Arbitrary division between Core and Extras. Decision to move everything into one build system, one compose system, one distribution. Many reasons: spread out the load, many eyes -> shallow bugs. If anyone can create a “spin” of Fedora. Red Hat does a good job of engineering the core, the build, the methodology. By enabling Fedora to be accessible to many, specialists can turn Fedora to their uses.
1. Getting source code control merged.
2. Getting packaging standards up-to-date. For example, SysVInit, conditional builds depend on files that haven’t existed forever.
3. Adding new features to the build systems. Currently using Plague. (ref. Jessie’s talk this morning)
4. Punji – “anyone can compose a Fedora distribution”
New feature: simple ways to customize a distribution. Punji only lets you pick packages. This is more parameters: ex. start some services by default. Changing the default desktop. Could be done now with a Kickstart file, but that’s an ugly post-build solution.
Fedora Update System: Luke Mecken (?) is working on this. In-house, developer pushes a fix, with a short description, what bugs are fixed, etc. Depends on internal build system at Red Hat. Hope to target Fedora, Extras, and EPEL (Extra Products for Enterprise Linux).
User-oriented features: wireless support. Next-gen wireless support “DeviceScape.” (whole list on the wiki). Nuevo device support. nVidia support a licensing issue – only 2-D freely distributable. Nuevo project is reverse-engineering 3-D support.
Fast-user switching, he says “about time, Windows has had this for years.”
Lots of little fixes: noisy apps the wake up the kernel 10 times a second, startup stuff that times out waiting for hardware you don’t have. A “tickless” kernel. R&R 1.2: from the X.org community. Online detection of device installation and removal, rotation and resizing.
KVM virtualization, which Jeremy talks about later.
Q&A: how about a Migration Wizard that notices that you’ve got FireFox and Thunderbird on your NTFS partition and offers to migrate the settings to Linux? (A: Hmm. NTFS may be an issue, but the code is out there.) How about full disk encryption for laptops (partial answer: no, portions only: home, swap, data, using ecryptfs.) Q: Version numbers: will we ever see 10?
Project Smolt: instrumenting what applications are used. Very sensitive to privacy issues. The LHCP is a larger-scoped project for building databases of hardware compatibility.
SysVInit replacement: there’s a horse race with several competitors. Opinions about what’s what appear to be all over the map. A replacement has to be compatible, support thrd-party packages and broken init scripts, and maintain LSB compatibility. Q whether LSB compatibility is met for init scripts and discussions on their vagueness.
How close is the Core-Extra merge? In process, targetted for Test 2.
Q: on the build system, which dissolved into indecipherable acronyms.
Tag: fedoraboston2007