00:00:27 (http://www.ksl.stanford.edu/projects/dql/syntax.shtml is a syntax url i found recently) 00:01:15 returns sets of bindings. 00:03:40 want to be able to specify justification.... 00:05:51 ian: impleemnting the entire thing (e.g. queries with cycles) is hard. some restritions not so bad 00:05:53 I made a front page for rdfq test case repository, http://www.w3.org/2003/03/rdfqr-tests/ 00:06:16 ...haven't any content (test cases) yet, but gathered pointers to all existing test stuff i can find 00:06:19 ...surveys etc 00:07:22 ericP - could you do a quick one-pass squish-like pass? 00:07:33 benjamin - yes, even if the backend was complex 00:07:46 ericP: wouldnt lose perfortmance on simple queries 00:08:07 Mike: can assert premises 00:09:02 ericP: variable feature: may bind, must bind, whether the node is a vraiable. also - find all results or work for a while and then return. 00:09:37 ericP: on the reporting - which variables do you want to see. 00:10:05 ericP: = 3 areas so far. wold like input. would like to do a taxonomy of query charactristics 00:10:25 josD: the same query can have different proofs. 00:10:55 ericP: in algae, you'll get multiple rows - different reasons 00:11:23 [missed some] 00:12:05 josd thinks this is a whole different dimension 00:12:15 ericP: exhaustive search/ result grouping 00:12:38 josd: 2 dimensions 00:13:08 raphael: kaon project - QL for ontologies (one binding is RDF) 00:14:09 ...different datamodels for rel dbs, RDF dbs, ontology stores. 00:14:58 ...retrurns not tuples but classes and properties - things that correspond to the datamodel (of onbtology) 00:15:12 ....modelled by a set of datalog ruless 00:16:10 ...easy composition of orthogonal operators 00:16:43 ....can rewrite e.g. to SQL where posisble 00:17:33 benjamin: can oit bind against uri or an arbitrary literal (R: yes) - sounds like ruleml 00:18:23 ...has things like, SOME, INVERSEOF, .... 00:18:40 timecheck -- how much longer do we have? Does the meeting run until 7.30pm? 00:19:09 that's right yep 00:19:21 oh, only 10 mins more... 00:19:55 ....and, or and not, where not is negation-as-failure 00:22:06 ian: can;t express cycles in non-distinguished variables. 00:23:49 Raphael: 'oneof' is expressed as a ! 00:25:04 cf http://www.w3.org/TR/owl-ref/#EnumeratedDatatype (hmm, probably not the best ref for owl:oneOf) 00:25:37 ....http://kaon.semanticweb.org - can download it. 00:25:56 ericP: example query now in http://www.w3.org/2001/11/13-RDF-Query-Rules/ 00:27:20 we decide to continue till 8 00:27:35 said talking about ruleml 00:27:50 ...close to 40 participating groups 00:30:03 ...several ruleml engines and translators avilable, eg mandarax. some free 00:30:28 is this channel still being logged? 00:30:37 that's a good q 00:30:43 I can get it if not 00:30:56 rssagent is still there 00:33:01 rrsagent, help? 00:33:01 I'm logging. Sorry, nothing found for 'help' 00:33:05 ok cool 00:33:21 but on http://www.w3.org/2003/03/06-swarch-irc only got partial 00:33:31 oh wait, maybe it's 'tomorrow'? 00:34:01 coudl you check danbri? e.g. http://www.w3.org/2003/03/07-swarch-irc (forbidden) 00:35:27 ...plan to submit usecaes to alberto's and andy's repository (http://rdfstore.sourceforge.net/2002/06/24/rdf-query/) 00:36:13 ....rulebase - GEDCOM - family relationships 00:36:22 ...created by Mike Dean 00:37:11 josd: how do those things cope with unique names assumption? very important, as we found in testcases 00:37:31 mikeD: one source of data, so implicit unque names assumption 00:38:00 josd: could do daml:differentFrom based on syntaxtic diffences? 00:38:16 mikeD: could do, but don't. should preobably ahve somethign more 00:40:12 thanks :) 00:40:24 world ACL 00:41:02 Said: workign on an ecommerce demo 00:41:58 ...on website soon 00:42:27 ericP: could we use the rules for rule languages? 00:43:20 benjamin: yes you can use this [...] but often yopu want actions triggered, which is beyond the QL scope 00:43:35 Said: can use outside services. 00:45:33 [??] why doyou can it object orientated? 