SPARQL Working Group Teleconference

21 Apr 2009


See also: IRC log


Chimezie, Bijan




<kjetil> I've told the secretary to check the Bristol Hotel and book me there, and then a lift up to HP would be nice :-)

<trackbot> Date: 21 April 2009

<scribe> Scribe: LeeF

<scribe> Scribenick: LeeF

<kasei> Zakim just hung up on me :(

<AndyS> Maybe. zakim did not speak the full messages

<SimonS> four

<PrateekJain-WSU> Hi This is Prateek Jain,937 775 4638

AxelPolleres: plan today is to get through the rest of the features from the wiki and go over Web survey
... survey will be open for 1.5 weeks or so, to give us an idea of where to go from the F2F topic on


PROPOSED: Approve minutes at http://www.w3.org/2009/sparql/meeting/2009-04-14

RESOLUTION: Approve minutes at http://www.w3.org/2009/sparql/meeting/2009-04-14

next meeting: one week from today, 28 Apr, will talk about F2F details

scribe for next meeting: Ivan M


AxelPolleres: rdf:text is basically finished, not sure when it will go to Last Call

<AxelPolleres> http://www.w3.org/2007/OWL/wiki/InternationalizedStringSpec

AxelPolleres: if we want to review it, it would be great

<AndyS> I volunteer (not exclusively)

<SteveH_> tentative volunteer, but I can't promise

<scribe> ACTION: AndyS to review rdf:text [recorded in http://www.w3.org/2009/04/21-sparql-minutes.html#action01]

<trackbot> Created ACTION-8 - Review rdf:text [on Andy Seaborne - due 2009-04-28].

<scribe> ACTION: SteveH to try to review rdf:text [recorded in http://www.w3.org/2009/04/21-sparql-minutes.html#action02]

<trackbot> Created ACTION-9 - Try to review rdf:text [on Steve Harris - due 2009-04-28].

AndyS: there will be substantive issues based on what I've seen

AxelPolleres: RIF WG had F2F in Cambridge last week
... plan is to go to LC by end of May
... will appreciate SPARQL WG reviews then

ericP: HCLS group is doing stuff with federated queries

<Zakim> LeeF, you wanted to ask about 90 min. teleconference?

<JanneS> I'll drop out after 60, sorry

introduction, Prateek

PrateekJain-WSU: PhD student at Wright State work wtih Amit Sheth
... research is in the area of query rewriting with emphasis on SPARQL
... trying to exploit semantic relationships within a knowledge base to automatically rewrite SPARQL
... interested in rdf serialization of queries and path queries

feature survey

<AxelPolleres> http://www.w3.org/2002/09/wbs/35463/features/

AxelPolleres: Each organization can fill out the survey once
... we will probably have around 8 features which we will aim for in the working group
... the survey lists 31 features that survived the "interested for anyone" criteria
... format of the survey was limited by what the WBS survey gave us
... options for each feature are ranks 1 - 31 and "don't mind" and "don't want"
... do not rank all features
... rank up to the first 8 of your choices

AndyS: are you going to enforce the limit?

LeeF: we will ask anyone who ranks more than 8 to adjust their choices to only rank 8

SteveH_: Don't agree with only ranking 8 - if my top 4 don't get done, i don't get to express an opinion about the bottom half of things
... voting shouldn't have any different weight just because you rank 4 vs. ranking 30

<Zakim> kjetil, you wanted to ask if the rank algorithm can't account for people ranking all

kjetil: if we use ranking algorithm, people can rank as many as they which

<john-l> I prefer using a ranking algorithm.

LeeF; I was concerned that organizations interested in 25 features should not be able to cast 'more' votes and influence things more than someone who casts less

AndyS: Concerned that everyone be playing by the same rules

AxelPolleres: i think if we use a threshold like 12 or so we can compromise

<SteveH_> LeeF, things like Condorcet don't give any advantage to ballot stuffers

<kjetil> Here's a site we can use for the final ballot: http://www.cs.cornell.edu/andru/civs.html

<john-l> I propose that we have every organization rank ALL of the features, and then use a Condorcet system to eliminate all but 8-12 winners.

<SteveH_> or, alternative: http://plugin.org.uk/rdf/condorcet/

LeeF: I'm not comfortable at all with approving a specific ranking to drive things forwards

<SteveH_> we don't need the threashold

SteveH_: with condorcet you're only voting against yourself
... rank the features you want in the order you'd like them and then we can analyze the data

<kasei> hearing lots of interference on kjetil(?)

<SteveH_> for ref. http://en.wikipedia.org/wiki/Condorcet_Method

<ericP> no tactical voting? i quit!

