GNU/Linux Makes Another Move In Canada

This week has been phenomenal for GNU/Linux page-views on StatCounter. How about that, rising from ~1.8% to 2.58% in a week? A year ago, GNU/Linux was at 0.88%. I’ll take it… This level leaves USA behind and is quite competitive with Europe. It’s all good.

I wonder how big the “May Blip” will be this year.
2013 – 4.53% on 8 May
2014 – 7.46% on 24 May
2015 – ?
I’m thinking ~10%, about four times the average value at the time. Time will tell. 😉

UPDATE I updated the graph to reflect another couple of days’ climb. That’s five of the last seven days climbing. There’s only a few percent chance that’s random.

UPDATE It seems the rise lasted a week. In one week the average daily level changed from 1.8% to 2.2%. It’s a huge migration or shipment of GNU/Linux PCs.

So, the change has persisted. Canada is now at a new level for GNU/Linux page-views. It wasn’t a statistical fluctuation but a measure of a real movement. Way to go, Canada!

About Robert Pogson

I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.
This entry was posted in technology and tagged , , , , , , , , . Bookmark the permalink.

51 Responses to GNU/Linux Makes Another Move In Canada

  1. oiaohm says:

    Interestingly, that was the precise strategy that Linux used to demolish AIX, HP-UX and Solaris. All three of which were better operating systems, btw.
    Cost and audit-ablity.

    So, Fifi, what then? How is Linux supposed to cannibalise an OS product that is equally free at the point of sale, and yet is considerably more familiar to the consumer?
    This arguement presumes Linux is not already doing this. The on going growth of Linux yes slow says its already doing this. The question is more what is required to speed it up.

    BotNets, as oiaohm suggested?
    You miss read me. BotNets explains some of the spikes. Not the on going growth. The massive growth from the Feb 20 lines up with superfish in fact even the countries showing the biggest alterations are countries superfish had contracts to show advertising in and the countries going along status normal superfish did not have any contracts for advertising. So a percentage of Microsoft numbers are infection question is how much.

    https://www.shadowserver.org in there blogs cover how botnets are taken down DrLoser. 24hours is the normal method.

    Remember for a Long time Linux users said the less than 1 percent was bogus. Its looking more and more the only reason Linux was showing less than 1 percent was simply infection of Windows machines.

    By the way DrLoser the change is 0.5 not 0.4. There are 30 other things out there like superfish that Windows defender and other Anti-virus software ignores.

    Its looking more and more Linux has been around 2 percent+ for a long time. The false data causes by webstats and infections has resulting in software not being released for Linux.

    No, you’re absolutely right. The basic corporate strategy at Microsoft seems to be to maintain an almost total lock-down on the desktop. What Windows 7 lost, Windows 8 picked up. When XP and Vista dwindled, Windows 8.1 picked up the slack. And when Windows 7 and Windows 8.1 run out of steam, one can assume that Windows 10 will sail blithely on, having cannibalised both.

    In fact Windows 10 will cannibalism both Windows 7 and 8.1 but the result will not be much income for Microsoft. Also more uses are going to portable devices like Android.

    The other reality is very few users have moved from Windows 7 to 8. So Windows 7 has grown to a point and held. Most of the 8 growth has been people replacing XP.

    Something else to remember if 7 and 8 drop and replaced by 10. The result could be nice and interesting with Linux ending up in OS number 4 position. Those positions may not seam important but some CEO are idiots and ask to have product on top 5 operating systems.

  2. DrLoser says:

    It seems the rise lasted a week. In one week the average daily level changed from 1.8% to 2.2%. It’s a huge migration or shipment of GNU/Linux PCs.

    In two weeks? Stupendously unlikely. It would show up somewhere else, other than the much-derided StatsCounter info (derided when they don’t conform to expectation, that is).

    Page views are page views, Robert. They don’t necessarily correspond to boots on the ground.

    BotNets, as oiaohm suggested? Home-based web-crawlers?

    Well, whatever. Bumping up a tiny number by adding 0.4% over a tiny window like two weeks is considerably more likely to be a software artefact than a hardware artefact, if you ask me.

  3. DrLoser says:

    Botnet destruction is normally the command and control server hijacked and the botnet forced to internally self destruct. So 24 hours the botnet dies.

    Any further info available on this particular “going nuts with bull crap,” Fifi?

    I hate to interrupt your weekend. There’s stubble to shave off, fishnet stockings to repair, and I imagine that competition under the lamp-post is quite fierce at commercially sensitive times like this.

    But, you know, just the tiniest smidgeon that your fertile imagination can dredge up in re the half-life of botnets …

    … That’s all I ask, really.

  4. DrLoser says:

    Lets go nuts with bull crap.

    Oh, I wouldn’t restrict yourself to bull crap, oiaohm. Practically any everyday thing can set you off.

    And you didn’t do so well with that prize $1 million bull, did you?

  5. DrLoser says:

    Drloser apparently you cannot copy paste… [Yiddish follows]

    Put that into google translate and notice not a single word changes.

    In a Platonic sense, Fifi, you are correct. After all, what are words, except symbolic representations, cast against the wall of a cave and interpreted through the hazy smoke of a stuttering fire?

