Odograph.com

I’ve been busy getting a software release out at the day job. I haven’t heard of anything new happening on the OS front anyway. In the meantime, I did join the ACM and sign up for their Digital Library. It feels a little wrong on two levels. It’s expensive ($188 for a year) and I can’t share what I read with everyone else on the net. There are various movements on the net that argue scientific papers should be freely available. The middle ground, that papers should appear first in paid journals, and after a short time (a year?) appear on the web, seems reasonable to me.

I read a little from “Mastering the Dynamics of Innovation” last night. As a longtime Apple-watcher, this paragraph caught my eye:

Apple Computers remained the sole major holdout from the IBM-driven movement. Apple, too, offered the standard outward features of monitor, keyboard, disk drive, operating system, processing unit, and bus, but its steadfastness in maintaining a closed architecture, proprietary operating system and bus, and reliance on Motorola microprocessors isolated Apple from the larger universe of DOS-based, Intel chip machine users. Despite the greater elegance of the Apple Macintosh machine, the company’s share of the market in 1993 remained stuck at 13 percent. At the same time, intense price competition within the entire industry cut deeply into its profit margins.

I guess Apple has opened a bit since then, moving to the PCI bus, but they’ve more or less ridden the same plan down from 13 percent market share to 3 or 4 percent.

The book provides an interesting angle on this well-known story. Apparently whole industries are often re-aligned when a dominant design arrives.

It seems obvious, when you put it in that perspective, that the “grey box” x86 PC has become the dominant hardware design of the PC era. It remains to be seen if anything can *ever* replace it.

The NSA works for you …

The National Security Agency (NSA) released a new prototype for Security-Enhanced Linux on the 23rd. There are a few papers on-line. Interesting stuff, but I can’t help wish that (good) security could be a little less processing (and data) intensive.

ReactOS

This OS was rev’d on the 17th, a little before I started this log. The homepage describes the OS thus: ReactOS is an Open Source effort to develop a quality operating system that is compatible with Windows NT applications and drivers. I find that a little boring, YMMV.

I got a new book in the mail today: Mastering the Dynamics of Innovation, by James M. Utterback. I enjoyed The Innovator’s Dilemma, by Clayton M. Christensen. I just came across that book a few months ago. I was somewhat surprised to read that the seemingly self-descriptive term disruptive technology actually has a formal definition (see also Linux). It will be interesting to see how the two books compare.

It becomes pretty obvious, if you watch the computer field for any length of time, that technology is not the only factor driving things. It is a game of technology and strategy played out in the market. The strategy part seems as interesting as the technology part.

Hobbyist OSes

Maybe the flip side of the decline in systems research (see below), is the rise of OS writing as an avocation. Even if academic projects are in decline, I think the sheer number of OS hackers must be increasing. The USENET newsgroup alt.os.development seems pretty busy, and the movement has generated a pretty amazing FAQ. I suspect the primary motivation is education, that people dink around with bits of OS code just to be sure they understand it (I know I have). Perhaps this wider developer base will lead to more innovation in the future …

AtheOS

Slashdot ran an item Friday on AtheOS, a free alternative operating system. AtheOS seems to be a new implementation centered on balancing POSIX compatibility with new features. The GUI is new, and looks quite nice. System APIs are presented in C++ and C.

It bothers me how poorly current systems (or at least popular systems) handle software upgrade; at the system, library, and application levels. Clemens Szyperski’s book, Component Software: Beyond Object-Oriented Programming, has a lot of interesting discussion on the topic. The book also surveys a number of existing component systems. Dr. Szyperski is now at Microsoft, working on their Component Applications research. I wish them luck, but wonder how much they can do without running afoul of Microsoft’s strategy tax.

Cutting Research

The Wall Street Journal: PC makers, squeezed by price wars, have slashed spending on product research, leaving much of the important technology development to Microsoft Corp. and Intel Corp. The software and chip giants have much larger profit margins than the computer manufacturers they supply but are themselves struggling to deliver meaningful-enough innovations to keep users regularly upgrading their PCs and programs. [more]

That bit about “to keep users regularly upgrading their PCs and programs” nails it on the head. The industry has become centered on revision-ware. Most people move from Windows to the next revision of Windows, and from a Pentium to the next revision of a Pentium. That’s not all bad, but it does seem to spawn a monoculture .

Rob Pike, of Bell Laboratories, gave a talk on why “Systems Software Research is Irrelevant.” The PostScript slides for his talk may be found here, or PDF here. It is interesting stuff. The presentation covers a number of problems in creating (and finding acceptance for) alternative system architectures.

See Also…

There are certain classes of things that gain value as more people use them. The classic example of this is the FAX machine. If you have the only FAX machine, you aren’t going to get much benefit from it. You gain a benefit (more people you can call) each time someone else gets a FAX machine. It expands the utility of the device.

This network effect seems to be a very strong factor in operating system usage. Indeed, it has become an issue in the Microsoft anti-trust trial.

Regardless of the short-term outcome (with respect to Microsoft), it seems that network effects are the dominant force driving the evolution of computing platforms.