IRC log of sparql on 2010-10-19

Timestamps are in UTC.

13:37:37 [RRSAgent]
RRSAgent has joined #sparql
13:37:37 [RRSAgent]
logging to http://www.w3.org/2010/10/19-sparql-irc
13:37:49 [Zakim]
Zakim has joined #sparql
13:38:00 [AxelPolleres]
trackbot, this will be sparql
13:38:00 [trackbot]
Sorry, AxelPolleres, I don't understand 'trackbot, this will be sparql'. Please refer to http://www.w3.org/2005/06/tracker/irc for help
13:38:08 [AxelPolleres]
Zakim, this will be sparql
13:38:08 [Zakim]
ok, AxelPolleres; I see SW_(SPARQL)10:00AM scheduled to start in 22 minutes
13:38:19 [AxelPolleres]
Chair: Axel Polleres
13:39:19 [AxelPolleres]
regrets: Steve Harris, Ivan Herman
13:39:33 [AxelPolleres]
Agenda: http://www.w3.org/2009/sparql/wiki/Agenda-2010-10-19
13:40:10 [AxelPolleres]
regrets: Steve Harris, Ivan Herman, Olivier Corby
13:48:48 [AxelPolleres]
regrets: Steve Harris, Ivan Herman, Olivier Corby, Lee Feigenbaum
13:48:55 [ericP]
heya Axel, i'll can come to the conf around 10 after
13:48:59 [ericP]
'zat work?
13:50:45 [cbuilara]
cbuilara has joined #sparql
13:52:48 [AxelPolleres]
eric, should be ok, that's when we just have finished admin stuff
13:52:57 [AxelPolleres]
... ideally, don't be later ;-)
13:53:26 [sandro]
partial regrets, sorry. I'll be able to call in for a bit in the middle.
13:58:21 [Zakim]
SW_(SPARQL)10:00AM has now started
13:58:28 [Zakim]
+??P3
13:59:03 [MattPerry]
MattPerry has joined #sparql
13:59:26 [Zakim]
+kasei
13:59:47 [bglimm]
bglimm has joined #sparql
14:00:05 [Zakim]
+MattPerry
14:00:29 [AxelPolleres]
who is P3?
14:00:45 [cbuilara]
me I think
14:00:56 [Zakim]
+AxelPolleres
14:01:07 [cbuilara]
I was the first in joining the call
14:01:16 [AxelPolleres]
Zakim, ??P3 is cbuilara
14:01:16 [Zakim]
+cbuilara; got it
14:01:28 [cbuilara]
thanks
14:01:33 [Zakim]
+??P12
14:01:36 [AndyS]
zakim, ??P12 is me
14:01:36 [Zakim]
+AndyS; got it
14:01:54 [chimezie]
chimezie has joined #sparql
14:01:56 [AxelPolleres]
Zakim, who is on the phone?
14:01:56 [Zakim]
On the phone I see cbuilara (muted), kasei, MattPerry, AxelPolleres, AndyS
14:02:26 [AxelPolleres]
Paul, will you be joining? (I have you on the list for scribing :-) )
14:02:38 [AxelPolleres]
trackbot, start meeting
14:02:40 [trackbot]
RRSAgent, make logs world
14:02:41 [Souri]
Souri has joined #sparql
14:02:42 [trackbot]
Zakim, this will be 77277
14:02:43 [trackbot]
Meeting: SPARQL Working Group Teleconference
14:02:43 [trackbot]
Date: 19 October 2010
14:02:44 [Zakim]
ok, trackbot; I see SW_(SPARQL)10:00AM scheduled to start 2 minutes ago
14:03:28 [AxelPolleres]
Zakim, who is on the phone?
14:03:32 [Zakim]
I notice SW_(SPARQL)10:00AM has restarted
14:03:38 [Zakim]
On the phone I see cbuilara, kasei, MattPerry, AxelPolleres, AndyS, bglimm, Chimezie_Ogbuji, Souri
14:03:50 [bglimm]
Zakim, mute me
14:04:04 [Zakim]
bglimm should now be muted
14:04:04 [AxelPolleres]
Zakim, who is on the phone?
