[KLUG Advocacy] Linux Outpacing Macintosh On Desktops

[Robert G. Brown]

I read the article and found that this whole business about Linux having
a larger number of seats (or market share of desktops) was only mentioned
once, in passing. The article as a whole has a lot of [other] interesting
things to convey to readers.

Bruce wrote:
>> Personally I don't put too much stock in those numbers anyway.  It seems
>> like it'd be very difficult to make number of Linux estimates since it's
>> free and legal to copy.  What are they going by?  

This is always a lot of fun. I've been involved in a number of discussions
of this nature, and it's very subjective stuff. There are lots of ways to 
sample this if you're serious about it. IDC is one of the serious samplers
of these things, not that everyone ought to take what they're saying as
the Revealed Word, but it isn't stated (even inpassing) frivolously.

There are three problems in estimating this stuff:
 1. How many copies are distributed?
 2. How many copies are installed?
 3. How many of the installed copies end up being removed?

How this is determined depends on the product, means of delivery, etc.

>>Results about browser info from web server logs?  
I suspect not, unless they're looking at the number of Linux dowlnloads.
You can bet that if you know about the problems, so does any serious-mined
sampler. 

>I prefer stats that are generated more like a neilson rating statistical
>analysis.  These are at least slightly honest about the +/- of being
>accurate.  
This is probably better than most informal surveys, not as good as a well-
worked out sample. The Nielsons work best where there are only a few
choices and a fairly undiferentiated market; these characteritics are 
probably don't apply to software in general (although it mightbe true of 
OS'es).

>I have not seen any numbers recently though.
No point unless you can have some confidence in how the numbers were
developed. I will bet (small amount of money or a commodity) that any
study good enough to be accurate is also proprietary.

>Trying to get a 1000 person slice of computer users has got to be a 
>nightmare to do.
Why? It's doesn't appear to be any harder than what Gallup or Neilson
does when rounding up a similar number of TV viewers or Democrats.

>Consider home and office users?
Yes, and home-office users, too. There are other slices to take, too.
Often (gaming is perhaps a good example), the choice of OS really depends
on other factors, and that has to be taken into account as well. A good 
sampler does some small scale studies on an isolated population, then
includes that on a broader survey and applies some statistical techniques 
to distill the numbers out of the aggregate.

>Do you count me as 2 for home and office??  
We all count you for a lot, Dirk! :)
OK, OK... at least two, maybe more, in an agregate survey. When you fill 
out the form, tell us, unless you have to sign an NDA. :)

>Depends on the poll, desktop users vs. desktop systems.
Yeah, you have to read the fine print carefully. A number of these studies
are pretty conditional.
							Regards,
							---> RGB <---