oreilly.comSafari Books Online.Conferences.


Growing with Gentoo

by Danny O'Brien

When Daniel Robbins recently spoke to 30 or so students and professors at Stanford's Computer Systems Lab, he didn't have to waste much time explaining what his baby, Gentoo Linux, was. Most of the audience seemed to have at least one Gentoo playbox each. And those who gathered around the project's founding father after the talk seemed pretty happy with how Daniel's distribution has been shaping up.

Instead of going into the details of Gentoo that have made it so popular -- the distribution's preference for source packages compiled by the user over binary packages, its ultra-configurability, the heavy automation of Gentoo's emerge build system -- Daniel sat down and spoke on the lesser known aspects of Gentoo: where it came from, what problems and opportunities Daniel was seeing these days, and his opinions on where the future might take it.

Gentoo's genesis was, like so much open source software, as academic as the talks' setting. Robbins was working as a system administrator for the School of Medicine at the University of New Mexico when he first started playing around with Linux.

He began with Debian, later moving on to work with others on developing an optimized distribution for the Stampede project. Stampede was an open source endeavor that suffered from personality clashes and, says Robbins, a poor management approach that did nothing to allay those problems. Robbins decided to move on (Stampede is no longer maintained), and started working on rolling out his own distribution.

A lot of Gentoo's philosophy and appeal comes from this time, Robbins says. Portage, Gentoo's famously automated build system, "was necessary, because I was the only person working on Gentoo, and I had time only to do everything once."

Learning Lab TigerLinux/Unix System Administration Certification -- Would you like to polish your system administration skills online and receive credit from the University of Illinois? Learn how to administer Linux/Unix systems and gain real experience with a root access account. The four-course series covers the Unix file system, networking, Unix services, and scripting. It's all at the O'Reilly Learning Lab.

Robbins also set out to make the distribution's development as open as possible, to prevent the short-circuiting of command that beleaguered Stampede.

These days, Gentoo is a little bigger than just Robbins. The ebuild tree holds 7,000 packages. Development and support is spread across dozens of mailing lists and IRC channels (with the main list, gentoo-users, receiving in excess of 100 messages per day). Robbins guesstimates that over 50,000 people have a Gentoo installation. ("We have no figures. It's over a thousand, I know that.")

With such an explosion in the size of the project, Robbins admits that the distribution is struggling to adjust. "I delegate. But the biggest problem is finding enough time to even delegate."

Much of his design is coping well with the shift in scale, and those elements that do not are, he says, receiving attention; USE flags, for instance. USE flags determine high-level, global compilation settings in Gentoo, such as whether all packages should be compiled with GNOME support, or with SSL turned on. With so many packages contributed, the list of USE flags has mushroomed in size. Robbins is looking at ways to turn this into a hierarchical list.

The biggest problem, though, is deciding in an open manner which direction Gentoo should go.

Gentoo has always worked on what Robbins describes as a minimally bureaucratic system. Gentoo developers, for instance, have CVS access to the whole Gentoo tree, allowing them to tweak packages that are not their own primary concern.

Fitting users into this model has been difficult, because until recently the stereotypical Gentoo user was a developer. And coordinating the future development of key utilities (like emerge, Gentoo's build utility, and catalyst, its distribution creator) with so many offering suggestions has grown impossible.

But Robbins has made a conscious decision to avoid the political governance model of Debian, which revolves around elected officials and fixed, voted-on policy. Instead, he's been looking at informal frameworks like Python's PEP system, which has an open period of solicitation for new feature suggestions from users, and then a closed, fixed time period -- this is when the main Gentoo developers work on implementing what has been decided.

Perhaps as a nod to his audience, Robbins also admits that Gentoo could do with some traditional project management. "There are a lot of us [in open source] who don't have formal software development skills. There's an arrogance that comes from getting something as big as this to work, and it can be bruising to one's pride [to admit that we need to know more] ... that we should have paid more attention to that class in college."

O'Reilly Open Source Convention.

But Robbins is most concerned that Gentoo preserves its reputation as the most open and flexible of distributions. When asked whether he thinks Gentoo's size will lead him into making it as inflexible as other distributions, Robbins answer is simple: Gentoo's developers won't let him.

Gentoo has other reputations, as well. Some are old and outdated -- you no longer have to spend a week compiling glibc before you can get anything to work. New customized installs (called "Stage 2" and "Stage 3") can let you skip the most time-consuming builds, or even skip to a complete pre-compiled suite of the most common packages.

Others have a germ of truth to them. Robbins acknowledged twice in his talk that Gentoo users have a reputation for pestering upstream open source developers with bug reports. Some have been legitimate -- the idiosyncratic configurations permitted by Gentoo often shook out obscure problems in the most stable packages. There's a general feeling among some developers that Gentoo users are identifying problems caused not by upstream bugs, but by aggressive optimization or other poor configuration choices that the users themselves have made.

Robbins wants to stop this, and emphasizes that if Gentoo users have a problem, reporting bugs upstream to the software's original author is the wrong place to start, and that they should let the Gentoo development team know first.

You sense, listening to Robbins (and hearing him listen to others after the talk), that the aspect of Gentoo that he is striving hardest to preserve is the sense that there is always somebody you can reach to help you solve your problem. This is what Robbins believes lies at the heart of Gentoo's appeal.

"We started out as an IRC project ... I'd release it, and people would give me instant feedback on bugs. No formal release cycle, just instant development. I really enjoyed getting feedback, and then fixing it and saying 'here you go.'"

"This is the heart of Gentoo: a technology where if someone doesn't like how it works, we can change it, or teach them how they can change it themselves and make their tool work for them. The heart of Gentoo is that concern for others, and the joy we had helping others."

And what of the future? Daniel keeps diplomatically quiet on most of the fine points, but nods at the barrage of suggestions from the Stanford audience. One wants cross-compilation so he can build Gentoo for his older Sparc systems without it taking weeks to create. Daniel says that cross-compilation of a sort will be in 2004.1, the version due to be released next month. "It's aimed primarily at embedded systems developers so they can create a distribution on the host machine, then transfer it over."

GPG-signing is another request. Gentoo etree releases are sometimes MD5-signed, but the system as a whole isn't necessarily secure (a hostile attacker could fake the MD5s as easily as insert new code). With an eye to Gentoo's increasing use in the enterprise, Daniel says this will be another incoming feature.

Whatever happens next to Gentoo, Robbins' place at the center of it seems assured, at least. He revealed that he has a Gentoo-inspired salary these days, thanks to an anonymous German benefactor who called him late one night and offered him money to continue hacking Gentoo.

Even without the financial help, you get the feeling he'd carry on regardless. And when one fan tells him that Gentoo has "been the most fun I've had with Linux since 1996," Robbins looks like he can only agree.

A Windows Media stream of Daniel Robbins' talk at Stanford's Computer Systems Laboratory Colloquium is available online.

Danny O'Brien is a journalist and a cofounder of the Need to Know newsletter.

Return to the

Linux Online Certification

Linux/Unix System Administration Certificate Series
Linux/Unix System Administration Certificate Series — This course series targets both beginning and intermediate Linux/Unix users who want to acquire advanced system administration skills, and to back those skills up with a Certificate from the University of Illinois Office of Continuing Education.

Enroll today!

Linux Resources
  • Linux Online
  • The Linux FAQ
  • Linux Kernel Archives
  • Kernel Traffic

  • Sponsored by: