[rdfweb-dev] Re: Parsing FOAF in perl?

[Jim Ley]

"Bill Kearney" <wkearney99 at h...>
> >From: "Jim Ley"
> >
> > Attribution exists in the foaf database, however, people seem unable
to
> > sign their foaf accurately, so there's not a lot of point, at the
moment
> > there are the same number of people with bad signatures as there are
with
> > good (although it's only uri's that have been updated through the
> > interface that are included) without signing no provenance exists, so
> > it's pretty irrelevant trying to build it into the interface.
>
> While being able to sign your own foaf is handy, there's nothing that
says the
> signed foaf is any more authoritative than an unsigned one.

Of course it is!

Jim at j... saying the person with the email address
jim at j... has the name "Fred Bloggs" has considerable more weight
than danbri at w... saying the person with that address has the name "Jim
Ley". Signing provides the self-assertion, you're right in that non
self-asserted parts are pretty irrelevant (it's part of the general level
of trust you need to put into the system, and there I think it's likely
to have an effect.)

> > We could easily make a version that only allowed information from
trusted
> people,
> > and only allowed "self description" of all information, but I don't
> > really see that such a tool is useful.
>
> I'd think some sort of 'authoritative self description' concept is
important to
> most folks.

The current system would be pretty useless then, I couldn't mark-up my
photo album, and say who was in pictures, I'd have to convince the other
people to author some descriptions saying the same. Only being allowed
to talk about yourself is not useful, you need to be able to talk about
other people, how you trust that information depends on the use. Given
that even self description can contain lies, or simply mistakes, you can
never have a fully trustworthy system.

> > And exists without a problem, it's just not exposed in the interface
(and
> > there's no point even thinking about it until people actually start
> > signing accurately their foaf!)
>
> Again, how is signing going to make it anymore authoritative? Unless
the signed
> file can contain only one "this is me" sort of foaf:Person I'm not
entirely
> clear on the value of signatures.

Hopefully the above description has made it clear, a signature identifies
the person who made the assertions, if the assertions are also about that
person - you have the the self assertion you're looking for, identifying
this at the overview level of foafnaut is a data problem that I don't
think is very feasible, it's the wrong sort of interface, it needs to be
"trust-everything" in the view. However, as things evolve, we'll also
want to find some way to get trust into the interface, either by
filtering it so it's "the foaf universe according to fred, barney and
wilma" or by enabling a mechanism to say "this is a bad triple, remove
all the triples from the same place/person" and getting a refined view
that way. Both of those though have considerable archetecture problems
in the current foafnaut/data.

Jim.