distribution habits and change

Here's another one for the "new workstation series."

Until now my linux-usage has been very Debian based. It's good, it's stable, and the package management system is really intensely wonderful. I was happy. And then I was somewhat less happy. Ubuntu--the distribution that I'd been using on my desktop.

Don't get me wrong, Ubuntu is great and I'd gladly recommend it to other people, but... with time, I've found that the system feels clunky. This is hard to describe, but a growing portion of the software I run isn't included in the normal Ubuntu repositories, some services are hard to turn off/manage correctly (the display manager, various superfluous gnome-parts,) and I've had some ubuntu-related kernel instability.

My server, of course, runs Debian Lenny without incident. There's something really beautiful about that whole stability thing that Debian does. I considered running Debian (lenny/testing) distributions on my desktops because of the wonderful experience I had Lenny, and so I tried running various flavors of Debian and I either found that for the software I wanted to run things were either too stale or too unstable. This is totally to be expected, as Debian's singular goal is stability, and getting a fresher/less stable operating system that's based on Debian is always going to be a difficult proposition.

In the past I've dragged my feet with regards to upgrading operating systems because I take a "don't fix it if it ain't broke," approach to maintaining computers, and all my systems worked so, so until I was faced with this work computer--despite my dissatisfaction--I'd never really seriously considered the mechanics of changing distributions, much less the prospect of having to interact with a linux distribution without the pleasures and joys of apt-get.

But then I got this new work station and...

I switched. At least for one system. So far. ArchLinux uses a system called "pacman" for package management. Pacman is really nifty, but it's different from apt: it's output is a bit clearer, it's just as fast and "smart," and the packages are fresh.

And then there's the whole business of the way Arch approaches management; Arch uses a 'rolling release system" where rather than release a version of an operating system, that has a given set of packages at a given moment, Arch packages are released when they're "ready" on a package-by-package basis (with an awareness toward interactions between packages,) and pacman has a system for easily installing software that isn't in the main repositories as packages. (Which makes upgrading and removal of said packages later much easier.)

This sounds complex, maybe too complex, but some how, it's not. When I started thinking about writing this post, I thought, "how do I convey how totally strange and different this is from the Debian way." By the time I got around to actually writing this post, I've settled into a place of stability and I must confess that I don't notice it very much. It's wild, and it just works.

I was going to go on this whole schpeal about how even though the functional differences between one "flavor" of GNU/Linux and another are pretty minimal, it's interesting to see how different the systems can "feel," in practice. Here's a brief list of what I've noticed:

  • I find the fact that the flags that control operations with pacman to be non-intuitive, and frankly I'm a bit annoyed that they're case sensitive so that: pacman -S is different from pacman -s which leads to a lot of typos.

  • I've yet to set up a machine in Arch that uses wireless. I'm just wary, mostly, of having to set up the network stuff "by hand" in Arch, given how finicky these things can be in general.

  • The ABS (arch build system, for installing packages that aren't in the main arch repositories,) took some getting used to, I think this is more about learning how to use a new program/tool and the fact that the commands are a bit weird.

    Having said that, I really like the way the package building scripts just work and pull from upstream sources, and even, say, use git to download the source.

  • I'm impressed with how complete the system is. Debian advertises a huge number of packages and prides itself on its completeness (and it is, I'm not quibbling,) but I run into programs where I have a hard time getting the right version, or the software plain old isn't in the repository. I've yet to find something that isn't in the repository. (Well, getting a version of mutt with the sidebar patch was a bit rough, and I haven't installed the urlview package yet, but that's minor.)

    I think this is roughly analogous to the discussions that python/ruby people have with perl people, about actual vs. the advertised worth of the CPAN. (eg. CPAN is great and has a lot of stuff, but it suffers from being unedited and its huge number of modules is more an artifact of time rather than actual value) So that, while Debian (and CPAN) have more "stuff" than their competitors, in many cases the competitors can still succeed with less "stuff," because their "stuff" has value because it's more edited and they can choose the 80% of stuff that satisfies 94% of need, rather than having 90% of the stuff that satisfies 98% of need. Diminishing returns and all that.

    It's lightweight and smooth. While the hardware of my work computer is indeed impressive, particularly by my own standards, the hardware of the "virtual machine" isn't particularly impressive. And it's still incredibly peppy. Lightweight for the win.

Next up? More reasons Arch Rocks.

See you later!

comments powered by Disqus