RE: Comment on SPARQL 1.1 Test Cases

Hello

As a small tip from a fellow implementer of this specification: I downloaded all test cases automatically using a web site downloader. (This one, but there are probably many others: http://download.cnet.com/WinWSD-WebSite-Downloader/3000-2377_4-10562531.html)

/Peter

From: Rob Vesse [mailto:rvesse@yarcdata.com]
Sent: den 18 juni 2012 22:39
To: public-rdf-dawg-comments@w3.org
Subject: Comment on SPARQL 1.1 Test Cases

Hi All

I have several comments on test cases.

Firstly there appears to be a discrepancy between the test cases for STRAFTER and STRBEFORE.  Updating my implementation to pass the newly introduced tests causes the old tests to no longer pass.  So either I have read the specification incorrectly (in which case it probably needs further examples or clarifications) or the old tests are still testing for the old behavior while the new tests are testing for the new behavior.

If the latter is the case either the old tests should be marked as Not Approved or the test cases updated so they match the current specification of those functions.

Secondly when will there be available a ZIP download of the complete test suite?  Currently I have to go and manually browse through the directories downloading the individual files necessary to obtain the new/updated tests which is error prone and time consuming.

Finally I understand that there is some discussion about skipping the CR implementation report gathering phase of the standards process because there are already various implementation reports available.  While this is true I believe this to be misguided given the significant changes to the specification around property paths and introduction of the VALUES clause.

Particularly in the area of property paths many existing implementation reports may be inaccurate (certainly dotNetRDF's is) because the WG changed the expected results of the existing tests rather than deprecating old tests and introducing new tests.  This means the current implementation report on file for dotNetRDF has a bunch of tests reported as passing (because they did under the old semantics and test cases) which in reality no longer pass now my test suite is updated to match the latest version of the WG test suite I.e. The current percentage compliance is actually higher than reality. I would like to see implementers given more time to actually implement the upcoming changes (which you have yet to formally publish) and report back comments on them as necessary rather than these changes be presented as fait accompli and risk formal objections to the spec delaying eventual standardization.

Rob