Re: Update

good morning;

> On 2015-10-11, at 23:00, Karima Rafes <karima.rafes@gmail.com> wrote:
> 
> Hello
> 
> Before to change the SPARQL test suite, there is a more urgent
> problem. I noted a lot of problems in the implementation of SPARQL
> protocol in the triplestores (the federated query with SPARQL is of
> the science fiction between editors of triplestores). Each editor has
> his own definition of SPARQL protocol.

it would certainly facilitate federation to include a mechanism, by which one declares an implementation’s participation in the form of an endpoint which is intended to be available as a remote federation location and to test the combinations.
if the tests were run by a service, as here proposed, the eventual result could be a cartesian map of respective interoperability.

> 
> […]
> 
> If I develop these new tests without participation of editors, no
> triplestores would pass these new tests. So I'll let you decide if
> it's useful to develop new tests for the protocol of SPARQL. If some
> editors(minimum 2) are OK to converge, I can propose implement quickly
> these tests via TFT [1].

in what sense would editorial authority be necessary?
even just a single test on the order of "SERVICE test 1” would be valuable
if run across implementations, if nothing else, it would draw out the current (low ?) watermark.

> 
> For info, the service http://sparqlscore.com/ uses the W3C's
> repository in github. If you update the test suite SPARQL in your
> repository, you will see the results after 24 hours on my service.

in the six months, since we attempted to use tft, github has evidently improved their http conformance and no longer serves turtle as octet stream content.
this much improves the chances that a service would be worthwhile.

best regards, from berlin,
---
james anderson | james@dydra.com | http://dydra.com