Venezuela Rebuilds The Desktop Amid The Chaos

Politics be damned. This weekend, GNU/Linux grew dramatically in Venezuela, according to StatCounter. Thank you, “7” and “8” for staying home. The last few weeks has seen major events that pushed the global share from ~1% to near 2%.

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.

52 Responses to Venezuela Rebuilds The Desktop Amid The Chaos

  1. oiaohm says:

    Loser did not get the meaning of this line.
    Gets funnier. Redhat was born in Carnegie Mellon University large number of professor at Stanford.
    Just like today majority of Stanfords Professors are from Carnegie Mellon University. So yes a large section of their network was Sun Ultra hardware but it also was running Linux by 1996. Redhat Linux to be exact. So the claim they did not have a network of Linux computer is another bogus claim by the Loser.

    Why run Linux simple so students could do homework at home on their x86 machines. Yes the reason why it was developed on x86 and Sun Ultra was that the x86 was is home computer. x86 home computer running Linux was normal for students at Stanfords or Carnegie Mellon University in 1996,

    Linux was not exactly techincally superier that much in 1996. Linux was cheap and would run on cheap hardware and that was the killer. Early Linux apache deployments is 1994-1995.

    Google redhat addiction in their server core was documented in a Linux Conference Video where they were merging deb packages into a redhat system to get newer versions. Yes they also state in that video that backrub was done on Redhat. Yes a video by the author of Backrub.

    Sorry Loser you have dug yourself into a hole if you want a finishing blow push me until I bring the video link in here from the Backrub author completely showing you are full of it. Yes real information who was at Standford in that time frame.

  2. DrLoser wrote, “1996 would be a ridiculously early point to develop a server system on Linux”.

    Apache web server was there. So was GNU/Linux. Apache was the most popular HTTP server on the planet then and ever since. VA Linux and many others were cranking out GNU/Linux computers. RH 4.0 came with Apache in 1996. “If you installed the Apache web server (from the apache package), then your Web service is already up and running!”
    see the User’s Guide

  3. DrLoser says:

    Do try and keep up, oiaohm.
    From the link that you didn’t actually bother to link to (funny how you do that so often):

    BackRub was written in Java and Python based “on several Sun Ultras and Intel Pentiums running Linux.”

    Pausing only to admire the weird choice of Java and Python in 1996, and I commend Sergei and Brin for their, ahem, insane bravery … (wot, no configure/make/make?)
    Allow me to observe the following fairly obvious fact.
    In 1996, Stanford University did not have a network of Linux servers
    The corollary to this fact is that, in 1996, Stanford University did indeed have a network of Sun Ultras.
    1996 would be a ridiculously early point to develop a server system on Linux. (Note: I am not saying that it was ridiculous, a few years later. For some value of “few,” possibly even two years.)
    So, here’s the deal. Unless you can dig up an obscure, yet meaningful, link, I propose we assume the following:
    1) Sergei and Brin used the Stanford University network to boot their search engine up.
    2) That network was exclusively Sun Ultras.
    3) At some point in that twelve plus period, they looked into the possibility of scaling the engine up.
    4) At which point they decided, and it doesn’t really matter why, that Linux was the way to go. Now, I say “it doesn’t really matter why.” I elide that question. I’m sure you have an expert opinion on that question, oiaohm, and I am equally sure that it will have no basis whatsoever in fact.

    As I say, the only interesting question here is why Andy Bechtolsheim didn’t just offer them $100,000 worth of free rent on Sun’s in-house servers. I mean, the man is rich. The man is even richer from that original $100,000 investment.
    But, not only would he have become immeasurably richer, based on resources he could actually afford to give away (free, as in beer) at the time:
    He’d have saved Sun from that silly little episode when the Pony-Tailed Twit sold the company down the river via FOSS to … Larry Ellison, of all people.
    Isn’t computer history fun?

    In this case, it’s an object lesson to commercial companies. When you say, “The Network is the Computer,” you’d better damned well understand what “The Network” is.
    I have no problems at all with Google running on Linux. But, let’s face it, Linux got lucky. That’s how these things work.
    Not through some farcical and completely unjustifiable notion of “technological superiority.”

  4. oiaohm says:

    Loser 100000 of cheap hardware turned in to Linux servers is faster than 200 000 of sun servers back when Andy Bechtolsheim funded Google. To be enough you are not talking double you are talking x5. So 100 000 dollars vs 500 000 dollars to get the same result.

    Gets funnier. Redhat was born in Carnegie Mellon University large number of professor at Stanford.

  5. oiaohm says:

    http://en.wikipedia.org/wiki/History_of_Google
    Quote of Email from author of backgrub.

    BackRub is written in Java and Python and runs on several Sun Ultras and Intel Pentiums running Linux. The primary database is kept on a Sun Ultra II with 28GB of disk. Scott Hassan and Alan Steremberg have provided a great deal of very talented implementation help. Sergey Brin has also been very involved and deserves many thanks.

    Loser there was a technical reason why Google choose Linux not FreeBSD. Broader hardware support. Backrub the first google prototype run on Linux. Never run on Solaris.

    Please stop being incompetent there is absolutely no reason to guess when the quote is on the wikipedia. Simple reality here Loser never does his homework. Takes wild guesses and hopes they are right.

    Please note he mentions that I talked about proxies about 3 years why is because the OpenNet Initiative report in 2011 gives a list of countries that statscounter and other major net stats are 100 percent blocked but they are showing collected data. Either their Geoip data is wrong or they are only showing the usage by a select few in those countries. Reality the data says that web stats are no where near a random sample so can be gamed in huge ways.

    Without reading the reports of the ground conditions you have to be very careful of confirmation bias.

  6. DrLoser says:

    Facebook, Google and Instagram are not built with M$ software at all. If Linux did not exist, neither of those would either. Food for thought.

    Mildly diverting for about tens seconds for anybody who thinks for a living, possibly. Otherwise, not.

