The thing to keep in mind is that 2.0 kernels are the choice of 90+% of
Linux users. A "boring" (highly reliable) system is exactly what most
Linux users are after. If someone is doing a patch that they want to be
immediately usable (or testable!) by the vast majority of Linux
users/enthusiasts it only makes sense to do it against a late-2.0 kernel.
Anyhow, nifty patches like the RAID stuff will certainly end up being
re-worked for 2.1/2.2.
<Side note to all kernel developers: thank you for making Linux great!!>
-- Adam
On Tue, 8 Jul 1997, linux kernel account wrote:
>
> On Tue, 8 Jul 1997, Gadi Oxman wrote:
>
> > The driver is released as a kernel patch against Linux 2.0.30.
> > raidtools-0.3 can be used to create and repair a RAID-1/4/5 set.
>
> I've noticed that most intresting new development is done on 2.0 kernels.
> I assume that this is because developers would like to have a stable base
> for sanity reasons and because they are released less often.. (Hard to hit
> a moving target).. Unfortunatly this makes it more difficult to add these
> patches to 2.1 kernels (exception handling etc.), also, it detracts from
> overall kernel testing with 2.1.x kernels.. (I have a few machines that
> are for testing.. But only one of them do I really have the time to
> pound).. This also encourges new feature to be added to 2.0 kernels (since
> they are available there only) which is bad for stibility.
>
> So, maby it would be a good idea to declare a
> 'Mostly-Stable-Unstable-Kernel'.. People making funkey new patches for
> linux (I.e. genetic fork routienes, soft raid-5) can target for that
> kernel.. Then it's less of a leap to the rest of the development.. Also,
> maby when the MSUK version changes maby some could produce a changelog to
> help developers switch versions..
>
> Would 2.1.36 be a good canidate?
>
>
>
>