RE: erratum Re: resources and URIs

OK, thanks for that. I agree that the letter/address/envelope analogy 
makes perfect sense, and is consistent with the architecture. 
Unfortunately it is not consistent with the other claim, that URIs 
*denote* resources, that 'identifies' means 'refers to' rather than 
'can be used to access' (in some suitably general form). So, to 
return to my original request for clarification, can we please have a 
clear statement one way or the other about the intended meaning of 
"URI indicates resource"?? Is this like "address on envelope 
indicates where to send, and bring back the reply" or is it like 
"name refers to entity"? It can't be both at once, sorry.

>It is tough to say the address on the envelope
>is the content of the letter, but that is the
>way the post office sees it.

Actually that is not how the post office sees it, nor the legal arm 
of the government. It is a federal crime to open a sealed letter if 
you are not the recipient or the recipient's agent. The address is 
all that the post office needs to concern itself with, agreed: but it 
is not the content of the letter, and God and a good lawyer help any 
mailman who gets the distinction confused. The analogy with URIs and 
meanings is almost perfect, in fact, even down to calling a letter a 
'representation'.

>Someone opening
>the envelope doesn't.  Use depends on the handler.
>
>The system conflates address with content.

The post office doesn't; are you saying the Web does? I see no reason 
for this claim: can you elaborate on it?

>That sleight of hand is fundamental to the
>web architecture which uses mapped locations
>as addresses

Yes

>then addresses as names.

Where? How? I would be grateful if you could explain why this is 
fundamental to the architecture. If you are right, then I may be 
missing something important.

>It works
>and all of the failure modes are predictable.
>
>It only quits making sense when a URI has to be
>meaningful independent of the content.   Then
>the babbling about 'resource' starts, but actually,
>it is still an address.   The post office system
>can sensibly have addresses for empty lots and
>for lots which no longer exist because they
>have been merged, overtaken by flood, whatever.
>The address still makes sense to everyone but
>the letter carrier.  He can't drop off, and
>since nothing is mailing from that address,
>there is no problem because he returns it as
>undeliverable.

Actually, in Florida, he just throws it away and doesn't tell anyone. 
Unless its a magazine he wants to read, of course, then he keeps it, 
reads it, and then throws it away. The address I gave did not make 
sense. The 1400 block of W. La Rua St. makes sense, but not 1404. 
You could make guesstimates of roughly where it would be if it did 
exist, but there is no such tract on the city plans: that block was 
always a cemetery and has never been subdivided.  Another example 
would be 1927 Ivy Lane, Palo Alto, Ca. 94303, which could not 
possibly exist, since there is no space between 1916 and 1945. It is 
easy to create addresses which are not the address of anything, just 
as it is easy to create URIs which are not the URI of anything. 
Nothing is gained by inventing nonexistent things for them to 
designate when they do not designate anything.

>  A user of the address
>only has to determine valid uses (operations).
>
>That system is as reliable as can be expected
>sensibly.  It isn't 100% but the letter carrier
>does have defined operations for failure modes:  404, etc.
>
>There isn't any better definition than
>some resource is on the web if it has
>a URI and is not on the web if it doesn't
>because the web is and always has been
>fundamentally, a system using an addressing
>system.

Let me paraphrase.  A URI is a URI if it is a URI. Does that amount 
to the same observation? Seems to me to say about the same thing, 
since 'on the Web' means 'having a URI', but there doesn't need to 
actually be anything which is the resource which has the URI (notice 
how the careless use of 'exists' leads us into nonsense very quickly) 
so to say that being 'on the Web' just means 'being the resource of a 
URI' is actually misleading, since the resource (which our design 
requires to be identified by any URI) might in fact not exist; and if 
it does not, then we just invent one to be the missing resource, by 
fiat, even when there is no such resource, so that we can say it 
exists *on the Web* even if it doesn't really exist.

In your excellent analogy, this is like analyzing the postal service 
by saying that every address *must* be the address of some unique 
delivery box. When faced with the manifest falsity of this claim, the 
response is to say that even if there is no delivery box at that 
address, we will just invent a virtual delivery box to be the 
delivery box of that address, because every address *must* have a 
delivery box, by fiat, even when there is no box there to be the 
delivery box: it is required by the architecture.

>Saying it is on the web means it
>has a valid address and a route to that is
>possible if not guaranteed.

A route to something nonexistent is not possible, pretty much by the 
meaning of 'exist' and 'route'.

>Uniqueness of identity is proven in terms
>of a process of identification; it makes
>no useful sense otherwise.  It is a provable
>or unprovable property but it is only a
>problem in the question of the proof.
>The SW has some real issues with proof.
>And it always will have if any provable
>property relies on analogical proofs.

You have lost me there. Im afraid. There is no need to prove 
uniqueness of identity: that amounts to the observation that x=y 
implies x=y.