    Leaving aside the question of what M$ server OS might have been available in 1997 (clue: none at that scalability), let us exercise our noggins on this tricky little what-if.

    What if, for example (to quote from your cite on Google history), they hadn’t been able to run BackRub on Stanford servers for more than a year in 1996? Those weren’t Linux servers. I would take a guess that they were Solaris.

    What if Andy Bechtolsheim had doubled his original $100,000 stake and offered another $100,000 of hosting on Solaris? I assume you are aware of Andy Bechtolsheim’s position at the time.

    But let’s leave all that to one side. Was there a suitable, cheap/free OS on which to run Google servers at the time, other than Linux?

    Have you heard of FreeBSD, Dougie? Because that’s the one.

    Ironically for your tale of po’boys in a garage, Google was actually started by a couple of well-off, highly intelligent kids with PhDs who went to Stanford. You may recall that Stanford is not merely elite, but also rather elitist.

    If they’d gone to the hot-bed of FOSS radicalism in the 1990s, ie Berkeley, they would presumably have chosen FreeBSD. There was no technical reason not to do so. Nor was there a problem with cost.

    Well, ten seconds of thought and about two minutes of explaining that thought in sound-bites suitable for the not very bright. I live to educate, Dougie, I live to educate. Even you.

  7. dougman says:

    The only liars I know in the IT industry is M$, time and time again they obfuscate the truth against Linux and open source in general.

    Truth be told, why wouldn’t they? Linux affects their bottom line and total profits annually.

    Now the trolls will probably say something to my comment, but lets be honest shall we?

    Facebook, Google and Instagram are not built with M$ software at all. If Linux did not exist, neither of those would either. Food for thought.

  8. DrLoser wrote, “It’s blatantly obvious, Robert, that if you “choose” a sub-set to observe, you are already guilty of onfirmation bias.”

    Nonsense. Why should I promote boring statistics? It’s just not interesting to read that no one uses GNU/Linux when we know that it is widely used. I am exposing the inconsistency of the web-stats industry. GNU/Linux has been growing quickly for well over a decade yet the web-stats industry would lead one to believe GNU/Linux doesn’t work for anyone. At least StatCounter lets me dig stuff out. NetApplications let me do that until it became too embarassing, e.g. Sunnyvale, California swinging the whole USA while huge school divisions using GNU/Linux registered nothing. What do you call it when the web-stats industry ignores a phenomenon that is thriving but not used universally? That’s a selection bias if there ever was one. It is an outright lie that GNU/Linux is not working for many millions of people. I expose that using the web-stats industry’s own tools because I am willing to make the effort to dig out the details. Rather than face the facts, DrLoser would rather attack the messenger while ignoring the message.

    The countries/regions I have pointed out involve millions of people and significant percentages of those people use GNU/Linux and love it for the price/performance and independence from Wintel.

  9. oiaohm says:

    http://en.wikipedia.org/wiki/Internet_in_Colombia

    Loser Colombia is one of the places with government enforced Internet filters. Sorry my squid proxy answer here is not baseless. Its the dominate filtering proxy used. You have never wanted to admit to the existence of this. Complete ISP’s in Colombia are invisible to statscounter and other web stats counting sites.

    Basically allowing statscounter to work could be a breach of Colombia law so some ISP’s there are going on the side of care.

    This case I can bring in absolute proof that statscounter cannot be trusted to respective of Colombia because over 50 percent of there Internet population is invisible to it. This is not the only country with privacy laws forcing ISP to filter out items like statscounter.

    Sorry without knowing the conditions on the ground of a country you don’t have any idea what the numbers are telling you.

    Loser you have never wanted to admit that some of the reason why statscounter numbers are so dead is filtering proxies. ISP and Country level filtering.

  10. dougman says:

    Hey Douchebag, instead of being a twit and nitpicking everything, perhaps you should go outside and find a new hobby.

    Don’t worry, Linux will still be around for you come winter and steadily growing, ready for more of you negative attitude.

  11. DrLoser says:

    I choose places where the governments have had the spine to defy M$.

    It’s blatantly obvious, Robert, that if you “choose” a sub-set to observe, you are already guilty of onfirmation bias.

    Let’s be honest, shall we? You didn’t pick these countries because their governments had “spine.” You picked them because the temporary statistics looked good, didn’t you?

    The very definition of confirmation bias.

    If I were in charge of any of those countries I would ban M$ outright as a criminal organization.

    And if I were in charge of any of the two you quote, I would concentrate on establishing a stable constitution, Rule Under Written Law, and the best rights I could acquire for children, women, homosexuals, and apostates.

    We can agree to differ on priorities.

    Robert Pogson May 31st, 2014

  12. DrLoser wrote, “is there any reason why you would concentrate on no doubt very nice and charming places like Venezuela and Ethiopia?”

    I choose places where the governments have had the spine to defy M$. If I were in charge of any of those countries I would ban M$ outright as a criminal organization.

  13. DrLoser wrote, “Why not, for example, Tajikistan? I picked that one at random, honestly. I can tell you why not. It’s obscure, but it has absolutely zero Linux Desktop penetration whatsoever.”

    You lose! Actually StatCounter lists “Top 7” (nice round number, that) just so GNU/Linux doesn’t show up in the graph but it’s there in the downloaded CSV:

    Actually, Tajikistan showed up to 1.4% but the average was 0.54% from April 1 to May 30. Digging deeper, I find the peak was 15.47% on 2009-Nov-16 (just a glitch, though). The average over all StatCounter’s days was 0.68%.

  14. DrLoser says:

    And may I humbly point out that “conformational bias” is presumably a tendency during stereoisomerism to pick the wrong sort of glucose?

    It certainly has nothing to do with “conformational bias,” which I have not exhibited in any way (remember, my original argument was even-handed and had nothing to do with OS platforms, just about Big Data).

    You, on the other hand, have confirmation bias — feel free to cut and paste it; it would make your pitiful insults a little more relevant though no more accurate — written through you like a stick of Blackpool rock.

