Chatlog 2009-11-13 Ent

From SPARQL Working Group
Jump to: navigation, search

See original RRSAgent log and preview nicely formatted version.

Please justify/explain all edits to this page, in your "edit summary" text.

13:34:18 <RRSAgent> RRSAgent has joined #sparql-ent
13:34:18 <RRSAgent> logging to
13:34:26 <ivanh> rrsagent, set log public
13:42:25 <ivanh> zakim, room for 10 at 14:00Z for 90 minutes?
13:42:27 <Zakim> ok, ivan; conference Team_(sparql-ent)14:00Z scheduled with code 26632 (CONF2) at 14:00Z for 90 minutes until 1530Z
13:42:47 <ivanh> ivan has changed the topic to: ad-hoc conference at zakim, with code 26632
13:47:49 <AndyS> AndyS has joined #sparql-ent
13:55:54 <sandro> sandro has joined #sparql-ent
13:56:02 <sandro> zakim, who is here?
13:56:02 <Zakim> Team_(sparql-ent)14:00Z has not yet started, sandro
13:56:03 <Zakim> On IRC I see sandro, AndyS, RRSAgent, Zakim, ivan
13:56:05 <sandro> RRSAgent, pointer?
13:56:05 <RRSAgent> See
13:57:03 <davidcharboneau> davidcharboneau has joined #sparql-ent
13:58:22 <bglimm> bglimm has joined #sparql-ent
13:59:56 <Zakim> Team_(sparql-ent)14:00Z has now started
14:00:03 <Zakim> +[IPcaller]
14:00:15 <AndyS> zakim, [IPCaller] is me
14:00:15 <Zakim> +AndyS; got it
14:00:31 <Zakim> +Sandro
14:00:32 <Zakim> +bglimm
14:01:13 <Zakim> +dcharbon2
14:02:02 <bglimm> Zakim, whi is on the phone?
14:02:02 <Zakim> I don't understand your question, bglimm.
14:02:13 <bglimm> Zakim, who is on the phone?
14:02:14 <Zakim> On the phone I see AndyS, Sandro, bglimm, dcharbon2
14:02:56 <bglimm> ok, then we'll strat
14:02:59 <bglimm> start
14:03:47 <bglimm> scribe: bglimm
14:04:03 <bglimm> was that the right command?
14:04:15 <AndyS> yes
14:04:32 <bglimm> Issue 28: Entailment regimes vs. update?
14:04:37 <sandro> zakim, who is here?
14:04:37 <Zakim> On the phone I see AndyS, Sandro, bglimm, dcharbon2
14:04:38 <Zakim> On IRC I see bglimm, davidcharboneau, sandro, AndyS, RRSAgent, Zakim, ivan
14:05:19 <sandro> bglimm: Do all systems have to implement UPDATE?    Maybe we can leave entailment + update undefined?
14:05:55 <bglimm> Sandro: Clearly there will be read-only end-point. 
14:06:16 <bglimm> ... we do want to alow for the possibility of systems doing both entailment and update
14:07:48 <sandro> bglimm: with an OWL disjunction, how could you choose?
14:08:06 <bglimm> ... AndyS: Is there entailment in the pattern used in the query
14:10:10 <bglimm> Sandro: If the modified graph contains entailments, we probably don't want to use entailment, but if the modified graph is target of an insert that could be ok
14:11:23 <sandro> sandro: I think the main thing is: if the updated graph contains inference, implementations may reject the update, if they can't do it (the usual case).
14:12:17 <bglimm> AndyS: I would leave it to systems and not put a complete spec part into the entailment doc
14:12:20 <sandro> sandro: I think we should have a sentence like that somewhere.
14:13:04 <AndyS> David, have we covered the point behind the issue?
14:13:33 <sandro> bglimm: Okay, settled.
14:13:41 <AndyS>
14:13:51 <sandro> topic: issue-34
14:14:01 <bglimm> [ISSUE 34]: How do entailment regimes interaction with
14:14:01 <bglimm> aggregates, grouping, and blank nodes?
14:14:33 <sandro> bglimm: For RDF and RDFS it's clear from the spec now.
14:14:50 <sandro> bglimm: For OWL DS I'm just adding the ent. reg. spec.  It's pretty much the same.
