URIs in RDF (Was: New Intro to RDF)

Tim, while I seldom contribute to semantic-web@w3, I feel the need to
respond to your concerns below.

On Tue, 2005-10-04 at 11:02 +0100, tim.glover@bt.com wrote:
> Secondly, I would like to make a comment about the use of URLs as unique
> identifiers. This idea is central to RDF and is always included in a
> discussion of its benefits, but this has been an area of some
> controversy, and I think counter arguments deserve a mention. Some
> counter arguments are
> 
> 1. Names only have to be unique IN CONTEXT. For example, I can write a
> program using variable x without any danger of interfering with your
> program, also containing variable x. And in natural language, the phrase
> "I am going to try to catch the plane" has a different meaning in the
> context of an airport and the context of a woodwork shop, but there is
> no difficulty in using the same word for two different things because
> the meaning is clear from the context. 

The web does give you names unique in context - that would be relative
URIs. Additionally RDF and XML use namespaces an QNames for a similar
purpose. So the human users may see short names and, if necessary, go
see what the context for that name is (its namespace or base URI); and
the machines will always fold the context into the name, making it a
global name. So I think we have both usable ambiguity for people and
uniqueness without ambiguity for machines.

> 2. Using URLS does not guarantee uniqueness. Many people may choose to
> use the same URL to mean different things. 

Only the creator of the URI has the authority to choose the meaning of
it. On the other hand, if that meaning disagrees with what many people
perceive the meaning to be, it will probably lead to the change of the
"official" meaning or just change of the URI.

So I cannot choose to use http://w3.org/ to mean my car, and I don't
think anything on the web gives anybody the opposite idea.

> 3. By avoiding the problem of using the same word for different things,
> you multiply the problem of using different words for the same thing,
> and this problem is probably more difficult to resolve. 

owl:sameAs means to solve exactly this problem. IMO ambiguity is more
difficult to resolve than multiple names for one thing.

> 4. Using URLs makes many people believe that the URL is an address of
> some useful information, and this is not the case. They are just names.
> The URLS can be completely fictitious. More dangerously, the content of
> a URL can change over time. 

While the URIs can be completely fictitious, people on the web already
know not to assign http://weather.example.com/Chicago to identify some
species of penguins. So you are afraid of a problem that shouldn't be
very significant. It's like saying roads should always have a fence in
the middle because otherwise cars in opposite directions will keep
colliding, but they don't because the drivers know better.

And if the content of a URI changes over time, that doesn't mean the
meaning has changed, like the www.news.com example presented by somebody
else. And sometimes the meaning does indeed change, but often enough the
"official" meaning is changed to be closer to what the public perceived
the meaning to be in the first place.

> 5. URLS make RDF difficult for humans to read and understand. The
> problem is compounded by that fact that prefixes can be used in some
> places but not in others. 

What you're saying here is that RDF/XML is a mess, and you'll find few
people disagreeing with you. Note how n3 is much more consistent on this
account.

> In other words, URLS are not a silver bullet, and I think the advantages
> should not be overstated. 

I agree, but URIs should not be underappreciated either. URIs are
definitely not a silver bullet, but they seem to be tremendously useful
in Web-related applications.

Best regards,

Jacek

Received on Saturday, 8 October 2005 15:19:48 UTC