IRC log of dawg on 2005-10-04

Timestamps are in UTC.

14:28:41 [RRSAgent]
RRSAgent has joined #dawg
14:28:41 [RRSAgent]
logging to http://www.w3.org/2005/10/04-dawg-irc
14:29:14 [DanC]
RRSAgent, stop
14:30:35 [Zakim]
+??P21
14:30:38 [AndyS]
zakim, ??P21 is AndyS
14:30:38 [Zakim]
+AndyS; got it
14:30:42 [Zakim]
+DanC
14:31:24 [DanC]
DanC has changed the topic to: RDF Data Access http://www.w3.org/2001/sw/DataAccess/ 4 Oct. scribe: AndyS
14:31:30 [DanC]
Zakim, take up item 1
14:31:30 [Zakim]
agendum 1. "Convene, take roll, review records and agenda" taken up [from DanC]
14:31:37 [DanC]
Meeting: RDF Data Access Weekly
14:31:41 [DanC]
Scribe: AndyS
14:31:43 [DanC]
Chair: DanC
14:32:07 [JanneS]
JanneS has joined #dawg
14:32:48 [DanC]
Regrets: SteveH, LeeF, DaveB, JosD
14:32:53 [DanC]
Regrets: SteveH, LeeF, DaveB, JosD, SouriD
14:32:58 [Zakim]
+[IPcaller]
14:33:13 [SerT]
Enrico and myself are dialing right now
14:33:18 [DanC]
Zakim, IPcaller is JanneS
14:33:18 [Zakim]
+JanneS; got it
14:33:18 [JanneS]
Zakim, IPcaller is JanneS
14:33:19 [Zakim]
sorry, JanneS, I do not recognize a party named 'IPcaller'
14:33:39 [Zakim]
+Enrico_Franconi
14:34:25 [DanC]
Regrets: SteveH, LeeF, DaveB, JosD, SouriD, PatH
14:34:44 [AndyS]
Regrets: EricP
14:35:00 [patH]
But PatH is lurking on IRC.
14:35:12 [DanC]
hi
14:35:15 [AndyS]
Regrets: Kendall
14:35:25 [DanC]
Regrets: SteveH, LeeF, DaveB, JosD, SouriD, PatH, KendallC
14:36:17 [DanC]
Zakim, who's talking?
14:36:28 [Zakim]
DanC, listening for 10 seconds I heard sound from the following: JanneS (79%), Enrico_Franconi (9%), DanC (74%)
14:36:36 [DanC]
Zakim, mute JanneS
14:36:36 [Zakim]
JanneS should now be muted
14:36:59 [Zakim]
+[UMD]
14:37:21 [DanC]
Zakim, who's on the phone?
14:37:21 [Zakim]
On the phone I see AndyS, DanC, JanneS (muted), Enrico_Franconi, [UMD]
14:37:23 [bijan]
bijan has joined #dawg
14:37:29 [AndyS]
zakim, UMD has Bijan
14:37:29 [Zakim]
+Bijan; got it
14:37:30 [DanC]
Regrets: SteveH, LeeF, DaveB, JosD, SouriD, PatH, KendallC, EricP
14:38:25 [JanneS]
I've got headset on now - echo should have gone.
14:38:28 [JanneS]
Zakim, unmute JanneS
14:38:28 [Zakim]
JanneS should no longer be muted
14:38:48 [bijan]
Zakim, who's here?
14:38:48 [Zakim]
On the phone I see AndyS, DanC, JanneS, Enrico_Franconi, [UMD]
14:38:49 [Zakim]
[UMD] has Bijan
14:38:51 [Zakim]
On IRC I see bijan, JanneS, RRSAgent, Zakim, franconi, SerT, patH, AndyS, afs, ericP, DanC
14:38:59 [DanC]
Agenda: http://lists.w3.org/Archives/Public/public-rdf-dawg/2005OctDec/att-0001/_ag4Oct.html
14:39:12 [DanC]
agenda + DAWG schedule
14:40:26 [DanC]
agenda order = 1, 6, 12
14:40:34 [DanC]
agenda order=1, 6, 12
14:40:39 [DanC]
Zakim, agenda order=1, 6, 12
14:40:39 [Zakim]
I don't understand 'agenda order=1, 6, 12', DanC
14:40:43 [DanC]
Zakim, agenda order is 1, 6, 12
14:40:43 [Zakim]
ok, DanC
14:41:14 [DanC]
scribe for Oct 11?
