Re: Resources and URIs

(To cut to the chase, go to @@)

>>Skolemized URIs work just fine. The existing Web protocols and 
>>conventions are perfectly adequate, in fact, to handle this. They 
>>are BETTER than the current RFC 2396 prose says they are.
>
>You have said that a few times, but the problem isn't what RFC 2396 says
>but rather what you assume it means.

It was, and is, the best I could do by reading the text as carefully 
as I know how. I didn't assume anything: I tried to follow what it 
was trying to say.

>You asked for a definition of
>identity, and I gave it, at which point you asked me to define all
>of the words within that definition.  That is nonsensical.

Im afraid I do not regard it as nonsensical, if the vocabulary used 
by your definition is more obscure than the original, which it was.

I share your frustration with what could be an endless lexicographic 
discussion, which is why my suggestion was to remove the potentially 
contentious wording, rather than to attempt to define or clarify it.

>Furthermore,
>your entire argument boils down to that we should consider changing
>the definition, which of course is why the issue is on the issues list
>and was listed as "pending".  Everything that has been said in this
>discussion is just a verbose variant of what was already said and
>summarized within that issue.  That is why I have been and will continue
>to be brusque in my responses: I am sick of reading fifty paragraph
>mails for a request that has already been conveyed in one paragraph.

I must apologize for contributing to the some extended email 
interchanges which were aside from the main editorial point. I will 
focus on issues of wording.

>To answer your real question: the definition in RFC 2396 means exactly
>what you wish it to mean.  Resources are things that have identity --
>not one unique identity, but an identity that is being referenced for
>the sake of denotation.

Oh dear, that makes things worse. Now I REALLY do not know what you 
are talking about.

>  That is a common English word and the way I
>used it in the sentence does not allow for any other reasonable
>interpretation

I have no idea what your intended interpretation was (or is).

>, including the one Miles described about the identity
>axiom in mathematics.

It really isn't enough to say that it is a common  English word. In 
that role it has many meanings, and several of the more common, as in 
the useage 'identity theft', apply only to persons, so become 
meaningless when applied to a larger domain of just about anything. 
(What would it mean to apply identity theft to my toaster, or to a 
sodium atom?) And I do not know of ANY sense in which something can 
be said to have more than one identity, except perhaps metaphorically 
("Alphonse, SuperSpy, Man of Many Identities")

>  After reading this discussion, it appears that
>you grasp the concept without any problem

Not at all. I honestly do not know, even after a great deal of 
reading, what you mean. The best I can do is  to assume that you are 
trying to say several different things at the same time; but even 
when I get the picture reasonably clear, the ground seems liable to 
shift unexpectedly. The above, for example, suggests that you think 
of things as having many possible identities, and that one refers to 
one of these rather than to the thing itself. This comes out of left 
field for me. I find this an extraordinary concept, and I have no 
idea what kind of ontological framework would arise from taking it 
seriously. So I'm afraid that I still have no confidence that I am 
really grokking you.

>, but you are being faced with
>other arguments that fail to grasp it or are using the multiple potential
>definitions of the term "identity" as some bizarre excuse to claim that
>a URI defines the resource's identity.  Okay, that was Miles point as
>well, so we can replace use of that term with the definition.  However,
>I am not going to proceed recursively down the ontological tree until
>everyone is satisfied -- I just want to get it to the point where the
>intended definition is distinguished from those that are not intended,
>even if I can't understand how people could confuse them now.
>
>I do not accept the argument that nothing should be said about resources
>in the spec.  The URI-WG tried that originally and failed miserably --
>dozens of identifier schemes and related projects were created based on
>false assumptions about the limits of resource.  Years of effort were
>wasted in attempts to create "better identifiers" that didn't suffer
>from the imagined limitations of existing identifiers.  That is why
>the definitions were added in 1997, and I'd rather hash this out in
>one place than have to explain it again in every single working group.
>
>In fact, I don't think the problem you are describing has to do with
>confusion over "identity" at all, but rather what it means to "identify"
>a resource.

That is a separate question, though some folk apparently understand 
the first to be closely related in meaning to the second. I have 
worries about both of these, in fact, and I am not sure whether or 
not the document intends to imply that they are related.

>In that sense, it is unfortunate that they are called
>identifiers, but I can improve that definition as well.  In fact, I
>already did that yesterday:
>
>http://www.apache.org/~fielding/uri/rev-2002/rfc2396bis.html#rfc.section.1.1
>
>If anyone has comments on that text, please be specific.

