Semantic Web Interest Group IRC Chat Logs for 2005-09-08

00:00:34 <danbri> done

00:06:30 * danbri & # zzz

00:50:18 <crschmidt> danbri: for the record, whatever you did didn't kick it

04:28:15 <Nick1> anyone by the keyboard here?

07:12:19 <michel_p_> michel_p_ is now known as michelp

11:53:47 * ericP sees some life on the ESW Wiki page "SOAP Encoding RDF"

11:54:08 <ericP> hmm, "brunette anal teen"

11:54:19 <ericP> i don't recall puting that there. do you danbri?

11:54:28 <crschmidt> I admit, it was me.

11:54:40 <ericP> ahh, thanks for fessing up

11:54:42 <crschmidt> I thought I'd spice up the life of the W3C with some random porn

11:55:44 <ericP> so is "celebrities lesbian" a codepiction experiment?

11:56:14 * ericP worked hard to find one that wasn't too profane for this hour

11:57:21 <_aniasis> _aniasis is now known as aniasis

11:57:54 <crschmidt> heh, heh

12:03:48 <crschmidt> I'm considering dumping out Julie's current database and starting anew. Wonderign if anyone has any thoughts on that

12:04:13 <ericP> julie, how do you feel about that?

12:04:51 <crschmidt> ^status

12:04:53 <julie> I currently hold 2118666 triples, 100 namespaces, and 69 stored commands. I have been running for 18.866 hours, and have used 0.063 minutes of CPU time. Stats on current machine (debian.localdomain): 08:05:08 up 14 days, 13:41, 8 users, load average: 0.03, 0.06, 0.07

12:05:02 <crschmidt> that's not very informative, julie

12:06:45 <Pike_> crschmidt: what did you expect? 90-60-90?

12:08:18 <crschmidt> is that the non-amemrican version of 36-24-36?

12:08:43 <Pike_> it's the SI version of it

12:09:23 * crschmidt nods

12:09:30 <Pike_> the american version could use the same number in inches, though, I guess

12:09:32 * Pike_ ducks

12:10:02 * ericP sends someone to sit on Pike_

12:10:17 <Pike_> ouch

12:10:36 <ericP> go ahead, try and get out from under that 90

12:13:17 * Pike_ melts home

12:27:09 <crschmidt> Hm

12:27:13 <crschmidt> sparql.org not happy

12:27:27 <crschmidt> http://sparql.org/sparql?query=prefix+%3A+%3Chttp%3A%2F%2Fwww.openmobilealliance.org%2Ftech%2Fprofiles%2FUAPROF%2Fccppschema-20021212%23%3E%0D%0Aprefix+rdf%3A+%3Chttp%3A%2F%2Fwww.w3.org%2F1999%2F02%2F22-rdf-syntax-ns%23%3E%0D%0A%0D%0ASELECT+%3FTY+where+%7B+%3FP+a+%3ASoftwarePlatform%3B+%3ACcppAccept+%5B+%3FLI+%3FTY+%5D+%7D%0D%0A&default-graph-uri=http%3A%2F%2Fwww.blackberry.net%2Fgo%2Fmobile%2Fprofiles%2Fuaprof%2F7280%2F3.3.1.rdf&stylesheet=xml-to

12:27:34 <crschmidt> Linked from http://esw.w3.org/topic/DawgShows

12:36:51 <dajobe> I guess the api changed

12:38:31 * crschmidt despams the wiki

12:44:50 <dajobe> and the spammer returns

13:50:43 <DanC> crschmidt, pls save the list of URIs that julie has read somwhere

13:50:52 <crschmidt> DanC: it's already in www-archive

13:51:04 <DanC> ah. excellent.

13:51:31 <crschmidt> Or should be, anyway: can't remember when I sent it off the top of my head

13:52:38 <crschmidt> http://lists.w3.org/Archives/Public/www-archive/2005Apr/att-0003/files

13:53:11 <crschmidt> Also, I'd keep the data around, possibly even still have web access to it

13:53:22 <crschmidt> but there's a large chunk of data which was added to julie solely for the sake of increasing the triple count

13:53:31 <crschmidt> and which has never been queried against

15:45:45 <DanC> timbl, I'm interested to discuss recent comments on SPARQL. Do you have bandwidth now?

15:46:03 <timbl> OK

15:46:31 <DanC> oops... now I need good pointers...

15:47:19 <DanC> pfps's comments have lots of good examples... http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2005Sep/0036.html

15:47:26 <DanC> BLURB: recent feedback on SPARQL

15:47:26 <dc_swig> A: recent feedback on SPARQL from DanC

15:47:50 <DanC> A: [http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2005Sep/0036.html|pfps's comments] have lots of good examples

15:47:50 <dc_swig> Added comment A1.

15:48:17 <DanC> do you see the 1st example, starting with G1, timbl?

15:49:08 <timbl> Yes.

15:49:45 <timbl> It is not friendly to replt with two results, one of which entails the other.

15:50:26 <DanC> ok, so you're swapped in. PFPS, Enrico, and Bijan have all argued that it's not friendly.

15:50:40 <timbl> I have been thinking about this, in writing up N3 logic.

15:51:53 <timbl> I'd agree. Cwm actually tests* before it adds some bound conclusion from a rule to the knowledge base that the KB doesn't already entail that result. This saves some of that. * used I think still does!

15:53:13 <DanC> right... PFPS points out that this is called "leaning" in RDF semantics, and that as SPARQL is defined today, and implementation that leans won't interoperate with one that doesn't.

15:53:39 <DanC> so you have some implementation experience with it. Good. that was a concern of mine.

15:54:03 <DanC> Andy had some concerns about streaming... and I wonder about how it works with SQL backends.

15:54:54 <DanC> next we have the issue of OWL entailments that don't fit the log:conclusion model...

15:55:10 <DanC> logger, chump A:

15:55:11 <DanC> A:See [http://www.ilrt.bris.ac.uk/discovery/chatlogs/swig/2005-09-08#T15-55-10|discussion]

15:55:11 <dc_swig> Added comment A2.

15:56:21 <DanC> A:[http://lists.w3.org/Archives/Public/public-rdf-dawg/2005JulSep/0376.html|jos's msg on subgraph/entailment, including a prover9 proof of the OWL disjunction worker example]

15:56:21 <dc_swig> Added comment A3.

15:57:49 * DanC waves to bijan, EliasT , invites them to catch up from logs of the last few minutes

15:58:14 * DanC hunts for a better statement of the worker example....

15:58:20 * EliasT is reading the logs as we speak.

15:59:00 <DanC> A:<timbl> It is not friendly to reply with two results, one of which entails the other.

15:59:00 <dc_swig> Added comment A4.

15:59:56 <bijanp> logger, pointer

15:59:56 <bijanp> See http://www.ilrt.bris.ac.uk/discovery/chatlogs/swig/2005-09-08#T15-59-56

16:00:01 <timbl> The query in that case is defined to be on the closure under OWL DL inference of the dataset?

16:00:03 <DanC> A:[http://lists.w3.org/Archives/Public/public-rdf-dawg/2004JulSep/0069|the worker example in Enrico's msg from July 2004]

16:00:03 <dc_swig> Added comment A5.

16:00:38 <DanC> well, sorta, except that "the closure under OWL DL inference" is an ill-formed definite description. There is no single graph that is the closure under OWL DL inference.

16:01:07 <AndyS> Sometimes you want the two results (e.g. to add a triple with the bnode later) even if the better information answer is lean

16:01:27 <dajobe> (lean: http://www.w3.org/TR/2004/REC-rdf-mt-20040210/#deflean )

16:01:55 <DanC> A:[http://www.w3.org/TR/2004/REC-rdf-mt-20040210/#deflean definition of lean from RDF Semantics]

16:01:55 <dc_swig> Added comment A6.

16:02:02 * bijanp qualifies his "not friendly" point...it can be friendly if you are manipulating the abstract syntax of the document

16:02:27 <DanC> A6:[http://www.w3.org/TR/2004/REC-rdf-mt-20040210/#deflean}definition of lean from RDF Semantics]

16:02:27 <dc_swig> Replaced comment A6.

16:02:33 <DanC> A6:[http://www.w3.org/TR/2004/REC-rdf-mt-20040210/#deflean|definition of lean from RDF Semantics]

16:02:34 <dc_swig> Replaced comment A6.

