W3C

- DRAFT -

RDB2RDF Working Group Teleconference

14 Jun 2011

Agenda

See also: IRC log

Attendees

Present
+3539149aaaa, mhausenblas, Ivan, cygri_, +49.153.6.aabb, boris, privera, +1.781.273.aacc, MacTed, Ashok_Malhotra, +1.314.394.aadd, dmcneil, nunolopes, EricP, +1.603.897.aaee, Alexandre, +1.603.897.aaff, Seema, Souri, +575737aagg, juansequeda, LeeF, EFranconi, Michael, Boris, Percy, Ted, Ashok, David, Juan, Lee, Enrico
Regrets
Marcelo
Chair
Michael
Scribe
ivan

Contents


<trackbot> Date: 14 June 2011

<mhausenblas> RRSAgent make logs public

<mhausenblas> scribenick: ivan

<scribe> scribe: ivan

<MacTed> :-)

admin stuff

<mhausenblas> PROPOSAL: Accept the minutes of last meeting http://www.w3.org/2011/05/31-rdb2rdf-minutes.html

mhausenblas: any comment on the minutes?
...
...
...

RESOLUTION: Accept the minutes of last meeting http://www.w3.org/2011/05/31-rdb2rdf-minutes.html

semtech debrief

mhausenblas: who wants to start

dmcneil: i made a talk on r2rml and spider
... fairly full room
... well received
... there were several questions, people followed along with the ideas
... the pressing question was the question i posted on the mailing list
... how to code a uri as a resource
... i got questions on the representation of the mapping itself,
... some people pushed back on the verbosity of turtle
... one gentleman proposed a specific syntax for the mapping

<mhausenblas> Michael: Enrico just send the following to the WG list: "Still stuck on the train with an horrible delay. I can't make it on time :-("

Ashok: what sort of syntax?

dmcneil: data representation, concise to r2rml
... and translate that

<betehess> funny enough, ericP just posted a abstract model this morning

juansequeda: was good to see the state of sw technologies
... i realized that people see that they need rdb to rdf
... eg people from golden sachs
... people from big it companies are waiting for us

mhausenblas: anyone else to report

<mhausenblas> Ivan: Cray has a big offering re iron

<ericP> dmcneil, here's a concrete syntax based on the example at the top of the r2rml spec: http://www.w3.org/2005/01/yacker/uploads/R2RML?lang=perl&markup=html#prod-R2RML-R2RML_Mapping

<ericP> (that'd different from the abstract syntax)

mhausenblas: any other reports regarding semtech
... we should, in general, report when a wg member is at an event, conference, etc
... use that to collect feedback
... getting to last call needs that

action items

mhausenblas: should we go through the action items or jump on the issues
... i know the enrico is delayed, but he may join us later
... i would propose to start with issue 22

issue 22

issue-22?

<trackbot> ISSUE-22 -- Support for database vendor specific SQL statements. -- open

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/22

Souri: i was going to make a comment on issue 18, I have started to update the document on the resolution of that one, should be done for next week

mhausenblas: so, issue 22, who can help?
... dmcneil, you brought this up?

<cygri> Proposal: http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2011May/0179.html

action-133?

<trackbot> ACTION-133 -- Eric Prud'hommeaux to write proposal for issue-22 -- due 2011-06-07 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/133

ericP: the prudent way of handling mine is to see a resolution on whether we flag an extended sql with anything
... we have some use to flag an extension with an identifier
... the debate is whether we want to flag extensions in the first place

mhausenblas: what is the issue you are referring to>

dmcneil: i think that is issue 22

ericP: the metaissue is that if we have a syntax to distinguish between sql 2009 or an sql with vendor specific
... if we have a flag for that, I have a syntax to do that

Ashok: a question
... eric, you were speaking of a flag like a binary one
... or are you speaking of an attribute with values

<mhausenblas> ignore cygri

<mhausenblas> ahm

ericP: the original proposal was a flag, i proposed an attribute
... the dispute comes from cygri who feels it is not necessary

<dmcneil> +q

ericP: the use cases that are enabled with an attribute are not compelling to richard
... so i would like to see a decision on the basic question

Ashok: if we have an attribute, that could be a string, what you end up choosing for, how you actually act based on that, we do not specify
... ie, there is no harm for this

ericP: per richard it makes the spec more difficult to read

cygri: there are a number of issues with this; one is that it puts additional burden on the authors, 'cause they have to say which sql dialect they are using, and I think most of the authors have no idea

