Saturday, August 22, 2009

Some thoughts on a month with Ubuntu Linux

So after I brought home the computer mentioned in the previous post, I repartitioned it with a small Windows XP install (of minimal use -- I think I only used it for a couple of setup tasks) and 25GB of Ubuntu. I picked Ubuntu for a couple of reasons, primarily because the circle of people I can turn to for support are overwhelmingly Ubuntu users, but also because I don't really like the current direction that Knoppix (the distro I ran on the old PC) took in version 6 by being a complete rebuild with LXDE instead of KDE. (I do not actually mind LXDE; it's the fact that Klaus Knopper decided to turn Knoppix 6 into a testbed for a couple of of his pet projects without really acknowledging that he'd created an entirely new product in the process that bothers me. Well, that and the excessive use of special effects in the GUI.)

So where to start? I already had an older version of Edubuntu, but I definitely wanted newer software, so I grabbed the latest version (9.04/Jaunty Jackalope, x86-32) and installed it. The first thing I noticed about it is that it is very overwhelmingly orange; this comes largely from the default theme (the Human-Clearlooks theme), but it's still a drastic change from the grey of MacOS and the blues and greens of Windows XP and Vista. Installation is beyond trivial, which is good since it seems to be the largest issue for new Linux users to get hung up on. The standard desktop is GNOME, which is odd territory for me since I've mostly stuck to KDE on my system, but it's not actually bad. However, it's pretty much the whipping boy for this entire review since, as many people less than enamored of GNOME can tell you, there's a lot of stupid mistakes.

The issue essentially comes down to control vs usability. I've been a Mac fan for over fifteen years now (my first solid experience being System 6 in my old high school) and I've always liked the way the Mac culture enforces a consistent interface between applications. But GUIs do inevitably come in for criticism, since it's very hard to make a spatial/gestural command language Turing-complete; even a macro system like QuicKeys doesn't do more than record keystrokes and mouse events, which is why for many professional Mac developers back in the day, the Macintosh Programmer's Workshop package, despite its high price, was the system of choice for its high scriptability and flexibility in automating the build process, something Think C and Think Pascal (the hobbyist IDEs of choice in the early 90s, before CodeWarrior) had only to a very limited extent. When Apple shipped AppleScript, with an event model that made it remarkably easy to operate applications like marionettes and the ability to use OSAXen (plugins similar to Unix command-line tools), the MacOS finally hit the sweet spot of full graphical control along with powerful automation and access to virtually everything a properly-written application could do.

GNOME... well, the desktop wars are probably the latest flashpoint in a long, long history of religious wars in the Unix world. GNOME has a leg up in one regard, since it's the desktop of choice for Solaris and Fedora Linux, but there are many people (KDE fans in particular) that will tell you that GNOME is a toy interface that hides necessary functions from you. It certainly fails at its attempts to be Mac-like, with a menu bar at the top that does not actually hold application menus (those, like any other X environment, go in the app windows themselves).
And as is often the case in Linux distributions, Ubuntu gives you just a little too much without ever really giving you quite enough, and the Synaptic interface for apt-get is, while usable, rather nonstandard in design and even rather clunky. I suppose that's how it is in the open source world, but did they really have to leave out a device manager app (trust me, lspci is not even close to what is needed) and include a screensaver control panel that doesn't let you adjust anything?

That said, it's not at all the clunky mess older versions like early Slackware or Red Hat were. There is a decent selection of application software included, including OpenOffice and The GIMP, and it's no harder to get up and running than Windows or MacOS. Nautilus is as good a file manager as you're going to find, mostly similar to the MacOS X Finder though it lacks the column view, and accessing network facilities is, if anything, even less annoying than it is on the Mac. Synaptic is nonstandard, but if you've got a sufficient internet connection largely painless.

Ubuntu is doing a valuable service for Linux users by making a concerted effort to create a system that anyone can manage. But it does irk me that after all this time, there are still so many rough edges and roadblocks. Considering the time it's been out there, shouldn't it be a little closer than almost there?

