Semantic Web Interest Group IRC Chat Logs for 2003-02-27

00:00:19 <timbl123> { ?F a :crawled; log:includes { [] :p [] }. ?p.log:racine log:semantics ?G} => { ?G a :crawled }.

00:00:34 <timbl123> this log:foAll :p

00:01:10 <timbl123> The validator does a form of crawling -- swap/util/validate.n3

00:01:26 <DanC> where's your foafwho file, danbri?

00:01:52 <timbl123> use --chatty=25

00:02:19 <danbri> see http://rdfweb.org/people/danbri/rdfweb/webwho.xrdf

00:03:02 <DanC> $ python cwm.py --chatty=25 test/crawlSeeAlso.n3 --rdf http://rdfweb.org/people/danbri/rdfweb/webwho.xrdf --think

00:03:12 <DanC> xml.sax._exceptions.SAXParseException: <unknown>:22:76: not well-formed (invalid token)

00:03:29 <DanC> hmm... semanticsOrError to the rescue...

00:04:22 * DanC borrowing from family time...

00:04:24 <danbri> or http://rdfweb.org/people/danbri/rdfweb/danbri-foaf.rdf

00:05:08 <DanC> Reading and parsing scutterplan.rdf

00:05:14 <DanC> xml.sax._exceptions.SAXParseException: <unknown>:22:76: not well-formed (invalid token)

00:05:41 <DanC> semanticsOrError doesn't seem to catch xml errors

00:07:23 * DanC checks in swap/test/crawlSeeAlso.n3 , heads out

00:07:24 <timbl123> Oops. It only catches BadSyntax and IOError IIRC. Adding xml.sax._exceptions.SAXParseException should be easy

00:07:51 <timbl123> It is tempting but probbly unwise to ctache everything.

00:07:55 <timbl123> catch

00:08:08 <DanC> yes. don't mask bugs

00:08:55 <danbri> are lots of the foaf files bad rdf/xml?

00:10:54 <danbri> fwiw once you've got one of the scutterplan RDF files, you should be able to find most all the rest of the known foaf docs out there...

00:11:37 <mdupont> mdupont is now known as mdupont-zzz

00:11:55 <mdupont-zzz> goodnight

00:19:03 <danbri> [[

00:19:03 <danbri> danbri@fireball:~/cvs/w3.org/WWW/2000/10/swap$ ./cwm.py

00:19:03 <danbri> Traceback (most recent call last):

00:19:03 <danbri> File "./cwm.py", line 35, in ?

00:19:03 <danbri> from why import FormulaReason

00:19:04 <danbri> File "why.py", line 25, in ?

00:19:06 <danbri> from thing import Namespace

00:19:08 <danbri> File "./thing.py", line 129, in ?

00:19:10 <danbri> if map(int, version) < [2, 2, 0]:

00:19:12 <danbri> ValueError: invalid literal for int(): 3a0

00:19:14 <danbri> ]]

00:19:16 <danbri> freshly cvs'd Cwm

00:19:21 <danbri> python -V

00:19:22 <danbri> Python 2.3a0

00:23:15 <reagle> oh, running 2.3 eh, the bleeding edge :) (actually the generators are rather cool...)

00:23:44 <danbri> I forget why I upgraded, almost certainly cwm-troubles!

00:23:52 * danbri commented out the check, for now

00:24:53 <danbri> crap! I typed 'make tested' and it seems Cwm has auto-committed itself...

00:25:09 <danbri> i only wanted to run the tests...

00:25:38 <danbri> [[

00:25:39 <danbri> danbri@fireball:~/cvs/w3.org/WWW/2000/10/swap$ make tested

00:25:39 <danbri> python cwm.py crypto.n3 --rdf > crypto.rdf

00:25:39 <danbri> cvs commit -m "Automatic: see Makefile" crypto.rdf

00:25:39 <danbri> Warning: Remote host denied X11 forwarding.

00:25:40 <danbri> Checking in crypto.rdf;

00:25:43 <danbri> new revision: 1.3; previous revision: 1.2

00:25:45 <danbri> done

00:25:47 <danbri> ]]

00:47:10 <DanC> that auto-commit is evil.

00:47:16 <DanC> or wierd, at least

00:48:03 <mortenf> night, thanks all.

00:51:06 <DanC> how about fixing the scutterplan, danbri?

00:51:13 <DanC> s/&/&amp;/g

00:51:20 <danbri> what's the url?

00:51:31 <DanC> . http://swordfish.rdfweb.org/discovery/2001/08/codepict/scutterplan.rdf

00:52:10 <danbri> libby's box, I could probably blunder in and change it but i think the file is written from script. I'll talk to her tommorrow...

00:53:27 <DanC> ok, I'll make cwm more defensive...

00:54:08 <danbri>http://www.perceive.net/xml/googleScutterNoChatlogs.rdf

00:54:08 <dc_rdfig> A: http://www.perceive.net/xml/googleScutterNoChatlogs.rdf from danbri

00:54:34 <danbri> A:|google'd list of RDF start points for FOAF harvesting

00:54:34 <dc_rdfig> Titled item A.

00:56:15 <grault> grault is now known as earle

00:57:00 <DanC> ok, cwm survives the sax error...

00:57:10 <DanC> wow! it finished

00:57:20 <earle> wow, I'm fifth on the list. benefits of leaping wholeheartedly onto the bandwagon.

00:59:29 <DanC> so I've got the results in ,aboutDanBri.rdf

00:59:38 <DanC> it's xmlwf happy.

00:59:50 <DanC> is that sorta what you meant by a crawler, danbri?

01:00:26 <DanC> <web:Description>

01:00:27 <DanC> <web:type web:resource="http://xmlns.com/wordnet/1.6/Person"/>

01:00:27 <DanC> <s:seeAlso web:resource="http://rdfweb.org/people/danbri/rdfweb/danbri-biblio.rdf"/>

01:00:27 <DanC> ...

01:00:37 <DanC> <foaf:dateOfBirth>1972-01-09</foaf:dateOfBirth>

01:00:45 <DanC> <foaf:mbox web:resource="mailto:danbri@w3.org"/>

01:00:54 <DanC> ...

01:00:56 <DanC> <foaf:depiction web:resource="http://rdfweb.org/people/danbri/2000/01/01/Image1.gif"/>

01:00:58 <DanC> ...

01:01:01 <danbri> I have a crawler that does v little except, um, crawl. It reads an rdf doc, parses it for seeAlsos, and moves onto the next one, maybe outputting a bit of info found in each file (names etc). A more traditional crawler might add the parsed info to a persistent store (SQL db or whatever)

01:01:11 <danbri> that looks like me, yup

01:01:21 <DanC> <foaf:workplaceHomepage web:resource="http://www.w3.org/"/>

01:01:52 <DanC> hmm... no foaf:knows stuff in there

01:02:02 <DanC> gee, isn't that sorta the point?

01:02:23 <danbri> to hang onto the data once you've gotten it? i guess so...

01:02:37 <DanC> no, to say who your friends are

01:03:29 <danbri> I prefer doing that implicitly through photo metadata etc, rather than asserting it flatly... collaborations etc too

01:03:45 <danbri> someone asked me what the point was on fork recently, so i made up an answer...

01:04:22 * danbri rummages

01:04:44 <danbri>http://www.xent.com/pipermail/fork/2003-February/017708.html

01:04:44 <dc_rdfig> B: http://www.xent.com/pipermail/fork/2003-February/017708.html from danbri

01:04:51 <danbri> B:|FoRK thread on FOAF

01:04:51 <dc_rdfig> Titled item B.

01:05:29 * danbri wonders why no 'knows' stuff showed up

01:05:38 <danbri> danc, what do i do to replicate what you just did?

01:06:39 <DanC> lemme check in my patch to make semanticsOrError survive...

01:06:45 <danbri> ok

01:07:21 <DanC> llyn.py,v 1.72 2003/02/27 01:07:08

01:07:56 <danbri> btw did you see my 'make tested' scare earlier? i wasn't expecting it to commit anything :(

01:08:04 <DanC> $ python cwm.py --chatty=25 test/crawlSeeAlso.n3 --rdf http://rdfweb.org/people/danbri/rdfweb/webwho.xrdf --think >,aboutDanBri.rdf

01:08:23 <DanC> yes, scary

01:08:43 <danbri> should i be worrying about backing out changes?

01:09:11 <DanC> you only committed crypto.rdf, yes?

01:09:15 <danbri> yup

01:09:25 * DanC doesn't see how 'make tested' would invoke cvs commit

01:10:25 <DanC> wow!

01:10:27 <DanC> .n3.rdf:

01:10:27 <danbri> via [[.n3.rdf:

01:10:27 <danbri> $(PYTHON) cwm.py $< --rdf > $@

01:10:27 <danbri> cvs commit -m "Automatic: see Makefile" $@

01:10:28 <DanC> $(PYTHON) cwm.py $< --rdf > $@

01:10:28 <DanC> cvs commit -m "Automatic: see Makefile" $@

01:10:28 <danbri> ]]

01:10:35 <danbri> yeah

01:11:04 <DanC> timbl made the mess; he can clean it up

01:11:10 * danbri nods

01:12:04 * danbri picks another starting point from http://rdfweb.org/rweb/wiki/wiki?FOAFBulletinBoard

01:12:43 <danbri> is the output the aggregate of all the triples encountered?

01:13:06 <DanC> yes... tweak to cmdline:

01:13:12 <DanC> $ python cwm.py --chatty=25 test/crawlSeeAlso.n3 --rdf http://rdfweb.org/people/danbri/rdfweb/webwho.xrdf --think --purge-rules >,aboutDanBri.rdf

01:13:18 <DanC> i.e. added --purge-rules

01:13:36 <danbri> what does that do?

01:13:57 <DanC> throws away the rules so the .rdf isn't screwed up

01:14:11 <DanC> code for serializing rules in .rdf (i.e XML) is hosed

01:14:31 * danbri notices that Edd has a seeAlso to an encrypted rdf file (http://heddley.com/edd/foaf.rdf)

01:15:24 * DanC tries starting at http://eikeon.com/foaf.rdf

01:15:25 <AaronSw> for foafbot, presumably

01:15:33 <DanC> this is kinda fun!

01:15:59 <danbri> I really enjoy the fact that there's rdf actually in the web, even though there isn't much yet, and we're not quite sure why we're doing it... ;)

01:16:57 <DanC> 7551 xmlwf-happy bytes result from crawling from eikon

01:17:13 * danbri trying from Edd's, it seems to hang with 'Reading and parsing foaf.rdf' oh no, as i type it came back to life.

01:18:08 <danbri> <!-- using base file:/home/danbri/cvs/w3.org/WWW/2000/10/swap/--># next char: '-'

01:18:08 <danbri> # Warning: ID=author on statement ignored

01:18:13 <DanC> danbri, you realize we can make this a smushing-crawler almost for free, right?

01:18:15 <danbri> ...in the XML output

01:18:20 <danbri> right :)

01:18:31 <danbri> I noticed the smush() function in diff.py

01:18:49 <DanC> it's easier than that. remember sameDan.n3?

01:18:51 <danbri> and now Cwm has SQL smarts, there's somewhere we could stash all this data...

01:18:53 <danbri> vaguely...

01:19:22 <DanC> look at test/sameThing.n3

01:19:24 <danbri> problem is we need to smush across the entire harvested dataset, probably more than we want loaded into one cwm instance

01:19:44 <DanC> :t1049 a test:CwmTest;

01:19:44 <DanC> test:shortFileName "sameDan.n3";

01:19:44 <DanC> test:description "dealing with multiple descriptions of the same thing using log:lessThan, log:uri, daml:equivalentTo";

01:19:44 <DanC> test:arguments """sameDan.n3 sameThing.n3 --think --apply=forgetDups.n3 --purge""".

01:19:56 <DanC> (from the regression suite)

01:20:36 <DanC> I thought the sun would burn out doing sameThing thinking, but I was pleasantly surprised.

01:20:57 * eikeon waves... and catches up with logs.

01:21:05 * danbri turns off --chatty, gets wf XML output from Cwm again

01:21:22 <DanC> it can, for example, take the groups.rdf file of all the W3C groups/chairs/activities/staff contacts and smush on contact:mailbox etc in a few minutes

01:22:01 <danbri> (aside: wonder how Euler performs on that...)

01:22:39 <DanC> probably about 200x better. er... hmm.. not sure how to purge in euler...

01:22:53 <DanC> euler is good if you know what you're looking for

01:23:29 <danbri> so looking at the output from a few of these test runs, i've not noticed any of my test runs outputting an RDF result thats noticably larger than the initial RDF doc I started it at...

01:24:05 <DanC> yeah, I thought I'd see more in the output too...

01:24:53 <DanC> oh! there's nothing to get the ?WHAT log:includes { ?S ?P ?O }. rule to grab onto

01:25:51 <danbri> yes, ,scuttered.rdf has 77 triples, vs 68 in eikeon's foaf file.

01:26:09 <danbri> and some of those are forAlls

01:27:00 <DanC> er... no...

01:28:57 <DanC> ?S doesn't work in {} or something...

01:29:22 <danbri> ?S being shorthand for wrapping in forAll...?

01:31:37 <DanC> yes... but it interacts strangely with levels of {}

01:32:37 * danbri feeling better this evening but thinks its bedtime nevertheless... better go! this is fun though, hope you can fix this, is too Cwm voodoo for me to help with...

01:33:23 <eikeon> <DanC> just predeclare xml: to http://www.w3.org/XML/1998/namespace as a work-around, timbl

01:33:53 <eikeon> I have had to do this to... and submitted a bug report / patch. But am not sure it was acted on ;(

01:34:43 <DanC> night, danbri

01:34:46 <DanC> get some rest

01:35:09 * danbri waves, heads off

01:36:06 * DanC gets lots more now...

01:36:19 <DanC> Grand total of 37 new statements in 2 iterations.

01:36:20 <eikeon> timbl: [[[parser = make_parser()

01:36:20 <eikeon> # Workaround for bug in expatreader.py. Needed when

01:36:20 <eikeon> # expatreader is trying to guess a prefix.

01:36:20 <eikeon> parser.start_namespace_decl("xml", "http://www.w3.org/XML/1998/namespace")

01:36:20 <eikeon> parser.setFeature(handler.feature_namespaces, 1)]]]

01:37:14 * DanC finds another bug...