16:02:44 <AndyS> An OWL editor for example :-)

16:03:11 <DanC> yes, there are 2 sides to the "not friendly" bit. I just wanted to note that timbl was sympathetic with at least one of them... in particular, the one that's different from SPARQL LC.

16:03:15 <bijanp> Yes, though I'm willing to believe that sparql is the wrong tool for a remote editor, for owl at lest

16:03:38 <bijanp> maybe "inclined" is better

16:03:41 <AndyS> This applies to local as well as remote

16:03:47 <bijanp> yes

16:04:04 <bijanp> s/remote// :)

16:04:20 <AndyS> Remote has the serilization anonymising the bnodes anyway

16:04:21 * tlr waves from the ground floor of HP Labs

16:04:32 * DanC thanks the planets for lining up and getting lots of interested parties here now.

16:04:40 <bijanp> Thinking about owl editors....there are two that I know off of hand that make use of triple representations directly

16:04:43 <bijanp> Or that I beleive

16:04:48 <bijanp> Triple20 certainly does

16:04:51 <bijanp> pOWL probably does

16:05:07 <DanC> tlr, note that UNSAID is at risk; we're getting request for formal soundness/completeness semantics, which suggest to me that I should re-open the SOURCE and UNSAID issues.

16:05:24 <DanC> bijan, pointers, please?

16:05:27 <bijanp> Protege, Swoop, OilED all certainly don't

16:05:35 <bijanp> Sure...

16:05:42 <bijanp> Maybe a separate chump

16:05:44 <DanC> I know where to find protege and swoop, and maybe oiled

16:05:48 <DanC> sure, separate, either way

16:06:03 <AndyS> UNSAID is already out isn't it?

16:06:03 <bijanp> BLURB: Editors and their underlying toolkits (triples or other)

16:06:03 <dc_swig> B: Editors and their underlying toolkits (triples or other) from bijanp

16:06:20 <tlr> UNSAID is out, but you can emulate is with !isbound.

16:06:27 <tlr> DanC, any pointer to that request?

16:06:30 <EliasT> A:[http://www.swi-prolog.org/packages/Triple20/|Triple20 -- An RDF/RDFS/OWL visualisation and editing tool]

16:06:30 <dc_swig> Added comment A7.

16:06:37 <DanC> UNSAID is the issue name. ISBOUND is the language feature that we used to address it.

16:06:57 <bijanp> B:|[Protege|http://protege.stanford.edu/] uses internal format (and DIG to interact with remote reasoners]

16:06:57 <dc_swig> Titled item B.

16:07:02 <bijanp> sigh

16:07:06 <bijanp> been a while

16:07:13 <bijanp> B:[Protege|http://protege.stanford.edu/] uses internal format (and DIG to interact with remote reasoners]

16:07:14 <dc_swig> Added comment B1.

16:07:28 <bijanp> B:| OWL editors and their underlying toolkits (triples or other)

16:07:28 <dc_swig> Titled item B.

16:08:12 <bijanp> B:[http://www.swi-prolog.org/packages/Triple20/|Triple20] uses triples directly. (So it is an OWL editor by editing the RDF triples of an OWL document)

16:08:12 <dc_swig> Added comment B2.

16:08:23 <EliasT> B:[http://powl.sourceforge.net/|pOWL - Semantic Web Development Plattform]

16:08:23 <dc_swig> Added comment B3.

16:08:34 <DanC> A:[http://lists.w3.org/Archives/Public/public-rdf-dawg/2005JulSep/0330.html|Enrico's comments] ask SPARQL core with formal semantics exclude GRAPH/BOUND, which suggest re-opening the [http://www.w3.org/2001/sw/DataAccess/issues#SOURCE|SOURCE] and [http://www.w3.org/2001/sw/DataAccess/issues#UNSAID|UNSAID] issues.

16:08:34 <dc_swig> Added comment A8.

16:08:58 <DanC> A8:[http://lists.w3.org/Archives/Public/public-rdf-dawg/2005JulSep/0330.html|Enrico's comments] ask for a SPARQL core with formal semantics exclude GRAPH/BOUND, which suggest re-opening the [http://www.w3.org/2001/sw/DataAccess/issues#SOURCE|SOURCE] and [http://www.w3.org/2001/sw/DataAccess/issues#UNSAID|UNSAID] issues.