<Souri> agree with Richard about SQL writers not being aware of compatibility

cygri: most would have no interest or the knowledge of what to use
... it is quite a complicated thing with different version of sql
... the other issue is that i would expect most of the implementation will be specific to a specific database system
... they know what sql dialect they are connected to
... at that point the assumption is that if the author binds to an oracle specific database, then it is a reasonable assumption that the author will expect that the sql dialect is oracle's
... the information is redundant, adds no value

dmcneil: i suggest that there is a support to add a binary indicator
... and then the specific dialect may be decided later

cygri: it is the same problem, the author may not know that either

Ashok: eric, can you live with what richard recommended

<EFranconi> i'm trying

ericP: his recommendation is not to have a flag or an attribute

<Souri> if we include it, we could make it optional, more like a comment, just informative "works on: Oracle DB 11.2.*"

ericP: if we had a flag, i was interested for an attribute

<dmcneil> +q

ericP: if we do not have flag, I can live with that

dmcneil: we say in the spec that we will use sql 2008
... but richard says, effectively, that the user can put anything there

cygri: we define r2rml/sql2008

<MacTed> was this flag to be mandatory? it seems a strong value-add if known/included, so having it and making it optional seems a win...

cygri: if someone wants to use r2rml with sql lite
... then they would be using r2rml/sql lite
... it is fairly clear what the behaviour would be
... but we do not necessarily say that
... and we can add something like 'you know what you would be doing...'

dmcneil: in our implementation we rely on that flag
... i can definitely say it is useful for us

MacTed: it seems problematic to make mandatory, but if the information there it may be useful
... so if it is optional, that is fine

cygri: i like to have this thing as an optional
... i am not particularly opposed to it
... what is the process to get the concrete identifiers for the dialects
... would we define those

mhausenblas: i would interpret eric's action to draft this with the optional attribute
... eric, is that fine?

ericP: I can do that, i want to do one for sql2008
... and the community can define to

mhausenblas: we can then resolve it next week

cygri: i am o.k. with that way of making progress
... what is the chance that there will be some sort of a standardization of these values
... to make things interoperable
... eric just said that he would propose one value

<Souri> I'd say cardinality should be 0-n (works for this, works for that, works for that too , ...)

MacTed: once one is define, the rest of use can throw in for others

mhausenblas: let eric start the process and let the others propose theirs

cygri: i am not sure how to address every dialect
... anyway, we can talk about that when we have something on the table

<Zakim> ericP, you wanted to say we define 1

<ericP> _:sMapping r:language <http://foo.example/tupleRelationalCaluculs> ; r:SQLQuery "Contacts : {company} | Contacts.name='Sue'" .

<mhausenblas> PROPOSAL: re ISSUE-22 the WG decides to have an OPTIONAL flag. this does not yet resolve ISSUE-22

+1

Souri: is that mean that it is compatible to XXX, yes, or no?

<dmcneil> +q

Ashok: the idea is that you have an attribute and you can use that as you wish
... the spec does not spell out the details

dmcneil: the issue I see we say sql queries for sql2008
... if you go to an r2rml process you expect that to run
... we would then water down
... it is implicit in there that there is vendor specific sql and do not use that

<ericP> ivan: i read this as "if i use something other than 2008, then i might make it explicit with this flag"

cygri: I think it will be common to have implementations that will use a single or a small number of sql engines

<dmcneil> PROPOSAL: re ISSUE-22 the WG decides to have an OPTIONAL flag. this does not yet resolve ISSUE-22, if the flag is not used then all SQL is expected to be SQL-2008 compliant

cygri: these implementations will accept queries in a particular dialect
... i may have sql lite
... that would make use of the flag
... i am opposed to have a flag that does not add any value
... flags the individual queries is comletely wrong that do not address a specific scenario

<Souri> +1 to Richard: we should make the common case less verbose (if SQL2008 => say it, if silent => it is NOT)

mhausenblas: one side, like dmcneil , who says that it is useful, another implementor says it does not really make sense

cygri: i am not disputing that it can be useful
... i expect that it will be common to write the queries that are not sql 2008
... and they do not want to flag

<Souri> if SQL2008 compatible: say it (if unspecified: may not be SQL2008 compatible)

cygri: what I said is that we do not define that in our spec
... we would say that the flag is of a small scope
... what the proposal say that we define portability
... i am not o.k. having to flag for non sql all the time

<LeeF> I have no idea when I'm writing compliant SQL or not :)

cygri: it would be stupid to ask for this flag

<MacTed> flag is optional -- if it's present, it gives information about the SQL conformance/dialect. if it's not present, the information is unknown.

