W3C home > Mailing lists > Public > semantic-web@w3.org > June 2007

Re: What if an URI also is a URL

From: Pat Hayes <phayes@ihmc.us>
Date: Wed, 13 Jun 2007 13:56:30 -0500
Message-Id: <p06230920c295e0f0aa87@[]>
To: Tim Berners-Lee <timbl@w3.org>
Cc: "M.David Peterson" <m.david@xmlhacker.com>, "r.j.koppes" <rikkert@rikkertkoppes.com>, "Yuzhong Qu" <yzqu@seu.edu.cn>, "Sandro Hawke" <sandro@w3.org>, <semantic-web@w3.org>, <swick@w3.org>, "John Black" <JohnBlack@kashori.com>

>On 2007-06 -11, at 17:09, Pat Hayes wrote:
>>Tim, as this discussion gets to the heart of 
>>what Ive been trying to argue for several 
>>years, please take the comments below as 
>>intended in a spirit of analysis rather than 
>>just pins and angels.
>Ok.  [takes a deep breath].  I'll go down this path with you.


>>>On 2007-06 -11, at 13:53, John Black wrote:
>>>>Tim Berners-Lee wrote
>>>>>On 2007-06 -09, at 21:22, M. David Peterson wrote:
>>>>>>On Sat, 09 Jun 2007 07:13:52 -0600, Tim Berners-Lee <timbl@w3.org>
>>>>>>>No. It cannot identify both a document and a person.
>>>>>>Tim: Will all due respect... WTF?
>>>>>I am using the 'identify' in the strict sense of 'denote'.
>>>>>The semantic web is like a logic language in which URIs are symbols.
>>>>Do you believe that by claiming to use the 
>>>>strict, logical sense of the word 'denote' 
>>>>you thereby cause or require such denotations 
>>>>to be absolute and unambiguous? Where do 
>>>>think denotations (or identifications) come 
>>>The architecture is that each URI is owned.
>>>  With HTTP URIs, this happens through the 
>>>domain name system and often  delegation 
>>>within a domain. Unlike a word, a URI has an 
>>>owner.  The owner attempts to make enough 
>>>information available that the URI can be used 
>>>by others without ambiguity in practical 
>>Well now, several points. First, do owners in 
>>fact do this? I havnt seen a great deal of such 
>>information-providing going on, myself. But OK, 
>>fair enough: perhaps you mean, they SHOULD make 
>>such information available. But second, more 
>>seriously, HOW would they do this? Take your 
>>>For example, W3C owns 
>>>and has delegated to me the right to say what 
>>>that URI stands for.
>>OK. So, what DOES that URI stand for? How will 
>>you tell someone what the referent is that you 
>>intend it to denote, so that they know what to 
>>use it for? Now of course, you and I being 
>>smart human native English speakers who are 
>>reasonably tech-savvy can look at this and 
>>figure out that it is probably meant to refer 
>>to you. But really, that does depend on us 
>>being this smart and savvy.
>Well,  I put it in the Tabulator and I get out (among other stuff):
>  Tim Berners-Lee
>	Assistant		Amy van der Hiel
>	HomePage		http://www.w3.org/People/Berners-Lee/
>	Work
>		Address
>			City		Cambridge
>			Country		USA
>			PostalCode		02139
>			Street		32 Vassar Street
>			Street2		MIT CSAIL Room 32-G524
>		Phone		tel:+1-617-253-5702
>		Latitude		42.361860
>		Longitude		-71.091840
>	Organization		expandfetchWorld Wide Web Consortium
>which to many people gives a fairly lear indication of what is identified.

No argument. Of course, as I said, people are 
smart enough to figure it out. But (1) machines 
aren't; and (2) for some purposes, this is still 
ambiguous. If someone or some thing makes a 
distinction between person-as-citizen and 
person-as-biological-system (say), then they 
might consider the above to still be ambiguous. 
No doubt, being competent human reasoners, they 
would not be unduly bothered by this, but the 
ambiguity of reference would still be there.

