Re: Making encoding out of an authentication cipher

[pgut001@cs.auckland.ac.nz]

Green alien space slime made ogren@cris.com ("David F. Ogren") write:
 
>1. It is slow.  This method would appear to be approximately the speed of 
>MDC.  And MDC (using SHA, what appears to be the most secure hash) is (very 
>roughly) 5 times slower than Blowfish and 3 times slower than IDEA.  And 
>although MDC is faster than 3DES in software, 3DES could easily outpace MDC 
>in hardware.
 
It depends.  On a PC (the most common type of computer hardware) you're 
limited by the bus speed.  Most encryption cards are still ISA (so far I've 
managed to find a single PCI DES card, everyone is still shipping ISA cards 
because they're the lowest common denominator).  This means that even with a 
15 MB/sec DES chip you can't get more than 1 MB/sec throughput.  I think the 
breakeven point for MDC/SHA vs the ISA bus is either a P5/90 or a P5/100 (I 
don't have access to either of them to check this right now).  This isn't just 
for MDC, virtually anything on any recent PC is faster than the ISA bus (in 
fact from Eric Youngs libdes figures there are CPU's which will do software 
3DES faster than hardware 3DES on an ISA bus).
 
Peter.