RE: Documents, Cars, Hills, and Valleys

>It's not my position that all URIs need to be disambiguated. My position
>is that URIs only have a referent in context, and that sometimes the
>context is sufficient to eliminate ambiguity, and sometimes it isn't.
>When it isn't it makes sense to add more context.

I agree. If locator-rooted URIs aren't regarded as ambiguous without
contextual qualification, then the SW won't be interoperable with the
existing web. There is also the issue of making people change their habits
and comply with whatever the unambiguous definition happens to be for this
Looking Glass Web.

Ok, if I'm talking about the size of http://www.microsoft.com am I talking
about the number of employees or the number of characters on their home
page? If I was using a web metrics schema then probably it's the former, if
I'm using a commerce schema then it's probably the latter, but it wouldn't
be ambiguous because the type (in a semantic sense) of the property would be
defined in the schema from which the property comes.

This may seem like more work for the agents, but going the other way, if the
URI only referred (e.g.) to the web site, then ok, if I want to make
assertions about the site then I can use the URI. But if I want to refer to
the company, then I have to go up a bnode. I haven't played with this, but I
can imagine that to make a lot of statements it might be necessary to chain
back a whole load of bnodes, whereas resolving the ambiguity in the
namespace of the property doesn't need any.

Incidentally, the Semantic Web (as identified by
http://www.scientificamerican.com/2001/0501issue/0501berners-lee.html) is
now pay-to-view.

Cheers,
Danny.

Received on Wednesday, 24 April 2002 06:01:38 UTC