On 2/14/07, Thomas Backlund <[EMAIL PROTECTED]> wrote:
Thierry Vignaud skrev:
> "Gustavo De Nardin (spuk)" <[EMAIL PROTECTED]> writes:
>
>
>>> How to say this in a diplomatic way...the issue is, well,
>>> procedural. Everyone who was voiced an opinion agrees that the
>>> change should be made. However, those who are responsible for
>>> making the change appear to be busy with other things, of whose
>>> nature we know nothing since they never communicate.
>>>
>> FWIW, I mostly disagree with automatic kernel updates, unless,
>> maybe, if they are security-*only* updates. Current Linux kernel
>> development unmodel makes me very wary of changing a working kernel.
>>
>> Of course, a -latest kernel scheme, which one can *choose* to use,
>> is ok..
>>
>
> the current solution used in contrib doesn't automatically update the
> kernel, the old one is still there, ...
>
>
That's a good thing !
(and BTW, neither does current model)
What if the new kernel has a regression, and the old one was removed...
you would end up with an unbootable system...
Of course, over time there will be many kernels installed, and some of
them should be removed, but that is less important that getting fixed
kernels installed IMHO...
If I remember correctly, that was mdk 9.x behaviour.
We should then find a solution to clearly tell users at the end of
installation "You have installed a new kernel. To use it, you need to
reboot your machine. At next startup, if you can't boot anymore for
whatever reason, choose menu entry "2.6.X.Y" : it's your previous
kernel which have been preserved. Feel free to report any regression
in this new kernel to http://qa.mandriva.com";
