The Long View on Linux, Part 2
N.B. What follows is part two of "The Long View on Linux", a conversation between Doc Searls and Phil Hughes. For part one, go here.
Doc: I remember when the Net first became obvious to business, right after Netscape released the next incarnation of Mosaic. Nobody knew what to do with it, because no company owned it. The thing was as free and open as air. Now, it seems like every business on earth has .com after their name, and the Net has utterly changed the context of their work. Are we seeing the same thing happening with Linux? It seems to me that the Net and Linux are two aspects of the same thing: a new world that works for three reasons. 1) nobody owns it, 2) everybody can use it, and 3) anybody can improve on it. This turns out to be great for business -- as long as the business doesn't try to throttle this public goose that lays golden eggs.
Phil: There is certainly a similar evolution. That is, Linux was ignored by many for a long time, even though it really was viable. A few years ago, SCO did a FAX SPAM about how you could get a $50 trade-in on your Linux distribution to "upgrade" to SCO UNIX. Well, anyone using Linux at the time was aware that Linux was a better product than SCO UNIX. It's pretty obvious this effort didn't work, as we can tell by SCO's market share.
The one thing that really has surprised me is there aren't any companies that decided to seriously bundle their product with Linux. Six years ago, I was talking to a company that had cross-development software. I suggested they make a Linux port of their product, and then just sell a complete package -- Linux itself, plus their product on a CD. It is unlikely that someone is going to use the same computer for office tasks and software development, so it wouldn't matter if the OS on this system was different from everything else.
I talked to the engineers and the marketing people about this, and they were all excited. I thought it was a sure thing, but it never happened.
Back to your points: yeah, Linux is a tool that can make money for a lot of different people, because it is open and free. Many companies have played along very well to help the goose continue to produce the eggs. The one company I believe got off track for a while was Red Hat. They seemed to want to make their name mean Linux. It wasn't that they did not continue to give back to the Linux community; the problem was their marketing wasn't showing Red Hat consumers that there was a lot more to the Linux community than Red Hat. Sorta like Microsoft trying to make people think that they were the only game in town.
At this point, I think the situation has corrected itself. You see books on SuSE, Caldera, Debian and other flavors of Linux in the stores. That means the distributions can compete on how well they fit the needs of each consumer, rather than on a visibility issue.
Is this change for the better? If you believe Linux should grow in market share, yes. We used to be a bunch of geeks doing what we want. Today, Linux is business, and to keep it growing, we need to keep getting more business people on board. It is just that along the way, there will be various frustrations. The most common is when a vendor doesn't understand the point of open, and finds a way to get around licensing such as the GPL.
Doc: What I see here is a deep conflict between very different ways of conceiving the Net (where Linux plays a fundamental role) as a context for business. Geeks conceive the Net and open-source OSes like Linux as environmental. They're growing spaces that nobody owns, everybody can use and anybody can improve. Business guys, including "new economy" get-rich-quick dot-com companies, see the Net as territory to dominate, as a medium for entertainment and advertising, as a new retail distribution system, or as something else borrowed from the familiar world of industry. They don't get that Jerry Yang and Jeff Bezos didn't make the Net. They only took advantage of what the geeks made for everybody.
It seems to me that, the way the geeks see it, the Net and everything that makes it good is basically a big new world, built like a UNIX system. It's a brand-new environment that still needs to be built out, but in a way that should work for everybody and not just those who, in Walt Whitman's words, are "demented with the mania of owning things." This is why the two groups are at such odds over the patent issue. Geeks know in their bones that the whole idea of software or business methods being ownable is simply ludicrous.
The GPL is hard to talk about, because it is meant to foster the virtues that produced this paradise. It so happens that it succeeded by discouraging ownership and encouraging sharing. These are rather anti-industrial business concepts, but they're not anti-business. Every smart business since the dawn of civilization has understood that what makes a market a market is the open public spaces that surround every business.
Phil: No argument here. I feel that Microsoft tried to fill in all those spaces, which worked for a while, but all that is changing.