SteveH_: with Condorcet there's no advantage at all to ranking fewer or more choices, nor to ranking two things the same

kjetil: it's just about the relative preference

LeeF: it's important to me that "all 1 votes" doesn't mean "everything is super important!" but instead "i don't care which of these we do, they're all equally important" - it sounds like people are on the same page about that

AxelPolleres: suggested deadline for filling out the survey is May 1

<JanneS> the vote page has April-28 set as the deadline

<ericP> i would like to propose a new voting scheme

<ericP> it uses parameterized owl entailment

LeeF: encourages everyone to fill out the survey as soon as you feel ready to

-> http://www.w3.org/2009/sparql/wiki/SPARQLX

<AxelPolleres> strawman from bijan: http://lists.w3.org/Archives/Public/public-rdf-dawg/2009AprJun/0089.html

<AndyS> That is SPARQL algebra in XML , not SPARQL AST

ericP: i've had people use XML version of queries for debugging in conjuncgtion with XSLT, can see some use of it

AndyS: abstract syntax need to be formally addressed if we do pragmas

<JanneS> http://www.w3.org/2009/sparql/wiki/Feature:Pragmas

AndyS: TopQuadrant uses an RDF serialization to store queries

<ericP> +1 to two

<SimonS> +1 on two

<ericP> i'm 0 on XML and -1 on RDF ('cause it's so concentious)

<AndyS> My requirement is to get the abstraction right and do XML, JSON, another

<SimonS> +q

<Zakim> ericP, you wanted to say i've had people use an XML expression of queries

Prateek: we are looking at SPIN for some of our work

SimonS: can see use cases for RDF serialization of SPARQL, don't have any use cases for XML

<JanneS> was the motivation to query sparql via sparql?

ericP: i've seen this be contentious before - in particular the expression of a graph pattern in RDF - difficult space to work in

<Zakim> ericP, you wanted to warn of contentious issues

<SimonS> kind of. Rather composing SPARQL queries at runtime from RDF data

<ericP> upshot: whatever we do, we don't want people expressing their data in RDF

AxelPolleres: straw poll on RDF serialization of SPARQL queries

<ericP> -1

<SteveH_> -1, it's a minefield

<AndyS> 0

<kjetil> 0

<PrateekJain-WSU> +1

<LukeWM> -1

<JanneS> sounds like closure to the extreme i.e. sparql query result could be a valid sparql query


<john-l> 0

<AlexPassant> 0

<kasei> 0

<JanneS> 0

<AxelPolleres> 0

<SimonS> 0 would be nice, but difficult

AxelPolleres: straw poll on XML serialization of SPARQL queries

<john-l> +1

<AndyS> 0

<LukeWM> 0


<SteveH_> 0, could be useful, but not huge usecases for us

<PrateekJain-WSU> +1

<ericP> 0

<JanneS> 0

<AxelPolleres> 0

<kasei> +1

<SimonS> -0

<kjetil> 0

<AlexPassant> 0

bijan is a +1 by proxy, I'm positive


-> http://www.w3.org/2009/sparql/wiki/Feature:FunctionLibrary

<AxelPolleres> Related here http://www.w3.org/2005/rules/wiki/DTB

<AxelPolleres> http://www.w3.org/TR/xpath-functions/

<AxelPolleres> http://jena.sourceforge.net/ARQ/library-function.html

LeeF: this feature is about spending the working group's time expanding the core set of functions that query writers can expect to be interoperable between implementations

ericP: we did spend some time testing last time to make sure that extension functions work in a sane way

<SteveH_> +1 to ericP

<AndyS> Significant value. Reuse F&O where possible. Fix a set of functions expected everywhere - not too large to ensure universal coverage.

<SteveH_> +1 to AndyS too

AxelPolleres: the question here is whether the WG should expand the list of built-in functions
... straw poll on working on extending function library

<AxelPolleres> 0

<SteveH_> +1

<kjetil> +1


<kasei> +1 but low priority

<john-l> 0

<JanneS> +1

+1 even

<PrateekJain-WSU> +1

<ericP> 0

<SimonS> 0

<AndyS> +1

<LukeWM> +1

<AlexPassant> 0

<kjetil> (but yeah, at the end)

<AndyS> http://www.w3.org/2009/sparql/wiki/Feature:CreatingIrisAndLiterals


<AxelPolleres> http://www.w3.org/2009/sparql/wiki/Feature:FullText

kjetil: most Web sites have search boxes... if you want to use a triple store for an application and SPARQL on top of that
... you need some way to communicate a search down to SPARQL endpoint
... we could standardize in several ways here
... could have a function in function library
... could also use xpath/xquery text functions
... users might want to search more than just one literal
... lots of possibilities but think this is important for interoperability

