W3C

- DRAFT -

RDB2RDF

13 Dec 2011

See also: IRC log

Attendees

Present
Ashok, David, Juan, Nuno, Richard, Souri, Ted, Seema, Eric
Regrets
Michael, Ivan, Joerg, Boris
Chair
Ashok
Scribe
dmcneil

Contents


i can scribe

<Ashok> scribenick: dmcneil

Accept the minutes of last meeting http://www.w3.org/2011/11/29-RDB2RDF-minutes.html

Accept minutes of Dec 6 http://www.w3.org/2011/12/06-RDB2RDF-minutes.html

<MacTed> Dec 6 was cancelled

<Ashok> Sorry, minutes for Nov 29 ...

RESOLUTION: Accept the minutes of last meeting http://www.w3.org/2011/11/29-RDB2RDF-minutes.html

david requested that we add an agenda item to talk about process

w3c process discussion

w3c process discussion

<Ashok> W3C Process document: http://www.w3.org/2005/10/Process-20051014/tr.html

per eric the process has evolved a bit since the link above

<cygri> see also ivan's mail: http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2011Nov/0070.html

we are currently working on last call comments

next stage is to go to a candidate recommendation where we ask if people have implemented the spec

we have not started this yet

it requires 1) defining test cases and 2) talk about what implementors need to report re: test cases

once we have at least two implementation then we can request that it become a recommendation

this would be version 1.0 of the new specs

possibly we would go on to define 1.1, etc.

it is quite reasonable to omit features from 1.0 in the interest of getting this to a recomendation and allow us to think about features for 1.1

anything to add eric?

eric: there is also a PR, proposed recommendation

there will also be implementation reports from implementors

probably we would wait for more than two implementations

working group would track what comments we are not accepting

once you exit CR, it is a short path to "rec"

we also have to tell the world how to run our tests

cygri: does the test suite need to be finalized to go to CR?

ericP: no, we get to define the details of what the exit criteria are

ashok: how formal is the process?

ericP: the level of formality varies, it is function of the test suite

+q

cygri: so there is the CR stage and then the PR, right?

ericP: CR to PR is hard, have to prove implementations

PR to rec is typically easier

cygri: question is: how to get to CR

we know how we want to do the test suite and it is starting to take shape, collecting more test cases

is that sufficient to get to CR, or is more needed?

ericP: technically we could probably go to CR immediately, but it is better to have a story more straight before that

because going to CR generates a buzz of activity from the community

we also need to define what interoperablity means

cygri: we also need to address all last call comments to get to CR

david: where should I look to see changes since last call?

ericP: at each stage there will be a new doc produced and editors provide a diff

<cygri> Editor's Draft: http://www.w3.org/2001/sw/rdb2rdf/r2rml/

cygri: end of editors draft list CSV commits with comments

+q

<cygri> http://www.w3.org/2001/sw/rdb2rdf/wiki/Last_Call_Changes_to_R2RML

i would hope that all my changes were in response to last call comments, well reasoned, etc

cygri: that last link is a list of relevant changes, things that would affect implementors working on the spec

those resources for R2RML should give a pretty good picture of what has changed,

plus there is the issues list where resolutions should be recorded

<ericP> dmcneil: i'd like a text diff; perhaps i have to create that myself

<ericP> ... is the editor's draft completely at the whim of the editors?

<ericP> cygri: i believe my edits respond to WG decisions

ashok: the editors should only write things agreed by WG

ericP: technically the W3C obligation is only at publication time

otherwise it is up to the team to agree

<ericP> CVS "Blame"

cygri seemdsto agree that since we are in last call now, that all changes should be linked to working group / last call comments (?)

ashok: how do we account for each line in the spec, that is what WG decision led to this text?

cygri: could dig up the CVS commits

<cygri> ACTION: cygri to email the group regarding cvs history and diffs [recorded in http://www.w3.org/2011/12/13-RDB2RDF-minutes.html#action01]

<trackbot> Created ACTION-178 - Email the group regarding cvs history and diffs [on Richard Cyganiak - due 2011-12-20].

ashok: the issue is most relevant for handling formal objections

<cygri> souri, it should be: cvs diff -r 1.162 -r 1.165 Overview.html

david: how can i see a diff between last call and now

cygri: csv committers can use cvs diff

there is also a way to generate html diffs

not as useful after a large number of changes

will make a diff of the entire thing to see what it looks like, but probably not useful

<cygri> http://www.w3.org/2001/sw/rdb2rdf/track/products/2

he can also produce specific diffs that we think are interesting

david: are there more changes coming

cygri: that link to remaining issues... each is expected to produce another change to the doc

ashok: will there be one commit for each?

cygri: generally, yes, but not always

ashok: can take two versions and understand the changes in light of the working group changes

david are you good with that?

david: sounds like i will need to read two version side-by-side

ISSUE-75

<Ashok> http://www.w3.org/2001/sw/rdb2rdf/track/issues/75

ashok: this is about tableName syntactic sugar

<Souri> Whenever editors commit, they can email a summary of the diffs produced by the "cvs diff " command for others to see.

david had a proposal on this

<Ashok> [ PROPOSAL: Resolve ISSUE-75 by removing the tableName syntactic sugar and simplifying the R2RML schema documentation and property table to reflect this. ]

any discussion?

