From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hans Reiser Subject: Re: Status of reiserfs in Redhat 2.4.7-10 kernel ? Date: Wed, 16 Apr 2003 19:58:30 +0400 Message-ID: <3E9D7DA6.8040507@namesys.com> References: <3E9D0744.9090907@namesys.com> <1050499043.10791.137.camel@tiny.suse.com> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: list-help: list-unsubscribe: list-post: Errors-To: flx@namesys.com In-Reply-To: <1050499043.10791.137.camel@tiny.suse.com> List-Id: Content-Type: text/plain; charset="us-ascii"; format="flowed" To: Chris Mason Cc: ahorn@deorth.org, reiserfs-list@namesys.com Chris Mason wrote: >On Wed, 2003-04-16 at 03:33, Hans Reiser wrote: > > > >>I am afraid that I can only say that Redhat kernels are not standard >>kernels, and non-standard kernels should not be preferred for use in >>production systems. They are not as well tested (or as debugged) as >>Linus/Marcelo kernels. Linus/Marcelo kernels are the official kernels, >>and Linus and Marcelo, not RedHat or any other distro, are the Linux >>kernel maintainers. I personally try to avoid any use of a non-standard >>kernel in systems I care about because I know that no distro has the >>testing ability of those who test the Marcelo kernels, and Marcelo's >>level of skill and caution exceeds that of any other maintainer I am >>familiar with in detail, but I am more conservative than most. >> >> > >Grin, Hans and I have had this debate before, you might want to read >through the reiserfs list archives for all the details. Just look for >the thread where Hubert Mantel most recently posted. > >The short version is that I would pick a current vendor kernel over a >current vanilla kernel any day for heavy production use. In terms of >performance and stability the vendor kernel will usually do better. The >kinds of QA done on large hardware and heavy workloads is significantly >better before a vendor release than a vanilla release. > >The problem is in defining 'current'. I haven't followed the rhas >releases enough to know if they still qualify. > >Anyway there's no need to start that debate again, neither Hans nor I >have changed our minds ;-) > >-chris > > > > > > Test suites cannot compete with real users. Marcelo gets lots of real users before he releases a final version, and the distros don't. However, I would also add that a conservative person would wait until a few weeks after Marcelo releases, and confirm that it really was stable according to those on the linux kernel mailing list. -- Hans