RE: Spamming a URI

Yes, suppose you have Jim and Sally; you could have the following
situations:

1. A statement is attributed to Sally, unverified.  (Someone says that
Sally says 'x')
2. A statement is attributed to Sally and is verified by Sally.  (Sally
says 'x')
3. A statement is attributed to Jim, saying that Sally says 'x',
unverified.  (Someone says that Jim says that Sally says 'x')
4. A statement is attributed to Jim, saying that Sally says 'x',
verified by Jim.  (Jim says that Sally says 'x') 

Your level of confidence in each of these statements would vary based on
what you think of the people involved.  Obviously in statements 1 and 3,
you would not trust the statement much without additional assurance.

Note that exactly the same situation happens on the WWW.  For example,
if you access a web page, and the web page has an SSL certificate
registered to microsoft.com, you can probably assume that the page was
made by someone at Microsoft.  If that page made claims about IBMs
products, you would have to decide wether or not to trust Microsoft's
word about IBM products in that particular instance.

Of course, this is a very simple example.  For starters, the "verified
by Sally" does not just mean "SSL certificate".  There are many types of
assurance -- some people might trust something based on the source
domain name, others might require a PGP signature, and so on.  The level
of assurance you want depends on the nature of the data.  Also, you can
imagine scenarios where the trust is much more transitive.  For example,
"friends of friends recommend this restaurant", or "people who bought
this book also bought...".  The ebay and epinions rating systems, the
thawte "web of trust" are examples of transitive trust systems.  Just as
there is not a single level of assurance that works for all data, there
is not a single degree of transitive-ness that is applicable in all
situations.



> -----Original Message-----
> From: Lisa Seeman [mailto:lisa@ubaccess.com] 
> Sent: Wednesday, September 22, 2004 7:02 AM
> To: Joshua Allen
> Cc: www-rdf-interest@w3.org
> Subject: RE: Spamming a URI
> 
> What about a trusted source that logs RDF trust levels of RDF 
> found at different servers/ authors etc.
> 
> 
> Lisa
> 
> > -----Original Message-----
> > From: www-rdf-interest-request@w3.org 
> > [mailto:www-rdf-interest-request@w3.org] On Behalf Of Joshua Allen
> > Sent: Wednesday, September 22, 2004 6:39 AM
> > To: Matt Halstead; www-rdf-interest@w3.org
> > Subject: RE: Spamming a URI
> > 
> > 
> > 
> > The semantic web is really no different than the regular 
> WWW in this 
> > respect.  We regard it as a virtue that *anyone* can publish a web 
> > page, even if the web page contains lies.  We also regard it as a 
> > virtue that anyone on the web can link to any other page, 
> even if the 
> > owner of the target page does not know or support the 
> person doing the 
> > linking.
> > 
> > You assess trust by considering the source, checking for digital 
> > signatures, etc.  One way to add assurances to RDF:
> http://xmlns.com/wot/0.1/
> 
> 
> > -----Original Message-----
> > From: www-rdf-interest-request@w3.org
> > [mailto:www-rdf-interest-request@w3.org] On Behalf Of Matt Halstead
> > Sent: Tuesday, September 21, 2004 7:28 PM
> > To: www-rdf-interest@w3.org
> > Subject: Spamming a URI
> > 
> > 
> > I realize there is 'trust' in the semantic web cake[1], but I
> > am intrigued to understand how this is envisaged to work at 
> > even a simple RDF level.  If we have something as simple and 
> > useful as a semantic web crawler, e.g. swoogle [2], then how 
> > do we ignore the work of spammers which inappropriately 
> > attribute properties and values to, or reference in any way, 
> > a particular resource URI?
> > 
> > 
> > [1] http://www.w3.org/2004/Talks/0412-RDF-functions/slide4-0.html
> > [2] http://pear.cs.umbc.edu/swoogle/index.php
> > 
> > cheers
> > Matt
> > 
> > 
> > 
> 
> 
> 
>

Received on Wednesday, 22 September 2004 18:05:43 UTC