RDB2RDF Working Group Teleconference -- 12 Jan 2010

<trackbot> Date: 12 January 2010

<mhausenblas> scribenick: Souri

ahmed: happy new year to everybody
... regrets from Marcelo

angela: yes

wolfgang: yes

mike: minutes approval

<mhausenblas> http://www.w3.org/2009/12/15-RDB2RDF-minutes.html

<mhausenblas> PROPOSAL: Accept the minutes of the 15 December 2009 telecon

<angela_UNITN> +1

<mhausenblas> RESOLUTION: Accept the minutes of the 15 December 2009 telecon

<mhausenblas> actions?

Juan?

<mhausenblas> issues?

when you speak, please first identify yourself to help me scribe

<mhausenblas> trackbot, issues?

<trackbot> Sorry, mhausenblas, I don't understand 'trackbot, issues?'. Please refer to http://www.w3.org/2005/06/tracker/irc for help

<jsequeda> yes, it was me. When i find out more about the sparql to sql patent from microsoft, I`ll let everybody know

ahmed: Oracle SQL-based approach, view definition specified in SQL
... The other approach was to use new RDF vocabulary to specify the views
... how to proceed from here
... requiring lot of changes in DB cannot be a requirement

<mhausenblas> ISSUES?

ahmed: but you get a lot of perf out of that
... look deeper to find some support at the client level or config level
... what will the WG do from now on ...
... Linked Data area will be led by Michael
... waiting for Michael to send more on that
... start debating ...

<Zakim> hhalpin, you wanted to deadlines

<mhausenblas> Michael: notes that http://www.w3.org/2001/sw/rdb2rdf/wiki/Use_Cases_and_Requirements exists

<MacTed> (bug report for Zakim / trackbot maintainers -- Zakim doesn't make connection between "this will be" and "this is" and doesn't accept "this is" before start-time ... so trackbot should use "this is" and must say so *after* the session start-time ... or some other fix which accomplishes same)

Harry: tight timeframe, just 2 yrs, some proposal should be ready by April

Michael: agree with Harry

<hhalpin> Again, should be no problem extending things for a few months, but the main issue is "how to do it"

ahmed: use case needs to come from discussion

<hhalpin> I am a bit neutral on this one...

ahmed: how to go about creating the user cases?

ashok: SQL-based approach is not Oracle-specific

<hhalpin> ACTION: hhalpin to update web-page and schedule [recorded in http://www.w3.org/2010/01/12-RDB2RDF-minutes.html#action01]

<trackbot> Created ACTION-18 - Update web-page and schedule [on Harry Halpin - due 2010-01-19].

ashok: there are other approaches that are very similar: Triplify, UltraWrap

<mhausenblas> actions?

<mhausenblas> issues?

ashok: Gather all these groups (Triplify, Oracle, Ultrawrap) together to write a document

orri: agree, SQL-based style could be the most aligned to DB
... challenge on impl side would be to take the sparql and rewrite it into sql
... worried about the efficiency of the translation

ashok: is sparql-to-sql translation relevant for this WG?

soren: strongly support those parts that are not covered by database systems
... how sparql query can access the data stored in the database tables

<hhalpin> We should take the most minimal starting point, but one that can extended. Perhaps the mapping first from tables and views, and then if we have time, then see what we can do re SPARQL->SQL mapping.

soren: one page syntactic description to access the base tables
... would like to see how the translation from sparql to sql would work
... rdbms to rdf on one hand, and sparql query to sql query on the other hand

orri: on-the-fly mapping from sparql to sql is more complex

<hhalpin> trackbot, ISSUE-2

<trackbot> Sorry, hhalpin, I don't understand 'trackbot, ISSUE-2'. Please refer to http://www.w3.org/2005/06/tracker/irc for help

<hhalpin> trackbot, ISSUE-2?

<trackbot> Sorry, hhalpin, I don't understand 'trackbot, ISSUE-2?'. Please refer to http://www.w3.org/2005/06/tracker/irc for help

<angela_UNITN> (then sparql endpoints...)

<hhalpin> ISSUE-2?

<trackbot> ISSUE-2 -- R2RML serializations -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/2

ahmed: mapping rel data to RDF is simpler, but in our env we need on-demand kind of support

if we don't do on-demand, we have twice the cost

orri: if we don't do on-demand, we have twice the cost

<Ashok> Souri: We are thinking of looking at RDB data as RDF data. Data need not be materialized

<Ashok> ... most impt part is the mapping part so we can help user with writing the SQPARQL query

<Zakim> hhalpin, you wanted to doing two -distinct steps, first mapping then possibly SPARQL to SQL

Harry: I am hearing consensus from the group to do the mapping first (R2ML language)

<hhalpin> The mapping is declarative...

Harry: then investigate the SPARQL to SQL translation approach.

<hhalpin> I agree that ETL isn't that used.

orri: the mapping is most important

Juan: liked what ahmed said at the beginning about writing a document

<hhalpin> +1 on writing proposal for each "family" of approaches

Juan: would like to see document regarding how to map from RDB to RDF using the two approaches -- SQL-based view def and RDF vocab (such as D2R) based view def

<hhalpin> Souri: Writing a SQL view is one approach, to go with SQL language to specify a query as a view definition, and then use that with some declarative mapping

<hhalpin> Souri: using that to define the mapping, the advantage of that is that the relational data is completely specifiable.

<hhalpin> Souri: the vocabulary-based approach is to add new words with RDF, express the view, but using the RDF as the underlying way to view the relational data.

