Re: OT - Supercomputers

[Gonz]

I guess if youre really paranoid, 2048 bits might be taking a risk at 
some point.  But I'm truly skeptical.  10^12 is a big factor.  I suspect 
that most of our computational performance is now going to come from 
massive parallelism, and not processor frequency.  We still have a 
little room in processor frequency, but not a whole lot, certainly not 
1000x, much less 10^6.  If you made 2048 bit (or more) processors, that 
would certainly help alot, and if you could build lots of them in 3D 
instead of close to 2D like we currently do, then that would buy you 
some more power.  But getting a million million times more I just dont 
see it in 25 to 30 years.  More like 200 to 300 years.  But technology 
is always finding new ways to keep up with Moore's law isn't it?

rg
[EMAIL PROTECTED] wrote:
Gonz mused:
Just sit back and enjoy the security.  That little 36.1 Tera-FLOP is 
just a drop in the bucket compared to the amount of power it would take 
to crack the 2048 code, much less the 4096 one.   36.1 Tera-FLOPS is 
36.1 x 10^15 flops.  If you remember that post before, if each quark in 
the universe was able to run at 10^15 flops (1 Teraflop), then all the 
quarks in the universe could not crack the 4096 (or 2048) code before it 
went cold.

You might want to check *your* figures again, too.
4096-bit DES is computationally unattainable, but to crack 2048-bit RSA
inside a year is only beyond current state-of-the-art computers by around
a factor of one million million.  Assuming that processor speed increases
continue at approximately the current rate, we'll probably get there in
25 or 30 years, more or less.
That suggests 2048-bit keys should be secure for maybe the next 20 years.
Adding two bits per year to the key length should more than keep up with
the increases in processor speed, if you're really being paranoid.