On Wed, Sep 10, 2003 at 11:15:49AM -0700, Simon J. Gerraty wrote:
> Robert Elz wrote:
> >client at just the right time.   This libc really good random ID stuff
> >makes no difference to anything at all that matters for DNS queries - it
> >simply adds overhead.
> 
> And serious overhead at that.  The suggestion to use this new random
> id thing for DNS queries raised a red-flag for me when I saw it,
> but I couldn't follow-up at the time.  I agree with Robert that this
> is likely excessive overhead for the benefit it may provide to DNS
> clients (in the real world).

Note that the libc version is doing a lot of 64bit divides and a
system call for every number....

The 'fast modular exponentiation' routine isn't particularly fast!
That function can easily be written without a single divide (for the
values of 'mod' in that routine).

	David

-- 
David Laight: david@l8s.co.uk