Softpanorama

May the source be with you, but remember the KISS principle ;-)
Home Switchboard Unix Administration Red Hat TCP/IP Networks Neoliberalism Toxic Managers
(slightly skeptical) Educational society promoting "Back to basics" movement against IT overcomplexity and  bastardization of classic Unix

Nikolai Bezroukov. Portraits of Open Source Pioneers

For readers with high sensitivity to grammar errors access to this page is not recommended :-)


Linus ceased to be a volunteer

Very soon after starting work on a Linux kernel, Linus got a work at the university (as a research assistant or something like that) and until early 1997 (as he admitted himself) he enjoyed almost complete freedom and was essentially paid for working on Linux kernel. In his 1994 to interview to IT4 he stated:

"Q: What do you consider a normal day (e.g. what takes your time, normally: school, Linux, spare time)

A:  Linux takes up my "working hours" - even just reading email takes a minimum of two hours a day, and to actually react to that email and do development fills up the rest of the day quite nicely indeed, thank you very much.. However, I do take time off for hobbies etc, and I can essentially do linux at my work at the university (they know I do Linux development, and they allow for that fact)."

That means that actually the volunteer period of working on kernel was limited to approximately the first two years. After that Linus should be considered to be a paid developer of the informal university-funded project (the essence of his situation was that he did not need to support himself with any other job and as such can not longer be considered a volunteer in a strict sense of this word). And after his move to the USA  he actually became one of the most highly paid Unix developers.

There was instant market for Linux as a replacement of pretty expensive SCO Xenix that was widely used on PCs. That's why Linus started to get some flow of small gifts and checks even earlier that he got his university job.  On 386sx computers (without 386 floating point co-processor) Linux was an instant hit and the only game in town.

The Birth of Commercial Open Source:
Linux Distributors change the course
of fight with FreeBSD (1993-1994).
 

Linux has emerged as the winner
because it was most easily turned into the perceived winner.
In an industry where perception is more important than reality,
that edge made all the difference.
Nicholas Petreley

In 1992 Linus made a very shrewd political move (and a risky one in view of Richard Stallman's personality) by sidelining FSF. Creating a complete Linux system by compiling the kernel, utilities, initialization routines, and other software, is a non-trivial task and is best left to competent programmers.  Before Linux it was assumed that FSF became a copyright holder and a principal distributor of the OS based on Linux kernel. But Linus Torvalds blessed independent commercial  distributors (essentially breaking RMS monopoly and creating "commercial open source")  and boldly refused to transfer copyright to FSF.  That move preserved his central position, but put RMS into tremendous political disadvantage of  "supplier" of Linux utilities, the role he resented so much but was unable to change.  Here is one historically important letter about the change where he explicitly stated that changing one of the motivations of changing license to GPL was to allow commercial distribution:

Newsgroups: comp.os.linux
From: [email protected] (Linus Torvalds)
Subject: Linux commercial use (was Re: BBS Project -- Linux?)
Message-ID: <[email protected]>
Organization: University of Helsinki
References: <[email protected]>
Date: Sat, 13 Mar 1993 22:10:58 GMT
Lines: 43

This is only a partial follow-up to the "BBS Project.." article - it's
mainly a response to all the other posts about linux commerciality that
I've seen (not that I read them all: the only thing on c.o.l I don't
read are the non-technical flame-fests).  Mild flames follow. 

While I can understand that people prefer free software (both in the "no
cost" and in the GPL sense) over commercial programs (I personally
certainly do), actually flaming somebody over trying to make some money
is silly: it does not limit you in any way, and only gives you more
options to choose from.  If you don't want to pay for your software,
fine, but there are actually people out there that find it convenient to
be able to buy programs that work and are supported. 

In the BBS case, it wasn't even a question of somebody "taking advantage
of the GPL to make a quick buck" as somebody so diplomatically put it in
some unrelated post, but of a program that was actually developed by
people wanting to make some money on it.  While I think GPL is a great
idea, I also acknowledge that I might actually have to work for a living
when out of university, and selling programs you write is one way.  Yes,
I know that you can make money on GPL'd software (hey, I've made a very
modest amount myself on linux, thank you), but I think almost everybody
agrees that it's easier to do it the "normal" way. 

