 |
|
Rebel Code, by Glyn Moody, is an excellent recount of free software (both free as in free beer and free as in free speech), open source, and Linux. It is wonderfully balanced in an area where extreme positions seem de rigueur. It is sympathetic towards Linus Torvalds and Linux, Richard Stallman of Emacs and GNU fame, and several Microsoft executives -- a journalist probably should not alienate his sources. In the end, I came away with more respect for all the major figures.
The first half of the book is an easy read. It is principally about people. Richard Stallman, the Old Testament prophet of free software, gets a chapter. He comes across as the archetypal hacker: brilliant, driven, awesomely prolific, and without a life. For 12 years he slept in his office at MIT's AI lab. Linus started Linux in 1991 for the same reasons I use it now -- he wanted a UNIX that would run on a PC that he could afford. Four chapters chronicle Linux's initial development as an OS for standalone PCs running in text mode, its integration with the X Windows system, the addition of networking, the advent of commercial distributions, the porting to a second architecture, and the cross fertilization with BSD UNIX (Berkeley System Distribution). Eric Raymond gets a chapter for his contribution of some fine code and his insights into why Linux's development process succeeds. If Stallman is the Old Testament prophet, Eric Raymond is the New Testament's Paul, carrying the good news far and wide.
I bogged down at about the halfway point. The chapter entitled "Low-Down in the Valley" is where the flame wars start. The multitude of players, voices, and arguments overwhelmed me, the bazaar became too large and too noisy. If I had noticed the extensive index at this point, it would have helped. The volume and multitude of players continues in "Free the Lizard," about the opening of the Netscape source (code-named "Mozilla," a contraction of Mosaic, the first graphical Web browser and Godzilla).
The next several chapters chronicle open source's entry into the mainstream of software development -- Red Hat's Linux distribution tied with Windows NT 4.0 for Infoworld's Product of the Year for 1996. The following year, it won the award outright. This was the era of IPOs and a staggering influx of money. It also was a time of an escalating battle between the purists and the pragmatists -- most specifically GNOME versus KDE. Here Moody's even hand shows both sides presented in their best light without the smoke and heat that accompanies most discussions on this topic.
Having lived through the Mindcraft benchmark flame war, it is interesting to read Moody's take on it. I am clear that on my hardware, Linux runs faster and more reliably than Windows. What he makes clear is that Linux did not scale well on multiprocessor systems under heavy use. It is true that on the first benchmark, Mindcraft did not make much effort to level the playing field. However, even after leveling the playing field, Linux with the Apache Web server did not run as well as Windows on four processor systems. This was a real coming of age for Linux. If Microsoft had not considered Linux real competition, they would not have commissioned the benchmark. The multiprocessor tuning had been done with microbenchmarks. Not many hobbyists have four processor systems, but they are routine in enterprise-level systems. In three months, the performance problems were fixed and Red Hat announced that Dell's servers running Linux were "more than twice as powerful as Windows 2000 on the same four processor system." Linux was entering the big time.
The last two chapters cover the year 2000 and look into Linux's future. Much of the chapter, "Tomorrow's Hothouse" covers IBM moving from supporting to embracing Linux.
"On 10 January, IBM announced that it 'intends to make all of its server platforms Linux friendly, including S/390, AS/400, RS/6000 and Netfinity servers, and the work is already underway.' The release continued, 'IBM has released source code modifications that enable Linux to run S/390 servers. These modifications can be downloaded free of charge from the IBM developerWorks Web site. IBM is not providing service support or maintenance for this source code. However, IBM is currently conducting a joint customer study to gauge customer interest and support requirements for Linux running on S/390 servers.' Official IBM support arrived on 17 May 2000, when it was also announced that SuSE and TurboLinux would be offering GNU/Linux for the S/390 as a product."
Linux was now in the big time. Moody's insight into why IBM was embracing Linux is obvious in hindsight.
"The overarching logic of the decision not withstanding, the practical benefits for IBM are clear. Possessing, as it does, the widest range of disparate hardware systems -- from portable computers up to supercomputers -- the company more than any other has had to live with the consequences of incompatible software. By making GNU/Linux run on all its hardware, it immediately unites all its systems into a single family and offers its customers perfect scalability."
Oh. Yeah. Why didn't I see that. A recent article in our local paper (reprinted from the Washington Post) adds another reason. There are still plenty of mainframes out there, but the baby boomers that program and maintain them are retiring. The younger generation has little training or experience on IBM mainframes, and it is getting hard to find replacements. However, schools are turning out plenty of UNIX and Linux programmers and sysadmins. The scarcity of human talent is the resource that limits a company's ability to grow and prosper. Six months ago, my wife asked about my employer at the time who had both money and products, "What will happen if they can't keep their programmers?" For many companies, human talent is their major asset; the hardware is replaceable, the people are not. The rest of this chapter is about how GNU/Linux is getting an expanding part of the developers' mindshare and how other nonWindows OSs are being starved for talent.
Moody's opinions, observations, and speculations form a more obvious part of the last two chapters. These are the informed opinions of someone who has talked with some of the brightest minds in software development, a welcome break from the narrow, dogmatic rants I've recently seen in e-mail lists. Mature wisdom is refreshing.
The last chapter, "Beyond the Market," is speculation about the future -- the possibility and consequences of project forks in core software, Microsoft's attacks on Open Source, and who might replace Richard Stallman and Linus Torvalds. It doesn't really have an ending, it just stops -- but then, the open source and Linux revolutions aren't over yet either.
Glyn Moody is described as a UK-based journalist, however the spelling is American. Some parts are from articles he has written, but the seams aren't obvious. This is a fine book, but only time will tell if it will become the classic history of this era, although it certainly will be in the running.
-- Jeffrey Taylor (jeff.taylor@ieee.org)