Re: 2.0.31 : please!

Frohwalt Egerer (froh@twins.iconsult.com)
17 Jul 1997 16:59:28 +0200


jlewis@inorganic5.fdt.net (Jon Lewis) writes:

>
> > > reading this group, by and large, are the people you speak of. But
> > > people who aren't hackers, and who are interested in Linux as an
> > > operating system, don't care why there isn't a stable kernel -- they
> > > just care that there isn't. We've seen how easy it is to be blown off
> >
> > 2.0.30 has been fairly stable for me, admit-ably it has it's quirks, and
> > I've even had apply a patch or two for it, but overall it's remarkably
> > stable. I've got all my machines running on 2.0.30 with uptimes approaching
> > one month now. Most of my downtime has been hardware related. I did need
>
> Repeat after me. There are stable kernels, there are stable kernels, ....

There are. But how should Joe Random User know which one. Oh, and don't ask
for every user on the world to read the kernel mailing list, that is not a
possible solution.

Wo is the bad boy here?

- The developer who does ot have resources to test the kernel on a gazillion
platforms with a gazillon kernel options?

- Joe Random User, who looks at the 'stable' 2.0.x source tree and picks 2.0.30
because he thinks "that is the latest stable kernel"

- The makers of linux distributions, who by evil forces of marketing, are
pressed to incorporated every new feature into their distributions? Who
build distributions with 2.0.30 that break horrbily. (Argh, I'm trying
to install a 2.0.30 based distribution on a 5MB 386. Means I have to build
2.0.29 and modules and put them on the boot disks and and and ...)

A long time ago there was a similar problem, at that time the solution
was splitting the kernel into development and stable version. Maybe it
is time to reconsider the way stable kernels are updated.

I don't know the solution. Maybe new stable kernel revisions should be
dubbed 'beta' versions, the 'beta' being removed when the kernel
proves to be stable in public usage for a few weeks. Maybe never
releasing a new version without a plethora of pre-patches is it. Or
even re-releasing 2.0.29 as 2.0.31 and 2.0.30-pre2 as 2.0.32-pre, thus
removing 2.0.30 from the 'market' solves the problem for now.

Just both sides yelling at another to do something is not a solution. Don't
tell each other to do something, sit togather and think of something
creative. It's a challenge, not a crisis.

(This is aimed at all the participiants in this thread in general)

Froh

P.S.: 2.0.30-pre2 + autofs runs rock solid here, the tulip driver
driving our network cards more stable than 'commercial and
beta-checked' Solaris 2.5.1 does.

-- 
Frohwalt Egerer

Or if you meant a new way to confuse lusers, just ask them their names and they'll inevitably go apoplectic, saying something similar to:

"I don't quite know what you mean... African or European? I don't knooooooooow..." -- Peter Gutmann