00:45:42 (harold will talk about it later) 00:46:06 danbri: this event-triggering rules seems very difefrent from timbl's rules... 00:46:26 danbri worried about scope - 'if and then' 00:47:21 said: didnt really start off as a rules language 00:47:59 benjamin: built-ins are very common in commercial rules sytems. 00:49:50 harold talks about object-orientated ruleml 00:50:28 ....ruleml and rdfs overlap 00:50:41 ...can use oo ruleml as an rdf ql and rles language 00:51:00 I just added a bunch more test-related links to http://www.w3.org/2003/03/rdfqr-tests/ 00:52:23 ...the subject and object are detrminted by their position 00:54:53 [sorry scribe missed if that was the positional one or the oo one] 00:55:24 ...the two versions can be translated using xslt 00:56:35 ...harold works us through an example 00:56:43 (url?) 00:57:38 ...can express e.g. which page was accessed by person. in ruleml, queries are aq special case of rules, including only the body. 00:58:28 ...there's an issue with bnodes 00:59:28 ....need to give bnodes ids? [scribe missing soem of this, coudl be wrong] 01:00:21 ...can do conjunctive queries: use 2 atoms 01:00:28 ...now generalize to bnodes.... 01:02:24 scribe fading... :( 01:02:33 imho ruleml includes a proposal for a new RDF syntax 01:03:23 ....seesm to use generated nodeids 01:04:02 ....model theory can bbuild on ruleml's rdf-xml integrating data model via flogic or triple 01:04:19 josd: a very diffreent datamodel to RDF. 01:04:31 harold: higher-order syntactic sugar 01:05:08 josd: why 'object-orientated'? 01:05:41 harold: as RDF is OO. also all the descriptions of objects clustered togetger 01:05:53 said: not the same as a definition of an OO language 01:05:59 timecheck! 01:06:04 we should be winding up... 01:06:07 benjamin: does not rely on positionality - has a name for the variable. 01:06:09 yep 01:06:14 'night folks 01:06:28 harold 'subject-oriented' 01:07:23 we are 10 minutes over time 01:07:34 banjamin says: 4 minutes! 01:07:47 he is outlining note draft on ruleml 01:08:05 ....has also been discussed in the joint committee 01:08:22 ....requiremeents, play nicely w the rest of SW, and also ws, xquery 01:08:43 ...different tpes of rules 01:09:07 2 minutes gone 01:09:18 ...those that derive new beliefs (like RDF query). also action rules - actiosn, get info. transfoemation can be vieweed as derrivation.... 01:09:43 ...some RDF qs return bindings, some graphs 01:10:21 ...vraious chaacteristics of rules 01:10:37 ...OO-ness (non-positional) 01:11:07 ...lits of first-order expressiveness... 01:11:13 s/lits/lots 01:11:48 ...in teh markup need to talk about things being derrived. this KB derrived this other stuff 01:12:36 ...using several KVs 01:12:43 KBs even! 01:12:56 ...complimentary doc on usecaese in teh joint committee 01:13:51 ...questions? 01:14:09 benjamin proposes we head to bar, declaring victory 01:14:32 ericP: this would be good foddder - clarifying different parts of query 01:14:43 b: 3-4 weeks, public draft 01:14:51 ...will post to RDF rules 01:15:29 ---scribe declares victory. ajourned..... 02:06:10 em-lap has joined #swarch 02:49:39 las has joined #swarch 03:55:53 Tantek has joined #swarch 03:56:25 Tantek has left #swarch 10:44:27 DaveB has joined #swarch 11:39:15 AndyS has joined #swarch 11:40:45 AndyS has joined #swarch 11:51:35 AndyS has joined #swarch 13:20:16 em-lap has joined #swarch 13:20:28 ack... no opps 13:20:41 Zakim has joined #swarch 13:20:53 em-lap has changed the topic to: semweb arch tech plen meeting - 2002-03-07 13:26:17 ericP, you here? 13:47:23 PStickler has joined #swarch 13:49:49 jhendler_ has joined #swarch 14:09:05 jhendler_ has joined #swarch 14:12:17 danbri has joined #swarch 14:12:38 (we got stuck in traffic) 14:14:46 JosD___ has joined #swarch 14:15:02 pfps has joined #swarch 14:15:06 danb_lap has joined #swarch 14:15:57 uj 14:17:33 timbl: interested in scoping new work areas, how much time things would likely take, etc 14:17:48 ...seems to me from way layers are developing, Query is next item ready for standardisation 14:18:07 ...i read thru various bits and pieces 14:18:31 ...has anyone read thru this? (not many hands