[RC5] DES 2 (in preparation of)

Mary Conner trif at serv.net
Wed Jan 7 14:01:49 EST 1998


On Wed, 7 Jan 1998, Skip Huffman wrote:
> Why?  Who cares if you have some unprocessed blocks on your system? 
> They are not going to get processed any faster sitting on the main
> keyserver.  Eventually they keyspace will be exhausted and the
> keyserver will start reissueing the oldest blocks that have not been
> turned in.
> 
> So.   Don't worry about it.

Okay, here's the scenario.  We'll assume the one we were originally
talking about, an 80 block in-buffer, the client does 30 blocks a
day, they are faithfully flushed once a day, either by hand, or by
having a 30 block out-buffer.  The keyspace is exhausted, the blocks
are reissued, and those 50 stale blocks are done by someone else.
But then either the computer's owner goes on vacation, or there is
some network problem, and the client goes on with doing those 50 blocks
which have already been done by someone else.  Better to be doing random
blocks than 50 that have already been done.  I don't expect this to be too
big a deal with DES II, but I've heard that with rc5/64, when the 0x64
keyspace is exhausted, once return of 0x64 blocks trickles off, those
that have not been returned will be reissued in order to close out the
keyspace.

In order to avoid this situation you can do one of three things.
1) don't run with a in-buffer bigger than your out-buffer, and don't
manually flush the client on a regular basis.
2) if you do decide you want to do this, then when you start seeing
keyblocks from outside the current subspace let your in-buffer go empty to
clean out those blocks.
3) or, once you start seeing keyblocks from outside the current subspace,
you can delete your in-buffer to ensure that your "dead space" will
not have blocks that will eventually be reissued before you're likely
to get to them.

Trif
Official Whale Exploder
Official dbaker Irritant


--
To unsubcribe, send 'unsubscribe rc5' to majordomo at llamas.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list