Is the Sarasota ES&S source code audit on the wrong track? Am I confused regarding

[philb]

what they are asking to do? Would what they are asking for be useful?



My understanding of what is going on with the source code audit may be wrong, but I don’t understand why this would be useful if we are talking about a general source code rather than for the programming on the machines where the problems occurred.

The problem does not appear to be with a general source code but with something specific to the machines where the undervotes were happening. It did not affect all machines. And it was not due to the voters voting on those machines as the SOE has suggested.


The Ballot Definition Files and PEBs are programmed individually. Checking a general "source code" will not tell you what was going on at the machine where strange things were happening.

Based on lots of experience all over the U.S. , the switching and disappearing votes was not in general due to voter error, the machines doing the strange things were doing it for all voters until they were reprogrammed or taken out of service; and it wasn't because a general source code used for all machines was problematic.

The problems only happened on select machines. And the cause of the problems on those machines can only be determined by auditing the screen layout, BDFs and PEBs, compiler, etc. for those machines. A general source code that was said to be used for all machines is not likely to tell you much.

If that is what we're talking about.



An official expanation of "what caused the problem"
http://www.heraldtribune.com/apps/pbcs.dll/article?AID=2006612050604

that analysis may be all or part of the story, regarding problematic ballot design.

But are you convinced that the voters who said they saw other designs were mistaken? I heard it was known that there were more than one ballot designs, and voters reported more than one design.

It would have been easy to determine if that was the case, Check the screen layout and BDFs and PEBs and compiler programming for each problematic machine.

But someone seems determined to not do that, but speculate instead. That’s an old story repeated elsewhere.

But for certain, the ballot design was not all of the story,

the reports by voters of "disappearing votes" was consistent with well documented occurrences of similar throughout the U.S.
and in S. Florida in 2004 and 2006 for example.

Those had nothing to do with ballot design. That kind of thing is caused by programming.

Some places had Defaults to a person or blank programmed in that were hard to override, so those who weren't extremely diligent lost their vote.

Note: default to blank is a fairly commonly used programming method that gives the right result if everyone is able to override the default. But this may not happen. I expect that Sarasota machines did use the default to blank design, but was the default hard to override on some machines and if so why?

But some just had misprogramming - either intentional or glitched

Its not clear which it was in Sarasota, but that could have likely been determined fairly easily if the effort had been made.

A real audit of the machine that is "disappearing" votes will almost surely determine what the problem was.
Is there no plan to make this happen??

[Wilms]

1. I think you're correct.

Not that I'd discouragre a look at source code. But the lack of a review of the the Ballot Definition File is curious.

I wonder if Jennings, et. al., have asked for it.

[FogerRox]

2. What about the extra machines that showed up?

And were they the machines that suffered from the undervotes....

[OnTheOtherHand]

3. the undervotes were ubiquitous

They were higher in some precincts than others, but most precincts had double-digit undervote rates.

[philb]

5. 3 precincts had over 25% undervotes; 6 more had over 20%, most had under 10%

There was something badly wrong in the precincts with over 25% undervotes,
and I don't think it was ballot design
There was at the minimum extreme negligance by officials/poll workers if there was this much problems and
the voters weren't warned/helped.

And over 25% undervotes in a precinct might have been 100% for some machines;
not all the machines in the precinct would have had the same pattern.

But similar for the 6 precincts with over 20% undervotes;
some machines likely had extremely high UVs

There was something very strange going on there

http://www.heraldtribune.com/apps/pbcs.dll/section?CATEGORY=Election03

[philb]

7. Apparently there were 5 precincts with over 25% undervotes in Dist 13

[OnTheOtherHand]

8. well, those numbers aren't correct

It's a nifty map, but since the symbols overlap, it's not a great way to estimate descriptive statistics.

Overall, out of 156 regular precincts, there were five precincts with FL-13 undervotes over 25% (all ballots together) -- eight with election day undervotes over 25%. There were 48 with overall undervotes under 10% (35 on election day). The medians were 12.4% overall, 13.0% on election day. So, as I said, most had undervote rates over 10%. By the way, the undervote rate in early voting was even higher.

Apart from one precinct that had 0% undervotes -- but only about 8 total voters! -- every precinct in Sarasota County had an Election Day undervote rate larger than the overall undervote rate for absentee ballots.