  15. DrLoser says:

    Yes Loser without knowing the countries methods and picking a country at random is going to tell you lead you to absolutely the wrong idea about what is going on.

    As opposed to making things up out of whole cloth, and lying through your teeth about personal knowledge of weekend migrations in Venezuela?

    I’d rather be me, oioahm. At least I can still divide 8 by 2 and get 4. You’re still left grasping at six points of phantom “spikes.”

    Nice to see you falling back to your “squid proxy” argument, though. Farcical and spurious as it is — you really have no proof at all, do you? — it’s an oldie but goldie. The last time you trotted it out was, what, three years ago?

    Go ahead. Pick a random country of your own. I will honestly believe you if you tell me it’s random. Anywhere between 10 million and 200 million population.

    Now tell me what the results were, between 1st April and 28th May.

    You lose, little outback dingbat.

  16. oiaohm says:

    Loser you are suffering from conformational bias in a big way. General rule when using statistics is work out a set of test cases to check of the numbers are responding how you expect them to be. So the numbers are not absolute garbage. Looking at what happens during and after known migrations confirms if the numbers are reacting how you expect. Like basic idea suggest more Linux installed increased percent right. But due to deployment designs you see some very strange things. More Linux can equal lower Linux counted in online-stats.

    Only countries with counter friendly policies in IT departments can you be sure are showing you anywhere near reality.

  17. oiaohm says:

    Note I did not say they were stacked I said they are placed in the location of do or be sacked.

    Columbia go back to 2010 when Columbia Education migrated(yes you can get exact time-lines they documented it). Its also very strange. You see no rise. You see drop.

    How each area interacts with statscounter relates to their IT methods. Due to know interactions of methods used in Columbia Linux could be like 60+ percent and still show up as nothing. Mandated policy to install privacy protection in squid proxies. So after Linux migration the Linux machines disappear from being counted by Statscounter or any of the other on-line stats counting. Yes the magical disappearing computer.

    There is a general patterns in numbers that appear all the time. Then long term patterns.

    Outside base level movement spikes don’t happen every weekend.

    Yes Loser without knowing the countries methods and picking a country at random is going to tell you lead you to absolutely the wrong idea about what is going on. Linux growth in Columbia is going to be absolutely nothing if their was growth their it would be a breach of their deployment methods. Basically their method is the adblock effect on steroids. Columbia is not the only area like this.

  18. DrLoser says:

    Why not, for example, Tajikistan? I picked that one at random, honestly. I can tell you why not. It’s obscure, but it has absolutely zero Linux Desktop penetration whatsoever. Oh well, presumably StatsCounter bias.
    Let’s try somewhere nearer to, say, Ethiopia, shall we? Again, at random (honestly), Rwanda. Looks promising: I think the GDP growth rate is above 5%. Linux desktop on StatsCounter? A lamentable 0.5% or so, with a small “spike” around 24th April.
    Or then again, what about Colombia? Now, this one is not just a random pick. I am picking it before I even look at the numbers, because it seems like a fair comparison against Venezuela (although without the forceful “politics be damned” government backing). I genuinely don’t know what this is going to show me.

    A total flat-line at 1%, and you really wouldn’t want to know how well WIndows 7 is doing, and I don’t want to hurt your feelings.

    Perhaps it’s all down to Yanqui Imperialism?

    Or perhaps, in oiaohm’s fanciful “statisticses” language, “a sort of flattened spike that occurs every other weekend in between major government department migrations when people are sacked and then re-hired the weekend after, because Mrs Elsie Prong of Moonie Ponds wrote a nice letter in purple ink?”

    One thing I know is that oiaohm can excuse everything, apart from his own ignorant ineptitude.

  19. DrLoser says:

    Over the last couple of months something huge has happened. It’s too big to be just a rollout or a stampede to GNU/Linux but it’s something and StatCounter calls it “page-views”. Perhaps M$ has stopped bribing web-stats firms… 😉

    Or perhaps you’re just letting confirmation bias get the better of you, Robert, since we both know you can do a better job of statistical analysis than I, once you put your mind to it.

    I’m none to sure what this “too big” thing is, although you appear to believe that the thing in question is for some reason a multiplier of Linux desktop growth per se — why you would believe that, and what that would mean, is beyond me.

    But let’s get serious here. If there’s a world movement towards adoption of the Linux desktop being shown here, is there any reason why you would concentrate on no doubt very nice and charming places like Venezuela and Ethiopia?

    Come on, be honest, you wouldn’t. There is only one possible reason to focus on these otherwise excellent but teeny tiny markets, and that is confirmation bias.

    Speaking of which, you can prove you don’t have confirmation bias by promising to review the counts in Venezuela at some future time of your own choosing.

    I’m not picky about that time frame, but frankly, if you don’t set it out as a plan before you see the numbers, you are essentially admitting confirmation bias.

    Naturally, you have every right to admit confirmation bias and leave it at that. But I see you as having more of a competitive, honest, and moral nature than that.

    Have at it!

  20. oiaohm says:

    Loser the problem is you are still saying like cosmic rays and political. History from 2008 to today is anything like that has not turned up before also fails basic logic test. Exactly why would cosmic rays and political effect windows users different to Linux users.

    There is one migration possibility that could be political. If a key building was destroyed or rendered not usable requiring a new office to be setup. Google office migration effect. Big swings other than to zero in statscounter over and over again trace to the same thing. Migration of some form.

    Loser for people wanting to prevent FUD you are putting out a lot of FUD guesswork. If you are going to hold others to high standards you need to hold yourself to equal or you are a laughing stock.

  21. oiaohm says:

    Loser basically there are clusters of it in 2013 and 2010. This is why I am saying its not abnormal. Not common. But not abnormal.

    Venezuela if you had gone into the number properly Loser you would have had a proper answer. Instead when I give a proper answer for the most likely you just laugh at me.

    DrLoser exactly what post that I could not divide 8 by 2.

  22. oiaohm says:

    DrLoser there are a total of 22 2 percent spikes not counting Roberts one. Those all align with known migrations. So exactly why would this one be any different.