    So then, we’ve proven that the Google translator is completely worthless when it comes to Yiddish, what with doing nothing more than a strictly character-by-character translation. What a wonderful advertisement for FLOSS that is!

    As to your main point, evidently I can copy and paste. Rather successfully. Even your inability to spell in English shines through remarkably well, Fifi.

    What was your point again?

  6. DrLoser says:

    DrLoser compare apples with orange here. It took less than 4 months after 8.1 release for Windows 8.0 to loss half it market share to 8.1. There is no reason not to except Windows 7 or 8.1 not to suffer the same kind of fate with Windows 10.

    Genius, Fifi! And there was I, falsely suspecting that all you do, all day long, is to sit in your northern NSW fox-hole and reminisce about how 9/11 totally wrecked your life!

    No, you’re absolutely right. The basic corporate strategy at Microsoft seems to be to maintain an almost total lock-down on the desktop. What Windows 7 lost, Windows 8 picked up. When XP and Vista dwindled, Windows 8.1 picked up the slack. And when Windows 7 and Windows 8.1 run out of steam, one can assume that Windows 10 will sail blithely on, having cannibalised both.

    Interestingly, that was the precise strategy that Linux used to demolish AIX, HP-UX and Solaris. All three of which were better operating systems, btw.

    So, Fifi, what then? How is Linux supposed to cannibalise an OS product that is equally free at the point of sale, and yet is considerably more familiar to the consumer?

  7. oiaohm says:

    kurkosdr
    Maybe FOSSies think that if Linux manages to get double-digit percentages, industry support will come.
    You like to say this is bogus. The reality is the number of developers from Intel and other Major companies for Linux is partly linked to market share.

    Which is bogus. Hardware manufacturers and software publishers only support a platform if it makes them money.
    This is wrong ish. Linux does make Intel money. Level of support Linux gets from Intel is linked the the amount of market share Linux holds or will possible hold in the future. Size of market share effects the amount of money Intel can make from their hardware.

    Market Share indirectly equals Money.

    Which is also bogus, because even if Linux “wins”, some company will release a distro laced with valued-added proprietary software, which would save in proprietary formats, converting Linux into vector of spreading proprietary formats.
    Bogus rant. Governments are demanding documented formats. Even Microsoft is forced to document there next generation MS Office formats. The future does not include proprietary formats.

    For example, the lock-in Google has on Android users with Hangouts, Play Store, Play Music and Play Movies
    Lets go nuts with bull crap.

    Play Music and Play Movies is available by chrome as extentions.
    Google Play Movies for Chrome
    Google Play Music for Chrome
    These two extensions also work in chromuim. They also work in the chrome browser of IOS. For Music there is also the Google Music Manager application.

    Hangouts also has a chrome application. Google Play store other than thens android applications that one kind make no sense to offer on other platforms everything else of the Google Play store is open to most other platform.

    kurkosdr so everything bar the android applications will work fine from iOS or a Ubuntu Phone. Why they have chrome. Only platform really being picked on is Windows Phone. Chromuim is in fact open source. Microsoft could in fact port it themselves. Apple helped port Chromuim to OSi.

    Sorry your lock-in claim what Google has done is really the worst attempt at vendor lock-in possible.

    The no coverer claim limited. The music you can straight download because google uses a watermarking system on music. So yes you can download the music and use what ever player you like.

    Google Play Movies being able to download these for offline playback is harder. Not major problem for Ubuntu Phone when it has android emulator and you can run the android application.

    Windows Phone has some major problems. kurkosdr please stop thinking that the Linux phones will have the same problems.

    You should have said Windows Phone and Firefox phone even then its iffy. Play Music is full operate-able by any decent browser. So Google Play Music should have never been included.

  8. oiaohm says:

    Windows 8.1 has fairly much stopped growing. I am expecting 7 and 8.1 after 10 release to go the way of Windows 8.

    Much in the way that Win95 gave way to Win98? And Windows 2000 gave way to Windows XP? And Windows Vista gave way to Windows 7?
    LOLOL Win95 users majority had to buy 98 the upgrade. Windows 2000 users majority had to buy XP to upgrade. So speed of change is way slower.

    Windows 7 and 8.1 are free upgrades to Windows 10 for all bar enterprise customers. Windows 8.0 has given way to Windows 8.1 very quickly. It was a free upgrade.

    DrLoser compare apples with orange here. It took less than 4 months after 8.1 release for Windows 8.0 to loss half it market share to 8.1. There is no reason not to except Windows 7 or 8.1 not to suffer the same kind of fate with Windows 10.

    Drloser apparently you cannot copy paste.

    די ספּרעעד אין פאָרעם פון די 20 אויך שורות אַרויף מיט די ספּרעעד אין באַזייַטיקונג צייַט רעכט צו דורכפאַל פֿאַר די באַזייַטיקונג כסימע צו זיין רעכט דיפּלויד .

    Put that into google translate and notice not a single word changes. You intermediate is wrong because you cut and pasted it wrong. When you learn how to use cut and paste properly for languages written the other way try again.

  9. kurkosdr says:

    I am never going to understand why FOSSies are so obssesed about market share. Can’t they just enjoy their Linux desktops without worrying?