Monday, August 10, 2009

ATX: Why it must die, why it never will

About three weeks ago I picked up a used Dell Dimension of indeterminate model. (Long story, but the motherboard (as far as I can tell, a Pentium 4/Northwood board with an Intel chipset) is substantially older than the case. Why and how, I don't know, as long as it contains what the guy who sold it to me says.) It's nothing to get excited about -- I had to cannibalize my old PC for a sound card, and the video card is a complete joke -- but it's sufficient to run Ubuntu and will probably hold me at least until I can afford a MacBook. I actually like the design -- it's designed in the same vein as the fliptop/drop-down cases that Apple used to be famous for until the G5/Mac Pro case. It's completely screwless, and the inside also replaces some inside screws (for example, for the PCI slot brackets) with latches -- you just lay it down, flip it open from the back, and have at it. I'm not a huge fan of Dell and likely would never buy one new (they tend to play fast and loose with standards and Michael Dell is basically a dullard who had a good idea once), but this is a pretty awesome case, and it rather sucks that their current cases are pretty much the same crappy single-ply ATX cases that everyone else uses on their budget systems.

There really aren't enough easy-access cases like this out there; I imagine they're rather expensive, and, well, you get right down to it, there's this slight problem with cable length and a few generally ignored issues about heat transfer and cable routing and the fact that working in the average ATX case is like building a model train in a goddamned BATHTUB and the few times I've ever seen a workable drop-side ATX case it was a complete cheezy disaster and WHAT THE BLOODY HELL WERE THESE PEOPLE THINKING anyway. The Dell case (which I have dubbed the "butterfly" case) seems to solve the cabling problem pretty effectively, since you lay it down and open it from the back, meaning you don't have to worry about yanking out a drive cable or something like that because the drive cages are right on top of the headers.

So every few years someone, usually Intel, comes out with the latest specification meant to supplant ATX, which is now closing on fifteen years old and exists in several vaguely compatible revisions to compensate for increasing demand for cooling and board neatness. There was NLX, which was supposed to replace the poorly-specced LPX for small form factor systems; there are a few NLX boxes out there, but it wasn't that popular. There was WTX which was supposed to replace ATX for higher-end boxes and had sophisticated thermal management; it went nowhere. There was BTX, which got some traction with some of the larger system builders like Gateway and Dell, but went nowhere on the DIY front and is apparently now a zombie standard. ATX, however, continues.

It was good for its time, don't get me wrong; it was relatively friendly for full-size desktop cases, but no one uses those anymore. And it might have been pretty easy to ignore the lessons of Apple's industrial design; this was during the Spindler/Amelio days, when the Macintosh was still a joke among informed techies and they blew most of their ad budget on product placement rather than actual effective advertising. But the first PCI PowerMacs should have served as a lesson to someone out there -- they were very nice to work with, as were the 630/6x00 series cases with the slide-out motherboards, and Apple went one better with the blue and white G3 models, which could actually open up while the system was still live. (If there had ever been any real market in internal USB or FireWire devices, this would have come in very handy indeed.) But Intel missed that lesson when designing WTX and BTX -- wouldn't it have made sense to put the drive headers on the left or right side of the board so the cables don't stretch in a drop-side or fliptop configuration or have to be, you know, YANKED OUT to get at anything in a regular case? Dell's butterfly case is the only workable solution to the problem of front-mounted headers I've ever seen, and nobody seems to have bothered to copy it.

But that's the disappointments of the PC world -- in a commodity business, even when someone gets something utterly and completely right, it's still based on some obnoxious compromise somewhere, and often isn't cost effective because no one else will cut into their profit margins. So Dell went back to the ATX bathtub like everyone else, because it's what everyone else knows and works with. And that comes back to the title of this post: ATX must die because it's half-baked and barely adequate, but it never will because it's so strongly standardized that it will probably be a standard for at least another decade. That's inertia for you.