Ian Murdock’s Weblog

Jason Perlow: “It’s pretty darn clear to me that to make mojo, SuSE Linux Professional needs to look deep into its roots and re-birth itself as a public, open source project similar to Fedora.”

More about the DCC here. Turns out it’s difficult to keep a secret in the open source world (and, no, I wasn’t the source on this, just as I wasn’t the source on the original story). The only addition I have is this: It’s Debian, and it’s Core, but it’s not a Consortium, at least not in the traditional sense of the word.

Alvaro Lopez: “I’m used to having all those facilities [of apt-get] available, and it is really hard living without it after having used it for a while. It’s like travelling in a time machine: using Slackware back in 1995. The proposal I’m about to expose tries to fill this huge usability gap. It tries to bring all those facilities to the OpenSolaris world in order to make Linux users feel at home when they switch to OpenSolaris and, at the same time, drastically improve the new users experience. My proposal is to begin a new Debian architecture based on OpenSolaris.” (Via Tim Bray.)

Matthew Garrett: “In 2001, you released a distribution that was not binary compatible with Debian. Why did you make that decision at the time, and why do you now believe it to have been the wrong decision to make? Do you honestly believe that having the same glibc version as Debian would have made Progeny 1.0 more of a success?”

We made the decision to track unstable for the same reason Ubuntu presumably did: We wanted newer packages. We saw that as a big draw: “Debian with fresher bits” was how we thought of it at the time. The “stable is old” problem was compounded back then because testing did not yet exist, so the choice for users absent Progeny was binary (no pun intended): stable or unstable.

To be honest, we didn’t think a whole lot about binary compatibility (we did, however, view working within the existing Debian community as critically important). If that’s your point, you got me. However, you have to put this in the proper context: This was before the explosion of Debian derivatives (Progeny was among the first, after all, along with Corel and Stormix), so worrying about compatibility then would be a bit like Gentoo worrying about compatibility now.

It was the wrong decision to make in the context of what’s happened with Debian derivatives in the intervening four years. That’s my point. And, no, it wouldn’t have made a lick of difference with respect to Progeny Debian 1.0’s fate. The real question is: Had Progeny Debian 1.0 been successful, would Debian be where it is now? I don’t know.

That’s actually a very interesting question. Let’s take a step into the bizarro world and imagine the outcome had been different: Rather than being a failure, Progeny Debian 1.0 is a huge success. Developers begin building packages on Progeny Debian 1.0. The packages these developers build begin to circulate, and before long, “Debian” packages are out there that don’t work on Debian because they’re linked against Progeny’s incompatible libc. Meanwhile, Progeny is seen as having the momentum, so more and more developers move over to target it as their “Debian platform”. This ultimately takes away energy that otherwise might have been put into the community project.

At the end of the day, though, Progeny is not a community project—it’s a commercial effort, and the money behind that commercial effort eventually dries up because there’s no sustainable business model (that, as you point out, was predetermined, regardless of how successful Progeny Debian 1.0 otherwise might have been). Developers are stranded, and all the energy that might have gone into the community project was wasted on the commercial effort. Most critically, in the intervening confusion, perhaps someone other than Debian emerges as the platform of choice for specialized/localized distros, the source of a lot of Debian’s current strength from my point of view (and this is one area where I think we agree completely).

Perhaps I’m the only one, but I see more than a few parallels here, the main difference being that we haven’t gone full circle with the current incarnation of Progeny Debian 1.0 yet.

Matthew: Perhaps you missed this (CNET News.com, October 15, 2001). In short: Forking from Debian was a bad strategy. It cost us a lot of money, and at the end of the day, the market wanted Debian, not Progeny. The universe did nearly end—for us. There’s a lesson in there, if you choose to take it.

eWEEK: Details Emerge about Debian Linux Plan.

Note: I am not confirming the story by linking to it (nor, by the way, am I denying it by adding a proviso). I simply find it interesting how positive the analysts are about Debian’s chances against Red Hat and Novell/SUSE, if we can get our act together (Laura DiDio’s comment about the vulnerability of Novell is interesting, and something I’ve been pointing out for a while).

Some may read the remarks and wonder why I see them as positive (after all, about the best we get is “it might work”). Simple: Taken in the context of what these folks normally say when asked whether a distro can successfully take on the incumbents, “it might work” is a ringing endorsement, and in any case, it’s pretty clear the general consensus is if anyone can do it, it’s Debian.

Dan Kusnetzky: “With the right investment, of course, anything is possible.”

Here’s a shout out to my big, sometimes dysfunctional, extended family, currently gathering together in Helsinki for this year’s DebConf, Debian’s annual developer gathering. Sorry I can’t be there with you, but I’m in San Francisco next week. Too bad, too, as there are so many interesting things brewing in the Debian world, and I’m sure there is going to be lots of good talk about compatibility and Debian derivatives and other issues critical to our future success. Since I can’t be there in person, I’ve sent Jeff Licquia, who is fully prepared to speak on my behalf. Furthermore, I’d love to join in by telephone in any meetings concerning topics I’ve written about here (release cycle, compatibility and standards, derivatives, etc.). If you can find a room with a telephone, send me the number, and I’ll call in, my dime, even if it’s 3am wherever I’m at.

Stephen O’Grady: “While I’m sure the tactical arguments against continued investments in the JDS Linux platform are easy to make, they probably were too for the folks that shelved Solaris x86 several years back, a decision that Sun still laments.”

I’m with Stephen here. Tactically, it makes a lot of sense for Sun to distance themselves from the enterprise desktop space, given their focus on network infrastructure; and from Linux, given their reemphasis on Solaris/OpenSolaris as the platform of choice at the OS layer of the software stack. Furthermore, where Sun has historically had a desktop story, it’s been centered around developer workstations and thin clients, so retargeting JDS as a solution for these areas makes a lot of sense to me.

Strategically, though, I’ve got this nagging feeling Sun is about to commit another royal blunder, akin to the failure to capitalize on Solaris x86 (which, let’s face it, would have killed Linux before it ever got any traction if Sun had pushed it at all in the late ’90s) and Java (which resulted in BEA and IBM reaping most of Java’s financial rewards). Linux is here to stay, and if I were Sun, I’d want to make sure I had a good Linux story, even as I’m telling my good Solaris/OpenSolaris, developer workstation, and thin client stories.

Want to say the game is over in the enterprise Linux space, with Red Hat and Novell having the lion’s share of the ISV support? The market will probably buy that. Want to compete against these companies with your newly open-source Solaris operating system as an enterprise solution? Given Solaris’ market position and strengths, I think you’ll do well. But what about all those other devices, like server appliances and point-of-sale terminals, where Linux is increasing its footprint and where Solaris doesn’t play at all? Those are part of the network infrastructure too. Furthermore, the game is far from over in a lot of these other markets outside the enterprise space, and the momentum is clearly behind Linux. The time to position yourself to do well here is now, and if Java is the platform of choice for developing new applications, why does it matter if Solaris or Linux are underneath?