As to people wondering about somebody actually selling linux and other
GPL'd products for "big" sums - rest assured that everybody who has
written a major piece of software and put it under the GPL has been
aware of the "problem".  In fact, linux originally was under a much
stricter copyright that allowed no selling at all: the reason it is now
under GPL is that I *wanted* to allow the kind of thing that SLS,
Yggdrasil etc are doing.  And while the other linux kernel developers
have been "forced" to use the GPL due to my decision, I haven't actually
heard any complaints from anybody who has worked on major parts of the
kernel. 

So please, don't bitch about commercial uses just because they are
commercial: find something better to complain about.  If it has some
technical failings, feel free to point them out just as you would any
program, but money per se is not a reason for flames.  It's ok to be a
bit more lenient on programs you can get for free ("hey, what do you
expect for $0"), but that does not mean that you should use unreasonably
higher standards just because somebody makes money off a program. 

			Linus

The first Linux distribution was probably SLS that started in mid-1992. The second and very important distribution was Yggdrasil. Initially each distribution was a one-man shop but they have huge influence of the development of Linux because they have huge stakes in Linux success.

With the emergence of  The Cathedral And The Bazaar fairy tale that became a manifesto for the "commercial open source movement" (Raymondsm) RMS also lost his role as a "philosopher-in-chief" to  Eric "surprised by wealth" Raymond.

All-in-all this single decision paid handsomely both politically-- preserving Linus position of a project leader as well as financially (due to commercial distributors Linus later became a multimillionaire and beneficiary of  several sleazy Linux IPOs).

Multiple Linus distributors, such as Yggdrasil, Slackware, Red Hat Software, and others as well as voluntarily organizations such as Debian started to appear to simplify packaging of Linux kernel and other software for the average Linux user. The cornerstone for such distributors were usually one (like is the case in Slackware and Red Hat) or a couple of highly qualified developers.  Often they later became the founder, or more often co-founders of their own company. The result of compiling and packaging a complete Linux system is known as a distribution. And due to the size and complexity this pretty large collection of various software packages can be sold on CD as a commercial product. I think emergence of CD readers and writers for PC in early 90th was one of the most important factors of Linux success: it made possible the commercial success of Linux distributors.

The support of commercial distributors was crucial,  as approximately in late 1994 496 CPUs became dominant AT&T lawsuit was settled and Linux lost advantage of being able to run of cheaper hardware and now needed to complete with BSD "face to face". At this moment BSD still enjoyed much higher technical quality  and that's why it was the operating system of choice for ISPs. Linux distributors essentially became a trench solders who fought and won this fight for Linux acceptance. IMHO in final outcome of the fight between FreeBSD and Linux was decided not by Linus and his team of developers, but by Linux distributors as a foot solders of this battle.

In November 1992  Yggdrasil released the first CD-ROM distribution and for the reasons, outlined above, I think this was  a turning point for the Linux: conversion of a previously traditional project into something new:  "commercial open source".  Due to availability of CD ROM distributions Linux user base became the fastest growing segment of Unix community as CD ROM distributions permitted more or less straightforward installation. Linux has an important edge over BSD in that it served larger segment of users -- it was much more PC friendly and required less expensive hardware (386sx with 4M was minimum). Other distributors emerged very fast -- it was like throwing a match on dry leaves. Soon Linux CD ROM distributions became widely popular, financially successful to the extent that each of them were able to employ several talented developers. As I mentioned before, the key year for Linux distributors seems to be 1994 when they became a small, but important industry with more than a dozen players. This particular year they managed to sell around 50,000 CDs a month (~$30 a CD) and also got some positive press. Selling Linux CDs was a small but profitable business. Volunteer part of the development became intermixed with hired guns in the distributors stuff.  That permitted the creation of a pretty complete Unix-like system attractive as a desktop for PC-oriented Unix enthusiasts. One can add FTP downloads to that. In 1992 alone users interested in desktop Unix were downloading about 8 terabyte of Linux distributions per month from various WWW and FTP sites (most Linux distributors maintained an FTP sites.). Caldera that was founded by Ray Noorda (of Novell fame) gave Linux some corporate credibility with its version 1.2. This was the first version that has a nice installer and can automatically configure X Windows on a typical PC.

With commercial firms hiring top Linux talent, the role of Linus became more of a manager and a coordinator of kernel development and a PR figure. Everything that can be outsourced was probably outsourced in this period. From this point he probably start to understood less and less about several less important areas of Linux kernel, but that's ok as long as he preserved the control over the kernel as a whole.  At this time several Linux developers became partners in new firms (Red Hat is a famous example).