SteveH_: I read the xpath full text specification - more complicated than I thought it would be

<ericP> xpath full text examples

SteveH_: this is more complex than just referring to xpath spec, because there's a lot there that is xpath-specific
... on the other end of a scale, LIKE syntax is just syntactic sugar over regex

<AndyS> SteveH_: similarity to LIKE

SteveH_: would need to understand what we're talking about - the LIKE syntax is easy to standardize, full text feature will be a lot of work just to figure out what parts of the XPath full text doc are relevant and which aren't

<ericP> http://www.w3.org/2009/sparql/wiki/Feature:FullText doesn't mention regex

<ericP> what uses cases does regex not support?

kjetil: Virtuoso feels that it is harder to implement regular expressions than full text

<SteveH_> ericP, stemming for one

AndyS: i think XQuery full text is too big, too complicated
... it's important to ensure that existing tools, not just lucene, can be used

<Zakim> SteveH_, you wanted to ask about XPath fulltext v's regex

<Zakim> AndyS, you wanted to talke about XQ full text and to say it's not a library function

AndyS: this is not just a library function - it's not a restriction, since it's generative from an index
... "find me all the things that match 'x'"
... "find me all the URIs of documents that contain the following string"
... concerned that scripting engines end up with a real burden to implement this

<SteveH_> not just smaller implementations, the only impl. of XPath fulltext i've seen is huge

<Zakim> LeeF, you wanted to ask about xpath compared with lucene and to note potential contention

<ericP> the scripting people can be incomplete. doesn't really matter except for bragging rights

<kasei> ericP: it matters for portability between scripting and the bigger impls.

<ericP> kasei, i think that's an argument for then being sound, but not complete

<ericP> i was arguing that if the burden of implementation keeps some scripts from being complete, that's not a big cost. however, we'd like to see them interoperate even in their subsets

LeeF: i see a strong case for interoperability here, but also see a huge amount of work to standardize this well

AndyS: there are other things to consider as well, such as scoring of results

<SteveH_> I don't want to see a world where we end up just standardising lucene syntax

AxelPolleres: seems that xpath/xquery might be too heavy for us, as opposed to aligning what existing implementations do

<AndyS> ack to SteveH_ -- starting point only

<SteveH_> ericP, it's not no

<Zakim> AxelPolleres, you wanted to ask about redundancy wrt regex

<ericP> we'll have a hell of a fight if we want to standardize a non-xpath function if there's an xpath function nearby

<AndyS> Maybe good to add some clear syntax to make it easier to understand for query writers

<JanneS> (I need to leave, will be voting -1 for the full fledged proposal with XQuery and XPath Full Text 1.0, +1 if we add some syntactic sugar to ease interoperability with Lucene type of implementations)

<ericP> burden will be on us to prove that it's not covered by subsetting ftcontains

SimonS: most implementations I know are very close to or based on Lucene - that seems to be what people want
... syntax extensions are similar as well

<SteveH_> or, what implemetors found it easiest to build

<JanneS> (uh, BasicFederatedQuery deserves the protocol extensions is my +1 for both, no comment on LimitPerResource yet) - bye

ericP: I think that if we try to standardize something analogous to an XQuery function (e.g. lucene:contains) then we will need to prove to the world & XQuery WG that we were not able to subset ft:contains to meet our needs

AndyS: to be pragmatic, i don't want to put an xquery parser inside my sparql impl

<ywang4> see you next time :)

SteveH_: it would be tough to separate e.g. how it refers to rdf literals rather than xml nodes

<AndyS> --> http://www.w3.org/TR/xpath-full-text-10/#id-grammar

AxelPolleres: straw poll on full text

<Zakim> SimonS, you wanted to say most existing implementations seem quite similar

<kjetil> +1

<ericP> 0

AxelPolleres: straw poll on full text

<SteveH_> 0

<AndyS> +1

<kasei> -1

<AxelPolleres> 0

<john-l> +1

<AlexPassant> +1

<PrateekJain-WSU> 0

<SimonS> +1

<LukeWM> +1

0, against my better judgment which says +1

<ericP> nice

<ericP> i'm surprised. usally "better judgement" is aligned with discresion


<AlexPassant> http://www.w3.org/2009/sparql/wiki/Feature:LimitPerResource

AlexPassant: proposal is to find a way to limit solutions not by tuples but by distinct instances of a resource
... example syntax on wiki page

kjetil: this is the single most important feature for us
... you don't know in advance how many rows you expect back

LeeF: is it true that if sparql has subselects then limitperresource is syntactic sugar?