16:08:58 <dc_swig> Replaced comment A8.

16:09:15 * DanC is interested to know if timbl has swapped in the OWL DL worker example

16:09:54 <bijanp> B:I believe pOWL works on triples but don't know for sure.

16:09:54 <dc_swig> Added comment B4.

16:10:30 <bijanp> B:pOWL being a database based editor might be the best example for SPARQL.

16:10:31 <dc_swig> Added comment B5.

16:10:40 * AndyS notes Bijan's email "I think there is no question about this [having syntactic query/ed]. Or, I hope there isn't :)

16:10:40 <AndyS> "

16:10:53 <bijanp> I think it's important in general

16:11:03 <bijanp> We have a "no reasoner" mode in swoop for example

16:11:33 <bijanp> And without an insert/delete modal in genearl, the told triple version is easiest

16:11:57 <timbl> The OWL DL worker example is then a question about OWL reasoning or about SPARQL?

16:11:58 <tlr> thanks danC

16:11:59 * DanC can't parse "modal in genearl". resend?

16:12:17 <bijanp> B:[Swoop|http://www.mindswap.org/2004/SWOOP/] uses the OWL API, not triples

16:12:18 <dc_swig> Added comment B6.

16:12:31 <DanC> The OWL DL worker example shows that inference and query aren't separable in all cases.

16:13:31 <DanC> i.e. you can't decompose the worker problem into (a) compute the log:conclusion and (b) do a query on the resulting graph.

16:13:51 <bijanp> B:[OilEd|http://oiled.man.ac.uk/] I believe, with very good cause, to not use triples. It uses DIG to interact with remote reasoners.

16:13:52 <dc_swig> Added comment B7.

16:14:28 <bijanp> B:Note that Swoop, Protege, and OilEd all have a base vs. full inference mode

16:14:29 <dc_swig> Added comment B8.

16:14:56 <bijanp> bijanp is now known as bijan

16:15:12 <DanC> B:cf [http://www.w3.org/TR/rdf-dawg-uc/#u2.17|SPARQL use case 2.17 Building Ontology Tools]

16:15:13 <dc_swig> Added comment B9.

16:16:16 <DanC> A:The worker example seems to conflict with [http://www.w3.org/2003/12/swa/dawg-charter#rdfs-owl-queries|section 2.1 Specification of RDF Schema/OWL semantics of the DAWG charter], "if OWL DL semantics are supported by a service, that may be evident in the description of the service or the virtual graph which is queried, but it will not affect the protocol designed under this charter.'

16:16:16 <dc_swig> Added comment A9.

16:17:20 <bijan> B:[OntoTrack|http://www.informatik.uni-ulm.de/ki/ontotrack/] I believe and would be shocked if it used triples directly. It doens't have a "told" mode. All editing is done with full reasoning all the time.

16:17:20 <dc_swig> Added comment B10.

16:17:27 <bijan> B:Which, I've said to the authors, is nuts :)

16:17:28 <dc_swig> Added comment B11.

16:18:08 <DanC> eek... that's like XML editors that never allow you to get the document into an invalid or non-well-formed state.

16:18:14 <bijan> Yep

16:18:21 <DanC> they don't exactly build huge user bases quickly ;-)

16:18:23 <bijan> Worse :)

16:18:32 <bengee> B:afaik, pOWL uses [http://www.wiwiss.fu-berlin.de/suhl/bizer/rdfapi/|RAP]'s persistence layer (which is statement-based)

16:18:32 <dc_swig> Added comment B12.

16:18:43 <bijan> You can't edit anything, even if correct, that your reasoner can't handle :)

16:19:20 <timbl> Well, you could argue that paul a Worker; friend [a Employee; freind [ a manager]]. is in the log:conclusion. It just ain't practical to evaluate the whole log:comclusion which would presumably not be finite, but you can still argue for the charter about the set of things which can be deduced, even if it isn't finite.

16:19:46 <bijan> Y'all are replaying my argument with enrico, and pat's argument with me

16:19:53 <timbl> You might want to rephrase it. but I don't see where you now show the language has to be different in which you ask the question.

