[Previous] [Next] [Index] [Thread]

Re: what are realistic threats?

>From: SZABO @ SMTP (Nick Szabo) {szabo@netcom.com}
>Date: Wednesday, October 05, 1994 4:24PM
>
>
>Dave Kearns:
>> But who would guarantee the statement that "No major security
>> holes have been found"? Are we simply to take XYZs word
>> for it?... No, but we need to 'guarantee the integrity' of the
>> SIGNER.
>
>What on earth do you mean by "guarantee of integrity"?  A legal
>contract promising you your money back?  Legal liability?
>Any old stranger claiming that he is making a "guarantee"?
>Specifics, please!
>

GUARANTEE: an assurance of the quality of or of the length of use to be 
expected from a product.

No legal liability, no 'money-back' provision, just the assurance that a 
given statement is true, to the best of the guarantor's knowledge. Please 
feel free to use the term CERTIFY if it makes you feel better.

>> No, hierarchies allow for standards based rules for issuing
>> certificates and 'guarantees'.
>
>It's quite possible to issue certficates without any sort
>of heirarchy: an example is the widely used public-key cryptography
>system, PGP.  And here's another place we need to be more
>precise: does "heirarchy" do we mean a single-rooted tree, a directed
>acyclic graph, a cyclic graph, or what?  What specific constraints
>are being set by the standards?   My argument was against
>single rooted trees.

"Hierarchy" simply means that each guarantor is guaranteed (or 'certified' 
if you prefer) by a
higher ranking guarantor - where "higher ranking" is a subjective judgement 
on the user's part.
Conceptually, I guess I see it as a "multi-rooted, distributed tree", or 
even trees - since the root of one tree might or might not be a branch on 
another tree.

>
>> The important point, to me, is that there exists a path I can follow
>> to establish the credentials of the Guarantor and satisfy myself
>> as to the reliability of whatever it is I'm about to access.
>
>I agree, but I'm hardly willing to follow some ill-defined "guarantee",
>or trust somebody merely because he's called a "Guarantor".

You don't trust someone simply because he's a guarantor (that's a 
tautology), but because you have knowledge - either personnally or through a 
higher guarantor - of his integrity and competence.


>I want each cryptographic step to be precisely defined, and
>each claim in a certificate be specific and highly credible.
>A system based on ambiguous "guarantees of integrity" wouldn't
>provide anything even approaching a guarantee of integrity.
>
Who defines these "precisely defined" steps?

 -dave
Follow-Ups: