Re: GRDDL Test Documents - moving all the tests, including approved tests?

Chimezie Ogbuji wrote:
> pass on what I had done so far to a webmaster to make sense of.  In 
> particular, the grddl-wg/td directory is a hodge-podge of alot of 
> content, only a fraction of which has been 'approved' and referenced in 
> the grddl-tests.html, I have so far. 

A very familiar problem - I had exactly that problem with the OWL Test 
Cases - the zip solution was the one I went for. I believe the rationale 
was as follows:

a) at some level the directory structure on the server is an artifact of 
the server, not the web. From the point of view of the web, we jkust 
have a collection of test resources with URIs. The fact that they all 
start with the same prefix is opaque (Web architecture).

b) the collection of test resources with URIs starting 
http://www.w3.org/2001/sw/grddl-wg/td/ is a mess.

c) if you want to access that mess, currently you can using the URI
http://www.w3.org/2001/sw/grddl-wg/td/

d) we can change
http://www.w3.org/2001/sw/grddl-wg/td/
(by editing Overview.html)
to hide the mess.

e) we can list a coherent set of tests - e.g. the ones for the test 
document by listing their URIs.

======

There are two routes forward that work for me:
1) the zip route - tried and tested with OWL and RDF
2) create a shiny new test directory, and migrate the approved tests there.

I think what Dan said about requiring a WG vote on every test that is 
moved is both procedurally incorrect, and not a big obstacle.
The test editor, and others at WG discretion, are empowered to make 
editorial changes to the tests. In OWL this was explicit see:
http://www.w3.org/TR/owl-test/#change
[[
The editors may make editorial changes to approved and proposed tests. 
This includes:
...
Moving tests to conform with document naming conventions.
]]
Obviously a log of any such changes to approved tests should be sent to 
the WG list.

The whole of
http://www.w3.org/TR/owl-test/#testProcess
is a useful an explicit summary of the process for test creating 
approval and modification.

What doesn't work for me is to have a set of tests with every copy of 
the test document. This is mainly to do with the complexities of 
relative and absolute URIs, particularly given that some of the test 
documents are namespace docs, and these should be referenced using 
absolute URIs.

Jeremy

Received on Wednesday, 21 March 2007 11:18:58 UTC