OK, will do.

The first sentence of Resource says
A resource is a source of supply or support within the context of a 
given system.

This seems to go off in an entirely new direction. The only way I can 
understand this, it would rule out all kinds of things from being 
resources. I can refer to entities which have never existed, or which 
have existed but do not exist now, or which will exist in the future, 
or which are too small, too large, or too remote to serve as 
'sources' of anything.  So this seems to directly contradict the next 
sentence, if that is being understood in the way it has been up to 
now.

Taken together, the first three sentences of your current draft seem 
to clearly indicate that the term 'resource' is reserved to mean only 
things which can be in some sense causally influence the system, 
things it can in some (perhaps broad) sense interact with, that the 
system can 'use' (fourth sentence). So anything which it cannot 
causally interact with cannot be a resource? That seems like a much 
more restrictive notion than the one we were discussing previously. 
Is that really your intent? If so, thanks for being clearer, but I 
would urge you to reconsider.

The second paragraph seems to contain a number of presuppositions. 
Please allow me to list them in order to clarify whether or not they 
are intended:

1. Resources are represented at times
2. Resources endure through (all?) time, retaining their identity.
3. A single, enduring, resource may (must?) have different 
representations at different times
4. There are mappings, called 'conceptual mappings' from resources 
(at times?) to representations.
5. The system 'views' a resource as being a conceptual mapping from a 
(from that?) resource to representations.
6. resources 'supply' representations to the system.

Now, I have a lot of trouble understanding exactly what this all means.

First, where did this idea of representations come from? This section 
is supposed to be about resources. Does 'representation' here refer 
to *any* kind of representation, including for example descriptions, 
graphical renderings, etc., or is it restricted to a particular class 
of representations (which class?)

Second, there seems to be an implicit assumption here of a kind of 
traffic between the resource and the 'system', involving 
representations. Resources are represented by representations, but 
they also emit representations and supply them to the system, 
providing a 'view' of resources (of themselves?). I can make out 
here, obviously, in rather blurry outline, a sketch of the Web; but 
the document would benefit by this picture, if it is indeed intended, 
being made more explicit, since it could be interpreted in other 
ways.  But also, if I do interpret these paragraphs in this way, the 
notion of resource becomes, if anything, murkier: for the word seems 
to be being used to refer both to the Web sources of information - 
the functional entities which return information when pinged for 
example by a GET - *and* to the entities that the resulting 
representations are 'about', *and* to the mappings from the former 
things to representations (and we are not told about the relationship 
of these to resources).  Does it in fact refer to all of these, or is 
the text meant to suggest that these are all the same category?  But 
surely the whole point of the Web is that it is used to convey 
information about far more than its own servers.  (I know  that this 
is returning to ancient debates which run on for ever.  My point is 
that the wording being used here does not help us avoid these 
debates: it only serves to forever cause them to be re-opened, since 
it can be understood ambiguously in apparently conflicting ways, and 
it can also be read as asserting that these categories are the same. )

Third, I (still) find the sentence "This conceptual mapping of 
resource to representations is what the system identifies as a 
resource."
quite opaque in meaning. It seems to say that a resource *is* a 
mapping from itself (at a time?) to a representation of itself, which 
is simply incoherent. Read more carefully, perhaps it only says that 
this is the systems' view of the resource: but what does that mean?

Fourth, what 'system' are we talking about here? (The Web? a Web 
browser? Something else?) If resources can be almost anything, what 
has 'the system' - in fact, ANY system - got to do with the identity 
of a resource? Please don't tell me that 'system' is just an English 
word.

Fifth, why bring time into it? This seems to add needless complexity 
to an already complex issue, and also to impose a particular 
ontological perspective on the entire discussion which might be too 
limiting. There are already deployed industrial ontology standards 
(eg. EPISTLE) which reject, for good technical reasons, the 
'common-sense' picture of things enduring through time without 
changing their identity; and you are immediately embroiled in 
potential debates about whether such endurants can be said to endure 
for ever, or when it is appropriate to decide what the temporal 
boundaries might be, etc. etc. (There's enough discussion of this 
stuff in the technical ontological literature to fill a library, 
never mind philosophy and linguistics.) In any case, there may well 
be transient - non-enduring - resources which it is still useful to 
refer to and exchange information about. At the very least, I would 
suggest choosing words which do not seem to imply that *all* 
resources must have this kind of enduring temporal identity.

