RDB2RDF Working Group Teleconference

20 Jul 2010


See also: IRC log


+1.562.249.aaaa, +1.603.897.aabb, MacTed, +3539149aacc, seema, mhausenblas, +49.322.222.0.aadd, Ashok_Malhotra, nunolopes, juansequeda, cygri, Marcelo, soeren, boris, hhalpin, Souri, +1.512.471.aaee, Lee_Feigenbaum, [IPcaller], dan, angela, EricP
seema, mhausenblas


<trackbot> Date: 20 July 2010

<mhausenblas> aacc is me

<MacTed> hhalpin - do you have the new UK number?

<mhausenblas> +44.203.318.0479 (London)

<mhausenblas> MacTed see above

<MacTed> mhausenblas - not for me, for hhalpin :-) +44.203.318.0479

<mhausenblas> scribenick: seema

<mhausenblas> aaee is dan

<mhausenblas> http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Jul/0030.html


<hhalpin> +1

<mhausenblas> PROPOSAL: Accept the minutes of last meeting, see http://www.w3.org/2010/07/13-rdb2rdf-minutes.html

RESOLUTION: minutes of last meeting http://www.w3.org/2010/07/13-rdb2rdf-minutes.html accepted

Meetings in July/August

<mhausenblas> http://doodle.com/u6qg5z7gmfhc2zak

R2RML Semantics

<hhalpin> Notes according to schedule we were supposed to get first draft of mapping language out in June...

<hhalpin> I'll ask for a 3 month extension, but even something very rough out by end of August would be ideal.

<mhausenblas> http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Jul/0028.html

Harry: would like summer meetings so that we can continue to work on the mapping language charter

<juansequeda> https://docs.google.com/document/pub?id=1APTqD2lJLRjwV6gmPKqDRqC3aT8bozHF0udIXndMNWQ

<juansequeda> http://www.w3.org/2001/sw/rdb2rdf/wiki/Database-Instance-Only_and_Database-Instances-and-Schema_Mapping

<LeeF> FYI you can often get HTML from google docs that can be pasted into the wiki after the fact


<juansequeda> Datalog, SQL, RIF, Eric - https://docs.google.com/document/pub?id=1APTqD2lJLRjwV6gmPKqDRqC3aT8bozHF0udIXndMNWQ

<juansequeda> http://www.w3.org/2001/sw/rdb2rdf/wiki/Database-Instance-Only_and_Database-Instances-and-Schema_Mapping

Marcelo: Idea is to have a running example in our document. DB instance mapping is the default mapping
... the 3 rules can also be generated automatically
... no ontology used in case 2 and case 3. hence the rules can be generated automatically

<alexdeleon> The IPcaller must be me

Ashok: why start with an example, instead of starting with the rules of the language

Marcelo: the example is just to make the rules easier to undestand

<Souri> which case are you explaining?

Richard: can we use aggregates in queries?

Marcelo: yes, the predicate can be replaced with aggregates

<hhalpin> good question - we need to figure out what fragment of Datalog aggregates is in, and if we can do that in RIF as well.

Souri: In datalog, is there a limit to the language we can use ?

<hhalpin> Using Datalog is still up for debate I think.

<ericP> datelog has aggregates?

Marcelo: up to us to decide what to include.. maybe a simplification..no recursion, no aggregates ..most basic form of SQL

<soeren> +q

can someone fill in me on who is speaking ? is it MacTed?

<juansequeda> its harry

<ericP> hhalpin


<LeeF> Can't people use SQL to produce whatever views they want before applying RDB2RDF standards *regardless* of what the syntax of the RDB2RDF mapping standard looks like?

<Souri> +1 to LeeF

<nunolopes> LeeF: +1

Harry: do we assume that users will know SQL ? so that we allow additional SQL power to the language ?

<LeeF> on DAWG & SPARQL WG, SteveH is fond of telling me that SQL does not at all line up with relational algebra (though the details are way beyond me)

<cygri> +1 to ashok

<hhalpin> LeeF, true and that's one way to do it but then what would RDB2RDF do precisely...

<hhalpin> the issue is that most people will be much happier probably using SQL than either RIF or some weird custom XML syntax IMHO

<mhausenblas> scribenick: mhausenblas

Marcelo: I agree with juansequeda

<MacTed> note that SQL-92 (aka SQL-2) has levels -- Entry, Intermediate, and Full ... SQL-99 (aka SQL-3) and later, have similar levels

<MacTed> SPARQL is a bit less mature, but I think the same should be coming

<scribe> scribenick: seema

<MacTed> grrr

sorry, got kicked out of irc

<dan> im q+

Marcelo: whatever is represented in Datalog can be represented easily in other languages.

Richard: agree that datalog has well defined semantics. However SQL is richer.
... do you want to consider different implementation by diff vendors and base a standard on that ? Would not agree with that.