16:20:54 <DanC> yes, replaying arguments in various overlapping communities is a big part of the consensus process.

16:20:56 <bijan> Note that the deductive closure argument just reintroduces entailment

16:21:10 <DanC> "the deductive closure argument"?

16:21:28 <bijan> DanC, as is providing continutity to earlier plays :)

16:21:43 <bijan> Sorry, the deductive closure technique

16:22:31 <DanC> the deductive closure technique doesn't introduce entailment into the place that's at issue, i.e. in SPARQL pattern matching.

16:22:51 <bijan> The deductive closure of a theory (in these cases) is exactly the set of entailments for that theory

16:23:03 <bijan> I understand that part of the argument as well

16:23:16 <DanC> yes, but the hypothesis is that SPARQL can be design such that that's Somebody Else's Problem.

16:23:41 <DanC> i.e. the hypothesis in the charter.

16:24:19 <DanC> might want to rephrase which, timbl?

16:25:07 <DanC> I don't think the issue is finiteness, timbl...

16:25:29 * DanC needs a pencil... aka emacs buffer...

16:25:32 <bijan> They I would argue that the hypothesis needs to be tested with one, preferable two, concrete examples

16:25:51 <DanC> yes.

16:26:01 <bijan> In particualr, one with significantly different characterstics

16:26:04 <DanC> I think tests with RDFS are going well, but tests with RDF and OWL are going poorly.

16:26:20 <bijan> (E.g., since RDFS is essetnailly like RDF, you need something that must consider multiple models)

16:26:27 <bijan> Thats probably the biggest distinction

16:26:34 <DanC> right.

16:26:47 <DanC> is paul a Worker; friend [a Employee; freind [ a manager]]. really entailed?

16:27:04 <bijan> Assuming not typos, yes

16:27:05 <bijan> :)

16:27:17 <bijan> From the theory enrico gave

16:27:44 <bijan> You coudl test that in pellet

16:27:49 <bijan> Via the web form

16:27:56 <bijan> You coudl put the entailment in the second box

16:28:15 <DanC> hmm... so there is some intermediate graph that works.

16:28:32 <DanC> but I suppose there's a feedback loop to find out what it is.

16:28:32 <bijan> It depends on what you mean by "intermediate graph"

16:29:09 <bijan> If it includes encodings for such formulae (which is pretty easy in this case) then you are on your way

16:29:26 <bijan> (You do have to worry about those "syntax" triples)

16:30:08 * timbl has to go

16:30:24 <DanC> ok, more later, I hope, timbl.

16:31:47 <DanC> bijan, have you found a design that you like in this entailment space? Do we parameterize queries by a URI that refers to an entailment relationship? Or do servers make that part of their service description?

16:32:43 <DanC> parallel question for bijan, do you know how to specify soundness/completeness for GRAPH and UNBOUND?

16:33:00 <bijan> Er..no and no :(

16:33:04 <bijan> Not yet at least

16:33:13 <DanC> I see.

16:33:18 <bijan> haven't thought about the latter

16:33:28 <DanC> ugh... so if I open these issues, I accept a pretty unbounded schedule risk.

16:34:33 <DanC> hi steve. TimBL's time ran out.

16:34:56 <DanC> the chump is a reasonable summary. http://swig.xmlhack.com/

16:34:58 <swh> Hi DanC, I was actually hoping for a summary :)

16:35:05 <bijan> As for the former I jsut haven't been thinking about those designs

16:35:05 <swh> great

16:35:06 <chimezie> B:->swig,owl,editors

16:35:06 <dc_swig> Set keywords for B.

16:35:21 <bijan> So, it's more that neither of these aspects of the issue have at all come to my attention

16:35:52 <bengee> B:[http://www.appmosphere.com/en-owlchestra|OWLchestra] uses a relational db for OWL models. I managed to build a SKOS editor on top an RDF store + API + SPARQL queries, not sure if that'd work for OWL as well...

16:35:53 <dc_swig> Added comment B13.

16:36:20 <DanC> yeah... and Enrico's investigation concluded with "and I don't see how to give semantics for GRAPH/UNBOUND so let's leave that out of the core for now"...