SteveH_: you also need grouping/limiting operations
... in other cases besides http://www.w3.org/2009/sparql/wiki/Feature:LimitPerResource#Related_Use_Cases.2FExtensions you need a grouping operator

AxelPolleres: Alex & Kjetil would you be happy if this was subsumed by subselect?

kjetil: yes, if we do subselect

<AxelPolleres> let's do a strawpoll conditional to subselects

AlexPassant: yes, it's the capability itself that is important

<ericP> 0

<kjetil> +1

<AlexPassant> +1

<kasei> +1

AxelPolleres: straw poll - would you want LimitPerResource GIVEN that we do not do subselects

<AndyS> -1

<john-l> +1

<AxelPolleres> 0

<SteveH_> -1


<LukeWM> 0

<PrateekJain-WSU> 0

<SimonS> 0

<SteveH_> [but we do do this a lot, but doing it without subselects would be crazy]

Basic federated queries


AndyS: goal is to find minimal features needed to make federation happen
... what's the minimal needed for one sparql endpoint to call out to another to get some results back
... related thing for sending CONSTRUCT query to another processor as part of the FROM clause to get data in - can do that now with long URLs
... very related to query parameters
... very related to subqueries

<Zakim> LeeF, you wanted to ask about relationship with subselect

AndyS: overall task in federated query is to find the right place to get a certain piece of information
... missing piece is the ability to actually make the call to a remote service, that's the minimum required piece
... ARQ does with SERVICE keyword, Virtuoso does it with pragma attached to subselect - mechanism less important than the feature

<AxelPolleres> Axel: syntax to "execute federated query plans, yes?

<Zakim> ericP, you wanted to say that he added the encode-this-and-append-to-graph-url functionality on the command line

ericP: HCLS group does a lot of query federation using my command line stuff
... useful but not top priority

SimonS: we define special named graphs that are evaluated remotely and then do subqueries against one or more remote endpoints
... a lot of people use it

<Zakim> SimonS, you wanted to say he does this using special named graphs

AxelPolleres: straw poll on basic federated query

<kasei> +1

<kjetil> +1

<LukeWM> -1

<SteveH_> 0, useful, but very very scary

<AndyS> +1

<ericP> +1

<SimonS> +1

<john-l> -1

<PrateekJain-WSU> +1

<AlexPassant> 0


<AxelPolleres> 0

<ericP> john-l, i'm curious about your -1. issue of priorities, or serious concearns?

ericP, you're not concerned about Luke's -1?

<john-l> ericP, Just priorities.

<ericP> and LukeWM?

query by reference and parameters

AxelPolleres: doesn't parameters need query by reference?

<ericP> LeeF, shoudl i talk about SPARQLfed grammar?

SimonS: Parameters are useful without query by reference, for example for distributed joins

<LukeWM> ericP, my -1 was because I've just not come across any use cases for it day to day

<ericP> roger

<ericP> tx

LeeF: query by ref might depend on parameters but not vice versa - params got some support, query by ref got no support (no +1s)

SteveH_: are you going to change the survey to not say "make 8 votes"?

AxelPolleres: yes, I will change the text to not be restricted to 8 votes
... AOB?


<SteveH_> bye

Summary of Action Items

[NEW] ACTION: AndyS to review rdf:text [recorded in http://www.w3.org/2009/04/21-sparql-minutes.html#action01]
[NEW] ACTION: SteveH to try to review rdf:text [recorded in http://www.w3.org/2009/04/21-sparql-minutes.html#action02]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2009/04/21 15:36:14 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.135  of Date: 2009/03/02 03:52:20  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/Date:/ Date:/
Succeeded: s/Liaisons:/Liaisons/
Succeeded: s/I'm not/LeeF: I'm not/
Succeeded: s/upshot:/ upshot:/
Succeeded: s/someway/some way/
Succeeded: s/ericP:/ericP,/
Succeeded: s/???/SimonS/
Found Scribe: LeeF
Inferring ScribeNick: LeeF
Found ScribeNick: LeeF

WARNING: No "Present: ... " found!
Possibly Present: AlexPassant AndyS Axel AxelPolleres Garlik JanneS LeeF Lee_Feigenbaum LukeWM P19 P3 P39 P9 PROPOSED Prateek PrateekJain-WSU Scribenick SimonS SteveH SteveH_ aaaa aabb dnewman2 ericP john-l joined kasei kjetil sparql trackbot upshot ywang4
You can indicate people for the Present list like this:
        <dbooth> Present: dbooth jonathan mary
        <dbooth> Present+ amy

Regrets: Chimezie Bijan
Agenda: http://www.w3.org/2009/sparql/wiki/Agenda-2009-04-21
Found Date: 21 Apr 2009
Guessing minutes URL: http://www.w3.org/2009/04/21-sparql-minutes.html
People with action items: andys steveh

[End of scribe.perl diagnostic output]