Although technically weaker than BSD, the existence of multiple Linux distributions each of which adapted Linux to a particular market and caters to their particular category of PC users,  provided an important advantage for Linux over BSD. Linux managed quickly pass FreeBSD in the number of CD sold and the number of downloads. It's interesting to note that this happened despite the fact that at this time FreeBSD still was of higher quality (especially in networking). That once again demonstrate that Bill Gates was right and that whose who commands desktop controls the market. Linus had chosen the right strategy --  on the initial stages of OS entrenchment server business does not decide everything, desktop does, and popularity of the desktop is in its turn the Trojan Horse that ensures server success too.

Some Unix enthusiasts at this time resented that fact that Linux is more popular that BSD. They correctly pointed out that BSD has a genuine Unix ancestry that makes it automatically more mature than Johnny-come-lately Linux: well known Linux's weaknesses included its virtual memory system and its scheduler. It was true, but they forgot that Linux kernel caters to the most popular hardware platform that that alone overweight many other considerations. Before 1998 generally there was a lot of hard evidence that BSD was more reliable and/or performed better than Linux, and there were surprisingly few claims to the contrary from the Linux camp. But did not matter much because due to its PC desktop orientation (and GPL licensing) Linux manage to get a political support as the major independent Unix. 

GNU license was also very probably an important political factor at the initial stages of Linux adoption (much more than in XXI century, when it was partially discredited by sleazy Linux IPOs and then bust of techno bubble, as well as Red Hat perversions ;-). Paradoxically GPL promotes such "capitalist" feelings as envy and "code stealing" paranoia very effectively.

One important side effect of commercialization was the emergence of Linux press and it was an important ingredient of Linux success:  more and more businesses invest or even start making money selling it and they naturally tend to advocate Linux in order to win your business trying to forge connection with the commercial press. Moreover "sexual appeal" (new system with a cute mascot and an interesting story being it -- some unknown Finnish student blah-blah-blah..) make Linux an interesting story by its own right.

Specialized commercial Linux press emerged in 1994 when Robert Young and the ACC Corporation published the first two issues of the Linux Journal. Phil Hughes became an editor. In March 1994 SSC -- the publishers of Unix Pocket References became a publishers of the Linux Journal and produce an excellent publication, although interestingly enough it was not available for free. Several book publishers like O'Reilly felt that Linux books could be profitable and rushed to the new market.

We can call this idea of commercialization supposedly free code complexity-based commercialization. Linus probably intuitively understood that at some point C-code became close to binary code due to excessive complexity and all you need is to withheld architecture and algorithms.

Early Marketing Efforts

If you can bear to hear the truth you've spoken
  Twisted by knaves to make a trap for fools.

Kipling,IF

It is important to understand that what Linux created, technically speaking, was not an operating system. It is simply the traditional-style monolithic kernel of the Unix-compatible operating system. But this is key component of OS: like any kernel Linux handles system processes, hardware access, and many other critical tasks. that means that the operating system we call "Linux" can encompasses much more than the kernel and  includes GNU GCC compiler and utilities, some BSD software, the X Window System (under X license) for windowing and graphics, and several key applications. Among them Netscape Communicator, TeX, Star Office, Word Perfect,  or other text processor, some databases like MySQL and Postgress, WEB server (usually Apache), Sendmail (or hotfix), bind etc. And from the user standpoint the results can be quite different. So there was enough commercial differentiation to support several companies, while preserving to a certain extent a single kernel (later Red Hat forked the kernel, to fend off competing Linux distributions).

In 1991-1994 Linux kernel has an important edge for the low segment of server and desktop market -- it can work without floating point co-processor and that mean it was usable on the cheapest 386sx motherboards and on PCs with just 4 Meg of memory.   And that's was exactly space where most action was and where commercial Linux distributors can flourish. Therefore an early and significant marketing edge was Linux 386sx orientation.

