The high level decision makers in our company insist on RedHat for our production servers because it “comes with support” but I don’t think we’ve ever used their support. At the testing and development level we run CentOS which essentially the same thing without support and the price.

Mr. Riley,

I am not quite as old as you are, and consequently, my programming experience does not go quite so far back, a mere thirty years or so, off and on. Briefly, my experience is: Victor Comptometer 4800 programmable calculator, TI-58 programmable calculator, IBM 1130, mainframe (System 370), and personal computers, of course. Languages: Fortran (Fortran II and WATFIV), PL/I, some SIMSCRIPT, and C, of course. Basic as well, on the personal computer, but that’s really just an version of Fortran. Also, a little S/370 assembler, and rather more Intel 8088 assembler. I’ve learned a bit of computer science on the side, but I was trained as a mechanical engineer (Engineering Science). I have never filed a patent; however, in 1985-86, I did a piece of work which I thought at the time was purely theoretical in character, and published it in a linguistics journal. I was working in the tradition of Noam Chomsky, under the supervision of a leading member of the neo-Chomsky-ian community, building algorithms which were potentially powerful tools and models of the human brain at the same time. Several years later, someone else patented the same mechanism. The patentee got very upset when I eventually pointed out the similarities to him. So much for the patent office’s vaunted examination requirements. I don’t know whether you would count that as experience of invention or not. I realize, rationally, that you, personally, were not the patent attorney involved, but at the same time, this does effect my views.

For the last twenty years, the computer industry has been in a state of incipient decline– not actual decline, but incipient decline. Total dollar volume has been kept up by more people getting computers, but for twenty years, there has been a softness in demand at the top of the market, among the kinds of users who pioneer the way for other users. My own personal maximum expenditure on a computer was in the year 1989. In the year 1985, it was self-evident that one would like to have a $50,000 desktop computer– then called a workstation– even if it was equally self-evident that one could not afford anything like that. One actually settled for something less– say a $2000 computer. By contrast, in the year 2008, it is very difficult to make a credible case why one should buy a $1000 desktop computer instead of a $300 desktop computer (*). Microsoft tried to force the issue, by designing Windows Vista to effectively require a $1000 computer, but the gambit backfired, people tried to run Vista on $300 computers, and Vista got a reputation as a lemon. Some years back, I found that Windows 3.1 ran effectively on a computer about ten times as powerful as the one it came bundled with. The “overbuilding” of Windows is not something new– it is just that this time, it backfired.

(*) Laptops don’t count, because a $2000 laptop is effectively a $200 computer buried inside of $1800 worth of portability. You are not paying for number-crunching ability, but for the expensive screen, etc.

This state of incipient decline has implications in software. If you look at Red Hat Linux, you will find that it has very few features which were not available in super-high-end software for workstations or dedicated mainframes by 1988. By that date, people were using stuff like CAD/CAM in selected places, in selected companies which represented the best practices. The big transition is that such software has gone from being custom-produced, and extremely expensive, to being mass-distributed. The $300 computer of 2008 is roughly equal in performance to a high-end Amdahl mainframe of 1985, with the graphics adapter being comparable to a Cray, and a comparable machine can support software of roughly equal sophistication. Circa 1990, Microsoft was cloning was cloning practically everyone in sight (Apple, Lotus, Borland/WordPerfect), and it was working to break IBM’s dominance of personal computer hardware architecture (remember the Micro Channel Architecture bus episode?). Microsoft got sued a couple of times (on the dubious basis of “visual copyright” rather than on patents), but the lawsuits came to nothing in the end.

This period of incipient decline is of course the period in which patents have been granted for software, but I think patents are a result, rather than a cause of the decline. Filing for a patent is in the last analysis an unproductive act, an act of blowing time and money on lawyers rather than getting more engineering work done. People who are actually engaged in productive work start to chase patents when they reach a certain level of collective anxiety, when the industry is closing down around them. The basic problem is just the law of diminishing returns. The most urgent stuff gets done first, and what gets left to a later time is less urgent. WordStar 3 really was a big change from a typewriter. The latest new version of Microsoft Word is much less compelling. I know at least one person who persistently refuses to stop using WordStar 3, because the benefits of the change do not seem to outweigh the difficulties. I do not go quite that far, myself, but my current word processor has a good many features which I have not troubled to learn to use. On my Windows machine, I am still using meaningful quantities of old MS-DOS software, with file dates ranging from 1988-1992. By 1992, there were DOS word processors which had all the features necessary for scholarly and literary writing, and which were “Windows aware,” meaning that they could use the mouse and access the Windows clipboard. Most of the equivalent applications on the Linux machine are simply not a quantum leap above the old MS-DOS applications.

Imagine that IBM had responded to anti-trust complaints, circa 1980, by open-sourcing, rather than by unbundling, its actual response. Let us imagine further, that the Pentagon had chosen at an early date to subsidize hard-core open-source, ie. people like Richard Stallman, and licenses like the GPL. If we had been using open-source software all the time, we would be able to point to particular files and ranges of lines of code which had gone unchanged for twenty years, and detailed documentation (ie. bug reports) explaining such changes as had taken place, and most patent claims would be comparatively easy to defend against. Of course what actually happened was that everything got cloned a couple of times before open-source emerged. Significant sections of open-source software are beginning to reach this point of historical transparency, and there will be more every year.

If you could come up with credible applications for a $2000 computer, it would be more or less inevitable that the code would be buggy for the first ten years or so, and you could happily go along selling upgrades, just the way they did back in the 1980’s. You would presumably be able to file good patents, but that would not much matter. The same thing which applies to software applies to electronics in general. For example, there is a lot of industry “buzz” about transmitting broadcast television over the internet. Akamai and Cisco are arguing about who owns the “buzz.” However, television over the internet is very unlikely to be transformative in the sense that the VCR was transformative. A VCR meant that you could record a program, and treat it more or less like a book. Devices like TIVO’s have not really extended on this, but they have merely tried to limit it with Digital Rights Management. An extraordinary number of people are filing patents for DRM, which is basically stupid, because DRM does not create value for the customer, and the customer is paying the tab.

You are blaming open-source for your real problem, which is that you cannot come up with credible new applications which would persuade large numbers of people to spend large sums of money. The active development in software is going into things like videogames, and those are necessarily a rather limited market.So tell me, why should I buy a $2000 desktop computer?