22 September 2005

Operating System (2)

(Part 1)

There are actually a very large number of available operating systems, although the number is generally shrinking. OS's specialized to run on a particular model of hardware naturally fold when the computer architecture becomes obsolete. Those with a small or negligible body of useful applications, of course, don't last either. The success or failure of individual OS's is frequently determined by the corporate strategy of big players; for example, Microsoft's conquest of the market was due not to technical superiority, but selection by IBM. A highly-acclaimed system engineered to maximize the potential of the Power PC chip used in the post-1995 Apple Macintoshes, BeOS, was thwarted as a result of Apple's acquisition of breakaway software developer, Next. Restored CEO Steve Jobs insisted on a unitary Macintosh platform, with one vendor, one system architecture, one OS, one API, no clones, etc. There was no compelling technical reason why it chose a NextStep over BeOS, or even why it did not allow the two to coexist. The decision was political.

Operating systems fall into different categories based on designer philosophy. In a few cases, those that were too limited in their versatility or adaptiveness simply became obsolete as their hardware ceased to be made; or else dragged down the entire system by being uncompetitive. But OS's at either extreme have survived. One involves the ideal of the computer as an appliance so easy to use, even the most tech-averse person could use it happily: the computer as toaster. Such computers involve no choices for their user, come ready to use out of the package, and offer little scope for user maintenance (e.g., micro-ITRON). At the opposite extreme are open systems, that require a tremendous amount of comprehension on the part of the user, including batch file editing; that offer a myriad of choices, such as in data security utilities; and that have many different user interfaces (e.g., for Unix: X-Windows, Solaris, NextStep, Common Desktop Environment, K Desktop Environment, and so on.).

Major Issues of Consideration
The tendency of an OS to become widespread or survive depends upon platform availability (are there machines that can run the system?) , platform durability (are the machines running the system going to be around for a while?), and institutional selection (are vendors or institutions going to chose the system? Are they going to make the investment of time and money to make it run as needed?).

The most obvious example of this is the case of MS-DOS, which was designed for a specific microprocessor and chipset, which was packaged into a system distributed by the world's largest computer company. IBM's awesome market power ensured that anything it released would dominate the market. Naturally, this inspired a lot of wistful speculation: what if IBM had emphasized connectivity instead of speedy development? The importance of its decision imposed a path of technical development that sidelined many promising technical developments.

Other systems, like the PDP-11 and S/360, have been influential by virtue of their great duration. The immense cost of database maintenance, accumulated over many decades, has led many institutions to soldier on with one of these systems 40 years after they were originally built (although I expect anyone still using a PDP-11 is either a hobbyist or a curator). Systems that survive a long time have the fascinating feature of spawning programs that are later adapted ("ported") to other technologies; for example, the previously-mentioned PDP-11, with its popularity among universities, became a testbed for numerous operating systems including Unix, VMS,* and lots of computer science Ph.D. dissertations. Of course many of these OS's matured on subsequent systems; that's my point. The ability of OS's to leap from architecture to architecture led to a parallel zoology of exotic software that existed in emulation shells, but also would re-appear as the basis for something very prominent, such as the Java programming language.

The third principle—the importance of strong institutional support—is illustrated by the demise of BeOS. BeOS had some great talent and brilliant ideas behind it. Originally developed for the AT&T Hobbit processor (Hobbit BeBox prototype), the team quickly ported it to the Power PC (PPC 603 BeBox).** This made it a candidate for the new Mac OS, and in fact, Mac & Be signed a deal allowing Be to distribute clones of the Mac bundled with Be OS. Later, when Apple acquired Next, and Steve Jobs resumed his leadership at Apple, all licenses were terminated and Be was no longer allowed to bundle its software onto Macintoshes. Be then ported its OS a second time, to the Pentium chips (since they allowed ready dual-processor implementation). Still, without any larger institution committed to the BeOS, the company folded in 2001; it was acquired by Palm for $11 million, then launched a lawsuit against Microsoft that bagged another $25 million; subsequently, an applications developer for the BeOS platform, Gobe, took over the publication of BeOS itself. This lasted less than a year before control of the BeOS license passed to a German start-up, yellowTab, and then to Magnussoft. While BeOS users are a passionate, committed lot, they are not an institution; and therefore, prone to fractiousness. Some other BeOS enthusiasts created Haiku, which dilutes the user community.

* VMS is now known as OpenVMS (since 1991). Since it was released the same time as the VAX-11/780, i.e., in 1978, it had to be developed on a PDP-11. Here is a comprehensive list.

** Be abandoned the Hobbit because it had problems interfacing with the digital signal processor. AT&T later ditched the Hobbit, which was evidently developed to run PDAs.

ADDITIONAL READING & SOURCES: "Overcoming a Standard Bearer: Challenges to Nec's Personal Computer in Japan" (PDF), David Methe, Will Mitchell, Junichiro Miyabe, & Ryoko Toyama; "Competing through Standards: DOS/V and Japan's PC Market" (PDF) Joel West and Jason Dedrick; "How open is open enough? Melding proprietary and open source platform strategies" (PDF), Joel West;

"What is Mac OS X?" Amit Singh.

Technical Specifications for Mac Pro; save any work you have before opening this page, because the page crashed my sessions of Mozilla Firefox.

Labels: ,


Post a Comment

<< Home