Lists Home |
Date Index |
Mike Champion wrote:
> >If you did _that_, then people might actually find out about your
> >_quickly_ ... alternatively you could publish your discovery in some old
> >fashioned print only journal, and 2 years after the 3 rock experts in the
> >world had reviewed your description, it would be published in "Rocks" the
> >journal ... in the meantime someone else, had published an HTTP URI and
> >_their_ description has become the definative description of this new
> ... and in the meantime your ISP has gone bust and the megacorp that
> their assets uncaringly invalidated your old URL ... or some other
> sued you for "squatting" on their "rocks-r-us.com" domain so you no
> longer own the root of your URI space ... :~)
> or one of the other things that make URLs less permanent than they
> could/should be happen .
Let's not throw out the baby with the bathwater ... is all this pushback on
HTTP and DNS some kind of revenge of the true hypermedia gods which has been
festering since 1991?
> I fully agree that abstract "resources" and "namespaces" should have an
> authoritative description on the web. It just seems a conceptual muddle
> (as the 2+ years of this unresolved discussion prove, IMHO) to equate
> the permanent name of the abstract resource with something as
> ephemeral in practice as an HTTP URL.
Add a timestamp to the HTTP URI. Can't you already get a snapshot of the Web
at any point in time somewhere online?
> I'm still not seeing why
> the URL of the description can't be a property of the resource in some
A literal piece of text or XML can be the object of an RDF statement, but
that isn't the issue.
a) How can I share ontologies on the Web?
b) How can I merge ontologies?
These are a couple of the real world issues surrounding ontologies that,
well, HTTP is sort of helpful in solving. OWL, the successor to DAML, is not
just another ontology definition language, OWL stands for "Ontology (for
the) Web Language" Its key characteristic, beyond the work already done on
description logics, and all sorts of other ontology languages, is that
ontologies can be shared on the Web, merged together, and extended. HTTP
enables that, and makes it difficult at the same time but without HTTP I
can't imagine doing it without creating something like HTTP.
> It doesn't solve the bootstrapping problem, but reliably
> resolving URNs seems like a solveable technical problem whereas the
> "HTTP URL == identity" equation inevitably leads to the circularity
> we've been wrestling with: how does one know whether the URI identifies
> the abstraction or just the location of a representation of it?
> Your original post clearly outlines the dilemma this leaves us with.
The URI always identifies the resource, never the location of the
representation of the resource. That's how you know, you read the RFCs.
Honestly. The "Semantic Web" as it stands today is not concerned with making
statements about representations of resources or locations of
representations of resources, in the "Semantic Web", or more specifically
using RDF one makes statements about resources. Always. Read the REC.
> I thought more about the option of retreating to a mountaintop or
> bunker to await Enlightenment rather than choosing either horn
> of this dilemma.
Seriously, this isn't a mystical question subject to interpretation. In RDF
you always make statements about resources. It is only an issue when you try
to mix concepts from the current Web and the Semantic Web -- and when you
fail to distinguish between the representation that is returned on GETing a
URI, and the resource that is identified by the URI. Now if you chuck HTTP
URIs and use only URNs, then there is no issue because you can't GET a URN,
but then you have no Web as we know it. And if you provide software to GET a
URN then you have _exactly the same issue as when you GET a HTTP URI_. In
any case RDF doesn't concern itself with representations so it could care
less what scheme a URI starts with.