But
The big "but" is that, according to Dan, the 1.7% figure accounts for "paid shipments only," which Dan readily admits may only include a small fraction of all desktop Linux installations. To begin with, he says IDC's research shows that for every 10 copies of Linux sold, approximately eight copies are downloaded from the Internet for free, and that there may be something like 15 copies made, on average, from each downloaded or purchased copy. Not only that, Dan says he personally knows a corporate IT manager who made not just 10 or 20, but 6,000 (yes, that's six thousand) copies from a single set of downloaded Linux CDs. Dan doesn't know how many of those copies went on servers and how many went on desktop machines, but either way, it doesn't take many corporate IT people installing a few thousand homemade copies of their favorite Linux distro to throw all the numbers out of whack.
Exact numbers aside, Dan says Linux desktop users "are still a tiny minority" in the world. Windows is still the Big Dog, with around 94% of all paid desktop operating shipments in the last year. Dan says IDC's figures show Mac's percentage is down slightly -- from 4.17% last year to 3.9% this year, and that "all other" operating systems account for less than 2% of the desktop total.
Remember, all the "hard" numbers are for paid shipments only. It's entirely possible that there are more *BSD desktop installations than IDC counted in its public figures.
Public vs. private figures
IDC is in the business of selling studies and reports, not giving them away. Dan says anyone who wants to see results of detailed operating system use surveys, as opposed to just shipment data, needs to either buy reports IDC has already written or commission his or her own study. (If you want to buy some of IDC's reports on Linux trends, you'd better make sure your credit card limit is measured in thousands of dollars, not hundreds, before you order.) Dan hints that some of the private reports may have figures notably different than the tastes the public gets of simpler, more easily obtained data. Perhaps so. Still, Linux is nowhere near "world domination" on the desktop. Dan says that Linux's biggest impact so far is, unquestionably, on servers. Indeed, Dan's own home office network runs a Linux server, while his primary "work" desktop runs Windows 95, which he says is still IDC's corporate standard, with a gradual companywide switch to Windows 2000 now in process.
Should Linux be counted as Unix?
Dan claims he's heard this request from more than a few Unix vendors, because on the server front, Unix and Linux combined have a market share equal to Windows if not greater. But to be officially called "Unix," Linux would have to be certified as "Unix" by The Open Group, keeper of the Unix copyright flame. So far, Dan says, "No one in the Linux community is interested" in obtaining Unix certification.
It's rather unlikely that, say, a whole lot of Debian or Gentoo developers are going to wake up one fine morning next year and suddenly decide they want their favorite Linux distribution certified as a jen-you-wine flavor of Unix(TM). But what about UnitedLinux? After all, Caldera-turned-back-into-SCO is one of the driving forces behind this consortium, and SCO is far more of a Unix vendor than a Linux vendor. Is it possible that all those Gentoo and Debian developers -- and all other Linux users and developers all over the world, could wake up one morning and find that UnitedLinux was really a kind of Unix, but that other distributions weren't?
This could create an interesting situation, possibly even a true Linux fork, except that UnitedLinux is working toward full compliance with the Linux Standard Base -- along with Red Hat, Debian, MandrakeSoft, and lots of other companies and non-profits including The Open Group itself. Would a UnitedLinux certification as "Unix" apply to other Linux distributions? If so, would anyone care besides a few SCO resellers and some of their stodgier customers? All this is just speculation, not the sort of thing Dan Kusnetzky or any other sane industry analyst would want to comment on even if we asked them.
The real Linux count
Is there such a thing? There's always the venerable Linux Counter, but it has fewer people registered than download each new Mandrake release in the first week it goes live on the mirrors, and there's a nice big "My guess at the number of Linux users: Eighteen million" note on The Linux Counter main page, with the link going to another page where the admittedly flawed methodology behind this guess is explained.
Linux is sneaky
Despite all those IBM ads showing Linux saving the game or whatever, and all the articles that get written about Linux, and all the Web sites -- including this one -- devoted to it, Linux is still an elusive, essentially non-mainstream phenomenon. It's kind of like one of those groups that comes out of nowhere and suddenly sets album sales records even though it's gotten little or no radio play.
The "stealth mode" aspect of Linux market penetration, especially on desktops, has disadvantages. If more hardware vendors saw Linux as a true up and comer, we'd see more Penguins on product boxes and laptop labels, and we'd certainly see more commercial software ported to Linux.
But on the other side on the coin, as long as Linux remains so elusive on the desktop and is so frequently undercounted, it means that its spread can easily be pooh-poohed and half ignored by a certain company that dominates the desktop operating system market so heavily that the U.S. government has taken it to court over its improper use of that monopoly. Personally, I wouldn't mind Linux keeping a low profile -- at least on the desktop -- for several years, and suddenly jumping into prominence, with ads saying, "Dude, you're getting a Dell running super-cool Linux," appearing all over the prime-time airwaves all of a sudden, and full-color ads suddenly showing up in every Sunday paper advertising the latest "Linux laptop blowout" at MegaloComputerMart locations all over the world.