14:04:08 [Zakim]
+pgearon
14:04:10 [Zakim]
On the phone I see cbuilara, kasei, MattPerry, AxelPolleres, AndyS, bglimm (muted), Chimezie_Ogbuji, Souri, pgearon
14:04:10 [chimezie]
Zakim, mute me
14:04:17 [Zakim]
Chimezie_Ogbuji should now be muted
14:04:24 [AxelPolleres]
scribe: Paul Gearon
14:04:45 [AxelPolleres]
http://www.w3.org/2009/sparql/wiki/Agenda-2010-10-19
14:04:49 [AxelPolleres]
topic: admin
14:05:00 [AxelPolleres]
PROPOSED: Approve minutes at http://www.w3.org/2009/sparql/meeting/2010-10-12
14:05:30 [pgearon]
seconded
14:05:31 [AxelPolleres]
RESOLVED: Approve minutes at http://www.w3.org/2009/sparql/meeting/2010-10-12
14:05:39 [AlexPassant]
AlexPassant has joined #sparql
14:06:00 [AxelPolleres]
Next regular meeting: 2010-10-26 ...
14:06:06 [AndyS]
Regrets for next time: AndyS
14:06:51 [pgearon]
AxelPolleres: Already new comments on draft publications
14:07:16 [pgearon]
goal to get to last call by the end of the year
14:07:17 [AxelPolleres]
make clear in thadvertisements that we want to get to LC by end of the year
14:07:34 [AxelPolleres]
... to be discussed within chairs, suggestions for further advertisement welcome.
14:07:51 [Zakim]
+[IPcaller]
14:08:06 [AlexPassant]
I just joined
14:08:17 [AxelPolleres]
Zakim, IPCaller is AlexPassant
14:08:17 [Zakim]
+AlexPassant; got it
14:08:28 [AxelPolleres]
topic: Federated Queries
14:09:08 [AxelPolleres]
1) the "must be bound issue" (certainly, potentially bound
14:09:13 [AxelPolleres]
2) error behavior (be it for {SERVICE t P} a) t not a service, b)
14:09:13 [AxelPolleres]
call to t fails/returns error/query refused, or c) t unbound),
14:09:13 [AxelPolleres]
3) incomplete/missing semantics definition (particularly for the case
14:09:13 [AxelPolleres]
when t is a variavble
14:09:13 [AxelPolleres]
4) informal style
14:09:14 [AxelPolleres]
5) "UNDEF" (i.e. UNBOUND/NULL) ... for BINDINGS is strange/nowhere
14:09:16 [AxelPolleres]
explained
14:09:18 [AxelPolleres]
6) further editorial comments by Greg
14:11:10 [AxelPolleres]
http://lists.w3.org/Archives/Public/public-rdf-dawg/2010JulSep/0433.html
14:11:16 [AxelPolleres]
(greg's mail)
14:11:28 [AxelPolleres]
greg: examples are very domain specific
14:11:52 [kasei]
SERVICE <sparql?default-graph=foo>
14:12:33 [AxelPolleres]
7) allow to add dataset as parameter?
14:12:44 [pgearon]
kasei: don't see any reason not to allow this, since it already allowed in the protocol
14:12:57 [AndyS]
Good idea -- use FROM, FROM NAMED (else need to construct SERVICE uri dynamically in SPARQL?)
14:13:49 [pgearon]
AxelPolleres: interested in feelings on these topics, rather than solutions at this stage
14:14:28 [ericP]
Zakim, please dial ericP-office
14:14:28 [Zakim]
ok, ericP; the call is being made
14:14:30 [Zakim]
+EricP
14:14:40 [pgearon]
AxelPolleres: on topic 2. Currently, if there is an error, then the entire service should fail. Perhaps this is not the best approach
14:14:54 [kasei]
i'd want OPTIONAL { SERVICE <s> { ... } } to not cause the entire query to fail if the service call fails
14:15:00 [AxelPolleres]
ad 2) ... SERVICE SILENT?