Doc: I've found it very hard to talk rationally about Microsoft inside the Linux world. Too many distinctions get collapsed. For example, the literal distinctions between open and closed source get collapsed with moral distinctions between right and wrong, and with qualitative distinctions between good and bad software. Eric Raymond is fond of saying he just likes software that doesn't suck. But a lot of the world likes closed-source software for which there are no alternatives, whether that software sucks or not. And most Microsoft users, on the whole, don't think Office sucks.
Phil: I think two things happened here. First, for many people, the first word processor they used was Microsoft Word. They learned it, and see anything different as something new they have to learn. So, whether it's sucky or not, they know how to use it. When I was learning vi, I had been using ned, a screen-oriented editor on UNIX. ned wasn't available for the new computer SSC had bought. I hated vi and loved ned. What happened, however, was in about two weeks, while I still hated vi, I realized I was editing faster than had ever been possible with ned.
The second is the charm of the GUI. Microsoft had the "in" that allowed it to produce the best GUI-based editor, because they had the inside track on the GUI. For the casual user -- someone who might write a letter a day -- Word was adequate, and easier to remember how to use. Even I can use Word; it's just that I can get more work done in vi.
The person I started SSC with, Irene Pasternack, left to "get a life" that included raising a family -- something you don't want to do when you are starting a company. Irene has consulted at Microsoft for close to 15 years doing tech writing. A year or so after she went to work for Microsoft, her boss told her that she did eight times as much work as his other employees and asked if she had any idea why. She said, "they use Microsoft Word, I use vi."
Doc: I'm not defending Microsoft here. I'm just trying to get some perspective on the popularity of its goods and the worldview those goods tend to teach -- quite aside from the company's obviously nasty hardball business practices. It's important to get perspective for several reasons, one of which is what I believe is why Microsoft has peaked and will soon be re-positioned as a specialty supplier. It's that Linux and the Net are rapidly forcing the software business to grow up.
The result, I think, will be something much more like the construction business. Construction is a trillion-dollar-plus business, and it has no Microsoft. Almost nobody outside the industry can name the leading construction companies or their biggest suppliers. That's because the real leaders are the architects, builders, contractors, subcontractors and volunteers who do the real work, just like we see in open-source software.
One reason construction comes to mind is that we make liberal use of construction and real-estate language in the software business. We architect, design, develop and build with software. On the Web, we build sites. Those sites have locations. I don't think this is just coincidental; I think these metaphors come easily because the practices are similar. Not identical, but similar enough to give us a model for the way the software business really ought to be. It is interesting to note that the construction business is about as open source as a business can get, and nobody is any the worse for it. I mean, there are no secrets to making nails or two-by-fours.
A friend of mine used to program for a familiar dot-com company that runs on Windows NT. He said he constantly had to deal with memory leaks in that OS, and added, "we're putting up 60-story buildings and we don't know if there's rebar in the concrete." Then he added, "Actually, we know there isn't rebar in the concrete, and we're going ahead anyway, knowing we just have to fix problems constantly."
Open-source software doesn't put geeks out of business; it just gives them more room to do their business. And gradually, Linux is proving that. Which is why I think that in five or ten years, the software and construction businesses will look very similar, with many more builders, architects and designers. Plus many more suppliers. The main difference will be that no one pre-fab supplier will control the business any more.
Phil: The UNIX community grew up as computer capabilities grew up. The 30-year history of UNIX means there were people making those buildings without foundations, because there wasn't any room for the foundation. But they recognized the problem, and when room became available, they added as much foundation as would fit. Thus, the best possible building was produced at each step along the way, and we built a work force of people who understood how to grow with industry changes.
On the other hand, the MS community knew the foundation was missing and there was nothing they could do about it, so they realized the whole building was disposable. Of course, being disposable means you can sell the new product next year. I think I am beginning to realize that the mess being created was not a plot -- just circumstance.
Doc: You've probably watched the climbing Linux adoption curve closer than anybody. Why is Linux finding universal adoption where other free UNIXes (such as the BSDs) did not?
Phil: As much as Richard Stallman and others would like to credit this adoption with the difference between the GPL and the other software licenses, I just don't think it is the case. Linux got out there, and it worked. While the various BSD camps were fighting over which was the one true BSD, people were using Linux and seeing that it was a solution.
