1993-11-17 - Re: Key Servers

Header Data

From: Derek Atkins <warlord@MIT.EDU>
To: “L. Detweiler” <ld231782@longs.lance.colostate.edu>
Message Hash: 8a29750fa9e9a33e0fad70b9c6e657e524a1eae9a7c98bc643f1dcd96f99a1eb
Message ID: <9311170639.AA02048@toxicwaste.media.mit.edu>
Reply To: <9311170445.AA26434@longs.lance.colostate.edu>
UTC Datetime: 1993-11-17 06:41:01 UTC
Raw Date: Tue, 16 Nov 93 22:41:01 PST

Raw message

From: Derek Atkins <warlord@MIT.EDU>
Date: Tue, 16 Nov 93 22:41:01 PST
To: "L. Detweiler" <ld231782@longs.lance.colostate.edu>
Subject: Re: Key Servers
In-Reply-To: <9311170445.AA26434@longs.lance.colostate.edu>
Message-ID: <9311170639.AA02048@toxicwaste.media.mit.edu>
MIME-Version: 1.0
Content-Type: text/plain


-----BEGIN PGP SIGNED MESSAGE-----

> Seems to me, MR EICHIN, that many people might be FLABBERGASTED to find
> out that people are using PGP key servers for PSEUDOSPOOFING.

No, it seems that only *you* are flabbergasted.  As has been stated
numerous time, the Keyservers exist solely for distributing keys.
Thats it. No more. No Less.  They are not existant to police anyone.
They make no policy decisions.  Anyone who wants to publish a key,
under any name, may do so.  That has always been the policy.  That
will always be the policy.

> why is it that the policy that ANYTHING GOES is NOT MADE CLEAR in
> KEYSERVER POLICY DOCUMENTS?

Why should it?  As I said, anyone can add any key.  The Keyserser
serves keys.  It doesn't, and I believe that it *shouldn't* make any
verification about the keys it serves.  That is the jobs of
signatures.  The Keyservers (by which I mean the Keyservers at
pgp.mit.edu and elsewhere) *don't provide any authentication*.  They
never have.  They never will.  And I don't believe it is their purpose
to do so.

> I'm writing this in hopes there are still some sane people left here.

There are many sane people out here.

> It's quite shocking to hear such blase arguments about identity from
> the people who have erected the current key server system. At the very
> minimum, these people should make their policies about pseudospoofing
> clear in policy documents -- anyone listed here could be imaginary,
> there are no guarantees.

I don't think that there are any arguments about identity.  In fact,
the only person I know who's brought up the issue of identity is
yourself.  The Keyservers have been erected for one, AND ONLY ONE,
purpose: TO SERVE KEYS.  If you can tell me which word you do not
understand I will be happy to explain in excruciating detail what I
mean here.  The Keyservers exist so people can request the PGP key for
some identity.  It does not matter whether that identity owns one or
one hundred keys.  If you want to insure identity, sign the key!

> 5) I remember a P.Metzger argument with M.Graff on some mailing list
> about using DNS as a kind of key distribution system. The argument
> boiled down to the point: can *anyone* insert entries? Coincidentally,
> this is the critical question in a True Name vs. Toxic Waste Dump databases.

Actually, there was more to this argument than just that.  The problem
is not how to insert entries.  Rather, the problem is how to
distribute the system so that the who system knows about every key.
The problem is that each key has more than one name: it has its
userID's associated with it, and it has a keyID, which is inherent to
the key.  The problem is how to distribute the database so that
lookups by keyID can be accomplished.

The question was *never* about who had the ability to enter keys in
the database.  The question was how to implement the database on top
of software that currently didn't allow for what was required.  The
final decision, as it stands, was that the current software was not
capable of performing what was required.

> 6) The software already exists to have a separate network of True Names.

Absolutely correct.  It is called a Digitial Signature Hierarchy (ala
PEM).  It can also be easily implemented in PGP as well.  In fact, I
was planning on doing this!

> I'm sure all the Psychopunks will come
> up with their clever arguments to sabotage any True Name debate or
> database system. 

I'm not trying to sabotage the debate.  I'm trying to say that there
are better, more effective ways of ensuring a True Identity, than
creating a placebo system of keyservers.  It is much more secure to
create a system of signators that will digitally sign keys that belong
to True Names.  If you see such a signature, then you are assured that
it is a True Name.  If you receive mail from a keyserver, it is quite
possible that the Keyserver response was intercepted and changed, or
that the Keyserver itself has been altered by some evil cracker, or
whatever.

As has been stated a million times: The keyserver provides keys.  The
signatures on the keys provide authentication.  If you want to
authenticate something, check its signatures.  If it has no
signatures, it should not be trusted.  If it has untrustworthy
signatures, it should not be trusted.  Which concept do you not
understand?

- -derek

-----BEGIN PGP SIGNATURE-----
Version: 2.3

iQBuAgUBLOnHGjh0K1zBsGrxAQGVHgLECYxXO/kDfttEY4KOyaQB9b+wLqFy2Omv
2q0CEaralDXJ2SZtJRZM4QhUWDoDvYYG23TeGZ3GTsgQxPccfWzSx+qv/qSpVfpn
9pZWBQ/RgG3zKPyV+Kd3YFk=
=e6Cm
-----END PGP SIGNATURE-----





Thread