I'm a Windows Sysadmin at a public university. I like to patch & reboot my servers the same day that critical updates are released, but I often run into resistance from my customers. I believe we're all on the same page as to what "critical update" means, but they'd like assurance that these patches have been fully tested before installing them on the production servers. I've always told them that Microsoft does way more QA testing of their patches before releasing them than I could ever hope to do, but I don't really know this to be true.
Could you give us some info about your QA process? What does your security team do, specifically, to test your patches before releasing them to the public? Do you test them on systems that have different 3rd party apps installed, such as antivirus & backup software? Do you have any lists of applications that you always test for compatibility with the new patches?
also, if you have any ideas on how to fix the problem of e-vote machines reporting inaccurate results -- and having no verifiable way to figure out what the actual results were, i'd love to hear them. do you at least acknowledge that this is a problem we should try to fix? i believe that fair elections & accurate results are a core necessity of democracy.
i understand your point. there are some complete morons out there. but i still think, especially if they are training on this specific issue, they will be able to correctly change a roll of paper, or there will at least be 1 person working at each polling station who will be able to figure it out. also, how many voters could they possibly expect at a voting precinct? they could just design the machines to come equipped with a large-enough roll of paper that the issue would never come up.
what of the similar scenarios i brought up? at some point, there's got to be a level of common sense that you do believe everyone has.
and i still fail to see how a claim that "poll workers might not be able to change a roll of paper" is an argument of any weight against using e-vote machines with a paper trail so that we can verify our election results -- especially since we have many documented / proven instances where non-receipt printing e-vote machines reported inaccurate results and we have no way to find out what the actual results were. we're talking about our democracy here - and we have proof of these no-receipt machines reporting inaccurate results. these machines contribute to the overall count and decide who our elected leaders will be. i think the argument for verifiable voting far outweighs any argument against joe public's ability to change a roll of paper. the necessity is real, and unless you can provide a compelling case that requiring a paper trail would make official results even more wrong, i don't see how you even have an argument. the mutual goal here should be to ensure accurate election results.
so the ATM's either detect the error and can continue, or they error out. program the machines to hault until the paper the changed, or let it error out, replace the paper, then reboot. you'd want to fix the later condition in the code, but seriously, come on.
am i willing to take the risk that a poll worker, who has been trained on how to deal with this exact situation, is able to properly change a roll of receipt paper? hell yes. if a poll worker is able to understand enough to explain how to fill out the voter mailing card and to require proof of residence -- as they are trained how to do ahead of time and have successfully done every time i've gone to vote -- i'm pretty sure he/she can change a roll of friggin paper. i mean, everbody is able to change the roll of toilet paper when it runs out, right? and it's not like anybody had to go through a training session on how to do it.
with the chad incident, i'm not sure they had any established procedure. with changing receipt paper rolls, there will be. have some faith in your fellow humans.
it's not "screwing up" when some percentage of the people you talk to all share a common trait. it's actually accurate for your polling place because your method of selecting people is systematic & unbiased. it's "screwing up" only if you try to make claims about the entire state (extrapolate) based on the numbers from your polling place without weighting for known demographics with known voting trends in that state.
until you show that the percentage of people polled who are some_trait is different from the state-wide percentage, and that this group of some_trait people all tended to vote a certain way that differs from a purely random sample of people in that state, there is no reason to ask the question -- you have no reason to believe that weighting that demographic of some_trait people will show any change in your final exit poll results. exit polling has been accurate for decades AND we already know that some e-vote machines reported inaccurate official results. the next logical step is to see if any more e-vote machines from ES&S, UniLect or Diebold reported inaccurate official results -- but we can't because there is no paper trail.
if we had proof that the official results were 100% accurate, then we would go back and look at how the exit poll weights missed some demographic or if a majority of pollers were intentionally not following the interview selection criteria (ie. every Xth person, or however they choose their interviewees)
furthermore, exit polls were out-of-sync with official data in only 3 states: Florida, Ohio & Pennsylvania. In every other state, exit polls matched offical data within their margins of error. so, even if we only interviewed happy people and most happy voters tended to vote for candidate X, and we didn't weight our exit poll data apropriately, why were the exit polls only off in these 3 states?
but forget the exit polls for a minute - my main purpose here is to get people to acknowledge that Diebold, UniLect & ES&S e-vote machines have all been shown to submit inaccurate results. no one's debating that, but that's what matters the most here. this is our democracy. all votes must be counted accurately, and we need a paper trail to ensure this is possible.
how would you be able to tell and why would it matter? terms like "liberal" and "conservative" are subjective, and have no place in an exit poll. the things exit pollers are interested in are gender, ethnicity, and who they voted for. these are all objective data.
