CONFIG RANDOM (compromise?)

A Shelton (ashelton@yallara.cs.rmit.edu.au)
Thu, 23 May 1996 12:09:37 +1000 (EST)


Just a couple of (possibly stupid) questions about /dev/random.

* Using the same name for a weaker device seems dangerous, but
isn't it valid for /dev/random to return ENO??? to indicate
that the program must supply its own method.

16K plus entropy gathering overhead seems significant for a
fairly specialized function.

perhaps another /dev for a weaker but always present random
device?

* I assume it is impossible for the entropy pool to become full,
and thus swap-able if no calls are being made on it? That would
be because it is too expensive for the entropy gathering elements
to check for status on the pool?

and, just in passing, there is no way a user can verify that he
is using a real /dev/random in a source available world is there?