[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: F8 kernel-

David Boles wrote:
Callum Lerwick wrote:
On Fri, 2008-03-07 at 11:15 -0500, Jarod Wilson wrote:
On Friday 07 March 2008 10:51:25 am Benjamin Kreuter wrote:
On Thursday 06 March 2008 19:29:23 Chuck Ebbert wrote:
Sorry, we had to release with known bugs. A new kernel will be in
updates-testing very shortly.
Why did you have to release with known bugs? Why not just wait until the
bugs are fixed?  The last three kernel updates broke suspend for me...
Uh... If we waited until all the known bugs were fixed, we'd never release *any* kernel... :)

Despite this kernel making my own iwl4965 unusable, I was fully in favor of releasing it. In theory, we fixed more problems than we caused, and you're always welcome to keep running the prior kernel. (I'm actually running a slightly modified now).

Yes, the real issue here is not all bugs, but regressions. Regressions
are a major problem for Aunt Tillie. Kernel regressions can result in an
unbootable, unusable system. I can't imagine ever deploying Fedora on
Aunt Tillie's machine for exactly this reason, kernel regressions.

Use case: Aunt Tillie diligently keeps her Fedora machine up to date. A
new kernel results in a regression with her hardware. Maybe it doesn't
even boot. What does she do? Can we really expect her to know how to
boot the previous kernel? How is she to even know it is the kernel that
broke? Does she even know what a kernel is? How does she fix it? Booting
the old kernel in GRUB is a one time deal. How does she make it stick?
How does she blacklist the broken kernel? What does she do when 6 more
broken kernels come through the update pipe?

What do *I* do to prevent this? Tell her to not update, and risk
security issues? Should I have blacklisted updating the kernel before
leaving her alone with the machine? Which still leaves the kernel
potentially vulnerable.

This is not theoretical, I ran into this very kind of problem in F7. F7
ran perfectly, initially. A kernel update (a bump from 2.6.21 to 2.6.22,
mind you...) resulted in a reboot loop on my wife's eMachines m6805
(x86_64) laptop. I even bugzilled it right away, though bugzilla's
wonderous search functionality is refusing to find it right now. Many
months and many kernel updates went by, all of them broken. It finally
got fixed when the bug was discovered in the rawhide kernel and ended up
on the F8 release blocker list.

This is a terrible user experience for *me*, let alone Aunt Tillie. I
can't imagine subjecting Aunt Tillie to this without help.

Now, I'm not saying I have the solution to this, and I'm not saying the
solution is easy. But IMHO this really needs to be addressed, somehow,
if Fedora is to ever truly be "ready for the desktop".

I don't think that "Aunt Tillie" should be using a bleeding edge Linux
distribution such as Fedora provides. And if "Nephew Johnie" installs it
for her and she has problems with it that she can not deal with herself I
think it is "Nephew Johny's" fault for installing it for her. What do you

Bzzz, wrong answer, regressions are bad . <period that is> Kernel regressions are worse.

Just because Fedora is trying to be up2date with all the latest and greatest stuff happening in FOSS land, does not make regressions all of a sudden OK.

You (we?) really need to stop thinking this way if we ever want Fedora to be a serious distro, thinking this way inevitably leads to reducing Fedora to nothing more then a toy distro.



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]