Internetworkers: http://www.ibiblio.org/internetworkers/

["Tanner Lovelace" <clubjuggler AT gmail.com>]

On Jan 7, 2008 6:57 PM, Roger Austin <raustin3 AT nc.rr.com> wrote:
>   I have been a programmer/analyst since the early '80s. I don't
> understand what is so complicated about a voting system.

Which pretty much means you haven't written one or even looked at
the challenges inherent in building one.

> I've been
> involved in very complex systems development in a highly regulated
> (FDA, EPA, DEA, etc.) environment so I know a little about building
> systems that are built to be robust.

Just because you've built one complex system doesn't automatically
make you qualified to build a different complex system with completely
different requirements.

>   There is no way that a computerized voting system should be more prone
> to error than a medical device or military system. It may be true that
> some systems crash without warning, but there is usually a good reason.

In general, as I see it (and I will admit, I've never implemented one either,
but I have looked into the challenges needed to overcome to implement one),
is that the biggest problems with implementing a computer based voting
machine are the following:

1. Conflicting directives.  (On a computer, it's generally easier to store
  as much data as you can so you can refer to it later.  However, in voting,
  you specifically do not want to do this.  You want voting to be anonymous.
  So, if voting is anonymous, how do you then later handle recounts?  In
  Diebold systems, recounts were as simple as "reprint the totals stored
  in the db" and that was WIDELY and rightly criticized. )
2. Not having control of the complete underlying system. (The main problem
  with a lot of the voting machines, and, in fact what I believe Dr. Felten 
was
  referring to, is that they're implemented on top of Microsoft Windows which
  is known to be unstable.  Yes, it is perfectly possible to make a
stable system
  if you're taking care to make sure you know everything that is happening in
  the system but by using windows you can not know that.  And, couple that
  with my point #3 and it's not going to be pretty.)
3. Voting machines operate at the extremes of the bell curves. (Most 
programmers
  today don't give a second thought to adequately debugging their
programs.  Open
  source software is even worse!  [Motto: release early, release often
... let the user
  report the bugs.]  Voting machines, however, have to deal with input
from a HUGE
  variety of people, many of them computer illiterate, and the
programmers cannot
  anticipate all the problems that will crop up.  The "drag and drop"
problem is a
  perfect example of #2 & #3 working together to bite you in the ass.)

> If a blue haired lady trying to vote crashes a voting machine, you are
> using the wrong development environment and/or people to build it.
> Plus, you are probably making it much too complex so there is more
> risk to failure in the system.

I actually agree with this.  I *do* think the system is "much too complex".
In fact, I will go so far as to say that Direct Record Entry (DRE) voting
machines are completely too complex and should be replaced with optical
scan machines.  Optical scan is much, *MUCH* less complex and less
prone to failure, which, I believe, was one of the points of the article.

>   Crappy software written on any OS will stink. I sounds like the
> vendor management looked at this as a revenue stream and ignored
> the developers. Nah, that never happens...

Again agreed, but I think the cost to build a decent system would
have been prohibitively expensive.

>   Okay, I am through ranting now. Get back to work.

BTW, from your "rant" above, it looks like you have no clue who
Ed Felten actually is. Considering the fact that he's actually a world
renowned authority on electronic voting machines (google him) and
you're not, your rant comes off a bit lacking.  Perhaps a bit more
digging next time might be in order?

Cheers,
Tanner

-- 
Tanner Lovelace
clubjuggler at gmail dot com
http://wtl.wayfarer.org/
(fieldless) In fess two roundels in pale, a billet fesswise and an
increscent, all sable.