On 2000-07-05 23:18:10 -0400, Horst von Brand wrote:
> You never noticed it happening all the time, it seems. But kernels from
> different series _are_ different inside, so backporting is _hard_ once they
> have diverged a bit.
If something is a genuine bugfix, it should be backported immediately -
Not 6-12-24 months later with the next kernel series. This does go on
to some extent, but there sure is a lot of stuff in 2.3 that is quite
stable (sometimes more so than 2.2 if the functionality even exists
there) but has to wait months and months before its in a "stable"
kernel simply because other areas of the devel kernel aren't quite to
that stage yet.
With changes like those proposed to the VM subsystem, in my limited
understanding its basically self-contained, hence the interface isn't
likely to change (unlike dcache) and its fairly straightforward to
backport.
> > I realise large structural changes could prevent this in a lot of cases,
> > but it'd be neat if the 2.4 VM was "good enough" and got replaced later
> > down the track with the rearchitected version.
>
> I'd say that is impossible for something that central in the OS.
There's not too many hard concepts in the VM... there's one or two entry
points and the effect the resulting computations have on the system,
with a whole lot of juggling in the middle. My guess is that the new
architecture planned is going to keep the same entry points, just fix
the way the juggling is done so the kernel isn't dropping its balls :)
(or herrings, toasters, flaming torches, steak knives...)
Now, for something like dcache you are right... but that change involved
the interface with various drivers. _that_ makes it non-trivial to
backport. There's bound to be things like this that will consume a lot
of a development cycle to stabilise as some smart cookie comes up with
better ways of doing things that need such far-reaching overhauls. All
I'm saying is that other areas that don't have such baggage shouldn't be
held back by these ones for such extended periods of time.
-- Matt- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Jul 07 2000 - 21:00:18 EST