Message-Id: <9307200600(_dot_)AA02487(_at_)TIS(_dot_)COM>
Subject: Re: PEM WG Meeting Minutes
Date: Tue, 20 Jul 93 02:00:26 -0400
From: Stephen D Crocker <crocker(_at_)TIS(_dot_)COM>
The point I was making is that you can achieve high speed using the
1-loop scheme if you demultiplex the steam. This provides the speed
you want if you use the maximum amount of hardware, but it is also
compatible with current EDE-CBC hardware.
Steve,
the point I was trying to make was that you can not parallelize the stream
on encryption, in the 1 loop case. (On decryption, you can split it into
an arbitrary number of pieces and decrypt all of them in parallel, saving
the XOR for a final pass. You could even have 1 chip per 64-bit block.)
In order to do an encryption in CBC mode, you need the IV or a previous
ciphertext with which to XOR the plaintext *before* starting any DES
encryption. When you first start, you have only the IV available, so you
get no parallelism. To be able to start on the second plaintext block of
64 bits, you have to have completely finished the first block of 64 bits
and that gives you only enough ciphertext for starting just the second
block -- so there is no parallelism possible. This isn't an implementation
choice. It's inherent in the 1-loop structure.
If this isn't apparent, we should take this off line (perhaps by telephone)
and spare pem-dev the discussion.
- Carl
- <<Disclaimer: All opinions expressed are my own, of course.>>
- Carl Ellison
cme(_at_)sw(_dot_)stratus(_dot_)com
- Stratus Computer Inc. M3-2-BKW TEL: (508)460-2783
- 55 Fairbanks Boulevard ; Marlborough MA 01752-1298 FAX: (508)624-7488