Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
Linux Software

The Internet Operating System Counter for 4/99 88

Posted by Hemos from the counting-them-up dept.
AshNazg writes "The Internet Operating System Counter project has published results for April 1999. This is very important because it's one of the few ways of reliably counting Linux hosts. Linux is show to be the leading OS in the survey with 31.3%. All Windows versions combined account only for 24.3% "
This discussion has been archived. No new comments can be posted.

The Internet Operating System Counter for 4/99 More

The Internet Operating System Counter for 4/99

Comments Filter:

  • Re:The method in this servey is wrong (Score:1)

    by Anonymous Coward
    My guess is that machines hidden behind firewall like thingies are represented by the random sample of exposed machines.

    I've got a couple of personal anecdotes that both back this up and tear it down...

    First, behind the local cable modem service provider's firewall (a FreeBSD deritive), all the server are unix based and many are of the free/OSS variety.

    Second, the only acceptable way to secure co-located NT boxen, that have varying levels of system administrator attention, is with a good (non-Microsoft) firewall product.

    So, there you go... 6 of one / half a dozen of another.

    We all know what is happening to the industry. It doesn't take samples or statistics to verify the success of our Jihad and might of Natural Selection.

  • Definition of Big Mainframe Iron? (Score:1)

    by Anonymous Coward
    You're on the right track, but I think your definition of "Big Mainframe Iron" is a bit off. These are the systems that run IBM's OS/390 such as IBM and Amdahl.

    The HP, IRIX, AIX, and DEC are essentially the traditional "UNIX minicomputers" and can be lumped in with Solaris.

    At least, to me it does. Anyone disagree?

  • Re:The method in this servey is wrong (Score:1)

    by Anonymous Coward

    Good point. And to add to this, it doesn't account for how much work these machines are doing. One HP/UX system can do the work of several Linux 486 boxes.

    Also, almost 13% of the hosts queried were not recognized, and over 6% were identified as Linux 1.2 system. Over 6% of the servers on the internet are Linux 1.2? I find that a little difficult to believe.

  • Linux up 2.8%. BSD down 0.4% (Score:1)

    by Anonymous Coward
    The stats show that Linux, Apple, and Netware were the only OS's to gain market share. Linux was the winner, followed by Apple, and then Netware. All other OS's have lost market share. Here is an executive summary of the gains and losses:

    OS ........... Change
    -----------------------------
    Linux ..... +2.8% ... gain
    Apple ..... +0.5% ... gain
    Netware ... +0.1% ... gain
    -----------------------------
    Win ....... -0.1% ... loss
    AIX ....... -0.3% ... loss
    BSD ....... -0.4% ... loss

  • Posted by keyva:

    While these are great stats, and I realize this is a global audience, it should be noted that "For this OS count the RIPE Host Count was used to collect host addresses"... RIPE [ripe.net], from what I can tell, is a coallition of European IP hosts, so I don't think any US data is included here...
  • Posted by Soco:

    There are like 20 of these throughout the internet. They are lame and don't measure shit. In some of them, OSes like BeOS and AmigaOS are kicking everyones asses, and other ones like Windows are kicking ass. The problem comes with dialup users who visit the site quite frequenty (most are probably windows and macos users) who get a different ip.
  • On the other hand, exposed machines are where the most important work gets done. These machines run the Internet; boxes behind a firewall may help an individual or a company, but they do nothing for the Net at large. It's hard to prove what's getting used behind the masqwalls/firewalls. What can be shown is that microshaft a has 25% share of the machines that power the Internet. That says a lot, and for people who wish their internal network worked as well as the Internet, it forces a close look at technology choices.
  • I would like to see a breakdown in a table or something about how this survay compares to the one done by Netcraft [netcraft.com], because they seem similar to me (and I don't have the time to do a detailed analysis myself).

    Anyone see the obvious diffrences?

  • Shit... I wish there was a way to kill your own comments. I just woke up 15 minutes ago, and it took me like 5 seconds (or less) looking at the two side by side to see. Netcraft only checks out httpd's where this one estimates OS's by checking at least httpd/ftpd/news...
  • Perhaps there were no OS/2 www, news or ftp servers on the boxen they surveyed? And as for DOS and Win31... I'd suspect there's practically no such thing as a webserver running DOS or Win31 that can be categorized as something other than a "because I can" project.
  • here http://www.netcraft.com/Survey/ [netcraft.com]
    also the other method is not correct. you cannot really run a "commercial site" using windows 95/98 as your webserver/ftpserver/nntpserver...
    --
  • A power failure, to ruin the uptime ;)
  • Very nice post. I am biased, being from Nigeria, but just as it was _very_ sweet to read about the Anglo-Nigerian university cooperative for computer connectivity based on Linux (it was in a recent Linux Journal), I am always glad to see how open-source empowers those you might not expect it to: from intrepid third-world citizens who haven't left in the Great Brain Drain to beleaguered victims of government oppression (I'll have to find an article about the Chiapas/E-mail thing). All I can say is, why wan't such a post moderated up? Viva la revolucion --Uche

  • I looked and looked but I could find no sign of a link to asses... and what do donkeys have to do with it anyway?

