Ideas & Analysis

There have been several attempts down the years to make Windows unnecessary. The most audacious is doubtless ReactOS, which cuts to the heart of things and wants to be a complete Windows XP-compatible OS. Needless to say, this is no small project and will take a long time to complete; right now, I’d call it somewhere between completely useless and intriguingly experimental. (It runs Skype, at least.) I’m also concerned that if they ever do get it anywhere near useful completion, Microsoft will stomp on it hard.

That’s certainly the high road. But how necessary is it to clone the whole damned OS? A Windows app, after all, is just a block of x86 machine code that makes calls into one or more APIs. If you can clone the APIs in an acceptably clean-room manner, you don’t need to duplicate the entire architecture, kernel and all.

And that brings us to one of the oldest and oddest ongoing projects in open-source computing: Wine, begun in 1993 by Bob Amstadt and Eric Youngdale. Wine provides a compatibility layer consisting of clean-room DLLs implementing the Win32 APIs, plus whatever magic is necessary to make the deeper host OS machinery look like Windows to the app. This is easier than implementing a whole OS, with the further advantage that if done properly, Wine can act as a Windows compatibility layer over several Unix-like OSes, rather than only Linux. Currently, Wine can operate over Linux, Mac OS X, FreeBSD Unix, and x86 Solaris.

After 16 years of dogged work, Wine actually works pretty well. Part of its success is due to a remarkable cooperation between the Wine project and a commercial software house in St. Paul named Codeweavers. Codeweavers sells a $40 deployment/management utility for Wine called Crossover, which basically makes Wine noob-friendly. (Naked Wine is pretty stark.) Codeweavers also tweaks Wine itself to improve app compatibility, and contributes those tweaks back to the Wine project under LGPL. Some financial support is also provided to the otherwise volunteer-based Wine project. Wine’s founder, Alexandre Julliard, is an employee of Codeweavers, where he works full-time on Wine development.

Codeweavers focuses mostly on big-market apps like Microsoft Office, and doesn’t officially support apps beyond a relatively short list of “gold” software. However, I’ve found that a great many Windows apps install and run just fine under Crossover whether they’re on the list or not. InDesign 2.0 is listed on the site as “known not to work” but apart from a minor display glitch, it seems to work as always. (I haven’t tested it deeply so far.) Most Microsoft apps work beautifully (especially older ones) and I’ve been using Office 2000 and Visio 2000 under it without incident since last fall.

Wine implements a sort of runtime environment emulation for Windows called a “bottle.” More than one bottle may be created on a single host OS, and each bottle has its own emulated C: drive and Registry. By giving each Windows app its own bottle under Wine, apps are prevented from interfering with one another in the dreaded “DLL Hell” effect. Because it’s not a VM, the performance hit for running Wine/Crossover is very small, and most important, you do not need to have a legal copy of Windows running in the VM. On the other hand, a bottle looks enough like Windows to be infectable by Windows malware, though one bottle probably can’t infect other bottles on a Linux system, or the underlying system itself. (From what I’ve heard, the low-level system tricks played by many malware packages keep them from running or at least running completely.) There are known conflicts between WGA and Wine, so don’t install WGA if you can avoid it.

Bottom line: If Wine supports all the Windows apps you absolutely must use, you do not need Windows at all. I haven’t tested all the Windows packages that I use here (next up is MapPoint 2004) but for Office and Visio 2000 it’s been nothing short of magical, and I’m guessing InDesign will come along eventually. In a mature software market, time works in our favor: One by one, existing apps will be installable under Wine, and each time that happens, Windows slips a little bit deeper beneath the waters of irrelevance.

Next up: For the hard cases, there’s always virtualization.

In the past two week’s I’ve installed Ubuntu Linux 9.10 (Karmic Koala) three times: twice as upgrades, and once as a clean install from a CD. The combo LiveCD/installer ISO came down with uTorrent in 8 minutes 5 seconds (!!) and has given me trouble only once, and that with the partitioner: I tried to install 9.10 on Carol’s old HP laptop, but the partitioner could not determine the size of the existing Windows partition, and thus the resizer slider would not appear. On an SX270 with a similar size hard drive and an existing XP partition, the partioner resized XP without any trouble. (I’m wondering if there’s some weirdness in the HP BIOS, but have not gone after it yet.) I’ve spent a great deal of time in Ubuntu in recent days, and beyond that one little glitch with the laptop, I’d say Karmic is the best one yet.

Apart from its contrarian (and purely optional) brownness, Karmic’s default GNOME desktop is a great deal like Windows XP. The taskbar functionality is divided between two panels, one at the top and one at the bottom. Now that 20″ displays are common, I use two lines for the Windows taskbar as well and don’t begrudge GNOME the extra line–and you can put both panels at the screen bottom if you want to. The Home folder stands well for My Documents, with familiar subfolders for documents, music, pictures, video, and downloads. Nautilus looks enough like Windows Explorer to pass, especially given that most of us custom-configure Explorer after awhile and not everyone’s configuration is identical. Left mouse button selects, right brings up context. In short, all the basic ideas of the Windows UI are present, in recognizable form. There’s some bumping-up-against-habit in switching from one UI to another, but after even a little time exploring in GNOME you won’t be lost anymore.