  23. DrLoser wrote, “Because I was under the impression that your argument was that the variable in question was far from random. In fact, I understood you to infer that, rather than being random, the variable in question actually meant something.”

    It’s not the variable that’s random but the measurement of it. One cannot measure anything with exactitude unless it’s the count of eggs in a carton or some such thing. Page-view share has many random inputs: the users, the sites, the network, events in 1000 million people’s days, etc. Counting anything over a long period of time, say, a day, will inevitably be something +/- some variance. That’s why the curves aren’t smooth but dancing up and down in jumps day by day. If this thing called “share of page-views” were able to be measured precisely, all the curves would vary rather more smoothly as everyone would use their PC every day to visit the same set of sites at the same rate. That doesn’t happen. The only kinks would be the deaths/maturity of some users, the deaths/maturity of some PCs, huge rollouts etc., variations of the order of a few percent per annum. Over the last couple of months something huge has happened. It’s too big to be just a rollout or a stampede to GNU/Linux but it’s something and StatCounter calls it “page-views”. Perhaps M$ has stopped bribing web-stats firms… 😉 Perhaps StatCounter has gained serious market-share of sites. Perhaps some governments have 1 IP address per machine. I think that’s what gave Google such an impact in Sunnyvale, CA. A few thousand users swung the stats for the whole USA. What’s with that?

  24. DrLoser says:

    With all due respect, Robert, this is quite obviously and definitionally a time series. Whether it is a continuous function or a discrete function rather depends upon your view of the epsilon in question.

    What is beyond question is that your assertion of a random deviation being in the same direction from the mean for fourteen days is demonstrably false, unless by fourteen you mean six, in which case there’s one stretch of six under and another stretch of six over.

    And if it is indeed “a random discrete variable” being measured, or even “a random continuous variable” being measured, then what’s this discussion all about? Because I was under the impression that your argument was that the variable in question was far from random. In fact, I understood you to infer that, rather than being random, the variable in question actually meant something.

    I’m still waiting for some sort of proper statistical evidence of this: even given the very obvious fact that, as I say, the fifty-odd data points are nothing but a consolidation of an underlying mass of far more interesting data.

    Perhaps oiaohm’s hilarious attempts to reach back to 2010 and 2013, thus pulling up another couple of 2% spikes (might be to do with weekend migrations, might be to do with cosmic rays, might just possibly have to do with the fact that oiaohm has proven himself unable to divide 8 by 2, let alone 6 by 2).

    OK, I’ll accept that my previous arguments are quite possibly utter tosh. It’s your right, as the man in charge, to assert that. But with that right comes an obligation.

    I’m not going to specify the exact period, but I think it’s only fair that you commit to revisiting Venezuela at a future time of your own choosing for a similar period, let’s say between three and eight weeks.

    And this time, tell us ahead of time. The great thing about a proper experiment like this is that there is no confirmation bias involved, either way.

    You won’t be able to throw the results away because they don’t agree with your assumptions. Equally, I won’t be able to disagree with your results because they don’t agree with mine.

    Still looking forwards to oiaohm’s farcical dalliances with an imagined Venezuelan past, however. Dusky beauties … naughty pharmaceuticals … superb music … and of course ridiculous numbers, as always with oiaohm.

  25. oiaohm says:

    By the way something interesting is that if you map known migrations against country data you find that each country has a different migration response in the data. Please note known Windows, Linux…. Migrations. But there are similarities based on how similar the migration methods are.

    Items like this are only known to people who look at the long term time frames.

  26. oiaohm says:

    Loser you can download a cvs file from Statscounter of the numbers from 2008-today on any region. Guess what I had done then run a backwards compare. The most exact match is 2010 for rapped size change. But anyone who had do this should have also know of the 2013. Then checked for alignments in numbers and events. That you asked for 4 years backwards to name another event the same either 1 you had done your homework or 2 you had not. If I played I did not know the 2013 I would find out what one. Loser don’t comment on these things without doing your homework I will catch you if you have not. I am not past making a mistake myself to confirm if the other person knows something or knows nothing. This case you know nothing.

    So since you say you don’t have the real numbers you did not spend the effort to grab the background data. Shut up on this topic. Yes the cvs of the data is publicly accessible. The only thing when you look at the complete set of numbers is that Linux is still over all trending up not down. And it does wobble about 1 percent per day.

    Migration events depend on size cause different size temporary bulges greater than the 1 percent general motion. Windows 7 and XP also have a general wobble motion also has bulges in there numbers with migrations. The behaviour in the numbers is normal.

    You are just looking at the graph on this page that means you are not looking at the raw data to form your own point of view. Instead form a idea from a subset of data that will be completely wrong Loser.

    I told you are done. If you keep this up I will pull in the 2010 and 2013 graphic data and show the similarities. Both of them were way larger migrations. Particular events have signatures.

  27. DrLoser wrote, “it’s not a very good way of looking at a continuous time series at all.”

    There is no continuous time-series here. It’s a counting experiment where StatCounter counts from one day to the next how many GNU/Linux page-views happened. It’s a random discrete variable being measured. It’s random because Johnny may be at the beach on Tuesday or Jack may not visit one of StatCounter’s sites or a site might be down the moment a GNU/Linux user thinks to visit. Assuming no one changes OS nor buys a new PC on a given day, the measurement can rise and fall. The most probable score is the mean value. Excursions of a few standard deviations from the mean are quite possible. Huge variations are improbable. Having huge variations all in the same direction day after day is extremely improbable. After a few days, one should accept that something real has changed, either StatCounter’s universe of sites or the Internet’s usage of GNU/Linux or both. It is unlikely that a huge rollout of GNU/Linux happened or we would have heard something, but statistics is quite a different argument. The thing is that the folks who heaped scorn on GNU/Linux when the numbers were sub-1% now are claiming “statistical fluctuation” when the numbers are several percent and “sticky”.