01:39:23 * DanC waits much longer for cwm this time...

01:40:56 <DanC> seems hung at

01:40:58 <DanC> Reading and parsing foaf.rdf

01:41:02 * DanC wanders off...

01:41:07 <DanC> unstuck...

01:46:57 * DanC watches cwm go totally nutso learning about all these people...

01:47:16 <DanC> Concluding definitively P->foaf:knows S->_g637 O->_g649 _g7->0_work

01:47:16 <DanC>

01:47:16 <DanC> Concluding definitively P->foaf:knows S->_g637 O->_g650 _g7->0_work

01:47:16 <DanC>

01:47:16 <DanC> Concluding definitively P->foaf:knows S->_g637 O->_g651 _g7->0_work

01:47:17 <DanC>

01:47:19 <DanC> Concluding definitively P->foaf:knows S->_g637 O->_g652 _g7->0_work

01:47:20 * danbri wanders past, grins

01:47:20 <DanC>

01:47:22 <DanC> Concluding definitively P->

01:53:35 <DanC> it's still going nutso, learning all this stuff...

01:53:56 <DanC> luding definitively P->foaf:firstName S->_g1965 O->"Carrie" _g7->0_work

01:53:56 <DanC> Concluding definitively P->foaf:firstName S->_g1966 O->"Michael" _g7->0_work

01:54:45 * DanC wanders off...

02:09:20 <jlinden> jlinden is now known as jlinden_away

02:09:51 <jlinden_away> jlinden_away is now known as jlinden

02:17:44 * timbl must install a IRC back end so one can invite a running cwm session to an IRC session.

02:18:16 <timbl> <> | grep timbl | grep mess :-(

02:18:25 * timbl starst to ctach up

02:23:21 <earle> evening timbl.

02:29:18 <timbl> Hmmm. make tested has to make sure the package is up to date. That involves making the RDF/XML equivalents fo a few files. they shouldn't be commited though.

02:30:45 <eikeon> timbl: Not sure if I pointed this out before, but you may be interested in looking at irc.AsyncClient to implement an IRC back end.

02:31:33 <eikeon> I have pulled the version I use into hypercode land: http://redfoot.net/2003/01/01/irc

02:32:07 <timbl> Hyperecode land - you mean red_importable?

02:32:32 <eikeon> Yep... fun eh? :)

02:32:39 <timbl> Definitely.

02:32:53 <eikeon> Also think this is the last CVS we where working on it from: http://cvs.plesh.net/infogami/irc.py

02:33:18 <timbl> Works for stand-alone modules, presumably.

02:33:28 <eikeon> Seems to be fairly stable... it is what I use for eikco (my logging bot... in RDF of course).

02:36:55 <eikeon> Stand-alone? Are you refering to irc.py. If so, yeah can build a stand-alone bot pretty quickly using it and a few lines of code.

02:39:45 <timbl> No, red_import.

02:40:05 <timbl> What happens when you start building up a system out of red_imported modules?

02:41:05 <timbl> IIRC its somethinmg like red_import("http:///dsfasdghdsjhg...dfasd.rdf", as="localModuleName")

02:41:27 <timbl> Presumably if the red_imported module imports anything, it has to red-import it.

02:41:41 <eikeon> Ah yeah. red_import and import do not integrate as tightly as they could. I had an earlier version where I was doing stuff with Python's import hooks etc.

02:43:01 * s1n is back (gone 05:03:20)

02:43:22 <eikeon> A red_imported module can import things using regular import iff it is installed already on your Python path.

02:43:50 <timbl> search paths consiered dangerous.

02:44:00 * eikeon agrees

02:44:39 <eikeon> Would be nice to be able to use import http://... though

02:45:02 <timbl> One needs to combine red_import with a persistence cache.

02:45:14 <timbl> nice: you mean to ambush the actual python install?

02:46:02 <timbl> import <http://www.python.org/version2.2/sys> as sys

02:46:46 <eikeon> Sure... think that is the approach I started to explore/use for Redfoot... standard modules do not get installed... just redfoot.py. Everything else gets pulled into the persistent cache when it first gets used.

02:47:29 <timbl> s/python install/python import/ but I was sthinking of install

02:48:15 <timbl> yes, install(uri) would recusrively load it into the persistent cache and do eth saem for any imported modules.

02:49:29 <timbl> At the moment, import xx will by default work for a module in the same directory. That form would survive.

02:50:38 <timbl> It musthave been at a similar point tht C ended up with the strage differnce betwen #include <sys.h> and #include "sys.h".

02:51:41 <timbl> There is the problem of how you disambiguate "import sys" as being a reference to the standard system libraries or to the local dircetory. That is the main place wheer the search paths are used.

02:52:02 * timbl is noodling aloud, looking for transition paths to webify python

02:52:06 <eikeon> Found my old attemps in the attic: http://redfoot.net/cvs/redfoot/redfootlib/Attic/

02:53:12 <eikeon> Think URLImporter was the second one I did... and http_importer the first.

02:55:06 <eikeon> Ah... they where from before I pushed the py into the rdf.

02:56:36 <eikeon> Wonder why I did not integrate with the import stuff when moving to storing the py in the rdf.

02:57:37 <eikeon> Must have gotten sick of the import stuff... not one of Python's best features.

02:58:22 <eikeon> I think the import functionality got worked on a fair bit moving to 2.3... have not to see where it landed.

02:59:02 * rhizo ponders python in an rdf channel

02:59:14 <rhizo> any particular reason it's python and not eg perl ?

03:00:12 <burtonator> can anyone recommend a THIN Java RDF library with a high level API?

03:00:21 <burtonator> which is x-platform

03:01:42 * timbl123 perl? you need the xml channel for that ;-)

03:02:03 <eikeon> Good answer... lol :)

03:03:05 <rhizo> timbl123: search.cpan.org for RDF :)

03:03:42 <rhizo> RDF::Core, RDF::Notation3, RDF::Service to mention just a few..

03:03:42 <dc_rdfig> Label RDF not found.

03:03:57 <eikeon> Python needs an cpan that is URLs and import <http://> etc.

03:04:08 <eikeon> s/is/uses/

03:04:54 <eikeon> Modules should flow p2p like along with the rest of the data.

03:05:29 <timbl123> Weblike interconnects.

03:05:34 <rhizo> eikeon: checkout CPAN shell sometime you're feeling adventurous ;)

03:07:14 <eikeon> rhizo: Have used it once... was impressed. Now if only Python had a similar system, but using URIs etc.

03:07:44 <zool> cpanplus, RDF/ even FOAF interfaces to search.cpan.org, aiming interconnection not techwars ;)

03:08:11 <rhizo> eikeon: sounds like a fun idea

03:19:47 <eikeon> zool, rhizo: Is there info on how cpan / cpanplus deal with security issues?

03:21:13 <rhizo> eikeon: perldoc <module name> is a good place to start, else check cpan.org and of course, there's always the source :)

03:22:42 <zool> the perl community is so inherently trusting and trustable, is there any need for Kwalitee assurance? ;) or cf http://search.cpan.org/author/KANE/CPANPLUS-0.042/lib/CPANPLUS/Backend.pm but i'm not the best person to ask

03:22:43 <DanC> wow... cwm is _still_ going.

03:22:54 <DanC> Concluding definitively P->foaf:mbox_sha1sum S->Ciaran McCann O->"f98ce916" _g7->0_work

03:23:55 * zool must try out cwm some point ... i am still hanging in anticipation for bijans muttered swi-prolog/OWL code

03:25:38 <DanC> ircAsync is cool... I integrated it with cwm to make swBot

03:26:34 <eikeon> DanC: Yeah... it seems it has become fairly stable. Did you use the version that Aaron and I hacked some more on?

03:27:28 <timbl123> DanC, is swbot in the swap space?

03:27:55 <DanC> no, swbot is in dev.w3.org

03:27:56 * timbl123 has been making his way through homework - the RDF query use cases in N3.

03:28:15 <DanC> .google swboty

03:28:16 <datum> swboty: http://old.cruisingworld.com/boty_2001/BOTYpreview/swboty/

03:28:17 <DanC> .google swbot

03:28:17 <datum> swbot: http://dev.w3.org/cvsweb/2000/scribe-bot/

03:28:38 <DanC> swboy (gm) connolly

03:28:42 <DanC> swbot (gm) connolly

03:28:45 <timbl123> Thanks, datum

03:28:47 <DanC> sigh. ruined the joke

03:29:35 <DanC> (gm) is googlemark... it means the world-according-to-google nominates your thingy as the top hit for the term

03:29:46 <timbl123> Maybe (gm) will become self-reinforcing to be more stable than (tn)

03:29:55 <timbl123> s/tn/tm

03:30:10 <DanC> wow... you don't have googlemark on timbl

03:30:37 <timbl123> (tm) and the trademark icon are tradeamarks of the World Intellectual Property Organization.

03:30:51 <timbl123> .google timbl

03:30:52 <datum> timbl: http://ilk.kub.nl/software.html

03:31:08 <DanC> sombody's got a big foaf database out there somewhere... module.php?mod=network&op=foafrdf&uid=18922

03:31:27 <DanC> cwm is wandering around in it

03:33:16 <earle> .google downlode

03:33:17 <datum> downlode: http://downlode.org/

03:33:52 * timbl123 adds the word "timbl" to hi shome page

03:34:09 * DanC checks in 1.2 2003/02/27 03:33:29 of crawlSeeAlso.n3 ; the one that has kept cwm busy for the last hour or so

03:34:27 * eikeon cheated to get his gm... picked a name that was not in use so that I would not keep nick colliding with other on irc back ~1995.

03:34:40 <earle> holy smokes, I'm 8th for "earle".

03:34:56 <timbl123> 6/10 ... some of these use cases are quite fun. How about you propose the seeAlso crawler as RDFQ use case?

03:37:50 <timbl123> cwm is probably retracing its steps a few times in the crawler.

03:39:42 * eikeon is curious what others thought of having an rdfs:Schema class (re: email I sent to rdf-comments). Seems somewhat ironic that we do not (yet;) have more metadata withing out RDF documents about them.

03:42:46 <timbl123> What is a member of rdf:Schema?

03:42:49 <DanC> yes; I harvested :InterestingProperties from the foaf schema... it seems to do: for each interesting prop: for each seeAlso doc: if doc includes { :S :P :O}, conclude {:S :P :O}

03:43:40 <timbl123> 7/10

03:43:53 <eikeon> timbl: Either the documents that contain the definitions of the Classes and Properties etc... or namespace that the schema uses.

03:44:03 <DanC> how about committing, timbl?

03:44:52 <DanC> { ?X log:racine ?DOC. ?DOC log:semantics [ log:includes { ?X a r:Property } ] } => { ?DOC a :Schema}.

03:45:06 <DanC> { ?X log:racine ?DOC. ?DOC log:semantics [ log:includes { ?X a r:Property } ] } => { ?DOC a :Schema. ?X s:isDefinedBy ?DOC. }.

03:45:26 <DanC> same for ?X a s:Class

03:46:07 * DanC wishes cwm would put the 'for each interesting prop' loop on the inside

03:46:50 <DanC> 100:08.82 cpu so far.

03:46:54 <DanC> but only 4% of memory.

03:47:11 <DanC> 37m RES

03:49:47 <eikeon> Motivation for an rdf:Schema Class (however one is defined) is to elevate rdf schemas past the point where you need google to find them... but can use the SW to find them.

03:50:24 <eikeon> At the moment the notion of an rdf schema simply does not exist ;)

03:50:28 <eikeon> Rather ;(

03:51:30 <DanC> timbl, a nice --chatty thing would be the sizes of llyn's queues.

03:52:19 <DanC> ala... 'concluding P S O. 456 items pending in queue'

03:53:31 <DanC> timbl123? share the 6 or seven you've got done?

03:55:21 <DanC> chuckle... http://logicerror.com/cwmTogether

03:55:44 <eikeon> DanC: Was your n3 above in response to my rambling re: identifying schemas

03:56:10 <DanC> yes, translation of "Either the documents that contain the definitions of the Classes and Properties"

04:00:32 <DanC> what tool does this style of layout again? http://rdfweb.org/foaf/examples/edd-foaf.jpeg

11:23:07 Topic now RDF & SemWeb hack-n-chat http://rdfig.xmlhack.com/

11:23:07 Users on #rdfig: logger_1 libby AndyS xover nmg Mutiny ericP_ hugo grault burtonator s1n Brad`HeliX eikeon timbl-gone2 kham sbp` pfps AaronSw mdupont-work timbl-gone dmiles fknow jordan dc_rdfig Wack shellac SeeTemp rajiv _joshua DanC em MarkB aml TrustBot Paul|trampsoc deltab grove jllykifsh sandro tav danbri soccos|away jang datum space xower

11:23:07 <ChanServ> [#rdfig] This channel is logged and blogged: http://logicerror.com/rdfIRCWelcome

11:23:20 <libby> so, I'm working through my rdfquery faq, and I get to: Why isn't there a W3C working group or taskforce for RDF query?

11:23:25 <libby> anyone know?

11:23:28 <libby> not that I want one

11:26:06 * danbri gets that "I've been in IRC too long" feeling

11:26:07 * danbri grins

11:28:12 <libby> damn you!

11:29:33 <libby> cheers for fixing logger dajobe

11:29:57 <dajobe> pc crashed

11:30:25 * dajobe blames nvidia binary (bleeugh) kernel drivers

11:30:38 <libby> yeugh

11:30:58 <libby> crashy crashy

11:31:49 <dajobe> it just stopped at 04:02, hey ho

11:31:51 <dajobe> computers--

11:32:31 <libby> yeah

11:32:36 <danbri> lsmod| grep nvidia

11:32:36 <danbri> nvidia 1468032 10 (autoclean)

11:32:37 * libby hates them

11:32:39 * danbri uhohs

11:37:05 <dajobe> How's this for an idea: I make an rdf/xml file with every RDF & RDFS property & class + datatypes, sending it to webont to see if it is OWL/DL/lite/whatever

11:37:37 <dajobe> it should validate as *some* OWL (Full?) I think

11:37:42 <dajobe> whatever validate means for that

11:37:59 <dajobe> hmm, wonder if they would add it to their test suites

11:38:38 <libby> they missed an oportunity to call it DLite there

11:38:43 <dajobe> lol

11:39:06 * libby cant believe I didnt wsee that before

11:40:17 * danbri neither

11:40:44 * danbri muses on 'grove is in the heart', but DL folk haven't heard of SGML Groves

11:40:47 <danbri> (mostly)

11:41:04 <libby> heh

11:41:11 * dajobe resists

11:52:04 <jang> grove is in the house

11:52:24 <jang> by funkmeister miller and dj dajobe.

