Re: normal?

DES Challenge Lists (deslists@dopey.verser.frii.com)
Tue, 10 Jun 1997 02:02:57 -0600


> From: "Charles E. Novitski" <c.novitski@cmich.edu>
>
> >> On Sun, 8 Jun 1997, Jason Buchanan wrote:
> >[..]
> >> > However, the 4th machine i'm running it on is a SS2 and it said "Key
> >> > found" or something like that... and since I wasn't paying close attention
>
> 1. Does anyone have convincing evidence that this "key not found" report
> (above) is just a misaprehension, rather than a real result that didn't get
> sent back to the server? If not, is it not worth (and/or possible)
> rechecking the keys sent to that site?
>
> [snip]
>
> Charles

Let's try to lay this puppy to rest once and for all...

Although it seemed highly improbably, I contacted the person running the
"SS2" client. He gave me the IP address of the machine in question and
the approximate time and date he thought he saw this message.

I checked the logs and found that there were a series of "key not found"
reports from the client to the server, extending a few hours before and
a few hours after his observation.

Except for the Mac PPC client, which doesn't behave identically to the
other clients, if the key is found, the client will send a coded "key
found" message to the server. The client will also display a "key found"
message on the user's terminal. The client will then terminate.

If the server receives a "key found" message, it logs the message; and
sends a "shutdown" message to the client. The server continues to run,
since the server's policy is to distrust the clients. However the server
spawns another process to act upon the allegedly correct key. No "key
found" message appeared in the logs, and no such process was spawned.

Note that all clients (except Mac PPC) will terminate after they have
displayed the "key found" message, even if the server misunderstands the
"key found" message, and even if the server fails to send the "shutdown"
message to the client.

In the case in question, the server received nothing but "not found"
messages, it responded by assigning additional keyspaces, and the
client did not terminate. This is very strong evidence that the user's
machine did not actually find the key.

The messages are just *slightly different* for bitslice clients. I
presume the user misread the message. [The confusion that seems to have
developed is a good argument for keeping all clients as identical as
possible.]

-- Rocke