14:16:16 [AndyS]
+1 to kasei
14:16:44 [pgearon]
ericP: yes, wrote that a service failure should result in total failure, else results will not be as constrained as clients think they are
14:17:04 [AxelPolleres]
OPTIONAL { SERVICE SILENT would do that job, greg?
14:17:21 [NickH]
NickH has joined #sparql
14:17:24 [pgearon]
ericP: wrote that the query fails, . Perhaps say that the GroupGraphPattern in which the service appears fails to match
14:17:46 [pgearon]
AxelPolleres: so it would leave all the variables unbound
14:17:51 [AxelPolleres]
"fail" would leavea ll vars unbound, i.e. = {}
14:18:04 [Zakim]
+??P38
14:18:09 [kasei]
{}, not {{}}
14:18:18 [NickH]
Zakim, ??P38 is me
14:18:18 [Zakim]
+NickH; got it
14:18:52 [pgearon]
ericP: no solutions, as opposed to one solution with no bindings
14:19:21 [pgearon]
ericP: makes it more complicated, but there are use cases that justify it (I think)
14:19:31 [AxelPolleres]
editors should make a proposal to the list...
14:19:51 [pgearon]
AndyS: editors write something up and propose to the list, rather than detailed design here
14:19:56 [pgearon]
ericP: seconded
14:20:25 [pgearon]
AxelPolleres: no explicit actions yet, but will provide a general action to do this
14:20:27 [ericP]
http://www.w3.org/2009/sparql/docs/fed/service#defn_service
14:20:51 [ericP]
[[
14:20:52 [ericP]
If E is of the form SERVICE VAR {P}
14:20:52 [ericP]
Then
14:20:52 [ericP]
Let G := Join(G, Service(VAR, G, Transform(P)))
14:20:54 [ericP]
]]
14:20:58 [pgearon]
point 3 - incomplete semantics
14:21:10 [ericP]
[[
14:21:14 [ericP]
Let R be the empty multiset
14:21:14 [ericP]
foreach i in Ω(?var->i)
14:21:14 [ericP]
if i is an IRI
14:21:14 [ericP]
R := Union(R, Join( Invocation( i, vars ∩ bound, P, Bindings(G, vars) ) , Ω(?var->i) ) )
14:21:17 [ericP]
else
14:21:20 [ericP]
exection fails.
14:21:22 [ericP]
the result is R
14:21:25 [ericP]
14:21:27 [ericP]
]]
14:22:01 [pgearon]
ericP: copied what was happening in the GraphGraphPattern. Hope AndyS can look it over and confirm that it appears OK
14:22:32 [pgearon]
ericP: made it a binary operation
14:22:35 [AndyS]
I'll take an action to look at it when the editors say "it's ready for you" (if not now)
14:22:58 [pgearon]
ericP: will take this to text. In email
14:23:07 [kasei]
q+
14:23:44 [pgearon]
point 5 - UNDEF not defined clearly
14:24:06 [AxelPolleres]
axel: 5) another approach from "UNDEF" would be "," ?
14:24:08 [pgearon]
kasei: UNDEF doesn't bother me, but it requires a definition
14:24:17 [AndyS]
,, no!
14:24:28 [AxelPolleres]
greg: UNDEF doesn't bother me... just explanation missing
14:24:54 [AndyS]
not adjacent commas for absence
14:25:07 [AxelPolleres]
greg: highlight again the examples are too specific.
14:25:14 [pgearon]
kasei: informal style may change over time, but domain specificness is a current issue
14:25:32 [AxelPolleres]
... easier to grasp use cases would be better.
