Re: [PATCH] Cryptographically secure transaction IDs

On Wed, 30 May 2007, Brad House wrote:

> Would it make sense to check for the existence of /dev/urandom or
> /dev/random on non-windows machines and just read the necessary bytes from
> that device? And fall back to rand() in a worst-case scenario?

I think we should.

> Have to look at OpenSSL to see the basic search paths they use.

Yeah, we could probably get some good inspiration from their RAND_*()
functions...

-- 
   c-ares -- my preferred DNS asynch resolver library