Re: EARL 1.0: trying it with hypertext instead of reification

Hi Dan,

Many thinks for the comments; some notes inline.

>      <rdfs:Resource>
>       <rdf:type rdf:resource="[...]#WebContent"/>
>       <reprOf rdf:resource="http://example.org/net/bob/"/>
>       <date>2001-05-07</date>
>      </rdfs:Resouce>

or equally:-

   <WebContent>
     <reprOf rdf:resource="http://example.org/net/bob/"/>
     <date>2001-05-07</date>
   </WebContent>

The attribute values make for less lines, but I guess that the above
is neater, and fits in better width-wise. I'll modify the draft
specification accordingly.

> Now, we try taking out the reification by putting the
> quoted content into a separate document: [...]
>  <asserts rdf:resource="doc2.rdf"/>

Note that we actually discussed this approach fairly recently (these
are Wendy's notes, and IRC logs mushed together):-

[[[
sbp: could do: everything in this doc asserted by... in rdf, implicit
defn of what's going on in doc
[...]
sbp: loads of people reviewed one site. assertor elements. people
rating the page. alternative: that info for the doc itself. implied
assertor element for the doc. easier to parse and create, but less
specific. only one person can assert per doc. if wants loads of
people, ... earl dependent on context info
NK: Why .. only one person can assert per doc?
sbp: not even sure technically possible or sound to move in that
direction, but make it easier. the way things happen in rdf, you have
statements - ground facts. reified facts in the doc. e.g. this page
passes x x passes or fails guideline x says that y says... question
is: should we have quoting or not. if don't have quoting, can only
have one person saying something - the author that's what's impiled in
rdf benefit, don't use reification
[...]
sbp: don't think the benefit outweighs the advantage of reificaiton
CMN: no, you can have multiple people who each say "I think something"
about a single subject
NK: .. where "that" may not necessarily be very well-defined
NK: ('cos yer average surfer doean't speak our jargon)
[...]
JL: I think we need multiple "people" saying things within one Earl
document, I can't see how it would be useful otherwise
SBP: Yes, I agree with that
]]] - http://www.w3.org/WAI/ER/2001/12/03-minutes

In other words, the idea got flamed down pretty quickly. BTW,
"asserts" as the predicate above would have to mean "asserts the data
that the document parses to", cf. log:semantics in CWM. The problem
with this is that it's less natural to say things about the root
context, unless we invent a predicate in EARL which relates a document
to its formulae (er... well, we'd just use log:semantics).

The added problem is that you'd have to keep the documents separate...
it would be possible to devise rules to merge the documents such that
they end up like EARL in its current form, but if you're going to be
doing that, why not just stick with the current form?

Of course, I'm not the one that you have to convince... the people
implementing it are Jim, and Nick, et al. If they find this solution
more viable to implement than the current means of doing things, then
we should probably change; but the indication from when I raised this
is that the approach is not liked.

> Am I misunderstanding EARL? I moved testMode into doc1 as
> it wasn't clear what it was attached to.

testMode simply points to the test mode of an assertion (as you
suspect). The current ranges are Auto, Manual, and Heuristic, but it's
open for extensions. I feel that it belongs off of the assertion
itself, because "heuristic" is a bit of a wild-card.

> Note that bits of the RDF weren't reified ("scare quoted") in the
> first place: the original RDF file claims that there exists some
> webcontent with a date, reprOf property etc. What we've done here
> is in effect more conservative than the original RDF, since the
> date/reprOf etc content ascribed to the WebContent is written in our
> 2nd RDF file, which is where we've put the stuff that "Bob says...".
> In the original, the main RDF file makes these assertions without
> any qualification or quoting.

Interesting point; the only thing that is reified is indeed the
assertion itself, not the data pertaining to the assertion. The design
decision behind this was that these are usually non-controversial
facts - if a resource is text/xml than it is clearly text/xml - there
is no particular reason for anyone to lie, or any which in which an
inaccuracy could creep in. So (goes the logic) there is no particular
reason for wanting to reify these statements. Of course, that doesn't
stop people from reifying them if they want to... EARL just assumes
that people will want to tell the truth about self-evident things, and
perhaps lie about the more subjective things - the tests themselves.

Of course, this logic could be flawed, but I remember this coming up
somewhere on WAI ER before (can't find the reference...), and we
haven't changed it yet. Once again, this does not stop us from
changing it if the overall consensus in the group is that it should be
changed.

These are all very useful points, and I have no doubt that we'll be
discussing them on the list/calls. Many thanks for the input,

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

Received on Thursday, 13 December 2001 18:46:51 UTC