November 4th, 2005 by Mark Ramm
OK, so I just got out of a BOF session with Mark Shuttleworth and the LTSP guys. And there was a lot of good information in there, which is definitely worth sharing.
To start the session, Mark talked a bit about why he got interested in working with the LTSP guys, and why he thinks LTSP is important. The whole thing started as they thought about what is happening in the open source world that is unique and special, and one of the first things that came to mind was LTSP. Not only that, but it was important to one of Mark’s other projects which had to do with providing a cost effective way to supply computing infrastructure to Schools in South Africa.
Mark went on to say; “I think LTSP is a good solution for a whole host of problems, but it is starting where infrastructure and support costs are critical.” It seems that for Mark getting into the cost-critical sections of the market will provide great stories that drive the technology into other more change averse market segments.
Another critical piece for Mark is his agenda to unify the user experience across various technologies. “It works the same when it boots from USB, or from the Live CD, or from the hard drive, or from the network.” If we are going to see open source take off on the desktop users have to believe that they know what is going to happen, and that the same thing is going to happen every time; “Predictability and reliability are key.”
The Ubuntu team worked hard to make sure LTSP works with the same bits as the live CD, witch works with the same bits as the hard drive install, and that works the same as if you had booted off of a USB key or USB hard drive. They also worked hard to put out a proof of concept release edubuntu, which is great, but now the focus has to shift into making this easy to customize for individual educational needs.
I think Mark is exactly right on all of these points, and that the integration of LTSP into has produced something that is easier to manage, and it is only going to get cleaner and better over the coming year.
November 4th, 2005 by Mark Ramm
Matt Zimmerman, Oliver Grawert, and others have put a lot of work into integrating LTSP into Ubuntu, and creating a new flavor of Ubuntu called edubuntu. And this week Jim McQuillan, Oliver, and a bunch of other folks have cut the boot time of a Ubuntu LTSP in half, and come up with a bunch of other improvements.
And today the LTSP conference is officially getting started. Already even before the conference got started officially, there has been lots of cool stuff going on with local sound, support for local USB devices, an easy to use gui based network PXE boot manager, etc.
Jim says “By this time next year, Ubuntu will have touched more people with LTSP than we have up to this point.” So it looks like 2006 will be an even bigger year for LTSP than 2005 — and 2005 was a great year, with first ever distro integration, best in show at Linux World, and a lot of cool stuff.
November 4th, 2005 by Mark Ramm
OK, so this is the blog entry that I have been trying to write all week.
Over at Toyota and Honda they design cars with an interesting model which I’ll call the “Lean Product Development” model. This is model is very different from the traditional processed used in product design, and the studies show that Lean Product Development is significantly more productive that the traditional processes. In other words Toyota and Honda can design more cars with better quality per year than their traditional counterparts, and they do it with significantly smaller design teams.
But, what the heck does this have to do with Open Source development? Well, in several important ways the open source development methodology mirrors the Lean Product Development methodology. This relationship is a huge topic, and one I can’t claim to have understood fully or to be able to cover completely today.
So for today, I’ll focus on the idea of Set Based Development. Toyota designs some specifications for interfaces between components, and then they design multiple engines, multiple transmissions, multiple alternators, etc concurrently. Then they document the various trade-offs for each of the components, and they bring them together in a series of integration events, where they refine the interfaces, and eventually select the best engine from the pool for this particular model.
This has several advantages, Toyota can work on a revolutionary hybrid engine, with almost no risk, because should the hybrid engine not work or fail to meet the design goals of the car, there is always another choice.
So, this morning Martin Pool showed of bazaar-ng, which is important because it is an example of open source set based development, and because it is an great tool for doing set based development. About a year ago, Canonical decided that they would like to have better tools for distributed revision control. They then started two projects bazaar and bazaar-ng. Bazaar was a conservative project based on incementally improving the best available tool, and Bazaar-NG was research project to see what revolutionary improvements were possible.
This is what Ubuntu Linux does all day do all the time, they have various pre-release events, where all the various packages which might be in the distribution, and if some new package isn’t cutting it they can roll back to something more stable.
For example, it turns out Bazaar-NG is now faster than Bazaar, has a better interface, is easier to extend, and is generally better. So after a lot of testing, Canonical is switching entirely over to Bazaar-NG.
I’ll write more later about why bazaar, and distributed version control, is such an important for set based development.