[KLUG Members] [Fwd: [NFS] Has anyone run the Connectathon Testsuite recently?]

[Bryan-TheBS-Smith]

Adam Tauno Williams wrote:
> They stink.  2.4.12 is a welcome chunk of seeming solid code.
> Everything from 2.4.4 to 2.4.11 has been a disaster, they should
> be viewed (IMHO) as extensions of the 2.3.x kernel tree.

Oh, I totally agree from the standpoint of the _stock_ Linus kernel.

But RedHat does a heck of a job patching and testing the crap out of
the kernel for months, and SGI seems to take this even one step
further with the VM (out of necessity for XFS compatibility).

> I think it is 2.4.0 that was rushed out the door

Naaahhh, really?  ;-PPP

> in order to provide decent hardware support (USB, etc...) and
> "enterprise" features like LVM, as you simply were not going to
> see these features in 2.2.x.

Actually, some of us had 2.2 kernels patched with these features. 
Yes, kernel-space USB allows things that user-space didn't.  And LVM
is more "integrated" in 2.4.  But LVM still has a lot of work left.

But I do agree with you, plus ReiserFS also went into 2.4.1 because
of this.  But in Linus' defense, SuSE had well over a year of
ReiserFS release time in its distros, and had to go with what they
felt was "production quality".

> 2.4.x had just got to the point where it seemed too long in coming,
> even Open Souce people have to ship eventually or credibility begins
> to erode (especially if the trade rags are focusing on it).

Actually, Linus renigged on his 6 month promise when they started
addressing the scheduler and VM in 2.3.  He then stated it wasn't
going to be months, but _years_.  But that didn't prevent the trade
mags from telling everyone it was 18 months late.

> I've still got servers happily running 2.2.x and I think they will
> continue to do so for some time yet.

I'm still running 2.2.16 with Ext3 0.0.2f in full data journaling
mode.  Nothing but clean filesystem!

> This isn't meant to criticize the kernel developers.  I don't think
> most people realize just how much changed from 2.2.x to 2.4.x.  The
> code is alot cleaner and more modular.

Every Linux kernel release has issues.  2.2 has mega-issues with
kNFSd until 2.2.18 (when Cox finally accepted the Trond+Higgen
patches).  2.4 has its VM nightmare.

Linus really "broke ground" in challenging conventional CS views on
what a monolithic kernel can really do.  It takes time for these new
approaches to "pan out."

> Some VFS issues remain, in part due to the introduction of journalized
> filesystems, not all of which agree on "how" to journal.

Or do quotas, ACLs, etc...  This won't be addressed until 2.5
either.

> And the VM had some admittadly glaring flaws, which it seems to have
> taken them much longer to fix than expected.

The problem with 2.2 was the simple scheduler/VM approach that was
awesome for single-user systems, but could cause servers to
seemingly "stall" for 15-30 seconds on buffer/cache flushes.  I
know, I ran into this all the time.

-- TheBS

-- 
Bryan "TheBS" Smith    mailto:b.j.smith@ieee.org   chat:thebs413
Engineer  AbsoluteValue Systems, Inc.  http://www.linux-wlan.org
President     SmithConcepts, Inc.   http://www.SmithConcepts.com
----------------------------------------------------------------
The US National ID is an enhanced Social Security Number.  It
will give those who abuse it more information than ever before.
And just like the SSN, they will ignore all the regulations.