Most Read Technology Reporter For More Than Two Decades

Maureen O'Gara

Subscribe to Maureen O'Gara: eMailAlertsEmail Alerts
Get Maureen O'Gara: homepageHomepage mobileMobile rssRSS facebookFacebook twitterTwitter linkedinLinkedIn


Related Topics: Apache Web Server Journal

Apache Web Server: Article

Bruce Perens on UserLinux

Get ready for a bulletproof distribution

Bruce Perens, cofounder of the Open Source Initiative and long-time leader of the Debian GNU/Linux distribution, announced plans at the November 2003 Desktop Linux Consortium event in Boston to start a project called UserLinux. UserLinux is to be a Linux distribution based on a subset of Debian that will target large and small business desktops and servers. Bruce is currently continuing negotiations with his customers while also beginning to put the first broad brush strokes on UserLinux as a technology.

A key aspect of the UserLinux strategy is to build on a solid open source software core and then augment that core with customer-funded engineering work to support further software development, certification, support, and service.

Bruce and Paul Nowak discussed what UserLinux is and what it means to software users and to open source developers.

LWM: What does UserLinux mean to corporations and other organizations that will be paying for the software engineering and using the software?
Perens:
Linux vendors like Red Hat really are not in a position today to be able to support the software they distribute. If you look inside Red Hat, they have 300 engineers but only a tiny fraction of those actually support the software.

UserLinux is taking the approach of "let's have a lot of support companies working together as equals on UserLinux, so that you can find the expert that you need, and so that competition drives quality up and prices down. Let's encourage service providers to differentiate themselves by specializing in niche markets that they know well. I want there to be so many UserLinux service providers that you'll be able to find a company that specializes in supporting dentists in Minnesota. And I don't want to own any part of that company - I just want to be its equal partner in developing the UserLinux system." And when you think of it this way, it turns out to be an approach that is particularly good for the more technically challenging markets because those are the markets that a Red Hat or SUSE can't go to. Red Hat is bound by strategies that enhance shareholder value, so they have to focus on the big market.

But, because software follows Pareto's law, which says 80% of the users are easy to satisfy, we should find that UserLinux is very good for a lot of users, not just the challenging markets that we initially build UserLinux for.

LWM: What are examples of technically challenging vertical markets?
Perens:
Industries that first have special hardware are often the most challenging. Industries that require special software applications are often the next most challenging. Finance, science, and the creative arts are examples.

Existing Linux vendors are driven to look at the large body of users - the 80% that are easy to satisfy - and because of that, the users with technically challenging requirements may never have the issues that are important to them addressed.

LWM: Users are one important aspect of UserLinux; developers are another. What does UserLinux mean for developers?
Perens:
For developers, UserLinux means a fair partnership. If Red Hat had not come out with Fedora, I would have described what a fair partnership means, but with Fedora there an example of what an unfair partnership is so I will focus on that.

Fedora is Red Hat trying to implement a project like Debian. Only Debian has been around for over 10 years - longer than Red Hat - so Red Hat is just reinventing Debian but with the corporate constraints of Red Hat layered into the policy surrounding the project.

Mike [Michael K. Johnson, technical lead of Fedora and Red Hat employee], is a good person, but Mike represents Red Hat and when a decision needs to be made that runs counter to the needs of Red Hat, Mike is limited in what he can do. This is already happening with the Fedora Leadership Document.

On December 2, 2003 (two days before this interview), Red Hat came out with the second draft of the Fedora Leadership Document (http://fedora.redhat.com/about/leadership.html). In this draft, Red Hat eliminated voting because they want control, which they freely admit in the document is because they don't think voting works. Debian has over 10 years of successful voting, so I think the record shows that voting could work on technical and political issues if a Red Hat corporate agenda were not involved.

With Red Hat in total control, developers should ask themselves if they want to be an unpaid employee of Red Hat - that is what working on Fedora means. It's not an equal partnership. It's a relationship in which Red Hat sets the agenda, and the technology developed through the Fedora community will end up in Red Hat products.

It's almost humorous that Red Hat has never been able to get package management to work as well as Debian's apt-get. Debian has over 12,000 packages, several times more than Red Hat has RPMs, but Debian's system works because all software either resides in the Debian repository or links to the repository for all its dependencies.

UserLinux is going to have Debian as a base but with 12,000 packages, we cannot hope to support all of Debian, so UserLinux will focus on a subset of Debian and certify that we can support that subset. To do that, we are going to have to make choices. We are going to have to pick a GUI, and that's going to be a tough decision with a lot of strong views involved. If a customer later wants us to support another GUI, we will if that customer is willing to pay for the work involved. We are going to have to pick one Web server, and Apache II looks like the choice there. We are going to pick one mail transfer agent. Things like this. Debian is really high-quality software throughout, but our intent is to make this subset of Debian something that we can make bulletproof and something we can support.

When we engineer software or modify code as part of UserLinux, our intent is to merge that back into Debian. If for some reason we can't get it back into Debian, we will create our own separate repository under UserLinux.

[Editor's note: UserLinux has matured since the December interview and is now testing working distributions and iso images. Core software packages have been selected for Server, Enterprise Desktop, and SOHO/Home developer targets. Core packages include GNOME, Apache, PostgreSQL, Python (as the primary scripting language, but PERL is included too), and more. For more information see www.userlinux.com/cgi-bin/wiki.pl?PackageFramework.]

LWM: What will the UserLinux policy be on closed source software? Specifically, what if the video hardware folks insist on close sourcing the software drivers for their video cards?
Perens:
We are in a terrible situation with the 3D video drivers today. I've discussed this with one of the UserLinux customers and they feel the same way. For example, for over one year, we could not get the gamma, which is like contrast, to adjust correctly on a 3D video driver. This is a 15-minute fix in open source software.

So the problem with NVIDIA and ATI is that we cannot properly maintain drivers with complete APIs and correct optimizations. Most of the trade secrets on video technology are in the hardware, but I think part of the issue is that the video card folks do not want to open the source because they think it will enable people to see what patents they are using in their products.

It's not a good situation, but new entrants are coming into the 3D video market that may make someone want to open the source. One of our customers needs 3D video so we will need to address this issue.

LWM: Enterprise software from Oracle and other providers has been a big boost to the adoption rates of commercial Linux distributions like Red Hat. How does UserLinux view these relationships or partnerships with large enterprise software vendors?
Perens:
Of course it's better to have Oracle customers when you approach Oracle.

If you really look at Oracle, the message Oracle is sending when they advertise and market Oracle on Red Hat or Oracle on SUSE is that Oracle runs on Linux. Oracle is really pairing up with Linux, not Red Hat or SUSE.

It's also not that big of a deal for Oracle to integrate into a new Linux distribution. Oracle is on a distribution called "Miracle Linux" from Japan and Oracle certifies that distribution - I think Oracle owns 50% of Miracle Linux.

.  .  .

Bruce and I did not have time to cover the way UserLinux will interact with existing service and support vendors or with a range of other aspects of the UserLinux effort. For more information, please review the UserLinux white paper at http://userlinux.com and the LinuxWorld.com article about UserLinux at http://linuxworld.com/story/38270.htm.

More Stories By Paul Nowak

Paul Nowak first used Linux in 1995 while migrating from Sun to Linux at the University of Michigan. He used Linux in subsequent IT projects including web, telecom, telemetry and embedded projects and is currently CIO of a small professional association based in Washington D.C.

Comments (12)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.