Could Linux be more in use than IDC or Gartner think 6808

mlw

...

What about those of us who will be using more than 1 OS next year: I'm intending to rejig my home network slightly so that my usage with be spread more evenly between FreeBSD, Windows and Linux. (Currently one Linux machine is used the most as it's running local services that I intend to shift to the FreeBSD machine.)

...

That's the problem with everyone who has a point to make.

I take all polls with a pinch of salt. May I suggest you read "How to lie with Statistics" by Darrell Huff (ISBN 0-14-021300-7); then you may not have so much trouble dismissing it.

For a more accurate poll, you need to randomly select a cross section of your universe (in this case: computer users in whatever countries in which you are interested).

By putting up a poll and asking visitors to vote you are putting in a big bias to begin with: you have automatically rejected all those who don't, or can't be bothered to, vote, or aren't vocal enough to be bothered to give their opinion, etc.

If I was to put up a poll asking "if visitors never opined in a poll" with two options 'Yes' and 'No', for complete accuracy there would have to be 0% who voted 'Yes' - leading to the conclusion that 100% always do - but that was a biased sample as those who don't opine have deselected themselves from responding! (And a biased question as anyone who never does opine in a poll will not opine in that one.)

Where you advertise your poll site is also another bias.

You're also having to reply on the truthfulness of your respondants - as do Web sites that use the Agent string of the browser to collect their stats.