W3C home > Mailing lists > Public > www-tag@w3.org > September 2004

RE: Information resources?

From: <Patrick.Stickler@nokia.com>
Date: Sat, 11 Sep 2004 08:33:11 +0300
Message-ID: <1E4A0AC134884349A21955574A90A7A50A1CE2@trebe051.ntc.nokia.com>
To: <joshuaa@microsoft.com>, <Norman.Walsh@sun.com>, <www-tag@w3.org>

> -----Original Message-----
> From: ext Joshua Allen [mailto:joshuaa@microsoft.com]
> Sent: 10 September, 2004 20:52
> To: Stickler Patrick (Nokia-TP-MSW/Tampere); Norman.Walsh@sun.com;
> www-tag@w3.org
> Subject: RE: Information resources?
> > methodologies which help humans manage the URI space due to 
> > mnemmonic structure.
> Nobody is suggesting such a thing.  My point would be just as valid if
> we were talking about words instead of URIs (which are just 
> the words of
> the SW).  I am not suggesting a mnemonic device, I am simply 
> saying that
> people should be very precise when speaking.  If we left URIs entirely
> out of the picture, I would say:
> "Hoary_Marmot" when you want to talk about the animal Hoary_Marmot
> "The word Hoary_Marmot" when you want to talk about the word
> "A picture of a Hoary_Marmot" when you want to talk about a
> representation
> And that *unless* you disambiguate between these three, your 
> words have
> no meaning at all (or at least you should not complain when people
> assume that you mean the 1st sense).

I agree. And that's what the semantic web is for. To actually
say what the various URIs *mean* -- not to simply provide access
to representations associated with whatever that "thing" is
denoted by the URI in question.

My example was to show how one could mint a set of URIs which
could be used to denote the kinds of resources that have been
within the scope of this discussion, in a way that allows for
all of those resources to have web accessible representations,
exploiting the existing web machinery, yet still allowing 
consistent and coherent statements to be made at the semantic
web layer about those resources.

The mnemonic qualities of the URIs in question are not essential.
Some of the syntactic qualities of the URIs in question are
essential for exploiting the rewriting/redirection functionality
provided by existing web servers to serve the representations.

> > But insofar as the archtecture is concerned, URIs are opaque, 
> > and any valid URI can be used to denote any resource. Period. 
> Yes, I agree.  But that's not a very meaningful statement. 

Eh? I consider it to be one of the cornerstones of the web architecture
(and also point out that this cornerstone is explicitly presented in
 AWWW). It is an *exceptionally* meaningful statement and one that should
be embraced, understood, and respected in all other aspects of web
archtecture and design.
> That is like
> saying, "any word can be used to denote any thing, period."  
> It is true
> that URIs are opaque, just like words are opaque, but this is 
> not nearly
> as relevant to this discussion as people seem to think.  Leaping from
> that true statement, to "I should be able to use a URI to refer
> simultaneously to a word, a thing, and a representation of 
> the thing" is
> just muddy-brained thinking.

I said nothing of the sort. I never suggested that the same URI
would be used to denote more than one resource. That is IMO
a cardinal sin on both the web and (especially) semantic web

> And to the contrary, people *should* feel that they should not use an
> http: URI to conflate semantics like this.  

You've misunderstood me.

> Just as it is possible for
> Michael Jackson to say "bad" when he means "good", there is nothing
> preventing people from attaching any wacky semantics they 
> want to a word
> (or URI).  But people should be discouraged from doing this.  It is
> antisocial.

Firstly, URIs are not words in a natural language. Secondly, there
are more than one language. There have be numerous threads discussing
this issue. Previous common examples are: what does "chat" or "pan"
mean? Well, it depends on who is using them and what language they
are speaking, etc.

URIs are not supposed to suffer from such multi-language overlap. Each
URI is supposed to mean one and only one thing (to name one and only
one resource). But the lexical nature of the URI is irrelevant, and
the minter of the URI is free to introduce linguistic properties specific
to their perspective into the URI to help them recognize and manage
that URI -- yet no'one should presume anything about a URI based on
its lexical nature.

If Michael Jackson were to use some URI http://example.com/bad to
denote the concept of "goodness" then that is his business and you would
be wrong to presume anything about the meaning of that URI based on
the fact that it contains the character sequence 'bad'.

And there's nothing necessarily "antisocial" about such a URI.

> > http://example.com/someDog               physical resource (dog)
> > http://example.com/someDog/index         abstract resource 
> (web page)
> > http://example.com/someDog/index.html    representation
> > http://example.com/someDog/index.txt     representation
> > http://example.com/someDog/index.jpg     representation
> Now you are confusing your own self by ignoring your own 
> insistence that
> URIs are opaque.  

Not at all. And I clarify that later in the posting.

> In fact, even to the web architecture, there is
> nothing stopping http://example.com/someDog/index.txt from returning a
> JPEG. 

True. Read on. I'm presenting a methodology, not a model, where
the lexical qualities of the URIs facilitiate *both* resolution
of URIs to representations as well as (limited) human understanding
of the meaning of those URIs (not that the latter is guarunteed
or licensed by the web architecture -- but still possible within
a community using a common set of known URIs).

>  To HTTP, all of those URIs are completely opaque, and 
> could very
> easily all return the same thing.  Consider hypothetically 
> that they are
> all mapped to the same web page and all return the same bits. 
>  What have
> you proved?  Even if you have a single unified URI, you are 
> still unable
> to disambiguate between the thing, it's representation, and it's
> representation dispenser.
> I would argue that all you have been successful in 
> demonstrating is that
> http: URIs in practice denote representation dispensers.

Insofar as the web layer is concerned, I wouldn't strongly disagree
with you there -- though I would present it somewhat differently.

It is quite true that all the web machinery can provide are representations,
and one cannot reliably or definitively obtain information *about* the
resources denoted by the URIs. Humans may be able to guess some things
correctly about those resources by analyzing the representations obtained,
but they can *never* be completely sure, insofar as the web machinery
is concerned. Interpretation of representations is completely outside
the scope of the web.

That is where the semantic web enters the picture, to help folks know
what those URIs actually denote and what the nature of those denoted
resources actually is.


Received on Saturday, 11 September 2004 05:33:32 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:28 GMT