Re: [gentoo-dev] Re: [gentoo-core] on the matter of security and cryptography

[Jean-Michel Smith]

On Tuesday 06 August 2002 02:00 am, Ryan Phillips wrote:
> On Mon, 5 Aug 2002 23:18:52 +0200
>
> Lukas Beeler <lb-lists@××××××××××××.org> wrote:
> > * Ryan Phillips <ryan.phillips@××××.edu>:
> > > Not entirely the case... Daniel could call me, and since I trust
> > > daniels PGP key I could have him sign a quote that I state over the
> > > phone.
> >
> > And how exactly do you make sure, that you are talking to Daniel?
> > You dont have a Passport, which officially states who he is, you
> > cant compare the foto. Signing over a phone line is completely
> > worthless, and destroys the web of trust.
>
> I'm guessing I would meet him at Linux World...

Not only that, but you *can* verify someone is who they say they are over the 
phone, if you've spoken to them in the past and can recognize their voice.  
Now if you want to get into ultra-paranoid concerns about people synthesizing 
or mimicking someone elses voice than go right on ahead, but I would point 
out that it would be easier for a cracker to make a fake passport with his 
picture on it and claim the developer's identity in person, so even a face to 
face meeting isn't foolproof.

The moral: if the CIA or the FBI are intent on 'cracking' gentoo, then they 
are probably going to succeed.  However, it is extraordinarilly unlikely that 
they would have any reason or incentive to do so, and the deception will 
collaps anyway as soon as the real developer emerges or discovers things are 
being signed on his/her behalf.

In a more general sense, there are two webs of trust that need to be 
addressed, separately.

One is the web of trust between developers.  Spider, drobbins, et. al. need to 
be able to be confident that things signed by each other and other developers 
are authentic.  This web of trust should be built carefully, ideally by face 
to face meetings or a key-signing party, but snail mailing a hard copy of the 
public key fingerprint, and verifying the credentials over the phone once it 
arrives, is more than likely very sufficient.

The other is the web of trust between the developers and the world at large.  
This can be a little looser, and should probably take the form of a public 
keyring of all the developers' public keys, downloadable from a separate 
server than the rsync and tarball mirrors are downloaded from, made available 
on several diverse, independent public key servers, and probably sold on 
CDROM for a nominal price for companies and organizations that want to be 
really, really certain they've got legitimate keys.

Nothing is perfect, but this is a solid foundation for building up a good web 
of trust.  Solid is really all we can ask for, and I would suggest it is a 
mistake to refrain from doing anything merely because perfection is 
impossible.  In a world of billions, we can't ALL attend key signing parties 
and meet the people we correspond with face to face, just as we cannot verify 
every signature, on every letter we receive.  This is no reason for people to 
start passing unsigned checks, however.

Jean.