<trackbot> Date: 23 March 2010

<scribe> scribenick: mhausenblas

Admin

<MacTed> trackbot, start meeting

<trackbot> Sorry... I don't know anything about this channel

<trackbot> If you want to associate this channel with an existing Tracker, please say 'trackbot, associate this channel with #channel' (where #channel is the name of default channel for the group)

<MacTed> grr

<MacTed> but it's not done it right

<MacTed> else Zakim would have reported my join

<Ahmed> Are we ready to start?

<ericP> harry will not be here today

<Souri> Anyone has the URL for the public-rdb2rdf-wg email archive?

<LeeF> Souri, http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/

PROPOSAL: accept minutes of previous telecon http://www.w3.org/2010/03/16-rdb2rdf-minutes.html

<whalb> +1

RESOLUTION: group accepted minutes of previous telecon http://www.w3.org/2010/03/16-rdb2rdf-minutes.html

Ahmed: seems like the discussion on the mailing list seems rather minimal
... we need to discuss the idea of the two-team approach

<Souri> We have sent an email with links to extended version of our example with a Part 3 that has SPARQL Query and its SQL translation (using the SQLdefString).

<Souri> http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Mar/0087.html

<Zakim> LeeF, you wanted to ask a beginner's question

LeeF: do you view the two approaches as mapping language approaches or implementation specific?

Orri: IMO a matter of syntax
... primarily, yes, as of the preferences of the people

LeeF: that was my understanding as well
... but got confused when I reviewed the WG's material

<juansequeda> I think it is Souri

<Seema> it is Souri

thanks ;)

Souri: its about the schema, we need some glue to express it
... view definition should be rather simple

Ahmed: Souri, the mapping the query is an orthogonal issue - we're about mapping data

<LeeF> Souri, the approach you explained sounds very similar to the D2RQ approach, where the "glue" is given via RDF but the ultimate expressivity is given by SQL fragments... i guess the difference is the SQL approach starts with a full-on SQL query/view?

Souri: we look into mapping a SPARQL to SQL query, re implementation

Update from both SQL & RDF mapping approach teams

<soeren> +q

Orri: Virtuoso is somewhere in the middle of the two approaches
... experience shows that a SPARQL2SQL mapping is essential similar to D2R approach

<LeeF> My target audience is definitely _not_ SQL people.

Orri: the SQL syntax is easier to sell (taking the target audience into consideration)
... use cases will make the requirements apparent

<cygri> +1 to everything Souri and Orri said

Ahmed: can be done either way (put load on user or database)

soeren: I think the difference between both approaches is not that big
... in SQL-based approach the view creation is part up-front
... if it can be done in SQL, why redo it?

Ahmed: agree with Soeren - but you don't need to create a view

<Souri> +q

Souri: we are *not* creating a view

Ashok: want to point out that we're a chartered to write a mapping language
... sure we need to have a ML implemented but not SPARQL2SQL

<Souri> +q

<LeeF> The charter gives 2 overall use cases for this group's work:

<LeeF> 1) Dumping RDB data into a triple store

<LeeF> 2) SPARQL to SQL translation

<LeeF> so it seems a little weird to ignore the 2nd overall use case

soeren: you need a SPARQL2SQL which yields the same

Ahmed: let's step back - JDBC as an example; all about decoupling

ericP: from the spec POV it's nice to say: here is how our mapping works
... but from the uptake , it needs to be implementable
... then, performance becomes very relevant

<Ashok> Lee, where did you see "2) SPARQL to SQL translation" ?

ericP: Richard, can you confirm that for D2RQ can map always 1:1 in terms of queries?

<Ashok> It is not in the scope bullets

cygri: AFAICT, it should be possible to always translate 1:1

ericP: do you think the underlying algebra (re D2R) exists?

cygri: for this you need a formal representation for the ML (but this does not exist for D2R)
... if you want to be sure that everything translates 1:1 we'd need it
... would be great to have

ericP: I think we will not have real adoption if we lack this formal semantics

Orri: Virtuoso uses proprietary extensions in some cases

<LeeF> I don't think the question is "can it be done" - the question is, how can I know when I've efficiently handled all cases

Souri: the aspect 'is it translateable' is not in scope (could be in a certain form)
... proofing that a SPARQL query can be translated into x SQL queries, is out of scope

juansequeda: need the semantics of the mapping language; we just guarantee that it works, not how to implement it

<Souri> +q

Ahmed: we should ensure that the ML is semantic valid

Souri: I disagree
... we want to expose relational data as RDF

(explains a simple example)

Juan: we need both syntax and semantics, so we know what the language means

<Marcelo> I agree with Juan

Souri: with SQL approach, we use the SQL semantics which is well-known ... the semantics of the glue is very simple

soeren: how to be sure that a SPARQL query over your mapping into a SQL query is always possible?
... that's independent from the mapping, not part of charter of this group

s/Soeren/Souri/ in above

<MacTed> can you guarantee that any SQL query can be executed?

<MacTed> especially, within reasonable time?

(Ahmed and Souri discuss SQL details)

ericP: we have a rough equivalence between SPARQL and SQL, beside the fact that the latter is much more epxressive
... for example SQL views
... if I use a trivial representation of relational data as, say XML

(ericP has lost the scribe and will fill in the missing bits of his interesting example after the meeting)

Souri: what is the mapping language, really?

juansequeda: 1:1 is simple, do we have 1:many use cases?

MacTed: does it make sense to give such a query-mapping gurantee?

soeren: every valid SQL is; we should at least define which ones are guaranteed to be mapped

MacTed: not saying it is efficient - might take forever

ericP: I ended up owning the UC doc

<cygri> soeren has already dropped from the call

<scribe> ACTION: ericP to review the UC and report back to the WG [recorded in http://www.w3.org/2010/03/23-rdb2rdf-minutes.html#action01]

<trackbot> Sorry, couldn't find user - ericP

<LeeF> ACTION: eric to review the UC and report back to the WG with requests for more details [recorded in http://www.w3.org/2010/03/23-rdb2rdf-minutes.html#action02]

<trackbot> Created ACTION-36 - Review the UC and report back to the WG with requests for more details [on Eric Prud'hommeaux - due 2010-03-30].

<scribe> ACTION: Orri to sum up today's discussion [recorded in http://www.w3.org/2010/03/23-rdb2rdf-minutes.html#action03]

<trackbot> Created ACTION-37 - Sum up today's discussion [on Orri Erling - due 2010-03-30].

[adjourned]

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)

WARNING: Bad s/// command: s/Soeren/Souri/ in above
Found ScribeNick: mhausenblas
Inferring Scribes: mhausenblas
Default Present: [IPcaller], +1.512.471.aaaa, +49.322.222.0.aabb, Ashok_Malhotra, whalb, +3539149aacc, MacTed, mhausenblas, cygri, nunolopes, LeeF, EricP, Souri, Seema, Soeren
Present: [IPcaller] +1.512.471.aaaa +49.322.222.0.aabb Ashok_Malhotra whalb +3539149aacc MacTed mhausenblas cygri nunolopes LeeF EricP Souri Seema Soeren
Regrets: Harry
Agenda: http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Mar/0074.html
Found Date: 23 Mar 2010
Guessing minutes URL: http://www.w3.org/2010/03/23-rdb2rdf-minutes.html
People with action items: eric ericp orri