<mhausenblas> queue

<hhalpin> Souri: So the D2RQ based approach has been popular, but not sure about coverage

orri: having considered both SQL and RDF vocab sides, even with the latter with some annotations it is possible to translate

soren: the query that defines the view does not need to be standardized, its only the mapping of view cols, names to RDF terms

souri: fully agree

harry: the mapping could be specified as XML

<hhalpin> so we can outline how to do the SQL to get view, then we can do the mapping via the XML file.

ahmed: do you kind of agree with Souri's point?

<soeren> +1

<angela_UNITN> +1

Michael: sounds ok to me

<hhalpin> I mean, the real question would be could the XML file also be used for more D2RQ ETL sort of things

ashok: looks fine, the only question is we are speaking of two approaches, right?

<hhalpin> I would agree, but not sure of the details, as I haven't implemented any of this myself yet.

ashok: there does not look like any conflict.

ahmed: summarize

ashok: rel to rdf vs. sparql-to-sql?

ahmed: converging into sql-based views and then mapping the names (view, view-col, constraint) to RDF terms

<hhalpin> so what I am saying is that it's unclear if we need the view, although that's obviously the sensible way to do this!

harry: can we just use XML to do the mapping?

<hhalpin> and debate over the listserv as well :)

Michael: propose give one week to the WG members to think about it

<hhalpin> We can create a W3C WBS poll

Michael: may be a proposal to consider

<hhalpin> but I'd like a bit more clarity about what options to put in the poll

<hhalpin> or the exact questions

<hhalpin> before I volunteer to make the poll.

Harry: a poll would make a lot of sense

Michael: the wording needs to be created for a poll. Souri?

souri: I'll do it.

Michael: draft it on the wiki? or whichever is better ...

how do I do action?

<mhausenblas> ACTION: Souri to draft a wording for the basic direction the WG takes [recorded in http://www.w3.org/2010/01/12-RDB2RDF-minutes.html#action02]

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

<mhausenblas> ACTION: hhalpin to put Souri's draft into a WBS poll and notify the WG [recorded in http://www.w3.org/2010/01/12-RDB2RDF-minutes.html#action03]

<trackbot> Created ACTION-19 - Put Souri's draft into a WBS poll and notify the WG [on Harry Halpin - due 2010-01-19].

soren: eric ...?

<hhalpin> EricP is definitely a SPARQL->SQL translation

<soeren> I can not hear anything anymore :-(

ashok: do not get very hung up on the XML requirement

<hhalpin> no worries ashok, I just want a public draft, I am happy to XML-ize whatever language people use.

$souri to do create a proposal text about the sql-based approach that the WG may want to vote on

<mhausenblas> actions?

ahmed: XML is not a critical issue today, agree with ashok

<mhausenblas> Michael: I agree with Ashok

ahmed: any situation where creating a view is not possible

orri: in some cases in practice it is not obvious, but can be done

ahmed: like that comment, mapping language is another level, Does the vendor need to know the exact RDF to RDF mapping?

orri: no

ahmed: so it can be done either way
... vendor specific features
... any thoughts? at what level primarily the support should be present?

Soren: you can use vendor-specific extensions may be used to define the view

ahmed: some vendors have sql extensions beyond standard, we should allow

soren: Triplify solves this problem by allowing users to define output processing functions
... functions can for example translate html to text and so on

orri: the functions should be reversible?

soren: why reversible?

<hhalpin> Works for me!

souri: reversible is good for performance, but not essential

<hhalpin> I would like to return to some draft proposals...

<hhalpin> It can be changed, and can be very very drafty, that's OK.

ahmed: send the text to the WG via email and then we refine it and finally post it for a poll
... timeframe for first draft

ashok: first draft by April

ahmed: when would it be posted? May or June?

ashok: depends upon the kind of agreement or disagreement on the first draft. Estimate for posting about 3 months after the first draft is created.

orri: do we need a requirement doc?

ashok: highly recommended

Harry: clarifying ... first draft need not be a complete agreement, more like a heartbeat check
... if we put together a proposal in three months and keep debating that is ok
... requirement document will be a W3C Note

<angela_UNITN> +q

Harry: Use Case doc will need to be created

<hhalpin> I think the use-cases should be available as soon as possible, but the XG report has some use-cases already in it.

Harry: ahmed: can we write the proposal before we have a Use Case document

ashok: we can do in parallel

<hhalpin> +1 ahmed

ahmed: a set of people will start writing the proposal and share with the group ...

<mhausenblas> Michael: I take care of editing the UC

<hhalpin> Souri: will agree to help draft a proposal.

<mhausenblas> Michael: I will collect stuff from http://www.w3.org/2001/sw/rdb2rdf/wiki/Use_Cases_and_Requirements

<angela_UNITN> i can write a UC ...i have some questions about that maybe i can discuss with the future UC group

<hhalpin> We need at least 2, ideally 3 editors ASAP.

ahmed: who will volunteer? let us decide as a group

<hhalpin> Could we do the write-up on the wiki?

<mhausenblas> yes, hhalpin

ahmed: share with the group as it is written to get feedbacks from the group

<hhalpin> I can help with formatting, no problem.

ashok: editing etc.

<hhalpin> In general, we need 2-3 editors.

ahmed: let us decide on the participants as a group

harry: 2-3 editors, comments from everybody, address the comments, and so on

Michael: for the recommendation track need to track and address comments
... at least 2-3 editors
... ahmed could ask for volunteers ...

AOB

Michael: send out email to the group asking for volunteers
... next time Mike will chair and send Linked Data related stuff

<mhausenblas> [adjourned]

- DRAFT -

RDB2RDF Working Group Teleconference

12 Jan 2010

Attendees

Contents

AOB

Summary of Action Items

Scribe.perl diagnostic output