Groundrules for processing EARL?

I'm a little disturbed by what Jim showed us today with the merged
EARL reports, preserving the root node of the first report and
appending the second.  In terms of the picture, we saw Page Valet
as assertor, and had to look quite hard to see that some of the
assertions came not from it but from Nadia's tool.  Presumably
the reverse could equally well have been the case?

I'm open to persuasion, but as of now I'd feel more comfortable
with a scheme under which we could be assured of an equivalence:
  A + B === B + A
(excuse my improvised ASCII notation)
Under RDF this is trivial, but if A or B defines the root node of
a merge, we seem to have lost it.

Would it, for instance, make sense to define a merged report as a
setwise union of assertions?  Is this sufficient to define a
baseline (called it a Level 0) merge.  We can then consider
additional formalisms for a smarter merge that deals intelligently
with relationships between different assertions (is-equivalent,
modifies-confidence, supersedes, or whatever is needed).

Or have I just latched on to something that was only ever an
artifact of the RDF validator?

-- 
Nick Kew

Site Valet - the mark of Quality on the Web.
<URL:http://valet.webthing.com/>

Received on Wednesday, 20 February 2002 14:40:16 UTC