W3C

Results of Questionnaire How should SPARQL 1.1 resolve ISSUE-29, the handling of negation?

The results of this questionnaire are available to anybody.

This questionnaire was open from 2010-05-07 to 2010-05-17.

14 answers have been received.

Jump to results for question:

  1. Which resolution to ISSUE-29 does your organization prefer?

1. Which resolution to ISSUE-29 does your organization prefer?

LeeF's mailing list message of 2010-Apr-19 summarizes the most recent state of discussion of the options for resolving ISSUE-29. Full statement of question. This survey includes most of the options outlined in LeeF's mail of 2010-Apr-06.

Summary

ChoiceAll responders
Results
Include MINUS and NOT EXISTS as graph pattern operators. Also include NOT EXISTS and EXISTS as FILTERs. 2
Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs. 9
Include MINUS graph pattern operator only.
Include NOT EXISTS graph pattern operator only. Also include NOT EXISTS and EXISTS as FILTERs. 2
Include a NOT EXISTS graph pattern operator only. Define it only for the cases that have the same results for both approaches. Explicitly note that implementers may choose which style to implement for the divergent cases. 1

Details

Responder Which resolution to ISSUE-29 does your organization prefer?Rationale
Garlik (Steve Harris) Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs. While correct, we don't feel that Andy's observation around FILTERs being mobile is really a significant problem, but the understanding overhead to any solution with both MINUS and NOT EXISTS as graph pattern operatorrs is far too great.

Garlik objects to any outcome that has both MINUS and NOT EXISTS as GP operators. A single operator with the semantics that are compatible with both would be fine.
Rensselaer Polytechnic Institute (Gregory Williams) Include MINUS and NOT EXISTS as graph pattern operators. Also include NOT EXISTS and EXISTS as FILTERs.
[DUPLICATE] Cambridge Semantics (Lee Feigenbaum) Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs. This approach allows users to choose between two reasonable negation "styles". To me, the confusion from having two graph pattern operators for negation outweighs the utility of having a form of the NOT EXISTS negation semantics that does not "move" to the end of a group the way filters do.
Talis Group Ltd (Andy Seaborne) Include MINUS and NOT EXISTS as graph pattern operators. Also include NOT EXISTS and EXISTS as FILTERs. The lack of implementation experience of MINUS in SPARQL is a concern but given the different styles, Talis can agree to accepting both forms. The WG should not remove the common case with succinct syntax and so it should offer NOT EXISTS without requiring FILTER. The WG is already making MINUS syntax easier to use (by not requiring {} for the left-handside which is required for UNION in SPARQL 1.0).

Option 5 is not acceptable because MINUS can't naturally test for the present/absence of a ground triple making any compatibility rules complex.
University of Oxford (Birte Glimm) Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs. I don't think that having two graph pattern operators, which behave slightly different, is a good option. After working my way through the examples and the semantics, I think MINUS is a reasonable choice as a graph pattern operator. This also has the benefit that EXISTS/NOT EXISTS can be used as filters without having the same name for a graph pattern operator and a filter and I would like to include EXISTS/NOT EXISTS as filters.

Option 5 is my by far least preferred one.
Paula Gearon (Paula Gearon) Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs. MINUS is a single-word operator whose semantics seem clear. While it might be desirable to leave it there, we get greater functional coverage if we also provide NOT EXISTS in filters.
W3C/ERCIM (Ivan Herman) Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs.
Clark & Parsia LLC (Kendall Clark) Include NOT EXISTS graph pattern operator only. Also include NOT EXISTS and EXISTS as FILTERs.
Institut National de Recherche en Informatique et en Automatique (Inria) (Olivier Corby) Include a NOT EXISTS graph pattern operator only. Define it only for the cases that have the same results for both approaches. Explicitly note that implementers may choose which style to implement for the divergent cases. Simplicity.
Profium (Tommi Koivula) Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs.
Cleveland Clinic (Chimezie Ogbuji) Include NOT EXISTS graph pattern operator only. Also include NOT EXISTS and EXISTS as FILTERs. This is more in line with a stable model (answer set) semantics for RDF
Oracle Corporation (Matthew Perry) Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs. I think MINUS belongs as a graph pattern operator because it is a set operation. EXISTS / NOT EXISTS are really filtering operations, so they should appear in FILTERs rather than graph patterns.
INSIGHT - The Centre for Data Analytics (Axel Polleres) Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs. This is DERI's proefered solution (submitted by chairhat-off-axel after discussing with Alex).
IBM Corporation (David Charboneau) Include MINUS as a graph pattern operat. Also include NOT EXISTS and EXISTS as FILTERs.

More details on responses

  • Garlik: last responded on 8, May 2010 at 07:01 (UTC)
  • Rensselaer Polytechnic Institute: last responded on 9, May 2010 at 16:37 (UTC)
  • [DUPLICATE] Cambridge Semantics: last responded on 10, May 2010 at 03:24 (UTC)
  • Talis Group Ltd: last responded on 11, May 2010 at 10:35 (UTC)
  • University of Oxford: last responded on 11, May 2010 at 16:04 (UTC)
  • Paula Gearon: last responded on 11, May 2010 at 21:25 (UTC)
  • W3C/ERCIM: last responded on 13, May 2010 at 12:27 (UTC)
  • Clark & Parsia LLC: last responded on 15, May 2010 at 13:07 (UTC)
  • Institut National de Recherche en Informatique et en Automatique (Inria): last responded on 17, May 2010 at 06:20 (UTC)
  • Profium: last responded on 17, May 2010 at 08:57 (UTC)
  • Cleveland Clinic: last responded on 17, May 2010 at 13:36 (UTC)
  • Oracle Corporation: last responded on 17, May 2010 at 13:54 (UTC)
  • INSIGHT - The Centre for Data Analytics: last responded on 17, May 2010 at 18:05 (UTC)
  • IBM Corporation: last responded on 17, May 2010 at 19:06 (UTC)

Everybody has responded to this questionnaire.


Compact view of the results / list of email addresses of the responders

WBS home / Questionnaires / WG questionnaires / Answer this questionnaire