    Mac users, in the days before the Intel switcheroo, were using a platform with low market share too. But you never saw this desire to evangelize their platform and attempts to “convert” people. In fact, they had this attitude of “we are awesome, we are an elite club, in fact if you don’t fit the mac culture, you should not be in”.

    Maybe FOSSies think that if Linux manages to get double-digit percentages, industry support will come. Which is bogus. Hardware manufacturers and software publishers only support a platform if it makes them money. For example, Android makes money for hardware companies but not for software companies, so it has good driver support but the software support is meh, despite it’s 80% marketshare. Also, the Mac platform had solid software support even during it’s darkest times, because making Mac software made the publishers money.

    Maybe FOSSies think that if Linux “wins”, coworkers will stop sending them files in annoying formats like WMV, WMA and DOCX, and everybody would stop using Exchange. Which is also bogus, because even if Linux “wins”, some company will release a distro laced with valued-added proprietary software, which would save in proprietary formats, converting Linux into vector of spreading proprietary formats.

    For example, the lock-in Google has on Android users with Hangouts, Play Store, Play Music and Play Movies is much stronger than Microsft could ever have via proprietary formats and Exchange, because there is no “converter” you can build to move anything purchased in there to another ecosystem (say Ubuntu Phone or Windows Phone). Ah, the joys of the “open-source” Android OS (read previous paragraph again).

  10. DrLoser says:

    Botnet destruction is normally the command and control server hijacked and the botnet forced to internally self destruct. So 24 hours the botnet dies.

    Fascinating, Fifi, yet also totally specious.

  11. DrLoser says:

    Windows 8.1 has fairly much stopped growing. I am expecting 7 and 8.1 after 10 release to go the way of Windows 8.

    Much in the way that Win95 gave way to Win98? And Windows 2000 gave way to Windows XP? And Windows Vista gave way to Windows 7?

    That’s a very bold prediction you’re making there, Fifi.

    Don’t worry. I’m not going to trouble you for a cite.

  12. DrLoser says:

    The spreed in shape from the 20 also lines up with the spreed in removal time due to failure for the removal signature to be properly deployed.

    You’ll be pleased to note, Robert, that I have translated this (naturally via Google Translate) into Yiddish and back again. Trust me, it makes a lot more sense if you do that. Don’t believe me? Here’s the intermediate Yiddish:

    די ספּרעèדאַ ספּרעèדאַ אין די פאָרעם פון די 20 אויך שורה אַרויף מיט די כסימע אין די באַזייַטיקונג צייַט רעכט צו דורכפאַל צו זייַן רעכט דיפּלויד פֿאַר די באַזייַטיקונג

  13. oiaohm says:

    The machines to which I refer are Lenovo’s backdoored bastards. When the news broke, I bet quite a few were taken off-line for delousing and most of them would be running that other OS. It’s a negative influx.
    Robert you make a mistake here. Superfish starts installing 2012 the program includes auto updating. Lenovo machines were in fact a very small percentage of Superfish infested machines. Yes more XP infested than Windows 7 or 8.1.

    And I use IE11, Chrome, Safari, and occasionally others, Robert. So what?
    The result of this DrLoser is you are being counted multi times. Comscores unique user count is per browser. So on a unique user count you are counted at least 3 times. There are other users who only use 1 browser who will only be counted once. This makes it very hard to get number of machines from webstats. DrLoser some days you would not use all 3 browsers right. Some days you would not browse as much.

    Statscounter is more a on-line activity counter. So the more active the user due to their actions or malware the more represented they will be.

    So far its not following a company or government migration. A migration drop day of migration followed sees a high spike follow by steep drop back to new stable. Jan 15 to Jan 21 has the signature marks of a migration.

    The cause the first few days after migration users don’t know how to do things so browse more then they get a backlog of work then are too busy to be on-line. This applies to Windows and Linux migrations.

    DrLoser reality is the Linux moves will be a mixture of different things. Some will be botnets destruction some will be malware and some will be migrations. 20 on does not follow the shape for botnet destruction. Botnet destruction is normally the command and control server hijacked and the botnet forced to internally self destruct. So 24 hours the botnet dies. Also after a botnet dies other parties try to grow their botnets and take up the contracts the prior botnet had. Bot net deaths normally cause a 0.1 to a 0.2 move in Linux numbers. 0.4-0.5 is way too much for a botnet. Yes the fact other parties with botnets will take up the contracts the dead botnet had means the traffic value does not move much.

    20 on does not have the migration or botnet shape. So 20 on looks like malware removal. if it malware removal it will basically level off at the new levels and go along like status normal. Interesting point is Linux is coming up on Windows 8 and Vista. So Windows 8 and Vista might be pushed into others.

    The spreed in shape from the 20 also lines up with the spreed in removal time due to failure for the removal signature to be properly deployed.

    Windows 8.1 has fairly much stopped growing. I am expecting 7 and 8.1 after 10 release to go the way of Windows 8.

  14. DrLoser says:

    That’s really bad for an OS that used to ship a million units per day… That’s why their shelf-space at Walmart has all but disappeared. 6.5% on Wikipedia…

    It’s not obviously bad when you consider that, in the same period, Windows 7 went from 43.5% to 43.5%.

    Try again, Robert. You’ll get the hang of sounding vaguely convincing eventually.

  15. DrLoser says:

    Anybody here understand the theory behind double integrals?