>  And in fact for non-english speakers, there are 
>words to the effect that "Sir Tim Berners-Lee is 
>geboren in Londen in 1955. Hij wordt gezien als 
>de bedenker en grondlegger van het World Wide 
>Web (WWW), dat was geïnspireerd door HyperCard. 
>Hij werkte hierin samen met de Belg Robert 
>Cailliau. Als directeur van het World Wide Web 
>Consortium houdt hij toezicht op de ontwikkeling 
>van webtalen en -protocollen als HTML, XML, CSS 
>en HTTP. Berners-Lee schreef ook de eerste 
>webbrowser, net als het web World Wide Web 
>genaamd, en de eerste webserver." and
>Sir Timothy "Tim" John Berners-Lee, KBE (TimBL o 
>TBL). Nacido el 8 de junio de 1955 en Londres 
>Inglaterra, se licenció en Física en 1976 en el 
>Queen's College de la Universidad de Oxford. Sus 
>padres eran matemáticos. Trabajando como 
>investigador en el Laboratorio Europeo de Física 
>de Partículas (CERN) de Ginebra, concibió la 
>idea de un proyecto de hipertexto global, que 
>años más tarde se convertiría en la world wide 
>>It isn't absolutely obvious: and there are 
>>many, many web pages out there which I really 
>>have no idea what their owners would say they 
>(Web pages? Web pages are documents.  I wouldn't 
>say they denote things.  Symbols, like URIs, 

Sorry, I was careless. I should have said, there 
are many webpages out there which, if someone 
told me that they were intended to establish a 
referent for a URI, I would have no (or only the 
vaguest) idea what the referent was.

>The URI 
>denotes me, and the URI 
>denotes the web page: an RDF one, but a web 
>page.  That web page describes mainly me but 
>also to an extent other people and things.

Yes, quite. In fact it can hardly do otherwise, 
if you think about how RDF works.

>>>To use it for something else is an error.
>>But surely if you make this ruling then it is 
>>up to you to tell me enough about what it is 
>>supposed to be used for, so that I can know 
>>when Im making an error. HOW??
>Well,  do you want the human way of the machine way?
>- The human way is that you are tempted to use 
>it to represent my car but  you see the stuff 
>above, and you realize that for example, people 
>would be concerned about the weight property and 
>so on.
>- The machine way could be for example that you 
>operate with an OWL system which include the 
>belief that foaf:Person and dot:Car are distinct 
>classes, which will flag an error.

That allows some inferences to be done which can 
make some distinctions, true. But it still 
doesn't get it actually attached to YOU.

>>   And what about a URI that I own and wish it 
>>to denote, say, the planet Venus, or my pet 
>>cat? What do I do, to attach the URI to my 
>>intended referent for it?
>>>>In my opinion to denote (or to identify) is a 
>>>>verb, something that is done by the users of 
>>>>a symbol. After all, symbols (URI) are not 
>>>>agents, they don't wake up and choose to 
>>>>denote this or that.
>>>They have wonders which create them for a specific purpose.
>>wonders/owners I presume.
>>But the key question to me is, how can they 
>>communicate the purpose to the reader? How does 
>>one publish an intention? How does one indicate 
>>a referent, other than by describing it? How 
>>much description is enough?
>Experience shows.  This is engineering.
>That question is I suppose one agonized over by working groups all the time.
>Fortunately, there are social systems not only 
>for announcing that a UTI has been minted and 
>describing what it denotes, but also for getting 
>feedback from people who don't understand it, or 
>whose machines are not able to process it. This 
>feedback can lead to an adjustment of the 
>information out there, publication of tutorials, 
>and so on.

I'm not getting my point across. I shouldn't have 
said 'how much is enough', which I think you took 
as a genuine question but I meant rhetorically 
(because its never enough.)