11:52:28 <danbri> hi jang

11:52:37 <jang> hiya

11:55:17 * grove is in da house ;)

11:55:28 <danbri> lol

12:03:38 <libby> hey jang, how's it going?

12:28:33 * tim245 stretches, waves across the atlantic

12:28:53 <danbri> mornin'

12:29:03 <danbri> what's the '245' for?

12:29:24 <tim245> so maybe the f2f is a time to talk about the charter for rdf rules/query, then.

12:29:37 <tim245> tim245 is now known as timbl

12:30:12 <timbl> 245 was just a reaction to my nick being already used by something

12:30:13 <danbri> do we really want another rec-track group... hmm... i'm so torn!

12:30:27 <timbl> if www-rdf-rules working?

12:30:31 <timbl> s/if/is

12:30:59 <danbri> well, its a mailing list. Rules people wary cos it has 'rdf' in the name, query people wary cos it doesn't have 'query' in the name. But it gets used...

12:31:07 <danbri> was interesting thread libby started on test cases recently

12:31:40 <danbri>http://lists.w3.org/Archives/Public/www-rdf-rules/2003Jan/0000.html

12:31:41 <dc_rdfig> C: http://lists.w3.org/Archives/Public/www-rdf-rules/2003Jan/0000.html from danbri

12:31:54 <danbri> C:|RDF query testcases, thread on www-rdf-rules list

12:31:54 <dc_rdfig> Titled item C.

12:32:47 <libby> are you thinking about rdf query rec track work?

12:33:20 <timbl> Maybe we should make an alias which has query in the name, and replaces all occurrences of "rule" with "query"...

12:33:27 <libby> heh

12:33:32 <timbl> W3C is erveryone.

12:33:41 <timbl> I've been thinking it is due for a long time.

12:33:53 <danbri> C:Participants (so far): Libby Miller, Graham Klyne, Andy Seaborne, Alberto Reggiori, Jos De Roo, Jeen Broekstra, Dan Brickley, Eric Prud'hommeaux, Danny Ayers, Matt Biddulph...

12:33:53 <dc_rdfig> Added comment C1.

12:34:10 <timbl> eg http://www.w3.org/DesignIssues/diagrams/SemWave.png for example

12:34:14 <danbri> C:...I think this demonstrates some interest amongst implementors.

12:34:14 <dc_rdfig> Added comment C2.

12:34:25 <libby> working groups are no fun...

12:34:34 <timbl> eye there's the rub

12:34:39 <timbl> s/eye/aye

12:34:47 <danbri> I would rather do testing across existing RDFQ implementations to discover what folk have implemented in common, than start another huge WG effort.

12:35:12 <danbri> well, i presented a false opposition there...

12:35:24 <libby> that would be interesting to do

12:36:11 <danbri> If we do a WG, I'd rather it took simplest dumbest RDFQ as base, and extended only as people implemented it convincingly. I don't even feel any rush to have a W3C REC for query. Test cases describing what people have deployed would do fine for now, imho.

12:36:23 <libby> yeah

12:36:36 <libby> it's easy to get hung up on teh syntax too

12:36:40 <danbri> We'd need to coordinate with XML Query, OWL, RDFCore, XSLT2, plus probably would end up doing a protocol too

12:36:49 <libby> run away!

12:36:59 <danbri> ...since whats use in standardising a QL if there is no agreed way of sending queries to the server and getting results back?

12:37:12 <danbri> hence need to coordinate w/ web service stack too

12:37:16 <danbri> BIG JOB

12:37:31 <danbri> And not one that imho stands between us and the semantic web existing.

12:37:52 <danbri> On the other hand, if folk are queing up to do all the work, who am I to grumble? ;)

12:41:57 <timbl> Hmmm.

12:42:40 <timbl> We are just about to play with doing remote query by taking a query in N3, URI encoding it, appending it to the URI of a service, and dereferencing it.

12:43:11 <timbl> Then it becomes a newtowrk protocol.

12:43:43 * timbl 's fingers are clearly in newtown

12:46:42 <timbl> The funny thing was the remark of your's Libby, "The main issue for me is how to express the query itself.

12:46:42 <timbl> My inclination is to try and express queries as graphs with parts

12:46:42 <timbl> missing, that is, to interpret an RDF graph as a query, as Jos de Roo

12:46:42 <timbl> suggested [5].

12:46:44 <timbl> "

12:47:02 <timbl> in http://lists.w3.org/Archives/Public/www-rdf-rules/2003Feb/0000.html

12:47:15 <timbl> That is basically the need for a query langauge, no?

12:47:52 <timbl> It seems that defining the langauge for defining hte test cases is a sort of way of defining the langauge without telling anyone you are.

12:49:46 <libby> heh

12:50:12 <libby> well, the manifest part seems like the least controversial and the most inclusive.

12:50:25 <libby> I ducked out of the query part for now

12:51:12 * libby has a report to write, would liek to do something useful

12:51:51 <AndyS> The important use of standard query is between toolkits - just within one toolkit

12:52:00 <AndyS> isn't such a need (IMHO) for a standard.

12:52:08 <AndyS> Nice - nut not necessary

12:52:13 <libby> hey andys

12:52:23 <AndyS> s/nut/not/ - not the only finger trouble today

12:52:36 <libby> well I nearly called you dandys!

12:52:46 * AndyS waves to Libby

12:53:03 <danbri> hi dandy andys

12:53:26 <libby> it would be very neat to have interoperable query tests

12:53:45 <libby> course taht's not going to happen unless people actually use the samer (or similar) syntax

12:53:58 <danbri> it would certainly be healthy for rdf is app developers knew they could swap out database systems and have their app still work

12:54:11 * AndyS interested in timbl's remote query protocol

12:54:13 <danbri> instils confidence that they can invest time in an rdf app

12:54:35 <danbri> andy, does rdql allow for 'optional' triples in the WHERE clause?

12:54:52 <danbri> I'm starting to feel thats more useful than I ack'd previously...

12:55:05 <libby> oh, I was just writing an faq about that

12:55:12 <AndyS> no - RDQL is conjunctive - could easily dataflow a query to split into disjunctive parts but not done

12:55:14 <libby> oh yeah, I emailed you about it ;)

12:55:53 * danbri jealous of the 'easily' ;)

12:56:01 <AndyS> I see it as part of the "extract" problem - there is "locate" (find places in graph) ...

12:56:09 <danbri> yup

12:56:23 <AndyS> "extract" - gets stuff from graph and "present" how it looks to the application

12:56:36 <danbri> bibliographic search in z39.50 works the same way; search vs retrieval very different (too different, in their protocol...)

12:56:47 <AndyS> can eval many QL's in this simple framework.

12:57:16 <AndyS> re DB swapping - (I presume you mean toolkit swapping in SWEB case)

12:57:43 <AndyS> std query is the small bit - std API is the big bit !

12:59:43 <AndyS> Libby - just sent the email to www-rdf-rules you asked for

12:59:50 <libby> ooh, cheers

12:59:57 <libby> geoff's looks useful too.

13:00:06 <danbri> DB or toolkit, yeah...

13:00:13 <libby> hm, "Can you represent b-nodes in RDF queries?"

13:00:21 * AndyS off to meeting - back by 1500UTC

13:00:27 <libby> how does that work in RDQL andys?

13:00:29 <libby> arn

13:00:31 <libby> darn

13:01:21 <AndyS> "represent" bnodes? Can't write as constant values but can find by property - variables bound to Jena objects like Resource or Literal

13:01:40 <AndyS> and returned to program as such object - bNodes are resources

13:01:53 <libby> so you cant ask for a specific bnode, but you can find out if the result is ojne?

13:02:05 <AndyS> (Not complete story - can pass in prebound variables)

13:02:25 <libby> thanks andys

13:02:32 <AndyS> Yes - like finding a node you don't know the URI of ...

13:02:44 <libby> right, me too

13:02:56 <AndyS> Strangly, this has never been asked for.

13:02:58 <libby> can you contruct a query using objects in rdql?

13:03:02 * AndyS must go now ...

13:03:09 <libby> ok, see you later

13:03:30 <AndyS> Yes - a query is a Java datastructure that can be hand built or the parser used.

13:03:44 <libby> so you could alk for a sepcific bnode through an api

13:03:54 <libby> s/alk/ask

13:03:54 <AndyS> Yes

13:04:17 <libby> thanks andy, appreciate it

13:04:22 <AndyS> Yes you can ask for a bNode you already had found by some other means - like the rest of the API

13:04:33 <libby> right

13:04:43 <timbl> re mainfest: cwm regression tests are in n3 now, but novery reusable. Could be made more generic. http://www.w3.org/2000/10/swap/test/regression.n3

13:06:08 <libby> thanks

13:07:11 * timbl not sure whether Andrys is at a meeting or not

13:07:42 <libby> I think he's may have gone now, but he'll be back later

13:10:17 <timbl> I'll email.

13:11:45 <danbri> he timbl, re cwm regression tests... i did 'make tested' and was suprised to watch it do a CVS commit

13:13:30 * danbri discusses "Human publishers of RDF content commit themselves to the mechanically-inferred social obligations." w/ damian...

13:13:57 <danbri> we're both kinda skeptical, thinking about it...

13:14:10 <libby> C:there's an IRC chat about this later [agenda|http://lists.w3.org/Archives/Public/www-rdf-rules/2003Feb/0011.html]

13:14:11 <dc_rdfig> Added comment C3.

13:14:43 <libby> C:see also [later thread|http://lists.w3.org/Archives/Public/www-rdf-rules/2003Feb/thread.html]

13:14:43 <dc_rdfig> Added comment C4.

13:15:48 <danbri>http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Jun/0180.html

13:15:48 <dc_rdfig> D: http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Jun/0180.html from danbri

13:15:51 <libby> C:I tried to [summarise teh first thread in a wiki|http://esw.w3.org/t/view/ESW/RDFQueryTestCases]

13:15:51 <dc_rdfig> Added comment C5.

13:16:09 <danbri> D:|Earlier danbri / pat hayes notes on rdf-ms-assertion

13:16:10 <dc_rdfig> Titled item D.

13:32:04 <DanC> cwm crawl at 194:35.76 CPU time and still chugging

13:32:14 * DanC getting the kids off to school...

13:32:48 <DanC> rather, make that 625:56.78 of CPU!

13:35:40 <Paul|trampsoc> Paul|trampsoc is now known as Jibbler

14:01:29 * DanC sent another cwm-seeAlso-crawler to chase the first to see where these...

14:01:37 <DanC> Already read and parsed module.php?mod=network&op=foafrdf&uid=2084 to _formula

14:01:37 <DanC> Already read and parsed module.php?mod=network&op=foafrdf&uid=19227 to _formula

14:01:37 <DanC> Already read and parsed module.php?mod=network&op=foafrdf&uid=20390 to _formula

14:01:41 <DanC> ... are rooted...

14:01:44 <DanC> turns out to be:

14:01:53 <DanC> Content-type: '' for http://www.ecademy.com/module.php?mod=network&op=foafrdf&uid=1

14:02:13 <DanC> [[ Ecademy's cause is "to build the world's largest Trusted Business Network by connecting people to each other - enabling knowledge, contacts and opportunities to be shared for World Wide Wealth." ]]

14:02:43 <DanC> holy cow, danbri, did you know people are using foaf in money-making (or at least: money-seeking) enterprises?

14:03:42 <shellac> he does - currently having lunch, so I'll act as his representative-at-a-terminal

14:06:09 <shellac> we met julian bond (who is one of the founders IIRC) in London recently

14:07:23 <DanC> hmm... the foaf/blog synergy is kinda obvious... cool to see it taking off... but looking ahead... what if the journalism trade took the idea seriously? what if every person mentioned in a news story was described in RDF... or at least the authors.

14:07:58 <DanC> I can't seem to find documentation of what ecademy is doing with foaf...

14:08:18 <DanC> ... are they just getting it as a consequence of using http://www.newsisfree.com/ ?

14:08:48 <shellac> I think the foaf connection was Julian's idea

14:09:28 <shellac> he exposed their data in foaf - but IIRC there were some objections

14:10:39 <shellac> I'm not sure where Julian found out about it - you might well be correct

14:10:58 <DanC> ok, but clearly they've got a lot of data, i.e. the sun's gonna burn out before cwm finishes crawling it.

14:11:17 <shellac> :-)

14:11:47 <shellac> time to move it to assembler?

14:13:28 <shellac> I'm currently working on a rendezvous foaf tool - bit like the old unix 'finger' command

14:15:38 <DanC> wow... these ecadamy data are pretty rich! it's not just a zillion files with dead hotmail addresses or something

14:15:49 <shellac> The curious thing about ecademy using foaf is that foaf is so decentralised

14:16:03 <DanC> well, normally it is...

14:16:33 <DanC> but of course, centralized webs embed in a decentralized web just fine

14:16:41 <shellac> true

14:17:10 <shellac> I guess the see a chance to get more data, rather than a reason not to use the ecademy service

14:17:38 <DanC> amazing... they got all the little details right... escaped the &amp;'s, namespaces correct... these guys are sharp

14:18:03 <DanC> timbl, take a look at http://www.ecademy.com/module.php?mod=network&op=foafrdf&uid=1

14:18:17 <DanC> ecadamy is what's been keeping cwm busy all night.

14:19:06 <DanC> hmm... one detail that could be improved: each person has a separate URI-name in each foaf file they serve.

14:19:38 <DanC> in their centralized world, they could use unique names for things they know about

14:21:18 <DanC>http://www.ecademy.com/

14:21:18 <dc_rdfig> E: http://www.ecademy.com/ from DanC

14:21:30 <DanC> E:|ecademy -- connecting business people

14:21:30 <dc_rdfig> Titled item E.

14:21:42 <DanC> E:found it during foaf seeAlso crawl

14:21:43 <dc_rdfig> Added comment E1.