    Because, clearly, Fifi does not.

  16. DrLoser wrote, “Windows 8.1 has gone from zero in June 2013 to around 20% now”.

    That’s really bad for an OS that used to ship a million units per day… That’s why their shelf-space at Walmart has all but disappeared. 6.5% on Wikipedia…

  17. DrLoser says:

    So, the change has persisted. Canada is now at a new level for GNU/Linux page-views. It wasn’t a statistical fluctuation but a measure of a real movement. Way to go, Canada!

    Even mayflies have a longer-term view on “persistence” than you do, Robert. Have you tried slowly parboiling them? Let’s consider the “May blip” of the most recent year on record, to whit, 2014.

    Linux page-views started off at ~2% on 02/05/2014, shot up to an impressive peak of 5.75% (Yummy! Superfish for tea!) on 09/05/2014, and coasted along at 3% or above until 29/05/2014.

    Perhaps the Government Department responsible for a mass migration decided that, regrettably, Linux was not for them.

    Or perhaps, a la Fifi, several bot-nets were closed down on 29/05/2014, and normal business resumed as usual.

    Is there any really convincing reason to think that a small blip (basically, a doubling) in February is more encouraging than a larger blip (basically, a trebling) in May?

    I continue to marvel at your twisty turny inventiveness, Robert. To make a long story short — there isn’t.

    It was interesting to download the numbers over the last three years, though. Apparently Windows 8.1 has gone from zero in June 2013 to around 20% now. That ain’t no “blip” right there, Mister.

    Windows 8.1 seems to do quite well on the 5/7 test, too. I can provide you with the numbers, if you’re interested.

    (Me, I just regard them as the obvious expected outcome of a binomial distribution. But I think we’ve covered this, although you seem to be less convinced.)

  18. DrLoser says:

    Wikipedia for example sees -8.8 percent difference between Dec 13 and Dec 14 2014. Yes this is not a factor of 10.

    Yes this is not a factor of 10. OMG! I’m parroting an idiot!

    Yes factors of 10 in a single day has happened in webstats notice the difference with comscore report.

    An Englishman, a Welshman, a Chinaman, a Frenchman, a German, a Russian, an Australian (half Italian immigrant, half Aboriginal), a Bantu-speaker and a Small Furry Creature from Alpha Centauri walk into a bar.

    “Which one of you do I serve first?” says the barman.

    “I don’t know,” says the multiple statistical abnormality, “You’d better ask Fifi. He was the idiot who came up with this preposterous set-up for a joke in the first place.”

    It tells you how many visitors it counted. Guess when another big botnet dies.

    Your cite to that effect was distinctly unconvincing, Fifi. Even by your usual standards.

    Virus removal is able to move the number of visitors to sites by at least 5 percent.

    Five whole per cent! I can see we’re going to have to re-train you in basic arithmetic via the Danny Kaye inch-worm method, Fifi.

  19. DrLoser says:

    That’s quite silly. The Little Woman and I share the same computer. We have multiple browsers going depending on the latest features: Google Chrome, stable and beta, FireFox stable and beta, and sometimes opera.

    And I use IE11, Chrome, Safari, and occasionally others, Robert. So what?

    You’re as lazy at reading oiaohm’s posts (frankly you have my sympathy on that one) as he is on reading … well, anything.

    oiaohm’s theory depends upon a substantial percentage of desktop (etc) users switching between virtual machines and/or dual boots. Which quite honestly is so completely absurd that there is no possible way to justify it, unless you are going to divert attention to the person who pointed it out.

    Which is basically what you just did there, Robert.

  20. DrLoser says:

    Some people are slow learners. A while ago he was claiming it was a statistical fluctuation, not anything real.

    I’d defy you to find any single sentence in any one of my posts that claimed that the change is “not anything real,” Robert. I haven’t even gotten around to that particular aspect yet.

    I’m still trying to explain simple statistics to you, and simple mathematics to oiaohm.

    Unless and until we can settle on what the statistics are saying (numerically), there seems very little point in guessing at what the reality behind them is. I have one or two theories. So do you and oiaohm.

    But unlike either of you, I prefer to start with the numbers and work outwards, rather than start with preconceptions and twiddle the numerical dials to suit.

  21. DrLoser says:

    Now, this “Superfish delousing” thing in general.

    You seem to be assuming that a large proportion of Lenovo owners would actually bother to do anything. I’d be surprised if the number was as high as 10%. But, let’s ignore the real world and take your theory at face value: let’s say 100% of them did so.

    As you may recall, the Vicious Slave-Master Twelve Foot Tall Lizards of Redmond were somewhat dilatory on that “delousing” thing.

    It took them all of twenty four hours.

    Don’t you think that your week’s worth of numbers would show a little kurtosis, given that fact? It didn’t, you know.

  22. DrLoser wrote, “There appears to be a 1% rise in Linux page views over the period in Canada.”

    Some people are slow learners. A while ago he was claiming it was a statistical fluctuation, not anything real. That’s progress, I guess. Because it took place over a week, it’s a sure thing that it’s a number of activities: Lenovo’s crapware, GNU/Linux migrations, purchasing of GNU/Linux machines. All of these moves can take place in a week and they’re all good things.