<MacTed> attribute likewise -- if it's present, it gives more detailed information than a flag.

Souri: my comment is that I agree with richard, when I write a sql I am not really sure what version it is
... i would like to get the common case
... if I do not know what it is compatible is,
... but if I know, then we have the option
... if i do not say anything, then it may or it may not be compatible

<ericP> souri, a la { _:sMapping r:language _:dunno } ?

Souri: if I want to say then I can choose to use the attribute

<cygri> PROPOSAL: if attribute is present, we assume it's a specific dialect. if attribute is absent, we do not know what dialect it is. we only specify the behaviour of R2RML where the attribute says it's SQL-2008.

<MacTed> +1

<mhausenblas> PROPOSAL: Re ISSUE-22 the WG decides to introduce an OPTIONAL attribute. if attribute is present, we assume it's a specific dialect. if attribute is absent, we do not know what dialect it is. we only specify the behaviour of R2RML where the attribute says it's SQL-2008.

<cygri> +0.75

+1

<MacTed> it rounds up

<Souri> +0.75

<boris> +1

<betehess> +1

RESOLUTION: Re ISSUE-22 the WG decides to introduce an OPTIONAL attribute. if attribute is present, we assume it's a specific dialect. if attribute is absent, we do not know what dialect it is. we only specify the behaviour of R2RML where the attribute says it's SQL-2008.

issue 42

issue-42?

<trackbot> ISSUE-42 -- How is the direct mapping suppose to handle NULL values wrt Information Preserving -- open

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/42

mhausenblas: as you all know we have a deadline in september
... given that the group has spent about a month on issues 41 and 42
... i am inclined to resolve that asap
... i doubt that another month or even another year will solve this
... my position that it is not solved yet, we cannot crack it
... i would say we have two proposals on the table
... enrico's proposal is that the dm does not apply if there is a null value in the database

<Zakim> ericP, you wanted to make a no change proposal

mhausenblas: my proposal is to reflect the schema in the proposal and omit triples with null

<EFranconi> in the latter case to omit triples with nulls, we shoudl add the note: add a note saying that in this case it is not known how to relate this interpretation to the semantics of the NULL of the rdb it comes from

<betehess> ericP, there *would* be a change: adding the table information, which isn't there yet

mhausenblas: if people agree with the september deadline, we resolve it today

juansequeda: the current spec does not generate triples and the only thing to be added is the table information
... and add some explanation to the text

ericP: the depth you can model schema is arbitrary, one can add owl statements
... i can go deep into that
... or shallowest i can do, is to assume that the person issuing the query knows the schema

<juansequeda> PROPOSAL: Add to the Direct Mapping spec the generating of the schema triples and also add a note stating that "it is not known how to relate this interpretation to the semantics of the NULL of the rdb it comes from"

ericP: the range is no change at all to deep curation of the data model in the relation model

<Souri> Schema generation (based on table and constraint definitions), non-generation of triples corr to NULL cells in table, and a note sounds good to me!

Ashok: there is one other possible action, that EFranconi proposed with all of its details

EFranconi: i do not disagree with the proposal, but it is a message of failure of this group
... what you say that the mapping does not say anything about the semantics

<dmcneil> +q

EFranconi: a serious implementer would not take this seriously, it is a bad message

dmcneil: i agree with the conclusion of enrico, can we try to soften the message with 'unknown'

mhausenblas: i would suggest that we ask EFranconi to write a non rec track note that explains the problem space, and we could point to this non rec track note

<ericP> +1 to asking enrico to writing a note about the semantics of NULL

juansequeda: i have the suspicion that we can get the issue solved offline with EFranconi

<Seema> I agree that the note is too strong.

juansequeda: let us have this proposal as now
... i am sure that the next week with enrico we can solve this problem

<mhausenblas> PROPOSAL: Re ISSUE-42, the WG decides add to the Direct Mapping spec the generating of the schema triples and also add a note stating that "it is not known how to relate this interpretation to the semantics of the NULL of the rdb it comes from"

EFranconi: i do not think we can find the solution within a week, but i believe we can have a note that formulates the problem

mhausenblas: i hope that EFranconi would lead the effort to write this note

<Souri> +1 to the proposal (with a non-Rec track note)

<Zakim> ericP, you wanted to ask enrico for a concrete screw case with real-world example data

mhausenblas: it must be a note to not screw up planning

ericP: reading the issue of EFranconi and I would like to see a real world example data where our assumption of using non assertions breaks some intuitive queries people would have
... ultimately we will need this

Ashok: that is a reasonable thing to have in a note

<Souri> +1

cygri: I try to capture this discussion better

<cygri> PROPOSAL: Re ISSUE-42, the WG decides add to the Direct Mapping spec the generating of rdfs:range (and possibly other) schema triples and also add a note to the Direct Mapping spec saying: "The DM does not generate triples for NULL values. For a detailed discussion of this issue, see the following document" which points to a separate working group note

cygri: ...