14:22:07 <DanC> E:e.g. [one of their exported foaf files|http://www.ecademy.com/module.php?mod=network&op=foafrdf&uid=1]

14:22:07 <dc_rdfig> Added comment E2.

14:22:14 <DanC> E:it's keeping cwm busy!

14:22:14 <dc_rdfig> Added comment E3.

14:22:48 <DanC> E:I think cwm is doing an n*(n-1) search over their data. :_{

14:22:48 <dc_rdfig> Added comment E4.

14:23:15 * DanC away family time...

14:23:43 * DanC kills cwm, after 659:18.74 CPU

14:28:42 * AndyS returns

14:34:39 <AndyS> timbl: re remote query : what is the contract for the "service" the remote query is sent to? A document? A processor?

14:40:58 <AaronSw> hi bitsko

14:41:00 <AndyS> Kinda "read part document?"

14:41:48 <bitsko> hey AaronSw. glad I caught you. what bot is 'datum', or is that a creation of yours?

14:41:55 <AaronSw> my creation

14:42:11 <bitsko> how easy is it to set up?

14:42:44 <AaronSw> not too hard, edit a couple of variables and type python datum.py

14:43:05 <bitsko> just one script, then? can I have it?

14:43:16 <AaronSw> what for?

14:43:47 <Morbus> [ot] evil!

14:43:51 <bitsko> internal irc server

14:48:37 <AaronSw> ok

14:49:52 <danbri> <DanC> cwm crawl at 194:35.76 CPU time and still chugging

14:49:56 * danbri grins

14:50:10 <danbri> <DanC> holy cow, danbri, did you know people are using foaf in money-making (or at least: money-seeking) enterprises?

14:50:13 * danbri grins some more

14:51:04 <danbri> ecademy is interesting... apparently a fair number of words were exchanged when FOAF was first automatically exposed, since that contactbook data is quite precious to its owners

14:51:15 <danbri> I think FOAF is now an opt-in facility on their site

14:51:44 <danbri>http://www.guardian.co.uk/online/story/0,3605,870848,00.html

14:51:44 <dc_rdfig> F: http://www.guardian.co.uk/online/story/0,3605,870848,00.html from danbri

14:52:25 <libby> re scutter, mine does some avoids: $JAVA_HOME/bin/java -classpath $CP org.desire.rudolf.app.Scutter -f \

14:52:25 <libby> file:scutterplan.rdf -d \

14:52:25 <libby> "rdf:jdbc:postgresql://localhost:5432/codepict?auth=password&user=postgres&password=notneeded" \

14:52:27 <libby> -avoid irc -avoid iris -avoid london.pm.org -avoid pubs \

14:52:27 <libby> -avoid http://jibbering.com/rdfweb/1016746423488.rdf

14:52:30 <danbri> F:|Click to the clique, Ben Hammersley on Social Networks, Ryze, Ecademy, Friendster and FOAF in the Guardian, 2003-01-09

14:52:30 <dc_rdfig> Titled item F.

14:52:47 <libby> the guy at the bottom wanted it removed. would be good to have a list to avoid probably

14:53:43 <danbri> Also we really should start observing robots.txt more carefully, when rdf crawling

14:53:53 <danbri> hi Jeen!

14:54:01 <dajobe> hi

14:54:18 <Jeen> good afternoon everyone. Hello Dan

14:54:27 <AndyS> Hi Jeen

14:54:33 <libby> hey jeen, glad you could make it

14:55:15 <Jeen> cheers. glad to figure out in time how to work this irc thing :)

14:55:18 <danbri> E:For sites that expose lots of such data automatically, maybe a single large aggregated download file would make more sense?

14:55:18 <dc_rdfig> Added comment E5.

14:55:19 <libby> heh

14:55:40 <dajobe> dajobe has changed the topic to: RDF & SemWeb hack-n-chat http://rdfig.xmlhack.com/ - RDF query language testcases meeting 15:00 UTC

14:58:12 * danbri waves to bbcuser136

14:59:35 * mattb waves to danbri

15:00:09 <shellac> 16 or 32k version?

15:00:51 <libby> ok, guess it's time for RDF query testcases chat

15:00:59 <libby> hey josd, glad you could make it!

15:01:15 <libby> ----------RDF query testcases chat ----------

15:01:17 <JosD> hi Libby!

15:01:24 <libby> how you doing?

15:01:54 <libby> hey alberto!

15:02:07 <AaronSw> AaronSw has changed the topic to: RDF Query language test cases chat | http://rdfig.xmlhack.com/

15:02:20 <JosD> fine Libby

15:02:24 <libby> good

15:02:30 <libby>http://lists.w3.org/Archives/Public/www-rdf-rules/2003Feb/0011.html

15:02:31 <dc_rdfig> G: http://lists.w3.org/Archives/Public/www-rdf-rules/2003Feb/0011.html from libby

15:02:38 <areggiori> hi all

15:02:55 <libby> G:|draft agenda for rdf query testcaeses chat

15:02:56 <dc_rdfig> Titled item G.

15:03:24 <libby> G:"Aim of the meeting - To start to draft a manifest format for RDF query testcases which can be used with multiple query syntaxes and perhaps multiple resultset formats.

15:03:24 <dc_rdfig> Added comment G1.

15:03:54 <libby> G:[something to start with|http://swordfish.rdfweb.org/rdfquery/tests/query-results-manifest.rdf]

15:03:54 <dc_rdfig> Added comment G2.

15:04:27 <libby> G:[andy seaborne's version|http://lists.w3.org/Archives/Public/www-rdf-rules/2003Feb/0013.html]

15:04:27 <dc_rdfig> Added comment G3.

15:04:57 <libby> G:[geoff chappell's comments|http://lists.w3.org/Archives/Public/www-rdf-rules/2003Feb/0012.html]

15:04:57 <dc_rdfig> Added comment G4.

15:05:31 <libby> would everyone who is here like to introduce themselves to the chump (the robot that's logging to http://rdfig.xmlhack.com?)

15:05:38 <libby> i.e. people here for the chat

15:05:49 <libby> BLURB:IRC chat about RDF query testcases

15:05:49 <dc_rdfig> H: IRC chat about RDF query testcases from libby

15:06:03 <libby> H:attending:

15:06:03 <dc_rdfig> Added comment H1.

15:06:33 <JosD> think about using log:semanticsOf, as a web page is not the same as a query/result written on it

15:06:39 <libby> H:libby miller, [inkling/squishQL query implementation|http://swordfish.rdfweb.org/rdfquery/]

15:06:40 <dc_rdfig> Added comment H2.

15:07:23 <libby> andys, care to introduce yourself?

15:07:24 * timbl starts to catch up

15:07:33 <AndyS> H: Andy Seaborne [RDQL in Jena|http://jena.sourceforge.net]

15:07:33 <dc_rdfig> Added comment H3.

15:07:36 <AndyS> Like that?

15:07:46 <libby> yeah,

15:07:53 <libby> just so we get an idea of who is here

15:07:57 <libby> alberto?

15:08:01 <libby> jeen?

15:08:04 <libby> josd?

15:08:27 <areggiori> hello, I am Alberto Reggiori who prepared together with Andy Seaborne the [RDF Query and Rules|http://rdfstore.sourceforge.net/2002/06/24/rdf-query/] survey - I am also personally working the [RDFStore|http://rdfstore.sf.net] perl software

15:08:37 <Jeen> hugo: Jeen Brokestra, [Sesame/RQL,RDQL query implementation|http://sesame.aidministrator.nl/]

15:09:05 <libby> guys, try H: with no space between them

15:09:06 <JosD> H: Jos De Roo [Euler|http://www.agfa.com/w3c/euler]

15:09:07 <dc_rdfig> Added comment H4.

15:09:13 <libby> in capitals

15:09:18 <libby> like jos did :)

15:09:18 <AndyS> Everyone: the H: attached the text/link to the chump at http://rdfig.xmlhack.com/

15:09:23 <Jeen> H: Jeen Broekstra, [Sesame/RQL,RDQL query implementation|http://sesame.aidministrator.nl/]

15:09:24 <dc_rdfig> Added comment H5.

15:09:29 <libby> thanks

15:09:30 <AndyS> because this is entry "H" today - have a go!

15:09:30 <Jeen> there we go :)

15:10:03 <areggiori> H: Alberto Reggiori and Andy Seaborne [RDF Query and Rules|12http://rdfstore.sourceforge.net/2002/06/24/rdf-query/]

15:10:03 <dc_rdfig> Added comment H6.

15:10:09 <AndyS> It can be changed later : retype as H<number you got>: new stuff

15:10:20 <libby> just to say why I suggeste dthe meeting: I have to write a report!

15:10:50 <timbl> H: Tim Berners-Lee, [Swap, Cwm|http://www.w3.org/2000/10/swap]

15:10:50 <dc_rdfig> Added comment H7.

15:10:50 <libby> ...and I wantedto put effort into somethign useful rather than rep[eat cool stuff that's been done already, e.g. by alberto and andys and ericP

15:10:52 <AaronSw> the colors broke the link

15:10:52 <AaronSw> H6: Alberto Reggiori and Andy Seaborne [RDF Query and Rules|http://rdfstore.sourceforge.net/2002/06/24/rdf-query/]

15:10:53 <dc_rdfig> Replaced comment H6.

15:11:13 <AndyS> H: See also [ISWC paper|http://www-uk.hpl.hp.com/people/afs/Papers/ISWC%202002%20-%20SquishQL.htm] with Libby and Alberto

15:11:13 <dc_rdfig> Added comment H8.

15:11:17 <areggiori> saw that, thanks

15:11:44 <libby> and I think we can up with a manifest format at least, and that might get us some way torwards reusing each others' queries for example

15:12:01 <libby> so thanks everyone fopr comeing. shall I start on the agenda (finally?)

15:12:30 <AndyS> Yes Libby (noting Geoff's 1.1 <-> 1.2 comment)

15:12:37 <Jeen> please

15:12:40 <libby> heh

15:12:41 <danbri> H:[http://www.w3.org/people/danbri/|Dan Brickley], W3C. Stuff I've done... Co-authored QL'1998 papers on rdf - [http://www.w3.org/TandS/QL/QL98/pp/enabling.html|Enabling Inference], w/ Guha et al.; [http://www.w3.org/TandS/QL/QL98/pp/distributed.html|distributed query issues] w/ PeterV, [http://www.w3.org/TandS/QL/QL98/pp/queryservice.html|Query and Inference services] w/ Decker et al. Worked on the library that became Inkling w/ Libby

15:12:41 <danbri> her mutate Guha's RDFdb QL into Squish, worked on Squish implementations in Ruby, working on mapping between RDF query langauges and Mozilla XUL templates. Recently more interested in hooking soap services into distributed RDF query, and on rules/query relationship, and on whether we should start more formal W3C work in this area.

15:12:41 <dc_rdfig> Added comment H9.

15:12:46 <AaronSw> heh

15:12:52 * danbri too verbose :)

15:13:01 <danbri> H:...her mutate Guha's RDFdb QL into Squish, worked on Squish implementations in Ruby, working on mapping between RDF query langauges and Mozilla XUL templates. Recently more interested in hooking soap services into distributed RDF query, and on rules/query relationship, and on whether we should start more formal W3C work in this area.

15:13:01 <dc_rdfig> Added comment H10.

15:13:10 <libby> 1. Requirements for a RDF query testcase format.

15:13:39 <AndyS> H: Remote query experiment [Joseki|http://www.joseki.org/]

15:13:39 <dc_rdfig> Added comment H11.

15:13:44 <libby> 1.2 first.....

15:13:48 <libby> 1.2 What do we intend to do with testcases expressed in this format?

15:14:15 <libby> geoff says:

15:14:16 <libby> [[

15:14:17 <libby> Possibilities include:

15:14:17 <libby> - cataloging how different query languages perform similar queries as a

15:14:17 <libby> means of gathering structured use cases for a future common rdf query

15:14:17 <libby> language or query mechanism

15:14:19 <libby> - automated testing

15:14:20 * timbl assumes H is just for introductions, and new stuff should be chumped with anew

15:14:21 <libby> - query rdf data and retrieve results in a std format (this assumes the

15:14:22 <libby> testcase format specifies a common query and result format and would be

15:14:24 <libby> more a result of those aspects than of the testcase as a whole)

15:14:26 <libby> - ...?

15:14:28 <libby> ]]

15:14:44 <libby> yeah I guess so

15:15:00 <libby> re 1.2, I'd plump for automated testing

15:15:05 <libby> as my main reason

15:15:06 <AndyS> I'd like to use the format in Jena for the RDQL testing itself. Also export/import to other systems.

15:15:12 <timbl> I would use these cases as regression test cases when making new releases.

15:15:15 * AndyS notes timbl's comment

15:15:53 <Jeen> automated testing is one thing I'd like. also, the catologing bit is interesting to me, to see how qls compare in expressivity.

15:16:06 <libby> interesting jeen

15:16:08 <areggiori> Alberto says:

15:16:12 <areggiori> [[

15:16:12 <areggiori> - test query parser engine (query parser tests)

15:16:27 <areggiori> - test query engine results (query tests)

15:16:40 <areggiori> -> autogenerate testing programs (see DanBri ruby stuff or perl Test::Harness)

15:16:41 <JosD> I would use them as proofs and as continuations (for new resolutions)

15:16:56 <danbri> re 1.2, I have a semi-articulated assumption, ie. that there is, across many RDF query implementations, a common subset (conjunctive query w/ no fancy stuff), and that implementations of this could share more by way of test cases and even query syntax than they do to date.

15:17:03 <areggiori> -> real world examples and use cases with real data (see RDF Query and Rules survey)

15:17:15 * timbl notes that he has done N3 versions of AndyS/agreggoriori's tests 1-8 but not filled in the form for them

15:17:17 <jang> re: common subset...

15:17:19 <danbri> ...we can not this without belittling the contribution of other approaches to RDF query, eg. path-based stuff.

15:17:26 <jang> .. the subset may be smaller than you think

15:17:27 <danbri> ...or importance of adding the extras.

15:17:29 <areggiori> - test transformation capabilities

15:17:35 <jang> eg, support for blank nodes across APIs is vastly variable

15:17:36 <danbri> jang, can you elaborate?