You test them beforehand to ensure reliability. You look at how many people are registered to vote in the precinct and make sure enough receipt paper is on hand. You program the e-vote machines to detect printer malfunctions and out-of-paper conditions, and have the machine go to an error condition when/if that happens. You train poll workers on how to handle these minimized risks if/when they actually happen. It's not rocket science and certainly isn't a show stopper.
E-vote machines with a paper trail would be much like ATM machines. Have you heard about problems with ATM reciept printers malfunctioning / running out of paper? I sure haven't.
Or the pollsters eliminated the possibility of "skewing the data" by reporting Male & Female exit polls as separate numbers, and weighting the final results by official M/F proportions in that state. From the PDF:
"If males or females disproportionately participate, it doesnt matter. If the sample were 90% female, female preferences would still only be weighted for their share of the electorate; CNN and others released data as in Figure 1.1 showing male and female preferences separately and their appropriate weight (their percentage of the overall electorate)."
My main point, which I concluded my original post with, is that the e-vote machines used in '04 produced less reliable "official counts" than in previous elections. I've sourced 3 particular instances where e-vote malfunctions have been documented. All 3 of them support my main point. In Pennsylvania, this has already lead to the decertification of UniLect machines in that state. This proves that the demonstated malfunctions of at least 1 e-vote company's machines have lead 1 state to decide they are not trustworthy for our elections.
As for exit polling discrepencies, I only mean to suggest that the demonstrated e-vote malfunctions could be a possible explanation.
All I'm after here is getting reliable voting machines in place with a verifiable paper trail.
We have 2 main "impossibilities" from the Nov '04 election:
- The 662,00 to 1 probability that exit polls were that far off in those 3 swing states.
- The assertion, with no statistical or other supporting evidence, that vote fraud on this massive of a scale is impossible.
What do you believe? Careful analysis of statistics? Or an un-verificable assertion?
A paper came out shortly after the Nov '04 election showing how exit poll data differend from official tallies in Florida, Ohio & Pennsylvania. Exit polls in all 3 states showed a Kerry win. Official results has Bush winning Florida & Ohio, and Kerry winning Pennsylvania by a much smaller margin than exit polling showed. Given the long, accurate-within-a-margin-of-error track record of exit polls, the probability of the exit polls being that wrong in all 3 states is 662,000 to 1. http://www.buzzflash.com/alerts/04/11/ale04090.htm l
And who decides to not vote just because e-vote machines are in use? The method used to cast my vote at the polling station is the LAST thing on my mind when I go to vote.
Recently, UniLect had their e-vote machines decertified in Pennsylvania, thanks to the efforts of 1 citizen who coughed up $450 for a re-evaluation of their functionality. The results were pretty embarassing for UniLect, to say the least, and I'm baffled as to how this wasn't discovered BEFORE the election: http://www.bradblog.com/archives/00001364.htm
ES&S's explanation for the thousands of extra Bush votes counted by their machines in Franklin County, Ohio in Nov '04 was that the card reader they had hooked up their tabulation laptop was sending the data to the laptop too quickly for the laptop to process it, so some data got dropped. This is either a huge lie, or only demonstrates some magnificent incompetence in ES&S's development team: http://www.bradblog.com/archives/00001184.htm Either way, they should also have their e-vote machines decertified. Here's to hoping.
And the fact that Walden O'Dell, chief executive of Diebold Inc, sent a fundraising letter to Republicans in Ohio in 2003 saying that he is "committed to helping Ohio deliver its electoral votes to the president next year" casts doubt on the legitimacy of all reported results from Diebold machines in Ohio in Nov '04. http://www.commondreams.org/headlines03/0828-08.ht m
I realize that nothing that humans do is perfect, but these e-vote machines used in '04 show a definite trend towards "much less perfect" than in previous elections.
