Long Key Performance

[Werner Koch]

On Sat, 20 Apr 2002 08:06:36 +0800, Enzo Michelangeli said:

> Well, you may be right, but why not let the users make their own decision?
> Anonymous appears to have a point.

He can do so up to a certain size.  Going over this needs changes to
the RNG mdoule.  If he does not like the warning message he can remove
them or ask someone to do this.  He might even find a distribution
where this warning is not showed.  Hey, it is free software and
everyone has the freedom to change wahtever he likes.

> large number of messages). There was a time when exceeding the following
> limits may have looked "plain stupid":

> - 31 bits for time_t values

There is no such limit.  Most 32 bit OSes only use 32 bit for time_t
and becuase there implementation is based on signed integers (due to
easier implementaion of error returns) there are effectivly only 31.
Please note that this is not a hard limit and glibc will sonn be
changed to use a larger base type for time_t.

> therefore even very long keys might one day be proven inadequate. But if
> this were the real concern, we should stop using PK crypto altogether,

And I should definitely stop building Windows binaries - using them is
a really threat.

It is better to think about an entire system and not just about one
detail (length of the key).

  Werner