Greg and I had a good time at Linuxfest. We met lots of neat people, had some good discussions, got some good information out to the people, and got some really good feedback from those in the trenches.
The event was fun as usual, this was my fourth year there. There was room for probably 35 "booth" type things, and as usual the majority of these are for local user groups, distribution groups, special interest groups and the like. There was a few commercial folks there too, Google, Pogo Linux, Wirepath, Fibrecloud, Mysql AB, Oracle, and a few others.
The talk I was most interested in was the K12LTSP talk. This project takes what the LTSP project does, modifies it slightly, rebuilds the FC installer (and CentOS installer) to add a high level option to either install stock FC or to install K12LTSP. Depending on the choice some things get automagicaly done, such as networking, package sets, etc.. They've done a great job with it, and Greg and I both think we need to find some way to help out and to bring K12LTSP into the Fedora fold. They provide a great testcase for variant Fedora releases.
Greg's talk went very well, it was fun, it was informative, and it answered a lot of questions that people had, while paving the way to more questions. His talk brought up the subject of opening Core for outside contribution and outlined a couple possible ways of doing this. Either allowing some folks to contribute to the Core CVS, or moving all things Fedora outside into a big pile of packages that is Fedora, bluring the lines between Core and Extras, and allowing any distribution that is comprised of packages in this pile could be called Fedora. I personally like this idea, but there is a long road to get there. Greg outlined some of the things in process to get to that point, such as yum in Anaconda, external repos in anaconda, and finally variant spun distros.
Some good questions that came up were:
- When are you going to clean up the crap that is the current spec files in Fedora? Valid question, and I explained that all new packages into Fedora have to go through the review process that Extras created. This means that new packages should be clean. I also explained that we're moving to a new build system internally and that would require some clean up of existing packages, which makes it easier to make further clean ups.
- What the hell is up with the Fedora Directory Server? This question came up at least 3 times. First when Greg and I were visiting Gamehouse (Real Networks) in Seattle on our way to the show. An IT architect complained that he's fighting a battle currently with other IT departments on rolling out Active Directory to solve their centralized authentication system. And without Fedora doing something with FDS, it is really hard for him to show a projected plan to having FDS in a supported product. It was also brought up during Greg's talk, and finally at the bar across the street from our Hotel where we spent the evening of the show talking with various other folks that were staying at the hotel for the show as well. The folks that asked the question all seemed to be trying to solve the centralized authentication question, and really didn't want to go the LDAP route, nor the AD route, but had to make decisions soon. What is up with FDS? Other than the name, what does it have to do with Fedora? They don't use our wiki, they don't use our mailing lists, they don't use our IRC channels, they don't have packages in Extras or Core, they don't use our build system, nor our CVS system. What are they doing? How can we bring them into the fold, and start getting their software into the hands of early adopters and folks trying to make a decision? This is a question I think we really need to bring up again and again until it gets done.
- When doing variant Fedoras, what about alternative kernels? Or alternative kernel modules? This is a good and tough question. If we really want to embrace more Fedora variants, more corner user bases solving their problems and helping the project at large, we need to be flexible. We already have something of an alternative kernel in the form of Xen. If you want to run Xen instances, you have to run the base xen kernel. However we don't have an installer that runs this kernel, it isn't necessary. But lets say you need to install to new funky hardware, and the upstream and stock Fedora kernels don't have a driver for say your disk subsystem. The hardware vendor has an open source kernel module available, it just isn't upstream. Following Extras guidelines you package up the module for Extras and get it approved. How can you then use this module with the installer to be able to spin a variant Fedora that will install to your hardware? Even harder, what if its a newer version of an existing module? These are some of the questions we'll have to think about and solve as we move closer to the reality of user spun variant Fedoras.
- While on the subject of variants, we got asked if we'll be using rpath's rbuilder to do the variants. This is an interesting question that neither Greg nor I could really answer. Rbuilder is very neat technology, and whether or not it makes sense to use it should be investigated. Neither Greg nor I had enough information to expand beyond that.
- Whats the status of Stateless, and what is Stateless? Another great question. Havoc mentioned something about Stateless at last year's Boston FUDCon, but nothing has really happened since then. Bill Nottingham just submitted some updates to inittools for stateless purposes, but no real roadmap has been put out, nor has any discussion happened as to how this would effect things like LTSP. I hope to speak with Bill this week to maybe get something out there.
All in all I think it was a good show. I'd like to continue having a Fedora presence there and continue doing talks such as this that explain where we're coming from and where we're going. Folks travel to this show from all over the area, Canada, Oregon, and into the East as well. For those that can't make it to California for LWCE or some of the other shows, this is another chance to touch base with the community at large.