Practical Technology

for practical people.

Debian 4.0 finally arrives… does anyone care?


A few months late, Debian 4.0, aka Etch, has been released, but how many people actually will be running it?

That’s not a trite question. It wasn’t that long ago that a new release of Debian would have the Linux world excited and downloading it. Now… well, Debian Etch isn’t exactly being greeted with yawns — but excitement?… that’s not a word I’d use to describe the reactions I’m hearing.

What I’m hearing about instead, is problems.

Now, some of these are minor. I, for example, find it more than almost stupid beyond my powers of description that Debian’s free software political correctness has them renaming Firefox to IceWeasel, Thunderbird to Icedove, and Seamonkey to Iceape. But, OK, while this will mean that Debian’s Mozilla programs will forever more be not as up to date as all other distributions’ versions, this is petty anti-stuff.

Some of the other problems, however, are down right serious. For example, if you use two or more network interfaces on your Debian server or a Debian system acting as a router, Etch is going to flat out break at least one of your interfaces.

Etch comes with CONFIG_IP_ROUTE_MULTIPATH_CACHED (experimental) enabled in its kernel. This setting has long been known to break iproute’s multipath behavior. Iproute is a collection of TCP/IP network utilities. These utilities form the foundation for almost all Linux traffic control and multi-routing. The effect is if you’re running Etch on a system connecting to say multiple ISP (Internet Service Providers), no matter how you set up iproute, you will only be able to use one of those connections. The other one might as well be on Mars.

According to Don Armstrong, a Debian developer, writing on Slashdot, the reason that this “slipped through the release engineering for the new stable release is quite simply because no one reported it as a bug.”

While that’s understandable, it’s not a perfect explanation. After all, this problem with Etch, and its fix, had already been reported on other, albeit unofficial Debian sites such as Debian Help and Debian Administration.

Another problem I’ve been hearing a lot about is the state of the X Window System installer. I’ve heard the X Window System described as “State of the art for 2000.” I’ve worked with Debian Etch enough this morning to agree that that’s a fair description.

Now, in all fairness there are also reasons for this. In a Slashdot note, David Nusinow, a Debian developer, explained. “The Debian X Strike Force burned the entire release cycle moving first from XFree86 to Xorg, and then from the monolithic Xorg to modular Xorg. By the time it all that was finished, about a year and a half had passed and there was a few months to polish things up for the release.”

“During this time,” Nusinow continued, “essentially an entirely new team was built up (only one person from the team that worked on XFree86 in Sarge is still an active member) and there was huge changes as the entire codebase was repackaged for 7.0 and we moved from a private SVN [Subversion, a popular version control system] repo to, which was no small feat while we did our best to keep the updates coming at a good pace.”

I’ll add to that comment that on top of all that, Etch is actually shipping with an even newer version of the windows manager, version 7.1. Still, it is troubling that when so many Debian-based Linux distributions, such as the extremely attractive Xandros, can manage to deliver excellent and easy to set up graphical user interfaces, that the base distribution is still stuck in the stone tools and bunny skins age of X-based GUI setups.

What I find especially troublesome is that Nusinow began his note with: “We know it’s a pain, and it’s a major goal for the next release.” I’ve got three problems with this. First, Debian has always been about “We release it when it’s done.” This sure doesn’t look done to me. Second, the last version of Debian, 3.1 Sarge, came out in June 2005. Will it take almost two years to fix this?

Finally, if only one developer made it from one release to another on something as fundamental as the windowing system, can we take that to mean that in the all-volunteer Debian developer force the odds are lousy that a programmer is actually going to stick with the work? We already know that Debian had an enormous internal fight over some developers getting paid and that as a direct result of that dispute other developers slowed down their work on Etch.

With all this, how state of the art can Debian really be? How important is Debian really? It seems to me that while Debian in the past has been the solid foundation of such excellent distributions as Ubuntu and MEPIS, this release shows that openSUSE and Fedora are the community Linuxes of the future.

Leave a Reply