Re: digital signatures

[Steve Lipa]

On Apr 01 Przemek Klosowski (address@hidden) wrote:
> Steve,
> 
> I am not denying that digital signatures are good; I am just saying
> that there's not enough infrastructure in the world at large to use
> them exclusively. Doing only gpg sigs will leave out everyone who
> doesn't have the setup; I am arguing to do both!
> 
> You definitely gloss over the problem of not having the gpg/pgp
> infrastructure on the client end. It's one thing to have John do 'gpg
> --sign', but you then have to tell everyone to
> 
>        - get gpg if they don't already have it
>        - find and import John's public key
>        - check the signatures.
>

I'm sorry, but there is no free lunch. Unless you do the three steps that you
list above,  there is no way to know if the octave-2.x.x.tar.bz2 that you just
downloaded is the one that Dr. Eaton wants you to have or the trojan-infested
one that the hackers that rooted www.octave.org want you to have.

MD5 sums can guarantee it.  But you need to go through a procedure at least as
convoluted as the above to achieve it, because, as you pointed out earlier in
this thread, the MD5 sum must be "distributed in a way that does not allow for
surreptitious modification."

This distribution system is PRECISELY what gpg and pgp address. A gpg
signature of a file is just a hash of the file signed with the private key.
gpg just makes it easier to make sure you are using the right hash.

I can see how MD5 sums provided on a separate channel are useful. But I think
that MD5 sums provided on the server are worse than useless because they are
trivial for anyone who has hacked the server to change and therefore may give
a false sense of security to users who are too unsophisticated to realize
this.

Steve

-- 

Steve Lipa
address@hidden
gpg fingerprint = 8B68 77D7 9E09 9991 C97E  25FF 6A12 D2B9 EC7D 66C1