Once Linux had the inertia, and Linux Journal certainly helped build that inertia, Linux would continue to roll along. I also have to give a lot of credit to Linus' management skills. Getting hundreds, possibly thousands, of programmers around the world to all cooperate on the development was an amazing effort. Getting them all to do it for free was way beyond amazing.
Even before I met Linus, I was impressed by his maturity and sobriety. The guy just seemed to have this highly informed and calm perspective, like a judge. Not your average 20-something dude; not your average geek, either. Made me wonder what would have happened, say, if Bill Joy had stayed with Berkeley UNIX. Instead, Bill is doing fun stuff, but it's all stuff that's owned by Sun. Java. Gini. Solaris. Sun is a good company, but in their own way, they're as closed as Microsoft.
Linus has introduced a new model to the computer industry. Or, maybe I should say that Richard Stallman proposed a new model, and Linus was the right man to successfully implement it. In any case, it has worked, and that may help convince companies such as Sun to at least move in the direction of more openness.
Doc: When did you meet Linus?
Phil: I first met Linus in 1994. It was at a Linux-related party in suburban Washington, DC. When I got there, Linus was there along with a few others. They were talking about some technical Linux-related issues. What I saw was a person who could participate in the conversation, rather than act as the boss. At this point, I understood why Linux was moving forward so rapidly.
Doc: Linux has spread like a wildfire, and Linux Journal has grown right along with it. How have both Linux and Linux Journal changed since both came out at v.1.0? And are you happy with all those changes?
Phil: The short answer is that today we are both better products. But, more importantly, our focuses have changed. Linux went from a geek OS to a major contender in the business and commercial market. As that was happening, LJ went from being a developer magazine to offering a lot more that would appeal to the commercial user.
Over the years, we have received letters from readers who credit LJ with convincing their bosses that Linux was real. At times, I felt LJ was portraying a more professional image than Linux itself; at other times, I felt it was the other way around.
I can't say I am happy with all the changes, but I also realize they needed to happen. While it used to be more fun to play with Linux than to publish a magazine, I also realize that for Linux to move forward, it is necessary for it to become more commercial. We would not be seeing drivers for all the new hardware if we didn't have the kind of growth we have.
Doc: In the last year, Linux has become a very hot topic. It seems to me this was for two reasons. One was the popularity of the OS itself. The other was the popularity of new Linux companies in the stock market. These are two very separate concerns that were covered by the mainstream press as if they were one -- at least while Linux stocks were hot. Now we're seeing the topics start to separate a bit, now that dot-coms of all kinds (including Linux stocks) are dropping from the sky like a plague of frogs. Linux stocks were especially hard hit. It hardly seems fair.
Phil: The growth in the popularity of the OS has been amazing. Here, I think the US Department of Justice and Microsoft both get some credit. The DOJ anti-trust suit brought the concept of an operating system into the minds of much of the public. Prior to the DOJ action, most people only understood "computer" and "program" together, where "program" was commonly spelled Microsoft Word. They bought a computer, and added Microsoft Word or Microsoft Office in order to make it do things. That's all very different today.
The Microsoft credit has to do with NT/Windows 2000. A lot of companies were betting on NT or Windows 2000 offering the server side of their computing. The harder people look at these products now, the more they seem to realize that there are alternatives such as Linux, and quite bluntly, Linux has proven to be a better choice.
Windows 2000 is finally out there, but people are sick of waiting. If Windows 2000 works, fine. But if not, they are likely to pick an alternative, and Linux seems to be at the top of that list.
Doc: Here at Linux Journal we use Debian, which is the largest noncommercial distribution. Are we doing it for agnostic reasons (not wanting to favor any particular advertiser), because it's better, or both?
Phil: Note that we used to use Slackware. I like the idea that Debian seemed to be the best fit for us, because we can then be agnostic, but we chose it mostly because it was a better solution. Debian has always had a good dependency system, plus it was designed so you could upgrade without a reboot. While Debian development might fall behind the commercial distributions, I am not a fan of upgrades without a reason, so it has tended to keep up with our needs.
Doc: What do you run personally?
Phil: On my personal systems, I run a whole bunch of different distributions for various reasons. For example, I have SuSE on one laptop and Caldera on another.
