Selected user comments on: The Last Dinosaur and the Tarpits of Doom

Marginal responses are by the author,

If you go to and do a power search for German references to Linux from Feb 15 1999 to Feb 15 1999, (last full day), and the same for Windows, you'll see that Linux turns in more hits. Linux 14,000 Windows 13,000 obviously there is some rounding.

If you do the same search but use a wider date span, e.g., Feb 9 to Feb 16 then Windows wins. That sounds to me like Linux, "mind-share," "buzz," "hype," or whatever you want to call it just passed Windows in Germany in one of the last several days.

Ross Nesbitt
Tue, 16 Feb 1999

Interesting? I'm not sure just what this means -- might be a blip -- but I doubt it is bad news for for Linux.

I think this makes you the first person to imply that my Last Dinosaur essay was too pessimistic. grin

There are a few points about your story I can see holes in:
1991 - 1 user Linus
1992 - when I first started using linux, Linus estimated 40,000 users
1997 - 1998, according to some surveys I saw over the last few weeks, linux usage rose 212% in this period. Most of this was server installations of which linux now has a 17% share.

Also you can't extrapolate linux usage as a curve while ignoring the curves of all the other OS's. Windoze (which I loath) has had a steady usage of between 70% and 95% for several years... There is no evidence to say this 'curve' will drop to 0 in just a few years..

Anyway, nice article, I'm just not as optimistic as you... But it doesn't matter to me on a very personal level, linux is here for me to use and I'll use it until something better comes along.. MS is irrelevant, Windows is irrelevant.. :)

John O'Sullivan Piercom Ltd.
RedHat 5.2 - The Ultimate NT Service Pack

It is true that the next growth phase of Linux will involve qualitatively different problems than the last ones: The desktop market is not the server market.

What is perhaps not sufficiently appreciated, however, is that this has been true for every doubling period Linux has experienced -- and it has come up with appropriate solutions every time. For twenty-four consecutive doublings!


Because Linux has enormously powerful forces underlying it which have powered it through all these previous barriers, and are odds-on likely to continue to do so.

I didn't want to plumb this deep in the original essay, which was intended to be fairly light, but I believe that one of the lowest-level motors powering Linux is the classic economic maxim that for economic efficiency, the equilibrium price of a good must be set equal to the marginal cost of production.

In the software world, that means that software has to be made available for the cost of copying, which is fractions of a cent. To do otherwise is to fail to reap maximum returns from society's sunk capital investment in producing that software, which costs the economy all the additional productivity lost due to society not using the full potential of the software.

Market distortions are of course a fact of life, but when goods wind up being priced thousands to millions of times higher than the economically optimal price, enormous pent-up energy is created, available to be tapped by anyone or anything offering a path around the problem.

I believe we are now seeing an economy rooted in the production of pig iron slowly, slowly beginning to adapt to the new economic realities. Companies who find a way of leveraging reality (as Netscape did by distributing free over the net) ultimately displace those who attempt to deny it (as do those companies attempting to use copy protection to make software act like pig iron).

Linux has become a major social mechanism for capturing the economic potential of cheap software replication, and steadily larger segments of society are fighting for a share of the wealth this creates.

That isn't going to stop until everyone has fed their fill!

To answer in another vein, consider Moore's Law: Transistor density doubles every 18 months or so.

Critics have been correctly pointing out for years that the next round of density doubling will involve solving qualitatively new problems.

This is absolutely true! The micro-history of Moore's Law is a fascinating trail of qualitatively different challenges overcome by brilliant, unforseeable engineering solutions.

None of which has had any perceptible impact whatever on the steady exponential progress of microelectronics!

In Asimov's Foundation, Harry Seldon's psychohistory science allowed him to quantitatively predict mass human behavior even while individual behavior remained unpredictable, somewhat as statistical mechanics allows us to predict the bulk behavior of gasses even while the individual behavior of molecules remains unpredictable.

Moore's Law seems to me to be the closest approach we have yet to a working psychohistory, successfully predicting a facet of human mass behavior year after year, decade after decade, even while the brilliant individual engineering breakthroughs underlying Moore's Law remain completely unpredictable.

Sometimes, looking too closely at the trees really does keep you from seeing the forest, and tree experts can prove the least able to perceive the larger pattern.

Well said. thanks for your work! 

If you have the time, consider adding a GNU-style disclaimer that will
allow people to redistribute your work (of course, it's your choice!).

        -Bob Duke

Good point. Done!


I'm responding to your essay at As a fellow Linux fan, I agree with the sentiment expressed. But I think there's a key flaw in your essay which needs correcting.

I dislike articles that make unrealistic predictions about the anticipated inroads of Free Software, because they set users up to be disappointed. I'm worried that Linux is grabbing popular attention far out of proportion to its significance _at this time_, for the same reason.

The problem with extrapolating exponential trends is that exponential growth only occurs until a niche has been almost filled; then it levels off rapidly. (Think: bacteria in a petri dish.) So the most accurate way to forecast the growth of an exponential trend is to identify its niche, or niches.

One of Linux' niches is the server realm, where I have no doubt it will soon dominate. Anyone competent enough to be administering a server of any importance can also recognize the high quality of Linux, and is probably clueful enough to recognize the advantages of truly open software. In addition, such a sysadmin will most likely know UNIX to begin with. Thus, in this niche, Linux can spread as fast as sysadmins can install it, learn about the idiosyncrasies of Linux, and educate fellow administrators about the differences between Linux and other Unices.

Linux has another niche: student programmers. Such programmers desire a high-quality development environment, but can't afford proprietary Unix workstations. Again, here the competence level is high, the motivation is high, and Linux can spread as fast as the users can install it. Linux already dominates this niche.

There are other groups of highly competent computer users for whom Linux is appropriate; they are already - or soon will be - all using Linux, thanks to the infrastructure built by the student programmers and increasing numbers of professional developers proficient in Linux.

