisical

23:37:26 <lackadaisical> lackadaisical is now known as dje