go up) 14:18:42 ...ben and harold fed back some possible changes 14:18:47 DanC has joined #swarch 14:19:01 ...see http://www.w3.org/2001/sw/meetings/tech-200303/query 14:19:12 ...b/g includes AndyS and Alberto's use case repository 14:19:15 bwm has joined #swarch 14:19:19 ...can register examples 14:19:24 GuusS has joined #swarch 14:19:30 ...going over that, you get a pretty good idea of what folk are doing w/ rdf query 14:19:34 rrsagent, help? 14:19:34 I'm logging. Sorry, nothing found for 'help' 14:19:55 ...interesting. Most syntaxes are non-xml 14:20:15 ...half of them use the word 'SELECT', ie. SQL mind set, sometimes 'FROM', 'USING' etc 14:20:24 libby has joined #swarch 14:20:28 ...some have some punctation around a chunk of rdf, and an ipmlication sign 14:20:31 ...fall into various groups 14:20:58 ...seems that concrete syntaxes may have questions, but the abstract syntax there is much commonality 14:21:05 ...simplest might be versa, which is just a path 14:21:15 ...this is a subset of a general rdf graph matching template 14:21:35 ...PatH presented DAML query (DQL) at a DAML meeting, is basically of that form 14:21:48 ...if you look at the graph match, ie. rdf with holes, a question of what you can put in the graph 14:21:59 ...so some case to standardise a non-xml syntax for such abstract queries 14:22:11 ...obviously good to have an XML version of that (compare xquery, which did both) 14:22:28 ...also case for some overlap w/ rule stuff eg ruleml (?missed point) 14:22:33 ...abstract syntax pretty common 14:22:46 ...engines differ a lot re the kind of inferences they do underneath 14:22:51 ...but the querying language loioks the same 14:23:03 ...you query notional, possibly infinite, dataset 14:23:18 ...like simpler if we make thw QL simpler, leave service capabilities a separate problem 14:23:32 jos: you said Fwd inference 14:23:43 timbl: conceptually you are querying all the possible derrived data 14:23:50 ...but we're covering that up 14:23:52 ...i looked at ruleml 14:24:02 ....it has a language which has an xml notation 14:24:12 ...original goal to unify all tyhe various rule systems out there 14:24:20 ...typically those weren't webized, ie. use URIs 14:24:33 ....ruleml now extended to be URI capable 14:24:53 ...can convert eg RQL and ruleml, but if rules lack uris, you have to go add namespace uris etc before get interop 14:25:00 ....so ruleml was extended 14:25:08 ...u/stand theres a later unpublished version 14:25:24 (aside from danbri: http://www.w3.org/2003/03/rdfqr-tests/ ) 14:25:36 timbl, there are dtds to be found on ruleml site 14:25:49 ...another difference: some languages allow a template for what you want returned 14:25:54 ...this looks like a template language 14:26:03 (er sorry, entailment language, or something) 14:26:19 ...if you think of these as languages sent to services, can see result as bindings versus an rdf graph 14:26:36 timbl shows his table of pro/cons re graph vs query, see http://www.w3.org/2001/sw/meetings/tech-200303/query 14:26:53 ...mentions poss of generating an alternate(???) 14:27:00 ...if you think of more pro/con, let me know 14:27:08 ...i tihnk you probably need both 14:27:15 ...there are also syntax choices 14:27:18 ...SELECT yadda 14:27:25 ...sort of thing we do in a WG 14:27:33 ...arguing round brackets vs square ones 14:27:46 ...semantics of what happens underneath more of less orthogonal 14:27:51 ...deciding on builtins again orthogonal 14:27:56 ...has happened already in various places 14:28:05 ...would make sense to pick up XQuery work on this 14:28:21 ...Steve Reed from Cyc has a list of builtins they support, and those of xml query, compare/contrast 14:28:26 (@@Url/google anyone) 14:28:35 ...have to pick your favourite libraries 14:28:53 ...will be services that do/don't support these 14:29:06 ....we can make intelligent systems that add such things to dumber stores 14:29:11 ...didn't get into remote query 14:29:19 ...main reason we do stuff at w3c network centric 14:29:30 ...one way to do this is wrap the query, uri-encoded, and append it to an http URI 14:29:36 ...simple way of query, eg with a '?" 