Re: Semantic Web Hackings

Danbri did some stuff a few months ago working on his "friend-of-a-friend"
ideas to collect information about people. There is stuff in DC that could
probably be sensibly applied (and stuff that could probably be sensibly left
out - I am not sure if you can copyright a person, although I guess that
people will figure it out as soon as they have got the general patents over
people <grin/>).

There is a limit anyway - while people are essentially unique, very few of
them have a unique identifier (I believe my name is unique in world history,
but I am aware that that is not the case for many of my work colleagues, let
alone the rest of the world). It is really a question of establishing enough
context to be able to match, not-match, or not be able to decide. Besides, I
am not sure that there is an intrinsic need to disambiguate people perfectly
- since various government agencies such as the tax office have failed in my
case, and groups like credit rference agencies routinely make mistakes, it is
more a case of being able to establish how much you trust a particular
source. And an out-of-band requirement of society that mistakes can be fixed
- anyone seen "Brazil", or read the bit of the book "Bliss" (the one by Peter
Carey, about a guy called Harry Joy, who tells stories, ...) where two
identities are exchanged for a while?

I am pretty sure that danbri's stuff uses mailto: and web pages as "tags" for
a person. If you add a date to a statement then you can say
"{ in december 1997} refers to
{ in December 2000}" and then that either of those is
the same as ...


Charles (who hopes danbri joins this thread - no time to go check up on my
assertions here :)

On Thu, 28 Dec 2000, Sean B. Palmer wrote:

  > personally I know one couple reachable at don.deb@... [...]
  > However, there isn't much else around at the moment - 'mailto:'
  > looks promising.

  As the SW will be run on URI's, we have to choose some form of URI to
  represent people. We are limited to either http:// or mailto: really for
  these purposes, and I felt that using mailto: would have many advantages
  over http:// because it is often used more singularly. There is nothing to
  stop you having an "SW assertion" about a mailbox that has two people: just
  say that this mailbox is owned by person 1 (description) and person 2

  > the sitting-on-lap problem could be handled by additional
  > qualifying attributes.

  That's a good suggestion.

  > The general idea I reckon is basically valid - the URI doesn't
  > have to refer to the person, just their interface to the digital
  > environment.

  Yes! Very neatly put. A person is unique, and a URI is unique, so if the
  Person => Web interface has a single URI (mailto:) then there should be no
  logistical problems there at all: one person, one vote, one URI. If I had
  three mailboxes (I have a lot more than that), then I could say that all of
  these mailboxes are owned by the smae person that owns this first mailbox:-

   <rdf:Description rdf:about="">
    <asserts:name>Sean B. Palmer</asserts:name>
    <sig:signed>[Digital Signature]</sig:signed>

  To do that properly (binding further URIs to a root URI), requires
  security, which leads us nicely to:-

  > Security could be a bigger problem though, I would think the
  > overheads could get sticky - perhaps one could have a public
  > URI (not secure) and a private one (digitally signed)?

  Yes, digital signatures and security are a problem at this stage. I'm no
  expert on security, but because there is *such* a commercial and practical
  need for it, I'm sure it will be perfected in time. For now I don't thenk
  that there is much of a problem: for example, if I have an RDF assertion
  for my mailbox on my server, you can be pretty sure that it is going to be
  valid. In other words, if I make an assertion about an
  mail address on then you can be pretty sure
  that you can trust it. If it's at then you
  have to wonder...

  > boris.ivan.vlad
  > I believe there are already geneology schemas in circulation...

  Intersting. I reckon that would make for intersting property information
  once the URI is asserted as definitively being a representation of me...
  I reckon that this exposes a true practical use for the Semantic Web that
  can be implemented and worked upon right now. If someone were to come up
  with a "name" property that means "is owned by person, and asserts person"
  then the simple triple:- => #name => Sean B. Palmer (verified)

  Has a lot more depth and meaning. From this you can say whatever you want
  about and technically assert it about me as
  long I as verify it.

  Well there you go, I suppose the SW is practical/useful/+ after all... :-)

  Kindest Regards,
  Sean B. Palmer [ERT/GL/PF]
  "Perhaps, but let's not get bogged down in semantics."
     - Homer J. Simpson, BABF07.

Charles McCathieNevile    phone: +61 (0) 409 134 136
W3C Web Accessibility Initiative            
Location: I-cubed, 110 Victoria Street, Carlton VIC 3053, Australia
until 6 January 2001 at:
W3C INRIA, 2004 Route des Lucioles, BP 93, 06902 Sophia Antipolis Cedex, France

Received on Thursday, 28 December 2000 14:30:22 UTC