The Control Center is a great deal like Control Panel, with applets to manage most configurable options. One gripe about the Control Center is that there is a perfectly good applet to manage grub’s boot options, but it’s not installed by default and I only came across mention of it online by accident, while looking for something else. If you want it, go into the Software Center, and search for Startup-Manager. Install it, and you can specify grub’s default OS, the menu time delay, a splash background, and other things. I take back the grouchiness expressed in my entry for November 7, 2009, with the exception that Startup-Manager needs to be installed by default.

I still have a gripe with Linux generally that goes back a long ways: It needs centralized font management. Unless I’m missing it (and I looked pretty hard) no such applet exists for Control Center, and installing fonts is much more fussy than it needs to be. This may not matter much if you’re not a publisher or a graphic artist, but it matters a lot to me. I have a set of expensive Type 1 fonts that I bought in 2001 and use in all of my Copperwood Press books, but getting them into Linux was non-trivial, and not all apps that should recognize them do. Scribus and poor little Abiword picked them up immediately, but OpenOffice still can’t see them and I still can’t figure why. I know that X11 makes font management a little more complex than it is in Windows, but that’s no excuse for not having a font manager in Control Center.

Beyond that, few complaints. There is now a Safely Remove Drive context menu item for thumb drive mount icons that I don’t remember seeing in earlier versions. Videos that play without sound in Windows (due to obscure codec errors) play with full sound in Karmic. Bottom line: Ubuntu 9.10 implements all the fundamental GUI machinery that Windows does, and does it with enough similarity not to drive a newcomer to distraction. From that standpoint, Windows really isn’t necessary at all…but alas, GUI machinery is only one small piece of the larger Windows pie. More tomorrow.

Still dealing with neck problems here, but in the background I’ve been pursuing a long-term, low-intensity project aimed at discerning how necessary Windows is for my daily operations. Back in 2001 or so, Keith and I considered publishing a book called Dumping Microsoft, but after I looked closely at the Linux releases of the time, decided it would be premature. Windows 2000 was mighty good, and Linux had yet to break out of its all-hail-the-console, pain-is-good hacker culture. It may be time to reconsider the necessity of Windows (and perhaps that book) which is what my upcoming series here is about.

I’ve changed operating systems often enough in the last 30 years not to get too attached to any of them, and have no tribal/emotional investment in Windows 2000 or XP. An OS is just a workbench, after all: It does very little work by itself, and exists almost entirely to help applications do what they must do in the most productive way possible. Too much of modern OS versions is just gratuitous glitz, which eats machine cycles and doesn’t get a page laid out any faster. My reaction to Windows Vista was pretty simple: What’s in all this for me? The primary purpose of Windows Vista was to make itself harder to steal, which is something of a fetish over there–and all the glitz was tossed in to persuade people to upgrade. Furthermore, the damned thing was slow. No sale.

You’ve heard me say many times here that personal computing is mature. I help out people at our parish with their computers, and an astonishing number of them still use Office 97. They may use the latest IE or Firefox, but they haven’t seen the need to spend more money on word processing or spreadsheets. That’s because the need isn’t there. Office ’97 probably has 90% of everything useful in an office suite; if you want the rest, get Office 2000, which I myself have been using for almost ten years now.

It’s taken Linux longer to reach parity with Win2K/XP because it’s written by volunteer programmers and not highly paid engineers on continuous death-march. It’s also required the vision of a consumer-oriented distro firm to package up a version of Linux amenable to non-technical people. The testing I’ve done over the past year (in parallel with revising my assembly language book for Linux) tells me that Canonical’s Ubuntu Linux is ready for ordinary users, and if it were widely available as a preinstall would be considered no geekier than Windows. (Installing Ubuntu from CD is actually loads easier than installing Windows XP. Microsoft would be perhaps a third its current size if not for Windows preinstalls.)

What makes it urgent now is a creeping suspicion I have that Windows malware is unstoppable. I see articles on tech sites several times a week describing new and increasingly clever exploits of flaws in Microsoft and Adobe software. This is troubling on many fronts, from the technical–Why in hell do we still have buffer overflow exploits after all these years?–to the purely political: How do we know that exploits in closed-source software have really been fixed? Linux is not immune to infection (it’s secure at least in part because of its rarity) but the fact is that infections are difficult and rarely seen in the wild. I want to take advantage of that security, whatever its origins.

Hence the current project. I’ve talked about bits and pieces of it here and there in the past year, and it may be time to present what I’ve learned in more detail. Stay tuned.