14:41:39 [DanC]
Janne offers for 18 Oct
14:41:50 [DanC]
ACTION: EricP to arrange for 200 response from http://www.w3.org/2005/08/sparql-protocol-query [DONE]
14:41:58 [DanC]
ACTION: KC to revert query-request from xs:all back to xs:seq and note this is due to limiations of XML Schema. [DONE]
14:42:13 [DanC]
next meeting is tentatively scheduled for 11 Oct, contingent on recruiting a scribe
14:42:21 [DanC]
Zakim, next agendum
14:42:21 [Zakim]
agendum 6. "issues#rdfSemantics, issues#owlDisjunction" taken up [from DanC]
14:43:04 [Zakim]
+EliasT
14:43:22 [AndyS]
From PatH summary email
14:44:24 [AndyS]
.. Sergio has summary of change areas
14:44:57 [EliasT]
EliasT has joined #dawg
14:45:16 [EliasT]
Zakim, who's on the phone
14:45:16 [Zakim]
I don't understand 'who's on the phone', EliasT
14:45:22 [DanC]
Zakim, who's on the phone?
14:45:22 [Zakim]
On the phone I see AndyS, DanC, JanneS, Enrico_Franconi, [UMD], EliasT
14:45:23 [Zakim]
[UMD] has Bijan
14:45:55 [AndyS]
Enrico: entailment 1/ def of pattern solution : no Bnodes
14:46:07 [DanC]
-> http://www.w3.org/2001/sw/DataAccess/rq23/#PatternSolutions 2.4 Pattern Solutions
14:46:11 [AndyS]
.. 2/ patten matching include entailment
14:46:22 [bijan]
DanC: http://www.mindswap.org/~bparsia/rdfssemsparql.txt
14:46:30 [bijan]
(not perm)
14:46:32 [bijan]
er.
14:46:33 [AndyS]
Add skolemised and unskolemisation
14:46:34 [bijan]
DanC, http://www.mindswap.org/~bparsia/rdfssemsparql.txt
14:46:41 [EliasT]
Zakim, mute me
14:46:41 [Zakim]
EliasT should now be muted
14:47:12 [AndyS]
Enrico: skolemization found not to work
14:47:21 [DanC]
er... my browser doesn't wrap rdfssemsparql.txt very well
14:47:22 [AndyS]
(Proof?)
14:47:37 [SerT]
skolemization does work
14:47:46 [AndyS]
optional in UNION and SELECT for minimization
14:47:54 [bijan]
It's just text.../me tried ot move fast
14:49:02 [AndyS]
No bnodes in data =>none in results
14:49:55 [patH]
if I follow that, I disagree. Bnode in answer can be bound to a query var in query.
14:49:56 [AndyS]
Syntactic bnodes in dataset => skolemize to constants => remember process => entail => deskolemize with mapping
14:50:22 [AndyS]
(Don't understand why Pat's approach was found wrong)
14:50:31 [patH]
Whoops, sorry, withdraw previous comment.
14:51:36 [patH]
Pat doesnt follow why Pat is wrong, either.
14:51:49 [AndyS]
q+ to ask whether this mapping undone after base pattern or after query
14:53:01 [AndyS]
Enrico: may be a problem with revealing bnodes labels
14:53:19 [AndyS]
.. use case (Ron's ?- ed) not met in this case
14:53:22 [bijan]
DanC, http://www.mindswap.org/~bparsia/rdfssemsparql.html
14:53:39 [bijan]
I don't think you are wrong.