The desktop-using masses, however, are essentially computer-illiterate; in this niche, computer knowledge is spread by word of mouth. Because virtually none of these users know anything about Linux, there is no computer lore for them to draw upon. I believe that in this niche, the largest of all, Linux is much closer to the beginning of its growth curve than your article suggests.

You may post this response if you like.


Jeremy Radlow

Sensible comments! People have been saying similar things for years, but Linux keeps hurdling the next barrier. Sort of like the sensible yearly warnings about the end of Moore's Law scaling in microprocessors. Let's compare notes in 2002.

I just read "The Last Dinosaur and the Tarpits of Doom" and the hairs on the back of my neck are still standing up: it is brilliant writing - congratulations.

- Lewis

PS I have an Apollo DN3500 in the attic of my parents' house and when I visit them I turn it on for a few hours for old times' sake. If their Window manager were available for Linux I'd switch immediately.

Always cool to see the past live on! My personal computer museum starts in the CP/M era. :-/


I somehow doubt that the growth of linux is a purely exponential curve. I suspect that it follows more of a tanh around the origin shifted up 1 unit along the y axis kinda curve. (ie: it will go exponential until 40-50%, and then more linear, and then more logarithmic, until it sticks at around 70%-80%)

Just a thought, since real life models are rarely a simple function.

Scott Bruce

Absolutely! No real-life exponential can last indefinitely. If the current human population growth curve lasted about 5-8000 years, the entire observable universe would be nothing but human flesh.

And, obviously, Linux can't double up to 160% and then 320% market share. :)

So in the end, you always have the classical sigma curve. But I didn't want to clutter the essay with Greek words and scare people.

I just read your "Last Dinosaur" essay. I enjoyed it very much. I just have one comment - I think your web hits might be off. Searching for Windows on altavista could have a significant number of hits unrelated to MS Windows (X, window washing, ...). Searching for Microsoft yields 8,308,061 hits & searching for Linux now yields 2,013,421 hits. Incidentally, searching for '(MS or Microsoft) and Windows' yields 1,273,297 hits.

Harvey J. Stein
BFM Financial Research

Good point! I thought of that, but when making "radical" sounding arguments, it is always best to use conservative numbers, so I let it stand.

To make a little contrarian argument, Linux is currently in the middle of a honeymoon period with the press that probably overstates its mindshare a trifle.

We'll get a press backlash against Linux shortly, after it stops being newsworthy to support Linux, and then a quieter bit, and then the next wave of pro-Linux publicity.

One must always have sympathy for the columnists struggling desperately to keep the ads from meeting in the middle of the page in the trade press. :) :) :)


One comment about your last dinosaur rant in you web page ( regarding the start of Unix. You write:
In 1975 Bell Labs released Unix.

Unix had no support from its creator, AT&T: Buy the magtape and don't call us. (AT&T was legally barred from entering the operating system market.)

Unix had no support from any existing vendor: None had the slightest interest in backing, supporting or developing an alternative to its proprietary operating systems offerings.

Unix had zero customer base: Nobody had ever heard of it, nobody was requesting it.

Unix had zero marketing: Nobody had any reason to spend money building mindshare for it.

A one-sided competition?

Decidedly: Unix wiped all workstation competition off the map in less than fifteen years.

I would say that in compressing history you are ignoring some other reasons why Unix caught on other than just getting a tape from Bell Labs. I see some of the important points as:
  1. Printing of the K&R C manual and Lyon's commentary on the V6 kernel source. Without the documentation, particularly documentation that was accessible, and accurate (well mostly, K&R-1 had some nasty dusty corners), I don't think it would have ever gone anywhere.

  2. One of the Bell Labs pioneers (I think it was Brian Kernighan) went to Berkeley for a sabbatical right after the initial unix started, and proceeded to preach the Unix gospel. This act drew in more people, and like we've seen in the Linux world, it goes out in waves, and the people Brian preached, preached to others in turn.

  3. At the time, CS departments were starting to think about doing hands on OS classes, but colleges are always short of funds. With the advent of the PDP-11 (and VAX later) which was relatively cheap and by getting relatively free source (though it was source that they could only share with other licensees), they could teach the classes. With the CS departments teaching UNIX, it brought a round of people who when they graduated were loyal to Unix (and VAXes for that matter) in their later professional jobs. Microsoft/Intel have certainly realized the potential of this, and have been seeding the campuses in the last few years (during which the workstation vendors were mostly cutting back such funding). There are some however that maintain that UNIX has actually killed the college OS courses, since the courses tend to be tweaks of it, rather than creating new OSs from scratch.

  4. Sun microsystems is also important in the early days of UNIX, in that they effectively legitimized UNIX as a commercial entity. If they hadn't decided to use UNIX and went with a propritary OS, things would have been different, and UNIX would have stayed just a hacker's toy.

  5. The ARPAnet (for those colleges/companies rich enough to join) and USENET provided for cross communication between different hackers (and of course for the flames). It helped that in the early days of USENET, AT&T (and later DEC) were furitively funding the long distance transmission of mail/news through UUCP links. At one point, I remember a quote to the effect that DECWRL's monthly phone bill amounted to $100,000 and the system administrators were having a hard time justifying it to upper management. This is the forerunner to todays Internet where for most people, the cost is a fixed cost, rather than a rate based cost.
Of course there are some negative things as well:

  1. The UNIX camp has never been one to rally around one standard. While the players have changed over time, you always have had at least two camps (originally PWB vs. V7, then System V vs. BSD, then Sun/At&t vs. OSF, etc.). Especially when you add in a good dose of NIH on each of the camps.

  2. Backwards compatibily has never been a strong suit of UNIXen. This is a very worrisome problem for the Linux arena. Look at the required upgrade list for the 2.2 kernel (or for that matter, the 2.0 kernel, and the 1.2 kernel before that). I've been through 2 changes of library/compiler infrastructure as well (a.out -> ELF, and now libc5 -> glibc2).

