RE: Semantic Web Hackings

[Jason Diamond <jason@injektilo.org>]

> > Where are you going to put the RDF document [...]
>
> On the Semantic Web.

You're kidding, right?

> > If, however, you're hosting it in your own web space, how is the rest
> > of the world supposed to find it? They not only need to know your
> > URI (mailto:sean@mysterylights.com) but also the URL where they
> > can retrieve the RDF document that describes your URI.
>
> That's not the function that this is performing. We are not asking for
> information about mailto:sean@mysterylights.com directly, we are simply
> querying it on the Semantic Web and then fining a piece of RDF that
> describes that email address, which is a URI representing me.

Just how exactly does this query take place? A distributed system cannot
function without some sort of transport protocol. I'd even go so far to say
that it couldn't reasonably scale without some sort of distributed naming
system a la DNS. (Aren't you glad that you no longer have to download
hosts.txt?)

I'm a programmer. I don't care about the meaning of mozzarella or even how
fast Len gets his pizza delivered. If I'm going to implement something, it
has to be implementable. In this case, all I need is a function that takes
in a mailto URI and gives me back an http URI so that I can execute this
magic query you keep talking about.

For example, messages aren't magically routed to their destination just by
tacking a To: sean@mysterylights.com header at the top of it. The DNS
servers for mysterylights.com need to be queried for the MX RRs which are
then used to determine what host an SMTP connection needs to be made to.

So unless you plan on extending DNS, we have a problem.

> > Somehow, I'd need to let my engine know where to find the
> > description for your URI and all of my other friends. Do we
> > reify those statements and say that the objects' descriptions
> > could be found at the appropriate URLs? Yuck.
>
> What's wrong with pointing someone to a URL? Here is some
> information about
> me: <my URL/> That's how stuff works: if you want to find out something
> about me, look it up.

That doesn't look like RDF to me. How am I supposed to parse it? And how
would I use it within my RDF?

> I we're both missing something in each others conversations here; I'm not
> sure why you are refering to URLs as if they are unrecoverable, and maybe
> you can't see that I'm just using a mailto: as a name, not an address.

I understand that mailto is just a name. I just think that it's worthless in
the context of a distributed system without any method of resolving it.

Jason.