15:17:41 <jang> ^^ elaboration

15:17:43 <areggiori> - test inference capabilities (entailment)

15:17:48 <areggiori> ]]

15:17:58 <jang> also, there's a second point:

15:18:03 <timbl> The proof of the pudding is in software to convert from one format to another.

15:18:05 * danbri nods, yes I think having a set of test queries *and target dataset* will help flush out such differences

15:18:07 * JosD wondrs about a pointer to that path based stuff...

15:18:14 <libby> I think real world stuff is v important (and yours and andy'd stuff very cool alberto)

15:18:21 <jang> it's been really helpful - rdf core that is - in articulating the distinction between abstract syntax and inference/entailment

15:18:33 <danbri> JosD, I vaguely recall path discussions, there was an rdfpath list on yahoogroups eg

15:18:45 * JosD agrees with jang; very important

15:18:47 <timbl> A test typically needs a query in some langauge, a test data set and an expected result.

15:18:50 <jang> query implementations can sometimes blur that distinction

15:18:59 <libby> isn't versa like that? (path-y)

15:19:03 <jang> are these query languages all targetting the abstract syntax?

15:19:07 <danbri> timbl, we can convert between query formats fairly cheaply, but the engines may still perform differently

15:19:09 <areggiori> I have been thinking about using some ideas of RDF Test Cases manifest for our needs

15:19:12 <danbri> yes, I think so JanG

15:19:27 <AndyS> Can only convert if they have the same power => common subset

15:19:48 <timbl> By converting you end up classifying queries by hte sorts of engines which can handle them.

15:19:51 <AndyS> (at least to start with ...)

15:19:55 <jang> the common core of conjunctive triple-matching can't answer "is X a subClassOf y"?

15:20:01 <JosD> timbl; what about (transient) merge of graph A and graph B as in http://lists.w3.org/Archives/Public/www-rdf-rules/2003Feb/0003.html

15:20:29 <danbri> I do hope these efforts don't discourage folk from exploring the non-common-subset aspects of RDF query, eg. support for xpath or fulltext search or fancy datatyping relations etc.

15:20:30 <timbl> I think it mkaes sense to have a langauge be expressive, and then have a conformance level for one or more subsets.

15:20:31 <AndyS> Inference and query can be separate - graph (data+engine) does inference, then query asks enhanced graph

15:20:38 <danbri> ...there's much creative work still to be done.

15:20:48 <jang> exactly andy

15:20:53 <jang> I think that's a good approach

15:21:05 <danbri> ...even though the enhanced graph may never really exist, that gives us a conceptual model

15:21:06 * libby too

15:21:09 * JosD really danbri :-)

15:21:20 <jang> it means that a triple-matching API may expose inferencing ability

15:21:23 <AndyS> So we need entailing and non-entailing variants?

15:21:26 <danbri> queries are about asking questions; how we answer them is nicely separate

15:21:38 <timbl> jang: I think we can separatehte inference from the dataset search. You just express that the given query service serves (effectively) the decutive closure of DDD under XXXX .

15:21:55 <jang> that's exactly the conclusion that 'd come to

15:22:11 <jang> it seems the neatest way to divide up the problem

15:22:13 <timbl> Good. So the query langauge is not affected.

15:22:34 <danbri> I started trying to characterise the entailments I was using. eg. (i) none (ii) RDFS (iii) RDFS + identity reasoning / smushing (iv) OWL if I get time, ie never ;)

15:22:42 <jang> also, the QL remains expressive without having to try to hand-code multiple transitive closures of properties

15:22:44 <timbl> I think taht was what the DAML-Q breakout group at the last DAML PI meeting thought too.

15:23:19 <arnarl> hi

15:23:21 * jang thinks: so you do SELECT foo FROM rdfs-closure(g1) WHERE ... :-)

15:23:27 <danbri>http://www.daml.org/2002/08/dql/dql

15:23:27 <dc_rdfig> I: http://www.daml.org/2002/08/dql/dql from danbri

15:23:34 <Jeen> not completely, some query language explicitly specify that they use entailment. RQL does this, for a couple of its constructs.

15:23:36 <danbri> I:|DAML Query Language (Aug 2002 draft)

15:23:36 <dc_rdfig> Titled item I.

15:23:44 <areggiori> what about this: PostivieQueryParserTest / NegativeQueryParserTest for the query parser engine, PositiveQueryTest/ NegativeQueryTest for the query engine itself (about real results), PositiveEntilmentQueryTest / NegativeEntilmentQueryTest being inspired to RDF Test Cases

15:23:46 <mattb> H:Matt Biddulph, [bits of triple querying code|http://www.picdiary.com/triplequerying/]

15:23:46 <dc_rdfig> Added comment H12.

15:24:02 <danbri> I:"Each binding in a query answer is a URIref or a literal that either explicitly occurs as a term in the answer KB or is a term in DAML+OIL. That is, DQL is designed for answering queries of the form "What URIrefs and literals from the answer KB and DAML+OIL denote objects that make the query pattern true??"

15:24:02 <dc_rdfig> Added comment I1.

15:24:18 <timbl> The power of inference isn't only hte logical inference, its the arithmetic, string etc functions supported, and whether the engine will go fetch documents from the web and so on.

15:24:57 <AndyS> Can bindings not be bNodes in DQL?

15:24:59 * danbri wonders how much structure folk want from this meeting; I'm happy at this stage with a general brainstorm (loosly structured by the agenda)

15:25:10 * JosD has very negative experience with NegativeEntailmentTest stuff

15:25:15 * jang is about to ask whether the question of denotation versus addressing is aposite, and bites his tongue

15:25:17 <danbri> Jos, can you expand?

15:25:18 * mattb tries to follow and catchup while juggling bbc work...

15:25:29 <libby> thanks for coming along mattb

15:25:38 <timbl> I'd like this meeting to concentrate on defining the triple extraction langauge -- its test infrastructure.

15:25:39 * AndyS wants something to come out of meeting (as well as this great discussion)

15:26:24 <danbri> OK, so are we done talking around "1.2 What do we intend to do with testcases expressed in this format?"

15:26:29 * em agrees with AndyS

15:26:32 * em waves

15:26:37 <libby> hey em

15:26:38 <danbri> hi em

15:26:42 <em> howdy all :)

15:26:46 <em> sorry i'm late

15:26:55 <JosD> well Danbri, try to do it and you will realize that negation in an open world is impossible...

15:27:03 <areggiori> hello

15:27:05 <danbri> ah :)

15:27:19 <AndyS> em - intros to chump H:

15:27:22 <jang> negative entailment isn't negation,

15:27:38 <jang> it's just asking: does this document entail that one? Not according to these rules...

15:27:47 <jang> but it does seem slightly dodgy

15:28:31 <em> thanks AndyS, reading log/chump now

15:28:49 <libby> is perhaps 1.3 a bnice easy one?

15:28:52 <timbl> So far, we have parser test manifests, and for cwm we have rather cwm-specific test manifets which for example give cwm command line arguments. I'd like to more to a more generic vocabulary in which the processing to be done is described in an RFD graph.

15:29:06 <JosD> well jang, I want to see if I can't proof it (the n't involves negation)

15:29:19 <areggiori> I mentioned entailment thinking about basic RDFS inferencing i.e. rdf:type "magic" in triple-pattern

15:30:08 <danbri> So, how would folks feel about targetting an RDF query test format which tried to target the simple conjunctive structure shared by guha's ql, rdql, squish, algae etc., and which through doing so (i) helped us find out if there were in fact multiple implemenentations of this that behave the same (ii) helped us appreciate the value and implementation difficulty of the various other directions in which rdf query systems have gone?

15:30:22 <AndyS> RDF is the best syntax - else we look "odd"(er)

15:30:31 <libby> heh

15:30:45 <libby> danbri, +1

15:30:53 <AndyS> danbri: +1 - just the conjunctive triple match for the moment

15:30:55 <JosD> is --think an example of such an abstraction; e.g. as returing all answers?

15:30:57 <danbri> Yep, An RDF description of an RDF query is an ugly beast, but we can map to prettier syntaxes later

15:31:05 <timbl> By "RDF", do you allow extenstion to quoted formulae?

15:31:14 <danbri> nope, that's N3 timbl

15:31:15 <areggiori> danb, yes I agree

15:31:26 * danbri would like to hear from Jeen re RQL

15:31:36 <timbl> That means effectively you have to reify everyhting, which is tedious.

15:31:42 <Jeen> ...working on it :)

15:31:43 <AndyS> Actually, we need variable predicates so not pure RDF

15:31:58 <danbri> conjunctive triple query seems to be lead candidate for the thing that is most multiply-implemented...

15:32:05 <danbri> AndyS, yes thats what I meant about ugly.

15:32:06 <AndyS> Or we could start real simple and have fixed predicates and variables at nodes

15:32:16 <Jeen> yeah, it seems a good start, and it will help us understand how (if) RQL extends the basic pattern.

15:32:33 <danbri> If we don't have variable predicates, we can do Query By Example, or Decorated Query By Example (ie. attach var names to a graph)

15:32:48 <danbri> Jeen, thanks, that's good to know.

15:32:48 <libby> can you show us an example danbri?

15:32:53 * danbri rummages

15:33:10 <areggiori> In my RDQL implementation in RDFStore the triple-pattern is a quad like thingie: s,p,o,c - where 'c' is a context (statement group)

15:33:25 <danbri>http://lists.w3.org/Archives/Public/www-archive/2002Apr/0040.html

15:33:25 <dc_rdfig> J: http://lists.w3.org/Archives/Public/www-archive/2002Apr/0040.html from danbri

15:33:28 <Jeen> although I'm worried since the result set of an rql query is not necessarily a set of triples. it can be 'faked' though...

15:33:30 <AndyS> QBE-style? I saw this and it had predciate varaibles - will find URL for that group

15:33:41 * timbl isn't sure that Query By Example doesn't really use variables.

15:33:41 <danbri> J:|'Decorated' query by example, in RDF syntax

15:33:42 <dc_rdfig> Titled item J.

15:33:56 <danbri> some folks don't mind about lack of variable names, others do

15:34:07 <AndyS>http://www.csd.abdn.ac.uk/research/AgentCities/QueryByExample/index.php

15:34:07 <dc_rdfig> K: http://www.csd.abdn.ac.uk/research/AgentCities/QueryByExample/index.php from AndyS

15:34:21 <danbri> ie. you can take any chunk of RDF, consider bnodes as unknowns, and think of it as a 'what things fit into the bnode roles to make this true', or you can be more explicit

15:34:35 <AndyS> K:|RDF Query By Example

15:34:35 <dc_rdfig> Titled item K.

15:34:39 <libby> it's more useful top ahve the variables anmes

15:34:53 <Jeen> libby, I agree

15:34:58 <libby> so danbri, in that example you cant have preds as variables

15:35:10 * JosD thinks danbris explanation is just an entailment test, no?

15:35:16 <danbri> Yeah, I find the variables v useful, and I think in terms of SW education/outreach, it helps us to echo the SQL programmers paradigm

15:35:16 <libby> that's anoying, although still very useful

15:35:21 <AndyS> Need variable names for non-tree shapes - but do we care? ( I do!)

15:35:32 <libby> I do too.

15:35:43 <danbri> yes, agreed libby. to have variables as preds requires reifying the whole damn thing (a la DQL I believe)

15:35:44 <em> hugo: Eric Miller, started query work in [eor|http://eor.dublincore.org/] but playing with RDQL, Squish lately... interested in seeing if it makes sense if a query / API is ready for standardization for future semantic web [work|http://www.w3.org/2001/sw/Activity]

15:35:47 <libby> can we get a fair waywithout preds as variables?

15:35:52 * JosD also very much i.e. graphs

15:35:58 <em> H:Eric Miller, started query work in [eor|http://eor.dublincore.org/] but playing with RDQL, Squish lately... interested in seeing if it makes sense if a query / API is ready for standardization for future semantic web [work|http://www.w3.org/2001/sw/Activity]

15:35:58 <dc_rdfig> Added comment H13.

15:35:59 <danbri> em s/hugo:/H:/

15:36:06 * em thought he turned off autocompletion.. arrr

15:36:12 <libby> can't we cheat with preds?

15:36:43 * AndyS wonders if we are on agenda - but this is interesting!

15:36:53 <libby> heh, no, but this si better :)

15:36:59 * JosD Libby cheating?!

15:37:03 <libby> we are kinda on requirements

15:37:03 <jang> danbri: what's tha about preds?

15:37:05 <jang> you mean predicates?

15:37:12 <libby> yes jang

15:37:13 <danbri> quick straw poll: Who here has found frequent use for quantifying over predicates in their queries?

15:37:15 <danbri> yes i did

15:37:24 <jang> to have variables for predicates requires reification? nonsense

15:37:26 <timbl> yes i did

15:37:28 <JosD> I didn't

15:37:33 <libby> sometimes, esp fro getting all members of a list

15:37:46 <jang> can I offer an observation:

15:37:53 <libby> wouldnt call it 'frequent' though

15:38:02 <jang> support for inference and closure means that you can use superproperties as predicates

15:38:04 <AndyS> I see it but it is a lesser case - with inference I think it will be more important (subproperty matching)

15:38:11 <danbri> I _used_ to, before we had rdfs:member. I needed: ?rel(?c,?i) rdf:type(?rel,rdfs:ContainerMembershipProperty), but that use case went away with rdfs:member.

15:38:12 <jang> and don't need them in the query as variables

15:38:34 <libby> jang, explain again?

15:38:45 <AndyS> Apps may want to know how X connects to Y

15:38:51 <danbri> It seesm a shame if we're doing automated testing to not be able to exercise that aspect of RDF query systems, even though the cost might be that our sample query files are more verbose.

15:38:59 <libby> josd, I was thinking that esp in Ntriples, it doesnt seem so bad to have a blank predicate ;)

15:39:00 <jang> 99% of use cases for predicates being bound variables...

15:39:11 <jang> ...disappear if you follow PO chains

15:39:17 <dajobe> (aside: you can't have bnodes for predicates in rdf/ntriples)

15:39:17 <jang> subPropertyOf chains even

15:39:30 <JosD> agreed Libby, we have similar experience now

15:39:44 <danbri> But that's ok, cos we can always author then in another syntax, and generate the verbose one; or generate alternate-syntax views of the 99% of verbose ones that don't have variables in predicate role.