14:29:44 ...easy way of attaching to existing servers 14:29:50 ...or else do it one way or another in soap 14:29:58 ...some arbitrary design choices there 14:30:05 ...if i had to think of possible work, this is what came up 14:30:14 ...i would imagine that a wg would aim for an abstract syntax 14:30:38 see http://www.w3.org/2001/sw/meetings/tech-200303/query 'possible deliverables' section 14:30:46 [[ 14:30:47 So, if work were begun in this area, formally or informally, more or less in chronological order, one might hope to see: 14:30:47 Abstract syntax of query language - probably described in RDF. 14:30:47 Definition of a few conformance levels (monotonically increasing in features supported) 14:30:47 A common concrete syntax in compact (non-XML) form 14:30:48 Ontology for description of inference services provided by a service. 14:30:51 A set or sets of standard functions 14:30:52 A profile or profiles which combine the above to enhance interoperability, when experience with common engines is sufficient to define interop levels. 14:30:56 ]] 14:31:02 ...inference service profiles, so you 'know what you're getting' 14:31:07 A concrete syntax needed for network use between machines 14:31:32 [timbl talks through bullets above] 14:32:10 ...remote query: could be trivial, or ratholes 14:32:27 (i think lots of work there -- jdbc etc have lots of admin features, danbri) 14:32:45 timbl, adding DELETE etc would make a bigger job 14:32:53 ...how much work, is it time to do it as a WG? 14:33:18 alberto has joined #swarch 14:33:30 ...any general feelings re scope? 14:33:41 Harold: do you have general notion of queries that encompass rules? 14:33:49 hey alberto! 14:33:56 timbl: if you look at the rdf query, there is a rule language that is very connected to it 14:34:03 (@@url for logs for alberto to read?) 14:34:11 See http://www.w3.org/2003/03/07-swarch-irc#T14-34-03 14:34:11 hello - sorry I am late :) 14:34:13 ...clearly there is a connection 14:34:20 ok, going through it now... 14:34:39 ...for eg. RQL examples could all convert into ruleml 14:34:47 harold: rules are slightly more general concept than queries 14:34:56 ...can chain queries as a rule 14:34:59 alberto has joined #swarch 14:35:06 ...woudln't speak really of a chain of queries 14:35:20 ...could call it rules and query, or really just rules 14:35:26 timbl: we have www-rdf-rules list 14:35:28 I wouldn't call all chains of queries rules 14:35:32 ...some folk demanded separate mailing lists 14:35:43 ...i agree they're basically the same thing, but there are different systems, engines 14:36:00 harold: from last night, ruleml folk promised to submit more discussions to www-rdf-rules 14:36:09 beng: what would be way to distinguish non-query rules? 14:36:15 ...a lot of rule systems run fwd 14:36:22 ...query is pretty much backward 14:36:49 ...if you have procedural attachment, actions, that extends beyond basic semantics of query 14:36:51 (?) 14:36:57 ...two aspects that go behyond basic query 14:37:08 ...from a tech pt of view, there remains a v close relationship 14:37:19 ...you'd want same semantics 14:37:35 ...can have simple stateless way to define rules mechanics 14:37:37 (?) 14:37:43 ...v closely associated w/ a pure belief view 14:37:56 ...storing queries, having queries built from subqueries 14:38:11 ...wasn't emphasised 14:38:18 ...should think about when is the right time to get into that 14:38:28 timbl: some people mentioned desire to store queries 14:38:31 RalphS has joined #swarch 14:38:41 ...one motivation for sending rules w/ a query is to add 14:38:42 (?) 14:38:47 sanScribe has joined #swarch 14:38:59 ianh: from a tech pt of view, rules in general are nothing diff from what we have in the onto languages 14:39:12 ...std axioms we have in ontology, eg for subclass, is just a rule 14:39:30 ...but onthe other side, query languages normally have this special feature that you only get back answers w/ fininte set of things 14:39:56 ...when you say that the answer to a rule isn't all poss conclusions, it is just concrete answers 14:40:02 ...eg 'tell me all the people that live in ...' 