  • Just how are these hosts found? I can't imagine that a serious site would be run on 95/98, more likely just a hobbyist with a web server running on their system, who just happened to be connected to the net at the time their system was queried.

  • If you look at the real numbers, you see that Sun's up 32,000 hosts! It's just that other systems like Linux, BSD, and NT have grown faster.

    In fact, all the OSes have grown, even SCO, Digital/Compaq Unix had the slowest growth.

  • Agreed..

    If this survey showed "All webservers" as opposed to "dedicated" or "professional" webservers, then the hobbyist OSes would show artificially high results, although I'm sure that there are many more Linux and BSD "dedicated" webservers than 95/98.

  • They're not counting total computers, they are just counting those that act as servers on the Internet.
  • Maybe there's still some hard feelings after Microsoft's Hostile takeover attempt [interlog.com] a few years ago! ;-)
  • It's perfectly valid as a measure of the OSes that people trust their web sites to.

    It doesn't claim to measure all OS use, just webserver OS use.
  • Many of you seem to be missing that
    this is only a small subset of machines,
    namely those with host names www, ftp,
    and news. This will excluded most of
    the dial up users etc and result in
    a much smaller number of hosts.

    From their web site http://leb.net/hzo/
    "All host names beginning with 'ftp.',
    'news.' and 'www.' (case insensitive) were queried."

  • Am I reading the numbers wrong, or are 80574 servers (6.3% of the exposed total) still using Linux 1.2.xx?! C'mon, what's it going to take before they upgrade?
  • that survey is nice to read. although ASCII is a bit outdated, character-cell displays will always be around.
  • Do not be so fast on opinions. The survey is not as old and broad as netcraft, but it definitely has its value because it explains how did they run it.

    And the "survey problem" is not with the OS counter or netcraft, but the way people interpret them and quote them. For an example see the introductory to the link on slashdor ;-)
  • Queso miscounts some unixwares as reliant. Reliant is NOT that popular (in the open world).

    It still runs wuite a lot of the world GSM traffic ;-)

  • The method in this servey is wrong (Score:3)

    by arivanov ( 12034 ) on Thursday April 29, 1999 @07:26AM (#1911091) Homepage
    Despite by admirations to this survey (Its has actually been going for quite a while) I have to admit that its results are very wrong (as an OS count), because the system counts only exposed machines.

    Machines behind firewalls, on 10.x.x.x 192.168.x.x networks behind NAT (masquarading) are not counted. And most of these are WinXX.

    Overall, the results actually reflect not the OS count, but the server OS count on the net.


  • There is also some data for Africa and central Asia. One particularly nice fact is that in Chad, Niger and Liberia, 100% of all their reported web servers are running Linux.

    Of course, there are less than 25 total servers between these three countries, but in the days prior to Linux and Free BSD there probably would have been zero. Who can afford or get access to Windows NT Server in Chad? Microsoft's east Africa support site [microsoft.com] is littered with information on how to stop "piracy." The closed software model makes poor people interested in computing into criminals.

    These data from Africa are evidence of Linux's awesome potential to give developing countries the chance to make their voices and concerns heard. This is exactly what has happened in Chiapas, Mexico, where the lives of Native American activists have been saved because of their ability to send email to concerned people in the rest of Mexico and in foreign countries. These people in turn put pressure on the Mexican government to keep their abuses of human rights in check.

  • Like good sysadmins who have queso-proofed or OS-forged their machines to fend off OS-specific attacks. I think OpenBSD might have some of that built-in, but I can't remember for sure.

    -lx
  • well, like Linus is always fond of saying in regards to his kernel updates...if you don't really need to upgrade your kernel, you probably shouldn't. If everything on their system (the 1.2.x people) is running to the best of their satisfaction well...if it ain't broke, don't fix it :)

  • Why is that so hard to believe? There are lots of servers that have never been upgraded simply because "if it ain't broke, don't fix it".
    ---
  • And believe me - here people don't cough up 50 bucks for software every day (Red Hat 5.2). Most of them haven't bought any software even if they use a computer from like 10 or 15 years


    You don't have to pay US$50 for each copy of Red Hat. You can:

    1. Download it for free.
    2. Copy it and pass it to friends.
    3. Buy it from Linuxmall [linuxmall.com] for US$1.89/copy (+ shippping)
    http://www.linuxmall.com/

    Please be aware that there is no reason at all not to copy and give away the RedHat Linux you have.

