EARL: Identification Problem [was: Re: Suggested EARL Changes]

> > 03:15:33 <sbp> * Remove testSubject
> > 03:15:40 <sbp> * Add reprOf
>
> I'm concerned this moves away from EARL as being
> general purpose evaluate/report on anything language,

Allow me to alleviate your anxiety. Firstly, the property is not being
changed, it is being renamed. "testSubject" didn't really have a fixed
meaning based upon Web architecture terms as set out in RFC 2396 and
2616, and so the "what is identified?" problem started to creep is.
Secondly, the threading that you're worried is being *introduced* to
EARL has been there since Aaron suggested to use a resource and a
timestamp. A resource is a thread; it has a number of entities
corresponsing to it over time. We expect pages to change, as they get
corrected, as new items come in, or whatever, but we expect the
mapping from the URI to the resource (the denotation, the semantics of
the identifier) to be consistent, otherwise the Web would not be very
useful.

Thirdly, this does not stop people from evaluating anything; I agree
that that is a very big strength of EARL, and will always be careful
to preserve it. This change simply represents a clearer definition of
semantics when evaluating WebContent. You can still evaluate tool and
user agents, although I am thinking about how to do it. Let me run
through the various methods.

   EARL Evaluation: x asserts { y passes z } .

Let us just concentrate on the structure of "y", the TestSubject. The
TestSubject is not some stand-in for what is being evaluated, it *is*
the thing being evaluated. There are plenty of discussions about what
constitutes a resource in HTTP space... TimBL says that HTTP resources
are necessarily generic documents. Roy Fielding believes that they can
be anything. Al Gilman often points out that HTTP resources can be
better modelled as service pools. Because we want to be as consistent
as we possibly can with these views, we'll follow the "be liberal in
what you accept, and conservative in what you do" axiom, and accept
anything. That simply means, we do not ask people to use URI
references for the TestSubject when it's of type WebContent... but I
digress.

The W3C homepage http://www.w3.org/ is the concept of the W3C
homepage... it is always the home of the W3C. Usually, I will want to
say that a particular representation of it passes a certain
checkpoint. The test subject in that case will be:-

   :W3CPage earl:reprOf <http://www.w3.org/>;
      earl:date "2001-12-09";
      dc:format "text/html" .

This is the standard use for EARL; standard in that it will be most
common. Note that because the domain of "earl:reprOf" is
"earl:WebContent", we do not really need to state that explicitly
(although it may help if we do).

Now, we may also want to rate the homepage as a great piece of work.
In that case, we are not talking about any one particular
serialization, but the concept of the homepage itself. Well, that
already has a URI... the test subject is simply:-

   <http://www.w3.org/> a earl:TestSubject .

I don't think that this will be a very common use of EARL, but it's
there in case anyone needs it. We don't want to restrict people at all
in what they can evaluate.

The next use for EARL will be to evaluate things which aren't
necessarily generic documents, but are tools, and so forth. Let's take
a UserAgent called "MyUA", based as MyUA.org (with apologies to anyone
who comes up with MyUA). We can use any identifier for the tool, in
absence of knowing whether the homepage for the tool identifies the
tool or the homepage. I think that usually it will be the homepage,
because otherwise the tool would have no homepage... this is a murky
area, but we can create a property that bulldozes over the argument by
saying "if you suport x POV, then y, if you support q POV, then p".

   :MyUA a earl:UserAgent;
      earl:homepage <http://myua.org/> .

Where "homepage" is a special property (that may need renaming) to say
that the subject either identifies the tool described by, or the tool
at, as appropriate. Either way, doesn't matter now; EARL is firmly in
control :-)

So this simply clarifies what is meant by the important identification
properties in EARL, so that people can be much clearer about what they
identify. I hope that it is consistent with the views of as many
people as I could possibly make it, and most notably the URI and HTTP
RFCs.

Cheers,

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://purl.org/net/swn#> .
:Sean :homepage <http://purl.org/net/sbp/> .

Received on Sunday, 9 December 2001 12:18:32 UTC