14:40:11 las has joined #swarch 14:40:19 ...that gives a completely diff computational properyty to the language 14:40:23 timbl: does it change the syntax? 14:40:36 ...can we have different operationals but keep the ql the same 14:40:41 ian: maybe... 14:40:47 sandro has joined #swarch 14:40:53 ...my point was that we might allow some things in a QL that we don't allow in an assertional language 14:41:06 ...since QL resultsets have different characteristics 14:41:27 timbl: [draw attention to ian, ben ... work on OWL<->Rule mapping] @@url? 14:41:40 frank: in QL, constructing new tuples 14:41:53 ...would want a ql to support certain kinds of construction 14:42:06 ian: true, but elements in the tuples are things we know about in advance 14:42:28 [...] 14:42:39 frank: we need to be clear about the extent to which this is a restrictuion 14:42:51 ...if we have cities example, links between cities 14:42:58 ...q is: am i constructing paths 14:43:04 ...perfectly reasonable thing to do 14:43:18 ian: infinite answer in general, as can to/fro the same pair of cities 14:43:34 ...generally you disallow such queries, by saying 'gimme acyclic paths' 14:43:50 frank: is notion of a path, or instances of paths, what you're considering old things vs new things 14:43:50 (?) 14:43:56 shellac has joined #swarch 14:44:00 timbl: you may be talking about 'path' differently 14:44:31 http://www.w3.org/2003/03/07-swarch-irc 14:44:46 ian: if you allow infinte paths and infinite poss answers, thats when the wheels fall off 14:44:58 frank: am just trying to clarify what kinds of things we can get back 14:45:01 (he isn't) 14:45:04 sandro - no 14:45:06 (shellac i mean) 14:45:23 patrick: [...] if you havd a Q engine without rules, inference, you get back just ground stuff 14:45:33 ...if it does have such support, you get things back that are implicit 14:45:42 (photos - good idea) 14:46:10 patrick: whether underlying engine has inference is separable 14:46:21 ...what you get back isn't necc asserted/explicit in your kb 14:46:32 ian: i wasn't intending to say what ought be in/out of language 14:46:48 ...just note that computational properties of a ql vs an assertional language differ 14:46:57 ...because of this fact that you know in advance finite set of answers 14:47:18 timbl: proposal is that you can use the same QL in both contexts (albeit w/ diff comp properties) 14:47:26 ...use same lang to talk to it, results come back the same 14:47:45 ...hypothesis is that the ql can be the same 14:48:00 DanC: this sort of thing? http://swordfish.rdfweb.org/discovery/2002/10/calwk/calwk.jsp?pics=+calendar+workshop+images+ (slow...) 14:48:08 Volz: ... 14:48:46 timbl: qls i saw didn't allow (various kinds of) fancy path ql 14:49:04 spiffy, libby. that's more than I had in mind, but that's cool. 14:49:32 pathayes: here when you say Query Lnaguage, are you talking about the abstract pattern, or all the other additional features too 14:49:32 If a given query engine is not able to answer 'true' that doesn't mean the answer is 'false' 14:49:39 volz said he things query language should support expression of rules that can be used for inference - at least thats what I heard 14:49:59 If one engine does not have inference, and the answer is implicit, then it simply cannot say, but another engine with inference may be able to answer positively 14:50:05 timbl: ... [summaries earlier discussion for pat's benefit] 14:50:16 bwm, maybe later so i can say stuff! ok for now. 14:50:23 hmmm... the calwk photos aren't all from the workshop. I'm thinking of photos that give evidence that, e.g., I was here. 14:50:56 yes very useful work danb 14:51:28 ericp: [...] fact that you have a rule engine that might stop on 1st answer... is characteristic of the engine ... not of the engine 14:51:38 (some discussion about whether eric had this backwards) 14:51:41 s/engine/rule/ 14:52:10 ...if we can keep in mind there's a large commonality, and that you can express things as characteristics of the query or the report 14:52:22 (timbl is doing things on paper) 14:52:36 I don't expect "I want the first answer only" to be part of the query language syntax. hmm... . 