>>>>Nor do I think denotation is an attribute or 
>>>>property of a symbol, somehow built in or 
>>>>attached when the symbol is first conceived. 
>>>>It is more like a dance. I use a symbol to 
>>>>denote something expecting you to interpret 
>>>>it to denote the same thing. And this 
>>>>coordination, this synchrony of 
>>>>interpretation by both sender and receiver, 
>>>>is not always easy. It requires real effort 
>>>>to sustain it. The minter of a URI cannot 
>>>>make it happen by declaration, nor can a 
>>>>research group or a standards body just 
>>>>decree it so.
>>>In many cases, the URI is defined by 
>>>connection to already well-defined sets of 
>>>things.  In other cases, such as the terms in 
>>>the OWL ontology, there was a huge amount of 
>>>effort and discussion involved, and the 
>>>current term is supported by a lot of ongoing 
>>>tutorials and so on.  No one said it was easy. 
>>>But it is a different architecture from dance 
>>>associated with natural language words.
>>The huge effort was for defining the meanings 
>>of the OWL reserved vocabulary, and that is 
>>indeed about as defined as a vocabulary can 
>>get. But if we just take some OWL ontology and 
>>ask how well it constrains the meanings of the 
>>new URIs that occur in it, this is *very* much 
>>like the word dance that John refers to. In 
>>fact in all essentials it seems to me to be 
>>pretty much the same game. I publish (say) some 
>>stuff which I intend to mean something, you 
>>read (hear) it and get some insight into my 
>>meaning. You draw some conclusions and if you 
>>send (tell) them to me I can check to see if I 
>>agree and if not try to send (say) some more to 
>>you to make your grasp of my intentions more 
>Yes, of course the game happens.  But for 
>example when I download my OFX bank statements,
>then my computer processes them.  Millions of 
>people do this.  Very very few of them worry 
>about who
>did the dance, or do the dance themselves. 
>Actually, some people from banks, Microsoft and 
>Inituit etc did the dance.

Sure, in some cases it has all been worked out in 
enough detail and incorporated into existing 
protocols and so forth. I don't intend to deny 
that this can be, has been, done successfully. 
But take that example. Can this kind of thing be 
done for *all* referring names? Is there a 
*general*, *architectural* technique for 
attaching names to referents? Or do we need to 
wait for the astronomers to organize our ways of 
referring to things in the dark sky, and 
biologists to give us a biologists-agreed way of 
referring to cells and microtubules, and etc. for 
all the rest?

>  This is the "total cost of ontologies" argument.

I confess Im not familiar with that yet. I'll search for it.

>>It HAS to be similar, barring machine 
>>telepathy. There really isn't anything that 
>>either people or machines can do much beyond 
>>send symbols to one another: and symbols carry 
>>meaning only imperfectly, with inherent 
>Yes. In theory, there is always a very small 
>possibility that I completely misunderstand the 
>way you use a word, say "car", and a moderate 
>probability that in some small very rare corner 
>case we haven't come across yet, we in pedantic 
>mode would disagree about its meaning.

You are mocking my point as mere pedantry, and I 
agree it is pedantry; but its not mere. It is 
pedantry which is now built into the Web, because 
it arises from the way that SWeb languages have 
specified how they convey meaning, and the 
obscure kinds of ambiguity I'm pointing out are 
ones that arise immediately in actual ontological 
engineering practice. At this very time I'm 
having a parallel email argument with Barry Smith 
on another forum about the need (or not) to 
distinguish continuants from occurrents. This is 
pedantic, if you like, but its also engineering. 
Does your Tabulator output above tell me whether 
TimBL is a continuant or an occurrent? OK, 
neither you nor the bank needs to know: but the 
BOF foundational ontology (now being used in a 
lot of medical applications) does. Maybe at some 
point your life will depend on getting this right.

>The extent to which I can behave, and build 
>systems, as though in fact the terms had 
>well-defied meanings depends on the amount of 
>dancing which has been done.   As the dance is 
>done, the structures in my brain and those in 
>yours, while never identical, become very similar

An aside: that is often claimed or assumed as 
obvious, but there is lots of evidence against it.

