Agenda, 30 Jan @ 14:30 UTC

Hi everyone,

Most of the telecon tomorrow will center on the scope of blank node 
labels. Please try to come familiar with the issue so that we can discuss 
it and attempt to reach consensus on a decision.


0. Convene [1]RDF Data Access WG meeting of Tuesday, 30 January, 2007 
at 14:30:00 UTC
         + LeeF chairing
         + teleconference bridge: tel:+1.617.761.6200 
tel:+ tel:+44.117.370.6152 code:7333
         + on irc at: irc://
         + Scribe: PatH
         + Regrets: EliasT, Jeen
         + roll call
         + 16 Jan minutes[2] to approve
         + 23 Jan minutes[3] to approve
         + next meeting 6 Feb., SimonR to scribe.
         + agenda comments?

1. Review ACTION Items

These action are DONE:

ACTION: AndyS to reply to 
mentioning the possibility of banning the same bnode id from appearing in 
multiple BGPs in a query
  [ed: done by EricP in ]
ACTION: LeeF to look back through minutes and mailing list to determine if 
the group has made a past decision on blank node scope
  [ed: done in ]

Let's check on the status of the following actions:

ACTION: AndyS to reply to Bob M noting changes in examples in curent 
ACTION: EricP to run the yacker tool over and annotate the existing tests
ACTION: Jeen to mark approved tests as dawg:approved
ACTION: LeeF to remember that the wee, lost filter tests should be put 

2. Test suite

Jeen sends his regrets, so no work for the WG on tests this week. The 
latest from Jeen:

3. Blank node label scope

I'd like to try to make a decision on this issue this week. Here are the 
relevant mailing list messages:

+ EricP raised the issue in, 
noting that allowing the same blank node label to correspond to different 
blank nodes in a single query can be confusing. (Note that ?who in his 
example should be _:who.)

+ Eric and Andy explained Andy's proposal to prohibit the same blank node 
label from appearing in multiple BGPs in a query here:

+ Andy provided test cases for his proposal in
  [ed: I can guess which of these are +'ve and which -'ve, but I can't 
tell from the message itself.]

As the chair sees it, there are three proposals with various (not 
exhaustive) pros and cons. My apologies if this summary is not completely 
accurate, we'll discuss this all on the telecon:

1) Blank node labels are scoped to BGPs
  + pros: the previous Last Call design; extends well to DL.
  + cons: allows confusing queries as per EricP's example.

2) Blank node labels are scoped to the query
  + pros: blank nodes work just like (non-projectable) variables; editing 
a query (e.g. adding an OPTIONAL) can use existing blank node labels
  + cons: requires support in the algebra; does not extend well to 
disjunctive logics wishing to treat query blank nodes as pure existentials 
(e.g. DL)

3) Blank node labels are scoped to the BGPs and each label may only appear 
in a single BGP per query
  + pros: confusing queries are prohibited; extends well to DL
  + cons: invalidates some queries which are currently valid

4. Minimal test suite?

Simon suggested two weeks ago that:

I'm somewhat inclined to have a "designed" collection of tests that are a 
roughly minimal coverage of the features.  Those extra tests reduce the 
chance of a human ever actually reading them, which is highly desirable 
for correctness and understanding.

This was mostly discussed on IRC two weeks ago, and I promised an agenda 
slot to further the discussion last week. I'll keep this around until we 
have a chance to discuss it.

5. rq25 status

Thanks to Orri for an initial review of rq25:


Received on Monday, 29 January 2007 05:07:11 UTC