14:25:42 [pgearon]
kasei: readers should not be required to understand "computational biology" (the current use case)
14:26:07 [AxelPolleres]
... again something we leave to the editors...
14:26:35 [AxelPolleres]
FWIW, I have some SERVICE example using a FOAF File and DBLP in my tutorial slides, maybe a start.
14:26:45 [pgearon]
ericP: current use cases were cut-and-pasted because they were syntactically valid
14:27:35 [pgearon]
AxelPolleres: Point 6 - further comments from Greg can be addressed in mail. Point 7 can be discussed in mail as well.
14:27:38 [kasei]
q+
14:28:02 [AxelPolleres]
subtopic: point 7)
14:28:50 [AndyS]
Why not SERVICE <> { GRAPH <g> {...} }
14:28:55 [AxelPolleres]
axel: personaly, the use case I see is that FROM/FROM NAMED wasn't allowed in subqwueries and allowing it in SERVICE would allow to circumvent this
14:29:22 [pgearon]
kasei: as it stands right now, service keyword is restricted to the graphs provided by the endpoint
14:29:27 [AxelPolleres]
greg: only useful for service in the default graph
14:29:27 [Zakim]
-AlexPassant
14:29:44 [pgearon]
AndyS: this may not be the "default graph" but it certainly applies to services that provide a fixed dataset
14:29:56 [ericP]
http://www.w3.org/2005/01/yacker?name=SPARQL_11&replace=1&lang=perl#prod-SPARQL_11-ServiceGraphPattern
14:30:11 [pgearon]
+q
14:30:17 [AxelPolleres]
eric: this is how it would look like with dataset clause
14:30:41 [kasei]
q-
14:30:50 [AxelPolleres]
ack pgearon
14:31:07 [Zakim]
+??P40
14:31:10 [AlexPassant]
Zakim, ??P40 is me
14:31:10 [Zakim]
+AlexPassant; got it
14:31:20 [AxelPolleres]
paul: wondering about interaction between fed doc and protocol
14:31:22 [ericP]
Definition: Evaluation of a Service Pattern
14:31:33 [ericP]
[[
14:31:35 [ericP]
where:Invocation(IRI, S, P, B) is an implementation of the SPARQL protocol
14:31:35 [ericP]
* against endpoint IRI,
14:31:35 [ericP]
* with a SELECT of S, WHERE pattern P, and BINDINGS B,
14:31:35 [ericP]
* with no default-graph-uri or named-graph-uri (see SPARQL Protocol [SPROT] section 2.1.1.1),
14:31:37 [pgearon]
pgearon: does this issue require some interaction between the service keyword and the protocol document?
14:31:38 [ericP]
Execution failures cause the query to fail.
14:31:41 [ericP]
14:31:43 [ericP]
]]
14:32:13 [AxelPolleres]
D )
14:32:29 [AxelPolleres]
q?
14:33:46 [pgearon]
AxelPolleres: Carlos and Eric, please look into this. Will follow up in 2 weeks
14:33:56 [AxelPolleres]
ACTION: Axel to ping carlos/eric in 2-3 weeks for progress on Fed query issues
14:33:56 [trackbot]
Created ACTION-324 - Ping carlos/eric in 2-3 weeks for progress on Fed query issues [on Axel Polleres - due 2010-10-26].
14:34:27 [pgearon]
new topic: Certainly bound vs. potentially bound
14:35:02 [pgearon]
AxelPolleres: potentially bound is a syntactic definition. proposed to provide a stricter definition that requires binding
14:35:18 [pgearon]
AxelPolleres: Use the mailing list for further discussion. kasei has some opinion there
14:35:40 [pgearon]
new topic: test cases
14:36:01 [AxelPolleres]
topic: test cases
14:36:07 [pgearon]
ericP: how stable is the grammar right now?