16:36:43 <DanC> ... what's frustrating to me is that's exactly what I was trying to say to the WG in Helsinki, some months ago when the issue was open.

16:37:11 <DanC> it's really frustrating that we didn't get Enrico's input at the time.

16:37:12 <swh> true enough

16:37:40 <DanC> I tried to ring all sorts of bells about UNSAID and SOURCE; I went to the CG, and I went to the best practices WG and got explicit advice.

16:37:49 <bijan> "run for the hills"?

16:38:07 <bijan> If not, you should have come to *me* for advice

16:38:10 <bijan> I'm full of such advice

16:38:15 <bijan> And it's *allllllll* good :)

16:38:18 <DanC> no; the advice I got was from Welty, who said "yeah, that's pretty mature technolgy. It's what SQL does. It's not too scary."

16:38:56 <swh> I've seen the comments on GRAPH, buts whats with UNSAID?

16:39:27 <swh> I thought we'd agreed not to have UNSAID

16:39:36 <DanC> our solution for UNSAID is !BOUND; note that BOUND is a function of more than just its argument. it's a function of the bindings. it's wierd, formally.

16:40:03 * tlr still thinks UNSAID is cleaner than the !BOUND hack.

16:40:13 <DanC> rather... our solution to UNSAID is !BOUND combined with OPTIONAL.

16:40:29 <DanC> tastes differ on preferred syntax; I'm convinced they're equally expressive.

16:40:33 <swh> right, gotcha (was reading wrong log page)

16:40:52 <tlr> so am I

16:40:55 <tlr> (re convinced)

16:40:55 <DanC> trl, swh, have you guys met?

16:41:03 <tlr> In Boston, I think.

16:41:42 * tlr is in Bristol, but stuck in PRIME meetings until his Taxi to the airport comes.

16:41:47 * DanC wishes he knew more about the market window for SPARQL

16:42:15 <DanC> I know that my personal market window closed in about April. I'm really out of energy to chair this thing.

16:42:48 <DanC> when we accepted new requirements and undid the April CR milestone, it really took the wind out of my sails.

16:44:43 <swh> yes, we did meet in boston

16:44:52 * AndyS notes swh is in Southampton

16:45:24 * tlr thought he was close to the center of the semantic web world. ;)

16:46:32 * AndyS having trouble composing a 3rd party sentence to fit /msg for a 4th party subject

16:47:04 <swh> southampton and bristol are in the same place, from a non-uk perspective ;)

16:47:07 <raxor> raxor is now known as shepers

16:47:15 <tlr> right

16:47:23 <shepers> shepers is now known as raxor

16:47:38 <AndyS> TLR knows - the airports are different

16:47:51 * tlr hasn't been to soton

16:48:09 <swh> i'm wondering if id rather have the slightly borken (as i see it) specification of GRAPH as it is, or no GRAPH at all

16:49:07 <swh> tlr, youre not missing much

16:49:17 <DanC> i think what's requested is that GRAPH be a sort of 2nd-class thingy. optional.

16:49:38 <swh> yes. i dont really like that, from a spec. design p.o.v.

16:49:46 <DanC> nor do i.

16:50:25 <DanC> I'd like the WG to decide either that it's ready for standardization now or that it's not

16:50:40 <swh> agreed

16:52:25 <timbl> Yosi's work converting SPARQL into N3 rules seemed to still work with GRAPH, get a little trickier with OPTIONAL but turn into a whole new ball of wax with !bound.

16:52:44 <AndyS> Notes (sub-rec) of things outside a one-graph SPARQL?

16:53:59 * bijan wonders if the working group just likes having the *option* for it to be optional...

16:55:15 <swh> I was concerned about !bound but in my impl. it turned out to be easy, i can see how it wouldn;t be in some cases though

16:55:45 * tlr seems to recall that ericP didn't have much of an implementation problem with either !bound or unsaid.

16:57:59 <swh> his backend is somewhat similar to mine

16:58:18 * tlr leaves for hotel, then meeting's social event. See you tomorrow.

