>> Justin Coffman <jus...@coffman.tech> writes: >> >> Package: tf5 >> Version: 5.0beta8-5+b1 >> Severity: important >> >> TinyFugue, when compiled from upstream source against OpenSSL, is >> capable of the full set of expected ciphersuites (up to and including >> TLSv1.2), such as those utilizing AES-GCM and EC Diffie-Hellman. The >> version packaged in Debian, compiled against GnuTLS, is only capable >> of >> SSLv3/TLSv1 negotiation, and only then with servers that do not >> require (EC)DH negotiation. This could render the client unusable for >> servers that enforce more modern security policies. >> >> TinyFugue when compiled against OpenSSL: >> % Connected to (unnamed1) using cipher ECDHE-RSA-AES128-GCM-SHA256. >> >> TinyFugue when compiled against GnuTLS, same site: >> % Connected to (unnamed1) using cipher RSA_AES_128_CBC_SHA1.
> Unfortunately, it can't be compiled against OpenSSL and included in Debian > since the licenses conflict. (Which is why it's built against > GnuTLS.) It's GPL without any license exception, so such a package would be > rejected by Debian ftpmaster. > > Sadly, upstream was contacted about this in the past and doesn't feel the > problem warrants the effort required to correct this, so there's basically no > chance that an OpenSSL build will be possible in Debian. > > Presumably there's some way to make GnuTLS negotiate the correct ciphers, but > unfortunately I don't know what it is off-hand, and probably won't have time > in the near future to do the necessary research. Patches welcome! > > -- > Russ Allbery (r...@debian.org) <http://www.eyrie.org/~eagle/>> I tried my hand at generating a patch, but the patched version didn't exhibit behavior any different than current. I guess my GnuTLS-fu is not strong enough. The gotcha (I think) is in the way GnuTLS shims the SSLv23_client_method in its OpenSSL compatibility layer. The only other available shim is TLSv1_client_method, which seems to behave exactly the same way as it does currently.
