[117470] in Cypherpunks
Re: Testvectors
daemon@ATHENA.MIT.EDU (Anonymous)
Fri Sep 3 18:32:31 1999
Date: 3 Sep 1999 22:16:52 -0000
Message-ID: <19990903221652.10922.qmail@hades.rpini.com>
From: Anonymous <nobody@remailer.ch>
To: cypherpunks@cyberpass.net
Reply-To: Anonymous <nobody@remailer.ch>
Something purporting to be mok-kong.shen@t-online.de did verily spew, upon 02-09-99, the following:
// BEGIN SPEW
Sorry, I don't yet know what your position is. Mr. Koch has written
a 3DES program and he asked in a post whether there are testvectors
for that. I expressed my opinion that it would be sufficient to use
the test vectors of single DES to test the three component DESs,
since interconnecting these components can't offer in practice
any risk of introducing programming bugs, because these connections
are here of very simple nature.
// END SPEW
You don't do much commercial programming, do you? or, for that matter,
much programming on projects with more than one programmer.
"can't offer in practice any risk"
I'm no old gaffer, but I suspect that even the older folks have never heard a remark like
this from a practicing, successful professional.
You then go on to talk about certifying absolute correctness.
Bottom line:even if you test the pieces and each works,
you won't be in business for too long if you send the
first 10.000 systems out the door without testing
them as systems
Monger. Just Monger
