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

RE: Are generic resources intentional?

From: Williams, Stuart (HP Labs, Bristol) <skw@hp.com>
Date: Mon, 8 Jun 2009 13:24:44 +0000
To: Jonathan Rees <jar@creativecommons.org>
CC: AWWSW TF <public-awwsw@w3.org>
Message-ID: <233101CD2D78D64E8C6691E90030E5C832D1BD4358@GVW1120EXC.americas.hpqcorp.net>
Hello Jonathan,

> -----Original Message-----
> From: Jonathan Rees [mailto:jar@creativecommons.org] 
> Sent: 04 June 2009 20:53
> To: Williams, Stuart (HP Labs, Bristol)
> Cc: AWWSW TF
> Subject: Re: Are generic resources intentional?
> 
> On Wed, Jun 3, 2009 at 6:17 AM, Williams, Stuart (HP Labs, Bristol)
> <skw@hp.com> wrote:
> > Hi Jonathan...
> >
> >> -----Original Message-----
> >> From: public-awwsw-request@w3.org
> >> [mailto:public-awwsw-request@w3.org] On Behalf Of Jonathan Rees
> >> Sent: 28 May 2009 13:49
> >> To: AWWSW TF
> >> Subject: Are generic resources intentional?
> >>
> >> I've been puzzling over the question of how two generic-resources can
> >> have the same trace by virtue of a difference in meaning, e.g. the use
> >> case Tim gives where he and I both work at Burger King for a week and
> >> end up with identical-looking time sheets (same trace), that are
> >> really distinct generic-resources simply because of properties not
> >> reflected in the traces.
> >
> > I'm not sure that I can buy the argument that these two 
> > timesheets can have the same traces. There have to be 
> > material differences otherwise either or both of you would 
> > gte paid twice, or some random other(s) would be collecting 
> > your wages.
> 
> The trace consists only of the wa-representations (as indexed by
> parameter tuple, including time), and the traces are the same by
> hypothesis.

Fair enough...

> I asked Tim whether two distinct generic resources could
> have the same trace, and he said yes, and gave the time sheet example.
> As he's the authority on this ontology... well either this is just
> true by definition, or we have to persuade him that this idea is not
> useful.

:-)

> The difference that lets us be paid separately is not in the material
> evidence of the resource itself but rather in the state of the larger system.
> Perhaps I tell the employer that the URI of my time sheet is U1, while
> Tim says the URI of his time sheet is U2. U1 and U2 might have the
> same trace, but that's OK because the employer might not even notice
> that they do. The understanding of the two accounts resides outside
> the resources themselves.
>
> > Ok... lets allow that the wa-representations and indeed the 
> > traces of the timesheets carry no distinguishing marks that 
> > allow and observer to discriminate - I such a situation I 
> > think that you have only one resource that happens to have many names.
> 
> Personally I think this is a much more clear and sensible model than
> the generic resource model, and probably a more useful one. When I
> asked Tim the question I expected the answer you gave. But he said no,
> so it's not the one under consideration.

Ok...

> Perhaps it should be added to the mix though, as a new kind of thing
> to talk about.
> 
> However I should point out that in Roy's presentation there are many
> REST-resources that don't have any wa-representations at all
> ("abstract concepts"). These would all have the same trace (i.e. the
> empty trace). So this is more congruent to Tim's model than is your
> formulation.

I'm aware of that, though from a slightly different perspective. IIRC in the short "Principled Architecture..." paper the time varying mapping function that Roy speaks of is to set of 'equivalent' representations or to URI. I think this later is often missed, but I think provides at least for the 3xx redirection cases. So... in the "abstract concepts" cases the fielding mapping function may yield a URI instead of a wa-representation that be a reference to a descriptive resource that has more to say about the "abstract concept" without actually being it (equally it may yield a 4xx response or indeed potentially no response whatsoever).