15:39:56 <AndyS> Transitive props? Yes - as also subgraph from node (Dans point before we started)

15:40:01 <libby> right, though several sysntems dont follw the chains. but I see what you mean - youmean pred vars are being used to get over that hurdle

15:40:04 <JosD> agreed jang,

15:40:20 <timbl> Use case: { ?x ?p ?y. ?y ?p ?z. ?p a daml:TransitiveProperty} => { ?x ?p ?z }.

15:40:26 <AndyS> Simple start 1: RDf graph - bnodes, no bArcs

15:40:41 <jang> incidentally, when it cones to transitive closures:

15:40:41 <AndyS> Simpel start 2: cwm/N3 LHS

15:40:54 <danbri> how do those two differ?

15:41:08 <jang> I was convinced you could close on subPropertyOf, then on subClassOf; suddenly I'm not so sure, can't put my finger ona counterexample though

15:41:09 <AndyS> Start 3 (not so simple): cwm/N3 rule (formats output)

15:41:31 <AndyS> differ in allows predicate variable (+exist/uni vars)

15:41:46 <timbl> There is a big general question as to whether a query returns a set of bindings or RDF.

15:41:49 <AndyS> Also: timbl has a parser ...

15:42:02 <jang> timbl: you occasionally want both

15:42:12 <danbri> timbl, that's not such a big question is it? you can think of it as both...

15:42:14 <AndyS> Also : One RDF graph or one per match possibility.

15:42:19 <libby> timbl, people seem to like bindings, see recent rdf-rules thread

15:42:24 <jang> that is, the "set of bindings" return can be plugged into an "assert as rdf" side

15:42:44 <areggiori> timbl: yes, we should distinguish between those

15:42:49 <danbri> if you hang on to the orig query, and just have the bindings, you can get the graph; or if you requery the graph you can reconstruct the bindings. But both have a use.

15:42:49 * mattb has found uses for both bindings and RDF as results

15:42:55 <mattb> often within the same application

15:42:56 <timbl> differ: aslo in allowing literal as subject, variable as predicate, bnode as predicate, allowing formulae, etc in generality.

15:43:04 <JosD> I don't see why we can't have both; just describe it explicitly in the result

15:43:13 <AndyS> danbri - recalculation can be expensive (but is what Joseki does)

15:43:14 <jang> analogously: sql's SELECT gives bindings; SELECT INTO creates a new structure

15:43:14 <danbri> I really think it varies from app to app, and developer to developer. Yes, what JosD said.

15:43:17 <timbl> recent rdf-rules thread? missed that.

15:43:17 <jang> what a poor analogy

15:43:19 <mattb> bindings results lend themselves to very sql-ish uses

15:43:30 <areggiori> danbri, sure

15:43:39 <jang> also, there's a chicken and egg problem:

15:43:44 <timbl> I like returning RDF. So did the query by example folks I notice.

15:43:47 <jang> bindings give you an answer

15:43:57 <AndyS> bindings is about getting info out of graph

15:43:58 <jang> rdf graphs give you graphs... but how do you examine them?

15:44:03 <mattb> bindings help you translate graphs into square data

15:44:04 <jang> with a query, which gives you a rgaph...

15:44:05 <danbri> I really think theres a strong sell on the only-slightly-misleading SQL analogy, and that consequently we should be clear about preserving the bindings view, and on noting that the bindings plus the query gives a graph.

15:44:06 <AndyS> subgraph is for furthe rprocessing

15:44:09 <libby> I like return ing rdf too - easy to test. but not as usefyul for using the answers I find

15:44:10 <mattb> sometimes that'svery useful for exporting

15:44:22 <mattb> for personal preference, i like rdf results

15:44:24 <Jeen> as an aside we are working on Yet Another :) ql that offers both. SELECT queries return variable bindings, CONSTRUCT queries return (transformed) graphs.

15:44:34 <libby> oooh

15:44:37 <danbri> the cost for us is that we don't yet have an agreed RDF or XML or whatever representation of the bindings structure.

15:44:46 * danbri intrigued

15:44:46 <AndyS> Sounds good Jeen - any docs?

15:44:52 * jang would like to see someone (anyone!) who uses a synonym for select in one of these languages :-)

15:44:52 * timbl suggets jang type a lot before hitting return he is being broken up!

15:44:52 <libby> danbri, not sure that's v big problem

15:45:03 * danbri neither, but its more work ;)

15:45:05 <Jeen> it's early days, we have a draft circulating. I'll send it to the mailinglist next week ok?

15:45:10 <danbri> cool

15:45:15 <AndyS> Excellent!

15:45:16 <libby> well andys has a proposal

15:45:16 <timbl> There seem to be two sets of langauges , those which use SELECT and those which use -> or =>

15:45:27 <jang> timbl: yeah, used to a slightly different style of irc meeting.

15:45:28 <Jeen> actually make that two weeks, i'm in mallorca next wee...

15:46:38 <danbri> re <AndyS> Start 3 (not so simple): cwm/N3 rule (formats output)

15:46:44 <danbri> is anyone proposing that?

15:46:54 <danbri> if not, can we focus on 2 vs 1, and on whether they differ?

15:47:33 <libby> I propose using http://lists.w3.org/Archives/Public/www-archive/2002Apr/0040.html, i.e. RDF format for query, no reification, no variables on predicates

15:47:34 <AndyS> My pref: 3 (generates results), then 2 (close behind) then 1

15:47:39 <timbl> To me it is obviously quite important that starts 1 2 and 3 are subsets of each other.

15:48:02 <jang> you mean they intersect? or they're equivalent?

15:48:06 <AndyS> Does Edutella have anything to say?

15:48:21 <danbri> who is here from Edutella? Is that Jeen again?

15:48:21 <AndyS> I think (hope) 1<2<3

15:48:23 <JosD> Libby, that URI seems to be 404

15:48:33 <Jeen> nope, I'm not in Edutella...

15:48:50 <libby> works for me josd

15:48:57 * AndyS gets that URL OK

15:48:58 <libby> I was hoping wolfgang could make it

15:48:59 <danbri> re http://lists.w3.org/Archives/Public/www-archive/2002Apr/0040.html I would hope that a manifest format would support this, but that it would also support more sophisticated variants

15:49:03 <libby> from edutella

15:49:32 <libby> danbri, me too: we lurched from talking about manifests to talking about query formats

15:49:34 * JosD is sorry, now it's OK

15:49:35 <danbri> for rubyrdf I wanted a tree of tests where I'd author a test in whatever format, and generate the others accordingly

15:49:57 <danbri> maybe we should all hush and revisit the agenda? :)

15:50:11 * danbri entirely distracted by all the interesting comments...

15:50:17 <libby> it is v interesting

15:50:24 <timbl> Here's the snag I found in N3 wiht just using variable names whit "?". Its OK for levels 1 and 2 but then when you can nest formulae, you actually have to explicit about which formula a variable is quantified in. The forall has to be in there.

15:51:19 <timbl> My compromise was to allow "?" whith the constraint that it implies quantificatoin within the next level up, which handles RDF query level 1 and 2 style.

15:51:48 <danbri> yup... but that's relatively rocket science compared to the basic level of interop and query capability we're initially focussing on. Hopefully we can have a relatively non-commital manfiest format that wouldn't need changing once we move to testing such fancier facilities...?

15:51:50 <AndyS> For query (i.e. -filter), can we have one-level only?

15:51:51 <libby> so we had a lot of agreement for [[

15:51:52 <libby> <danbri> So, how would folks feel about targetting an RDF query test format which tried to target the simple conjunctive structure shared by guha's ql, rdql, squish, algae etc., and which through doing so (i) helped us find out if there were in fact multiple implemenentations of this that behave the same (ii) helped us appreciate the value and implementation difficulty of the various other directions in which rdf query systems have gone?

15:51:53 <libby> ]]

15:52:05 <libby> but it's important that it's extensible I guess

15:52:42 * libby notes that we have about 10 minutes left scheduled

15:52:44 <AndyS> to be clear - that is "simple start 2" - right libby?

15:53:19 <libby> I'd actually be inclided to go with 1

15:53:25 <libby> straw poll maybe?

15:53:28 <danbri> So we note (i) the initial design of the manifest format targets a to-be-explored common subset of RDFQ, anded triple query; (ii) we try to keep the manifest format noncommital re concrete syntaxes and deployment choices...

15:53:52 <libby> +1

15:54:11 <Jeen> agreed

15:54:23 <areggiori> +1

15:54:26 <timbl> ANy ideas about a default way to return bindings when the constructed RDF is not specified? As a list? As a copy of the matched bits of graph?

15:54:30 <AndyS> Manifest general - agreed

15:54:39 <danbri> libby, re simple start 1 vs 2

15:54:40 <danbri> <AndyS> Simple start 1: RDf graph - bnodes, no bArcs

15:54:40 <danbri> <jang> incidentally, when it cones to transitive closures:

15:54:40 <danbri> <AndyS> Simpel start 2: cwm/N3 LHS

15:54:49 <danbri> the distinction: is it just bArcs or not?

15:55:00 <AndyS> hasSolution [ rs:hasBinding [ rs:variable "u" ; rs:value ?u ] ; ]

15:55:26 <danbri> One could go with what I suggested but in practice author a heap of tests without bArcs, and use a syntax (eg. the decorated QBE) that doesn't even support them.

15:55:32 <AndyS> I think it is the bArcs

15:55:42 <libby> timbl, as a sort of RDF table? I think RQL takes that approach

15:56:09 * JosD timbl, that "within the next level up" for ?vars is I think not important (at least not for backchaing) but I think further...

15:56:12 <danbri> ie. it would be a real shame if our design of the manifest format stopped us testing bArc implementation...

15:56:18 <libby> danbri: I'd like to go with that, although the manifest format should be allow more complex approaches

15:56:19 <Jeen> pretty much, yes. RQL result set is a bag containing several sequences, one for each "table row".

15:56:33 <AndyS> Whatever it is can we encoded in RDF so can do graph match (isomorphism) please.

15:56:36 <libby> danbri, agreed

15:56:50 <danbri> Jeen, that's interesting...

15:56:51 <libby> andys - the resultset? makes sense...

15:57:01 <AndyS> Bag of results : +1 here

15:57:02 <areggiori> +1

15:57:10 <JosD> +1

15:57:10 <libby> +1

15:57:14 <libby> :)

15:57:16 * danbri proposes to extend the meeting until 15 past

15:57:20 <Jeen> +1

15:57:21 <AndyS> hasSolution = one row of result set (table)

15:57:21 <libby> seconded

15:57:27 <JosD> +1

15:57:30 <areggiori> ok

15:57:37 <libby> andy, where's your resultset propsal?

15:57:38 <Jeen> unfortunately I'll have to leave in five, but I'll read the log afterwards then.

15:57:41 <timbl> please please make the manifest format inherently flexible but then explicitly constrain it to the level you want.

15:57:44 * danbri not sure whether the +1s are for extending the meting or restultsets

15:57:46 * AndyS would like extension - realises may be a pain for some

15:57:50 <danbri> Yes, what TimBL said.

15:58:26 <danbri> We can say "Here's a useful general manifest format, handy (at least) for conjunctive triple query systems. And _here_ is how we intende to collaborate with it on www-rdf-rules to see if our tools behave similarly"

15:58:34 <areggiori> ok for the meeting extension

15:58:48 <libby> ok, could we perhaps spend the last few mins looking at the manifest itelf?

15:58:51 * timbl not very happy with bags, ok with collections

15:59:05 <jang> not collections, no order

15:59:07 <danbri> Let's do time of next meeting first, as some have to leave per original schedule.

15:59:14 <libby> ok, cool

15:59:17 <jang> just query -[hasSolution]-> result row

15:59:17 <jang> ...

15:59:23 <timbl> collections have a well-defined end.

15:59:26 <AndyS> So can we (1) write proposals (we have some already in email) [is www-rdf-rules OK?] and (2) apply to our systems?

15:59:26 <jang> hence, isomorphism to check results

15:59:27 <danbri> Next week is W3C tech plenary, some of us might be in Boston f2f, others won't be. How about 2 weeks from now, same place / time?

15:59:44 <danbri> www-rdf-rules is the best place for this, yes.

15:59:44 <libby> danbri's suggestion ok by me

15:59:59 <AndyS> jang - yes - that is what I meant : http://lists.w3.org/Archives/Public/www-rdf-rules/2003Feb/0005.html : bag was for RQL.

16:00:00 <Jeen> thu 13 right? ok with me

16:00:11 <danbri> or mailto:www-rdf-rules+query-lovein@w3.org would probably work too

16:00:19 <libby> :)

16:00:21 <areggiori> I will send an email to ww-rules about some alternative RDF/XML format to Libby's

16:00:30 <libby> alberto, can you make 2 weeks?

16:00:31 <danbri> cool, thanks alberto

16:00:37 <libby> jeen? josd?

16:00:44 <areggiori> yes, fine for the next meeting

16:00:51 * JosD will try hard...

16:00:53 <Jeen> yes, I can make it

16:00:55 * AndyS can make 13th - is time OK for people?

16:00:58 <libby> any objections to march 13th at 15:00 GMT?

16:01:05 <timbl> jang, where what was the subject of query -[hasSolution]-> result row

16:01:18 <areggiori> ...checking my iCal

16:01:20 <danbri> Regarding result sets, I may be perverse, but the only XML format I've used for this to date is SOAP graph encoding. I used an array of hashtables, each hashtable (redundantly) having keys as variable names, values as, er, values.

16:01:38 <AndyS> timbl - a bNode? Need something for query-execution-instance

16:01:50 <libby> perhaps we could gather together a bunch of possible resultset by email, as alberto suggess

16:02:12 * timbl 's iCal has a broken hinge and is in repair :-( I will tryto be there anyway

16:02:28 <jang> timbl: andy's proposal was to use a simple property to link from a "result" node to each row; not an ordered collection. The whole thing's then amenable to isomorphism checking for testing result sets

16:02:28 <AndyS> I prefer RDF for result set as checkable.

16:02:47 <timbl> jang: I htnak that if good.

16:02:56 <danbri> I don't think results are intrinsically ordered, so makes sense to me.

16:03:07 <timbl> jang: I think that just using a simple property is good.

