Re: Self-descriptive assertions

On Mon, 2004-03-22 at 14:07, Mark Baker wrote:
> Hi Dan,
> 
> On Mon, Mar 22, 2004 at 01:01:17PM -0600, Dan Connolly wrote:
> > On Mon, 2004-03-22 at 12:30, Mark Baker wrote:
> > > Hi,
> > > 
> > > I've got a simple requirement I need met, driven largely by my desire to
> > > build RESTful Semantic Web apps.
> > 
> > Could you be more specific? What sort of app has this requirement?
> 
> It's just your run-of-the-mill SemWeb app.

I haven't seen enough semweb apps to know what's run-of-the-mill.

>   We've just chosen to use
> the REST architectural style to develop it because we think it's a great
> guide.  So given the option of a self-descriptive solution to a problem,
> versus a non-self-descriptive one, we'll almost always opt for the
> former.  Hence my attempt to find a self-descriptive solution to this
> problem.

I'd like to understand the problem better.

> > If you GET some text that says
> > 
> > 	The world is flat.
> > 
> > You can sorta tell from common sense that it's not asserted...
> > unless it was written a _long_ time ago. But if you find
> > 
> >  	Four out of five dentists recommend trident
> > 	for their patients who chew gum.
> > 
> > You don't really know if it's in jest or serious without poking around.
> 
> Of course.  You'll get no argument from me that this ambiguity isn't
> omnipresent for humans.  But I believe it's problematic for automata.

And so... ?

> > Similarly, most RDF data that published in a way that the
> > author expects others to believe enough to act on has
> > dc:creator link or some such.
> 
> I don't think "believe enough to act on" has anything to do with the
> issue I'm raising.  Whether some agent wants to run out and buy SUNW
> based on my prediction that it'll reach $200 next month is none of my
> business as an RDF publisher.

Most likely, it is. The publishers/authors who are in the business of
making stock predictions gain and lose trust of the consumers
of that info depending on how well their info pans out.

Even if you're just making a prediction in a livejournal entry
or something, your peers are likely to change their estimation
of your reliability depending on whether the info you give turns
out to be right. And if you want to say "suppose SUNW
went up to $200" rather than "I predict SUNW will go up to $200",
you should either represent that in RDF itself or, as sandro
pointed out, make it part of the link metadata:

  You need me, as I hand you a URI, to say "this is true"
   -- sandro Mon, 22 Mar 2004 13:51:06 -0500


>   I'm just interested in communicating that
> I *made* such a prediction, versus, for example, just stating a
> prediction of somebody else, or serving up random triples.

I still don't understand enough about your application to know
why you need that.

>   AFAICT,
> currently, those two states are indistinguishable from a messaging POV
> (i.e. communicating both via HTTP GET & RDF uses the same message).
>
> If there's another way I should be handling stuff like that, I'm all
> ears.  I don't see that dc:creator is anything special, nor do I feel
> that any type of predicate should be special, because it itself may
> turn out to be a non-asserted predicate.

Such is life. Why do you want to constrain RDF to be more
self-descriptive than any other representation of information?
It's nice when a web page has links to provenance info to
show why you should believe it, but nothing in the HTML
spec says you must do that.

dc:creator is a/the way people connect RDF graphs to people
that assert them (or... at least: wrote them). You could try
specializing dc:creator to baker:asserter and let us know how
that works in your app.

The foaf community, among others, uses rdfs:seeAlso to
say "if you believe this document, you're probably interested
in that document over there too" and such.

The OWL spec gives owl:imports as a mechanism to say
"that document is asserted just as much as this one is".

And so on.


> > Er... you somehow want to get a widespread understanding that
> > some RDF is asserted, but you want to short-circuit the process
> > of getting widespread agreement. I don't see how to do that.
> 
> I only want to short-cut the mechanism, not the process, by declaring
> that all application/rdf+xml-described documents are asserting their
> graphs.

Hmm... ok, I've seen you declare that. I haven't seen rationale
that convinces me it's a good idea. And so the process continues
until consensus (or at least a critical mass of agreement) emerges...

>   That leaves the door open for other media types to be used to
> do things differently in the future.
> 
> Mark.
-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
see you at the WWW2004 in NY 17-22 May?

Received on Tuesday, 23 March 2004 12:13:42 UTC