Re: BSDstats: New High Water Mark: 25 000+ Hosts Reporting In

["Murray Stokely" <murray%stokely.org@localhost>]

[BCCed others]

On Mon, Nov 3, 2008 at 3:12 PM, Don Witt <witt%cylogistics.com@localhost> wrote:
> This is very cool.  What is up with NetBSD and OpenBSD.  Are these numbers
> accurate?

These numbers represent the number of people that have installed a
program to report usage, and are almost completely uncorrelated with
actual usage rates.  There needs to be a big warning at the top of
these very misleading reports.  Also, the massive spam to many
different lists is not ideal.  Any guess at actual "usage" numbers
would be many orders of magnitude larger, and PC-BSD would be a
rounding error to other BSDs used in large hosting environments
(nothing against the great work being done by PC-BSD, just a fact
based on desktop vs server focus).

I continue to believe that sending out these numbers which massively
undercount all operating systems is very counter-productive, but I've
said that before.

If you want to get better numbers you could try to survey all web
servers on the internet, identify the host operating systems by server
responses, tcp/ip timing characteristics, or other heuristics.  You
could alternatively mine google analytics  / webserver log data for
large websites if you want client numbers, or countless other data
sources that would give you far more data than this self reporting
mechanism, and with a much better sample than the very biased
mechanism used for these numbers.

                - Murray