  23. DrLoser says:

    The machines to which I refer are Lenovo’s backdoored bastards. When the news broke, I bet quite a few were taken off-line for delousing and most of them would be running that other OS. It’s a negative influx.

    A “negative influx?” Oh well. At least the concept is intelligible. But I don’t really think that “I bet quite a few” qualifies as anything other than wild speculation. Let’s try the Mount Fuji thing again.

    There appears to be a 1% rise in Linux page views over the period in Canada. (I’m simplifying to 1%, and as usual I’m doing so on the conservative side as far as the next calculation goes.)

    Assume uniform page-views per desktop, irrespective of OS.

    Assume that half the population of Canada owns or has the majority access to a desktop of some sort. (Please don’t argue about desktop sharing or thin clients. I’m not interested in that. I’m trying to reach a sensible estimate of Canadian desktops. And once again I’m being conservative.) I therefore estimate 18 million desktops.

    Got all that? Good. Now. One per cent of 18 million is 180,000.

    You are therefore suggesting, Robert, that Canadians, taken as a whole, have bought 180,000 Lenovo Superfished desktops, notebooks and laptops. (I’m saying this because your theory is based on “replacement.” At the ~1-2% level of Linux page-views in Canada, the “replacements” or “offlines” should theoretically dominate the numbers.)

    Lenovo would be over the moon if this were true — even with the horrible publicity. That’s a huge number of desktops (etc) sold in a very small timescale. What, four months or so?

    Crunch the numbers for yourself, Robert. It’s what Mount Fuji estimates are for. But however you do it (and I assure you, it’s a fun intellectual exercise), you’re not going to find the numbers backing up this particular theory of yours.

  24. DrLoser says:

    So influx of OS or Linux machines or a reduction in number of Windows machines showing up would provide identical movements in percentages.

    stochastic matrix can be applied to comscore numbers without resulting in what the hell disasters.

    I can take it that you didn’t find the (numerical) flaw in my argument, then, Fifi? It’s quite an easy one to spot. In fact the ratio between the actual OSX numbers and my “Mount Fuji” projected numbers should lead you to it immediately.

    You’re really not very good at calculating things or analysing actual numbers, are you, Fifi? Mind you, you’re exceptionally good at unintelligible blather.

    That particular effort, quoted above, was one of your finest.

  25. oiaohm wrote, ” Instead of a influx of machines the numbers move because the is a reduction in number of machines turning up.”

    The machines to which I refer are Lenovo’s backdoored bastards. When the news broke, I bet quite a few were taken off-line for delousing and most of them would be running that other OS. It’s a negative influx.

    OTOH, yesterday, the level was back to 2.58%, so it seems last week was not a blip but a real rapid shift in level. Lenovo’s blunder has been fixed by now, either by removing the malware or upgrading to GNU/Linux. It’s all good. Canada has caught up to the rest of the world and I like it.

  26. oiaohm says:

    suggests that a discontinuity in the denominator by a factor of ten in the matter of a scant few days is ridiculously improbable.
    http://reportcard.wmflabs.org/#
    Wikipedia for example sees -8.8 percent difference between Dec 13 and Dec 14 2014. Yes this is not a factor of 10. Yes factors of 10 in a single day has happened in webstats notice the difference with comscore report. It tells you how many visitors it counted. Guess when another big botnet dies. Virus removal is able to move the number of visitors to sites by at least 5 percent.

    The percentages change on Linux and OS X does not need more than a 8 percent flux in Windows Users.

    That’s a pretty good hypothesis. It correlates with the idea that an influx of machines is the cause.
    Robert virus remove is the reverse. Instead of a influx of machines the numbers move because the is a reduction in number of machines turning up.

    So influx of OS or Linux machines or a reduction in number of Windows machines showing up would provide identical movements in percentages.

    stochastic matrix can be applied to comscore numbers without resulting in what the hell disasters.

    There is a old saying garbage in garbage out. Statscounter percentages are that far removed that you cannot see any more what is really going on. Comscore numbers for the same time frame we would have a very clear understanding what happened. Either their would be increased Linux/OS X unique vistor numbers or there would be decreased Windows vistor numbers.

    Since Comscore would be numbers not percentages stats processing stochastic matrix works.

    Also statscounter has another issue.
    http://gs.statcounter.com/faq#page-views-uniques
    Yes statscounter counts page views not unique vistors. So this means 1 computer visiting a million pages is counted as 1 million in statscounter numbers.

    So statscounter moving by a factor of 10 heck even a million in a day is more than possible. The claim that there cannot be a factor of 10 move in a day fails to understand how bad statscounter numbers are.

  27. oiaohm wrote, “Linux and OS X are both acting oddly in the same kind of way the probability is that its something happening on Windows causing it.
    And if virus removal is causing it suggest Linux and OS X usage numbers have been higher for quite some time.”

    That’s a pretty good hypothesis. It correlates with the idea that an influx of machines is the cause. Most alert users would have deloused that other OS many times to be sure. It’s still a plus for GNU/Linux because everyone I know is really annoyed with malware by now if they weren’t a decade ago. One of the reasons for Android/Linux being popular is that even though it is demonstrably weak it’s a darn sight better than that other OS. If OEMs really wanted to grow their businesses they would trumpet the lack of malware for GNU/Linux.

