Re: I vote for updated RAID and KNFSD

Fred Reimer (fwr@ga.prestige.net)
Sat, 11 Sep 1999 10:57:00 -0400


On Wed, 08 Sep 1999, M Carling wrote:
> I agree that having to deal with old patches is bad, but it is a known
> risk in using non-standard patches, even if they are widely used. Because
> the RAID code changed in a way that was not backwards-compatible, a choice
> must be made between breaking a standard kernel feature in a stable kernel
> and leaving people using the currently non-standard (from a kernel
> perspective) feature with the hassle of patch incompatibilities. Both
> options are bad. Getting a stable 2.4 out with the new RAID code solves
> both problems.

The RAID code in the current "official stable kernel" is mostly
non-functional. I don't know a respectable administrator that would
use the existing RAID code in a production system while a very "stable"
newer RAID system is available. In effect, the RAID code in the
current stable kernel is "broken" and needs the bug fix as implemented
in 0.90. That this bug fix requires a backup and restore, which should
not be an issue for someone serious about RAID, is inconsequential. If
someone doesn't want to upgrade to fix the RAID bug, then let them stay
on 2.2.12.

> If Alan and Linus were to go along with my suggestion and accept only bug
> fixes into 2.2, there might be fewer 2.2.x releases to trip you up, and
> each one would have less code changed so the old patches would be more
> likely to work.

You seem to prefer holding back software updates that have been tested
and tested and tested simply to fit your belief in the old model of a
"stable" and "development" kernel. I don't think you realize that
>MOST PEOPLE< do not use the kernel releases that come from Linus.
Instead, they use the releases that their distribution company makes.
It's no longer the "old" way where everyone in the Linux community
would download and compile the latest kernel release as soon as it came
out. Our community has grown to include "others" that don't have the
necessary skill or time in order to test out new "stable" kernels and
rely on the distributions. In effect, there are three levels of kernel
"releases" now, developmental, stable, and distribution. The "stable"
kernel releases are no less stable than they have always been. The new
RAID code is not "developmental."

> > I would agree that perhaps it should be put in as a seperate CONFIG
> > option, and even marked 'experimental' if people are truely concerned,
> > but it needs to get updated with the regular kernel releases.
>
> I think this is a good suggestion, though I'd rather see the effort go
> into getting these features into 2.3, getting 2.3 frozen, then
> stabilized, and released as 2.4.

Why, so Linux won't have a "real" RAID system until sometime next year?
Perhaps you don't realize that the 0.90 code was release a LONG time
ago.

> M Carling

-
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/