    The GPL permits it, and if you were to start burning RedHat CD's and passing them to friends in Bulgaria, Red Hat's president, Bob Young, would actually be happy about it. This is exactly what Linuxmall.com is doing. The US$1.89 covers the cost of the blank CD and the burning.

    Red Hat's business model is to make money from support and brand-name marketing. Why US$50 for an official Red Hat CD/Book?

    1. Some support is included
    2. Many people would rather pay for a CD/Manual prepackaged than download it.
    3. Large businesses don't trust software they don't pay for.

    The department manager where I work actually said "We won't use free software because we can't sue anybody if it breaks." (Yeah, like we sue M$ everytime NT crashes halt the production lines...")

    Also, the corporate world doesn't trust anything they don't have to pay for, so Red Hat will happily charge for what they also give away free.
    Remember, Red Hat didn't write Linux. They collected the peices and put them together. If you want to start your own business burning and selling Red Hat 5.2 CD's in Bulgaria for $15 each, Red Hat won't try to stop you. (Just remember your first customer could do the same thing...)

    Heck, if you promise to sell/give away a few hundred copies, they'll probably send you a free one to get started. They would see you as an unpaid volunteer, rather than competetion.

    I just read Bob Young's chapter of Open Sources a few hours ago. Bob (president of Red Hat) talks about areas of the world that can't afford to buy software. It's a competetive advantage, because with a free product, he gets extreme market penetration as people pass copies around.

    As your business grow and need software support contracts, Red Hat will be there to supply them. The US$50 is only for a single user for a month or two. Rumor has it the big corporate support contracts run up to US$60,000 per year.

    Since Linux doesn't break, after the setup, it's lots of support/service contract money for no work.

    That's why Red Hat want's everybody using Red Hat.

    So don't worry about the copying. It's legal.

    Good luck!
  • Agreed...by restricting the search to hostnames of www or ftp or news, they have eliminated from consideration all workstations which are also, directly connected to the Internet. As a survey of Web, FTP and News Server OS's, the results may be a little more representative.

    However, not including .com, .org and .net (not to mention other top level domains) in the survey could also skew the results due to the differing roles the servers may play in different organizations.

    Sean
    Sean Brown
    Linux Evangelist
    "I'll let you be in my dream if I can be in yours." - Bob D.
  • It's still a fascinating report, but definitely follow the link to asses what's being reported.

    There are no reports available for the .com .net or .org domains, though there appears to be thorough reports for all of the top level country codes.

    bnf

  • The website [leb.net] for the project specifically says they only queried hosts with names beginning with "ftp.", "news.", and "www.".

    It's not intended to be a survey of computers connected to the internet. It's only a survey of computers serving content.

    -dP
  • Actually I distinguished Solaris for a couple of other reasons:
    1. First and foremost, AFAICT it runs on a much wider price range of machines, from lower end workstation, workgroup, through mid-range servers all the way up BIG IRON companies like HP, IBM, etc. would rather hang out. Which means that while a Solaris-based machine has traditionally not been the machine for "everyman", it is available at a reasonable price.
    2. Solaris "gets along" with the rest of the world, vs. M$ which is trying to take over and eliminate the competition with crappy products and predatory business practices.
    3. As you surmised, the third reason I split them out is that they have significant market share. If you look at the percentages a different way, Solaris has around 70% as many servers as the entire WinXX community.
    And although I don't own a Solaris machine, you know what I like about them the most? They do good work without badmouthing or FUD'ing the rest of us.
  • These statistics don't surprise me at all. Other posters have commented about machines not exposed to the survey, statistics counted or not counted, etc., but IMHO we don't need to split hairs in order to see the value of this survey. An analogy comes to mind...

    Consider the major OS's as combatants in a "virtual world war." On one side, we have BIG MAINFRAME IRON (HP,IRIX,AIX,Digital Unix), sort of like the USA's M1 tank. On another side we have the beast with it's armies from Redmond, sort of like a Naval task force -- alot of power but hard to move around. A third side represents Solaris (something like the Air Force (?) Big, fast, and powerful. (IMHO mostly friendly to the next group). Now then, the "everyday people -- mostly inexpensive machines running Linux & the BSD families. We're sort of like the infantry but all of a sudden the infantry has the software (2.2X kernels) and hardware equivalents (newer 300MHz + hardware) of 25th Century MechWarriors with extended air-support. Smart missiles galore (Samba, Apache, etc.), and the best radar on the entire battlefield (/., etc.) Built in the back alleys of the computer world and assembled for the battle on all fronts. Look who's ahead:

    1. Linux + the BSDs...........45.9%
    2. The Beast (M$).....24.4%
    3. Solaris..................17.7%
    4. Big Iron..................7.5%

      Total of #'s 2-4.......49.5%

    Which means that if Linux/BSD's capture just an additional 1.8% of the hosts [reached by the Internet Count] from these three sources, our "forces" will be as big as the next three groups combined .