  3. In the late 80's, you started to see the commercial companies (Sun, MIPS, DEC, etc.) edging out the hackers for control of the UNIX base. This resulted in less sharing. Even in organizations designed to share the code base such as the System V group or OSF, it was usually an uphill battle to get the companies to actually work together (I worked for OSF for 5 years, so I saw the tail end of things from the inside).

Michael Meissner, Cygnus Solutions (Massachusetts office)
4th floor, 955 Massachusetts Avenue, Cambridge, MA 02139, USA

You're of course absolutely right that behind the smooth Unix exponential I painted there's a fascinating microhistory of battles, people and events!

I read the Sun story oppositely to you, and in fact considered making a parallel in the essay between Sun (startup powered by Unix) and RedHat (startup powered by Linux). But I didn't want to obscure the main point with too many subpoints.

In my reading, it was not Sun which legitimized Unix, but Unix which gave Sun its competitive edge: There were lots of companies peddling proprietary solutions, and years ahead of Sun in doing so, but what Sun could offer customers that was new was a comprehensive open-standards soution top to bottom, from commodity Motorola CPUs to Unix to TP/IP and NFS.

Sun, as a desperate startup, was willing to do this when the established vendors weren't, customers wanted it, and the rest was history: Sun wound up virtually owning the workstation market.

An IEEE study some years back concluded that every paradigm shift in the computing industry required a new set of companies, since the established vendors each time just defended the existing status quo. DEC was the canonical new company establishing the minicomputer paradigm, and to me Sun was the canonical new company establishing the Unix/open-standards workstation paradigm, and RedHat is making a strong bid to be the canonical new company establishing the Linux/open-source paradigm.

As far as multiple Unix standards: Your point is good, and we may be seeing something similar with (say) OpenLinux vs RedHat Linux.

But there's a phenomenon in physics percolation theory in which, when the average crack length exceeds a certain limit, the average cluster size suddenly goes to infinity: Instead of a lot of little isolated spaces, you have one continuous space that reaches everywhere.

I personally suspect that the Internet has done something like that to the computing world: Connectivity has gotten so good that separate computing camps are fundamentally just not stable any more: The cost of isolation is just too high. I take the unilateral moves by DEC and BSDI to start supporting Linux binaries as evidence of that: Being incompatible with the critical mass of Linux applications is simply no longer acceptable.

But we can of course still have KDE vs GNOME religious wars. :)

I liked your rant on the demise of MS, I am mostly in agreement with your analysis.

I especially agree with the non-obvious conclusion that the process is 80% complete.

The interesting thing about exponential growth is how long it takes before anything noticeable happens. Like the old story of the guy who invented chess and the emperor says I'll give you anything you like. So he asks for a grain of rice on the first square and two grains on the second, four on the third....(64 squares on chessboard) The emperor won't notice anything amiss until he's gone through 1/4 of the board. The 2nd thing to realise is that the emperor isn't going to just let this bastard take all the rice in his empire, he'll execute him for his insolence (or if he's smart just tell the guy he'll comply but the inventor has to count out the rice himself). OSS has just reached about square 12. The people who dismiss it as an irrelevent fad are akin to courtiers saying, "So, he's getting a bucket of rice for each square now, so what, rice is cheap. What a moron, he was offered anything he liked, he could have asked for a golden necklace..."

I think there may be a dip in Linux adoption. Read "Inside the tornado" for an interesting analysis on high tech paradigm shifts. Initial fast growth occurs with early adaptors, usually followed by a dip in rate of growth as one reaches saturation in early adaptors market. This can be overcome by "bowling alley" strategy of selling to tightly defined groups of pragmatists with specific needs (Linux has already started making those advances, and the groups for which it is a pragmatic solution are growing fast).

However, whether it takes 2 years or 5 years, I see it as inevitable that Linux will destroy MS.

An interesting phenomena that you missed which I believe will accelerate the process will be pragmatic ISVs taking advantage of current state of market by releasing Linux software early. Some of this will be "GPL on GPL'ed OSs, all rights reserved on proprietary OSs" (I'm looking to do this myself). This will enable software companies to receive most of benefit of OSS (testing, users providing bug fixes and even enhancements) while retaining ability to make money from people who insist on staying with WinXX. One of the main advantages that MS has currently is that new software tends to be released first on WinXX and it is only later that Linux version is released or OSS equivalent is created. The first company to release a Linux version of a particular product may well be the gorilla in that market sector on Linux platform. Since developers generally prefer Linux anyway, and they know that large profits only make them a target for MS, the migration to Linux by ISVs will occur much faster than industry analysts expect.

All the best,

You make the intended scaling point better than I did. :)

The random fluctuations around the basic exponential are of course virtually impossible to predict. The usual parade of experts have been announcing that Moore's Law scaling has to slow down right about now, but in fact the doubling time has actually been shortening a bit of late. Go figure.


Great prediction essay on Linux! It warms the heart to realize just how close we are :) Anyway, I thought I'd mention that you may have made a typo on the "AOL" / "ICQ" thing. As far as I am aware, AOL has a proprietary protocol for "AIM", not "ICQ" (which is owned by Mirabilis).

Duane Johnson
Calgary, Alberta

Thanks, Duane!

Sounds like you might have missed Mirabilis being purchased by AOL for about $258M, a few months ago. Speaking as someone who was working for, and had stock options in, a direct competitor to Mirabilis (, selling Ding!, a technically much sweeter and more open alternative to ICQ), I remember the moment vividly. :)


I just read and was fascinated by your article "The Last Dinosaur and the Tarpits of Doom: How Linux Smashed Windows". I'm just a Treasury financial economist with an interest in how new ideas become accepted, and also an interest in the evolution of the computer+software industry. Anyway, I enjoyed your article and I was just curious to know in general terms a little more about your background. I guess I can't help being an 'easterner'.