Until then, maybe the desktop Linux market share numbers we see in the computer and mainstream media are accurate, and maybe they're not. I no longer care. The only desktop Linux market share figure in which I have total confidence -- and the only one I really care about -- is the number of Linux-running computers sitting on my desk, which just happens to be an easily calculated 100 percent.
Note: Comments are owned by the poster. We are not responsible for their content.
It's darn near impossible to order machines without a pre-sold operating system from Redmond or elsewhere. They are simply not available in practice. All of the desktop machines I've seen or handled since 1986 have had PC/MS-DOS, MS-Windows or Macintosh OS pre-installed. In otherwords, with a pre-sold operating system.
Yet, in the last three years, none of the incoming desktop machines in my dept. were ever booted from the pre-sold operating system that came on the hard disk. Instead, the very first start was from a Debian/RedHat/Suse/YellowDog install CD each time. On paper that looks like 0% Linux. In reality it is 100%.
It's darn near impossible to order machines without a pre-sold operating system from Redmond or elsewhere. They are simply not available in practice. All of the desktop machines I've seen or handled since 1986 have had PC/MS-DOS, MS-Windows or Macintosh OS pre-installed. In otherwords, with a pre-sold operating system.
Linux != GNU. The entire GNU system uses Hurd as the kernel. Linux just uses parts of the system (such as libc).
There is only one reason some people refer to Linux as GNU/Linux. RMS was mad about it stealing the thunder from his Hurd system. He started ranting about how everyone should say "GNU/Linux" because most (if not all) Linux distros use various GNU projects. That is true, but they also use projects from XFree86, IJG, Apache, etc. Why should GNU get special treatment?
Some people caved or agreed with him. Most just call it Linux and think Stallman is a jerk or laugh at him.
I have begun to think GNU/RMS is a bane on the free software community (free as in not charging money to use or copy it--not free as in "there can be only one license"). Yes, they wrote quite a few programs and libraries, however what gives them the right to try to dictate what everyone else does.
Would there really be no free compilers, libraries, system utilities written if GNU didn't exist? I doubt it. Maybe some things would be different, but I'm sure many people would have pitched in no matter if the license was GPL, BSD, Artistic or even POV-Ray's.
A little client that sends information back to a central server to signify a new Linux workstation is in use. You know, each client has a unique ID nuber and sends trivial yet unique data back to certify validity, like the MAC address of the primary interface or something.
A few counterpoints.
IDC and other commercial reports are largely interested in the business of GNU/Linux, which is to say, commercial shipments. Studies and methodology are going to focus on dollars. Not until they find a client who's interested in marketing to the installed base will you see solid efforts to determine how large it is.
GNU/Linux doesn't need tremendous marketshare, it needs a viable development and user base. Unlike every other competitor Microsoft has faced, GNU/Linux can't be priced out of the market. So most of the community side developers don't particularly care how large the base is. If it's big enough to give them a decent toolset and, preferably, employment, it works.
The GNU/Linux community and market aren't the only ones with an interest here. Don't think for a minute that Redmond doesn't have a tremendous interest in what the actual picture is. And they're as much in the dark as the rest of us. And are painfully aware that the statistics for MS Windows are inflated by their own marketing practices. Unless they're not, in which case, please share your numbers<nobr> <wbr></nobr>;-)
Numbers I'd be interested in seeing are interest queries to major ISVs for Linux versions of desktop products, notably Quicken and tax preparation software, which are widely used and frequently replaced. This is likely a strong gauge for both commercial viability and actual size of the desktop space.
--
<A HREF="mailto:kmself@ix.netcom.com">Karsten M. Self</a mailto>
<A HREF="http://twiki.iwethey.org/">TWikIWeThey</a iwethey.org>, an experiment in collective intelligence. Stupidity. Whatever.
We are offering a monthly prize to any user who
freshly registers their Linux installation via
the Internet - you can *register now* (clickable)
or go to http://www.registerlinuxandwin.com/ (made
up address) after installation.
You really need something fairly unique (such as
an MD5 hash of the LAN MAC addr - not too
clever if you change your LAN card though) to ID a
machine and you maybe could register more than 1
Linux distro per machine for multi-boot
machines (only one prize per machine though).
You'd need to give your e-mail address too so
that the prize could be claimed.
I don't think people would bother registering if
they didn't stand a chance of winning a prize -
the prize could be a copy of the boxed set of
each of the commercial distros who get involved
in the scheme. Anyone think this is a good idea ?
I think you'd get a lot more reg's than Linux
Counter does if the distros got involved and there
was a prize winnable for registering.
Counting Linux Users
Posted by: Anonymous Coward on September 19, 2002 06:28 PMThe Colonel
#