Re: EARL Semantics and Queryability [was: Re: EARL-producing testing tool]

Sean, I completely disagree on this question. I think that what should happen
if there are two reports made which conflict in their details is precisely
what you have outlined below. The resource (the thing being tested) hasn't
changed what it is, and nor has the EARL schema (probably). Some details may
have changed, but most URIs do not have any guarantee that they remain
byte-identical forever.

It is a question for the client processor what to do about conflicting
statements - a generating tool just produces those statements.

If a web page is not accessible on tuesday, and is accessible on thursday,
then it is up to a client to interpret the results and decide that if the
same person makes confliting claims, the best resolution is to accept which
set of claims has the latest date attached. (This is not new. It is like the
CSS cascade, and like principles of applying law that have been around for
hundreds of years). Alternatively, a client can make the decisions on the
author, if there is more than one and they differ. Or ask the user to tell it
how to decide in this case. Or point out that there are differing opinions.

None of these things will happen if you invent a new namespace for the thing
being tested each time you run a test - it is important that it has the same
URI if it is the same thing.

(otherwise we end up in the urn rathole - urns are a great idea, but the only
way to find them is to define a urn->URI mapping, at which point it makes
more sense to have some rdf using uris and be done with it...)

cheers

Chaals



On Fri, 15 Feb 2002, Sean B. Palmer wrote:

  > Hmm... I've changed it to produce that for now, but I'm
  > not sure I understand what this new and improved URI is
  > supposed to represent.  Does this mean that I have to make
  > my own namespace so that this can be translated into
  > something meaningful later?

  The main thing we're after is consistent information that can be
  queried useful in future. The more consistent your information, the
  more useful it is. Imagine the following report:-

  <http://www.w3.org/blargh> rdf:type earl:WebContent;
     earl:date "2001-10-15" .
  <http://www.w3.org/blargh> earl:fails :MyTest.

  and then a couple of days later, another report:-

  <http://www.w3.org/blargh> rdf:type earl:WebContent;
     earl:date "2001-10-17" .
  <http://www.w3.org/blargh> earl:passes :MyTest.

  Now, how do you expect to query it to check whether the page has been
  fixed, given that the URIs representing the resources are the same?
  It's broken. The rule is: one URI, one resource. [People get votes,
  URIs get resources!]

  So you're putting the "thing" in your own namespace for a few reasons:
  to separate it from other "things"; so that you can add data in the
  future; so that your data will be consistent; so that people can query
  it in the future; so that if you abandon or break it, you lessen the
  impact.

Received on Friday, 15 February 2002 21:04:48 UTC