14:52:42 jos: w.r.t. resultset, i think the results as a primitive form of proof/explanation in rdf, is wortyh considering in our experience 14:52:46 me neither, danc 14:53:00 jos: ...worth explaining in an rdf graph, facts/rules/queries/proofs all part of this 14:53:13 ...proofs as continuations, could be given to another engine to continue processing 14:53:23 timbl: this is a list of other poss extensions 14:53:40 (@@we should transcribe later. Have 'saving result set'; returning proof; so far) 14:54:18 brian: way we're looking at this in jena 14:54:22 ..rel w/ rules v query 14:54:35 ..you target a query at an engine, you identify the rules you want to apply by picking a target 14:54:45 ...implicit suggestion that the two differ 14:54:58 ...but QL could have a hole saying 'and some companion rules could go here' 14:55:57 danbri (barging in): this suggets a way that work can be punted from QL to companion APIs 14:56:25 ian: just to try to clarify point about computation, same thing raphael was saying... 14:56:29 :flipA log:semantics { :OtherFeatures is rdf:type of :savingResultSet, :returnedProof }. 14:56:43 ...is perfectly possible to take OWL and guarantee you can compute answers to all quewries in that lang 14:56:47 ...ie complete 14:57:03 ...same language as an assertional language, can show it isn't complete 14:57:09 (@@did i get that right?) 14:57:25 :flipA dc:description "(timbl is doing things on paper)". 14:57:31 ...you may not care about completeness, but point is you get diff comp properties 14:57:38 Query and Rule need not be different languages, but rather, Query is a subcomponent of the Rule language 14:57:47 harold: ... you can perhaps have two kinds of queries, lookup vs inferential 14:58:00 ...subqueries, could be lookup or further composite 14:58:12 ...in ruleml, we have an xml element 'query' with flags for lookup vs inference 14:58:27 timbl: thinks thats a shared model 14:58:38 mikedean: one big question... how much do q and rules overlap 14:58:42 ...am trying to do a venn diagram 14:58:54 ...what i have not quite ready to share but suggest useful technique 14:59:02 ...prelim, that most stuff is out on the edges 14:59:12 ...suggest interesection is pretty small 14:59:23 (@@is that good? ie. a small focussed workitem?! --danbri) 14:59:41 path: there is a std logical picture of the query process 14:59:46 ...and how query/rules overlap 14:59:54 ...rules are like horn clause implication 15:00:05 ...a query is a pattern put up as a candidate conclusion 15:00:07 (?) 15:00:10 ... 15:00:10 Specifying whether a given engine does or does not perform inference is a parameter to the engine, not a feature of the core query language 15:00:20 ...get this on table as the 'off the shelf' picture of the rel'nship 15:00:27 ...also Jos's point about returning proofs 15:00:50 ...nice story about poss responses, range from 'yes'! to getting entire proof, versus intermemdiate, getting bindings 15:00:55 I am in the queue after DanBri (though got skipped over already once) 15:00:57 ...might well be other things 15:01:01 q- 15:01:18 zakim, DanBri is danb_lap 15:01:20 sorry, danb_lap, I do not recognize a party named 'DanBri' 15:01:39 ack Pat 15:01:39 ben: to follow that up, jos and pat touched on proof 15:01:45 ack Ben 15:01:47 ...query as a concept in KR is something that any KR can have 15:02:04 I am in the queue on the whiteboard 15:02:05 ...you usually start conceptually from a KR, eg principles of sanctioned entailment 15:02:25 w/ proof, other actions, consistency, monotoniciity, syntactic violations, resource limits, max answers etc. 15:02:34 ...mechanical or complemenetary surrounding considerations 15:02:44 ...much of xquery focussed on such things 15:02:53 timbl: did xquery cover resource limits? 15:03:03 ben: eg don't try more than 1000 seconds on this 15:03:19 ...if you doing info integration across sites, spend only so much money/time 15:03:25 timbl: didn't think this was in xquery 15:03:44 ack PatrickS 15:03:44