Whitney Culbertson

Not sure what sort of info you're looking for, but you could start at

I'm best known in some circles for inventing the Citadel family of computer BBS programs.

I'm best known in some other circles for the Skandha[1-4] series of anatomical visualization programs powering the Digital Anatomist, which directly inspired the better-known Visible Human Project . (The contract was written with us in mind, but Colorado landed the grant and in our opinion butchered the cadaver, wrecking the dataset for most scientific purposes. Oh well.)

I'm best known in some other circles for running QwestMUCK for a few years, and for the aborning follow-on project Muq.

My parents are Canadian (as am I) and my grandparents were Brit, excepting for Granny, a full-blooded Scot (Stewart clan).

My wife and I are at present tremendously pleased over our new house.

That do it?

I'm gonna hafta question your math. You equate market share (a percentage) with users.

Your mathematics fail to represent that the userbase is growing. Doubling Linux's number of users over the course of a year does not equate to doubling its market share, since the market will be larger.

I'm sure there are several desktop market estimates floating around on the web. You could easily use one of them to "fix" your math.

--Mark Storer

You're formally correct, of course, but it is a second-order effect that doesn't affect the results much: With Linux doubling three times a year, even a 60%/year increase in market size would be lost in the noise. If Linux has to do one additional doubling per 18 months, that's less than one part in four: 2003 becomes 2004, say.

Not to spoil an otherwise entertaining rant, but I have to take exception to the following paragraph in your "Last Dinosaur" document:
"Another cheap prediction directly from history: Just as the death knell of minicomputer proprietary operating systems was a lawsuit requiring that Federal minicomputer purchases specify an open standard such as POSIX rather than a single-vendor sweetheart solution such as VAX/VMS, the final death knell of microcomputer proprietary operating systems will be a successful lawsuit requiring that Federal microcomputer purchases specify an open solution such as POSIX rather than a single-vendor sweetheart solution such as Windows."

I don't think so: Windows NT is already POSIX-conformant as far as the US government is concerned, having been officially tested and validated for conformance to FIPS 151-2, the Federal POSIX standard; thus in the event of your hypothetical lawsuit all the Federal agencies would have to do in order to comply would be to install NT as the agency-wide desktop operating system, which many of them are doing (or planning to do) already and which I'm sure Microsoft would be glad to have them do even sooner.

Federal agencies have already awarded major contracts for "POSIX desktop workstations" that were won by PC vendors and NT in competition with Unix vendors such as Sun; one of the major reasons Microsoft did a POSIX implementation in the first place (back in the early 1990's) was to allow them to bid on such contracts. Now of course probably few if any government agencies ever _use_ the POSIX capabilities of NT, but they've "checked the box" in terms of "open operating system standards" as far as NIST and the Federal government are concerned.

For Microsoft statements regarding Windows NT's POSIX conformance, see the document

If you can't access it, here's a relevant quote: "Windows NT version 3.1 Workstation and Windows NT Server have been tested using the official NIST PCTS [POSIX Conformance Test Suite] for Federal Information Processing Standard (FIPS) 151-2 and NIST has validated the test results. Windows NT version 3.5 is in the process of being verified for POSIX.1 compliance and will also be submitted to NIST for FIPS 151-2 certification. FIPS 151-2 incorporates POSIX.1 as a reference standard and also requires a number of the optional features defined in POSIX.1 to promote application portability among conforming implementations."

See also the relevant NIST list of validated POSIX implementations:

and search for "Microsoft".

So to sum up, the chances of a POSIX-related lawsuit like you suggest are virtually nil. A more relevant legal possibility would be a US government directive to promote the use of open-source software (like Linux) within Federal government agencies. I have no idea what the chances of this are though; I'd say relatively slim at this time.

Frank Hecker Pre-sales support, Netscape government sales

Time will tell!

It is differences of opinion which make horse races interesting.

I don't think anyone is under any illusions that Microsoft's supposed NT "POSIX compliance" means anything: POSIX was designed to provide rational guidelines for companies trying in good faith to achieve compatibility, leaving it to the market to sort out blatant cheats like NT's deliberately useless POSIX box.

IMHO, nobody has bothered contesting this particular bit of silliness because there hasn't been much point; Windows and POSIX systems haven't been competing directly, by and large.

Let POSIX get >10% of the desktop market, and some billion-dollar companies like IBM will start seeing massive profit in beating Microsoft over the head with serious POSIX compliance. Those companies will in turn start making the required "campaign contributions" and so forth to make it happen, and suddenly you'll start seeing meaningful action on the POSIX compliance front.

For better or worse, large-scale government isn't about logic, it is about adjudicating conflicts without gunfire. It works as long as government decisions favor the strongest power groupings. (If they don't, those groupings win the civil war and set up a new government whiche does favor them.)

Microsoft's monopoly benefits Microsoft and hurts everyone else, which means that it is terribly, terribly outgunned in these contests: No points for guessing the eventual result, no matter what individual maverick judges or legislators may do.

So despite current appearances to the contrary, I think there is a quite good chance of the above prediction coming true.

But I wasn't actually trying to make a strong case for it, I was really just presenting it as an example of a simple-minded prediction which can be read off directly from the parallel between minicomputer and microcomputer history, without doing any hard thinking. Lots of such predictions will of course be utterly silly.

Dear Cynbe,

You have explain very well in depth about how Microsoft will fail over Linux, and surprisingly I begin to believe that will happen, :)

I think I should get busy to learn Linux now...

Benjamin Woen

Actually, the math here undercuts the other very good observations. The mathematically correct way to extrapolate is not to expect to see continued doubling, per se, but rather to understand that the total number at any time is the integration under the technology adoption curve. It is the same technique (though admittedly, it is probably less precise) used to predict how much oil is left in a given reservoir by looking at the rate of change in production over time. _

If you integrate over a bell curve ./ \.

