Re: statements about resources vs. representations

Your message just shows how shabby my exposition was, so rather than
pick over what you wrote, I think I'll start over, with the aim of
answering the questions you posed.

I'm going to switch from book to journal article, and attempt to
deconstruct as objectively as I can.

You define a class that I'll call Edition:

> ... a Platonic abstract book, the kind of thing that can have an ISBN
> number; not a physical copy of the book but the 'edition' that said physical
> copy would be a copy of. This, plausibly, might have its own URI assigned by
> the publisher, say, which we can take U to be.

You also say that something is a "webpage", and I will assume that by
"webpage" you really mean what you called "accessible" in your "In
Defense of Ambiguity"
(http://www.ibiblio.org/hhalpin/irw2006/presentations/HayesSlides.pdf
- I can't seem to access the paper itself online without paying).
(Love that last slide BTW.) So I'll also define this class:

Accessible = there is a URI that provides a causal pathway to the
thing, mediated by the Internet.

In passing I will note some other classes of other things we've talked
about here:

"resource" sensu RFC 2616 - "A network data object or service"
"information resource" sensu AWWW (AWWWIR) - something having the
property that all of its essential characteristics can be conveyed in
a message
"abstract document" as used by TimBL - maybe something like what's
described in http://www.w3.org/DesignIssues/Generic.html ?
fftr:IR (David Booth) - mathematical mapping from Request x Time to Response
2xx-responder - a thing X for which there is a URI U such that (1) U
names X, (2) some GET on U yields a 2xx response

(having laid these out I don't want anyone ever to call something an
"information resource" again!)

I would venture to say that it would be difficult to argue a
distinction between Accessible and an RFC 2616 resource, so I will
cheerfully assume they're the same.

AWWWIR is important because it occurs in a W3C recommendation. I still
don't have a clue how to tell AWWWIRs apart from other kinds of
things, but neither do most people, I think, so I don't think I'm at a
disadvantage.

I am reliably informed that ordinary web pages (the things named by
200-yielding URIs whose naming authorities have said nothing in
particular about what the URIs denote) are AWWWIRs, so let us take
Accessible to be a subclass of AWWWIR. It's a proper subclass, since
it's pretty clear that not all AWWWIRs are Accessible ("can be
conveyed" vs. "causal pathway").

I think I sort of understand what Tim's getting at with "abstract
document" but I can't tell whether all, or even any, of these things
are Accessible. If they're "conceptual entities" then I don't see how
I'd get a causal pathway to one; you get the pathway to some apparatus
that wrangles a representation space, not to the resource.

I'm not sure how fftr:IRs relate to Accessibles. Again, I think the
classes are disjoint, since I'm not sure how I'd get a causal pathway
to a function.

httpRange-14 permits one to use http: URIs to name things that aren't
allowed by RFC 2616, but says that any 2xx-responder ought to be an
AWWWIR. However, incorrect use of the http: scheme so extended can
result in non-AWWWIRs that are 2xx-responders (e.g. Dublin Core URIs).

But all that aside... one question I have is whether the classes
Edition and Accessible intersect. When I forge a network connection to
a server, and obtain from it a copy of the Edition in the form of an
HTTP entity, have I accessed the Edition, or merely some realization
(instantiation, manifestation) of it? If the Edition is "Platonic"
then I don't see how there can be a causal pathway to it. But maybe I
am not being imaginative enough here. Is this an example of
destructive hair-splitting as described in your slides, or it an
essential aspect of web semantics, as you suggested in your last
email?

(end of putatively objective part)

In a way I prefer either AWWWIR or AbstractDocument as the restriction
to place on 2xx responses, rather than Accessible, since it gives me
the wiggle room needed to give 2xxs for things like Editions.
(Never thought I'd come to the defense of the AWWW IR definition.)
Clearly an Edition and an Accessible that delivers its content are
different things, and if both have URIs then the URIs must be
different. But maybe I would rather not say which one of these is
named by a 2xx-responding URI (ambiguity), or maybe I would like to
use a 2xx-responding URI to name the Edition, foregoing the ability to
use that URI to talk about the Accessible to which a "causal pathway"
is forged.

This question (should 2xx responses be allowed for Editions?) has been
argued before, I think with Moby Dick as the example. Sounds like
you've been on both sides of this debate. So have I.

I've been harrassed for trying to get a definition of "information
resource" that's less ambiguous than the one in AWWW. I know that many
people have not looked to others for this, but have simply made up
their own. I have tried hard to avoid this question. But I can't help
worrying about dissonance between all the various definitions of
"information resource" and how RDF is used in the wild with
2xx-responders as the subjects of statements. Clearly 2xx-responders
are being used as Dublin Core subjects, even though what's meant is
rarely an Accessible but rather some Edition (or another "abstract"
thing). Similarly for Creative Commons metadata, which may be the most
voluminous use of RDF outside certain high-volume RDF specialists - on
the face of it these statements are assertions about copyrightable
material (as defined legally), not Accessibles. While accessibility is
a great explanation of the relation between reference and the web, we
need to be careful that we're not tilting at windmills by saying that
all this RDF is wrong (the RDF is malformed, the vocabulary is wrong,
or the 2xx responses are out of protocol).

I look forward to being proven wrong!

-----

... The analysis into multiple classes sort of takes the wind out of
the question I started with, which was whether there is, or ought to
be any constraint on how resources relate to their representations.
Does P is-representation-of R nontrivially imply (or contradict)
anything else about R or P? Only in conjunction with other statements
(asserted by the naming authority, say). But I'm suggesting there is
probably *some* pattern that holds *some* of the time. I want to say
that if the resource (as described by the naming authority) has a
particular author, then none of its representations (as obtained via
GET) should have a different author. This may be full of type errors,
but I don't think they're irreparable, and I think we should be able
to find a way to say it that we're both happy with, and then explain
how one might account for such a state of affairs.

Best
Jonathan

Received on Sunday, 23 November 2008 03:38:33 UTC