Re: symbol catenation and montgomery

[Torbjorn Granlund]

Werner Koch <address@hidden> writes:

  There is a reason not to use GMP:  It is a complex software and
  GnuPG is already too large - every line of extra code is a security
  risk and by stripping down the math lib to the actually needed size
  we can make the GnuPg somewhat more secure. (Well, I know there are
  all the ugly options, but they are very straighforward implemented
  and should only have a low risk of unwanted sideeffects).
  
That's a good point.

We actually have a known buffer overrun in GMP 3.1 (shiver!).  It is
in the mpf function layer, which few serious cryptographers use.  But
a buffer overrun is always a buffer overrun.  :-)

We are now pushing GMP 3.1.1 out the door, which has only unknown
buffer overruns.

-- 
Torbjörn