14:53:41 [bijan]
pat
14:56:15 [AndyS]
.. preservation of told bnodes
14:56:39 [AndyS]
.. on abstract syntax, not lean graph
14:58:07 [AndyS]
q+ to ask about named graphs
14:58:23 [Zakim]
+SteveH
14:58:37 [AndyS]
Bijan: lean store = abstract syntax+entailment queries
14:59:26 [SteveH]
SteveH has joined #dawg
14:59:33 [AndyS]
How to specify the query mode: protocol, query flags, (per graph? per base pattern?)
15:00:03 [AndyS]
Per endpoint? Or on a per request basis (server can reject
15:00:32 [AndyS]
No proposal to change the QL at the moment
15:00:59 [SteveH]
zakim, mute me
15:00:59 [Zakim]
SteveH should now be muted
15:01:01 [AndyS]
Default is per service and include in description (scribe checking)
15:01:46 [AndyS]
New "entailment" -> "abstract syntax", "told bnode"
15:02:24 [patH]
It is impossible to follow this from the IRC record. Just a remark.
15:02:48 [AndyS]
DanC: suggest: say "redundant answers are OK"
15:03:40 [patH]
+1
15:04:10 [AndyS]
Enrico: only told bNode redundance ?
15:05:31 [patH]
Suggest not qualifying this, since removing redundacy is potentially high cost for service agent, independent of entailment issues.
15:07:41 [AndyS]
DanC: concern about introducing a new entailment
15:08:01 [AndyS]
.. alt: one kind of service - may get redundancies
15:08:33 [AndyS]
Bijan: service advertises abilities (a property of the service - single valued)
15:09:41 [AndyS]
(My server does it based on graph queried - not just service)
15:10:51 [AndyS]
DanC: Tension between abs syntax and servers that wish to forget it
15:11:03 [AndyS]
Bijan: has examples of both
15:11:18 [AndyS]
DanC: clients loose
15:11:39 [AndyS]
May be possible to have a server that only has lean graphs
15:12:12 [AndyS]
(Does cwm lean graphs?)
15:13:03 [DanC]
(yes)
15:13:11 [DanC]
(at least in some cases, I gather)
15:13:32 [AndyS]
When? During log:semantics?
15:13:35 [DanC]
I'm not sure
15:14:06 [DanC]
DanC: suppose we just say: clients lose; we don't require servers to support querying abstract syntax
15:14:14 [patH]
BUt imagine a RSS-feed type query service, required to be lean might be impossibly expensive and not important in use.
15:14:25 [bijan]
DanC, is an aspect or presumption of your position that we shouldn't specify answers for sparql against graphs under, e.g., rdfs semantics?
15:14:57 [DanC]
quite. I'm sticking to the idea that rdfs semantics is done by closure before query
15:15:02 [patH]
I'd be happy with that.
15:15:22 [bijan]
Oh. We're at completely different points
15:16:02 [patH]
Guys, consider that the RDFS case can be implemented in various ways even if spec describes result in terms of closure or in terms of inference.
15:16:24 [bijan]
Yes, I just want it specified
15:16:35 [bijan]
And in the protocol I want us to identify names for these modes (extensible)
15:16:40 [bijan]
Etc.
15:17:10 [patH]
But there are theorems showing rdfs-entail = simple-entail from rdfs closure. SO why does it matter which way we describe it?
15:17:36 [franconi]
because there is no way to characterise the told-bonode semantics
15:17:41 [bijan]
It doesn't, expect insofar as it prevents extensibily to owl
15:17:53 [franconi]
yes
15:18:09 [patH]
I calim it can be done in terms of scope. Thqats orthogonal issue to the rdf/rdfs/owl?/entailment issue.
15:18:52 [patH]
Bijan, we should be honest about not providing owl entailment. We don't, so lets not pretend otherwise.
15:19:19 [franconi]
Let's not start again ont that :-)
15:19:20 [DanC]
stepping back, this telcon has already been valuable to me; I think i understand Enrico's proposal much more clearly than I did before.
15:20:03 [AndyS]
+1
15:22:37 [bijan]
patH, yes, but I think there is a clear path forward and I would like it to be a simple extension
15:22:44 [bijan]
Rather than a rewrite of some part of the spec
15:23:56 [AndyS]
Test harness
15:24:00 [patH]
Bijan, OK, but we will need to wrap this around with warnings about not expecting too much, etc.. And there is an orthogonal concern, that other, different, generalizations will be blocked in order to keep the entailments pure.