  28. DrLoser wrote, “Octopodes might consider that possibility, Fifi. Humans do not.”

    That’s quite silly. The Little Woman and I share the same computer. We have multiple browsers going depending on the latest features: Google Chrome, stable and beta, FireFox stable and beta, and sometimes opera. Just last week I found we were using more RAM than I liked and found part of that was she and I were using different versions of FireFox. I’ve often used two different browsers simultaneously, sometimes on the same computer or one running on one and another on another computer. It’s easy with GNU/Linux. The browser isn’t part of the OS, just another application. GNU/Linux is a multi-user/multi-tasking OS and the display is easily networked.

  29. LinuxGentlesir says:

    This is great news! The world is discovering the joys of GNU/Linux!!

  30. DrLoser says:

    Ah, well, it’s becoming a feature.

    A person might be using 2/3/4 browsers at the same time on the same machine or virtual machines or whatever.

    Octopodes might consider that possibility, Fifi. Humans do not.

    BWAHAHAHAHA!

  31. DrLoser says:

    No idea why Robert has apparently gone off his Pure Love for Thin Client Chromebooks, btw. Given his methodology, he should be ecstatic!.

    I’ve just tabulated the world-wide ups and downs for Chromebooks on StatsCounter for the last month, and they’re pretty impressive, binomially speaking. (29/01/2015 to 27/02/2015.)

    * Fifteen “ups”.
    * Six “downs”.
    * Eight “no differences”.

    For the “five of seven” test, I’m going to assume that all “break evens” are in fact small negatives. Tolerance, you see. In which case we see:

    1. An unbelievable run of Nine Straight 5/7 Wins!
    2. A not really very demoralising run of Sixteen Straight 5/7 Losses.

    One after the other. It’s fascinating, isn’t it?

    If you have no clue whatsoever about how to judge number, that is.

  32. DrLoser says:

    Aha! I’ve immediately found the flaw in my argument.

    Let’s see if anybody else can.

  33. DrLoser says:

    Actually, oiaohm is correct to point out a serious discrepancy in the current analysis: to whit, OSX web page share. And as he correctly points out:

    OS X for new users need increase retail sales. No reports of OS X in any country selling like hot cakes to cause numbers to move.

    Now, about those retail sales. I believe they’re roughly at a plateau of 4.5 million worldwide. I am now going to perform some “How to move Mount Fuji” calculations which are (self-evidently) open to challenge.

    I’ll start by dividing the Mac market evenly between {Canada + USA} and {the rest of the world}. That doesn’t seem entirely unfair. (And, informally, I am going to assume a uniform distribution of sales between the USA and Canada, which probably overestimates Canadian sales.)

    That brings sales to the first of the two sets to around 2.25 million per year.

    Now we divide by the relative populations. (Again, this is pretty broad-brush. Canada is a lovely, lovely place, and outside fern bars in downtown Toronto is notably lacking in hipsters.) I’m going for roughly ten to one.

    That brings sales in Canada down to around 225 thousand per year.

    Now let’s assume that each Mac has an operational lifetime of, say, four years. That means that there are roughly 900,000 Macs out there, curling out page requests.

    I submit that probably half of the population of Canada has access to an Intel desktop, which would mean roughly 18 million of the things. (As you can see, I’m keeping the numbers in line just for the sake of innumerates like Fifi.)

    Which would mean that you would, in general, expect the ratio of PC-to-OSX page views to be 20:1. Or (again, just to help Fifi out) 5%.

    Interestingly, it is in fact 20%.

    I await bogus “statistical” explanations of this differential with cordial interest.

  34. DrLoser says:

    Hang on they are the exact same number of english people in a day.

    Evidently standard statistical methods are wasted on you, Fifi. Let me explain your conundrum using nothing but simple mathematics:

    1. One divided by one is one.
    2. One divided by ten is nought point one.
    3. One divided by one hundred is nought point nought one.
    4. One divided by one thousand is nought point nought nought one.

    Your two theoretical observations correspond to cases 3 and 4. However, you seem to be “privileging” the numerator at the expense of the denominator.

    Nobody cares about the singular numerator, Fifi. They only care about the vast discontinuity between the two denominators.

    The general behaviour of commercial systems — which include both your notional shop and web page views — suggests that a discontinuity in the denominator by a factor of ten in the matter of a scant few days is ridiculously improbable.

    Much like Robert’s weird comparison between the data as currently presented, and the purported sales/installation target of OLPC.

  35. DrLoser says:

    DrLoser you did not get it. A particular person might be 3-4 races at the same time. A person might be using 2/3/4 browsers at the same time on the same machine or virtual machines or whatever.

    It’s difficult to “get” a proposition that was not stated in the first place, oiaohm. Unlike you, my brain does not feed off Secret Microwave Information. And your revised version is a tom-fool analogy, as a moment’s thought will tell you.

    No matter. Let us assume that it is Golden. Explain, please, how it cannot modelled using a stochastic matrix.

  36. oiaohm says:

    Robert Pogson look again. OS X also abnormally rises from the 20. Windows 7 shape goes abnormal on the 23. A lot by then would have notices that anti-virus was not working properly.

