Re: What does this all mean?

andrew meggs (insect@antennahead.com)
Sat, 17 May 1997 16:58:11 -0400 (EDT)


On Sat, 17 May 1997, Louis C. LaCour, Jr. wrote:

> 1. Why is it that initially the key server assigned 2^22 pairs, then 2^26,
> then 2^28 then 2^30 pairs? And why is progress measured (on the web site
> and elsewhere) as X * 2^20 pairs checked? In other words, what's the
> "magic" in 2^20? Is 2^20 some special "unit" of labor??
>

2^20 is approximately 1 million.

> 2. What does each "progress dot" (if that's the right term) signify? That
> X pairs have veeb tested? Something else?
>

It means 2^21 pairs tested. I believe in the 64-bit bitslice version it
means more than that. In the upcoming MP-capabale MacOS client there aren't
any dots at all, but you'll see what I mean...

> 3. How does the program know empirically that a given pair doesn't (or
> does) work? i.e. what is the nature of the test it employs?
>

The exact way the deschall client works is confidential, but it goes
something like this: The contest rules specify the encrypted text and the
first few words of the plaintext. To test a key, decode the first block
of the encrypted text and see if it produces the first (known) block of
the plaintext. If so, we can decrypt the entire message and see what it
says. This isn't that farfetched a situation, because many files in the
real world have a known header.

> 4. What does this startup message mean, exactly:
> Foreground: run 30/sleep 0
> Background: run 6/sleep 0
>

You can read up on that in the pereferences file. It has to do with the
way Mac programs use cooperative multitasking, alternating between
computation, polling for user events, and yielding time to other programs.
The "run" parameter is how many ticks it computes between calls to
WaitNextEvent, and the sleep parameter is passed directly to WaitNextEvent.

> 5. Finally, not to be a cynic, since I support the effort, but one could
> argue that an encryption key which requires an effort of this scale to
> crack is, by definition, pretty damn secure; i.e. hope the project isn't in
> itself self-defeating.

As long as there's money involved, it's not self-defeating. ;)