  28. DrLoser says:

    A day after the start of that period, btw, the figure was 6.46% (a leap of 7% over the previous day), and a day before the end of that period, the figure was 6.39% (and the day after that leapt by 9.5%).
    I’m not going to be mean. I’m not going to say that April 1st was a bad day for Linux in Venezuela. Nor am I going to say that May 28th was a particularly good day for Linux in Venezuela.

    And I’m certainly not going to conclude, given that Linux in Venezuela went from 6.46% to 6.39% over almost all of this period, that it’s trending downwards.

    Because, just like you, just like the fantasist oiaohm, I genuinely have no insight here. I do not have access to the real numbers.

    And unlike you, and as always unlike the fantasist oiaohm, I refuse to guess.

    It’s undignified.

  29. DrLoser says:

    That’s not at all what we see here. It’s an upward trend lasting weeks. The samples are daily. What are the odds that a random deviation is in the same direction from the mean 14 days in a row?

    Glad you brought that up, Robert. The first thing we have to assume is that we are dealing with a binomial distribution, or otherwise a distribution where the calculation of a standard deviation works according to your formula.

    The second thing that needs to be noted is that there is an obvious difference between continuous functions and discrete functions.

    1 in 214, a tiny number. It’s not unusual for a family to have 3 consecutive girl-babies, but 14 would be a world’s record.

    Indeed, and here we see the canonical version of a discrete function with a binomial distribution. (We can lay to one side the minor detail that boy/girl births are not quite 50/50.)

    Unfortunately, it doesn’t apply here. In fact, it’s not a very good way of looking at a continuous time series at all.

    Here’s a quick thought experiment. Google’s stock price is at, say, $500 on Monday. It’s at $501 on Tuesday, $501.10 on Wednesday, $503.99 on Thursday, and $504 on Friday.

    In binomial terms, that’s four coin-flips and it meets a 1/16th expectation of happening. (cf your boy/girl example.) On the other hand, the end result wouldn’t really surprise anybody, because the stock has gone up by less than 1% in a week. Good, but not outrageously good. Whereas, on a discrete function (dealing only with ups and downs of the original unit, ie $500), the same binomial sequence would yield $8,000.

    I think we can all agree that $8,000 is a far more impressive number (in a week, yet!) than $504.

    Now, to the specific numbers on Venezuela between 1 April and 28 May. We are obviously reading these numbers differently. I’ve just fed them into my trusty free spreadsheet, and the mean for that period is 6.67.

    Here are the positive (above the mean) and negative (below the mean) runs:

    -3, +2, -5, +5, -2, +1, -6, +3, -2, +3, -1, +1, -4, +3, -6, +1, -1, +6, -2, +1.

    I don’t see anything extraordinary there, much less a +14 (or indeed a -14) of any kind. I can calculate it on a rolling mean, if you like. I very much doubt I’d see anything particularly impressive.

    And bear in mind that I am not choosing my own end-points: I’m using yours.

    Should I, for example, choose to look at an equivalent period, starting on 24/05 and ending on 22/07, I’d expect to see the same distribution around the mean.

    Of course, if you like, we can start talking about the third moment (skew) and the fourth moment (kurtosis), but I think I’ve said enough to reinforce the point I started with, which is:

    If we don’t have the petabytes of data behind this tiny number of data points, and if we don’t have the enormous amount of computing power to run thousands of Monte-Carlo simulations, or whatever … then we’re just blowing fairy dust.

    I’ll note that oiaohm is especially predisposed to blow fairy dust.

  30. oiaohm says:

    Loser a migration spike does not just happen in case of Linux to WIndows migration. It also happens in case of a Linux to Linux. If there is a radical enough interface change. Yes some ubuntu changes have been radical enough.

    There has been constant slow growth over the last 4 years. Data from 2008 to today is accessible. The behaviour of spike rise in middle 2010. In fact Loser there is a migration spike inside the 4 year window.
    Jan 2013.

    See Loser never does his homework. This 2 percent jump followed by fall is not abnormal. Its abnormal if you look a 1 year time frames. 5 to 10 year time frames its normal effect in these numbers. 4 year time frames is not normal way to process IT numbers when 5 years worth of numbers+ is on hand.

    There is numbers from 2008 to today that you can download Loser. 2 percent spike has only happened in alignments with Migrations. So exactly why have they not appeared from 2008 to 2014 for any other reason.

    The numbers don’t do 2 percent variance. The 2008-2018 tell you there noise value is 1 percent. Anything under 1 percent is background noise.

    A Linux to Linux Migration are still triggered. End of XP is a trigger for that.

    Sorry that you agreed there are non in the last 4 years and did not correct me Loser means you have not read the numbers and need to shut up before we do kick you to curb.

  31. pogson says:

    DrLoser wrote about “an occasional spike”.

    That’s not at all what we see here. It’s an upward trend lasting weeks. The samples are daily. What are the odds that a random deviation is in the same direction from the mean 14 days in a row? 1 in 214, a tiny number. It’s not unusual for a family to have 3 consecutive girl-babies, but 14 would be a world’s record.

  32. DrLoser says:

    (Needless to say, for those out there that do not understand these things, Robert and I are assuming a binomial distribution for the purposes of discussing standard deviations.

    (Your distribution may vary …)

  33. DrLoser says:

    The “straight-forward statistical variance” was meant to abnegate my obviously outrageous “claim” that Windows XP had pulled audience share back in two days, Robert.

    Obviously, a 2% variance when you start with a mean of something around 28% doesn’t have the exciting potential, measured against standard deviation, as a 2% variance when you start at a mean of 6-ish%.

    Which is partly why I “confessed” that my totally made-up experiment was bogus.

    But then again, at the 6% level, you’re by definition talking about small sample sizes. An occasional spike at four sigmas wouldn’t surprise me at all. More than one would be interesting, but that’s not what we’re seeing here.

    I repeat: if you don’t have the underlying petabytes of data, you’ve basically got nothing but wishful thinking.