Certainly if you are thinking of a completely empty trace, no wa-representations, no URI, then you pose an interesting corner case, akin to me being a member of the class of people "all of whose girl children have blue eyes" (I have only sons). I think with such an empty trace (as a special case) I think I would say that it is not possible to conclude sameness or difference for distnict resource references that yield the null trace.

But... I guess we are intested in Tim's framing here.

> Would you say that traces and these Stuart-resources are in 1-1
> correspondence, or are there Stuart-resources that don't have traces
> or traces that don't have Stuart-resources?

Hmmm... interesting questions. I think that there are resources that have material side effects on the world that are not directly observable in their traces. Eg. Two synchronised robotic arms that have document/resource oriented interfaces. Both may indeed exhibit the same trace behaviour and relative to their own origin each robotic arm may exhibit identical positions and orientation for their various joints. But they are different arms occupying different absolute temporal-spation positions. FWIW: I also think that way into the future the likelihood is that the robot arms would not remain synchronised and their traces would - in the long term - diverge, but that's more a pragmatic argument.

Tim's timesheet resources also have different side effects in the world (different accounts get credited with wages).

> >> Allow me to call this difference "intent" - I won't define this but
> >> Alan, don't jump all over me, build on what I say. It is the missing
> >> dimension, the resource's "essential characteristic" that is not
> >> conveyed in any wa-representation.
> >>
> >> I don't know what "intent" is ontologically; I use the word as a
> >> placeholder. It has to be quite broad, so that it allows me to say
> >> that the GR named by http://random.org/integers/?num=100&min=1&max=100&col=5&base=10&format=html&rnd=new
> >> has "intent" - namely the web site author's intent to satisfy the
> >> world's hunger for white noise. Generally, we have grandfathered all
> >> "web pages" just by saying that none of them are accidental
> >> (unintentional) - someone went to the trouble of registering a domain
> >> name, setting up a server, and deploying content. We also get Moby
> >> Dick, since someone went to the trouble of writing it - it didn't fall
> >> from the sky. And we get Finnegan's Wake and Beethoven's 7th for the
> >> same reason even though know one knows what message they convey.
> >>
> >> But I think "intent" can explain another use case Tim has given. He is
> >> adamant that numbers, such as the 46th Mersenne prime, are not
> >> information resources. I was puzzled by his refusal because it seems
> >> very clear to me that all of its essential characteristics *can* be
> >> conveyed in a message; I just did so.
> >
> > :-)... well not for me... I don't know what a "Mersenne 
> > prime" is let alone the 46th one, though that seem to clearly 
> > be descriptive of some ordinal position in a series. (/me 
> > heads for wikipedia).
> 
> Every definition - indeed every piece of communication - will require
> prior knowledge for processing. That you happen to not know the words
> I use does not make the definition different in kind from other
> definitions, since one can never depend on all readers understanding
> every word one uses in a definition. (E.g. someone might be a novice
> in the English language, and might have to look up even the simplest
> words in a dictionary.) In this particular case there is no argument
> among mathematicians or anyone else what is meant by the phrase, and
> it is easy for you to find determine what it means, e.g. via wikipedia
> as you say.

Yes... I understand. Somewhere there is a line between what is explicitly conveyed in a message and where one appeals to commonly held background knowledge... Eg with wa-representations we allow ourselves (I think) to admit media-type specifications as part of the background that are not conveyed explicitly (nor in general is there a machine readable articulation of a given media type specification... and even if there were...).

Wrt to numbers, I guess the question is more what their common characteristics are and what distinguishes one from another - eg each forms an equivalence class under the relation "has the same quantity as" or some such and at least if we're talking about integers each has a distinct predecessor and successor but framed that way the essential characteristics of one of them takes in at least half the set of integers (my number theory is evidently a bit rusty).

> Jonathan

Stuart
--
Received on Monday, 8 June 2009 13:25:45 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 8 June 2009 13:25:46 GMT