Note: I did not write a VDM Diary for the months of January and February 1999. See the note in the March 1, 1999 entry.
March 31, 1999:
I am being pummeled by news reports of the Melissa virus, which is a macro that hides in Word documents sent as email binary attachments. Double click on the document to open it, and the damned thing executes, first infecting every Word document on your system and then emailing itself to the first 500 names on your address book. I have a policy of nuking binary attachments unread from people I don't know, but the whole thing makes my skin crawl. Why do we need macros in documents anyway? I've never fully understood why this was useful, other than to add another pip on the product comparison charts in PC Magazine. Coping with smarter viruses probably requires stupider documents—and even stupider word processors. Bring 'em on!
March 29, 1999:
I finished my novel last night, after sixteen months and 134,000 words. I could not have done it without electronic word processing. The story line is immensely more complex than those I used to hammer out on my old Underwood Standard. In the old days, if I forgot to tie up a loose end somewhere in a sizeable story, I would have had to re-type hundreds of pages, and probably wouldn't have bothered. You can have the Fifties. I'll be happy to be middle aged if I can just have a Pentium and modern dentistry.
March 27, 1999:
I heard today that Sun is pulling the plug on Java Studio, a flawed but extremely forward-looking RAD environment in which everything is visual—including the "wiring" of components with events. I got the distinctest impression that Sun never really got behind the product, which came out of a research project in their labs. They never promoted it, and there were whispers that a lot of people at Sun just didn't like it. No idea why—but could it be that Sun's Unix culture couldn't abide the notion of a programming product that didn't ultimately cook down to a compiler filter translating text files into machine code? The hostility—nay, naked venom—that Unix people display toward visual environments has always puzzled me. This may be another instance. There's little sense in saying, "The best user interface is no user interface!" a slogan I have heard repeatedly from that crowd—unless you really like wasting your own time and working far harder than you have to. Java Studio challenged that bias right at the core of Unix-land, and has apparently paid the price.
March 26, 1999:
After almost four years of continuous use, my office machine, an early P-200, is being handed off to someone else. I get a shiny new Dell P2-400 to put in its place. In looking over the old machine's directories, I'm appalled at the number of orphan DLLs and system files that have piled up there over time. I have installed and then removed a goodly number of software packages in that time, and almost every single uninstall session warns me to leave one or more DLLs on the system, because although no one appears to be using these files anymore, the system simply can't be sure… Crap. The problem is that shared code is a sham and a bad idea, especially in an operating system that cannot absolutely guarantee its safety. Namespace collisions and versioning conflicts cause any such system to grow flaky over time, as mine has, despite its NT4 operating system and the registry that supposedly handles this. It doesn't. Executables should be one piece, always. Each executable should contain every byte of code it needs. With 10GB hard drives now standard, it's becoming more and more pointless to argue that sharing code via DLLs saves hard disk space. Although I can't prove it, I am almost certain that a good part of Windows' flakiness lies in its use of DLLs and its faulty control over what code uses what DLLs, and how DLLs are installed, uninstalled, and upgraded. This is yet another reason I write in Delphi and do not use ActiveX components in my Delphi apps. (Yes, even Delphi isn't perfect—the BDE is a separate module. But it comes way closer than anything else.)
March 24, 1999:
Here's an odd notion: Could there be an open source project aimed at creating a database? (I mean, gathering and organizing the data, not creating a database manager like MySQL.) My idea: Develop a simple app that allows people to gather system-specific information on various models of Intel computers and compatible add-in boards, and transmit that information to a central repository for merging with an installer parameters database. This information would then be available for download with a software Linux installer, which would install Linux on a system using the information stored in the database for that system. The trick is not the database, but how it would be created: By letting everyone with a system contribute specific data for that system to the database. The more who used it, the better it would get, and eventually, hardware manufacturers would contribute data to the database themselves for newly released equipment. If something like this is being done, please let me know who and where. And if not—yikes, guys, what are you waiting for?
