- From: Al Gilman <asgilman@iamdigex.net>
- Date: Mon, 10 Dec 2001 10:18:20 -0500
- To: "Jim Ley" <jim@jibbering.com>, <w3c-wai-er-ig@w3.org>
I believe the EARL proposal at the moment lacks a URI from which one can retrieve, verbatim, a copy of the representation instance that was evaluated in an EARL-reported test-exposure a.k.a. test-experience episode. It is all well and good to identify what was tested by URI (and related expressions). For precision one needs to be able to, as Jim suggests, re-test the old-date representation with the new tool if one wishes, after getting different outcomes with two test exposures which differ in both date and tool. This is exactly what I meant by trying to "localize the frontier in experienced outcomes." You want to go back and locally fill in a regular pattern of test exposures to see which difference made the difference. Is the pass/fail different because of the tool change, or because of the date change? Al At 07:08 AM 2001-12-10 , Jim Ley wrote: >"Sean B. Palmer" <sean@mysterylights.com> >> > Certainly, although I don't see how we, when we're >> > external to the resource, can know enough to state >> > that some resource is newer version of the old. >> >> The resource doesn't change, just the representations. > >That is a somewhat utopian view, but I take your point. :-) > >> > My point is [not?] that we're introducing anything new, >> > just that we're introducing the idea that testSubject/reprOf >> > is threaded [...] >> >> Well, the idea of a resource is threaded in a sense. That doesn't >> enable us to *necessarily* conclude that we should drop old >> evaluations, just that there is a thread of entities assocaited with >> that resource. > >There are evaluations of various states of that resource, there's no >definitive thread, just because some evaluation is dated later, it >doesn't have any intrinsic bearing on the former evaluation unless it's >an identical tool doing the evaluation. The meaning can all be infered, >and this includes threading issues. > >> To ground it in a scenario... the thing that started this was of the >> broken page that is fixed by an author and re-evaluated. If you put >> both of the evaluations into a database, you want to test if the >> representation of the resource has been fixed. You can do that by >> using the threading of a resource as an assumption. > >PageValet asserts <http://www.e-media.co.uk/>http://www.e-media.co.uk/ on 2001-12-10 fails valid >HTML >JimsValidator asserts <http://www.e-media.co.uk/>http://www.e-media.co.uk/ on 2001-12-11 passes >valid HTML > >If we make the threading assumption then we would conclude that it was >valid - rather than my validator is completely flawed (this would be a >more likely scenario with something less well defined than validity.) > >> There is nothing >> in EARL that says you must make that assumption, it is a function of >> the person controlling the database. > >Indeed, so we need to make no allowances for threading or a resource. If >we evaluate things, and describe their state (which we already do) then >all threading can be infered, but it requires knowledge outside of the >EARL about the resource, and the people making the assertions. > >The original scenario would seem to be better served by an assertion >about the previous assertion. > >theEvaluation: Sean asserts ( <http://www.e-media.co.uk/>http://www.e-media.co.uk/ fails validity ) >Jim asserts theEvaluation fixed > >So if we want to say something about the previous assertion, we say >something about the previous assertion, we don't try to impose structure >on the resource (which it may not have, or certainly won't externally >know what it is.) and then infer from that what we originally wanted to >say. > >Jim. >
Received on Monday, 10 December 2001 10:08:01 UTC