<juansequeda> +q

is this acceptable to the group? can we close ISSUE-75 with this proposal?

<ericP> HTML diff of LC to editor's draft

<Souri> +1 to removing this sugar

juan: is this something we expect to come back when users start using it?

ashok: who can tell?

juan: well richard has experience with it, but i am fine with removing it

cygri: ok with dropping it because it is a very minor issue, just one little thing for each triple map, not worth a big fight

ericP - yes!

<Souri> +q

ashok: any objections?

souri: so this means we will not have an rr:tableName hanging from triple maps directly, right?

ashok: no objections heard

RESOLUTION: Resolve ISSUE-75 by removing the tableName syntactic sugar and simplifying the R2RML schema documentation and property table to reflect this.

ashok: three others on the agenda, 72, 57, 56 and 13 minutes

not worth starting 72, because it will be long/difficult

others are 57 and 68

<Souri> URL please

ISSUE-68: Multiple PredicateMaps in a PredicateObjectMap http://www.w3.org/2001/sw/rdb2rdf/track/issues/68

souri: if we do this special case, why not others, we don't need this

<Souri> P{+}O, PO{+}, ...

cygri: with syntactic sugar, it is always a tradeoff and always subjective

but there will be some people who are quite happy to have it

there are many things that are effectively sugar, e.g. templates, etc.

we could drop this wihtout expressivity, but it would make mapping harder

the question is: is the WG convinced that the sugar is not too confusing or complicated and provided enough value, always a judgement call

cygri: this feature is a poor man's RDF schema inferencing

e.g. employee class from db

employee is a sub-class of foaf:person

map that to rdf and do rdf schema inferencing to pick up foaf:agent

without rdf schema inference engine as part of r2rml implementation

users can just specify all the types in the mapping file

with class you can do this, because multiple classes are allowed per triplesmap

but the same thing holds for sub-properties

<Souri> you can have a single object map URI, and then use multiple POmaps each referring to the same object map URI

it would be convenient to be able to list this like we can list multiple classes

ashok: what about souri's idea to also allow multiple objects?

cygri: that is not needed for rdfs inferencing

ashok: but as a question of symmetry?

cygri: symmetry not very important

souri: it is not as hard as you think to do this without the sugar

see souri's comment above

the main pattern of s, p, o... if we allow s, p+, o... why not s,p,o+

and once we add a shortcut, we can never take it away

this is version 1.0, we are not losing any expressivity, don't see the need for a shortcut at this point, in 1.0

cygri: turtle is a representation format, not a mapping language so the analogy does not apply

re: how complicated it is without sugar:

in D2RQ people use a mapping generator that produces a skeleton that is customized

really nice to go in there and just change the autogenerated property name

and just add another property name

many users don't understand D2RQ mapping language, but they do know how to change properties in a generated mapping

so it is quite easy to replace one name with a few names

but creating new resources and wiring them up is much harder and requires an understanding of the language

and of turtle

which is a much higher barrier to creating mapping

graphical editors will not exist overnight

also, implementing this feature is not hard, can just explode it out in the graph, it is not a big deal

we did this in D2RQ

souri: regarding the analogy to turtle

there are other uses and other perspectives

users spend much more time debugging their code, not the mapping

<MacTed> different kinds of "users"

<juansequeda> Let's leave any type of syntactic sugar and short cuts for R2RML 1.1

<MacTed> exactly -- implementors vs users of implementations

cygri: not worried about spec author's time, not worried about implementor's time so much, most worried about users time

if implementor can do some extra work to save a lot of complexity for users, then that is a good thing

<Souri> why not SPO{+} then? I do not find this "that" beneficial!

regarding predicates vs objects, the symmetry is not important

<Souri> I am not arguing for symmetry

<Souri> SPO{+} is important and pretty common

<MacTed> not just saving time -- also removing potentially large barrier to entry/usage.

<MacTed> I think the SPO{+} is also valuable, but not because it's symmetric

ashok: we are out of time

we can't resolve this today

juan suggested holding off on sugar until 1.1

we have this issue plus ISSUE-72 and ISSUE-57

we have to resolve these... please think about what you can do and what compromises can be made

<Souri> for ISSUE-72 please see URL: http://www.w3.org/2001/sw/rdb2rdf/wiki/Translating_Database_Values_to_RDF_Terms

<Ashok> Sorry, I introduced the word "symmetry" --- perhaps that was inappropriate

Summary of Action Items

[NEW] ACTION: cygri to email the group regarding cvs history and diffs [recorded in http://www.w3.org/2011/12/13-RDB2RDF-minutes.html#action01]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.136 (CVS log)
$Date: 2011/12/13 18:07:01 $

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/effecticely/effectively/
Succeeded: s/there/their/
Found ScribeNick: dmcneil
Inferring Scribes: dmcneil
Present: Ashok David Juan Nuno Richard Souri Ted Seema Eric
Regrets: Michael Ivan Joerg Boris
Got date from IRC log name: 13 Dec 2011
Guessing minutes URL: http://www.w3.org/2011/12/13-RDB2RDF-minutes.html
People with action items: cygri

WARNING: Input appears to use implicit continuation lines.
You may need the "-implicitContinuations" option.


[End of scribe.perl diagnostic output]