    Having said which, I’m aware that you revisit these sites at intervals, and I’m genuinely interested in the next set of results.

  34. DrLoser wrote, of Venezuela’s excursion in GNU/Linux page-views, “It’s straight-forward statistical variance.”

    That’s entirely possible but extremely unlikely. Standard deviation of counting statistics is like 1/sqrt(n) so the values near the beginning of the period, ~6.6% would be expected to vary over a few standard deviations of ~0.4PP, not ~2%. That would be five standard deviations, possible but extremely unlikely. Further, such excursions should vary +/- the average, and not make a mountain. This is a prolonged excursion from the status quo, not random variation. StatCounter collects data daily, so the excursion is lasting many days. Do you feel lucky? Go to Vegas. You will find the big losers betting on things going one way for too long a time.

  35. DrLoser says:

    (Platform migrations, not data migrations. Apologies.)

  36. DrLoser says:

    Well, that was probably malformed, or else Robert has correctly implemented security procedures. So here’s the link instead.

    Note that Venezuelan Linux “stat-countiness,” fwiw, has dropped to the monthly average.

    And here I could pull an oiaohm and make things up and basically lie to you all and point out that the Windows XP numbers have bounced right back up there.

    Why would that be? Is it because this mythical “weekend migration” was an abject failure, and was rolled back to Windows XP, and everybody in the department celebrated by hitting the Web, big time?

    OF COURSE IT ISN’T. It’s straight-forward statistical variance.

    But if you don’t have the real, underlying, data to hand, oiaohm, and you feel emboldened to make up pathetic little stories about “data migrations” based on a small number of consolidated data points, then you have to be open to equally silly stories, such as a temporary Windows XP resurgence.

    (Which resurgence, to make myself plain, I don’t believe in for a second.)

    This is why you leave the analysis of Big Data to the Big Boys, with Big Servers and Big Algorithms.

    And, trust me, they get it wrong often enough.

  37. DrLoser says:

    As an example of why we shouldn’t trust a bare graph with very few data points, I’m going to try to link to the latest StatsCounter graph on the current topic.

    Let’s recap: Robert sees it as a massive 2% spike in Linux interest, which it certainly seems to be.

    oiaohm sees it as evidence of a weekend migration. I think. Although he’s recently recanted. I think. These things are difficult to tell.

    I made the playful suggestion, in line with what Google or Bing would call “experiments,” ie they need to be tested rigorously, that it was something to do with student politics. But let’s look at the current graph, shall we?

  38. DrLoser says:

    There is a video out there of it.

    Which you have kindly failed to link to.

    And on the 26 now we see the predictable fall…

    Predicted by whom? Not by you, oiaohm.

    …and its normally over scaled. The first 7 days after a migration is always higher Internet usage.

    Do you mean, over-scaled as compared to the norm? Or do you mean “over-scaled as I predicted but didn’t bother to share?” Is there any statistical justification for this gubbins? I think not.

    And is there any evidence for this seven day spurt you insist upon? It seems unlikely in an office environment. And where would the scale come into consideration? So many questions, so few babbling answers.

    If you like it or not migrations have a predictable pattern.

    Whether you like it or not, oiaohm, there is no evidence that a migration actually happened.

    Over the last 4 years none.

    So, we agree it’s abnormal, then? Jolly good. Carry on.

    Also there have been no documented major migrations either in Venezuela…. Migration spikes are intermittent beasts.

    OK, I’m going to give up arguing with you now, oiaohm. Apparently the different voices inside your head are doing an excellent, if distinctly incoherent, job of arguing amongst themselves.

  39. oiaohm says:

    Venezuela IT administrators did document their processes of migrations Loser and present it at a Linux conference. There is a video out their of it.

    The thing do you think I am only looking at Roberts graph. I am also looking at statscounters current. It was not one data point by the time I commented.

    And on the 26 now we see the predictable fall and its normally over scaled. The first 7 days after a migration is always higher Internet usage.

    Sorry robbert might have been referring to a single data point but I was not.

    If you like it or not migrations have a predictable pattern. This one might be abnormal where its a migration that has happened and it returns to the same line. Over the last 4 years none. Also there have been no documented major migrations either in Venezuela. To see the last set of spikes you have to go back 5 years. Migration spikes are intermittent beasts. People only looking at short term data see them as something strange not something normal.

  40. DrLoser says:

    Migrations Venezuela as always weekends. To be correct always comes on line Saturday. Staff know by Monday they have to be able to use it. Historically the staff their don’t get paid for retraining. Fail to retrain fired.

    Ludicrous, oiaohm. You know this from personal experience, or is this just a risible fantasy?

    Even the size being around 2 percent is not exactly abnormal.

    As recent discussions on the subject of UTF-8 have proven, oiaohm, you are incapable of dividing 8 by 2 and getting anything like a sensible answer. So, let me ask you this: how many of these “not exactly abnormal” spikes do you think we can fit into Venezuelan desktop usage in the last four years?

    And if it’s not exactly abnormal, then why are we talking about it?

    There is a problem loser as well you asked for a months time.

    Listen, sonny jim, I really don’t care about your problems. Please stop trying to invent more of them: it would make your life happier and less paranoid, I think.

    It isn’t a problem at all. I don’t insist on a revisit in a month. I don’t insist on a revisit in two weeks. I don’t care if it’s six months. What I would expect is a revisit in whatever time frame seems to make sense to Robert. Otherwise I think we can safely see this as a one-off spike.

    If we’re going to narrow down a time frame, however, how about a year? That should be plenty of time to see another 2% spike, or maybe just an incremental month-on-month growth from 8% to 15%. I’m not going to pre-judge this. People like you, oiaohm, who spin torrid fantasies off single data points when they don’t even have the detail bebind the data point, are frankly pre-judging matters to the point of hysteria.

    Enough of oiaohm. There’s more, but rather than being unfair and cherry-picking the more insane bits, I thought I’d do it paragraph by paragraph until I got bored.

