RE: Qustions about UC&R 1.114

-------- Original Message --------
> From: Kendall Clark <mailto:kendall@monkeyfist.com>
> Date: 21 June 2004 16:18
> 
> On Mon, Jun 21, 2004 at 03:54:48PM +0100, Seaborne, Andy wrote:
> > 
> > Kendall,
> > 
> > 1/ 2.10 "Discovering Learning Resources"
> > 
> > As I read 2.10 "Discovering Learning Resources", the professor
> > submits a query to the web site, that is, as far as the prof is
> > concerned, there is one destination for the query.  The target is an
> > aggregation, hence the provenance information returned in the results.
> > 
> > In what way does this motivate "Aggregate Query" where there is
> > multiple named targets for the query?  The value seems to be in the
> > fact that the recommended web site has done the aggregation, removing
> > the need for the prof to know about several different places.
> 
> Andy,
> 
> Good questions. I hate to do the cheap thing and say "ask Dave" but,
> well, you should ask Dave. In other words, I'm not really sure *who*
> does the aggregate query in this UC. For my 2 cents, though, it
> doesn't have to be the professor's agent that does it in order for it
> to motivate aggregate query. If the other sites do it, that still
> motivates IMO.
> 
> > 2/ 2.11 "Finding Out New Things About People"
> > 
> > I didn't see the connection to motivate provenance. I can't see where
> > the crawler needs to know where any statements came from.
> 
> Crawlers *always* need to know where statements come from, don't they?

Not so clear-cut - the client may trust the aggregator;  The aggregator may
not want to reveal where the info came from.  I agree that often
aggregations are more useful if they do have provenance but it is not
automatic.  Web portals don't usually do this today - the value is in the
edits process of getting things onto the portal and trusting the editors.
Makes the content more valuable.

> Seriously, tracking provenance of statements while crawling is like
> the canonical example that motivates people to add quads or some other
> context tracking mechanism. I can think of examples of this from the
> spider that Krech wrote for  rdflib, to Matt Biddulph's FOAF crawling
> work, to others.

But in the this UC, the crawler doesn't do anything with the source for the
statement.  The UC is, "find new property", "look it up".  To do this, it
doesn't need provenance as faras I can see.

	Andy

> 
> > PS The CVS date isn't changing - its still June 7.  Timewrap?
> 
> Interesting. I didn't notice that. Maybe ericp can ask the CVS mavens?
> 
> Kendall

Received on Monday, 21 June 2004 11:31:20 UTC