Decide that SPARQL tests in rdf-tests will be for RDF 1.1.

[afs]

This is an issue to ask the community what it wants.

Do we, the rdf-tests community, want to maintain the SPARQL tests as RDF 1.1?
This includes PRs #76 and #83 (these are not necessarily all the changes possible).

Please "vote" by adding up (yes, RDF 1.1) or down (no, keep as an ad hoc mixture).

[afs]

Separately, there are other related issues such as the escaping rules which the strict definition in the spec and common practice differ. #67 (comment)

[Tpt]

My opinion: this repository should maintain tests for the latest specifications (so SPARQL 1.1 + RDF 1.1 and soon 1.2) but make clear it's not the testsuite from SPARQL 1.0 anymore. We should rename "data-r2" to something else to make the distinction clear.

Edit: it might be also relevant to remove the ol reports and rename the main directory for SPARQL sparql to make clear it targets the latest iteration of SPARQL and not SPARQL 1.1 when the REC was published.

[TallTed]

[@afs] Decide that SPARQL tests in rdf-tests will be for RDF 1.1.

I think you must (or should) have meant Decide whether tests in rdf-tests will be for SPARQL 1.{0,1,2} and/or RDF 1.{0,1,2}.

(I definitely think that thinking only about "only RDF 1.1" vs "an ad hoc mixture" is too limited a list of possibilities, and the latter leaves too much to chance.)

My personal wish would be that rdf-tests support all permutations of all of these (to later add whatever the future brings!), whether by maintaining distinct test sets (perhaps as forks) within one (or more) repo(s), or by maintaining one test set that is executed differently (e.g., with different flags or other option settings) to address different combinations of RDF & SPARQL versions, or otherwise.

[@Tpt] remove the ol reports and rename the main directory for SPARQL

I'm nothing but trouble, but I'd strongly prefer it if we could rename the reports and/or directories (e.g., to sparql-01-00, because we don't know how many major or minor versions we'll have over time, tho gosh I'm sure we all hope neither will exceed two digits!), and do relevant search-and-replaces therein/thereon, to keep their out-/inter-links working and to make any references to un-versioned SPARQL or RDF appropriately version-specific.

I would also like it if we could have the un-versioned resources redirect, optimally with options including all versions (i.e., 300 Multiple Choices "the user-agent can present the user with the choices and the user can directly choose such"), sub-optimally simply to the latest/current version (i.e., 303 See Other).

Why? Because those detailed reports may remain relevant for people trying to use newer tools which no longer support older RDF data and/or SPARQL queries, who need to refer back to the older tools for use with archival data, or reproducing query results for audits, among other reasons.

[afs]

rename

See #84

[TallTed]

#84 helps. It might be good to add #84 to the references in the initial comment on this issue.

[TallTed]

I still suggest that the title of this issue would be better changed from Decide that to Decide whether.

[afs]

This issue is for a vote on the principle.

Let's keep this clear cut so when people vote it is on the principle here.

[TallTed]

I believe I've clearly stated a third principle (neither ad-hoc nor solely RDF 1.1), which should at least lead to another straw poll...

[gkellogg]

One issue to consider is if any kind of maintenance is expected on older test suites, and at this point, SPARQL/RDF 1.1 are some 10 years old and work is currently moving to SPARQL/RDF 1.2. While keeping a named branch holding some previous state is always a good idea, people who need them probably expect to access through the GH Pages interface, so this argues for some form of simultaneous access.

Note that the JSON-LD Test Suite does maintain a specVersion property to allow for this type of thing, it has proven to be confusing to both developers and maintainers. But, I don't expect many 1.1 tests to actually be changed for 1.2, only new tests added. There may be some debate about this, but arguably, any changes other than for RDF-star and any other updates would be additive, so adding new tests to the current manifests (with appropriate tagging) might not be too bad.

I echo @TallTed's concern about maintaining access to historical implementation reports, so generally, I'd favor creating sub-directories containing the complete previous versions, at least back to SPARQL/RDF 1.1.

I do favor using this repository for ongoing test maintenance, but that is a decision to be taken by the RDF-star working group; we should be prepared for that. Having some -dev areas is also good to support active Community Group development. It's useful to have one place to go to find all the tests (JSON-LD and RDFa aside).

I generally support @afs's plan as stated in #83 and #88.

[afs]

detailed reports

They are copies of the original submissions plus any later ones (IIRC, maybe none).

They aren't up to date; they aren't reproducible, and never have been because it was not an objective at the time.

The WGs asked for self-reported coverage for the purposes of the REC transition.

So the implementation reports are not affected one way or the other.

The test have had commits since. The original work group material links:

• SPARQL 1.0 : https://www.w3.org/2001/sw/DataAccess/tests/r2
• SPARQL 1.1 : https://www.w3.org/2009/sparql/docs/tests/README.html

The SPARQL 1.0 have had some manifest fixes.
The SPARQL 1.1 tests have already (over several years) diverged in exact content (details in git).

[afs]

I've tagged the repository with sparql-mixed-rdf-version-tests so we can easily find the state as of today, before any of the possible #76, #83, #84.