You get an "S-curve":

(very bad ascii approximation -- however -- probably no less accurate than the numbers we are dealing with here.)

The bell curve represents the rate of adoption over time. The S-curve is the number of adoptees over time.

This is the same thing for adoption of air travel or VCRs or linux.

The fact that total linux adoption is doubling every X months, merely implies that we are somewhere in the first *half* of the cycle. Since the sample data are very low quality, only when the rate of adoption *clearly* begins to decrease, will we know that we have crossed the half-way point.

-- Steven Talcott Smith

You're right, of course!

But I didn't want to dilute the essay with charting and footnoting, and I don't think your more accurate analysis would change the results.

Linux is already at 2.5% of desktop market, and as you point out, not showing any signs yet of entering the second half of the bellcurve.

Linux only really needs to make it to 10% of the market to become effectively unstoppable, which is to say, two more doubling times. (Linux can't be undersold, can't be sued, can't be effectively FUDded -- if it gets to the critical 10% of market where it is being supported on a par with other OSes by third party software vendors, the game is over.)

To change the prognostication significantly, we'd have to see a huge increase in doubling time immediately, which would be nearly miraculous: It would require a virtually sawtooth "bell curve".

Anything is possible of course: Microsoft might get saved by the Second Coming in 2000!


Re your article "The Last Dinosaur and the Tarpits of Doom".

I guess we should know how things are panning out by the end of '99.

I recently saw an example of the sort of thing you are commenting on in the British Science Museum. Last century when electricity was coming in, gas companies were busy defending their 'patch' against the newcomer, trying to match electricity's versatility.

One of the last-stand inventions - a gas-powered room fan. That's right, if your room was too hot you lit this gas-powered device (creating more heat!) to power a fan blade via a small heat-engine.

I hope you can see some similarities to the Windows/Linux situation....

Jamie Cherryh


No, I hadn't neard that particular story, but the status quo never lacks for brilliant engineering talent making a game try to preserve the obsoleted paradigm.

Here's another example: After bronze blades doomed flint weapons, some master of the flint paradigm actually managed to build a wood-core, flintflake-edged competitor to the newfangled swords.

It was without a doubt a brilliant masterwork, one of the crowning achievements of neolithic technology.

And, of course, it made no difference whatever: Bronze still won.

Last gasp brilliance actually seems as much the rule as the exception: The status quo paradigm has the resources and trained talent to produce them, and the approach of the new paradigm provides the incentive.

The last clipper ships before steam were quite something also!

Microsoft today has a market worth of $102,000,000,000 and has hired some of the most brilliant established software talent on the planet, legends like Alvy Ray Smith. We may see some very pretty last-gasp stuff come out as the Last Dinosaur's nostrils disappear beneath the tar...

Reading your essay, I notice a peculiar issue that you don't appear to address: while it is relatively easy to capture, say, 2.5% of a market like this one, it is not easy to take real share away from a behemoth. In fact, it seems to me that the only reason why Apple has been able to double its share in the last 5 months or so is that -> people already know about Apple & and like the Mac. <- On the other hand, everybody is peeved at Microsoft now, or nearly everybody, and maybe Linux will be the perfect excuse for them to jump ship. On the third hand, unless and until there are hooks to permit people to run MS software under Linux, it does cost a hell of a lot of $ to jump ship, and that will also deter a great many people.

If you want to argue that it takes a behemoth to take market share away from a behemoth, reflect that companies like IBM, DEC and Sun are lining up behind Linux, and aren't exactly pipsqueaks themselves!

But I honestly think that megacorp backing of Linux is more symptom than cause of its success, past present and future.

One of the details I suppressed in the essay is that progress is rolling right along on WINE, the Linux utility that lets you run Windows executables directly on Linux. One of the consequences of Linux' astounding 212% growth last year in the server market is that corporations are starting to kick in cash funding of WINE development, which should give it a big kick in the pants. The developers currently estimate that they have about 95% of the Windows API calls implemented to some extent, and that they are about 20% of the way to a production release. Being able to grab Windows apps out of the box and run them on Linux will definitely be one of the crucial micro-details underlying the superficially smooth exponential Linux ramp.

I am, right now, attempting to build computers for three people who cannot afford to buy them, and equally cannot afford to buy software for them. Linux would be a fine solution, if there was a solid suite of the usual apps out there & debugged, and if these people were not already thoroughly used to Windows. (I may try to get at least one of them to go with Linux in any case, most particularly because it _is_ essentially free; she's _really_ dirt broke, as far as I'm aware.)

-- Jon Singer

You might want to check out SAL, Scientific Applications on Linux, which despite the name indexes hundreds (thousands?) of general-interest Linux apps ranging from spreadsheets to word processors -- as well as lots of cool scientific apps.

Good story. But linux will never take the desktop from microsoft UNTIL a solid gui standard across the platform is developed. KDE and GNOME are getting there but one of them has to win.

Probably right! At the moment GNOME looks the odds-on favorite to win, backed by RedHat and Debian &tc. KDE's sudden change of source copyright terms looks more than a little desperate.

Also the makers of hardware such as winmodems, printers, video cards, USB devices must throw away their NDA's and open up their hardware programing spec's so that linux can use these things! Until linux runs on ALL the hardware that windows does, it can't steal the desktop. And guess who is keeping the vital secrets locked up. Somewhere in Redmond.......

-- Kenneth Scharf

I have to disagree there: Microsoft is a nonentity in hardware design and sales.

To me, one of the most telling wins for Linux in the last year has precisely been seeing the hardware vendors starting to jump on the Linux bandwagon.

3Dfx wouldn't give Linux the time of day a couple of years ago: Today is it working closely with Linux driver writers, and by midyear it is promising to release native Linux drivers with the same level of support as its Windows drivers.