<dmcneil> +1

<ericP> +1

<Souri> +1

<Seema> +1 to Richard's wording

<betehess> +1

<MacTed> +1

+1

<EFranconi> abstain :-)

<privera> +1

<boris> +1

<juansequeda> +1

<LeeF> +1

RESOLUTION: Re ISSUE-42, the WG decides add to the Direct Mapping spec the generating of rdfs:range (and possibly other) schema triples and also add a note to the Direct Mapping spec saying: "The DM does not generate triples for NULL values. For a detailed discussion of this issue, see the following document" which points to a separate working group note

<ericP> for the record, i heard a +1 from ashok

<ericP> (verbal)

<Ashok> +1

mhausenblas: who would accept to take a lead this

EFranconi: for family reasons I cannot take a leading role
... i am happy to contribute

juansequeda: i am talking on behalf of marcello, we can take the lead on that

action on juansequeda ask marcello whether he is happy be co-editor of the NULL Note

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

<juansequeda> jsequeda

<scribe> ACTION: member to juansequeda ask marcello whether he is happy be co-editor of the NULL Note [recorded in http://www.w3.org/2011/06/14-rdb2rdf-minutes.html#action01]

<trackbot> Sorry, bad ACTION syntax

action juansqueda ask marcello whether he is happy be co-editor of the NULL Note

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

<mhausenblas> ACTION: ivan to learn Zakimish [recorded in http://www.w3.org/2011/06/14-rdb2rdf-minutes.html#action02]

<trackbot> Sorry, amibiguous username (more than one match) - ivan

<trackbot> Try using a different identifier, such as family name or username (eg. ivan, imikhail)

<LeeF> ACTION: juan to ask marcello whether he is happy to be co-editor of the NULL Note [recorded in http://www.w3.org/2011/06/14-rdb2rdf-minutes.html#action03]

<trackbot> Created ACTION-134 - Ask marcello whether he is happy to be co-editor of the NULL Note [on Juan Sequeda - due 2011-06-21].

<ericP> actioncolon...

<mhausenblas> [adjourned]

<mhausenblas> trackbot, end telecon

Summary of Action Items

[NEW] ACTION: ivan to learn Zakimish [recorded in http://www.w3.org/2011/06/14-rdb2rdf-minutes.html#action02]
[NEW] ACTION: juan to ask marcello whether he is happy to be co-editor of the NULL Note [recorded in http://www.w3.org/2011/06/14-rdb2rdf-minutes.html#action03]
[NEW] ACTION: member to juansequeda ask marcello whether he is happy be co-editor of the NULL Note [recorded in http://www.w3.org/2011/06/14-rdb2rdf-minutes.html#action01]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.136 (CVS log)
$Date: 2011/06/14 17:24:10 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.136  of Date: 2011/05/12 12:01:43  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/represntation/representation/
Succeeded: s/distinquish/distinguish/
Succeeded: s/shema/schema/
Found ScribeNick: ivan
Found Scribe: ivan
Inferring ScribeNick: ivan
Default Present: +3539149aaaa, mhausenblas, Ivan, cygri_, +49.153.6.aabb, boris, privera, +1.781.273.aacc, MacTed, Ashok_Malhotra, +1.314.394.aadd, dmcneil, nunolopes, EricP, +1.603.897.aaee, Alexandre, +1.603.897.aaff, Seema, Souri, +575737aagg, juansequeda, LeeF, EFranconi
Present: +3539149aaaa mhausenblas Ivan cygri_ +49.153.6.aabb boris privera +1.781.273.aacc MacTed Ashok_Malhotra +1.314.394.aadd dmcneil nunolopes EricP +1.603.897.aaee Alexandre +1.603.897.aaff Seema Souri +575737aagg juansequeda LeeF EFranconi Michael Boris Percy Ted Ashok David Juan Lee Enrico
Regrets: Marcelo
Agenda: http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2011Jun/0051.html
Found Date: 14 Jun 2011
Guessing minutes URL: http://www.w3.org/2011/06/14-rdb2rdf-minutes.html
People with action items: ivan juan member

[End of scribe.perl diagnostic output]