    I’m going to reiterate my basic point, which is that without having access to the petabytes of data behind a graph like this, all you can do is to make informed guesses. Robert’s informed guess, my informed guess, oiaohm’s wish-fulfilment guess … it doesn’t really matter. Without further analysis, it’s just a guess.

    And here I’ll concede Robert’s point: if we had access to those petabytes of data, we could test our informed guesses against reality.

    Sadly, I suspect the actual result of such access would be another oiaohm-friendly graph. Confirmation bias tends to do that to you.

  41. oiaohm says:

    Migrations Venezuela as always weekends. To be correct always comes on line Saturday. Staff know by Monday they have to be able to use it. Historically the staff their don’t get paid for retraining. Fail to retrain fired.

    Sorry Loser the timing of spike lines up exactly for normal Venezuela operations. Even the size being around 2 percent is not exactly abnormal. The question is what company or what department migrated their computer systems.

    When the numbers from the change settle down you will expect the graph to go on growing from between 7 to 8 percent market share. Instead of the 6 to 7 percent market share it was travelling on.

    Just because a country is having political issues does not mean its internet activity changes at all.

    There is a problem loser as well you asked for a months time. If this is government that department change will normally not be isolated. 2 to 4 weekends of following spikes and growth. Or how they moved from about 0 percent Linux to about 4 in 2004.

    It will be interesting to watch the next month or two. It will be more clear if it government or company. Company migrations there are exactly 1 day. Governments are many.

    It could be quite a ride.

    Real justification is hard to find out in areas with dictatorial management due to the political noise in area company doing migrations will not make news. Yes company avoids bad press from fired staff at moment.

    Even without website details there is other information you overlooked loser. Timing is the big one. Other thing is how their countries IT policies are applied.

    Yes to us the idea is office shuts Friday that night IT staff go nuts and re-image all the machines. Sat you are in office learning the system and debugging it and Monday you have to use it. Is just unbelievably savage. Venezuela that is normal.

    The other weekend before the 26 also had events that should have had people on the internet. No spike. This is why what we are looking at cannot be political directly. Political directly other events of the year would have created spikes.

    I suspect political indirectly. Frozen budgets partly to blame.

    Loser you have January today of base line data with the political disturbance. Using the base line data takes out lots of possibilities even without known what websites were accessed. Only things in Venezuela data known to give spikes like this is company or government department OS migrations.

    Please note those spikes are OS neutral. Windows and Linux migrations have both generated them. Its normalish behaviour for Venezuela. Most likely cause is XP end of life effect. So most likely will be more.

    Ethiopia number also match up to the pattern you would see when their government Migrated between versions of Windows. The only thing that is slightly odd is that is Linux.

  42. dougman says:

    Chavez even understood the importance of Linux back in 2006, before Snowden divulged the NSA crap.

    http://www.foxnews.com/story/2006/03/31/venezuela-chavez-touts-linux-as-microsoft-alternative/

  43. pogson says:

    oiaohm wrote, “Result after migration is increased internet activity…”

    In that case we are in for quite a ride because that “spike” was a weekend.

  44. DrLoser says:

    Correct me if I’m wrong, oiaohm, but doesn’t that mean that, absent detailed figures on who pinged which website and why, you are agreeing with me?
    This does not bode well.

  45. oiaohm says:

    The lots of spike like this appear in Venezuela numbers in 2004. Venezuela when they do Linux migrations in government or companies is not like Munich or other developed countries. The migration time frame is 24 hours for Venezuela. Venezuela companies and government departments are dictatorships. You use what you are provided with or you are fired. No debate no resistance is possible.

    Result after migration is increased internet activity as people teach themselves how to keep on doing their job on the new system. By the way the same effect happened in Venezuela every time they migrated Windows.

    History also says the spike halves after people have worked out how todo their job and the average growth rate of Linux increases minor-ally.

    The reality is only thing known to cause a spike like that in Venezuela numbers is a company or government migration. The loser was right a spike like this cannot be individuals.

    Not all government departments of Venezuela have migrated yet. Companies who depend on Venezuela Government contracts are also up the creek at the moment.

    Linux market share growth increases when ever there is financial problems in an area.

    Redhat in the global financial criss had the best year ever. This is just how Linux is.

    Depending on the IT implementation methods in a country government department and company OS migrations either a increase in the base rate growth or create spikes and increase base rate growth. Yes the the spike pattern you can make a list of countries that do it.

    Venezuela and Ethiopia are places where companies and government departments internally are dictatorships. Where staff must us exactly what they are told and you don’t have to bother about slow migrations.

    How migrations show up in numbers link to how internal policies are applied.

  46. DrLoser says:

    Still not seeing a real justification for the 2% spike in five days, Robert.
    Sorry.
    It’s a tiny market, I know, but I’d need to see the figures behind the spike.
    Can we see the same graph in, say, a month’s time?

  47. DrLoser wrote, “it’s futile to read anything into a five day spike”.

    I think you are not seeing the forest for the trees. A decade ago, this behaviour was ignored or laughed at by many in IT. For the past 7 years this behaviour was said to be impossible for all kinds of reasons by visitors to this blog, but it’s happening and rather quickly for all kinds of reasons: lack of money, political will, chaos in the economy/government, death of XP, etc. What I have been predicting for years is coming to pass. GNU/Linux is not a niche/1%/insignificant/amateur affair. It is the future of IT or at least a big part of IT. The narrative has changed from the dominance of Wintel to in what fraction of IT will Wintel remain relevant. That’s the right way for things to go. Instead of being slaves to Wintel, we can choose our hardware and our software and do what needs to be done without being forced to do what Bill Gates and friends want. On top of that the world is beginning to pay for the real cost of IT and not welfare for Bill Gates and “partners”. The world has started to invest in its future, not M$’s.

  48. oiaohm wrote, “This will just make the linux change in these places entrenched possibly too deep for Microsoft ever to get back market share.”