If you look around, you'll see similar stories everywhere: Linux has reached the critical mass at which it becomes worthwhile for hardware vendors to support it. They have no OS allegiance, they just want to move product out the door, and if customers are running Linux, so be it.

What I find it even more interesting that Intel has rammed through open disclosure of I2O, the intelligent I/O standard which is going to make it possible for hardware people to write a single driver that works with all OSes.

That will put Linux exactly on par with Windows on the driver front, even if hardware manufacturers don't lift a finger for Linux.

What is particularly telling is that Intel opened up the I2O specs midstream in a sudden course correction clearly designed specifically to help Linux, and did so despite Microsoft having a seat in the organization.

If the Last Dinosaur had the clout to shut Linux out of the hardware driver game, that was the exact time and place to appy it.

It didn't happen, and from where I stand, that front looks like clear sailing for Linux from here on out. This is one of the many underlying details I left out the the original essay to avoid clutter!

I thought that the reason for some hardware, such as winmodems not working with linux was MS making the vendors sign NDA's to get info on un-documented windows hooks to build such devices and then forcing them to make others do the same.

Oh! That sounds like Microsoft silliness, yes -- making life harder for everyone else in the hope of making more money itself.

But that's the Microsoft software API being locked up, not the hardware API -- the same hardware vendors are free to write Linux drivers for the same cards, and will be scrambling to do so as soon as Linux hits about 10% of the desktop market. Right?

Microsoft's problem remains that its business model only lets it win by making everyone else lose: Ethically if not legally, Microsoft is engaged in theft.

This means that of the Fortune 500, 499 of them see Microsoft as a problem to be solved -- and collectively, they can squish Microsoft like a mosquito, it just takes time for a disorganized group like that to agree on how and when.

Linux is turning out to be a very handy Mosquito-swatter which they can agree upon because it is vendor neutral -- it doesn't favor any of the Fortune 499 particularly.

The FSF's contribution could be lumped in with the other 'microevents' making up the exponential curve, in analogy to the disparate technical advances which have been holding Moore's law true. Still, the FSF's code base more or less instantly transformed Linux from an hobbyist 386 OS into a usable system, and that happened very early in its history. The instant Linux could run gcc, all of the GNU utilities and other programs were available.


I've actually made the point myself on occasion that "Linux" is at least as much FSF/GNU code as anything else. There are a number of reasonable kernels floating around, for example, but no realistic alternative to gcc, which was a significantly harder project. (Speaking as someone who wrote both something fairly similar to gcc, and also a 386 protected mode dos extender not far from being an OS kernel.)

Stallman's "Lignux" coinage is logically justified, but sets some sort of record for ugliness: I can't bring myself to use it.


I read your editorial on Linux overtaking Windows with great interest. However, in your editorial you state "at the accepted industry rate of $100/line" for a single line of source code. Where is this number documented as to its validity ? Is there a web site which I could reference where such is documented ?


Saif Warsi

I wish I had a good primary source for this number -- if you find one, let me know! For a typical reference in passing to it as an accepted industry number, see Reuse Metrics Calculator:

New Development Cost (Cost/LOC)

The historical cost to develop new software in your organization, in dollars per line of code. We recommend obtaining this value from your contracts or financial planning group; otherwise use the industry average of $100/LOC as a default value. Uses dollars ($) for units.

You make a compelling argument. However, it strikes me that the market for Windows is rather different than the ealier market for VMS, CP/M in that it is dominated by people with no computer expertise. So the triumph of Linux will require that it be factory installed, be at least as easy to use as Windows, and have a substantial suite of application programs. I suppose that will happen, but it may take a few years.

John Prothero

Yes, but all that is happening.

For example, Compaq is the top PC company on the planet, and will start shipping Linux factory installed next month.

The GNOME project is focussing on building a Linux desktop as easy to use as Windows or the Mac -- and now has commercial funding from companies dependent on the Linux market.

The WINE effort is working on allowing standard Windows applications to run on Linux, and also has commercial funding now from companies dependent on the Linux market.

Debian Linux now ships with 1800 programs standard, and both SAL and the Linux Applications and Utilities Page list thousands of additional Linux applications.

Well stated observations on an overall trend.

A refinement for Alta Vista hits (, any language) is Macintosh internet mind share ranking is more closely represented by "Macintosh" than "MacOS".

*       Windows     12,011,841
*       Microsoft    9,599,621
*       Macintosh    3,344,038    <--- Macintosh SYSTEM
*       Linux        2,219,764
*       Unix         1,967,390
*       Solaris        764,842
*       Amiga          561,532
*       VMS            480,770
*       MacOS          230,110    <--- "MacOS" LOGO
*       HP-UX          102,418
*       Ultrix          34,620
*       OpenBSD         26,900

The "MacOS" logo arrived late on the scene and has gone away to just a "Mac" LOGO ( ).

The "Macintosh" operating system shipped with the original Macintosh computer and the "Macintosh" operating system still ships today with a Macintosh computer.

A detail, but an order of magnitude... ???

Again, nice paper.

-- Marc Campbell

Very good! Your numbers are more consistent with other numbers indicating that Linux desktop share is currently third behind Apple (2nd) and Windows. "MacOS" probably undercounts, as you say, but "Macintosh" probably overcounts, since it is a hardware brand also. Hard to get good numbers without doing a lot of work.

Another nice follow-up is to compare results by language:


Microsoft 20434 
Linux      3443
Macintosh  3242
FreeBSD    1524


Microsoft 92394 
Macintosh 89823
Linux     35096
FreeBSD   19083 


Microsoft 36932
Macintosh 12384 
Linux      3461
FreeBSD    2746


Microsoft 49215
Linux     23211
Macintosh 14353 
FreeBSD     351


Microsoft 58287 
Macintosh 27010
Linux     17530
FreeBSD     926

There are clearly some minor mysteries in those numbers!