Slashdot Mirror
I'm a Windows Sysadmin at a public university. I like to patch & reboot my servers the same day that critical updates are released, but I often run into resistance from my customers. I believe we're all on the same page as to what "critical update" means, but they'd like assurance that these patches have been fully tested before installing them on the production servers. I've always told them that Microsoft does way more QA testing of their patches before releasing them than I could ever hope to do, but I don't really know this to be true.
Could you give us some info about your QA process? What does your security team do, specifically, to test your patches before releasing them to the public? Do you test them on systems that have different 3rd party apps installed, such as antivirus & backup software? Do you have any lists of applications that you always test for compatibility with the new patches?
also, if you have any ideas on how to fix the problem of e-vote machines reporting inaccurate results -- and having no verifiable way to figure out what the actual results were, i'd love to hear them. do you at least acknowledge that this is a problem we should try to fix? i believe that fair elections & accurate results are a core necessity of democracy.
i understand your point. there are some complete morons out there. but i still think, especially if they are training on this specific issue, they will be able to correctly change a roll of paper, or there will at least be 1 person working at each polling station who will be able to figure it out. also, how many voters could they possibly expect at a voting precinct? they could just design the machines to come equipped with a large-enough roll of paper that the issue would never come up.
what of the similar scenarios i brought up? at some point, there's got to be a level of common sense that you do believe everyone has.
and i still fail to see how a claim that "poll workers might not be able to change a roll of paper" is an argument of any weight against using e-vote machines with a paper trail so that we can verify our election results -- especially since we have many documented / proven instances where non-receipt printing e-vote machines reported inaccurate results and we have no way to find out what the actual results were. we're talking about our democracy here - and we have proof of these no-receipt machines reporting inaccurate results. these machines contribute to the overall count and decide who our elected leaders will be. i think the argument for verifiable voting far outweighs any argument against joe public's ability to change a roll of paper. the necessity is real, and unless you can provide a compelling case that requiring a paper trail would make official results even more wrong, i don't see how you even have an argument. the mutual goal here should be to ensure accurate election results.
so the ATM's either detect the error and can continue, or they error out. program the machines to hault until the paper the changed, or let it error out, replace the paper, then reboot. you'd want to fix the later condition in the code, but seriously, come on.
am i willing to take the risk that a poll worker, who has been trained on how to deal with this exact situation, is able to properly change a roll of receipt paper? hell yes. if a poll worker is able to understand enough to explain how to fill out the voter mailing card and to require proof of residence -- as they are trained how to do ahead of time and have successfully done every time i've gone to vote -- i'm pretty sure he/she can change a roll of friggin paper. i mean, everbody is able to change the roll of toilet paper when it runs out, right? and it's not like anybody had to go through a training session on how to do it.
with the chad incident, i'm not sure they had any established procedure. with changing receipt paper rolls, there will be. have some faith in your fellow humans.
it's not "screwing up" when some percentage of the people you talk to all share a common trait. it's actually accurate for your polling place because your method of selecting people is systematic & unbiased. it's "screwing up" only if you try to make claims about the entire state (extrapolate) based on the numbers from your polling place without weighting for known demographics with known voting trends in that state.
until you show that the percentage of people polled who are some_trait is different from the state-wide percentage, and that this group of some_trait people all tended to vote a certain way that differs from a purely random sample of people in that state, there is no reason to ask the question -- you have no reason to believe that weighting that demographic of some_trait people will show any change in your final exit poll results. exit polling has been accurate for decades AND we already know that some e-vote machines reported inaccurate official results. the next logical step is to see if any more e-vote machines from ES&S, UniLect or Diebold reported inaccurate official results -- but we can't because there is no paper trail.
if we had proof that the official results were 100% accurate, then we would go back and look at how the exit poll weights missed some demographic or if a majority of pollers were intentionally not following the interview selection criteria (ie. every Xth person, or however they choose their interviewees)
furthermore, exit polls were out-of-sync with official data in only 3 states: Florida, Ohio & Pennsylvania. In every other state, exit polls matched offical data within their margins of error. so, even if we only interviewed happy people and most happy voters tended to vote for candidate X, and we didn't weight our exit poll data apropriately, why were the exit polls only off in these 3 states?
but forget the exit polls for a minute - my main purpose here is to get people to acknowledge that Diebold, UniLect & ES&S e-vote machines have all been shown to submit inaccurate results. no one's debating that, but that's what matters the most here. this is our democracy. all votes must be counted accurately, and we need a paper trail to ensure this is possible.
how would you be able to tell and why would it matter? terms like "liberal" and "conservative" are subjective, and have no place in an exit poll. the things exit pollers are interested in are gender, ethnicity, and who they voted for. these are all objective data.
