Who says so? Cryptographic certificates to authenticate declarations of fact. [SEC=UNCLASSIFIED]

Hi all - I'm new to this list (new to RDF), so sorry if this has been
done to death already.

AFAIK, the semantic web consists of all RDF documents everywhere.
Looking through RDF: you can say something about two things [sky
has-property is-blue], but there doesn't seem to be an obvious way to
decorate the act of declaring such a thing, aside from making a note of
the url you fetched it from.

Firstly, such decorations are of semantic interest. You want to know
things like: who says so? How sure are they? Does this fact only apply
during a certain time, or in certain circumstances? Is it meant to be a
promulgation of a standard? A statement of commonly accepted fact? A
hope or a wish that something ought to be true? A hypthesis?

Secondly ... just let me toss in the word "porn spammer". If every
declaration of fact is the same as any other, and if facts begin to move
across the semantc web (caches and so on), and if it starts to become
important, then you are going to get commercial spam.

The other problem is liars. What happens when a liar publishes an RDF
document that [X has-propery is-gay], or whatever?

Now at present, it would seem that you *could* do this stuff in RDF,
with a little fooling about. You need a well-known verb that says "X
is-a-caveated-version-of Y", where Y is some other operator. 

sky has-property-005 is-blue
sky is-above-005 the-ground
has-property-005 is-a-caveated-version-of has-property
is-above-005 is-a-caveated-version-of is-above
has-property-005 asserted-at 1-Jul-07
has-property-005 asserted-by encyclopedia-britannica
is-above-005 asserted-at 1-Jul-07
is-above-005 asserted-by encyclopedia-britannica


Perhaps we could make this a little less bulky with inheritance
operators:
"A inherits-subject  B:  B D E -> A D E"
and maybe another one Y that allows you to reverse the arguments
"A inherits-object B:  B D E -> A D E"
Heck, why not complete it
"A inherits-verb B:  D B E -> D A E"

I'm not sure whether "is-a-caveated-version-of" is the same as
"inherits-verb". It is if you decide to drop all the caveats. In any
case:

sky has-property-005 is-blue
sky is-above-005 the-ground
has-property-005 is-a-caveated-version-of has-property
is-above-005 is-a-caveated-version-of is-above
has-property-005 inherits-subject  download-005
is-above-005 inherits-subject  download-005
download-005 asserted-at 1-Jul-07
download-005 asserted-by encyclopedia-britannica

====================================

The thing is, whenever you read a fact from some other source, there is
an implicit performative. My declaration "the sky is blue", when you
read it, nessesarily becomes "Paul Murray said in an email to the
working group at this date that the sky is blue". This wrapping happens
whenver a fact is transferred from one place to another.

However, in a semantic web where facts are moved from place to place,
archived, clustered and so on, the necessity of tracking "who says so,
how and when?" Means that our facts get wrapped in the details of the
transfers "database a says that archive b says that encyclopedia
britannica says that the sky is blue". Every time we import a fact from
anywhere, we have to create another verb.

There is a way of avoiding this: a trusted cryptographic signature
allows us to collapse these layers of packaging. Thus, when encyclopedia
britannica asserts a fact, it can be asserted with an attached
signature. If I get that fact from an archive or other
man-in-the-middle, and the signature checks out (that is: I trust the
CA), I do not need to keep the fact that I got it from an archive. As
far as I am concerned, it comes straight from britannica.

The details are somewhat tricky. Should each fact be individually
signed? Is there a way of bundling them, somehow? Of course, if you
bundle them then in order to check that the signature applies you need
the entire bundle. You could simply permit the entire RDF docment to be
signed out-of-band (eg: http authentication, an xml envelope), but that
means that someone getting your facts that you got from Britannica only
has your word for it. Unless they trust you to make declarations of the
form "I got this from britannica".

But the sematic web is going to need something like this, and it's also
going to need a well-known set of very basic verbs about other verbs: A
implies B, A means the same a B, A is the negation of B, and so on.

May I also take the opportunity to complain about the lack of an XSD?
Describing RDF in RDF is fine, but RDF is transferred in XML documents,
read in with an XML parser, and I'd like to be able to validate it at
that level.





 



------
If you have received this transmission in error please notify us immediately by return e-mail and delete all copies. If this e-mail or any attachments have been sent to you in error, that error does not constitute waiver of any confidentiality, privilege or copyright in respect of information in the e-mail or attachments. 



Please consider the environment before printing this email.

------