[rdfweb-dev] Re: XML.com 'Semantic Web: A Primer' article (the SW and you ;-)

[Dan Brickley]

On Mon, 6 Nov 2000, Aaron Swartz wrote:

> Dan Brickley <daniel.brickley at b...> wrote:
> 
> > This stuff also raises countless interesting legal/technical/political
> > issues surrounding unique identification of inviduals... Not to
> > mention privacy...
> 
> Which is true. It always bothered me that RDFweb used mailto:s to identify
> people. I think this is a bad idea, for several reasons:
> 
> 1) people are afraid to give out their email because of spam
> 2) email addresses change and become outdated
> 3) email addresses let you contact the person, but don't give any
> information by themselves

I should make this clearer in the next writeup. The aggregation algorithm,
and the modelling style encouraged for RDFWeb, adopt a more general
strategy. Resources can be identified either by intrinsic URI (where these
are obvious / agreed) or else by the use of a uniquely identifying
property. These properties are those which, for any given value, there is
AT_Most_One resource with that property/value combination. PersonalMailBox
is just one such property. PersonalHomepage and countless others can also
be used in this way. As you point out there are some subtleties, which we
can work around by defining foaf:personalMbox more carefully:

people share mailboxes; so we define it as the 'primary owner'.

people have multiople mailboxes; not a problem - each has atmostone owner

mail addresses get recycled; this is a real pain. We should figure out
whether we mean the current, or the original owner of any given mailbox.

re spam etc, I don't know if this is deployable, but a property like
some-function-of-personal-mailbox that turned mailbox strings into
gibberish consistently might work just as well. 

> I'd propose website URLs, with these benefits:
> 
> 1) plenty of people have websites
> 2) they already have lots of information on them
> 3) if they change, people can set up a PURL
> 4) if you don't have one, you can still set up a PURL, pointing to
> nowhere

These can work pretty well. I use a purl (though they get
annoying). HumanGenomeChecksum is another example I keep using but don't
expect to implement...

dan