    As Linus himself said (remember, he was joking at the time!!) "total world domination!!" Power to the people!!

  • OScategorygrowth'market'share
    -------------------------------
    Linux 39.24%42.28%
    Windows95/98/NT 26.50%24.39%
    BSDFamily 23.47%13.29%
    Solaris/SunOS 19.60%13.12%
    Mac/Apple 64.18%3.89%
    IRIX 9.77%1.94%
    NovellNetware 63.62%0.87%
    HPUX 12.90%0.50%
    AIX 6.13%0.40%
    SCOUnix 6.18%0.09%
    DigitalUnix 3.39%0.06%
    ReliantUnix/Sinix -17.37%-0.84%

    This uses the TT and BR tags, along with actual non-breaking spaces (option-space on IE/Mac -- YMMV otherwise).
    --

  • Maybe I just don't quite understand the survey but printers with there own IP address??

    "HP/JETdirect Printer (old model) 1480 "

    This could give a whole new meaning to spam. :)

  • Exactly. I have very little confidence in the
    results of this survey.

    When I run queso it almost always misidentifies
    HP-UX boxes. Sometimes it calls them BSD,
    sometimes it calls them Cisco routers.

    On the other hand, "nmap -O" always seems to
    get it right, even when I point it at weird
    devices like X terminals.
  • Thanks for cleaning that up, and the clue about real non-breaking spaces.

    I'll try that next time.

  • Interesting - look at the "market share" (Score:3)

    by AJWM ( 19027 ) on Thursday April 29, 1999 @10:11AM (#1911106) Homepage
    The numbers as presented don't show some of the more interesting things that can be learned from them. I took the numbers and worked out the percentage change for each OS, and then the change in each OS as a percentage of the total growth in the number of servers -- i.e, "market share" for the months since the last count. Here they are:
    OS______________category_growth__'market'_share

    Linux______________39.24%__________42.28%
    Windows_95/98/NT___26.50%__________24.39%
    BSD_Family_________23.47%__________13.29%
    Solaris/SunOS______19.60%__________13.12%
    Mac/Apple__________64.18%___________3.89%
    IRIX________________9.77%___________1.94%
    Novell_Netware_____63.62%___________0.87%
    HPUX_______________12.90%___________0.50%
    AIX_________________6.13%___________0.40%
    SCO_Unix____________6.18%___________0.09%
    Digital_Unix________3.39%___________0.06%
    Reliant_Unix/Sinix_-17.37%_________-0.84%

    (Sorted_by_'market'_share.)
    (Sorry about all the underscores, but $nbsp doesn't seem to work, neither does the PRE tag. (Rob, can we have the PRE tag, please?)

    The only OS that's actually lost sites is Reliant, but they only account for 0.84% of the change in the installed base. It looks like new server installs are choosing Linux at near twice the rate of all Windows combined, with BSD and Solaris up there too. Mac and Novell show major jumps relative to their previously installed base, but are still minor players in the server field (although Mac is catching up -- wonder how much (if any) of that is the Unix-based Mac OS X?)

  • The results [leb.net] has a breakdown of Responce details, and it mentions various versions of Linux, but there are NO version 2.2.x... To me that's rather strange.

    Maybe it's just the OS Counter that needs to be told about Linux-2.2. But does it mean that Linux-2.2 machines are categorised wrongly, or does it mean they're not counted at all???

    I hope to see even more Linux machines turn op here, and generally everywhere else!

  • OK, people, I'm from Bulgaria and the results for the .bg domain show:
    Windowses - 41.9 %
    Linux - 41.7 %

    OTOH I cannot help feeling like a pioneer because when I suggested to my boss to place the web server (www.protos.bg) on Linux I had 0 experience with any kind of Unix and I knew 0 people that could help me with that (the latter hasn't changed).

    I was just lucky that the web server and the ftp server ran right out of the box. (I'm still struggling without any success to start up a mail server.)

    And believe me - here people don't cough up 50 bucks for software every day (Red Hat 5.2). Most of them haven't bought any software even if they use a computer from like 10 or 15 years
  • Thanks for your post but I was aware of the legal stuff.

