Mailing List Archive: problems with 0.9.5b (CVS)

problems with 0.9.5b (CVS)

Apr 15, 1999, 5:01 AM

Post #1 of 3 (156 views)

I've just had a look at the current CVS versin of gnupg, and there
are a couple of problems:

- Validity calculations seem to have problems. Consider the
following public key (output from pgp 2.6.3in, reduced to the
relevant parts):

pub 1280/593238E1 1996/01/19 Thomas Roessler <roessler@guug.de>
sig!* CE6AC6C1 1998/01/05 Thomas Roessler <roessler@guug.de>
sig! 593238E1 1997/05/19 Thomas Roessler <roessler@guug.de>
Thomas Roessler
<Thomas.Roessler@Sobolev.Rhein.DE>
sig! 593238E1 1996/01/21 Thomas Roessler <roessler@guug.de>

The key is axiomatically valid/trusted. Nevertheless, gnupg gives
the following validity information:

pub:u:1280:1:3E2500EF593238E1:1996-01-19::3017:-:Thomas Roessler <roessler@guug.de>:
uid:q::::::::Thomas Roessler <Thomas.Roessler@Sobolev.Rhein.DE>:

That is, undefined validity. My guess is that gnupg _always_
ignores self-signatures for validity computations. This is ok,
except in the case of our own keys.

- GnuPG still requires a self-signature when it should add a key to
the public key ring. This breaks the key pair concept introduced
by the IN Certification Hierarchy: PGP 2.6.3in has usage
restrictions which are put into the key's user ID. Many people
generate _two_ public keys, a signing and an encryption key, the
encryption key's user ID signed with the corresponding signing
key, BUT NOT SELF-SIGNED.

The idea behind this is roughly similar to OpenPGP's key/subkey
concept. (The implementation is flawed, but that doesn't really
matter here - the keys are out there.)

Example:

pub 1535/E013E2A9 1999/01/08 Hanno Wagner (Rince) <rince@xlink.net> (SIGN,EXPIRE:1999-12-31)
Expire: 1999/12/31 SIGNature only
sig!* CE6AC6C1 1999/01/25 Thomas Roessler <roessler@guug.de>
sig!* 42F948E9 1999/01/22 Hanno Wagner (Rince) <wagner@fitug.de> (SIGN,EXPIRE:1999-12-31)
sig! E013E2A9 1999/01/08 Hanno Wagner (Rince) <rince@xlink.net> (SIGN,EXPIRE:1999-12-31)
pub 1535/34469C01 1999/01/08 Hanno Wagner (Rince) <rince@xlink.net> (ENCR,EXPIRE:1999-12-31)
Expire: 1999/12/31 ENCRyption only
sig!* CE6AC6C1 1999/01/25 Thomas Roessler <roessler@guug.de>
sig!* 42F948E9 1999/01/22 Hanno Wagner (Rince) <wagner@fitug.de> (SIGN,EXPIRE:1999-12-31)
sig! E013E2A9 1999/01/08 Hanno Wagner (Rince) <rince@xlink.net> (SIGN,EXPIRE:1999-12-31)

Bottom line: Due to the lack of a self signature under the
encryption key, gnupg refuses to import it into the public key
ring, making secure communications with many users of PGP 2.6.3in
unusable.

gnupg should at least have an _option_ to enable the import of
keys which have only non-self-signed user IDs.

tlr
--
http://home.pages.de/~roessler/

Re: problems with 0.9.5b (CVS) [ In reply to ]

wk at isil

Apr 18, 1999, 6:24 AM

Post #2 of 3 (156 views)

Permalink

Thomas Roessler <roessler@guug.de> writes:

> pub:u:1280:1:3E2500EF593238E1:1996-01-19::3017:-:Thomas Roessler <roessler@guug.de>:
> uid:q::::::::Thomas Roessler <Thomas.Roessler@Sobolev.Rhein.DE>:

Fixed (but a bad hack). Now that we have validity information for
every uid, I think it makes sense to print all uids and do not merge
the forst one with the "pub" line. Right or would it break any code?

> gnupg should at least have an _option_ to enable the import of
> keys which have only non-self-signed user IDs.

I could argue, that they are anyway only usable with RSA/IDEA. But okay,
I see that there is a problem. Maybe a solution is to allow this only
for RSA keys in v3 packets which do have a "ENCR" in it? Simply
allowing non self-signed sigs may produce some other problems - please
give me some time to think about it.

--
Werner Koch at guug.de www.gnupg.org keyid 621CC013

Re: problems with 0.9.5b (CVS) [ In reply to ]

roessler at guug

Apr 19, 1999, 2:35 AM

Post #3 of 3 (159 views)

Permalink

On 1999-04-18 15:24:00 +0200, Werner Koch wrote:

>> pub:u:1280:1:3E2500EF593238E1:1996-01-19::3017:-:Thomas Roessler
>> <roessler@guug.de>:

>> uid:q::::::::Thomas Roessler <Thomas.Roessler@Sobolev.Rhein.DE>:

> Fixed (but a bad hack). Now that we have validity information for
> every uid, I think it makes sense to print all uids and do not
> merge the forst one with the "pub" line. Right or would it break
> any code?

Umh... The parser code in the unstable mutt should be able to
handle this, the stable code won't. This is not so nice since the
new parser assumes a considerably different data structur which I
didn't want to back-port. Maybe you coul d make this an option?

> I could argue, that they are anyway only usable with RSA/IDEA. But
> okay, I see that there is a problem. Maybe a solution is to allow
> this only for RSA keys in v3 packets which do have a "ENCR" in it?
> Simply allowing non self-signed sigs may produce some other
> problems - please give me some time to think about it.

ok.

tlr
--
http://home.pages.de/~roessler/