March 20, 1999:
I've gotten in with an ISP called Speedchoice, local here to the greater Phoenix area. It's half of the solution to my longstanding problem with limited bandwidth out in cactus country north of town: There's an 8-inch dish antenna on my roof, pointed at the antenna farm atop South Mountain south of downtown Phoenix. The downlink is wireless, while the uplink is a 33Kbps Telco dialup. It's roughly comparable to the Hughes satellite-based service DirecTV with Internet. The downlink is quite fast, probably 600Kbps at its best. It has the downside that the speed is dependent on aggregate usage of the system, so in the evenings, when everybody is out surfing the Web and downloading MP3s and heaven knows what else (actually, heaven knows and I suspect), it gets pretty slow. Still, it's way faster than the 26 Kbps I typically got using traditional ISPs. I occasionally need a fast uplink for sending book chapters and other large files to authors, but at this point it's better than anything else I could get, and with cable TV companies something less than eager to run cables to this less-than-densely-populated area, it may be the best I'll ever do. www.speedchoice.com; currently available here in Phoenix and in Pittsburgh.
March 18, 1999:
It's one of those things you're surprised hasn't been done before, but there it is: Genitor Corporation has a new product called Surveyor that looks at a body of C++ code and creates a Web site for documenting that code. (Go to http://www.genitor.com/ObjectWebs/PaintLib/default.htm to see a Web site created with Surveyor.) You have to fill in some blanks to consider it documentation, obviously, but the relationships in code are represented by hotlinks, and the generated HTML makes intelligent use of frames to make navigating through the code base easy and intuitive. While currently it only runs on Windows platforms, the product is definitely in the emerging spirit of open-source development. I'm encouraging them to implement it for Linux—and would somebody please create something like this for Object Pascal?
March 17, 1999:
St. Patrick's day, and top o' the stack to ye! Being German, Irish, Polish, and French (and maybe Austrian) I get to wear lots of colors and eat lots of cool ethnic foods and have lots of special holidays. Mongrelization has its privileges…
March 16, 1999:
The more I use GNOME, the more disappointed I am in it. It doesn't crash, at least, but it's not all there yet. And I need to study the font support; for some reason, the fonts we have installed are truly ugly and hard to read. One of GNOME's biggest problems may be its protean configurability. There are hundreds of different tweaks, maybe thousands, if you multiply the number of options by the number of installable window managers—and there's probably yet as many things to tweak that I haven't even discovered yet. The need for a carefully-chosen set of defaults is desperate. Appearance is only my first objection; I installed GIMP (a free PhotoShop clone) and it would not run under GNOME. The flash screen came up and it loaded all its tinkertoys, and then just went away. "He dies, and makes no sign." Later I brought up KDE, and GIMP loaded and ran just fine. And application installation—aargh!—is a tale for another day.
March 15, 1999:
We got Red Hat installed on a new lab machine this morning, with the very latest kernel and GNOME 1.0. It took some fooling-with, but what astonished me was the relative ease with which Eric Johnson (our local Linux guru) downloaded and recompiled the whole damned kernel. (I've heard that doing a rebuild on NT4 takes twelve hours.) We put KDE up as well, and I'll be testing both and deciding how to cover both in our future book publishing program. The lab machine is creaky, greasy, and slow; it was in the truckload of heavily used office accoutrements we acquired with the assets of Ventana Press early last year and has been sitting on a shelf ever since. Yet KDE runs as briskly on that limping P-100 as Windows NT4 does on my muscular Compaq P-350 at home. I am most impressed.
March 12, 1999:
Aardmarks is still waiting for a really good data-aware treeview component for Delphi. Woll2Woll has announced one, (their First Class product) and once it ships I'll get down to some serious programming and begin to discuss the project here. Besides, by then I'll be done with my novel and ready to do some programming again!
March 11, 1999:
Here's an interesting question: Will open source software ever invade the Windows world? In some cases, it already is. Apache Server for Windows is out and well-received. (In fact, Coriolis Group Books has the only book on Apache for Windows that I know of: Apache for Windows Little Black Book, ISBN 1-57610-391-9.) GIMP (a PhotoShop competitor with a rabid following) is being ported to Windows. Things like Perl and Tcl/tk have been available for Windows for some time, as are many of the free compilers like Free Pascal. What we haven't really seen yet, however, are any horizontal applications that compete heavily with Microsoft products. If the K Office Suite (a Linux office app compendium for the K Desktop Environment) is ever ported to Windows, you might see wholesale installation of Linux and free software on hypothetical under-$500 PCs. That's when Microsoft's price-to-earnings ratio begins to slide, in my view. It'll happen, too. My guess: Middle of 2000.