16:58:57 <timbl> Yosi was converting the query into existing n3 rules predicates. !bound meant he had to compile an expression for the conditions for x to be not bound, which ended up involving all the possible ways that filters could fail because of type errors etc etc

16:59:13 <DanC> I'm satisfied with the implementation experience around graph/bound/optional. But there are these calls for formal completeness/soundness semantics. If we accept that requirement, we've got work to do. And I think there's sufficient critical mass that we have to at least consider that requirement as a WG.

17:00:17 <DanC> timbl, !bound should be reducible to log:notIncludes, if my hunch is right. I haven't thought it thru carefully, but it works in all the practical cases I've tried.

17:02:02 <timbl> Well, you need to to talk to Yosi directly. That is true in principle but it is vary hairy

17:02:24 <timbl> (especially when the log didn't follow de Morgan's laws)

17:02:50 <timbl> The filter expressions have to be negated.

17:04:20 <DanC> I think the WG is convinced of your argument re de Morgan's laws.

17:04:51 * bengee would like to see sparql move to next stage, even if a first version would be limited. With the currently existing features we can already do way cooler things than with rdql & co, and I know of at least 2 developers who stopped implementing the spec because they were tired of following the additions/changes...

17:05:15 <timbl> convinced - good. check that one off.

17:05:15 <bijan> Er...

17:05:26 <bijan> I don't know what y'all mean, so please don't count me as convinced :)

17:05:51 <DanC> well, be advised, read JeenB's message, bijan, and if you disagree, send something soonish.

17:06:00 <bijan> Pointer?

17:06:15 <DanC> Subject: boolean operators and type errors

17:06:15 <DanC> Date: Wed, 07 Sep 2005 17:13:41 +0200 (10:13 CDT)

17:06:30 <DanC> X-Archived-At: http://www.w3.org/mid/431F03A5.9040101@aduna.biz

17:06:57 <AndyS> bengee - examples of "limited"?

17:07:09 <timbl> There was a rule that false or typeErorr = false, which contradicts the normal false or x = x and would mask errors.

17:07:16 * bijan doubts he'd *dis*agree, esp. with DM laws!

17:07:45 <swh>http://lists.w3.org/Archives/Public/public-rdf-dawg/2005JulSep/0386.html

17:07:48 <dc_swig> C: http://lists.w3.org/Archives/Public/public-rdf-dawg/2005JulSep/0386.html from swh

17:08:27 * DanC sees his 12:00 meeting starting.

17:10:01 <dajobe> well I'm not convinced, it's tri-value logic so no clear DM law applies

17:10:13 <bengee> AndyS, just following the discussion a bit, seemed to me that people continue demanding more than currently available/spec'd

17:10:38 <AndyS> :-) Esp the words "GROUP BY"

17:11:06 <bengee> ;)

17:11:08 <bijan> (er...I believe there are demorgan laws for trivalant logics that are clear)

17:11:08 <DanC> bengee, that's the nature of the beast, i.e. the consensus process.

17:11:10 <AndyS> F && E => E ??

17:12:04 * AndyS having a tricky moment wondering why there is no truth table for && in rq23

17:13:19 <bengee> the thing I personally would have liked was this bnode referencebility (if that word exists), but otherwise, sparql seems to do enough for me.

17:13:37 <bijan> Hmm...http://en.wikipedia.org/wiki/Ternary_logic

17:14:18 <AndyS> Syntax <_:abc> is tacitally allowed. If your impl is nice to you.

17:14:24 <AndyS> But don't tell anyone.

17:15:13 <AndyS> as a custom URN scheme would (cheatingly) do the same.

17:15:23 <crschmidt> bengee: me as well, but I understand the reason for it not being included

17:15:33 * bengee secretly deletes blog post

17:17:37 <AndyS> F && E => F there

17:18:20 <AndyS> I choose to allow <_:abc> locally but not across the network.

17:19:01 <AndyS> May revisit and add network but the lifetime of the bNode label now matters much more

17:19:37 * bengee nods

17:20:37 <bengee> my remote access interface still needs a lot of work. it's still too easy to kill the mysql db etc

17:21:28 <bijan> dajobe, http://en.wikipedia.org/wiki/Ternary_logic seems to have the normal demorgan laws

17:21:48 <bijan> I just verified ~(P&Q) and ~P v ~Q

17:25:19 <AndyS> bengee - what's the DB schema (ptr?)

