W3C home > Mailing lists > Public > public-awwsw@w3.org > April 2009

Re: Next AWWSW telecon Tue 4/14

From: Jonathan Rees <jar@creativecommons.org>
Date: Tue, 14 Apr 2009 08:39:09 -0400
Message-ID: <760bcb2a0904140539w3b0c9d2fuff1ea3d34cc05472@mail.gmail.com>
To: Michael Hausenblas <michael.hausenblas@deri.org>
Cc: AWWSW TF <public-awwsw@w3.org>
On Tue, Apr 14, 2009 at 6:29 AM, Michael Hausenblas
<michael.hausenblas@deri.org> wrote:
>
> Jonathan,
>
>> If Michael is present: report on what Michael has done
>
> Yes, Michael will be present ;)
>
> Additionally I've got a new proposal [1], based on our early discussion re
> IR/NIR:
>
> As we seem to agree the term non-information resource is (at best)
> misleading and only relevant for certain applications (such as found in the
> Web of Data realm). Still, we might want to be able to differentiate between
> 'things available on the Web' and 'things, be it abstract things like
> concepts or real-world objects, which are *not* available directly on the
> Web, but only descriptions of them exist (that are accessible via the Web)'.

Any time you define any class, you are distinguishing between the
things in that class from the things that aren't.  Thus we have people
and non-people, countries and non-countries, etc. I've been trying to
steer the effort toward developing and refining as many definitions
and properties as necessary to capture the different things and
phenomena that are on the table. So if you can articulate what's in a
class, you'll automatically know what's not in it.

Defining a thing-on-the-web class might be useful (assuming the
definition were honed a bit), so let's try
to do that. But having done so we would then have to ask how that
class relates to the other classes we've been talking about.

I don't like your characterization, because there are many things that exist,
are not on the web, and are *not* described. There are also many things
that are just like things that are on the web, in that they *could* be on the
web, but just happen to not be on the web.

> Well, here is my proposal: as depicted in [1], let us define 'resource', and
> a sub-set of it, called 'dereferencable resource' where the former covers
> everything one could imagine and the latter is the base for all generic,
> FTRR, REST, etc. resources. This would enable us to talk about resources in
> general (that is in applications that don't have to or don't want to care
> about the difference) and point out the difference in cases where this is
> important (aka Web of Data).

Actually I'd say that one 'dereferences' a URI, not a resource
(generally), and that one 'accesses' a resource (an accessible
resource, in particular). Let's be extra careful about use/mention
confusions.

And please - no more definitions of "resource"! We already have at
least four of them (2616, 3986, AWWW, and RDF) and certainly don't
need another.

What you are describing is similar to what RFC 2616 calls 'network
data object'. That's fine but note that 'accessible' (or 'has
dereferenceable URI') is not at all the same as what TimBL calls
'generic resource' - in particular, while a 'network data object' is
on the web by definition, there are generic resources that are not on
the web. One thing we could try to figure out is whether there exists
(in the model we are developing) a 'network data object' that is not a
'generic resource'.

Accessible is also different from ftrr and REST. This is my point in
making all of these diagrams: "on the web" is just plain different
from the cluster of {REST, FTRR, generic resource, information
resource} because the latter terms are not defined with respect to the
web - the members of these classes include many things that are not on
the web (and, if you listen to Larry Masinter, they are *disjoint*
from
the class of network data objects - a network data object can *hold*
Moby Dick, so that you get (a copy of) Moby Dick when you GET the
network data object, but Moby Dick is not itself a network data
object).

As I've done with others, I'm going to push back on any attempt to
look at the class 200-response-permitted-for until we've understood
*all* class definitions and their interrelations more clearly. It
sounds like you want to go straight to that issue, and I cry foul.
We've got to understand the completely ordinary web-things whose URIs
are leading to 200 responses before we can possibly go beyond that to
semantic-webby things.

> Forgive me in case I'm rehashing old/already existing proposals - this is
> likely not due to my ignorance. Happy to learn if others had thoughts in the
> same direction.

I'm sort of frustrated because I feel like I'm saying the same thing
over and over and people aren't getting it (not that this isn't my
fault). I read what's been written, with special attention to what's
said "normatively", put all existing definitions in one place, try to
explain the differences, and ask for help making a coherent synthesis.
I thought that this accessible vs. generic-resource distinction had
been made clear, as both Tim and I have brought it up many times, but
I admit that the idea is not necessarily accessible (as it were). If
discussion to date is not recorded in a form that makes sense, that's
a problem we have to fix.

Thanks
Jonathan
Received on Tuesday, 14 April 2009 12:39:44 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 14 April 2009 12:39:44 GMT