March 10, 1999:
As I get out on the Web and start sniffing around for open source projects, I am continually astonished at how many new things are springing up—virtually every single one targeted at one graphical user interface shell or another. Damned little is being developed for strict text mode anymore, even in the Unix world. Someone cautioned me that we're seeing a lot of version 0.1.0 things out there—in other words, two guys have spent a weekend cobbling something together and two commands work, so here it is and have fun! On the other hand, I feel today like I'm standing in a cornfield in mid-June. There's nothing to eat in sight…but there are a lot of ankle-high plants. And even if half of them die before the harvest, boy, are we gonna eat like kings come September!
March 9, 1999:

One of the odder things I saw at LinuxWorld was a family of machines known as microservers. One is the Cobalt Qube, a bright blue 7 inch cube that comes with Linux and every damned server you could imagine preinstalled. It's so small you could throw a hat over it (and I have lots of hats) but so sharp-looking you couldn't bear to. About $1200. www.cobaltmicro.com.

The other one actually comes from Corel: It's the NetWinder, which is about the size of a desk dictionary, and also comes pre-stuffed and largely preconfigured for Web server operation. www.corelcomputer.com. The two machines are comparable in power and price, and if you've got the bandwidth (as many cable modem users now have) either of these machines will put you on the server side of the Web. Note well that neither machine uses Intel processors, so you can't load additional executables from the Intel world. Still, if I were to start again with a Web server on my cable modem in Chicago, I think I would pick one of these things up.
March 8, 1999:
I've received a neat Delphi VCL component library that does something crucial for my Aardmarks project: It accepts OLE drag and drop items from outside the Delphi application. My need is primarily to accept Web shortcuts dragged from Explorer and Netscape, and the component plugged into my testbed with ease and did the job with almost zero coding. First class. The product will accept nearly any data type via OLE drop, and encapsulates dropped data in appropriately typed Delphi objects. The product is UnitOOPS OLE Drag and Drop Components, from UnitOOPS Software in Orange, CT. $25 per developer without source, $75 per developer with source. www.unitoops.com.
March 5, 1999:
There was a guy in the "Greenhouse" microbooth section of the LinuxWorld show floor demonstrating a "universal Linux installer" gizmo, that generates a customized install wizard that will suck Linux off a CD and configure it to your hardware in eight minutes flat. I'd tell you more but I couldn't get anywhere near him. (I know the company name—I think—and will try to contact them in the coming weeks and report further.) This is what Linux needs: effortless installation, first of the OS itself, and then of the applications that run under the OS. Splitting a .tar archive down and piecemealing files into several directories by hand is appalling, and nobody but the nerd elite are ever going to consent to do it.
March 4, 1999:
Just got back from LinuxWorld. Hoo-boy, the ground is moving again, people. I haven't seen this much excitement since the Web first went mainstream in '94. And that's good…it was getting just a hair boring in bitland over the past year, I'm afraid. Now, Linux itself is nothing new—I had it loaded on a lab machine here in 1993—but what's new is that it's breaking out of nerd-land into end-user territory. And that is interesting indeed, especially to me, a student of visual technologies and interface design. I haven't used Unix in any form because it wastes too much of my time. Not my machine cycles—head time. It's dumb having to remember endless catalogs of command-line options. Hey, remembering is what computers are for. We're now getting some genuinely useful GUI shells for Linux, primarily the K Desktop Environment (KDE) and Red Hat's GNOME. And while Linux isn't ready for people who hate or are bored with computers, moderately motivated users with some Windows experience will be able to manage it. And as the shells mature and new GUI software appears, the magical level of required user expertise (RUE) will drop. More on what I saw at the show over the next few days.
March 1, 1999:
VDM Diary's back. Sorry for the lapse. During January and February our company changed ownership and I was out of town a lot. And to be honest with you, there wasn't a whole lot to write about. Microsoft has kept pushing Windows 2000 further out, and the vendor community isn't doing much in the meantime. So rather than bore you with progress reports on my novel or making up news and seeing if anybody would notice, I decided to take a break.