15:24:06 [AndyS]
What effect does this have on the tests?
15:24:33 [AndyS]
Bijan: impls not doing entailment
15:24:41 [AndyS]
(Andy -> no - 3Store)
15:24:52 [patH]
Bijan, I see this already in the concern for a semantic story for told bnodes. THere are clear practical use cases for this.
15:26:38 [bijan]
Yes, but I'm hearing that the test cases might be indifferent to told bnode redundancy!
15:27:03 [AndyS]
[] rdf:type rs:ResultSet ;
15:27:03 [AndyS]
rs:resultVariable "x" ;
15:27:03 [AndyS]
rs:solution [ rs:binding [ rs:value 1 ;
15:27:03 [AndyS]
rs:variable "x"
15:27:03 [AndyS]
]
15:27:04 [AndyS]
] .
15:28:01 [SteveH]
we have a count for queryies with ORDER BY
15:28:22 [SteveH]
(I'm muted, in a noisy room)
15:29:13 [AndyS]
So logical equiv data graphs lead to logical equiv result sets
15:29:49 [AndyS]
Err - no - because result set bnodes aren't redundant.
15:30:32 [patH]
Wait, we have to specify what semantics we are applying to the result sets.
15:30:55 [AndyS]
It's "logical equivance" - which may not be enough
15:30:58 [AndyS]
:-(#
15:31:23 [patH]
Is that defined anywhere for result sets?
15:31:42 [DanC]
yes, if you squint...
15:31:49 [AndyS]
Tests README
15:31:59 [Zakim]
-EliasT
15:32:51 [franconi]
patH: the semantics we gave in our document (conjunctive existential) would be the right one, I guess
15:32:58 [JanneS]
thanks for letting me catch up via just listening you today - have to go now for tonight, bye
15:33:09 [Zakim]
-JanneS
15:33:12 [JanneS]
JanneS has left #dawg
15:33:28 [patH]
Franconi, can you send the URI for that doc?
15:36:08 [DanC]
so http://www.mindswap.org/~bparsia/rdfssemsparql.html is progress on ACTION: Enrico to take a pass through the editor's draft, listing what will change with the new semantics understanding ...
15:36:35 [DanC]
it's done to my satisfaction... some discussion as to whether it's done to the satisfaction of the meeting [not to mention whether we're quorate to discharge actions anyway]
15:36:41 [bijan]
* Ditch the whole GRAPH/NAMED thing (or push it in soma appendix).
15:37:05 [franconi]
http://www.w3.org/mid/a80c7960501c216dc201b3552d81ab03@inf.unibz.it
15:37:10 [AndyS]
AndyS confused by that proposal - can it be in the algebra?
15:37:13 [patH]
Ta.
15:37:57 [DanC]
ACTION Bijan: estimate impact of "abstract syntax entilemnt" etc. on WG test harness
15:38:45 [DanC]
Bijan: ... algebra ...?
15:39:05 [DanC]
AndyS: GRAPH is just the same algebra with 4-tuples rather than 3-tuples
15:39:39 [DanC]
Enrico: yes, at that level, it's just another column in the results[?], but we don't have a semantics at the RDF level [er... he said it better]
15:40:45 [patH]
I know this is considered fighting words, but this sounds to me like the semantics police blocking a useful mechanism.
15:41:18 [DanC]
DanC: yes, months ago, [in Helsinki] I argued against SOURCE/GRAPH on the grounds that we don't really understand it formally... but now the users expect it
15:41:34 [patH]
And, more to the point, they are using it.
15:41:55 [DanC]
Bijan: so... can we put GRAPH in an appendix, or put a warning around it or something?
15:42:07 [DanC]
DanC: yes, I think that's a good idea...
15:42:09 [AndyS]
Proposal to add text to warn there are no formal semantics for GRAPH yet (or in time for rc)
15:42:11 [AndyS]
Proposal to add text to warn there are no formal semantics for GRAPH yet (or in time for rec)
15:42:13 [patH]
And, BTW, it does have a (trivial) clear semantics.
