Valdis Kletnieks <Valdis.Kletnieks(a)vt.edu> writes in xemacs-beta(a)xemacs.org:
On 04 Dec 1998 05:32:33 PST, you said:
> Once I did manage to get it built (and after generating a secure key
> on a Linux box with /dev/random and importing it to the BSDI box) it
> spits this at me when stuff is signed:
This is, in *general*, not the recommended procedure, for all the
obvious reasons having to do with people seeing the keys as they go by
on the wire or finding the floppy after you sneakernet it.
Sneakernet is out of the question.
cvs.xemacs.org is physically 200
miles away,
ftp.xemacs.org is physically 3000 miles away from me.
My development machine is no longer accessible via the 'net and would
make a perfect machine for signing except that I don't have a 'net
connection where I can transfer that many megabytes of data as often
as would be needed if distributions were being built there.
Now, if you use sufficient care (such as using SSH with good crypto
support), you can close those exposures, but... ;)
I am. Since moving the XEmacs ftp server to gwyn from UIUC, I'm only
talking to it over encrypted links, which is an improvement from the
straight FTP I was doing in the 19.15 .. 20.2 days.
> Um, what does having a weak RNG have to do with signing with a
key
> generated with a strong RNG?
Due to the above, the code is making an assumption that the key was
generated on the local machine. It has no way of knowing that you
built it on a machine with a working /dev/random and played crypto
games behind its back...
I got an answer back from one of the gnupg people. It is explained in
Schneier, _Applied Cryptography_[1]. Good random numbers are required
for DSA digital signatures. Double Sigh.
Footnotes:
[1] Which I suppose I should have checked first, sigh.