[foaf-dev] rough notes: badges, lists and groups in RDFa/OWL

Dan Brickley danbri at danbri.org
Mon Dec 20 23:09:16 CET 2010


On Mon, Dec 20, 2010 at 10:26 PM, Peter Williams
<pwilliams at rapattoni.com> wrote:
> I can see the need, in an assertive world FOR A "do-not-assert" mechanism. And, indeed, a do-not-re-assert" mechanism.


Yes. It's not easy. The technology lets you do certain things which
socially wouldn't help much. I could assert that I'm a member of the
class that is mutually disjoint with 'Crook', but doing that just
draws attention to the [non-]issue.

RDF really has two modes of operation; the mode where you stand back
and take a skeptical Web view, looking at it in terms of documents
making (possibly competing, conflicting) claims about the world. And
the view where you dive into "the world according to" some such set of
documents. SPARQL has some nice primitives that help with this switch,
"GRAPH" etc. As the years have gone by, the ability to work in this
latter, more skeptical, pluralistic, sophisticated etc mode have
slowly improved.

But all that said, far too many consumers of RDF just slurp everything
in, on the assumption that it's all tends towards being true-ish, and
that dealing with all that who-said-what stuff is a pain (and an
unsolved UI problem too).

One of the problems with having a machine-readable 'do not assert'
mechanism, or rather an 'if you suspect x, consider this evidence'
mechanism, is that you don't want to repeat rumours just to disprove
them. But related to that is the lack of a canonical form,  which
makes it touch to match things up. So Eddie Example might want to say
"I'm not a crook", and assert some RDF accordingly. Or link the topic
of Eddie's supposed criminality to some page with human or machine
information, nuanced debate etc. However if someone else was dealing
with triples asserted in terms of fraud or dishonest or whatever, it
might take a system like Cyc to link the two topics together.


> You cannot claim me as a member of YOUR group, just because you assert it to be so. It may be preducial to me.  Sorry, as much as folks want to, you may NOT put me in a listing of (US) Americans (because its strictly illegal for me to claim to be one, or half claim that status, or even to let there exist evidence that others are claim that state about-me).

So you'll want to watch out for those kinds of groups that you don't
get asked about joining. Facebook's new groups system has that
characteristic.  Twitter lists too.

eg. http://culturemap.com/newsdetail/10-08-10-how-mark-zuckerberg-found-man-boy-love-and-what-it-means-to-your-new-facebook-future/

"Thursday blogger Michael Arrington added Zuckerberg (his Facebook
friend) to the Facebook group "North American Man Boy Love
Association" (NAMBLA).  Facebook's new Groups feature allows users to
add their friends to various "friend groups" which can be open,
closed, or secret. Though the tool was designed to make the social
networking site more user-friendly, some are upset that friends can
add them to groups without their permission."

> This suggests: for my badge with a do-not-assert tag, the party asserting I have consented to listing in their group SHALL have some or other (technical) evidence of that consent - probably from the SSL or SSL+Digest handshake.

Yes, it would be good to have some non-paper papertrail showing
consent where it exists, and a general understanding (of the relevant
schemas, APIs etc) that just because some Web page carries a claim,
it's not necessarily true, and it's not necessarily asserted by the
party it claims to describe.

> The nice thing about the SSL-enhanced www-auth header's fields is that there is implicit evidence from the crypto and the crypto's timing and handshake - that then can tie HTTP handshakes to time and place. Used nicely, these might be re-purposed for *Discretionary* consent in semweb groupware, to deal with the kind of semi-trust/semi-formal controls needed in large societies.

The key there is "used nicely" :)

cheers,

Dan


More information about the foaf-dev mailing list