(These figures are ripped from an unofficial county canvass from election week, so they may have changed a bit, but not much.)

EDIT TO ADD: I'm not even trying to say what these numbers mean -- just sharing the numbers.

[philb]

26. 8 precincts had over 25% UV on election day and 13 precincts in Early voting

but the Early voting counts were dispersed to the precincts from a fewer number of
early voting sites- the largest undervotes occurred in early voting.

Were warnings about problems with the Dist 13 race the reason for the slightly better results on election day vs early voting.
If so, apparently most poll workers didn't do enough to get the voters attention-
for the problems that the voters could have resolved by more diligence.


12 more precincts had over 20% undervotes on election day
and 35 more precincts have over 20% undervotes in early voting

only 22% of precincts had less than 10% UVs on election day
and 6% of precincts in early voting

So its clear that there were few precincts where the race was
counted in a reasonable manner. SOE & ES&S clearly botched the election- either
negligently or intentionally
And its clear from the data and audit results that the majoriy who turned out to vote in this race favored Jennings.

I think Buchanan should concede- that result should be inevitable.

Its a virtual certainty that the cause of the disappeared votes can be determined if
there is a desire to do so by officials. To this point, they seem to be resisting the steps needed to reach the obvious conclusions however.

[FogerRox]

10. try this video

http://www.therighttocount.com/shorts

It should be noted that early vote locations are fewer than on election day.

http://www.srqelections.com/results/gen2006pct.htm

So these results should be looked at in at least 2 ways. EArly votes were cast on a different group of machines @ fewer locations than voting on Election day. In the above video pct 22 is mentioned, pct 22 has a very high early vote undervote rate & a hi turout overall, in excess of 51% IIRC.

[OnTheOtherHand]

13. and the early undervotes were even higher

I suspect that if there were huge variations across machines within precincts, I would have heard about it by now, but maybe not. (I'm not sure what the absence of such variations would prove; just curious.)

Dunno if you've seen this: http://macht.arts.cornell.edu/wrm1/htresponse.pdf

[FogerRox]

15. I am not sure I made this clear....You do understand

That the Early voting structure was probably 15 locations, not precinct oriented. vs Election day voting.

[OnTheOtherHand]

17. yeah, but I know my response was cryptic

I probably should've said "precincts/locations," but that would have been even more cryptic!

I think I saw something like seven early voting locations in Sarasota, FWIW.

The canvass does report early votes by precinct although people don't vote that way, which raises the question: if the problem was (in part) specific to particular machines, would we expect early-voting undervotes to be more uniform across precincts because (I'm speculating) in many cases they were actually cast on the same machines? Of course the early voting site in umm North Port probably drew mostly North Port voters.

[philb]

18. Is there data available by early voting locations?

[FogerRox]

20. None that I am aware of.

[philb]

16. In other races and counties with switching, disappearing votes- it occurred mostly on certain machin

es

And those machines were doing similar things for all voters

switching from one candidate to another
switching to blank (delayed switch to blank in some cases)


In Sarasota, some voters reported that the votes were disappearing
and others reported that the votes were switching from Jennings to Buchanan

That is clearly a programming thing, not a matter of misleading ballot design

and apparently not all of the machines were programmed the same way

I suspect that the machines with programming that switched or disappeared votes, in additional to the apparently
deliberately deceptive ballot design (according the the expert who testified-Stewart)
these were the ones with the high undervotes

The machines that just had deceptive ballot layout were the ones with moderate undervotes

Seems they thought they needed multiple advantages in this race

Its unlikely all of the machines had the same ballot layout and the same programming

[FogerRox]

21. Good reason to look at the Individual DRE tapes.

Dontcha think....

[philb]

14. what are the legal implications of using machines that were not zeroed out before using?

What law or rule requires zeroing out the machines? In Florida? In U.S.?

How can the number of votes on those machines be certified if not properly zeroed out?
What is supposed to be done when this occurs?
Who was at fault?
What has been the result?

Were there other precincts where the machines used were not zeroed out before using>?

[philb]

4. I didn't hear about extra machines; what are you talking about?

[FogerRox]

11. watch video in post #10

[Stevepol]

6. K&R

[eomer]

9. I think the source code is the more likely place to look.

Of course, they should look at everything until they find the answer. The reports from voters are overwhelming evidence that at least some machines were malfunctioning so they need to keep looking until they determine the cause and the ballot definition file is certainly one of the places they should look.