    The feb 11 spike lines up with a termination of a particular botnet.
    https://www.shadowserver.org/wiki/pmwiki.php/Stats/BotnetCharts

    There is a lot of webstats disturbances from Feb4 on that line up with virus termination.

    Robert Pogson its the tunnel vision. You are just looking at Linux. You did not look at OS X to notice its reacting in a very similar way. So if Linux and OS X are both acting oddly in the same kind of way the probability is that its something happening on Windows causing it.

    And if virus removal is causing it suggest Linux and OS X usage numbers have been higher for quite some time.

    OS X for new users need increase retail sales. No reports of OS X in any country selling like hot cakes to cause numbers to move.

  37. oiaohm wrote, “So the reality so number of Linux users may not have change a single bit. All that had to change was the number of Windows users turning up.”

    StatCounter give “% of page-views”. The “wiggliness” of the lines on the graph gives an idea of the actual numbers of users. You can see “7” is a much smoother curve than GNU/Linux’s curve. You can also see the “wiggliness” of the GNU/Linux line decreased this week, indicating it was an increased number of page-views due to GNU/Linux. It is extremely unlikely that page-views due to that other OS would decrease by such percentages in such a short time.

  38. oiaohm says:

    DrLoser you did not get it. A particular person might be 3-4 races at the same time. A person might be using 2/3/4 browsers at the same time on the same machine or virtual machines or whatever.

    The notable irrelevant detail is the “particular race,” since we can be fairly sure that no section of the human population is likely to breed sufficiently rapidly to make a difference in the timeframe under discussion.
    The particular race thing is key a multi race person might call themselves 1 race one day and a different race the next.

    The problem here you said breed. How fast can people change software. There are a lot of people who can quickly change between using Window or Linux options. Or reduced infection rates in Windows. Cause less Windows users to turn up to be counted so causing Linux percentages to rise.

    Reality we don’t have from statscounter raw numbers of how many Linux people they counted.

    Does a shop get the same number of people threw the door. I have a shop and one english person comes threw the door. Now if 1000 people come threw the door total on that day not english that is a 0.1 percent english right. I have 1 english person come threw door yet only 100 people come threw door total on that day its 1 percent right. 0.1 and 1 percent are not even close right? Hang on they are the exact same number of english people in a day.

    So the reality so number of Linux users may not have change a single bit. All that had to change was the number of Windows users turning up.

    Binomial Probability Formula don’t work for web statistics because you don’t know how many trails you are really dealing with to make the web statistic numbers.

    The person walking threw the shop door example and what percent of the days clientèle is a great way to think about the webstats.

  39. DrLoser says:

    Its like the statistics of a particular person being a particular race to walk through the door of a shop on a particular day. But its more complex.

    In other words, it isn’t like that at all. Now, what were you telling me about binomial probabilities? And, incidentally, I wasn’t the one to bring up the question of randomness — Robert was. I merely ran with the idea and offered concrete numbers.

    You, on the other hand, produced nothing but gibberish. Your initial stab at this can be described as “stochastic modelling decorated with irrelevant details.” The elements of the model are the man and the shop doorway. The notable irrelevant detail is the “particular race,” since we can be fairly sure that no section of the human population is likely to breed sufficiently rapidly to make a difference in the timeframe under discussion.

    Were one to model your scenario using a stochastic matrix, it would be mathematically equivalent to the binomial procedure I cited. You will note that I assigned both p and q the values of 0.5, for simplicity: feel free to plug in (using a suitable lemma) any two other values summing to unity.

    I have demonstrated, in my post, the results of assuming a random effect, which at least addresses Robert’s assertion that the observed phenomena are absurdly unlikely to be the result of blind chance.

    You, Fifi, have demonstrated nothing but a drunkard’s walk.

    You seem to be quite good at a drunkard’s walk.

  40. oiaohm says:

    Note web status is monitor multi doors. The probability that every counties stats will move in the same direction for 4 to 5 days running. across something like StatCounter basically zero without some form of global event. Its not like something majority interesting in news happened.

    The alignment does suggest a percentage of Windows numbers is infection.

  41. oiaohm says:

    DrLoser you are a statical idiot. Binomial Probabilities does not apply to webstats. Because web stats are not a pure random number event.

    Its like the statistics of a particular person being a particular race to walk through the door of a shop on a particular day. But its more complex.

    In the global numbers of Linux on statscounter you see a steep rise on the 24 with no exact explain. Mind you the global numbers were by normal should have dipped on the 20 but did not dip.

    Interesting its Windows 7 and Windows 8.1 that same to lose the market share Linux gains. Timing very much suggests superfish. Globsl number still have Linux only just over the 1 percent.

    The problem I have is how can global effect at that time. Yes some countries effect is bigger/smaller. Normally each market is going different directions and when they add together they end up moving some random pattern. Since 20 of Feb all stats numbers in all markets have seamed to be moving in the same direction.

    20 Feb global effect possibility superfish removal tool. So far I cannot find anything else global around the 20 of Feb that could explain a OS number move.

  42. DrLoser says:

    Dr Loser, being not quite as mathematically challenged as a casual Manitoban observer might suspect, refers said casual Manitoban observer to a site that assists the more statistically challenged amongst us with the calculation of binomial probabilities.

    You should avail yourself of the Wonders of the Internet more often, Robert. Sometimes ownership of a Beast, an abacus, and a backup notepad and biro is not enough to avoid simple mis-statements of mathematical fact.

  43. DrLoser says:

    So, you claim that rises six out of the last seven days are more probable than five? I have a bridge you should buy. It’s old but still carrying a lot of traffic…

    If it’s a bridge recommended by a man who cannot recognise the concept of over and under, Robert, then I have no wish to buy it.

    Your original (“update”) claim was as follows:

    There’s only a few percent chance that’s random.

    In the absence of an authoritative definition of “a few per cent,” I took the bold step of quantifying said “few per cent” on behalf of your readers. I do not consider 14.6% to qualify as “a few per cent,” unless by “a few per cent” you mean “more than zero and less than or equal to one hundred per cent.”

    While I was at it, it seemed only fair to your readers to calculate the sum overage percentage. I don’t really see why you find this objectionable.

    Let’s face it, if you’d hit a coin-flip of 6/7 or 7/7, you’d almost certainly have devoted an entire new post to this magnificent achievement, wouldn’t you?

  44. DrLoser, being mathematically challenged, wrote,
    “1. Exactly five out of seven: 16.4%
    2. Five or more out of seven: 22.6%”

    So, you claim that rises six out of the last seven days are more probable than five? I have a bridge you should buy. It’s old but still carrying a lot of traffic…

    If the trend were flat or had a very low slope, random would mean a 50-50 chance for a rise on any given day. 3 in a row might happen on average 2-3 of the time. That’s 12.5%. Five rises in seven can happen a lot of ways, strings of 2 and 3 or 5 but 7! is a huge number, 5040. That blows the numerator away. A weekend is coming up. The last few weekends have been new highs. The trend continues.

  45. DrLoser says:

    OLPC is doing a lot but their plan is to migrate ~9000 students in 2015, close but no prize …

    Further to my previous observation: in what possible way can “~9000 in 365 days” be considered to be close to “hundreds of thousands in 7 days?”

    Simply to ask this question is to make it very plain that you are almost certainly reading more into these statistics than the weight of evidence can support.

  46. DrLoser says:

    UPDATE I updated the graph to reflect another couple of days’ climb. That’s five of the last seven days climbing. There’s only a few percent chance that’s random.

    You know very well that this is not true, Robert. On a pure binomial distribution of ups/downs, and assuming no priors — which is the best definition I can come up with here for “random” — the chances are as follows:

    1. Exactly five out of seven: 16.4%
    2. Five or more out of seven: 22.6%

    Given two years’ worth of daily numbers — which stretches back to the last time that StatsCounter “recalibrated,” I think — you’d expect to see this pattern quite often. In fact, very often. In fact, over and over again.

    You know this, and I know you know this. So why are you trying to pull a fast one on your more susceptible readers?

  47. I wondered whether OLPC might be the cause of this ramp. It’s possible but not very likely. A move of this size requires hundreds of thousands of GNU/Linux PCs. OLPC is doing a lot but their plan is to migrate ~9000 students in 2015, close but no prize … unless those kids surf a lot. They are also working slowly, one or two communities each month, rather than an explosive distribution.

  48. luvr says:

    “There was always a story that Linux low showing in Webnumbers was caused by malware infecting Windows machine with advertising ware causing windows machines to abnormally high hit web stats like StatCounter.”

    If that’s true, then certainly that lends credibility to the claims that “Linux flies under the radar.” I have always assumed that it was simply a matter of market share (which would naturally underrepresent Linux, because you don’t necessarily “buy” every Linux copy that you run), though that left me wondering why those web statistics remained so low.

    Anyway, time will tell how much of these variations comes down to “random noise” and how much of them is significant. IF the variations turn out to be significant, then I will be most interested to learn how those Windozers will spin the story. Perhaps something along the lines of “Well, the Intellectual Poverty from Microsoft clearly remains as popular as ever!”

  49. oiaohm says:

    http://arstechnica.com/security/2015/02/windows-defender-now-removes-superfish-malware-if-youre-lucky/

    This recent spike seams to line up with feb 20. There was always a story that Linux low showing in Webnumbers was caused by malware infecting Windows machine with advertising ware causing windows machines to abnormally high hit web stats like StatCounter.

    Of course the fact Windows could be over represented in numbers is not something those anti-Linux want to consider.

  50. DrLoser wrote, “I believe you understand the concept of statistical variance, Robert.”

    Yes, I do. As you can well see, the lines are not “straight” but dance about. That’s just random noise in the signal. This move is many times the typical variation that we see. The bursts in May are much too large to be coincidences too. The typical variations are 0.1 – 0.2% and this move is 0.6%, day after day. It’s not likely a random variation. I could do some calculations to measure that but it would never please you so I won’t bother. These are on top of the earlier trend that held all through January as well. This is real.

    Chuckle. I just updated the post to reflect yesterday’s movement. That’s now five of the last six days climbing. There’s only about 3% chance of that being random.

  51. DrLoser says:

    How about that, rising from ~1.8% to 2.45% in a week?

    I believe you understand the concept of statistical variance, Robert.

    And here’s a good chance for you to prove that you do. Let’s both guess the figure for the month of March.

    My guess is 1.7% to maybe 1.95%.

    Your confident prediction would be, what?

Leave a Reply