Re: [foaf-dev] fidelity vs idiom

Hi Peter,

On Wed, Nov 24, 2010 at 5:37 PM, Peter Williams <[email protected]> wrote:

> -----Original Message-----
> From: foaf-dev-bounces-RyYwo1q5J+qsOXdr9/[email protected]
> [mailto:foaf-dev-bounces-RyYwo1q5J+qsOXdr9/[email protected]] On Behalf Of Dan Brickley
> Sent: Wednesday, November 24, 2010 8:12 AM
> To: Rinke Hoekstra; Ian Davis
> Cc: foaf-dev Friend of a
> Subject: [foaf-dev] Implying that two mentioned individuals are distinct...
> updating best practice from owl:differentFrom to
> owl:IrreflexiveObjectProperty?
>
> From discussion here at benelux semweb meetup in Amsterdam -
>
> In FOAF, if we say
>
> alice a foaf:Person .
> alice foaf:knows bob .
> bob a foaf:Person .
>
> this implies the existence of ...
> ...
> It's truth doesn't guarantee formally that there are
> ...
> and that this doesn't put the idiom into OWL Full.
>
>
> One of the things Henry and I used to argue about was whether a FOAF file
> (used in a FOAF+SSL context) was an assertion; similar to a SAML assertion.
> The argument for assumed that because the basis of FOAF (RDF) was formal and
> logical (and capable of making assertions), so too was FOAF; and a FOAF file
> signaled via FOAF+SSL is thus an assertion. At the time of initial
> communication using mainstream SSL, it's an authenticated assertion
> furthermore.

This is a good thing to discuss but I think you've keyed off of the
word 'idiom' in a way that wasn't quite intended.

By idiom, I'm referering to patterns and conventions for structuring RDF data.

I do believe RDF to be claim centric. Most of its formal machinery is
organized in terms of truth-preserving operations, once we have some
mapping from the abstract graph to some reality. But this only gets us
as far as saying that RDF documents can be considered to have
propositional, descriptive content. That doesn't mean that they
directly are *asserted* by someone, just that they could be.

One pattern we discussed here (about 10 years ago, eg
http://lists.foaf-project.org/pipermail/foaf-dev/2000-August/004213.html
) was to (a) pgp sign the doc (b) arrange for the doc to itself say in
rdf 'i'm a doc and my creator was <author> and I was made on
<date>...'. Which in the context of the fairly passive environment of
published Web docs, about as close as we can get to 'assertion'. Now
if you're designing your own full-fledged protocol, you can go much
further.


> If FOAF is only an idiom

"Out of the box", all RDF data is merely assert-able, rather than
necessarily asserted by some real world entity.

This was actually an issue raised during the RDF Core round of spec
revisions by TimBL, who argued
"RDF is not just a data model. The RDF specs should define a semantics
so that an RDF statement on the web is interpreted as an assertion of
that statement such that its author would be responsible in law as if
it had been published in, say, a newspaper."
http://www.w3.org/2000/03/rdf-tracking/#rdfms-assertion

The WG didn't endorse that view, since RDF doesn't have a built in way
of figuring out who is claiming something; it just shows you how to
write down the claims in a standard way.

>      (albeit one with intuitive semantics about its
> pseudo-assertions of identity when used in the FOAF+SSL context) then the
> proposed relationship attributes (when invoking the powers of less than
> OWL-FULL) are not "legally sound" - having not formal basis of truth that
> meets criminal law standards. They may be circumstantially sound, which may
> well meet the legal standards for truthful assertions used typically in
> resolving civil disputes (which much more interesting to the web, of
> course).

I think the working assumption is that most sensibly written RDF
should be the kind of thing you could pass to a 'detective /
scientist' and set them out to determine the truth/accuracy of what is
claimed. Since schemas sometimes have murky vague corners, that (as in
natural language) couldn't be an exact science. But we're certainly
dealing with descriptive technology here.

> Though forlorn, a middle ground of legal semantics exists: known as equity.
> It stands between criminal and common law. About the only place its relevant
> today is in really old-fashioned trust instruments, using archaic language
> mostly. These are often found in legacy social structures, such as mortages
> and land dealing. Posh lawyers trudged out of the American law schools and
> used to invoking the phrases of the   17th and 18th century seem to like to
> invoke these language forms, I've noticed; alluding to (and only "alluding
> to") legal reasoning powers that derive from the rather abstract equitable
> principles  referenced by all the code words.

I've not encountered that concept...

> So, if FOAF as it stands is suitable for the non-criminal law tradition
> (common law), and FOAF with OWL FULL would be suitable for the criminal law,
> then perhaps FOAF's idioms augmented with some OWL constructs might be
> suitable particularly for making statements about the world of trust and
> equity - statements that are "about identity" (merely).

The OWL aspect is interesting. If I assert some set of triples, I have
some relationship to those concrete things. But if the schema I used
evolves, or itself mentions other evolving schemas, and those things
are edited over time, we can't assume the original asserter will
always agree with subtly shifting meaning of all the implied OWL
triples. All the more reason that each schema should keep a safe
date-stamped version history (as is case for FOAF, Dublin Core, W3C
materials at least).

> Of course, FOAF is no longer an idiom - once augmented like this. To be
> honest, I don’t know the correct term for what kind of language FOAF is, if
> augmented with such equitable reasoning. Perhaps it’s now a "fidelity"
> language (vs a mere idiom).

By idiom I mean roughly 'higher level pattern of description'. Other
RDF communities around Dublin Core sometimes talk of 'application
profiles' eg http://dublincore.org/documents/dc-dsp/
http://dublincore.org/usage/documents/profile-guidelines/ as an
attempt to pin down this slippery idea of patterns of description.
It's not quite the same as schema or ontology, since a single
description might use 2,3 or 6 different schemas...

cheers,

Dan