15:42:35 [patH]
Even though it might not be in a W3C spec yet :-)
15:42:52 [AndyS]
Idea: formalise during CR period
15:42:54 [DanC]
... some discussion of saying that GRAPH is at risk during CR, with a CR exit criterion that we formalize it.
15:43:14 [franconi]
patH: what are we blocking exactly? My attitude is to formaise whatever users do. So, if I forgot something, let me understand what.
15:43:26 [patH]
we already did, in the named graph paper. BUt Id wlecome discusion/criticism.
15:44:01 [patH]
franconi, maybe I was reading too mcuh intot he enigmatic IRC record. The red flag was the calim that named graphs have no semantics.
15:44:14 [AndyS]
Many users want GROUP.
15:45:16 [AndyS]
I thought Pat had at least a first cut on its formalization - Enrico - do you have comments on that (later? email?)
15:45:45 [patH]
Lets do it by email, it may take time. Certainly details :-)
15:45:48 [franconi]
Aha, I don't claim that they can not have semantics, I only claim that at the current stage of SPARQL it may take some (long) time to have it in a complete form. So, it is a matter of timing, I guess
15:46:19 [patH]
OK, then Im more optimisitc. Lets talk about this by email.
15:46:34 [franconi]
sure
15:46:37 [AndyS]
Dan -- Schedule discussion
15:46:49 [AndyS]
CR ==> ~Oct 37th
15:46:57 [AndyS]
XQuery - got comments
15:47:01 [patH]
BTW, we worked iit out for the common logic propposal in some detail. I'll send you the URI.
15:47:18 [AndyS]
q+ to ask about XSD decimals
15:47:38 [AndyS]
SWBP review - some comments
15:47:39 [patH]
Pat has to leave, his trun to present the next paper :-)
15:47:46 [AndyS]
Thx Pat
15:47:52 [bijan]
ta, pat
15:47:57 [bijan]
ta, pat's ghost
15:51:34 [AndyS]
(I would need a clearer explaination of the issue re XQuery after all this time)
15:52:31 [bijan]
Ha!
15:52:47 [AndyS]
The chair is committed to get to CR.
15:56:42 [EliasT]
10AndyS: 01CR ==> ~Oct 37th
15:57:15 [DanC]
Zakim, take up item schedule
15:57:15 [Zakim]
agendum 12. "DAWG schedule" taken up [from DanC]
16:05:26 [Zakim]
-SteveH
16:06:53 [DanC]
ADJOURN.
16:06:58 [Zakim]
-Enrico_Franconi
16:07:01 [Zakim]
-[UMD]
16:09:28 [Zakim]
-DanC
16:09:30 [Zakim]
-AndyS
16:09:31 [Zakim]
SW_DAWG()10:30AM has ended
16:09:32 [Zakim]
Attendees were AndyS, DanC, JanneS, Enrico_Franconi, Bijan, EliasT, SteveH
16:09:51 [SteveH]
SteveH has joined #dawg
16:12:55 [DanC]
RRSAgent, make logs world-access
16:31:25 [AndyS]
rrsagent, please leave
16:31:25 [RRSAgent]
I see 3 open action items saved in http://www.w3.org/2005/10/04-dawg-actions.rdf :
16:31:25 [RRSAgent]
ACTION: EricP to arrange for 200 response from http://www.w3.org/2005/08/sparql-protocol-query [DONE] [1]
16:31:25 [RRSAgent]
recorded in http://www.w3.org/2005/10/04-dawg-irc#T14-41-50
16:31:25 [RRSAgent]
ACTION: KC to revert query-request from xs:all back to xs:seq and note this is due to limiations of XML Schema. [DONE] [2]
16:31:25 [RRSAgent]
recorded in http://www.w3.org/2005/10/04-dawg-irc#T14-41-58
16:31:25 [RRSAgent]
ACTION: Bijan to estimate impact of "abstract syntax entilemnt" etc. on WG test harness [3]
16:31:25 [RRSAgent]
recorded in http://www.w3.org/2005/10/04-dawg-irc#T15-37-57