<Ashok> +1 to Richard

thanks Michael

<LeeF> If the mapping language is a pretty thin wrapper around SQL & punts on the semantics of the wrapped SQL... that doesn't seem all too useful to me (off the top of my head)

<LeeF> +1 to going around in circles a lot :D

<LeeF> agree with macted

MacTed: SQL is not monolithic or simple. Lots of things in standard that leave specific implementation out to the vendor.

<juansequeda> I agree with the levels of SQL that MacTed is talking about. We should have levels of our mapping language

MacTed: we have slightly diff syntax per vendor
... however DB vendors can declare compliance to a SQL level

<mhausenblas> +1 to MacTed's simple and to-the-point view

<Souri> SQL *query* standards could be used as a guideline. But, are we saying that SQL??? features {x,y,z,a,...} are the only ones that are supported OR, going the other way, saying that SQL??? features {u,v,w,...} may *not* be portable?

Souri: the query portion of SQL is standard.
... we could have vendors state something along the likes of SQL quote above..some parts are portable, others are not

marcelo: datalog can be easily converted to SQL. then SQL can be enhanced.
... people in DB industry doing data integration use Datalog, not SQL

<juansequeda> Dan is talking about : https://docs.google.com/document/pub?id=1APTqD2lJLRjwV6gmPKqDRqC3aT8bozHF0udIXndMNWQ

<LeeF> Ashok, mhausenblas - if we were going to take a straw poll on this conversation now... what would the question be? i'm still not convinced we all agree/understand what the exact question under debate is

<Souri> Steps: 0) (not a step) you have a db schema present, 1) you want to transform it to a desired schema (=> do it using SQL query), 2) use a mapping language to map the interface of the queries to the OWL classes and datatype properties 3) use a mapping language to map the ref integrity constraints to OWL object properties

<hhalpin> the question is the meta-level approach.

<mhausenblas> Good question LeeF - I'm still trying to figure ...

+1 to Lee

<hhalpin> Datalog is just an "option" to bridge SQL and RIF/RDF approaches.

<juansequeda> hhalpin, what other options of bridges do we have? I'd be interested if somebody could propose something else

<mhausenblas> LeeF care to do q+ and ask this to everyone? :)

<Souri> So, a basic question is what should be the mapping langauge? Does user have to write the mapping in datalog? Do we have to write a parser for translating the datalog spec to SQL query?

<hhalpin> 1) Use all of SQL, basic mapping in XML.

<hhalpin> 2) Use a default mapping, then do mapping in RIF

<juansequeda> Souri, datalog as a syntax... but we don't have to use that. We can define another syntax if we want

<hhalpin> 3) Which part of SQL and RIF can we hold in common (Datalog ,which is just a semantics to my knowledge)

<LeeF> The fact is... we have SQL to transform between relational models.

<LeeF> And we have RIF (or SPARQL) to translate between RDF models (sort of).

<LeeF> So the only gap is that bit in between (creating URIs, data types, etc.)

<LeeF> so let's just define the simple simple mapping and call it a day

<LeeF> we talked a bit at SemTech about enumerating the list of things that would need to be in that mapping

Richard: address Harry's options : using all of SQL would be popular, though easy to understand

<hhalpin> or SPARQL Constructs rather than RIF.

Richard: option 2 addresses RDF audience..but they don't use RIF currently

<hhalpin> he said just using all of SQL would be popular.

<mhausenblas> thanks LeeF !

<juansequeda> cygri: RDF to RDF is something we shouldnt address +1

<MacTed> +1

<Zakim> ericP, you wanted to ask what piece of datalog we're talking about

<Ashok> Richard, could you write mail summarizing what you said and we can discuss next Tuesday

<cygri> Ashok: ok

<mhausenblas> thanks cygri

<LeeF> good observation, Dan!

Dan: as an audience for this document : 1 audience is us, 2 is folks not us but implementing the audience, 3) folks who have to live with this

<LeeF> curious who _is_ going to implement this

<LeeF> would Informatica implement this?

<hhalpin> The implementers are obviously database vendors.

<hhalpin> Therefore, we have to be *most* sensitive to their needs.

<mhausenblas> hhalpin I wouldn't be that sure (re obviously database vendors) - there are others ATM and there may be others

<Souri> what goes to the database has to be a SQL query

<hhalpin> Well, if database vendors don't implement it, it won't get into the hands of end-users

<mhausenblas> but I agree that we should be very sensitive re DB vendors, yes

<hhalpin> +1 Souri.

Marcelo: Language for mapping in commercial tools is often Datalog


<Souri> question is what is the mapping language to use? the spec of course needs to get translated to SQL before sending it to the database.

<dan> thanks - go t to go...

<hhalpin> So whatever mapping language we use has to either be SQL or be a fragment of SQL with some definite extensibility, or easily translated into SQL.

<mhausenblas> Michael: all the best in taking this further, see you in late August

<mhausenblas> ok, seema, let's wrap up

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2010/07/20 17:05:00 $