roy@phri.UUCP (Roy Smith) (03/08/86)
About a month ago, I asked "should net.general be deleted". The stream of votes has slowed to a halt, so I'm ready to announce the results. I had requested that people send mail to gen-vote@phri with a subject line that either read "keep net.general" or "delete net.general". I intended to use a shell file to sort and count the votes using grep and wc. As many of you pointed out (some politely, some less so), since the title of my original article was "Let's delete net.general", you had to change the subject to vote "no". Enough people were confused by this that I counted the votes twice. First, using the shell file, I got the following results: bad 13 delete 77 keep 94 total auto 184 The "bad" category consists of messages which didn't count as a valid vote either way because the subject line was missing or messed up. One poor soul sent a message with a subject that read "delete net.sources". I didn't think my instructions were that confusing! :-) Then, I went back and sorted the messages into "yes" and "no" piles by looking at them manually. Where there was any clear indication in the text of how the person wanted to vote, I used that regardless of what the header said. On occasional messages I had to make judgement calls because there was no clear indication one way or the other. I extracted several messages which were obvious attempts at ballot stuffing and put the in a cheat pile. The results of that count are as follows: cheat 39 delete 68 keep 74 total manual 181 The discrepancy between totals is due to two factors. Some of the cheat messages contained several subject lines and thus got counted more than once by wc (as was inteded by the sender, I assume). Some of the messages had garbled headers and may have been counted wrong, or not at all. The bottom line is that more people want to keep net.general around than want to delete it. In one way I'm relieved at the result; if it had been 2:1 in favor of deleting it, we would then have to start fighting over whether that constituted enough of a majority to do it. Many of the people who voted to keep net.general suggested that net.followup should be deleted instead (and I agree). That's another issue, however, and I don't plan on running a poll on that question. Somebody else feel like taking up the cause? Other suggestions included mod.general, and net.announce (which should really be called mod.announce; I think I remember being told that the name change is in the works) as alternatives for net.general. Some people thought net.misc should replace net.general. I was pleasantly surprised to not see all the "I vote {yes,no}" messages posted to the net. On the other hand, I was disapointed in how many people did not understand my directions for voting. Much of the fault for that, however, must be placed on the procedure I set up. I think automatic vote counting can work and hope that some day the news software will support it in some "official" manner. Until that happens (and it may very well never come about because of the nature of the net), I think there are some lessons that can be learned from my attempt. If you consider my procedure an experiment that failed, it still has value in the analysis of the failure. My biggest mistake (and it was unintentional; I didn't realize the problem until too late) was having the default subject line generated for a reply being the right syntax to count as a "yes" vote. It should be equally difficult to vote either way. One possibility would be to have two lines in the body, one of which must be marked with an "X". Alternatively, you might have the voter delete the line he disagreed with. I was undecided if each vote should be confirmed by return mail. In the end, I decided against it because it would double the mail traffic in and out of our neighboring sites. There might be advantages to sending replies, however. If a person made a clerical error, the reply would tell him how his vote was actually recorded. It would be nice to have some way to have a person send in a correction and only have the latest vote from a particular person be counted. This later bit sounds a bit extravagant, however. I don't know whether to be happy or sad at the number of votes I get. Initially, I expected a much greater turnout. On the other hand, several people have indicated that 180 votes is considered rather a lot for a poll conducted on the net. Some people asked about the geographic distribution of votes. I didn't to any careful statistics, but my impression from scanning the votes by eye was that 90-95% were from the U.S. If anybody would like to get a look at the actual votes (to do statistics or verify the count), please send me a tape with a self-addressed, pre-paid return mailer. The total volume is about 300 kbytes; too much to be mailing around by uucp. -- Roy Smith, {allegra,philabs}!phri!roy System Administrator, Public Health Research Institute 455 First Avenue, New York, NY 10016