RE: Some thoughts on effective access to "primary" vs "secondary" resources, consistency of descriptions, and bootstrapping the semantic web...



> -----Original Message-----
> From: ext Phil Tetlow [mailto:philip.tetlow@uk.ibm.com]
> Sent: 08 October, 2004 16:47
> To: Stickler Patrick (Nokia-TP-MSW/Tampere)
> Cc: public-swbp-wg@w3.org
> Subject: Re: Some thoughts on effective access to "primary" vs
> "secondary" resources, consistency of descriptions, and bootstrapping
> the semantic web...
> 
> 
> 
> 
> 
> 
> Patrick
> 
> FYI, I am now a member of the SWBP Working Group and, as you 
> will be aware,
> also share your opinions on the importance of bootstrapping 
> mechanisms for
> the Semantic Web. In particular I agree that the issue of 
> authority is one
> that should be investigated from the point of best practice.
> 
> Obviously there is much debate to be had here - and I’m aware that raw
> personal opinion is a much sharper, but far less effective, tool than
> collaborative study - but personally I am not sure that even 
> the authority
> specified in a URI itself can be taken as a valid source of ‘boot’ or
> ‘self-description information.

I expect that there would be several levels of trust. One
basic level would rely simply on the integrity of web authorities,
which probably will suffice for many applications.

Beyond that, signed graphs with all the necessary certificate
and authentication machinery can be applied, to both corroborate
and strengthen web authority.

I.e. one application may be satisfied with having accessed a
given graph from the same web authority of a given URI.

Another application may, in addition, require that the graph
be signed and that the authority in some way be both verified
and in agreement with that web authority, and the authority
recognized in some other way as "trusted".

Discrepancies of any kind will likely result in the graph being
rejected.

I don't see that a fundamental component of the general knowledge
interchange machinery such as URIQA should be expected to provide 
the full trust solution, no more so than that fundamental web
machinery.

See my comments about a related issue of context per

http://lists.w3.org/Archives/Public/www-rdf-interest/2004Oct/0103.html


Just as context issues should ideally be addressed by making
explicit RDF statements, so trust issues can and IMO should be 
addressed in like manner -- and URIQA will be a very effective
tool for the interchange of such statements. E.g. see the
discussion and examples regarding qualification of statements per

http://lists.w3.org/Archives/Public/www-rdf-interest/2004Oct/0105.html


Bootstrapping is all about getting to the point when statements
pertaining to issues such as trust, context, etc. can be evaluated
and the knowledge accepted or rejected accordingly.

But just as HTTP does not concern itself with how representations
are processed by a recieving client, nor should URIQA concern itself
with how descriptions of resources are processed by a receiving agent.

The lack of explicit functionality in URIQA for addressing issues such 
as trust or context IMO reflects a proper division in the functional
layers of the web and semantic web architecture, not a shortcoming of
the URIQA model.

> My preference now follows the work of Jon Klienberg 
> (Authoritative sources
> in a hyperlinked environment.). Its been a while since I read 
> Jon’s work
> properly, but if I remember correctly he proposes an 
> algorythmic scheme
> based both upon patterns found internally within a URI 
> (semantics,syntax,
> structure etc), 

That would seem to violate the principle of URI opacity. (???)

> patterns found within the external URIs that 
> reference, and
> are referenced by, the URI and the patterns  formed by the references
> themselves (i.e., the context in which the URI exists). This 
> provides a
> much richer picture of a URI than semantic self-referencing 
> on its own, and
> possibly provides a valid base on which also to establish 
> composite IFP
> schemes for automatically aggregating dynamic ontologies. Obviously,
> however, such schemes involve much higher levels of complexity than is
> currently common on the Semantic Web today. Nevertheless, for 
> me, authority
> is a holistic concept that involves varying degrees of trust 
> – an over used
> term if ever there was one. It cannot necessarily be 
> represented purely via
> formal semantics. The embodiement of stochastic based 
> algorithms, either
> within, or in layers above the Semantic Web may well be one route to
> achieving authoritative credibility for use by automated 
> mechanisms such as
> semantic web agents. My current position is that 
> standardisation on such
> algorithms could, ultimately, be of more value to the Web’s 
> future than
> applying best practice rules solely at the semantic language level. We
> shall see. Perhaps this is a valid subject for the SWBP WG?

For the most part, if I understood the above correctly, I agree
that trust is a complex issue that will involve numerous bits
of machinery to arrive at a complete solution. Also, different
applications will require different kinds or degrees of trust, 
and thus different complexities of solutions.

> Following on this line of thought, I consider your semantic 
> request “tell
> me about this thing”, to too generic, unfortunately, to be useful.

I see URIQA as one component which can facilitate such solutions.
But while URIQA does not provide a complete solution for establishing
trust for particular applications does not IMO mean that URIQA is in
any way lacking or unsuitable for its intended purpose.

That question "tell me about this thing" is certainly not the *only*
question that a given agent must ask to accomplish a given task. It
is simply one of the first, if not the first, question it might ask.

> Additionally, relying on default semantic schemes to define 
> default schemes
> of consistency, for me, appears to degrade the deliberately open
> descriptive mechanisms behind the Semantic Web. Perhaps a 
> better idea might
> be, “tell me how to find something about this think that I 
> can trust with a
> given degree of certainty. If you can’t provide me with such 
> a specific
> method, Ill default to my own preferred means which are….yada, yada”.

But this is precisely how signed, named graphs work. You ask,
tell me about this thing, and the graph (CBD) returned is signed
in such a way that all signatures and certificates are provided
and the methods for signing and authentication are indicated, 
and the agent then verifies the authenticity of the graph returned
and decides whether to trust it or not.

But most of that is entirely out of scope for the interchange
protocol (URIQA) and in fact, keeping it separate from the
general interchange protocol allows for more flexibility and
allows for trust mechanisms to evolve and improve without requiring
any changes whasoever to the interlying machinery used for general
knowledge interchange.

> Please forgive me for rambling, but I too consider this an 
> important area.
> I look forward to the day when autonomic issues like this 
> possibly become
> mainstream agenda items on a number of standards based working groups.

As do I. In the meantime, I continue to deploy such solutions,
specific to our own requirements and applications.

Regards,

Patrick

> 
> Phil Tetlow
> Senior Consultant
> IBM Business Consulting Services
> Mobile. (+44) 7740 923328
> 
> 
>                                                               
>              
>              <Patrick.Stickler                                
>              
>              @nokia.com>                                      
>              
>              Sent by:                                         
>           To 
>              public-swbp-wg-re         
> <public-swbp-wg@w3.org>             
>              quest@w3.org                                     
>           cc 
>                                                               
>              
>                                                               
>      Subject 
>              08/10/2004 11:44          Some thoughts on 
> effective access   
>                                        to "primary" vs 
> "secondary"         
>                                        resources, consistency 
> of           
>                                        descriptions, and 
> bootstrapping the 
>                                        semantic web...        
>              
>                                                               
>              
>                                                               
>              
>                                                               
>              
>                                                               
>              
>                                                               
>              
>                                                               
>              
> 
> 
> 
> 
> 
> 
> 
> I draw the SWBP WG's attention to some comments which I feel
> are relevant to the WG's activities:
> 
> http://lists.w3.org/Archives/Public/www-rdf-interest/2004Oct/0086.html

> 
> Regards,
> 
> Patrick
> 
> 
> 

Received on Sunday, 10 October 2004 12:35:48 UTC