On Thu, Sep 29, 2005 at 01:17:03PM +0300, guy keren wrote: > hopefully, not too fast - you should remember that there is no 2.7 kernel > series any more - all the development is now done in the normal 2.6 kernel > series, and they assume the distributions will do the stabilization work.
Not only distributions - there is also a team releasing 2.6.x.y versions which are bugfixes only. So if y is large enough, or the patch from y-1 to y was small enough, you can still feel quite confident about it. And we do recall that 2.4 (and 2.2 to a lesser extent) also wasn't the most stable kernel until around a year after the .0 release. > > as a result, there are quite massive changes in the newer 2.6 kernel > releases. thus, distributions tend to update slower now (i'm not talking > about 'cutting-user-fingers-edge distributions' - i'm talking about those > that try to be somewhat stable). > > this is just a reminder to y'all, to be more catious then before about > whether to upgrade your non-experimental machine(s) to the latest "stable" > 2.6 kernel. > > does someone know if the 'unstable work' on the 2.6 tree was official > started? i do know that some non-trivial changes were made in recent > versions (e.g. in 2.6.12, the SCSI sub-system supports plug-and-play - > this means that SCSI disks would disappear right under your nose if your > fiber-channel's HBA got disconnected from the network for a short while). I only read kernel-traffic, not lkml directly. As far as I understand, the latest decision was that the 2 weeks after a 2.6.x release are for adding new features, then only bugfixes with x+1-rc releases, then a x+1 release, and a release is followed, as I said, with 2.6.x+1.y fixes. I do not think they intend to maintain 2.6.x.y when 2.6.x+1 is released, which means each version will enjoy a somewhat short life, which might not be enough to stabilize it, and therefore distributions will probably backport patches to the version they chose instead of moving forward. But this is all relatively new - we should still wait and see how it works in reality. BTW, people still use phrases like "This is 2.7 material", meaning they do not expect the current methodology to be good enough for really large changes. -- Didi ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]