That said, the reported behavior does not seem to me like a problem with just the ballot definition file. Voters reported that the result that showed on the review screen did not reflect the selection they made on the voting screen. If there is any way you can configure the BDF to produce this result then the software program is defective -- the software (if it were not defective) should not produce that behavior no matter what the election worker puts into the BDF.

Also, the fact that the result on the review screen would eventually match if the voter redid the selection one or more times is another indication of something weird in the software. Otherwise the software would produce the same behavior (right or wrong) every time the voter progressed from the voting screen to the review screen. Once again, there should be no combination that a worker can enter into the BDF that would produce this kind of erratic behavior. If you can trigger this kind of behavior by way of some entry in the BDF then the software is defective for giving that ability.

[FogerRox]

12. Look at the video in my post #10

Machines not zeroed out to start the day.

[philb]

19. Based on other races and counties, not all machines switch and disappear votes

and so the problem is not likely with a universal source code, but specific to certain machines

In this case it appears that all machines had a deliberately deceptive ballot design (as described by the expert from MIT- Stewart)

But only certain machines likely had programming that was switching or disappearing votes
and some machines were swtiching (to Buchanan)
and some machines were disappearing votes (switching to blank)

[eomer]

22. There are various possible causes for only certain machines

switching and disappearing votes.

A few possible causes:

1. The malfunctioning machines were running a different version of the software.
   
2. The software was the same on all machines but had a secret trigger that was activated only on the malfunctioning machines.
   
3. Some other software component or script was run on only the malfunctioning machines and it changed the software binary instructions in memory.
   
4. Some other software component or script was run on only the malfunctioning machines and it changed the software binary instructions on disk. Another script may or may not have been run after the election to change the instructions back in order to cover their tracks.
   

It's interesting to think about where the cause may have been but the bottom line is they need to look everywhere until they find it. We know that certain malfunctions occurred. Unfortunately the Dept of State has already told us they're inclined toward a conclusion that the machines worked fine so it seems they are likely to stop as soon as they've done enough looking to generate corporate news media stories that say they looked into it and the machines were fine. They don't seem to be determined to keep looking until they find the cause(s).

[philb]

23. Still seems to me just checking the generic source code wouldn't resolve most on your list

[eomer]

24. Yes, we agree.

Just checking the generic source code is not enough. In addition they need to check whether the software running on machines that malfunctioned is the same version as the escrowed source code, they need to check the machines for other software components or scripts that should not be there, they need to devise more varied test scripts in order to duplicate the known malfunctions, and they need to do a bunch of other things to try to discover what caused the malfunctions. They need to look everywhere until they find the cause(s).

[philb]

27. An expert, Dan Wallach, confirms these needs and the fact they haven't been allowed so far

He says that the following access to equipment and results is needed to resolve the
cause of the undervotes

Dan S Wallach states that to resolve the question as to why voters were reporting "disappearing votes" and undervotes were extremely high in some precincts,

experts need to look at the "event logs" and "ballot image logs". This has not been allowed to date apparently. he says the logs need to be provided in digital form.

He also says that an effective audit would require digital copies of the ballot style style files for all nine ballot styles used in Sarasota County; and
every flat file that is loaded onto the iVotronic as part of the "ballot programming process", both in early voting and on election day(if different).

He said that at a minimum for an effective audit, you would need at least the machines which experienced the highest undervote rates during the election, along with their carrying cases, power adapters, and other related apparatus.

He also said that two supervisor PEBs, along with nince regular PEBs(configured like the 9 ballot styles), plaus a standard ES&S Communicatons Pack with related equipment would be needed.

And also access to the ES&S source code.
He said that he's worked with proprietary source code in audits before and any issue regarding that issue can be dealt with by signing non-disclosure agreements.

[philb]

25. No one has answered my question: what are legal implications of not running zero tape?

before the election. I thought this was a requirement of someone?
state? HAVA?

Is anything being done about the situation noted in the tape posted?


Also, does anyone have more definitive summary data on the precincts undervotes?
how many precincts had over 25% and over 20%
in total?
on election day?
in early voting?