The next paragraph, by the way, although clearly intended to be an 
clarifying example, seems to say that a single representation can be 
of two distinct resources at the same time. In fact, in the case 
indicated, it *must* be. (It might help if you could indicate what 
the 'conceptual mapping' is for this example. I presume it must be 
something like a function from times to camera views (?); but in this 
case the same function identifies both resources, apparently.) (BTW, 
issues of identity get very delicate when we start to talk about 
mappings and times. In particular, it simply doesn't make sense to 
say that a mapping takes times as an argument and also itself 
endures. This is like asking how fast time goes by: the question 
embodies a temporal contradiction. Either say that the mapping is 
itself temporal, so that the mapping of the camera changes every time 
a picture is taken, or say that the mapping - which itself now exists 
outside of time altogether - is from the camera and a time to a 
picture.)

Concerning the text of 'Identifier': can you indicate the intended 
distinction between "define" and "provide sufficient data to 
distinguish....from all other things..." ? I have trouble 
distinguishing these concepts.

Finally, the idea that a URI carries data to identify the referent 
seems wrong to me: surely what the URI identifies is the 'owner' of 
the URI, which is often the authoritative source of information 
concerning the intended meaning or referent of the URI.  But the 
distinction seems important.  This is why I have been howling at 
Graham and Patrick on your email list, because the idea that URIs 
always contain enough information in their syntax to uniquely specify 
a single *referent* is incoherent; but the idea that they do so 
uniquely identify a *source of information* is not only coherent, but 
seems to be close to correct. And of course, in some cases - as when 
the resource in question IS the source of information - the URI might 
indeed identify the resource. But not in all.

Pat Hayes

@@ PS. suggested modified text (based on 
http://www.apache.org/~fielding/uri/rev-2002/rfc2396bis.html#rfc.section.1.1

"Resource

The Web supports traffic in representations, and resources are 
whatever is indicated or referred to by a  representation.  In 
general, a resource could be anything. This specification does not 
place any limits on the nature of a resource.  In practice, a 
system's resources will usually include the things to which it can 
apply actions in the future. For example, if a system is performing 
an information retrieval action, then things that can provide 
information to that system are considered resources for the system. 
Familiar examples of a resource include an electronic document, an 
image, a service (e.g., "today's weather report for Los Angeles"), 
and a collection of other resources.  Not all resources are network 
"retrievable"; e.g., human beings, corporations, and bound books in a 
library can also be considered resources. For some purposes, 
resources may include abstractions, non-physical things or fictional 
objects.

A common misunderstanding of resources, particularly in information 
systems, is to equate a resource with a representation of that 
resource as observed at a particular instant in time. The same 
resource may endure though time and have different representations at 
different times; a change in the representation need not be 
considered a change in the resource. For example, a resource that is 
defined to be "a live camera view from my window" may continue to 
exist as the same resource, even though the view is changing.

Identifier

An identifier is an object that carries identifying data.  The data 
may identify a resource or a source of information about a resource. 
A URI is an identifier object consisting of a sequence of characters 
that matches the restricted syntax defined by this specification.

------

I have tried to keep to the spirit of the original without saying 
anything that would cause damage. I avoided the 'conceptual mapping' 
language in order to avoid getting into potential tar-pits about 
whether or not mappings endure through time, etc. etc., and tried to 
use 'might' and 'often' style language rather than 'do' and 'must' 
style language.

The two sentences starting "In practice, ..." clearly serve a useful 
role in the explanation, but I worry that by being used so centrally 
in the definition, particularly with the original 'however', they 
convey the implicit message that ALL resources are of this 
'actionable' variety. I have tried to add words to guard against this 
interpretation; but it might be easier to just cut this out 
altogether.

(It seems to me that to say this stuff properly requires getting into 
issues like one resource being a source of information about other 
resources, an area I know you have thought about more extensively 
than most people have. I guess the point is that this stuff needs to 
either be said explicitly or not said at all, since to kind of 
half-say it only creates more potential confusion than clarity.)

PPS. I feel reluctant to engage in writing someone else's prose for 
them, which is considered highly rude and intrusive behavior in the 
circles I move in; so please don't be offended by my rewriting your 
words for you. I gather this is the right way to behave here.
-------
-- 
---------------------------------------------------------------------
IHMC					(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
s.pam@ai.uwf.edu   for spam