Dear All!!
A core kernel developer NEEDS to get these features straightened out properly
in a clean fashion. People WANT reiserfs integrated and working with LVM and
RAID, with good quota support. These are 'Enterpise' level features.
It does not look like that much work is needed to fix the stuff the base kernel
and get better integration of these features. The rough edges on them make us
look like a bunch of amatuers.
Alan, will you look into it as a project???
Best Regards,
Matthew Grant
> > kernel is more broken than beta7 version, a lot of lvm operations in the
> > kernel version DON'T WORK!!!
>
> beta7 is also broken
>
> > What are the problems????
>
> Disgusting stuff like using the low 2bits of an address as flags.
>
> > Reiserfs quota support should be added to the kernel if it is simple also.
>
> its not worth doing. Thats based on the current, racey, broken quota code
> not the -ac quota code we need
>
> > there rather than having a lot of different ongoing backroom operations to add
> > the functionality.
>
> For the base kernel they need to work properly. Vendors may be less picky.
--
===============================================================================
Matthew Grant /\ ^/\^ [email protected] It's/~~~~\Plain where
A Linux Network Guy /~~\^/~~\_/~~~~~\_______/~~~~~~~~~~\____/******\I come from
===============================================================================
>
> A core kernel developer NEEDS to get these features straightened out properly
> in a clean fashion. People WANT reiserfs integrated and working with LVM and
Note _clean_
> It does not look like that much work is needed to fix the stuff the base kernel
> and get better integration of these features. The rough edges on them make us
> look like a bunch of amatuers.
Lots of work needs doing. The entire quota code needs rewriting and that won't
happen until 2.5. The reiser+nfs stuff is probably also 2.5
> Alan, will you look into it as a project???
Not interested. Now is stabilizing time, not stuffing crap into the kernel tree
time. Vendors can (and will) ship stuff which is workable but not clean, and
they will hopefully over time fix up the stuff they care about and submit it.
Alan
Matthew Grant wrote:
> Linus,
>
> Can't we have a controlled development tree branch for 2-3 months just to get
> the ReiserFS/LVM/quota/knfsd integration done _cleanly_ (this has significant
> short-term benefits NOW), with the brakes on all other changes, and another
> development tree where everyone goes wild with NUMA/Networking et al., where
> the changes from the other get merged???
We have -got- to get 2.4 stable and sane first. It's a psychological
thing, but if Linus opens 2.5 now, people will ignore the existing 2.4
problems that need attention.
If LVM et. al. is fundamentally broken in 2.4, then mark it experimental
and support it via external patches. That's the way these things are
always dealt it, when you need a feature set beyond the existing stable
kernel.
Patience, grasshopper :)
Jeff
--
Jeff Garzik | Sam: "Mind if I drive?"
Building 1024 | Max: "Not if you don't mind me clawing at the dash
MandrakeSoft | and shrieking like a cheerleader."