[
Lists Home |
Date Index |
Thread Index
]
Uche Ogbuji wrote:
>
> > Uche Ogbuji <uche.ogbuji@fourthought.com> wrote:
> > |[Paul Prescod:]
> >
> > |> RDF exists to solve a problem: associating metadata with web objects.
> > |> You can't do that without a strong notion of identity.
> > |
> > | So you mean the identity that comes with a string match of URIs?
> >
> > I can't speak for Paul, but I doubt that is what he meant. I'd say a
> > closer example would be the way in which ID attributes establish identity
> > in a document (i.e. "distinct existence").
>
> Hmm. This is not what I'd call a "strong notion of identity". The same can
> be said of any XML vocab that uses attributes of type ID. I'd be quite
> surprised if this is what I'm to understand here.
I'm sorry I caused so much confusion in my absence. ;)
People use RDF to make assertions about things called resources. It is
important to understand which assertions are definitely talking about
the same resource, could be talking about the same resource or could not
be talking about the same resource (with the default being "could be").
Resources are identified by identifiers known as URIs.
If your assertion framework is as simple as a single hard-coded tag that
says: "The title of that resource is X" you will run into the issue of
determining identity (what exactly are you giving a title to?). Anything
that deals with identity needs to figure this out (e.g. caching servers,
Google, ... ). Identity is especially tricky in a system with mutable
objects and caching. Another thing that complicates identity is the
issue of whether assertions can be made about things which are outside
of the computer. (i.e. asserting the title of a book, not a web page)
Sure, figuring these things out is painful for those watching RDF. But
comparing it to XML is not, IMHO, informative. RDF is solving a problem
that resolves around objects with identity. XML is solving the issue of
how to transport labelled trees. Simon asserts that some people think in
the one way and others in the other way. I would instead say that some
people have chosen to solve problems that require notions of distributed
identity and others have chosen to solve problems that have different
requirements.
Nevertheless, namespaces or not, XML would not have escaped the issue of
identity for long. Given an XML document it is necessary to figure out
what it means, often without any other context than the document itself
-- in other words element type identity. SGML used public identifiers.
XML uses namespaces. In either case you have nasty issues about whether
one identifier should survive for several versions, how to get from the
identifier to a schema, whether to use the identifier as a pointer to
something, etc.
That doesn't mean I think that namespaces are great. It means that I
think that "patterns" versus "identifiers" is a false dichotomy.
--
Paul Prescod
|
