Posted on 11/05/2004 7:47:45 PM PST by RogerWilko
Friday, November 05, 2004
FORT LAUDERDALE — It had to happen. Things were just going too smoothly.
Early Thursday, as Broward County elections officials wrapped up after a long day of canvassing votes, something unusual caught their eye. Tallies should go up as more votes are counted. That's simple math. But in some races, the numbers had gone . . . down.
Officials found the software used in Broward can handle only 32,000 votes per precinct. After that, the system starts counting backward.
Why a voting system would be designed to count backward was a mystery to Broward County Mayor Ilene Lieberman. She was on the phone late Wednesday with Omaha-based Elections Systems and Software.
Bad numbers showed up only in running tallies through the day, not the final one. Final tallies were reached by cross-checking machine totals, and officials are confident they are accurate.
The glitch affected only the 97,434 absentee ballots, Broward Elections Supervisor Brenda Snipes said. All were placed in their own precincts and optical scanners totaled votes, which were then fed to a main computer.
That's where the counting problems surfaced. They affected only votes for constitutional amendments 4 through 8, because they were on the only page that was exactly the same on all county absentee ballots. The same software is used in Martin and Miami-Dade counties; Palm Beach and St. Lucie counties use different companies.
The problem cropped up in the 2002 election. Lieberman said ES&S told her it had sent software upgrades to the Florida Secretary of State's office, but that the office kept rejecting the software. The state said that's not true. Broward elections officials said they had thought the problem was fixed.
Secretary of State spokeswoman Jenny Nash said all counties using this system had been told that such problems would occur if a precinct is set up in a way that would allow votes to get above 32,000. She said Broward should have split the absentee ballots into four separate precincts to avoid that and that a Broward elections employee since has admitted to not doing that.
But Lieberman said later, "No election employee has come to the canvassing board and made the statements that Jenny Nash said occurred."
Late Thursday, ES&S issued a statement reiterating that it learned of the problems in 2002 and said the software upgrades would be submitted to Hood's office next year. The company was working with the counties it serves to make sure ballots don't exceed capacity and said no other counties reported similar problems.
"While the county bears the ultimate responsibility for programming the ballot and structuring the precincts, we . . . regret any confusion the discrepancy in early vote totals has caused," the statement said.
After several calls to the company during the day were not returned, an ES&S spokeswoman said late Thursday she did not know whether ES&S contacted the secretary of state two years ago or whether the software is designed to count backward.
While the problem surfaced two years ago, it was under a different Broward elections supervisor and a different secretary of state. Snipes said she had not known about the 2002 snafu.
Later, Lieberman said, "I am not passing judgments and I'm not pointing a finger." But she said that if ES&S is found to be at fault, actions might include penalizing ES&S or even defaulting on its contract.
And some FReepers actually use linux, yanno....
No mystery to a programmer. (No excuse, either, but...)
The computer wasn't programmed to count down. That's just the way a computer will work when someone tries to add one more to a counter that is so at its maximum value.
Remember that computers use binary (base-2) numbers, not decimal (base-10) as we do. Imagine a car odometer that is at 999999 and you add one more mile. It will roll over to 000000. Oops.
Similarly, a 16-digit binary number is not uncommon on older hardware and it turns out that this means it will roll over at 32,768 (which is 2^15). Exactly what happens at that point can vary based on hardware, but one possibility is that it rolls over to 0, but then starts counting negative (-1, -2, -3, ...).
Which sounds suspiciously like what happened here.
The problem is that a programmer: 1) used a counter that is too small for the task at hand, and 2) did not check to see if the counter was at its maximum before trying to add one more to it.
This is mathematically possible when the voters were acting in negative space, or in an alternate universe. Say, voting for the Socialist/Progressive/Liberal/Dem named Kerry.
But the whole concept of "free software" is so liberal. Kind of like free health care, etc.
Any programmer worth his salt can figure that out from here.
Change the "signed short" to "signed long", and the problem's fixed (at least until you get a really really big turnout).
Why in the hell can't they just use a simple color in the circle paper ballot like the absentee ballots?!?!?
I wonder if Miss Harris would report voter fraud if it implicated Kerry.
Hmmmmmm. Affected only the absentee ballots, huh?
"0000 0000 0000 0011 = +2"
Oops, actually that's 3, not 2.
amazing. As a programmer, it is obvious that was the problem. It this was done by a government lackey, you could understand it; if it was contracted out, it was probably to an officials son. Broward, isn't that demoRat controlled? That would explain a lot.
Example: the computer reads a value of 25,000. We know that's bogus. We also know it was 'down-counting'. Therefore, the 32,768th ballot would show 32,766 on the meter; the next one shows 32,765, and so on.
Extrapolating this gives an actual count of 40,535. And given the uncertainty of things, I'd still call that +/- 2 votes. but of course there's no way anyone would ever certify such a calculation.
This is a non story. It only affects absentee ballots. The absentee ballots are run through a scanner. They were only supposed to run so many through at a time.
It has nothing to do with the electronic voting machines.
In 2004 there were 695000 votes in Broward county and it also went 2:1 for Kerry.
So in both 2000 and 2004 we have the same ratio of Democrat votes to Republican votes but in 2004 there is an increase of 130000 votes over the year 2000. How the heck those morons are even bringing this issue with machines problems if there are more votes cast in 2004 then 2000.
Delusional liberals will get further delusional. It will be fun to watch them in the next four years.
Somebody better tell this to the DUmmies! They think they got a hold of a HUGE story that's gonna give them back the election as they hand their cash over! HAH!
I know, I know... I just couldn't resist. :)
Incrementing a 16-bit signed int, the sequence would be
0 1 2 ... 32766 32767 -32768 -32767 -32766 ... -2 -1 0
16-bit ints haven't been in common use for about ten years now, ever since the introduction of Windows 95.
These days, the default would be a 32-bit int (goes up to 2,147,483,647). For a critical application like this one, you might use a 64-bit int (up to 9,223,372,036,854,775,807). More important, you would debug it before releasing it! And you should be required to publish the source.
Election Systems and Software is owned by the Omaha World Herald.
They are braindead idiots.
I should know, I've worked 6 elections with the morons...and that's what they are, morons.
It does not surprise me that they limited the ballot count to 32k per precinct. The same idiots also double counted the entire count of Sarpy County, Nebraska, though because Nebraska is a red state, the Dems don't care.
Meanwhile, I was stumbling around in the area of 32K array limits imposed on old languages like Pascal...
It also didn't affect the final total, so what's the big deal? That's the only number that matters.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.