Re: What is an RDF Query?

> > >we could just use assertions for everything:
> > >       "Ralph is in his office."
> > >       "I want to know whether Ralph is in his office."
> >
> > The trouble with this is that it invites the response: "So? Why
> > should I care what you want?"  which would be impolite in human
> > discourse, of course, but only because humans have a mutually agreed
> > set of social conventions about being helpful to one another, etc. .
> > Without some such conventions, nothing particular follows from your
> > needing something.
>Perhaps I should have phrased it a little better, as
>     "Tell me whether Ralph is in his office."
>which puts us where we are with current client/server systems, which
>pretty-much do their best to follow any instructions they receive
>(pending some authorization procedures we could discuss).
>I'd love to go a bit farther, to things like
>     "If you tell me whether Ralph is in his office within the
>     next 3 seconds, and your information turns out to be correct,
>     I'll transfer $0.10 to an account of your chosing."
>but that's probably more researchy than we should be talking about
>right now.


But there's no way to say 'tell me....' in RDF.

> > What we really need here is a set of agreed protocols for asking,
> > asserting, querying, whatever, which are related in systematic ways
> > to the contents of what are asserted, queried, etc.
>Sure.  The question I think we're debating in this thread is simply
>whether those protocols should use an RDF syntax at the bottom-layer
>or not.  We could define a query protocol with an S-Expression syntax,
>an ASN.1 syntax, a internet "simple" (a la SMTP) style syntax, etc.  I
>suggest we ignore the syntax and simply say we're using an RDF
>assertional graph (knowledge base) to convey the query (and its response).

Wait. The RDF graph can convey the *content* of the query, but it 
does not express the difference between *asserting* that content and 
*querying* it. That's what I meant by 'protocols' above.

<unofficial comment>The current RDF core WG consensus seems to be 
that RDF (in its present incarnation, anyway) is strictly an 
assertional language. </unofficial comment> The point of my little 
essay (with the joke in it) was to draw attention to this distinction 
and point out that RDF *isnt* being used for querying; it *could* be, 
but if it were, then it shouldn't be called RDF, because the same 
graph really would mean something different when used as a query from 
when used as an assertion. The English sentences "Joe is fat" and "Is 
Joe fat?" are closely related in meaning, but they aren't the same 
sentence. Writing a piece of rdf with <rqf>... </rqf> around it would 
be like adding a question mark to an English assertional sentence to 
make it into a query.

>Someone else (us on another mailing list, perhaps) can decide how the
>RDF assertions get securely passed from one agent to another.  Here,
>we can just talk about what kind of RDF statements "Agent A" would
>make to "Agent B" to get "Agent B" to do some useful query work and
>send some results back to "Agent A".

Answer: none. RDF doesn't convey that kind of thing. It just says: 
this is true, make of it what you will. Nothing of the form 
<rdf>....</rdf> can possibly, right now, request any kind of response 
from the reader.

> > >But as to existential variables: yes, if we can't get existential
> > >variables in the basic assertion language,
> >
> > Well, RDF has them, so relax :-)
>Thank you.   <big><sigh of relief /> :-) </big>

I said in the *assertion* language, mind. <;-/>


(650)859 6569 w
(650)494 3973 h (until September)

Received on Tuesday, 11 September 2001 16:48:22 UTC