14:36:21 [AxelPolleres]
andyS: ISSUW-59 might still change the grammar
14:36:23 [pgearon]
AndyS: issue-59 may change the grammar, but it is not super volatile
14:36:27 [Zakim]
-EricP
14:36:49 [pgearon]
AxelPolleres: fix the test case vocabulary
14:36:55 [AxelPolleres]
http://www.w3.org/2009/sparql/docs/tests/README.html
14:37:28 [AxelPolleres]
ut:request for update actions
14:38:02 [AxelPolleres]
graphstore state
14:38:12 [AxelPolleres]
ut:data ut:graphData
14:39:03 [AxelPolleres]
ut:graphData [ ut:data graph rdfs:label name ]
14:39:18 [pgearon]
AxelPolleres: ut:graphData can allow graphs to be named
14:39:35 [AndyS]
This second use of ut:data should be changed.
14:39:41 [pgearon]
AxelPolleres: some question on the domain of ut:data
14:39:53 [AxelPolleres]
... ut:graph instead ut:data ?
14:40:25 [AxelPolleres]
would that be ok?
14:40:35 [pgearon]
AxelPolleres: Andy had a problem with reuse of ut:data, so I'm proposing to move to ut:graph
14:40:38 [NickH]
nope
14:40:42 [AxelPolleres]
Can you still hear me?
14:40:44 [AxelPolleres]
:-)
14:40:58 [kasei]
as long as all of the applicable tests are changed at the same time, I'm fine with mostly anything.
14:41:17 [NickH]
sounds fine to me
14:41:22 [AxelPolleres]
paul: also for using a new predicate.
14:41:48 [pgearon]
AxelPolleres: will change the nested use of ut:data to ut:graph
14:41:56 [AxelPolleres]
ACTION: Axel to chang nested use of ut:data to ut:graph
14:41:57 [trackbot]
Created ACTION-325 - Chang nested use of ut:data to ut:graph [on Axel Polleres - due 2010-10-26].
14:42:24 [AxelPolleres]
http://www.w3.org/2009/sparql/docs/tests/README.html#entailevaltests
14:42:31 [pgearon]
AxelPolleres: updated test vocab will then be stable (IMO)
14:43:03 [AndyS]
q+
14:43:48 [pgearon]
AndyS: update vocab still. ut:result and others
14:44:12 [pgearon]
AxelPolleres: this is longer, so voer entailments first and get back to it
14:44:13 [bglimm]
seems ok
14:44:15 [bglimm]
Zakim, unmute me
14:44:16 [Zakim]
bglimm should no longer be muted
14:44:58 [pgearon]
AxelPolleres: different entailment regimes for different graphs, may be possible, but would fix regime for the whole document
14:45:13 [bglimm]
Zakim, mute me
14:45:13 [Zakim]
bglimm should now be muted
14:45:26 [chimezie]
breaking in and out, Axel
14:45:31 [pgearon]
AxelPolleres: update use case descriptions
14:46:17 [kasei]
AxelPolleres?
14:46:34 [chimezie]
Zakim, unmute me
14:46:34 [Zakim]
Chimezie_Ogbuji should no longer be muted
14:46:44 [AlexPassant]
yes chimezie
14:46:54 [NickH]
oh dear
14:46:58 [kasei]
heh
14:47:05 [pgearon]
AxelPolleres may be having IP troubles
14:47:56 [AxelPolleres]
AxelPolleres has joined #sparql
14:48:19 [AxelPolleres]
ACTION: Axel to add a sentence in README.html that typically suffix .rq .ru are to be sused for query and update request files
14:48:19 [trackbot]
Created ACTION-326 - Add a sentence in README.html that typically suffix .rq .ru are to be sused for query and update request files [on Axel Polleres - due 2010-10-26].
14:48:47 [AxelPolleres]
mf:result for update TCs...
14:48:54 [chimezie]
Zakim, mute me
14:48:54 [Zakim]
Chimezie_Ogbuji should now be muted
14:49:02 [AxelPolleres]
... ut:success and ut:FAILUREMESSAGE
14:49:34 [pgearon]
AxelPolleres: ut:FAILUREMESSSAGE is defined in the protocol document
14:49:37 [AndyS]
Note MIME registration for SPARQL Update suggests a extension.