Why is Microsoft so utterly dominant in Chinese? The low numbers vs large Chinese population lead me to suspect that the population isn't online, and we're primarily seeing paid advertisements, possibly.

Why is Macintosh head-and-head with Microsoft in Japan? I'd guess it is partly because NEC has until recently filled the Microsoft niche in Japan, and partly because the Mac presumably handles the triple Japanese writing system better.

Why is Linux way ahead of Macintosh in Spanish? Are Spanish- speaking countries more price-sensitive?

Why is FreeBSD neck-and-neck with Linux in Russian but back in the dust in French? I have no clue. Perhaps one volunteer did a Russian translation of the FreeBSD docs, but nobody did a French one.

Another follow-up would be to search by date:

         MS      Mac    Linux
1994       95     26      23
1995      464    650     134
1996     2609   2221    2208
1997    19486   4112    3705
1998    31998  16764    9889
1999     1847    432     393

Unfortunately, these Alta Vista query results seem to vary by as much as a factor of two to the same query given at different times, which does not inspire confidence. The results for early years may come more from misdated modern documents than actual early references? The parallel between the results for 1994 and those for 1999 suggest this to me, at least.


I read your article with much interest. I am in agreement with half it and in disagreement with the other half.

I do not agree with your application of doubling times to OS market share. Trends which have been exponential in the past need not continue to do so. That Linux has thus far grown exponentially does not imply that it will do so in the future. Perhaps Open Source is a paradigm appropriate for producing software for only certain tasks. Then Linux would grow exponentially until fills a market niche, and would then stop. I am not saying that this is the case. But there is no a priori reason to assume that continued exponential growth is guaranteed on the basis of past performance.

I agree wholeheartedly with your analysis of open standards, however. I am in complete agreement that Windows is doomed for a wide variety of reasons. Microsoft's money, PR power, marketing power, establishment, etc, are all totally irrelevant. People who focus on those things and claim they assure Microsoft's continued dominance are ignoring fundamentals while focusing on superficialities.

Microsoft is doomed, doomed, doomed.

Do inform me if you write anything else.

Tom Werges

"The race goes not always to the swift, nor the battle to the strong. But that's the way to bet it."

The past is the only guide we have to the future. Barring a clear drop-off in the Linux growth curve, the conservative assumption is that it will continue.

From:"RHB" Fir Oct 29 1999

Talk about momentum! Since your article, Altavista "hits" for Windows went to 12.25 million from 2+ million, Linux went to 5.7 million from half a million or so, and from zero, "OS/2 Warp" now shows 1.5 million pages. So windows growth rate was 500%, Linux was 1000%, and OS/2's, well, 1.5 million divided by zero is, what? A whole bunch.

"The race goes not always to the swift, nor the battle to the strong. But that's the way to bet it."

The past is the only guide we have to the future. Barring a clear drop-off in the Linux growth curve, the conservative assumption is that it will continue.

Here is an article suggesting Microsoft is actually fountaining red ink:
Microsoft Financial Fraud.

An internal Microsoft memo:
Jim Allchin, whose job these days is to pilot Windows 2000 into harbour, was a key Microsoft figure in the Caldera case email chains, and was seriously worried about Novell. On 9 September 1991, he emailed Bill Gates: "We must slow down Novell. ... As you said Bill, it has to be dramatic ... We need to slaughter Novell before they get stronger." Allchin was not generous enough of spirit to express any admiration for Novell publicly, but in an internal email on 26 March 1993 he produced a ringing endorsement: "I still don't think we take them as serious as is required of us to win. This isn't IBM.

"These guys are really good; they have an installed base; they have a channel; they have marketing power; they have good products. and they want our position. They want to control the APIs, middleware, and as many desktops as they can in addition to the server market they already own. We need to start thinking about Novell as the competitor to fight against - not in one area of our business, but all of them. If you want to get serious about stopping Novell, we need to start understanding this is war-nothing less. That's how Novell views it. We better wake up and get serious about them or they will eventually find a way to hurt us badly."

Another internal Microsoft memo:
Phil Barrett, who was managing Windows 3.1 in 1990, received an email from a subordinate about DR-DOS 5.0, which said: "You asked me for a user's view of DR DOS 5.0... I used DR DOS 5.0 with a huge number of apps. I found it incredibly superior to MS DOS 3.31 and IBM DOS 4.01. ... The most important reason to use any version of DOS is to run DOS apps. DR DOS 5.0 runs every DOS app I know. DR DOS 5.0 works successfully with Windows (2.11, Win 386 2.11 and Windows 3.0 and 3.0a). ... Conclusion: DR DOS is vastly superior to MS dos 5.0."

Microsoft's fondness for consumer choice on display:
At one time Microsoft proposed that DRI (Digital Research, which then owned DR-DOS) stop marketing DR-DOS and the companies cross-license each others' products.

Microsoft's fondness for promoting free consumer choice again on display. Also its occasional willingness to at least consider telling the truth:
David Cole and Phil Barrett exchanged emails on 30 September 1991:
"It's pretty clear we need to make sure Windows 3.1 only runs on top of MS DOS or an OEM version of it. [...] The approach we will take is to detect dr 6 and refuse to load. The error message should be something like 'Invalid device driver interface.'"

Microsoft had several methods of detecting and sabotaging the use of DR-DOS with Windows, one incorporated into "Bambi", the code name that Microsoft used for its disk cache utility (SMARTDRV) that detected DR-DOS and refused to load it for Windows 3.1. The AARD code trickery is well-known, but Caldera is now pursuing four other deliberate incompatibilities. One of them was a version check in XMS in the Windows 3.1 setup program which produced the message: "The XMS driver you have installed is not compatible with Windows. You must remove it before setup can successfully install Windows." Of course there was no reason for this.