14:15:04 <sandro> bglimm: close this issue?
14:15:34 <sandro> andy: I don't understand why this issue mentions blank nodes
14:16:10 <sandro> bglimm: It's kind of confused now yes.   At the time, we weren't sure about counting blank nodes.   Since we now limit bnodes to those that occured in original graph, that pretty much solves the issue.
14:16:27 <bglimm> [ISSUE 42]: TF-ENT What should happen for RDFS entailment in the
14:16:27 <bglimm> face of inconsistencies?
14:16:30 <sandro> andy: Yeah, at least makes it finite!
14:16:37 <sandro> topic: issue-42
14:17:14 <sandro> bglimm: the MUST raise an error could be a problem to implement with good performance.
14:17:32 <sandro> bglimm: such as in the face of join.
14:17:44 <sandro> bglimm: MAY raise it, SHOULD if they encounter it.
14:19:08 <bglimm> bglimm: I'll write an email to record the decisions/discussions so Axel and Lee can decide to close them
14:19:08 <AndyS> Check entailment: ASK { ?s ?p ?o }
14:20:52 <sandro> sandro: can we have a keyword or something so the client can demand a consistency check (to be accepted or rejected).
14:21:19 <sandro> AndyS: There are other things you want, like asserting primitives during the duration of the query.
14:21:38 <sandro> sandro: Fine.   This could be added later.
14:21:50 <sandro> AndyS: This is not the full API for a reasoning system.
14:22:00 <bglimm> [ISSUE 43]: should entailment-regimes be declared over the whole
14:22:01 <bglimm> dataset or individual graphs?
14:22:03 <sandro> bglimm: So we'll leave this for later.
14:22:16 <sandro> topic: issue-43
14:22:42 <sandro> bglimm: "endpoing does this entailment" in service description
14:23:07 <sandro> bglimm: But you might want to query direct subclasses.    one graph with simple entailment, another with rdfs entailment
14:23:49 <sandro> AndyS: I think this is about service description, not the Entailment Document.
14:24:50 <sandro> sandro: I think the Entailment Document should make clear we expect endpoints to often have graphs with inference and graphs without inference.
14:25:04 <sandro> sandro: Yes, this is something service description shoudl do.
14:27:19 <bglimm> bglimm: Axel can do next Friday, so we could do another telcon next Friday no RIF
14:28:29 <bglimm> Sandro: Some RIF dialects are not based on model theory
14:29:01 <bglimm> ... they are not really entailments, but they behave in a sense like entailments
14:30:25 <bglimm> Sandro: Sometimes you want an endpoint that does just RIF and sometimes you want to have entailment wrt a particular rule set
14:30:48 <bglimm> AndyS: This can be computationally hard. 
14:31:16 <bglimm> ... you can use different graphs in your data set that have the features you want. 
14:31:52 <bglimm> Sandro: Can we add something to the language or protocol to specify what rule set to use? So that the client can say what it wants. 
14:32:08 <bglimm> AndyS: That goes into content/entailment negotiation. 
14:32:30 <bglimm> ...: The server could declare what it can do and the client can decide whether that is ok or not
14:33:24 <bglimm> AndyS: It would require changes in several places to give the client the ability to ask for a particular rule set to be used
14:33:50 <sandro> sandro: not sure if it's important to have a way for the client to ask for particular inference, or just use service description.
14:34:54 <bglimm> Sandro: we should get a note for RIF to et RIF into a kind of graph format
14:34:59 <sandro> sandro: one way for now is by naming a graph....
14:35:29 <sandro> sandro: We can probably get a RIF Note with rif:import, so you can name a graph that has that.
14:35:35 <sandro> bglimm: solves part of the problem.
14:36:22 <sandro> topic: non-monotonic logics
14:36:32 <sandro> bglimm: I've been ignoring it.
14:36:47 <sandro> AndyS: I think that's our only terminating decision.
14:37:41 <sandro> bglimm: I'm happy to have a co-editor who can address non-mon.    Not good thing for me to address.
14:38:07 <bglimm> Sandro: I think there is only one person from RIF who could really do that
14:38:36 <bglimm> ... there are production rules which have no model theory 
14:38:56 <bglimm> AndyS: For production rules there is not even work on conjunctive queries
14:38:56 <sandro> andy: PRD: no notion of answering a conjunctive query.