14:50:09 [pgearon]
AxelPolleres: looking for someone to define protocol test cases, and also a vocabulary for that
14:50:13 [AndyS]
q+
14:50:32 [AxelPolleres]
Andy, please go ahead...
14:50:36 [bglimm]
yes
14:50:37 [chimezie]
i can
14:50:38 [NickH]
I can AndyS
14:50:42 [pgearon]
AndyS: don't think that's a good idea, for 2 reasons
14:50:46 [AndyS]
Axel can you hear me?
14:50:57 [pgearon]
Axel unable to hear Andy. Everyone else hears him
14:50:58 [AxelPolleres]
I will need to re-dail, I only hear my echo :-(
14:50:58 [chimezie]
Axel, if you can see this it seems as if the communication is one way on your line
14:51:05 [Zakim]
-AxelPolleres
14:52:52 [AxelPolleres]
:-( hardware problem on the telephone :-(
14:53:16 [pgearon]
AndyS - perhaps you can outline, and Axel will see it on IRC?
14:53:31 [AxelPolleres]
hmmm, sincere apologies...
14:53:38 [AxelPolleres]
let's try with chat...
14:53:51 [pgearon]
AndyS: 2 problems. First, we need format stabilized so people can write test code parsing against it
14:53:54 [AxelPolleres]
... seems I can't solve the phone issue quickly.
14:54:17 [AxelPolleres]
+1 on that.
14:54:40 [pgearon]
AndyS: we have a list of errors in the README that don't make sense to me
14:55:04 [pgearon]
AndyS: list - SUCCESS, REQUEST REFUSED, SYNTAX ERROR
14:55:19 [pgearon]
AndyS: SYNTAX ERRORS should be in syntax errors
14:55:31 [pgearon]
AndyS: what should the errors be targetting?
14:56:43 [AxelPolleres]
the errors in the doc came from protocol
14:56:51 [AlexPassant]
unexistent graph ?
14:57:26 [AxelPolleres]
I'd be ok to go without ut:success and the errors for the moment... but we'd want to have test cases for protocol as well, I assume?
14:57:32 [Zakim]
-Chimezie_Ogbuji
14:57:42 [pgearon]
AndyS: insert data, and possibly other operations do automatic graph creation. Are there cases where all stores must signal an error in some conditions?
14:57:52 [AxelPolleres]
we'd need someone voluteering for TCs and TC structure for protocol...
14:57:55 [AxelPolleres]
any volunteer?
14:58:38 [pgearon]
AndyS: is that an error for stores that can't tell the difference between no graph and empty graph
14:59:39 [pgearon]
AndyS: made assumption that loading a graph that doesn't exist to automatically create the graph.
14:59:47 [AlexPassant]
"In case no RDF data can be retrieved from documentURI, the SPARQL 1.1 Update service is expected to return failure."
14:59:52 [AlexPassant]
I meant does not exist *remotely*
14:59:53 [pgearon]
AndyS: propose taking this to email to get a definitive list
15:00:05 [AndyS]
ah! I understand now.
15:00:11 [AxelPolleres]
for now, would people be ok with the update TC format, if we just drop the ut:success/ut:FAILORCODE stuff for updateEvaluationTests?
15:00:48 [AxelPolleres]
agree to take this on email.
15:00:52 [AndyS]
Yes. Haven't read it with all the changes but looks OK.
15:01:00 [AxelPolleres]
sincere apologies for the technical problames again.
15:01:34 [AndyS]
BTW need different syntax tests for query and update.
15:01:49 [AxelPolleres]
Can others add support for that proposal with +1 ... "drop ut:success/ut:FAILORCODE for now for update TC format" ?
15:02:01 [AxelPolleres]
if no opposition, I will implement this...
