W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > January to March 2012

Re: summary on options for JP-4 Comment about the semantics of property paths

From: Andy Seaborne <andy.seaborne@epimorphics.com>
Date: Mon, 13 Feb 2012 14:46:21 +0000
Message-ID: <4F39223D.1060606@epimorphics.com>
To: public-rdf-dawg@w3.org


On 12/02/12 13:08, Axel Polleres wrote:
> Just for completeness and not to forget:
>
> Birte's mail [1] indicated that marking Property paths informative/non-normative could also be
> considered, also because of the things that remained unclear wrt. entailment regimes [2].
>
> I'd rather call this a "variant" of Options 1-3, since this seems to be orthogonal...
> We had discussed this briefly in the last Telco [3] but I have the feeling that the last
> word here has not yet been spoken. Opinion on marking PPs optional (i.e. informative)
> would be appreciated.

-1 : separate email

> It'd be great to collect more on that per email (thanks all for the useful discussion!)
> and we'will sum up in the upcoming Telco.
>
> Just to get together where I understand we are now (also regarding the "numbering" of options):
>
> Option 1... keep everything as it is in the grammar, and explain which DISTINCT path subqueries can be optimized
>              (which in my understanding was, explaining the there was a set of query transformations that are
>               equivalent to the possible semantics in the paper)

Worded like that, -1  Explaining DISTINCT can be used to control 
duplicates is not the same as providing an equivalence).

There are several problems but mainly I'm not clear as to the promises 
this is now making.

Suppose no such set of transformations exists - isn't that strong 
evidence the semantics are not an appropriate choice for a large class 
for implementations?

Suppose a close, but different, semantics is possible and more practical 
for implementation at scale?

Suppose such transformations do exist as do similar but slightly 
different ones.  Why choose one over the other to highlight?

> Option 2... Option 2 ... add DISTINCT around paths (which could be argued to be optimized, agreed?)

My preferred option is 2.1.

I recognize the time-cost of all option 2 schemes.

(Not sure what you mean by "optimized" here.)

> Option 3... leave things as they are, and pointing to a future WG regarding optimization.

Acceptable.

Add discussion material that property paths can be expensive on graph 
such as cliques.

>
> Variants of these Options could be marking PPs informative (we may call them Option1*-Option3*),

Another email.

	Andy

>
> Axel
>
> 1. http://lists.w3.org/Archives/Public/public-rdf-dawg/2012JanMar/0128.html
> 2. http://www.w3.org/TR/sparql11-entailment/#property-path
> 3. http://www.w3.org/2009/sparql/meeting/2012-02-07#line0084
Received on Monday, 13 February 2012 14:46:51 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:47 GMT