May 28, 1999:

I'm getting ready to "burn down the house" on my machine at home. In only 18 months I've managed to fill a 6.4 GB hard drive, and the constant depositing of DLLs in the NT system directory (which of course every uninstaller warns you not to delete!) has begun forcing me to uninstall things from the C: drive. I install and uninstall a lot of things in the course of testing products, so this effect is doubtless worse for me than it is for most people, but for me it's severe. A month or two ago, something started giving Netscape Navigator fits, causing it to crash irritatingly often. The year before that it was completely stable—so my guess is that something in the musical DLLs game poked a hole in it somewhere.

This is how it happens: Little by little your hard drive fills up, and little by little the whole system gets incrementally flaky—so I've ordered an 18.2GB hard drive, which I will plug into the box once I've backed out all data of value. I will then reinstall NT4 and every app I use, along with the data I've backed out onto ZIP disks. Burn down the house, and build it again—and it's a bigger house this time. But it's a shame (and my prime complaint about Windows NT) that you have to burn the house down completely to clean it. Shared code is a diabolically evil, bad, lousy, brain-dead idea! How often do I have to say this? 18GB hard drives for $300, and you still want to share code? I don't. Every app an island, I say—and the only code that should ever be shared are things (like standard controls) that are installed with the operating system. Burning the house down once a year is a bad use of my time.
May 27, 1999:
In studying up on the way assembly language is done today (rather than in 1992, when I last worked extensively with it) I was struck—nay, poleaxed!—by the resemblance of the x86 Tiny model to the 32-bit flat model. In a sense, the Tiny model (which is what you get in a .COM file, with a squiggle at the front inherited from the CP/M era) is 16-bit flat model. All the segment registers point to the same place, and, basically, you let the OS set them and then forget them. How is this different from 32-bit flat model, in which all the extended segment registers point to the same (enormous) segment and are then ignored? My guess: not much different at all. My hunch is that teaching Tiny model (rather than jumping in and tackling segments right away) is the best possible prelude to teaching 32-bit flat model. Little by little, segmentation is becoming a quaint compromise inherited from an earlier era of memory- and register-starved computers.
May 26, 1999:
I've signed a deal to update my assembly book, which has been in print now since 1992. The second edition of Assembly Language Step By Step will focus on NASM, the Net-Wide Assembler, which is free and will be included on the book's CD. I'm hoping to include the NASM-IDE product—a better version of the thing I created for the first edition, called JED—on the CD as well. The book will cover all the same ground as the first edition did, as well as add some chapters on 32-bit flat mode, programmed under Linux. NASM supports Linux as well as DOS (and actually supports Win32 console apps, though I don't think I'll cover those) and it's actually simpler to understand and use than either TASM or MASM. Watch for the book in the spring of 2000.
May 25, 1999:
Got a great little printer to replace my ten-year-old HP Laserjet II. It's another HP—why should I even think of buying a printer from somebody else?—and so far it has not disappointed me. The HP LaserJet 2100 is small, light, fast, and high-resolution. It feeds stacks of business envelopes (up to 15 at a shot) and produces some of the crispest-looking laser output I've ever seen. It's good enough to print your stationery when you print your letter—something I've always hoped for but have not tried—and prints very quickly even at the highest resolution. There's a Mac version (the 2100M) with a PostScript interpreter, and I've heard that that one works well with Linux. The old HP LJII still works fine, but it's so slow it's begun to get on my nerves. I'm hoping the 2100 will be as reliable over the long haul. Ask me about it in ten years.
May 24, 1999:
Check this Web site out: It's the dream of a very persistent visionary to create what amounts to a flying car-not an airplane, but a small Fiberglas VTOL vehicle that uses heavy computing—impossible until now—to basically fly itself from one point to another, guided by satellite GPS data. The blasted thing takes off and lands from a hover, and needs no runway. The propulsion system uses ducted fan engines that run much cleaner and more quietly than conventional aircraft engines. It has a range of almost 700 miles and cruises at up to 350 MPH. The first free flight demonstration could be this summer, and at that point commercial production becomes feasible. I hope to live long enough to have one—but even if I don't, the future will almost definitely contain vehicles like this. What a wonder to live in this era!
May 21, 1999:
I received the domain registration for not long ago, and I'll be describing the project in the September/October issue (#57) of VDM. The basic idea is to collaboratively develop a "virtual encyclopedia" by creating a bookmark manager with a very rich subject-oriented folder hierarchy, and then a system for allowing users to share their bookmarks (classified by dropping them into folders in the hierarchy) with a central database. The database combines all submissions into a single database, maintains them by looking for dead links, and makes them available to users of the bookmark manager. Maybe it'll work—and maybe not. But regardless of whether enough people submit classified bookmarks to make it a virtual encyclopedia, it'll still be a killer standalone bookmark manager, and there's plenty of value in that. The Web site ( should go live late this summer, ideally with a prototype you can download and play with. More in VDM as it happens.
May 20, 1999:
I saw Star Wars last night, on the first day. No, it wasn't for the very first showing (I'm not quite that nuts) but I figured I had a reputation to uphold. And it was well worth it, though not perhaps for the same reason the first Star Wars movie was worth it. The critics were mostly right: The acting was uniformly awful—but the worlds George Lucas and his armies created were almost beyond belief. I'm a sucker for starships and strange cities and beautiful places unlike my own. The fact that these astounding images were created on small computers is very gratifying to me. It means I backed the right horse in my choice of careers and industries. And I'm wondering if the acting was so bad because the poor actors were forced to act out the story on a sound stage empty of the beautiful things we saw all around them. (Yes, it's obvious that the focus of the directing was elsewhere.) Will we have to make these mostly digital movies twice in the future? Once with clueless actors overlaid upon the rendered images of the future, and the second time after the actors have had a chance to see themselves on the screen with their aliens and worlds seamlessly surrounding them? The next few years will tell—Episode 2 should appear by 2002.
May 19, 1999:
The Megido Web site is back online, and it's had a facelift, if not a significant amount of new information. No real explanation of why it vanished, other than some general muttering about network problems at the University of Tel Aviv.
May 18, 1999:
I've learned so much about Linux since installing Caldera's astonishing OpenLinux 2.2 that I've decided to write a followup to the cover story in the May/June VDM. Watch for it in the July/August issue. I'm systematically exploring the KDE environment item by item, and I'm amazed at the goodies I'm turning up.
May 11, 1999:
We're flying to St. Thomas for our spring sales conference today, and I won't be back until the 17th. See you on the flipside.
May 10, 1999:
Is it just me, or does the rendering performed by the Linux implementation of Netscape Navigator really bad? I've watched Netscape run on three different Linux systems now, and all of them have really ragged-looking text, with font proportions that don't match those rendered by Windows versions of what is nominally the same software. The Heading 1 font is much smaller than under Windows, making many pages (including my own personal page at look radically different. I haven't thought much about font management lately, but I know in thinking back that it's a very gnarly business, and I have no idea how fonts are handled under Linux. Windows font management is pretty damned good, and this may be one place where Linux does not equal Windows. (But why does the effect seem limited to Navigator?)
May 7, 1999:
I've discovered an interesting cultural artifact in Linux: Most Linux utilities consist of a command-line oriented back end that runs in text mode, with an optional front end that runs under one of the various GUIs. The RPM package manager is a perfect example. You can run RPM from the notorious Unix command line, or you can run a front end like KPackage (for the KDE GUI) that gives you menus and buttons and builds a command line that it hands to the back end based on your choices. For larger apps like word processors this model breaks down, but for compilers and other similar stdin/stdout utilities it's almost an obsession.
May 6, 1999:
I have finally sent my novel off to a publisher for consideration. Cross your fingers for me—and if you like action/adventure hard SF, you will not be disappointed.
May 5, 1999:
Half an hour of fussing and trying things have failed to make sound support work on my installation of OpenLinux 2.2. I'm clueless. If you've chased this particular goblin around the tree and caught him, please let me know the secret.
May 4, 1999:
After talking up the Megido project in the last issue of VDM, I am beginning to get reports from readers that the Megido Web site is down, and that the Megido list server has gone away. I've sent email to Michael Zayats, the project coordinator, asking for clarification. Megido is a Linux functional clone of Delphi. It's basically a port of the Sibyl IDE and class library from OS/2 to Linux, using FreePascal 32 as the underlying compiler. It would be a very nice thing to have, and I will be most annoyed if the project fails. Any news? By all means let me know what you discover.
May 3, 1999:

Good grief! I'm still not sure I believe what I just did: I installed Linux in fifteen minutes flat, from a dead stop, and not once did I ever see a command line prompt! (Or even text mode, yikes!) Read my next sentence twice: It was easier installing Linux than the last time I installed Windows NT4. The software creates a boot diskette for you, and you restart the machine from the diskette. If you have a bootable CD, it boots directly from the distribution CD and works even faster. A new autoprobe module (written by the Troll Tech Qt guys whose graphics libraries underlie KDE) goes out and detects your graphics card's parameters, so you don't have to fuss with video timing numbers. Once it gets the OS installed, it installs the K Desktop Environment GUI and almost a gigabyte of applications and utilities, running those installs in the background while it allows you to either set up user accounts or—egad—play Tetris. (On my machine, a 450Mhz PII, the installations happened so fast I didn't get a chance to try the Tetris game.)

I'm seeing on the alt.os.linux.caldera newsgroup that not everybody is having as easy a time of it as I am—and in truth, I haven't gotten the sound support to work yet, not that I use it much here at work. My guess is that if OpenLinux can understand your hardware, you won't have any trouble. And if it doesn't, you will. So this is news? My first install of NT4 on a noname box never could make its internal modem work. I'll be reporting on my Linux adventures in future entries. The future looks way more interesting today than it did yesterday. Stay tuned.