Brad Silverberg, the Microsoft exec who finally left the company last week, but who in an earlier life had been responsible for Windows 95, emailed Allchin on 27 September 1991: "after IBM announces support for dr-dos at comdex, it's a small step for them to also announce they will be selling netware lite, maybe sometime soon thereafter. but count on it. We don't know precisely what ibm is going to announce. my best hunch is that they will offer dr-dos as the preferred solution for 286, os 2 2.0 for 386. they will also probably continue to offer msdos at $165 (drdos for $99). drdos has problems running windows today, and I assume will have more problems in the future." Allchin replied: "You should make sure it has problems in the future. :-)", which is clear enough, and it should be noted that the pair were both high level Microsoft executives.

"Janine has brought up some good questions on how we handle the error messages that the users will get if they aren't using MS-DOS. The beta testers will ask questions. How should the techs respond: Ignorance, the truth, other? This will no doubt raise a stir on Compuserve. We should either be proactive and post something up there now, or have a response already constructed so we can flash it up there as soon as the issue arises so we can nip it in the bud before we have a typical CIS snow-ball mutiny."

Cole replied: "Let's plead ignorance for a while. We need to figure out our overall strategy for this. I'm surprised people aren't flaming yet, maybe they won't." Cole also sent an email to Silverberg suggesting a less severe message be used when DR DOS was detected: "A kind-gentle message in setup would probably not offend anyone and probably won't get the press up in arms, but I don't think it serves much of a warning. BillP made an excellent point, what is the guy supposed to do? With a TSR, the solution is to just remove it. With DR-DOS, or any others, I doubt the user is in a position of changing. He will no doubt continue to install. When he finds problems, he will call PSS. We will get a lot of calls from DR-DOS users.

"Perhaps a message in the phone system for Windows. It would say something like 'if you are not using MS-DOS or an OEM version of MS-DOS, then press ##'. Then give them the message." Silverberg replied: "What the guy is supposed to do is feel uncomfortable, and when he has bugs, suspect that the problem is dr-dos and then go out to buy ms-dos. or decide to not take the risk for the other machines he has to buy for in the office."

Posted 24/05/2000 12:20pm by Sean Fleming on

Linux is more popular than sex

Our thanks go out to Register reader Gunnar Isaksson who wrote in to inform us of the "unscientifical investigation" he has carried out into the growing popularity of Linux.

The unstoppable OS machine has been gunning for Windows for some time now and despite the small - but perfectly formed - market share Linux lays claim to it looks like it has reached an important milestone. Or maybe not.

Isaksson's research was simple - he ran the word Linux through the AltaVista search engine and then he did the same with the word Windows. The results were, erm, well they were as follows.

There were 11,313,520 Web pages found for the Linux search but only 10,755,265 found for Windows.

So there you have it, inconclusive proof that Linux is now more prevalent/important/hyped* than Microsoft Windows.

But worse is yet to come from Isaksson's research. When he put the word sex into the AltaVista search engine, only 9,589,620 were found.

So, could it be that not only is Linux becoming more popular than Windows and a real threat to Microsoft, but it is also becoming more popular than sex and that the human race might be facing extinction unless we abandon the open source OS?

Probably not, no. .

From: Ross Nesbitt Subject: linux takes mindshare lead To: Date: Mon, 03 Jun 2002 23:25:27 -0700 (PDT) search for windows Results 1 - 10 of about 46,300,000. Search took 0.06 seconds. search for linux Results 1 - 10 of about 50,900,000. Search took 0.05 seconds.

Wow -- That's definitely a milestone! I've appended your email to the lastdinos comments page (stripping your email address to save you from spambots) -- I hope you don't mind.

With the German government just officially announcing a switch from Windows to Linux and 2/3 of Microsoft's current customers refusing to accept the latest round of price increases, things are pretty clearly on course.

The convergence of a number of different productivity suite solutions for Linux (from StarOffice, OpenOffice and KDEOffice to Win4Lin and similar hacks for running Microsoft Windows on Linux) with the evolving critical mass of customers like the German and Chinese governments willing to pay for such solutions means, I think, that we're headed pretty steadily towards the point where Linux is 10% of the desktop market.

10% desktop marketshare is pretty much "Game Over" for Microsoft: That's the point at which all the third party software developers will find it profitable to port their stuff to Linux (any CEO is happy to show a 10% increase in sales to the board of directors and stockholders) which in turn means that all the consumers who have been (correctly) worried that Linux is Not Quite Ready For Prime Time from an application point of view will stop sitting on the fence and start cashing in on the savings to be had by getting out from under the Microsoft tax, leading to an unstoppable positive feedback effect. (This is hardly a prediction: It is an observation of what has already happened in a number of Linux markets, from servers to CGI graphics at the likes of Pixar and Industrial Light and Magic.)

Comparing "windows" to "linux" has to be an underestimate of real mindshare, of course, since "windows" has lots of non-Microsoft and in fact non-computing uses, while "linux" really does not. So out of curiosity I tried a measure of this: Asking Google for

windows -microsoft
produces roughly three million hits. So over 90% of the mentions of 'windows' on the WWWeb in fact refer to MS-Windows, seems like. That's pretty amazing, when you think about it. You'd think, regular old house windows being clearly more common than PCs, that they'd make a better showing. Are we all just a little computer obsessed on the web? :)

Anyhow, it suggests that the error due to this effect in the mindshare numbers is negligible in practice: We're looking at a roughly 50:45 (which is to say 10:9) mindshare lead for Linux over Windows -- about a 10% lead.

I should set up a little Perl script to poll these numbers and graph them, I suppose. Wouldn't be hard to do. Maybe this weekend. (I've just been getting into Perl seriously for the first time over the last six months.)

Thanks for forwarding this! I wonder if slashdot would be interested -- seems like it deserves wider circulation.

Life is Good!

-- Cynbe