17:27:15 <chaals> :-( no shellac

17:28:07 <bengee> .g bnode sparql

17:28:09 <phenny> bnode sparql: http://www.bnode.org/archives1/40

17:28:51 <bengee> I think..

17:29:09 <dajobe> bijan: good. well that's a good choice then, rather than some more micro-change form "make A | B now be C instead of D"

17:29:51 <bengee> yep, it's in the middle of the post, AndyS

17:30:37 <bengee> ah, no: http://www.bnode.org/archives2/36

17:31:42 * bengee a bit confused today, sorry, still in holiday mode..

17:33:17 <AndyS> np - lucky to be in holiday mode :-)

17:33:36 <bengee> heh

17:33:58 <AndyS> Scale? How many triples loaded into one graph?

17:34:21 <AndyS> (the $64K question :-)

17:35:21 <dajobe> ok, that tri-logic table exactly matches ericP's proposal (after timbl) given in http://lists.w3.org/Archives/Public/public-rdf-dawg/2005JulSep/0360.html

17:36:01 <bengee> low-scale, some 30000 statements in store. it's somewhere in the rewriter, or my naive approach to add owl enhancements to the queries

17:37:42 <bengee> I also get completeley different response times depending on the mysql version, though. some absolutely don't seem to like distinct+order in one query.

17:39:10 <bengee> spent two days trying different index designs, but that led to problems with other queries then...

17:43:47 <bengee> mysql seems to build complete result sets when queries use ORDER BY which leads to issues here when there are more than just a few joins.

17:44:33 <bengee> (basically what crschmidt blogged a couple of weeks ago: the ordering of the joins is not optimized etc)

17:44:47 <AndyS> Yes - we found MySQL can be "interesting" - subqueries seem to actively avoid the indexes

17:45:38 <bengee> yeah, richard sent me several helpful comments

17:46:27 <AndyS> Cool.

17:48:44 <bengee> I'll have to somehow get "remote access restrictions" into the interface, simply restricting the LIMIT value is clearly not sufficient for an open sparql endpoint..

17:52:44 <bijan> dajobe, excellent, thanks for checkign that

19:08:00 <Wikier> Wikier is now known as WikierOFF

19:20:25 <DanC> C:|boolean operators and type errors

19:20:25 <dc_swig> Titled item C.

19:20:30 <DanC> C:Jeen to DAWG 7Sep

19:20:30 <dc_swig> Added comment C1.

19:21:27 <DanC> C:part of the DAWG's discussion of [http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2005Aug/0030.html|Bug: "A value disjunction that encounters a type error on only one branch will return the result of evaluating the other branch." Tim Berners-Lee (Tuesday, 2 August)]

19:21:28 <dc_swig> Added comment C2.

19:29:28 <bamse> bamse is now known as KjetilK

20:30:57 <WikierOFF> WikierOFF is now known as Wikier

20:31:58 <edsu> edsu is now known as edzz

21:05:01 <Wikier> Wikier is now known as WikierOFF

21:24:23 <ows> hi there

21:28:24 <RU3> hello

21:28:59 <RU3> I just visited the Semantic Web pages on w3. I like it.

21:31:28 <michelp> RU3, drink the kool-aid

21:43:34 <RU3> what kool-aid?

21:47:22 <michelp> RU3, bad joke. there was a cult in the 70's that made its members drink a poisoned american drink sweetner called kool-aid

21:47:35 <michelp> google "Jim Jones Jonestown"

21:51:21 <RU3> what does that have to do with Semantic Web ?

21:52:19 <chimezie> join #swhack

21:52:30 <michelp> RU3 it's just a saying a modern coloquialism. It means welcome to the cult. you've drank the kool-aid. get ready for your fantastic adventure through time and space

21:54:46 <RU3> hah hah!

21:56:41 <RU3> After I study and learn RDF, is there anthing else I need to know to be productive with building Semantic Web apps?

21:58:17 <RU3> darn time to go home. :/ be back in an hour or two :)

22:21:55 <hex_> hex_ is now known as _hex_

22:22:29 <_hex_> _hex_ is now known as _hex

23:05:17 <RuebaiYaT> RuebaiYaT is now known as RU3

23:06:51 <RU3> hello