[Monotone-devel] Re: PGP key signing

[Graydon Hoare]

Bruce Stephens wrote:

> And (obviously) maybe a VCS could use some kind of similar idea,
> rather than trust always being binary.
>
> So maybe when I do "mtn update", I could give some indication of how
> lucky I feel, and then mtn could choose a revision that's either
> completely tested and signed by people I definitely trust, or perhaps
> a riskier one with possibly more features.

Despite being frequently lost in the noise of developing a working 
system, this exactly the reasoning that went into the current design: 
that a public key is most useful when it has signed *lots* of old 
material you consider good, not when it's attested through a complex 
PKI. This is why we issue such a great volume of certs, and why the 
update (good/no-good) decision is delayed until the last second, and 
even then consults user preferences via a trust hook.

Of course such a view of public keys doesn't prevent against "passive 
infiltration" attacks, where someone submits weeks or years of "good" 
signed material only to one day suddenly turn "evil". I don't know of 
anything *cryptographic* that protects against such a scenario, though. 
Certainly not having some magic rubber stamp from Verisign or the PGP SCC.

-graydon