    Good analysis. Thanks. One way or another M$ is coming down. I like it. Watching this slow-motion train-wreck is almost as much fun as I had hiking in the bush yesterday giving free rides to wood ticks. I enjoyed the fresh air and exercise. We saw a hare and heard a bunch of frogs and song-birds.

  49. oiaohm says:

    http://en.wikipedia.org/wiki/Category:Protests_in_Venezuela

    Remember Venezuela major Linux migration was 2004. There was no up kick in 2007.

    Venezuela up lift aligns with end of life action in schools and government agencies..

    Venezuela has had protests before with no GNU/Linux upswings. So theory that its student interest in political sites does not justify it. and removal of XP has not been mandated.

    Also the first protests start in January 2014 not may. This is the problem the 26 of may does not line up with the population doing Internet look ups. There were other events where we should have seen upswings as well.

    There is something to remember Policy Paralysis is the state Venezuela is in. This means your IT budget is zero dollars for new hardware or software. Make do with what you have or can get for free. In fact it impossible to bribe anyone as it not possible to perform a purchase order.

    And there is no sign that in the near future this is going to be resolved. IT officers still have to perform their job.

    So it should be normal to expect some third world to end up 50+ Linux. Events will happen where its impossible to sell closed source software and the people running IT resources have to make do with what they have or can get for free.

    Robert Pogson in this case its all about costs. The fact the school systems have invested in Linux is allowing them to keep operating with somewhat current software.

    Same kind of mess is possible in the USA if the budget does not pass and both sides get into a long dragged out arguement.

    Third world countries are more likely to have very bad political problems so prevent software acquirement with money. In fact the 2004 migration in Venezuela tracks back to a prior block in supply issue.

    Sorry some people just don’t understand that countries with greater than 50 percent Linux will happen. Reason for happening is no choice in the matter.

    There are governments around the world who operations collapse into these states quite often. Its a perfect storm. Microsoft does not have the money at the moment to throw in freebies. The government acquirement systems are stopped and the common closed source OS they have has gone end of life. 50+ percent Linux here we come.

    This will just make the linux change in these places entrenched possibly too deep for Microsoft ever to get back market share.

    This is why I said all Linux has todo is sit back and be patient. Linux will gain market share the natural political storms will come. Microsoft/closed source software model does not operate in unstable political environments . FOSS model where sales are not required can operate in unstable political environments.

    By this point those IT officers in Venezuela know they are in for long term funds disruption.

  50. oiaohm wrote, “Linux runs on the older hardware they have. So no hardware costs. Staff costs deploying their wages are bugger all in most cases.”

    Yes. It’s not just about costs but also about performance. By the time IT has grown up in these countries, GNU/Linux will be the “go to” OS, there will be no shortage of people skilled in GNU/Linux and they will have great performance. M$ has been enslaving and taxing the rest of the world in multiple ways for decades. The next billion users of IT will be freed from all that.

  51. DrLoser says:

    This is what I mean about pre-judging the numbers, Robert. I have a hard time believing that 2% of Venezuela’s desktop users have downloaded Linux over the last five days and made a permanent shift to it. But I’m prepared to admit that I could very well be wrong.

    And “politics be damned” is an unfortunate statement under the circumstances. In fact, I suspect it’s another pre-judgement, because it’s possible to hypothesise that this page-view jump is all about politics.

    As we all know, the Venezuelan government has invested heavily in getting Debian/Canaima into the school system. Which may be a good thing. However, it does suggest the following possibility: what if May 20th to May 25th saw an upsurge in student interest in political/news sites? The kids aren’t going to rush out of school to find another computer to browse off, are they? No, they’ll use the one they’ve got.

    There are doubtless other possibilities. My only point here is that, without the ability to look deeper into the numbers (specifically, the web sites accessed), it’s futile to read anything into a five day spike.

    Still, it seems to keep oiaohm happy. I haven’t seen the little fellow babble quite as many yes, yes, yesses before the inevitable “perfect storm” for quite a while. Shame he still makes no sense whatsoever.

  52. oiaohm says:

    Yes you are seeing the effects of the XP triggered Windows Market share collapse.

    These third world countries will be first. This is a result of a Microsoft anti-piracy push a few years back.

    UK and USA and other developed countries have been able to pay for XP extended support. Yes it was cheaper to extend XP for another 2 years than it was to upgrade to Windows 7 or 8. Ethiopia and Venezuela don’t have the money todo this. Also all most all their installed versions of XP were acquired for less than normal sales prices under anti-priacy plan of Microsoft. Part of this allows Microsoft to send personal to inspect. So the countries who agree to this cannot just go out and pirate Windows 7 or 8.

    Yes this is the funny one in the past neither of the countries could afford Windows without Microsoft discounts and today they still cannot and currently no Microsoft discounts on the table. Microsoft is not in as flush position to be handing out discounts left and right at the moment either. Yes when Ethiopia and Venezuela got their XP machines Microsoft had a decent Mobile Phone market and the PC market was not in decline. Microsoft had the idea give these countries PC’s this will cause wealth growth so in future they will be to afford to pay. Has not happened.

    Yes a majority of PC’s in third world countries happen to be government owned. Yes this is also the effect of global debt control where countries don’t have a unlimited allowance to keep on taking out loans.

    So it should be expected for a percentage of countries to go over 50 percent Linux due to have no other choice. Linux runs on the older hardware they have. So no hardware costs. Staff costs deploying their wages are bugger all in most cases. Some cases the migration cost in staff for them will be equal to 1 copy of Windows for us and maybe 10 for them.

    The important but here. The increase Linux user-base will bring more diversity to Linux development.

    These events have been predicted for along time. It was just a question when the perfect storm would form up. Traditionally when Microsoft was doing bad in mobile the PC was booming and vice verse. Of course Microsoft doing bad in both at the same time had to happen.

    Linux world is insanely patient. Linux could have spent huge amounts of money marketing and not get very far. Or be smart and just get ready for the upcoming collapse. Most were predicting that MS would have extended XP for another 2 years before throwing in the towel. So Linux world has been caught a little on back-foot but not much.

Leave a Reply