    What I was trying to say was that people here don't like to pay for commercial software because everyone else just copies it. At the same time I managed to convince the owner to buy Red Hat Linux just to help the movement a little bit. Because I really am impressed how much they return to the community.

    What is more is that we already had the whole 1-st CD downloaded. The only drawback was that we didn't know how to burn a bootable one with it.
  • Overall, the results actually reflect not the OS count, but the server OS count on the net.

    Agreed, the count does reflect servers, but I thought that was the whole point... You want to know what the OS of choice is in terms of commercial UN*X vs. NT vs. Linux vs. X, etc. And all of those are server systems.

    I don't think the problem of machines "hiding" behind NAT firewalls, etc., is what's skewing the results, it's the exclusion of the other domains that's making the results not entirely trustworthy ATM.
    Herbert von Kammerstein

  • No, these are servers, not clients. Their page says "all host names beginning with 'ftp.', 'news.', and 'www.' (case insensitive) were queried." Presumably very few or none of these are dialups with dynamic addresses.

    Mark

  • On this page, they break down their count by country domain (e.g. .uk, .de, .au, etc.) and they show stats for .edu -- but there is no break down for .com, .net, or .org.

    Does this mean they don't query those sites? While I like to see stats that raise linux and diminish Win, I don't think these stats can be reliable with these three TLDs left out.
  • It is very interesting to look at the breakdowns as opposed to the aggregate percentages.

    First of all, this survey leaves out .com, .net, and .org (as lots of people have already pointed out), which makes it incredibly inaccurate to model the US, at least.

    I'd argue (based soley on the .edu domain, which should have a high linux percentage compared to other predominately US domains), that linux does much better outside the US then within the US. I don't know why this is, exactly, but, for example, .de (Germany) had almost 45% linux. Whereas .edu had only 10% linux, with Solaris and windows neck and neck for the top.

    I'd guess that outside of the US (especially a lot of the poorer countries) price is more important than it is for a corporation in the US. Also, linux probably (not sure on this, because I have no experience with Windows) does a better job at internationalization than windows does; at least redhat ships with the howtos in more languages than I imagine Microsoft has translated there documents into.

    So, I don't think this survey can accuratly represent US/large corporate data (because I imagine large corporations outside the US also get .com domains), but I think it does show to some degree how strong the international support for linux is, which is something that, I, a US citizen, often forget about, and something a thing the US media is completely ignorant of, but which is an incredibly important part of the linux/open-source movement. It is truely global.

    Mike

    Mike
  • Legend has it that an ISP in my area had a networked HP printer and really weak security.

    They came in one morning to find ~600 pages of 36-point boldface "Your ISP Sucks"

    ...just wanted to share that with the class. :)
  • at work, our tektronix phaser 740 not only has an IP address but runs a webserver too! you can slam a four gig harddrive in the suckers for those really bad-ass postscript jobs -- but forget that -- i'm want use it to serve my MP3s...

    let the Y2K nazis find that!
  • well that's excellent news. less asp errors and unresponsive servers by the day! i'm currently moving all my web projects across to linux. it's great -- they don't crash anymore.

    unless of course someone rips all my pci cards out of the box and then reboots the box [all when i'm not looking] -- i guess "mad max" has some nice tips on how to deal with this sort of behaviour!
  • I looked at the main page and they have only counted a little over 1.4 million computers. That to me seems a little low. Heck, we have nearly 3k computers at work. And we are not that big.
    If they combine the data with a NetCraft (No, not the MS people) survey, then a closer count may be had. That would bump Solaris up a lot and unfortunately, the WinFamily. Linux wont take a huge hit though except in the percentages. It seems linux is more popular right now with the .edu set and the non-com people. This will change.

    RB
  • I don't think queso distinguishes 2.1.x from 2.2.x. Presumeably, the two lines have not been out long enough for a distinguishing characterisitic to emerge for OS fingerprinting. For instance, when my box was 2.0.x, it showed up as that. When it was 2.1.x, it showed up as that. But now that I've got 2.2.x, it stills shows up as 2.1.x.

    -Jaac (Just Another Anonymous Coward)
  • Probably, if they've been running it for this long on there system, it's "mission critical", and after this amount of time, has been tweaked so that it's so stable on that hardware that they don't feel any need to upgrade: it hides in the closet, and keeps on running.
  • I think that was the point. Nothing outside of Europe was included in the survey.

    That makes it a bogus survey in my mind.

Slashdot Top Deals

Talent does what it can. Genius does what it must. You do what you get paid to do.

Close