14:39:00 <sandro> sandro: Right....
14:39:17 <sandro> andy: I'd like to see some pre-work there.
14:39:33 <sandro> sandro: I agree...   someone who knows and cares about that stuff needs to do some work there.
14:39:50 <sandro> bglimm: it wont go into this round unless someone else comes in to work on it.
14:39:58 <sandro> sandro: okay.
14:40:10 <bglimm> bglimm: Anything else for RIF?
14:40:34 <sandro> topic: OWL entailment, Direct Semantics
14:40:52 <sandro> bglimm: Not all axioms carry semantics, but users want to query them.   eg annotations.   
14:41:34 <sandro> bglimm: Maybe: say use DS for axioms that carry semantics,and simple entailment for non-logic axioms.    So users can still query for annotations.
14:42:33 <sandro> AndyS: Sounds good to me.       Linked between the two partitions could be problematic.
14:42:43 <sandro> bglimm: need to do one part.
14:43:05 <bglimm> Sandro: Another way would be to let the client do.
14:46:11 <sandro> sandro: If it works, and PFPS is okay with it, then great.   Otherwise, fall back on the two-graphs approach.
14:46:49 <AndyS> I can't make next Friday.
14:46:57 <sandro> next meeting: tentative next friday, cancel if no business
14:48:16 <sandro> bglimm: I'll send e-mail summarizing these matters, and let the chairs figure out whether/how to close issues
14:48:32 <sandro> AndyS: OR, since this is time-permitting, and we're doing it outside the WG, ...
14:49:03 <Zakim> -dcharbon2
14:49:05 <Zakim> -AndyS
14:49:07 <Zakim> -Sandro
14:49:08 <Zakim> -bglimm
14:49:08 <Zakim> Team_(sparql-ent)14:00Z has ended
14:49:09 <Zakim> Attendees were AndyS, Sandro, bglimm, dcharbon2
14:49:21 <bijan> bijan has joined #sparql-ent
14:49:31 <sandro> nice timing, bijan :-)
14:49:44 <sandro> <Zakim> Team_(sparql-ent)14:00Z has ended
14:49:44 <sandro> <Zakim> Attendees were AndyS, Sandro, bglimm, dcharbon2
14:49:44 <sandro> --> bijan (bparsia@ has joined #sparql-ent
14:49:48 <bijan> Are you just wrapping up ?:)
14:49:54 <bglimm> yes, all done
14:49:57 <sandro> we hung about about 15 seconds ago.
14:49:58 <bijan> Sorry, my meeting ran over :(
14:49:59 <bijan> Ok
14:50:01 <bijan> Cool
14:50:02 <bijan> sorry
14:50:05 <AndyS> GMT Bijan.
14:50:23 <bijan> Yes, I'm still wrongtimed
14:50:23 <AndyS> You still on summer time? :-)
14:50:31 <bijan> Well, Zoe was in the states
14:50:34 <bijan> but they shifted too
14:50:36 <bijan> so i have no excuse
14:50:38 <bglimm> rrsagent, make records public
14:50:58 <bijan> I was talking to someone about experiment design and got distracted
14:51:28 <bijan> Oh well. I'll look at the minutes. sorry for flaking out
14:51:30 <bglimm> Sandro, are you still there?
14:51:33 <sandro> yes.
14:51:44 <bglimm> How can I get the minutes into the wiki?
14:52:01 <sandro> ummmmm.    I'll do it.   hold on.
14:52:06 <sandro> RRSAgent, pointer?
14:52:06 <RRSAgent> See
14:52:07 <davidcharboneau> davidcharboneau has left #sparql-ent
14:52:10 <bglimm> With the control centre, I tried URL but it did not work
14:52:36 <sandro> yeah, because this is on a different channel, it thinks this is a task force, and the minutes should go into a different series.
14:52:54 <sandro> I'm trying to decide whether to really make it a task force, or pretend it was a real sparql meeting.
14:53:06 <ivanh> ivan has left #sparql-ent
14:53:16 <bglimm> Hm,
14:53:25 <bglimm> is task force complicated?
14:53:44 <sandro> Not sure.   :-)   I'll try and see if I still remember how to do it.  :-)