>, specifically in that they develop an very 
>strong association between a class of real-world 
>object (car) or concept (TransitiveProperty).
>Philosophy studies the dance.  (Philosophy has 
>ion the past been distracted by a concern that 
>you can't measure the brain's structure, and you 
>can't measure an external TranstiveProperty 
>object.  The fact htat philsophers have found 
>this difficult doesn't stop the fact that it 
>happens, and now we have fMRI machines we are 
>even relieving some of this angst. But I digress)

Yes, you do; but so did I :-)

>The important thing is that as the dance is 
>done, the probability of major disagreement, and 
>the degree of pedantic disagreement,
>decrease very dramatically, to become negligible for engineering purposes.

This is where we disagree. A lot more dancing is 
needed for ontological engineering to work on the 
SWeb. Its not negligible, and it hasn't all been 
done yet. And as ontologies get used more and 
more, this extra dancing (to resolve previously 
invisible ambiguities) will need to be done by 
more and more systems and people in more and more 
contexts. Its already happening.

>>>It is different by design.  The semantic web 
>>>is an engineered system, not an observation of 
>>But it is a real system, and so it has to obey 
>>laws of nature. And there are laws of meaning 
>>as well as of physics.
>You bring up physics. Indeed.  As an engineer, I 
>design my coffee mug to stay in one place: with 
>no wheels. Does it?   It is a real system, and 
>so it has to obey laws of nature.  Physics tells 
>as that at any time, there is a finite 
>possibility that any object might just be 
>measured to be in a different place, such as a 
>foot to the right and no longer around my 
>coffee.  In fact if I study the construction of 
>a coffee mug from particles, all kinds of 
>questions arise, as the first few particles are 
>very difficult to manage.  I can't even build a 
>tool which will tell me where the stupid 
>electron is, without my being completely unaware 
>of its momentum.   Given that I can't actually 
>claim to have put an electron and proton 
>together with any degree of accuracy, how can I 
>claim to be able to build a mug which will stay 
>in one place.  Well, it turns out that as the 
>number of particles becomes larger, these 
>effects, while still true, just become ignorable 
>for engineering purposes.
>The analogy is limited, of course.

Yes, it is. I think I'll ignore it rather than demolish it.

>>>>The reason this matters is that since it 
>>>>requires this effort to create a 
>>>>denotation/identification in the first place, 
>>>>it is far more sensible, to me at least, to 
>>>>expect that the final disambiguation of a 
>>>>symbol be accomplished in the same way, by 
>>>>coordinated effort of the parties using the 
>>>>symbol, not by declaration of the W3C 
>>>>specifications that all URIs be absolutely 
>>>>This seems to me to be, as my grandfather used to say, a vain task.
>>>Your grandfather would perhaps have suggested 
>>>that an attempt to define the meaning of 
>>>common words, as the Académie Française is set 
>>>up to do were a 'vain task'.   Many would 
>>>agree.  But given that his water came to him 
>>>though pipes connected, possibly, by half-inch 
>>>British Standard pipe-thread connections, and 
>>>he rode on rails set a certain distance apart 
>>>by some committee, and his TV came for better 
>>>or worse in 525 or 625 lines as decided by 
>>>other committees, he may have respected that 
>>>the creation of standards is a very valuable 
>>>function, and an essential to progress.
>>>When people meet to define W3C specifications 
>>>they are not doing it out of vanity.   They 
>>>are performing coordinated effort of the 
>>>parties who would like to be able to use the 
>>>symbol.  They are, in general, users and 
>>>representatives of users of the symbol.   They 
>>>come together to allow those who follow them 
>>>to use it. They often work long hours, 
>>>receiving inadequate recognition for either 
>>>products shipped or papers published, the 
>>>conventional metrics of performance, so I 
>>>would not call it vanity.
>>The point surely is that URIs used to refer 
>>(not as in HTTP, but as in OWL) do *not* have a 
>>standardized meaning. Standards are certainly a 
>>chore to create, but they only go so far. OWL 
>>defines the meanings of the OWL namespace, but 
>>it does not define the meanings of the FOAF 
>>vocabulary, or the URIrefs used in, say, 
>>ontologies published by the NIH or by JPL.
>Ah, so you accept that the OWL working group has 
>effectively given meanings to the the terms in 
>the OWL namespace, but you don't think that the 
>FOAF friends gave meaning to the FOAF namespace, 
>or the NCBI to its ontologies? How curious.