Pat Hayes

>
>Since systems of human
>reasoning are always analogical in their
>initial process, the SW will be a lawyer's
>paradise.  See the EBay problem:  we can
>collect the names for the agencies easily,
>but proofs of identity are beyond the system and will
>rely on case-based evidence.  See the Judge.
>
>The rest of the stuff is other systems
>that make use of those addresses.  How
>they are used makes up the REST of the
>story.  Now one has to determine what of
>the REST of the story is the chapter on
>web architecture.
>
>len
>
>-----Original Message-----
>From: pat hayes [mailto:phayes@ihmc.us]
>
>OK, but that's my sense of 'link' and the more limited sense of
>'resource', what TimBL calls an information resource. I'm quite happy
>with that notion, of course, as long as we can agree to stick to it.
>
>>... I'd assume that it would return
>>a 404 ... BUT ... by creating the URI you actually do 'create' a resource.
>
>Really? You do? Well, OK, if you say so.  But if it is that easy to
>create a resource then what do we need the Web for? I can write a
>simple piece of code which will generate syntactically legal URIs at
>an astonishing rate: am I creating resources?
>
>Seems to me that a more useful notion might be "actual information
>resource", which requires rather more effort to bring into existence:
>something that really can actually emit representations when pinged,
>something that costs real money and uses real energy.
>
>>It does 'exist' as the resource whose URI is
>>http://this.does.not.exist.com/index.html ... now Pat uses an example of an
>>address *for the physical world* which may or may not physically exist. In
>>URI space, as long as the URI is valid, the resource 'exists'. There is
>>nothing preventing the owner of 'exist.com' from creating such a resource
>>... the only issue I can see here is that since you don't own 'exist.com'
>>the URI you've created is 'illegal'. URI space as opposed to 'physical
>>space' contains a point for every legal URI ... such points might be called
>>'resources', and for every URI there does exist a resource (in URI space).
>
>Well, that makes URI space into a purely theoretical construct,
>right? Now, what theory does it serve any purpose for?? What theory -
>and what *kind* of theory - requires us to hypothesise this virtual
>URI space of virtual resources? We don't need it for semantics - in
>fact, its does more harm than good - and I can't see why we need it
>for an architectural theory, since it plays no role in the
>architecture. In fact, it strikes me as just a piece of nonsense made
>up to seem like a theory, which has become a kind of doctrine that
>must be defended at all costs and is beyond the reach of reason.
>
>>We might similarly say that "1404 West La Rua St., Pensacola" is an
>>_invalid_ address per the rules of either the postal service or of
>>Pensacola, FL ... just as per the rules of "exist.com"
>>http://this.does.not.exist.com/index.html might be invalid.
>
>Right, but it is invalid precisely because if you follow the rules
>for dereferencing postal addresses, you will find that there is no
>such place (it would be in the middle of a large cemetery). In other
>words, the 'resource' - in this case, the building at that address -
>*does not exist*.
>
>BTW, this example is close to my heart because of an amusing
>incident. My local cable company mistyped my address and sent my
>bills to this non-address (mine is *east*, not *west*). The post
>office simply trashed them and didn't tell anyone (no 404 errors). I
>never saw them. The cable company sent demands, then threats, then a
>technician to cut off my service.  The technician, finding the
>graveyard, reported that there was no service to that address. Then
>the debt collection agency sent even nastier threats, all to this
>nonexistent address; I had no idea this was happening.  Eventually, a
>store in California refused to accept my credit, and the subsequent
>enquiries uncovered the trail of uncorrected confusions. It will take
>7 years to get this mess off my credit reports. Now, should I console
>myself by saying, Ah, but there really *was* a resource there - a
>kind of Platonic building in the graveyard that nobody could see,
>brought into existence in Address Space just by the typo on the
>envelope;  its just that those postal guys weren't able to do the
>dereferencing? Somehow, that doesn't seem to help.
>
>>...
>>>
>>>   >A citation of a book giving page number, paragraph and line number is
>>>   >still a link, even if the book no longer exists. Not all links can be
>>>   >traversed or dereferenced.
>>>   >
>>>   >"Creating a link to" == referencing, in some cases quite optimistically
>>>
>>>   Ah.  OK, with that usage, then I would agree; but that is a VERY
>>>   strange usage of "link", at least to one reader.  And I think it is
>>>   highly misleading, since "link" usually *does* imply at least the
>>>   possibility of dereferencing, in principle, in an ideal world, etc.;
>>>   whereas this sense of 'link' to simply mean 'refer to' clearly never
>>>   has, and never had, any such implication.  (What a referring name
>>>   enables you to do is to *think* and *talk* about something, not to
>>>   actually get your hands on it.)
>>
>>In the Web Architecture, you never can actually get your hands on the
>>resource, rather the representation.
>
>Is that really correct? Seems to me that you - or maybe the Web
>itself - does 'get its hands on' an actual information resource
>indicated by the URI, and that very thing then emits the
>representation. Isnt that (a sketch of) the Web architecture for all
>the ...TP protocols?
>
>>I don't see the problem with: "create a
>>link to" == reference.
>
>Well, I agree, that is OK also, although it makes a lot of the TAG
>document wrong.  But what isn't OK is conflating the earlier sense
>with this sense, because they are completely different and have
>different properties.
>
>>
>>In any case the URI "urn:isbn:xxx" does allow you to reference a book, and
>>using this URI you might be able to 'dereference' a physical copy of the
>>book i.e. via an indirection service such as amazon.com
>
>You might; but you also might not.  (Last year I tried to get hold of
>a book on modern Japanese prints, a present for someone. There are
>dozens of isbn#s for these books, but no copies for sale anywhere,
>not even on Ebay.) It has nothing to do with being a reference,
>though.  And almost certainly it is false to claim that there is a
>*single global* link from a URI to its corresponding resource in this
>sense, right? (The only example would be a case where there was just
>one copy of the book left in the entire world, in which case you
>almost certainly can't buy it on Amazon.)
>
>>
>>With this sense of 'link' it is
>>>   clearly just a mistake to think that there must always be a unique
>>>   thing 'linked to', for example: notoriously, one can refer to things
>>>   that do not and never have existed, and one can, and often does,
>>>   refer ambiguously.
>>
>>right an such resources (that never have physically existed) may indeed
>have
>>representations e.g. a picture of a unicorn, or a written description.
>
>Representations, yes: links to them (in any useful architectural sense), no.
>
>Pat
>--
>---------------------------------------------------------------------
>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@ihmc.us       http://www.ihmc.us/users/phayes


-- 
---------------------------------------------------------------------
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@ihmc.us       http://www.ihmc.us/users/phayes

Received on Monday, 21 July 2003 22:17:12 UTC