15:02:15 [bglimm]
Zakim, unmute me
15:02:15 [AlexPassant]
I'll be at ISWC
15:02:16 [pgearon]
telephone portion of this meeting appears to have wrapped up
15:02:17 [AndyS]
+1
15:02:21 [Zakim]
bglimm should no longer be muted
15:02:26 [AxelPolleres]
but as I can't hear you, I;d like some support on the ICR
15:02:32 [AxelPolleres]
s/ICR/IRC/
15:02:45 [pgearon]
AndyS: when are people leaving for ISWC? How many meetings will be missed?
15:02:47 [bglimm]
I leave on Friday 5th comming back on Mon 15th
15:02:55 [AlexPassant]
be away for 2 telcos (5->21)
15:02:57 [Zakim]
-Souri
15:03:06 [pgearon]
bglimm: going Friday, so only missing 1
15:03:16 [AlexPassant]
AxelPolleres: fine with your proposal
15:03:33 [Zakim]
-MattPerry
15:03:33 [AxelPolleres]
ok, I have two supports... I will go ahead with that and send around mail.
15:03:42 [pgearon]
thanks Axel
15:03:49 [AndyS]
ADJOURNED
15:03:54 [pgearon]
AndyS: we're now adjourned
15:03:59 [AxelPolleres]
ACTION: Axel to drop ut:success and ut:FAILURECODES from update TC vocab.
15:03:59 [trackbot]
Created ACTION-327 - Drop ut:success and ut:FAILURECODES from update TC vocab. [on Axel Polleres - due 2010-10-26].
15:04:35 [AxelPolleres]
Thanks for keeping on, we seem at least to have some format for UC and entailment tests now... approval hopefully next time.
15:04:58 [Zakim]
-bglimm
15:04:59 [NickH]
bye all!
15:05:00 [AxelPolleres]
Please, all TC providers, work through your unapproved TCs following this conclusion.
15:05:02 [Zakim]
-NickH
15:05:03 [AxelPolleres]
THanks all!
15:05:06 [AxelPolleres]
Bye
15:05:22 [Zakim]
-kasei
15:05:25 [Zakim]
-pgearon
15:05:28 [Zakim]
-AndyS
15:05:34 [cbuilara]
-cbuilara
15:05:41 [Zakim]
-cbuilara
15:05:46 [cbuilara]
bye
15:06:00 [Zakim]
-AlexPassant
15:06:02 [Zakim]
SW_(SPARQL)10:00AM has ended
15:06:03 [Zakim]
Attendees were kasei, MattPerry, AxelPolleres, cbuilara, AndyS, bglimm, Chimezie_Ogbuji, Souri, pgearon, AlexPassant, EricP, NickH
15:07:03 [AndyS]
Pls could someone agree (or disagree) to two comment responses. Should not be controversial.
15:07:27 [AxelPolleres]
AndyS, can you drop the links?
15:07:36 [AndyS]
2010OctDec/0117 and 0118
15:07:42 [AxelPolleres]
I am in another call now... will look into it right afterwards.
15:10:12 [AxelPolleres]
rrsagent, make records public
15:31:00 [kasei]
has there been agreement on an unbound variable in a GROUP BY leading to dropping of the group? don't recall how that worked out...
15:39:07 [kasei]
AndyS?
15:43:50 [SteveH]
kasei, I don't think it drops the group
15:44:07 [SteveH]
it evaluates to an "error", and can be key'd like any other value
15:44:18 [kasei]
oh. I was trying to make sense of the "Aggregates question" thread from a couple weeks ago...
15:44:45 [kasei]
in particular, this: http://lists.w3.org/Archives/Public/public-rdf-dawg/2010OctDec/0040.html
15:45:01 [kasei]
I thought I understood your answer to Axel as indicating that groups with unbound variables are dropped.
15:45:08 [kasei]
I misunderstood that, then?
15:45:09 [SteveH]
let me read...