Not at all. OWL has a (actually several, 
regrettably) semantics defined with mathematical 
precision, and corresponding operational systems 
related carefully to that semantics. The FOAF 
namespace is intended to refer things that have 
no mathematical description, and their meaning is 
some cases is not even loosely defined (what is a 
'friend', exactly?) Now, don't get me wrong, I'm 
not meaning to complain about this. It is fine 
for a vocabulary like OWL to have little more 
than the way its being used, as a specification 
of what its supposed to mean. In fact, I think 
this will happen more and more; in some cases, 
like dc:author, the "socially" defined meaning 
now takes precedence over the declared intentions 
of the authors of the spec. But one cannot then 
rationally claim that these URIs have no 
ambiguity in what they refer to. They are 
wonderfully, usefully, creatively ambiguous. That 
is one reason why they are so useful, because 
their meaning has NOT been tied down too tightly.

>  OWL does indeed only go so far.  
>FOAF and OFX and so on  go  bit further, but 
>also  only so far.  They all create new terms 
>which, modulo arguments afterward an so on, 
>become with time an increasingly stable 
>foundation for communication between parties.

Of course. I have never denied this. What I have 
been saying is that it is a mistake to think (and 
still more to authoritatively assert) that this 
stability of foundation for communication depends 
on, or arises from, the terms being unambiguous 
referents or identifiers. On the contrary, it 
arises precisely from the fact that they are 
often treated exactly like newly coined English 
words, rife with ambiguity and which slowly 
crystallize to a clearer (but still not unique or 
even necessarily fixed) meaning from the ways 
they are used (recent examples of this in English 
include "hacker" which in my youth was a term of 
respect, now having criminal associations, and 
more recently "wanker" which in US English has 
been stripped of its obscene UK-English 
connotation, and can be used safely in a variety 
of social settings.)

>>The only way those meanings can be specified is 
>>by writing ontologies: and finite ontologies do 
>>not - cannot possibly - nail down referents 
>>*uniquely*. No amount   of authority-delegating 
>>or standard-setting is going to change this 
>>basic fact.
>Fortunately, the trains continue to run, and 
>bank statements continue to generate graphs and 
>tax forms, more and more people say they 
>foaf:knows each other, despite the fact that 
>none of them or the relationships have been 
>defined really completely precisely to the 
>complete mutual understanding of all parties.

Of course. I don't deny this. But what this 
illustrates is precisely what Ive been saying, 
which is that terms do not need to be defined 
tightly, still less defined so tightly that they 
must be considered to be uniquely specific 
identifiers, in order to be used successfully in 

>This seems to me to be a very important bridge 
>between philosophy of microscopic linguistic 
>interactions of agents, and the large scale 
>world of communities with common terms which are 
>'good enough for government work'.  I suppose I 
>am surprised it isn't in the textbooks.

Im sure it is, somewhere. To be fair to 
linguists, many of them use a robust notion of 
"word meaning" which corresponds to your notion, 
in which for example "bank" has three meanings. 
But if you ask a computational linguist, they 
will probably tell you that "bank" has more like 
ten or fifteen distinguishable senses. And if you 
talk to someone who actually writes ontologies - 
you know, an ontology *engineer* - then the 
number of distinct things that could all be 
rendered by an English word like "bank" or 
(famous Cyc example) "cover", then you get 
answers in the dozens. This is mere pedantry for 
some people, but tough engineering practice for 
others. If URIs have to identify uniquely and 
unambiguously, and if URIs are universal and 
eternal, which of them gets to specify how finely 
divided those URI meanings have to be sliced and 


IHMC		(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32502			(850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
Received on Wednesday, 13 June 2007 18:56:49 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:41:57 UTC