Yes, technically the kernel was primitive and far from being polished (in some sources you can actually see how people learn C language :-) while Free/OpenBSD kernels can be called a legitimate child of the original Unix written by brilliant programmers from AT&T and Berkeley. From the point of view of heritage Linux status was somewhat fuzzy, but who cares when we have such a great fairy tail about its origin  "A poor Finnish student, in his spare time, etc." That PR attractiveness of Linux simplified marketing for commercial distributors and they quickly started to work on utilizing this PR advantage, effectively creating "Linus cult" (see below), playing Linus as a David against Goliath of Microsoft and commercial Unixes. That was a nice fairy tail story that from the point of distributors effectively promoted Linux as a unique player, and as such deserved a lot of ink. Linux got all of it...

Paradoxically commercialization of Linux had shown that this anarchistic license can be played in   several non-trivial ways, Stallman be damned.  First of all, GPL provided the foundation of Linux techno cult, and paradoxically from the beginning it had some anti-RMS dimension. That was actually pretty logical as with the success the movement needed a more centrist leader and Linus was a perfect replacement of RMS in this particular role.

Also GPL represented an important leverage against FreeBSD/Nets/OpenBSD because of all this "free software" blah-blah blah" as well as GPL incompatibility with almost any other license in existence: GPL proved to be a one way street -- Linux developers can borrow whatever they wish from a BSD-licensed OS, but the reverse case is more difficult and generally requires rewriting in order to avoid the trap of GPL license.  Moreover Stallman's hobby of picking weak targets for enforcing GPL and generally his "GPL purity crusade" served as a pretty effective Linux marketing tool  ;-). Another implicit marketing campaign launched by Stallman was connected with his desire to rename Linux into GNU/Linux. All this noise was actually a pretty effective marketing.

The second thing was more subtle. Amateur developers that are working outside research centers are usually unable to pick up with the pace of technical development in such area as OS kernel -- they have neither the necessary qualification or experience to deal with such level of complexity. But Linus, still being an amateur, managed to survive by restricting his role to "grand arbiter" for other developers and the chief configuration management specialist.  He retains control, as unless he blessed a particular change it would never get to the production kernel.

At the same time GPL provided a possibility for highly paid and highly trained corporate developers to use their talents outside, often strangulating for a real talent, corporate atmosphere. Although questionable, this actually might be considered to be one of the few advantages of GPL over BSD license. It's somewhat safer from the point of view of preserving your employment in a large firm to contribute to GPL-based project then to BSD-based project: you can always claim that your contribution should became GPLed because it is a part of the larger GPLed whole, and that might get you off the hook in some cases.

All this fairy tales about free hackers in the best style of "surprised by wealth" Eric Raymond should not conceal the fact that the major contributor of open source projects were highly paid and highly training corporate and academic developers, with many of them understanding their particular subsystem much better than Linus. And it is only "configuration management" job of "herding cats" for commercial distributors that Linus performed extremely well. And as we will see was later royally remunerated for.

Drunken walk principle as a technical direction

Linux never has a technical direction of its own. Its always tried to follow somebody's taillights be it FreeBSD or Microsoft. Linux often has been called the poor man's Windows, but initially its approximation of windows desktop functionality was pretty raw. But gradually random project emerged and some prospered to the level to attract the attention of distributors. In turn the adoption by distributors represented a strong shoot in the arm for any such project. That suggests that "drunken walk principle" might be a viable technical strategy for similar reimplementation projects. Some also suggest that this is a viable model for emulation of any entrenched commercial product: 

Myddrin - Subject: Drunken Walk... ( May 23, 2001, 19:09:27 )
Right now, Linux on the desktop is following the same pattern as the drunken walk problem*. As time goes on and we get apps then we will see a more distinct trend but to look for it now when the desktop apps are still in their infancy (despite what the ver numbers say) is getting ahead of yourself.

Given time and the leaner economic times, I think Linux getting a nice double-digit (20-30%) desktop market share in inevitable, after that who knows. We just need to give the various desktop teams time to get the real work done.

---

*For those of you who haven't taken a course of study that covers this (such as biology, statistical quantum mechanics, probability and so forth), the drunken walk problem refers to a drunk walking down the street. S/he has a probability to keep going, a probability to step left and a probability to step right.  If one direction is more probable (even slightly), then given time the drunks average position will drift further and further from the 0.

The connection here is that there is a small probability that a given user will step right (switch to Linux), but the probability that they will then step left (switch back to windows) may be lower.

If this is right model, we will slowly gain a small percentage of users each time period until such time as other effects come into play (killer app, backwards compatibility, pick your favorite weakness).

