It’s all part of M$’s technological evangelization. They will do whatever it takes to promote their stuff and denigrate others. Money can buy a lot of “X recommends That Ohter OS”.

They nicely collated the quotes proving Miguel de Icaza was one of the problems and is attempting to blame others for his failure to manage.

Gnome illness that started under Miguel de Icaza rain. Its meeks who is tasked with sorting it out.

Really is sad the people Microsoft will issue a MVP to.

If they are not monitoring the sites students will be visiting. So 100K machines in a school don’t exist along with many other groups.

Since there is not a enough coverage web numbers when you allow for the possible statistical error due to lack of data the number could be almost anything. In fact Microsoft could be losing and get the numbers the way they appear on web numbers. Basically worthless.

Statistical error is not the problem. They are consistently biased towards business use. That’s why Google shows up with ~10K PCs to move California when school divisions with 100K GNU/Linux PCs count as nothing.

Large proportion of traffic. Lets take 1 percent of the USA population could be the upper-class.

Statcounter is one of the bigger ones doing this. Even they are not large enough that there numbers can be trusted.

It could be like saying almost no one speaks mandarin because you don’t have mandarin sites in your counters. We would know this was false straight way because we have ground surveys that cover complete populations over what language they use.

No country wide census has every asked the question what Computer OS do you use. Some have asked if you have a computer. Maybe at some point in the future we might get some areas of usable numbers if the question gets on a few country wide census forms.

In fact I don’t trust any web counter. Web counter only tells you what you customers are using that is about it. Statcounter and others are doing a lot of reaching claiming their numbers tell something about the market. The base facts are not there to support that claim. Statcounter can track a change in the population of users they monitor. That does not mean that population is a representative. Detect a change yes. Know what is truly going on not possible.

3 million sites does not mean you will get correct numbers. How many websites are there.

620,132,319 by last netcraft.

Hmm 3 million sites is a sampling from less than half a percent of all the web sites in existence.

193,533,160 are recorded active by netcraft. So at best 1.5 percent sampling of active sites. With such a small sample of sites the margin for error is huge Ted.

Like you would not call a election won when only 1.5 percent of the vote was counted.

So from so possible statistical error is huge. So the allow for possible error. 1-2 percent to 90 percent. Linux is somewhere in their. That is how big the error could be due to the sample size being way too small on web numbers.

The reality is web stats are worthless because they are errored in collection. They don’t collect enough data from enough different sources.

So Linux could have a 20 percent market share and since you sampled 3,000,000 million non FOSS sites you see 1 percent. Yes this is more than possible with how small the sample sizes are.

Robert Pogson his numbers are small there is the problem.

Impressed by large numbers? I am not, particularly if they are meaningless. The first step in measuring anything is to know what you are measuring. NetApplications and others don’t tell us anything about what kind of sites those are. They could be M$-specific in large part, you know, “best seen at 800×600″ or such. Heaven knows M$ has tried to lock the world in in many ways.

Consider a university’s computer science department in Singapore. We know who will visit there, likely students, potential students and researchers. So, their stats will mean what those folks use.
Lo! and Behold! 15.7% GNU/Linux. How is that possible in Ted’s world where GNU/Linux is going nowhere? What are those students thinking? MacOS 8%. How can that be?

The incontrovertible signs you’ve lost the argument.

I take the 1-2% thing back – desktop Linux has dropped below “OTHER” worldwide. It’s no longer even worth its own line on the graph. Lumped in with Windows 9x, Windows 2000 and lowly Amiga. How the not-so-mighty have fallen.

Brazil again? Why not mention some larger markets, such as The United States? Is it because desktop Linux is below “Other” there too, even with Sunnyvale propping it up?? Or Europe? “Other” again. Gains in small markets will not affect the worldwide figures. Why not mention that the whole of South America is still only 1.1%? Gains in Brazil don’t even affect those of South America.

“Don’t say NetApplications, because they don’t publish proper stats just meaningless numbers. Who are the clients visiting the sites and what are the sites?”

I am reminded of what you so politely requested of me earlier. I will not, however, descend to your potty-mouthed and insulting level.

http://gs.statcounter.com/factsheet

“our stats are based on a geographically diverse sample of over 15 billion page views per month across more than 3 million global websites. This means we have large enough sample sizes in the majority of countries to be confident that our stats are broadly representative of real world figures” [emphasis mine]

3,000,000 websites.
Three and six zeroes.
THREE MILLION

For the hard of understanding, that’s a lot of websites. Rather too many to list, I’d say.

THEY STILL CAN’T ALL BE WRONG.”

without providing any, “You have no proof whatsoever of the original premise – that Linux installs outnumbered Mac OS in 2003.”

Shit. Read the web, twit.

Computerworld: “Linux captured the No. 2 spot as desktop operating system in 2003,” said IDC analyst Dan Kusnetzky in a recent interview.

Not all web stats report GNU/Linux as 1-2% as you can easily see by finding some:

A Brazilian site, http://www.areaseg.com.br is health-related subject matter not related to IT at all and the stats show 8.95% GNU/Linux. That’s cumulative since 2006 so it does not show current rates, just the share in the last six years. Since that year, the government has done much to promote GNU/Linux in the schools, government offices and by tarrifs.

Here’s another, a business that sells juice. It’s not much of a page, just an ad and a portal. That probably shows business usage at 9.5% GNU/Linux. So, where it the ~1% number for Brazil coming from? Don’t say NetApplications, because they don’t publish proper stats just meaningless numbers. Who are the clients visiting the sites and what are the sites?