16:03:17 <Jeen> I apologize, but I really have to leave now. see you all in two weeks hopefully...

16:03:22 <timbl> But also, just returning the matched graph is good too.

16:03:24 <AndyS> [] rs:hasSolution ... or <> rs:hasSolution .... and multiple hasSolution's

16:03:28 <areggiori> no Seq(s), repeated props are good

16:03:37 * AndyS notes timbl does not like "has" :-)

16:03:41 <danbri> yup, death to rdf:Seqs

16:03:44 <dajobe> lol

16:03:52 <timbl> Advantage of returning the matched graph is that it can just be incoporated into your knowledge base.

16:03:55 <libby> bye jeen!

16:03:59 <danbri> cu

16:04:15 <AndyS> danbri: are lists OK? I like test caes in same order each run!

16:04:22 <danbri> re matched graph, any friendly API will offer both views...

16:04:26 <AndyS> s/caes/cases/

16:04:28 <danbri> Yes, that makes sense AndyS

16:04:30 <jang> counteradvantage of returning a result set is that the returned graph can be constructed from it; and that testing is simpler

16:04:31 <libby> H:next meeting[13th match, 15:00GMT|http://www.timeanddate.com/worldclock/fixedtime.html?day=13&month=3&year=2003&hour=15&min=0&sec=0&p1=0]

16:04:32 <dc_rdfig> Added comment H14.

16:04:32 * timbl notes N3 has an optional "has" keyword for people who have to have it, but yes, Ithink th estatnadrd is a noun phrase for a property id

16:04:35 <jang> it's a simpler operation

16:04:45 <danbri> I just meant not to suggest that there was an official ordering to the bindings that come back

16:04:58 * JosD ... the result is indeed an RDF graph

16:04:59 <libby> i.e. the ?x ?y ?z etc?

16:05:01 <AndyS> jang - only is query exec'ed locally as well - may ship to remote server with no parsing

16:05:08 <AndyS> s/is/if/

16:05:25 <AndyS> NO ordering

16:05:38 <timbl> So then what is the object of "solution" ?

16:05:45 <areggiori> libby, +1

16:05:51 <AndyS> no ordering to the rows of the table

16:05:53 <danbri> So what happens in the meantime, before next meeting? Libby, do you plan to revise a proposal for manifest format?

16:06:06 <jang> object of "solution" is a "binding"

16:06:13 <libby> erm, I'm not sure if I know how right now

16:06:17 <areggiori> andys, the implementation will do that

16:06:20 <libby> irc is moving too fast

16:06:22 <AndyS> object of solution is a collection of bindings (one row) as name/value pairs.

16:06:36 <timbl> And "binding" mentions variable names?

16:06:57 <jang> actually, as a collection, it doesn't necessarily have to

16:07:03 <AndyS> areggiori - I hope so!

16:07:03 <libby> I could (a) collect togteher various resulsets formats (b) see if we decided anything today about manifests

16:07:18 <libby> I also promised to summarise this discussion to the list

16:07:29 <danbri> yeah, i need to go back and re-read this. I'm willing to have a crack at implementing something if theres a target...

16:07:37 <AndyS> yes - binding has variable name ( I don't like that bit but need to quote the variable so how)

16:07:51 <libby> me too - I tried on an easlier version and it was really helpful

16:07:56 <libby> improved code a lot

16:08:27 <danbri> Has there been any discussion about a manifest format describing not only queries, and test datafiles, but the combination of a query and one or more test files, along with expected results?

16:08:29 <AndyS> I will hope implment it as I need to work on testing. I can change again later to be "std" as it is internal only - not app API stuff

16:08:36 <libby> perhaps we alos need a page, a la http://www.w3.org/2002/12/cal/?

16:08:38 <danbri> that kind of mix'n'match seems useful

16:08:50 <danbri> libby, I could arrange for that...

16:09:08 <libby> danbri, how do you mean - more than opne source to query?

16:09:10 <timbl> So a binding only makes sense then in relation to a specific query. It isn't data which is simply true on its own.

16:09:21 <timbl> So the subject would usefully be the query.

16:09:26 <AndyS> Libby/danbri - you offering to upload stuff?

16:09:49 <timbl> This is very similar to the proof langauge design, I notice.

16:09:57 <AndyS> timbl - subject = query - yes (or query+sources if not in query)

16:09:58 <libby> danbri would need to deal eith it....I can help :)

16:10:00 <danbri> Yes, I mean in my system I want to be able to say: Test-12 is the application of Query-21 to Data-file12, Data-file15, Data-file-17 with RDFS and smushing closures.

16:10:18 <areggiori> I think we should allow in the manifest rdf:resource for input source, the query itself and the output - then use rdf:parseType="Literal" if we want to in-line those in the manifest

16:10:20 <AndyS> It is a proof design - Jos did it that way in euler

16:10:30 <JosD> Andy, what about just saying ?var = object (where = is meaning owl:sameAs)

16:10:49 <danbri> AndyS, yes we can host stuff on w3.org (this I think would be under W3C software license terms btw, so no fears of GPL infection)

16:10:59 <JosD> .. as in http://www.agfa.com/w3c/euler/etc5-proof.n3

16:11:07 <AndyS> Because in RHS of rule it will be substitues for value so "object = object" :-(

16:11:08 <libby> I think v useful to put test there, when we've decided

16:11:08 <areggiori> no GPL please :)

16:11:23 <libby> tests as gpl dont make sense

16:11:34 <jang> yes they do.

16:11:43 <danbri> Yep, we talked about relicensing the UoB Inkling tests under w3c license before, libby...

16:11:47 <jang> that is, you'd see how you could sensibly apply that license

16:11:47 <libby> well they make perfect sense _not_ as gpl

16:11:51 <timbl> danbri, so you want to say <foo.rdf> log:semantics F. F log:conclusion G. <bar.rdf> log:semantics G OWTTE

16:11:51 <jang> but better to be open

16:11:54 <areggiori> BSD-ish, then you can narrow down later on...

16:11:58 <JosD> Andy, there are none at the RHS (in the results)

16:12:10 <AndyS> These tests may be executable so GPL => GPL any system that uses them ????

16:12:14 <libby> I'd probably be w3c alberto

16:12:15 * timbl cannot make meeting on 13th at that time alas, internal project review about RDF in HTML.

16:12:21 <jang> andy: nope.

16:12:50 <jang> they're data, but a reduced bsd-compatible license is most helpful to everyone here I think

16:12:50 <libby> w3c license rather. which is not gpl, and doesnt assign copyright to w3c any more necessarily anyway

16:12:52 <AndyS> Jos - I see now - in your proof style I agree : in my cwm example no - output is rule RHS

16:13:21 <libby> so alberto, does more tahn one source file make sense to you?

16:13:30 <areggiori> fine for the w3c, not a big deal at the moment

16:14:00 <libby> i.e. in teh manifest

16:14:08 <AndyS> More than one source is either (1) combine then query or (2) query both sources (different sites) - no merge

16:14:10 * danbri forgot the project review meeting, I'll try to attend both since only one requires danbri.audio... rescheduling would be a pain

16:14:34 <AndyS> (1) isn't signifacnt functionality but (2) is distributed query

16:14:45 <areggiori> good to have one source, but i am not completely convinced that we could live with only one FROM "table"

16:14:51 * AndyS can make other days as well

16:14:56 <libby> what shall we do about an agenda? shall I see what we discuss on teh mailing list in teh meantime?

16:14:57 * JosD goodbye everyone, see you later

16:15:07 <libby> bye josd, thanks for coming

16:15:07 * AndyS waves bye to Jos

16:15:12 <libby> see you next week

16:15:16 <areggiori> bye

16:15:20 <danbri> yes, mailing list re agenda

16:15:35 * libby notes we're out of time

16:15:52 <AndyS> Can we get a "working draft" style consolidation - lots of ideas in different places at the moment

16:15:57 <danbri> multiple sources useful for me as I'm concerned with graph-merge and identity reasoning features, but maybe thats my own problem...

16:15:58 <libby> sorry if that wasnt very firmly agenda-managed all - it's interesting just to see people chat....

16:16:16 <libby> andys, a page on w3c for that might be useful

16:16:23 * jang wants to ask you about identity reasoning danbri, once we're done here

16:16:33 <libby> or, I could tryi to summarise to teh wiki

16:16:33 <danbri> i think we'll get a bit more focus, was good to have everyone together for starters, hence a bit of brainstorming inevitable

16:16:45 <danbri> yeah, wikis there for anyone to make notes in...

16:16:52 <libby> hope so :)

16:16:54 <areggiori> danbri, I am also investigating how to do smushing for some apps of mine - i.e. I need that

16:17:02 <AndyS> I was thinking not of a page of links but a "working proposal" - one place for people to comment as working consensus + email for new stuff

16:17:13 <libby> did you see the smushing tests alberto?

16:17:28 <AndyS> It is the agreed bit - not the discussion area

16:17:36 <areggiori> libby, can you elaborate?

16:17:36 <libby> andys, that would be good, unsure how to manage it

16:17:43 <jang> danbri: you said,"smushing closure rules". Do you actually use closure rules? Ie, generate the missing triples that are a consequence of sameThingAs?

16:17:50 <libby> damian steer came up with some tests. hang on a sec

16:17:51 <AndyS> I propose Libby is editor

16:17:56 <AndyS> :-)

16:17:57 <libby> hehhe

16:17:58 <jang> seconded

16:17:59 <danbri> nope, i was handwaving...

16:18:01 <jang> all opposed?

16:18:05 <jang> carried unanimously

16:18:09 * danbri proposes libby does all the work

16:18:09 <jang> well delegated andy

16:18:14 <jang> seconded

16:18:16 <jang> all opposed?

16:18:17 <jang> carried

16:18:23 * danbri proposes AndyS for delegatoror

16:18:24 <libby> dammit

16:18:25 <jang> nice one dan

16:18:31 <areggiori> yes, fine for Libby's as editor :-)

16:18:35 <AndyS> Jang carried out on stretcher?

16:18:40 <libby> well thanks :) I think.....

16:18:48 <AndyS> Seriously - Libby - would you be editor?

16:19:05 <libby> I guess it counts under swad-e :)

16:19:06 * jang learned everything about process from Brian :-)

16:19:07 <areggiori> but i can help out from next week (not home at the moment)

16:19:07 <AndyS> Say "no" if you want to

16:19:20 <danbri> Yeah, this can count as swad-europe work, definitely

16:19:21 <AndyS> (say "no")

16:19:42 <danbri> so long as we give preferential treatment to Europeans when scheduling meetings...

16:19:53 * danbri ducks, apologies to the californians...

16:20:23 <libby> I'll give it a go; I might (will) need help though...

16:20:24 <AndyS> Unusual approach then danbri (was on phone 6-9pm last week - twice)

16:20:40 <AndyS> We will ALL (hint hint) help

16:20:42 <libby> and most of the work will be done here or on the list

16:21:06 * libby wonders if andys woudl liek to be coeditor

16:21:07 <AndyS> Sounds good - I just think we shopud gather together and not leave scattered for future reference

16:21:23 <danbri> excellent :)

16:21:29 * AndyS should have seen that coming :-)

16:21:32 <libby> heheh

16:21:34 <danbri> AndyS, do you have CVS access to w3.org?

16:21:50 <AndyS> danbri - no idea

16:22:00 <danbri> You'd know if you did, believe me... ;)

16:22:10 <AndyS> What does it take? Did it get set up for SWAD-e (which does not fund me)?

16:22:15 <libby> the approach we took with the calendar schema, was to annouce changes to the list, and if anyone creams, rol back. we could do this with this doc too.

16:22:25 <libby> erk, screams not creams :)

16:22:31 <danbri> It used to be pretty fiddly, but should be relatively straightforward now. I'd need an SSH1 key (RSA?). Can dig out the details...

16:22:57 <AndyS> ssh works just great from here (I use it to SourceForge)

16:23:01 <libby> alberto, if this something you would be interested in too?

16:23:24 <areggiori> ok fulks, I am about to quit here

16:23:26 <areggiori> yes, why not ? ;-)

16:23:31 <libby> it would be particularly good if we could link in the tests with the work you did on examples.

16:23:33 <libby> with andy

16:23:36 <danbri> AndyS, if you're gonna co-edit this w/ Libby, I'll take an [action] to get you SSH/CVS access to a shared workspace...

16:23:52 <libby> danbri, if there any reason why alberyo shoudl not also?

16:23:53 <AndyS> can we divide up the problem space and get small work areas / task with people to push at them?

16:24:06 <AndyS> danbri - OK then.

16:24:21 <libby> (alberto, are you a member of a member organisation?)

16:24:25 <areggiori> I can work to intergrate/link the manifest with the RDF Q&R survey

16:24:25 * AndyS waves to Alberto - thanks for being here

16:24:26 * danbri wonders whether to extend that to alberto too... yes could do that, might need to jump through an extra hoop there if Alberto's not working for a W3C Member

16:24:54 <areggiori> I am a partner of a company now - not working for EC anymore

16:25:08 <AndyS> "Editors" and "CVS people" can be different

16:25:13 <danbri> Alberto, at this point I ought to be evangelising the merits of your company joining W3C. Come on, give us $5k...

16:25:19 <AndyS> if enough "CVS people"

16:25:35 <libby> well danbri, will you take an action to see if possible for alberto too, as invited expert maybe?

16:25:48 <libby> andys true

16:25:53 <timbl> Andys?

16:25:54 <danbri> OK, 2 editors seems plenty for hopefully a small doc, and yes we can make Alberto a CVS person... might need to do invited expert thing.

16:26:04 <areggiori> I will talk about with the others and write to you about it

16:26:14 * danbri hopes to be a CVS person too

16:26:14 <AndyS> I was thinking other way round, danbri

16:26:19 <libby> ok, liely that this one will run and run anyway

16:26:34 <danbri> okay, we can carve it up one way or the other... what did you have in mind Andy?

16:27:06 <libby> I think he meanty 3 of us editors, without nnec having to have cvs access for alberto if a problem

16:27:06 <AndyS> Nothing specific - just we are all busy and need to keep some momentum up else it will die

16:27:23 <danbri> ah ok

16:27:26 <danbri> fine by me

16:27:32 <libby> me too

16:27:37 <areggiori> I am happy to be a CVS persion or helper, whatever

16:27:38 <AndyS> Maybe (1) manifest (2) results (bindings, graph(s) and (3) query defn (hmm to last)

16:27:49 <AndyS> Write vocabaulries!