I don't have the figures, but I would expect the Linux server market to follow the same pattern till about 1996 or 1997 when there was that 200% increase in market share that started the Linux hype machine.

---
Myddrin
GLIMS Project -- glims.sf.net
Panda Thumb -- ptmb.sf.net

The other aspect of  "drunken walk principle"  can be called the "Linux Uncertainty Principle" It can formulated as "The more precisely the time of next version of kernel is determined, the less features and the usability of Linux of the version can be predicted"

Linux Uncertainty Principle
 The more precisely the time of next version of kernel is determined, the less features and the usability of Linux of the version can be predicted

Reinforcements arrive: Professionals from DEC start to contribute
to Linux by doing Alpha port (1994-1995)


He, who recognize the need
and promptly donate something to help the person,
essentially doubles the size of his gift.
Roman proverb ?

It's forgotten and largely ignored by "official" Linux historians fact that one of the most important early steps in the  development of Linux was the DEC port to Alpha. This was the second influx of paid professionals into the development of Linux and probably as important as the first one because of the quality of DEC engineers. It immediately increase technical quality of Linux kernel in several major areas and first of all in the filesystem area.  Moreover it dramatically raised the technical level of  Torvalds himself, who got an opportunity to overcome his status of  an obscure provincial Finnish student trying to cook a Unix clone on  a pretty primitive hardware to a specialist with the intimate knowledge of two important hardware architectures. Way back in the beginning, Linus didn't think that Linux would ever run on anything other than the 386:

I'd say that porting is impossible. It's mostly in C, but most people wouldn't call what I write C. It uses every conceivable feature of the 386 I could find, as it was also a project to teach me about the 386.

That change came about when Jon 'maddog' Hall, then working at DEC, got an Alpha-based system delivered to Linus in 1994. Actually understanding a second hardware architecture is an extremely important for all OS designers.  That was also the first chance for Linus to see USA. Here is how Jon ``maddog'' Hall recollected this events in My Life and Free Software:

As part of DECUS, there was a special interest group (SIG) called UNISIG which looked into ``everything UNIX''. Kurt Reisler, the group's chairperson, had heard of a small operating system that had been started in Finland and was now being developed all over the world. He wanted to invite the architect of this operating system to DECUS in New Orleans in May of 1994. After watching Kurt send many, many letters to potential funders of the ticket, I decided this would be worthwhile for our customers and convinced my management to fund the ticket and hotel costs. It was in New Orleans that I first saw Linux and met Linus Torvalds.

After DECUS was over, I took Linus out on the Natchez, one of the riverboats that goes up and down the Mississippi River. We had dinner and discussed Linux. I asked Linus if he had ever considered porting Linux to a 64-bit processor and a RISC chip, to make sure it was portable. Linus told me he had considered that, but was having trouble getting an Alpha system from the Digital people in Helsinki. Knowing Digital's procedures, I could imagine the Helsinki people were trying, but having difficulty. I went back to my office, called a few friends, pulled in a few favors, and an Alpha system was on its way to Linus' apartment.

As I was getting the system for Linus, I became aware of how many engineers inside Digital were already using Linux, and that a group in Digital Semiconductor was trying to port Linux to the Alpha, but as a 32-bit port, not a 64-bit port as Linus was doing. With a little convincing, I got them to join forces with Linus.

In January of 1995, the project actually got underway, and as it unfolded, I found many people in the Linux community who wanted to help. Some people who already had access to an Alpha immediately started working on the code. Others actually bought their own Alpha to help with the project. In November of 1995, the first distribution of Alpha Linux came out. I was proud of the fact that the Linux community thought so much of Alpha, Digital and their own Linux system that they had all cooperated on this project.

Windows 95 changed the landscape of both hardware and software

Most people know that Windows 95 was a huge success for Microsoft, but few people understand that it was extremely beneficial for Linux.  One thing that Windows 95 instructed is new Pc specification that included "plug and play" as well as more memory (16 and 32M became common in 1995) and cards that support 32 bit drivers, larger harddrive and new devices (CD ROM became common with windows 95). It drove PC hardware to the new level where running Unix became not a problem at all. That tremendously benefited linux.   Moreover the desire replicate Windows functionality created kind of surrogate agenda for linux and helped to promote certain individuals (Gnome and Miguel De Icaza came to mind).

Last modified: March 12, 2019