__/ [ Larry Qualig ] on Monday 13 March 2006 17:19 \__
> Roy Schestowitz wrote:
>> __/ [ Larry Qualig ] on Monday 13 March 2006 16:40 \__
>> > And how do we tell? Not that anyone here has the resources to
>> > personally determine this number but how could/should this number be
>> > computed.
>> What is a "Linux user"? How do you handle or count dual-booters? What
>> about a system administrator with an entire Linux cluster? Or one with 3
>> Linux boxes in the house? What about legacy hardware? If it's a 486, does
>> it count? Judging by figures I see in my Web stats, only 70% of the world
>> is using a flavour of Windows. The number of ways to identify Linux leads
>> to confusion, especially if you rely on user-agent strings. Let's not
>> forget that the majority of servers run *nix.
> Part of the question asks what a "Linux user" is? My opinion is that
> "users" should be people. Clearly there are two distinct numbers one
> could get. There are people-users and there are Linux-boxes. Dual
> booting should count as one of each. Each being whatever OS's that box
> can boot. But it should only count if it's actually being used that way
> meaning that if it's setup to dual-boot X and Y but one of the OS's
> only gets used 10% of the time then the figures should reflect that.
It doesn't account for actual /usage/. Does the number of heads count? Are
all people counted as equal? Which partition(s) is(are) used more often? How
long is it (/are they) up for? What happens when the computer is idle? Does
actual bandwidth or CPU cycles come into play?
I am not running away from you enquiry. I try to convery the fact that
different settings for such surveys will produce and lead to different
conclusions, whatever they correspond to.
Example: Symantec declared Firefox to be safer than IE a few days ago.
Previous figures suffered from many issues. They were not comparing like
with like (or "apples with apples", as they call it). There remain issues
like the ability to find flaws in Open Source software, as opposed to
identifying vulnerabilities in binaries, as well as the labelling of flaws
by the vendor that controls the product.
Also comes to mind: the vulnerability count in Windows and Open Source
platforms, which considered and aggregated all bugs (even identical or
related) from different variants. Even Apache was considered among the core
components. Go figure...
>> > The typical arguments made are that you can't use dollar figures
>> > because Linux is free. You can't use downloads because someone
>> > downloads a distro then makes several copies of the CD or installs the
>> > same CD on multiple computers. Other issues that complicate this are
>> > that an ISP can't simply use IP addresses because NAT and routers can
>> > and will translate multiple computers into a single IP addresss.
>> What is the answer so important to you? I am not putting your motives to
>> doubt, but I am genuinely curious. Do you seek acceptance? Are you afraid
>> that being a minority makes you weaker? Is it that "if fewer people choose
>> x, x must be worse"? Don't go by the cattle effect. History indicates that
>> it is a bad approach by all means.
> Nothing at all to do with seeking acceptance, being in a weaker
> minority or cattle effect. If anything it's more of a curiousity
> factor. There are all sorts of numbers and stats being thrown around
> here all the time. (Ex. IBM to migrate 5% of its workforce.) So what is
> the current usage state of Linux? Not that it matters to me but to
> someone like Dell computers it *DOES* matter. To a company selling a
> video/audio/network/video-capture/etc. card it does matter. They want
> to know if the market large enough to warrant supporting Linux? So in
> this regard how close the user base is to achieving "critical mass" is
> Any numbers published here are usually dismissed as being inaccurate.
> (Ex. Recent thread that had a "Linux user" counter and the number of
> registered users.) So the question I'm asking is what is the most
> accurate way of determining the *actual* number.
>> > So what is the best (read: most accurate) way of determining the real
>> > number of Linux users out in the world? Page hits at a generic
>> > web-site? Is this accurate and what is a "generic" web-site? Cnn.com,
>> > eBay.com, Google.com? Personally I always thought that email addresses
>> > are a better way of doing it. It won't tell you the exact number of
>> > systems but it will give you the number of users.
>> Linux appeals to IT-savvy people the most. Thus, you can't measure it
>> based on Web statistics or even E-mail traffic. Some people handle volumes
>> an order of magnitude higher than others, in terms of volume. Capacity is
>> often limited only by the tool at hand. One example among many: use IE,
>> forget about tabs -> lower bandwidth.
> I wasn't proposing measuring email/web-stats based on volume, but based
> on distinctness. One email-address gets one point. Whether you send
> 10-million emails or 10, you are only counted once.
What about people without an E-mail address? I know such people, my young
brother included. What about dead E-mail addresses? Or multiple E-mail
addresses (I have about 20 active ones).
>> > My email for example is accessed from Windows and Linux machines. This
>> > would count as "one vote" for both platforms. Perhaps a more accurate
>> > way would be to look at the *ratio* of emails that I send with one OS
>> > vs. the other and do a 60-40% split or whatever the number may be. The
>> > added benefit to using email is that we're really interested in "users"
>> > rather than "machines" so one (1) user with 4-5 machines would only be
>> > count once.
>> Again, why should this matter to you? If has no purpose to be gauged.
>> Statistics are meaningless and I think Culley's 'statistics' are a valid
>> proof of this on occasions.
> Again, for me it's a curiousity thing and not even all that high on my
> list. But let's put it this way... if there someone posted an URL that
> had the "actual and accurate value" of this number... I'd click on it
> to find out. Wouldn't you?
I would if the number has a meaning. My definition of Linux users will differ
from the perception of another. I haven't the slightest doubt about it.
Roy S. Schestowitz | Gas, brake, honk! Honk, honk, punch! Gas, gas!
http://Schestowitz.com | SuSE Linux ¦ PGP-Key: 0x74572E8E
5:25pm up 5 days 10:02, 7 users, load average: 0.52, 0.58, 0.61
http://iuron.com - next generation of search paradigms