DaveNet: Saturday, April 5, 1997; by Dave Winer.

Jan 14 Survey Spam Attack

The box got stuffed

When I checked the final numbers at about 1AM Pacific on January 15, I was shocked to see how dramatically the percentages had changed. An almost doubling of the vote in the last few hours.

I checked the log of IP addresses, and there was no clear evidence of spamming. But the direction of the results had changed dramatically, and there was no new announcement from Apple or anyone else that could account for a major change in direction.

Checking my email, I had messages from several people questioning the results. One person, mark@digiplace.com, had been following the results thru the evening (I had a dinner meeting, so I was not watching):

I'm sure some shenanigans happened during your survey. I had been following it on and off from the time voting opened, and a steady pattern emerged right away that persisted throughout the day.

That is until from about 7:30 pm PT to about 9:30 PT, when you got an almost unbroken string of anti Mac, anti-Win, pro-Next, pro-Unix, and pro-Be votes. I'm in the pro-Next camp myself, so it's not as though I'm complaining, although I do want the resounding victory we were headed for to be untarnished :-)

Right on Mark! So... Luckily, we had a scheduled script running that saved off the results to a folder every hour on the hour. I went back thru the results and found that the spamming did in fact start after 7PM.

The survey ended at 7PM

So, the wisest choice, I believe, is to say that the survey ended at 7PM, we'll use those numbers as the final results.

To the honest people who voted only once after 7PM: my apologies. Your votes should have been included in the survey, but because some person or people stuffed the ballot box, your votes were not counted in the final results.

It's clear that, if we're going to do another survey, we need to have a qualification process, and password access, that guarantees a single vote per developer. It's not possible to just trust that people will be honest. One clever net programmer can spam the process. We knew it was possible, I had hoped that because it was so easy to do spam the survey, no one would do it.

I hope the spamming incident doesn't become the story of this survey. As Mark points out, up to 7PM the survey was converging on a strong result. We have the hourly snapshots that back up this conclusion. By throwing out the after-7PM results, we got rid of the effect of the survey-stuffing.

Dave Winer