The only irony here is that you signed your post as "MrLogic".
You test them beforehand to ensure reliability. You look at how many people are registered to vote in the precinct and make sure enough receipt paper is on hand. You program the e-vote machines to detect printer malfunctions and out-of-paper conditions, and have the machine go to an error condition when/if that happens. You train poll workers on how to handle these minimized risks if/when they actually happen. It's not rocket science and certainly isn't a show stopper.
E-vote machines with a paper trail would be much like ATM machines. Have you heard about problems with ATM reciept printers malfunctioning / running out of paper? I sure haven't.
Or the pollsters eliminated the possibility of "skewing the data" by reporting Male & Female exit polls as separate numbers, and weighting the final results by official M/F proportions in that state. From the PDF:
"If males or females disproportionately participate, it doesnt matter. If the sample were 90% female, female preferences would still only be weighted for their share of the electorate; CNN
and others released data as in Figure 1.1 showing male and female preferences separately and
their appropriate weight (their percentage of the overall electorate)."
My main point, which I concluded my original post with, is that the e-vote machines used in '04 produced less reliable "official counts" than in previous elections. I've sourced 3 particular instances where e-vote malfunctions have been documented. All 3 of them support my main point. In Pennsylvania, this has already lead to the decertification of UniLect machines in that state. This proves that the demonstated malfunctions of at least 1 e-vote company's machines have lead 1 state to decide they are not trustworthy for our elections.
As for exit polling discrepencies, I only mean to suggest that the demonstrated e-vote malfunctions could be a possible explanation.
All I'm after here is getting reliable voting machines in place with a verifiable paper trail.
We have 2 main "impossibilities" from the Nov '04 election: - The 662,00 to 1 probability that exit polls were that far off in those 3 swing states. - The assertion, with no statistical or other supporting evidence, that vote fraud on this massive of a scale is impossible. What do you believe? Careful analysis of statistics? Or an un-verificable assertion?
I beg to differ...
m l
t m
A paper came out shortly after the Nov '04 election showing how exit poll data differend from official tallies in Florida, Ohio & Pennsylvania. Exit polls in all 3 states showed a Kerry win. Official results has Bush winning Florida & Ohio, and Kerry winning Pennsylvania by a much smaller margin than exit polling showed. Given the long, accurate-within-a-margin-of-error track record of exit polls, the probability of the exit polls being that wrong in all 3 states is 662,000 to 1.
http://www.buzzflash.com/alerts/04/11/ale04090.ht
And who decides to not vote just because e-vote machines are in use? The method used to cast my vote at the polling station is the LAST thing on my mind when I go to vote.
Recently, UniLect had their e-vote machines decertified in Pennsylvania, thanks to the efforts of 1 citizen who coughed up $450 for a re-evaluation of their functionality. The results were pretty embarassing for UniLect, to say the least, and I'm baffled as to how this wasn't discovered BEFORE the election: http://www.bradblog.com/archives/00001364.htm
ES&S's explanation for the thousands of extra Bush votes counted by their machines in Franklin County, Ohio in Nov '04 was that the card reader they had hooked up their tabulation laptop was sending the data to the laptop too quickly for the laptop to process it, so some data got dropped. This is either a huge lie, or only demonstrates some magnificent incompetence in ES&S's development team: http://www.bradblog.com/archives/00001184.htm Either way, they should also have their e-vote machines decertified. Here's to hoping.
The Miama Herald also reported this week that their ES&S machines counted more votes than voters in Nov '04: http://www.bradblog.com/archives/00001390.htm
And the fact that Walden O'Dell, chief executive of Diebold Inc, sent a fundraising letter to Republicans in Ohio in 2003 saying that he is "committed to helping Ohio deliver its electoral votes to the president next year" casts doubt on the legitimacy of all reported results from Diebold machines in Ohio in Nov '04.
http://www.commondreams.org/headlines03/0828-08.h
I realize that nothing that humans do is perfect, but these e-vote machines used in '04 show a definite trend towards "much less perfect" than in previous elections.