2.4.x release process comments
Hi all
I'm a little worried about the support for the 2.4.x kernel series. In the past few releases I've noticed a couple of things that I think are bad for the kernel.
- Kernel version releases (and -pre releases) do not happen often enough to keep up with recent hardware
Maybe it's the contrast with the speed of releases on the 2.5.x series, but while I understand the need for stability on the 2.4.x series, bugfixes and hardware support should be kept up to date in the stable series as well. Distributions need this, since every 6 months a new release is made (mandrake, redhat, suse). If it is not kept up to date, the distros start using 2.4.x-pre series to provide support for the most recent hardware on which the new distros are going to be installed.
I'd love to see regular (say once a week) releases -preX releases and no more than 10 -pre releases before a -rc. No more than 4 -rc's (released no more than 2 weeks apart) before a new version. Faster full version releases would also be fine with me.
- Marcello doesn't participate (much) on the kernel list
Again, not really a problem on its own, I'm sure Marcello has a lot of things to worry about, but somehow it detracts a little from the feeling of "trust" not to see regular posts form the kernel maintainer about bugs, patches and how/why patches are applied/ignored.
- Alan Cox seems to provide 90% of the patches to the 2.4.x series
Obviously Alan has responsibilities for redhat, so that explains a lot. However I get the feeling that patches for 2.4.x are more and more directed to him rather than Marcello...
My personal interest in this is that my laptop is not yet working 100%... (see http://margo.student.utwente.nl/simon/ongoing/jade8060.php)
BTW, I've been following the list for at least a year now (mostly via kernel traffic and kernel releases, but now I'm subscribed)
Cheers everyone and thanks a lot for the work on the kernel!
Simon
Quoted 10 lines
- Kernel version releases (and -pre releases) do not happen often enough to keep up with recent hardware
Maybe it's the contrast with the speed of releases on the 2.5.x series, but while I understand the need for stability on the 2.4.x series, bugfixes and hardware support should be kept up to date in the stable series as well. Distributions need this, since every 6 months a new release is made (mandrake, redhat, suse). If it is not kept up to date, the distros start using 2.4.x-pre series to provide support for the most recent hardware on which the new distros are going to be installed.
Distributions typically use modified versions of the standard kernel anyway. The kernel development schedule has never been based around the needs of any particular distribution or distributions.
Quoted 4 lines
I'd love to see regular (say once a week) releases -preX releases and no more than 10 -pre releases before a -rc. No more than 4 -rc's (released no more than 2 weeks apart) before a new version. Faster full version releases would also be fine with me.
It doesn't make sense to limit the number of -pre releases and release candidates - they are needed to make sure, as far as possible, that the actual release is stable.
Quoted 6 lines
- Marcello doesn't participate (much) on the kernel list
- Alan Cox seems to provide 90% of the patches to the 2.4.x series
However I get the feeling that patches for 2.4.x are more and more directed to him rather than Marcello...
I don't see a problem with this, (although we shouldn't just be lazy, and send all patches to Alan when there are more appropriate people).
Just because somebody is maintaining a subsystem or kernel tree, does not mean that they have to do the most work, or that nobody else is expected to edit that code.
Quoted 2 lines
My personal interest in this is that my laptop is not yet working 100%... (see http://margo.student.utwente.nl/simon/ongoing/jade8060.php)
I had a quick look, and it looks like it's 95% working, though :-).
John.
Quoted 6 lines · John Bradford
Quoted 3 lines
- Kernel version releases (and -pre releases) do not happen often enough to keep up with recent hardware
Distributions typically use modified versions of the standard kernel anyway. The kernel development schedule has never been based around the needs of any particular distribution or distributions.
I'm not saying it should, but it would be good from a PR perspective and as an element in the reliability feeling vector ;-)
Quoted 8 lines
Quoted 4 lines
I'd love to see regular (say once a week) releases -preX releases and no more than 10 -pre releases before a -rc. No more than 4 -rc's (released no more than 2 weeks apart) before a new version. Faster full version releases would also be fine with me.
It doesn't make sense to limit the number of -pre releases and release candidates - they are needed to make sure, as far as possible, that the actual release is stable.
The number of -pre releases shouldn't be limited for its own sake, but rather in the process of stabilising the kernel for release. So I mean after a couple of -pre releases start focussing on debugging and then finish with a few -rc's before the next cycle starts. That way the diffs between full versions are smaller and upgrading gets easier.
Quoted 4 lines
Quoted 2 lines
My personal interest in this is that my laptop is not yet working 100%... (see http://margo.student.utwente.nl/simon/ongoing/jade8060.php)
I had a quick look, and it looks like it's 95% working, though :-).
yup, I'm getting there :-) Would be nice to have the smartcard reader working as well... Some things I can't test because I don't have the hardware (IEEE1394, wireless lan, irda). power management is still quite magical to me though ;-)
Cheers
Simon
Hello,
Quoted 3 lines
I'm not saying it should, but it would be good from a PR perspective and as
an element in the reliability feeling vector ;-)
Not sure about it... People like it when a product looks stable, and having a -blah or -pre and so on once a week doesn't make me feel I have some stable product...
Quoted 8 lines
The number of -pre releases shouldn't be limited for its own sake, but rather in the process of stabilising the kernel for release. So I mean after
a couple of -pre releases start focussing on debugging and then finish with
a few -rc's before the next cycle starts. That way the diffs between full
versions are smaller and upgrading gets easier.
So, the question is to choose between :
- less releases with more changes
or
- more relaseases with less changes
Is that correct ?
Paul
Quoted 8 lines · Paul Rolland
Hello,
Quoted 3 lines
I'm not saying it should, but it would be good from a PR perspective and as
an element in the reliability feeling vector ;-)Not sure about it... People like it when a product looks stable, and having a -blah or -pre and so on once a week doesn't make me feel I have some stable product...
But that's only because the kernel is in public development, it's not a black box (which is a Good Thing (tm)). You shouldn't need to run a -pre kernel release in 99% of all cases, so having them available shouldn't detract from a feeling of stability (regardless of how often they come)
Quoted 8 lines
Quoted 8 lines
The number of -pre releases shouldn't be limited for its own sake, but rather in the process of stabilising the kernel for release. So I mean after
a couple of -pre releases start focussing on debugging and then finish with
a few -rc's before the next cycle starts. That way the diffs between full
versions are smaller and upgrading gets easier.
Quoted 6 lines
So, the question is to choose between :
- less releases with more changes
or
- more relaseases with less changesIs that correct ?
I guess so.
There's probably not a "right" way to choose between the two, but I'd prefer the latter option.
Cheers
Simon
Quoted 5 lines · Simon Oosthoek
- Alan Cox seems to provide 90% of the patches to the 2.4.x series
Obviously Alan has responsibilities for redhat, so that explains a lot. However I get the feeling that patches for 2.4.x are more and more directed to him rather than Marcello...
Seems to be that way.
Quoted 2 lines
My personal interest in this is that my laptop is not yet working 100%... (see http://margo.student.utwente.nl/simon/ongoing/jade8060.php)
The ipaq guys I think had some kind of driver for generic MMC cards. SD cards are full of DRM and off limits to anyone who doesnt wish to sell their soul