16:28:01 <areggiori> yes right :-)

16:28:45 * AndyS thanks Libby for organising this chat

16:28:48 <timbl> Write software as you go.

16:29:03 <libby> no prob, thank everyone for coming!

16:29:08 <danbri> yeah, cheers lib

16:29:12 <dajobe> all done?

16:29:12 <libby> yep to software

16:29:15 <danbri> and everyone for coming

16:29:17 <timbl> Yes, thanks Libby that was useful. A little fast sometimes, brainwhirring.

16:29:19 <areggiori> thanks to Libby for meeting and everybody else for the nice chat

16:29:22 <AndyS> I will try to extract the testing from RDQL/Jena (not yet RDF-ised)

16:29:31 * danbri needs to re-read it all :)

16:29:47 <areggiori> me too! :-)

16:29:52 <libby> I will try to summaraise this discussion to list, start a page about what we are doing (maybe in wiki for now)

16:29:57 <libby> me three :)

16:30:02 <dajobe> dajobe has changed the topic to: RDF & SemWeb hack-n-chat - http://rdfig.xmlhack.com/

16:30:18 <areggiori> talk via email fulks - cheers

16:30:27 <libby> bye now!

16:30:28 <AndyS> CU

16:30:31 <areggiori> s/fulks/folks/

16:30:32 <areggiori> ciao

16:30:36 <dajobe> I was wondering (given *sufficient* notice) of adding something to the chump page about upcoming chats

16:30:37 * libby tired now

16:30:42 <AndyS> s/fulks/folks/g !

16:30:55 <libby> good idea dajobe

16:31:16 <jang> danbri: still awake?

16:31:20 <libby> since otherwise regular irc people not on the list dont know what's going on

16:31:30 <danbri> yup

16:31:30 <libby> oh, smush tests

16:31:34 <jang> smushing...

16:31:39 <danbri> (tm)

16:31:41 <jang> ...do you really do a closure?

16:31:44 <jang> ie,

16:31:50 <jang> if you have a p y .

16:31:57 <jang> and you know x and y are the same thing

16:32:00 <jang> do you add

16:32:01 <jang> a p x

16:32:03 <jang> as a closure?

16:32:25 <libby>http://swordfish.rdfweb.org/rdfquery/tests/rdf/

16:32:26 <dc_rdfig> L: http://swordfish.rdfweb.org/rdfquery/tests/rdf/ from libby

16:32:36 <libby> L:|some rdf tests

16:32:37 <dc_rdfig> Titled item L.

16:33:09 <danbri> i probably would, except i don't. what i do do, currently, is a bit monstrous. I rewrite the identifiers in the aggregate graph...

16:33:17 * mattb does that too

16:33:23 <libby> L:shellac ones are smush tests. shoudl be used with [manifest file|http://swordfish.rdfweb.org/rdfquery/tests/rdf/query-results-manifest-nt.rdf]

16:33:23 <dc_rdfig> Added comment L1.

16:33:27 <danbri> ...even (bad bit) losing URIs, if the two folded nodes both have URIs

16:33:32 <jang> so you're guilty of confusing abstract syntax and inference layers then? :-)

16:33:37 <mattb> then rewrite triples with those identifiers as subject or object

16:33:43 * libby doesnt

16:33:45 <danbri> I intend to preserve the uris by attaching them as value of URI properties

16:33:57 * mattb hasn't finished writing the smusher yet, tho :)

16:34:05 <danbri> Nope, I infer a new graph, basically.

16:34:05 <mattb> it's a voyage of discovery

16:34:18 <mattb> for use internal to the smushing app, a new graph seems reasonable

16:34:20 <mattb> i retain provenance

16:34:27 <jang> <danbri> I intend to preserve the uris by attaching them as value

16:34:27 <jang> of URI properties

16:34:28 <danbri> It just happens to be mostly the same SQL DB that implements the new graph.

16:34:35 * libby uses internal ids which are never known to the outside world, so that smushing doesnt cause information loss. it'sa a pain to implementa nd sloe though

16:34:51 <jang> you mean, if x = y...

16:35:03 <jang> x <uri> uri:y ?

16:35:11 <jang> or

16:35:12 <jang> x

16:35:21 <jang> x <uri> "y"^^xsd:uri ?

16:35:30 <jang> or x <uri> y ?

16:35:56 * danbri tries to read that

16:35:58 <jang> or x <owl?:sameThingAs> y ?

16:36:01 <danbri> I mean

16:36:15 <jang> if x and y are uris that denote the same thing then...

16:36:18 <jang> (waits)

16:36:25 <libby> L:|some RDF query tests

16:36:25 <dc_rdfig> Titled item L.

16:36:32 <danbri> <Person rdf:about="http://example.org/uri1" foaf:dnachecksum="12345" foaf:name="Jan Grant"/>

16:36:51 <danbri> <Person rdf:about="http://example.org/uri2" foaf:dnachecksum="12345" foaf:myersBriggs="????"/>

16:37:34 * jang has his DNA checksum as the number on his luggage. Who'da thunk it?

16:37:45 <danbri> I'll end up with a node that has both property sets, ie the name and the myersBriggs, and also two 'log:uri' or whatever properties, one with value "http://example.org/uri1" the other "http://example.org/uri2" (yes, xsd:uris)

16:37:53 <libby> L1:shellac ones are smush tests

16:37:53 <dc_rdfig> Replaced comment L1.

16:39:09 * jang wonders if clark kent and superman would have the same dna checksum

16:39:11 <danbri> jang, see http://www.w3.org/2001/12/rubyrdf/pack/lib/squish.rb under def defrag for what I (shudder) actually do.

16:39:51 <AaronSw> well, if one knew clark kent's dna one could reasonably conclude that something was up ;-)

16:40:08 <jang> it's hard to get a sample

16:40:34 <jang> but if I knew his DNA checksum then I could start randomly cloning kryptonians until I got a copy

16:40:47 <jang> and the odds would favour it being superman I suppose.

16:41:06 <jang> they'd be kal em, kal en, etc.

16:41:46 <danbri> jang, do you know where I can find (eg. from literature, not by brute force) and example of two different chunks of data with the same sha1sum?

16:42:12 <danbri> ie. its known to be possible, but v hard to discover. are there some known examples anywhere?

16:42:38 <AndyS> Constructed or accidental hash clash?

16:42:49 <danbri> context: I currently say foaf:mbox_sha1sum is an owl:InverseFunctionalProperty, but thats bending truth slightly

16:43:14 <danbri> constructed or accidental, I was just looking for an example of a clash to use in a test case

16:43:28 <AndyS> It will depend on the reliablity needs of API - I want more proof than SHA1 before sending $1B

16:43:51 * danbri nods

16:43:56 <AndyS> Hypotetically - I don't have $1B

16:44:03 * danbri guessed ;)

16:44:04 <AndyS> Its the "ebay" problem

16:44:41 <AndyS> Maybe we could label with the implications as vocabulary annotation and let app decide

16:44:44 <AaronSw> danbri, no, there are no known examples. if there were, sha1 would be considered broken

16:44:54 <jang> dan, as far as I'm aware, the sha1 space is pretty uniform (ie, no weak hashes). I don't know that anyone's published a clashing could of plaintexts

16:45:25 <jang> couple, even

16:45:25 <AaronSw> there are some partial examples for MD5 (comes out the same after a few rounds, but not the full hashing process)

16:45:52 <danbri> But we know such examples exist. That is no suprise. So hearing of a specific clash shouldn't shatter our confidence. Hearing of a million clashes would be alarming, though...

16:46:02 <jang> really? why

16:46:11 <jang> 1,000,000 out of 2^160 isn't very big

16:46:11 <danbri> or billion...

16:46:13 <AaronSw> Well, apparently the way the system works is that if you know one clash you're able to generate a million

16:46:19 <danbri> or zillion.... ;)

16:46:26 <jang> yes, it's ablock hash

16:46:33 <danbri> ah, didn't know that.

16:46:44 <jang> if two prefixes hash to the same, you can just chuck the same suffix on and you'll get the same result, I think...

16:46:50 <AaronSw> jang, it'd be alarming because it would mean people could twiddle bits to cause clashes

16:46:52 <jang> that is, all the state ends up in the #

16:47:12 <jang> aaron: I wouldn't be alarmed. i use crc16 for all my vital information

16:47:22 <jang> :-)

16:47:38 <AaronSw> all the cool kids have upgraded to sha512 :)

16:48:04 <AaronSw> actually, I'm told that the SHAN, N>1 series is designed to be much more secure than the MD*, SHA1 family

16:48:54 <jang> really?

16:49:13 <AaronSw> yeah

16:49:18 <jang> how can it be secure against dumb luck?

16:49:58 <AaronSw> obviously it can't be secure against dumb luck, but it can be free of specific weaknesses in the MD5 hash that lower the amount of luck required

16:50:32 <jang> I'll only settle for _damn_foolproof solutions

16:50:50 * danbri is a damn fool

16:50:52 <jang> (as in, "you can make it foolproof, but...")

18:34:16 <eikeon> Anyone know of a property to use for storing info about parse error in a document. <http://...> ??? "info about error".

18:34:59 <danbri> nope. i'd attach it to a ParseEvent instead of the doc itself, easier if doc changes, gets fixed etc.

18:36:11 <eikeon> Sounds like a good idea. Do you happen to have such terms defined?

18:36:13 <mortenf> yeah, a lot of stuff depends on time...

18:36:55 <danbri> i don't have such a vocab, but i've often advocated event-centric modelling

18:37:05 <danbri> its probably time to start putting theory into action

18:37:14 <danbri> .google abc model

18:37:15 <datum> abc model: http://www.geocities.com/moto_bike/

18:37:19 <danbri> .google abc model rdf brickley

18:37:19 <mortenf> cyc does have some events.

18:37:20 <datum> abc model rdf brickley: http://jodi.ecs.soton.ac.uk/Articles/v02/i02/Lagoze/

18:37:46 <danbri> nah, not so keen on that version...

18:38:47 <danbri> ah http://ilrt.org/discovery/harmony/docs/abc/abc_draft.html is woolier and simpler

18:40:49 * mortenf added that to list of reading material...

18:41:49 <danbri> its pretty old, a quick skim would do it, don't worry about the detail (we didn't ;)

18:42:09 <danbri> one goal was to motivate RDF rules stuff to a Dublin Core-ish audience...

18:42:49 <mortenf> k, but I already have quite a backlog of quick skims (which is also what brought it about in the first place)...

18:43:06 <danbri> heheh

18:43:16 <mortenf> it's hard.

18:43:22 <eikeon> "Modification event => VersionOf relation" -- The resource I was looking to add a parse error property to is already a version of relation. So I may already be doing something similar.

18:43:56 <mortenf> when you think you understand something, you've really just uncovered 5 other things you don't understand, and that needs to be understood, etc.

18:43:58 * danbri tries to remember...

18:44:08 <danbri> thats what i love about all this...

18:44:18 <mortenf> yeah, but...

18:44:28 <danbri> but then i don't have a day job that requires me to actually build anything, any more...

18:44:40 <mortenf> lucky you!

18:44:50 <danbri> (not that I don't care, but yes, I remember trying to keep up in my spare time was insane...)

18:45:07 <mortenf> :)

19:04:35 * rajiv is back (gone 10:02:14)

19:50:59 * rajiv is away: phone call

20:20:53 <soccos|away> soccos|away is now known as soccos

20:24:01 * rajiv is back (gone 00:33:03)

20:54:30 * rajiv is away: gone

21:17:01 * rajiv is back (gone 00:22:31)

21:19:05 <grault> grault is now known as earle

21:19:18 <earle> good evening all/anyone

21:19:56 <soccos> soccos is now known as soccos|away

21:31:50 <Jibbler> Jibbler is now known as Jibbler|out

22:16:07 <LTjake> Hello.

22:17:37 <earle> Howdy.

22:19:47 <LTjake> anyone have any experience with RDF->RDBMS?

22:20:46 <earle> Nobody seems to be around at the moment...

22:21:28 <dajobe> LTjake: funny you should ask

22:21:47 <dajobe> try http://www.w3.org/2001/sw/Europe/reports/scalable_rdbms_mapping_report/

22:22:05 <dajobe> especially sec 5 & 6

22:22:18 <LTjake> Hey, thanks!

22:23:01 <dajobe> that was the plan; to answer these FAQs

22:30:03 <earle> earle is now known as grault

22:31:59 <LTjake> dajobe: aren't 3 & 4 more what i need? =)

22:32:06 <dajobe> possibly

22:32:15 <dajobe> but you want rdf to rdbms

22:32:23 <dajobe> mostly the 3&4 are the other way

22:33:30 <LTjake> oh? it's not storing RDF data in a DB?

22:33:40 <dajobe> it is

22:33:49 <dajobe> well, read the lot then and work it out :)

22:34:15 <dajobe> plenty of experience in those schemas

22:34:24 <LTjake> heh.

22:34:28 <dajobe> and even better, they are already implemnented so you don't have to

22:36:19 <LTjake> man, why are so many of these things in java? :)

22:36:37 <dajobe> lol

22:37:22 <LTjake> I'm stuck in ASP & MS SQL.

22:37:41 <dajobe> well at least you can get the schema ideas

22:37:49 <LTjake> *nod*

22:38:08 <LTjake> I have something implemented already. but it's tres basic.

22:39:32 <LTjake> Basically, i have book resources (from MARC records), and i want to store metadata on them for searching.

22:40:19 <LTjake> i've made a basic subject, predicate, value table where subject is always an id (to reference another table), and predicates are generally dublin core.

22:40:36 <LTjake> i can do easy searches on dc:title and whatnot.

22:40:58 <dajobe> right

22:42:48 <LTjake> Of course, now i'm looking into importing different collection -- and people want different sets of metadata...

22:44:13 <dajobe> that's the general idea, fixed property(column) schemas have limitations here, although optimised nicely for predicates you do know in advance

22:44:42 <dajobe> well, have fun reading, it's the end of my day...

22:44:43 <dajobe> bye

22:44:50 <LTjake> thanks again.

22:44:51 <LTjake> cya.