15:45:19 [SteveH]
it's not unknown for me to contradict myself :)
15:45:27 [kasei]
:)
15:46:10 [kasei]
but i couldn't make sense of Andy's follow-up later on in the thread...
15:47:20 [SteveH]
kasei, http://lists.w3.org/Archives/Public/public-rdf-dawg/2010OctDec/0058.html
15:47:28 [SteveH]
there's several things going on there
15:47:56 [SteveH]
but ListEval(unbound) = error
15:48:16 [SteveH]
however, projecting an error value, results in a dropped result
15:48:39 [kasei]
ah
15:48:54 [kasei]
so without projecting ?Q in that example, there would be two results?
15:49:51 [SteveH]
[vague waving of hands] yeah, I think so
15:49:59 [kasei]
haha. ok, thanks Steve
15:50:09 [SteveH]
Aggregate() doesn't remove rows
15:51:11 [kasei]
hrmm. wait. where does the ListEval in this case happen?
15:51:24 [kasei]
in the grouping, right?
15:54:11 [kasei]
I'm trying to sort out if the :group05 test is right.
15:54:55 [kasei]
it does: SELECT ?s ?w { ?s :p ?v OPTIONAL { ?s :q ?w } } GROUP BY ?s ?w
15:56:10 [kasei]
the results contain rows that have ?s bound and ?w unbound, but I'm not sure why that's different from Axel's example in email.
15:59:09 [SteveH]
you can still make a group of (<a>, error)
15:59:13 [SteveH]
AFAICR
16:02:23 [kasei]
but it's then projected, just like it is in the email example.
16:18:31 [SteveH_]
SteveH_ has joined #sparql
16:24:06 [AndyS]
kasei
16:25:25 [AndyS]
re: you can still make a group of (<a>, error)
16:26:14 [AndyS]
wrong way round: (error, {1,2,3})
16:26:15 [kasei]
so what is it that causes the group to be dropped in Axel's example in the email?
16:26:33 [AndyS]
No idea.
16:26:37 [kasei]
heh
16:26:43 [kasei]
ok
16:28:22 [AndyS]
He is proposing the opposite case but I see no reason why that's more valuable.
16:28:44 [AndyS]
I think my suggested modifications to the wording work.
16:28:53 [AndyS]
Any issues you see?
16:33:01 [AndyS]
What's your architectural issue?
16:34:16 [AndyS]
kasei, have you followed the limit-by-resource discussion on semantic-web@w3?
16:35:35 [kasei]
not sure about the architectural issue as I'm not sure I've fully understood the impact of errors on aggregates+projection yet.
16:35:46 [kasei]
I've followed the limit-by-resource discussion a little bit, but not in depth.
16:36:09 [AndyS]
I wonder if this might turn into a "easier to do than explain everytime" thing. LIMIT 10 by ?x, or other similar syntax, looks quite easy to do.
16:36:37 [kasei]
I tend to agree
16:36:52 [kasei]
not thrilled about adding more syntax/features at this point, but this issue does come up a lot.
16:37:48 [AndyS]
My feeling exactly. Hard to balance small features and overrun with freeze for <small n> years.
16:49:35 [kasei]
speak of the devil
16:49:42 [kasei]
someone wants to talk to you about it on #swig
17:02:52 [Zakim]
Zakim has left #sparql
18:31:37 [AndyS]
AndyS has joined #sparql
18:51:41 [AxelPolleres]
AxelPolleres has joined #sparql
19:21:05 [AxelPolleres]
AxelPolleres has joined #sparql
20:22:51 [SteveH]
SteveH has joined #sparql
20:43:27 [AndyS]
Scanning sparql.org log files. Recently rebooted to install latest joseki version : Of 2326 requests, 1715 involve SERVICE
20:44:36 [AndyS]
Some are used to add SPARQL 1.1 features to 1.0 endpoints!
21:16:45 [AxelPolleres]
AxelPolleres has joined #sparql