Doc: Richard Stallman tells me he'd like to see a clearer differentiation between GNU Linux and the commercial forms of Linux. Can this happen? Should it? Is there that much of a difference in kind between the various distros?
Phil: Linux, the kernel and hundreds of utility programs are virtually the same in all the distributions. Sure, SuSE will release a new distribution this week with a newer kernel than Red Hat. But next week, Caldera will be ahead, and so on. But they are mostly the same.
What they add to all this GPLed software is a combination of additional programs, such as an evaluation copy of some commercial software, and an installation method. All the distributions I have seen differentiate between GPLed software and commercial software, so I don't see this as a serious issue. As for the installation methods, most distributions have GPLed this code as well.
Doc: Let's talk about Linux on the desktop. When I first saw Linux running an Office suite on a GUI desktop, well over a year ago, I expected great things to happen fast. Yet today, there isn't a single non-proprietary desktop application suite to compete with Microsoft Office, and I don't see any on the horizon. Enterprise-customer types tell me Linux won't make it on the desktop until it has a truly competitive (and compatible) office suite. What will break the logjam here?
Phil: StarOffice, while not open source, is free and is forcing the issue of compatibility. Sun, the owner of StarOffice, needs a generic and compatible office suite in order to restrict Microsoft's ability to rewrite the standard every year or two. If StarOffice is successful in preventing the standards re-write, then I expect other packages will appear. The good news is that StarOffice is available for MS Windows, and as I understand it, E-machines is shipping it on their boxes. This should help market penetration.
Doc: If you see Sun as the only player with the clout to really compete with Microsoft in office suites, what do you make of Applix's and Corel's chances? And do you think it's possible that some group of open-source developers, KDE for example, will come up with a practical open-source office suite?
Phil: Corel doesn't have a complete suite yet, and I don't think they will until after the issue has been forced anyway. Applix has had an office suite out there for a long time, but I don't think they have the clout needed. Sun wins on image. Lots of people know that Sun systems are all over the Internet, and some people even know that Microsoft's hotmail.com site is running on Sun systems. Maybe it's BSD. In any case, Scott McNealy is out there picking on Microsoft, and that helps position Sun as the leader.
Doc: Sun really has nailed down a strong position as the primary big-time Net-native player. And Scott McNealy is funny as hell. When he was asked the other day about Microsoft, he said, "Which one?"
Phil: Well, the way I feel Linux will creep into the office desktop is on a company by company basis. Take a bank, for example. While there are lots of computers in banks, they don't do a lot of different tasks. Armed with a word processor, a spreadsheet and some program to access their proprietary software running on a server, you could quickly convert a bank over to Linux desktops. This will happen because of cost issues.
Doc: This is what lots of people want to see with Linux appliances. Should be interesting to see how that goes.
Phil: Besides cost, the ease at which you can add things to Linux and the fact that Linux for an appliance can be made really small are two more pluses in favor of Linux in this market. As much as BeOS is a great OS, I think it is going to be hard for Be, Inc. in this market because of Linux.
Doc: Do you give Be much of a chance in any case? How about if they open source the OS?
Phil: I think they bailed out of the desktop at the wrong time. For some high-end IAs, they may have a market, but I don't see them getting any significant percentage. Qt without X (this is new) makes Linux more viable here.
Doc: Let's talk about loves & hates. Why do you hate Perl and love Python? Why do you hate Emacs and love vi? These are often religious issues, but I'm interested in what makes for wise and forward choices.
Phil: They generally are religious issues. In the days of less-powerful computers, you could make the argument that vi is small and Emacs is a pig, but today, it doesn't matter. I learned vi in 1983. I don't even think about which keys to hit to do a task -- it is just magic now. I have no reason to change.
A newcomer should try both. Some people, for example, really can't handle a moded editor like vi. Others see the advantage of fewer keystrokes. Yes, we have an office full of vi users. but I expect some places have offices full of Emacs users.
Perl vs. Python is, to me, different. Historically, Perl evolved from UNIX commands such as sed and awk, whereas Python was designed. Python was designed to be an object-oriented language, whereas OOP features were tacked on to Perl. If you are an old UNIX programmer, then Perl may make sense to you; but for a non-UNIX person, I can't see it.