COPY/MOVE tests

I think we ran out of time on the telecon earlier this week before I got to the details of my issues with the copy-05 and move-05 tests, but I'd like to suggest that we drop them (or remove them from the manifest list, or simply don't approve them; whatever the process here is).

My understanding is that copy-05 and move-05 are attempting to test that an implementation follows the suggestions:

[COPY] "By default, the service SHOULD return failure if the input graph does not exist."
[MOVE] "By default, the service MAY return failure if the input graph does not exist."

since both tests attempt to use an input graph that doesn't exist in the dataset.

I believe Paul agreed that the difference in normative language here was a mistake. As far as I'm concerned, though, the only sensible solution is to change the SHOULD to a MAY. This follows from our support of both proper graph stores (for which a test of existence on a graph makes sense) and of quadstores (for which a non-existent graph can't be distinguished from an empty graph). Since the user may very well want to copy or move an empty graph, the existing "SHOULD" language means that in the quadstore case, the implementation may end up throwing an error on a perfectly natural request.

Because of the difference in how graph existence is treated between the two types of systems, I don't think that either copy-05 or move-05 makes sense to test.

thanks,
.greg

Received on Friday, 3 February 2012 18:46:22 UTC