My statement on draft 5.3.2
I have attached a copy of my statement to the TGDC. My statement
primarily focuses on the reliability, availability, and accuracy
requirements, although it also summarizes some of my other comments.
The reliability requirements in the current IEEE draft are the same as
in the FEC 1990 and 2002 standards.
In summary, I have found that the specifications are deeply deficient in
reliability and accuracy, and that the reliability deficiencies
correlate with the observed unreliability of voting systems reported by
citizen groups during the recent election.
Essentially, the specifications allow voting systems to be unreliable,
and the vendors build to the specifications.
The deficient specifications appear to be based on the authors of the
FEC standards having placed a much higher priority on minimizing the
cost of testing than on ensuring that voting systems are reliable. The
pitiful 163 hour MTBF appears to be based on an approach of capping the
level of testing at what is being done for other purposes (so that the
reliability testing can be concurrent with the other testing), and
setting the reliability specification at what can be accommodated under
the testing cap.
A similar strategy was used and is more clear and blatant in the area of
accuracy. The standards set a "target" of one error in 10 million
ballot positions, but only test to a requirement 20 times worse.
We can now understand all the reported reliability problems with voting
systems, so bad that in at least one case an entire election (in North
Carolina) will have to be rerun. If you don't mandate quality, you
won't get it.
Stan Klein
--
comment-memo-5.3.2.pdf