[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Wolverine & LVM



On Fri, 23 Feb 2001, Mike A. Harris wrote:

> I personally consider reiserfs in the "cool hacker toy" category
> for now, as it is damned useful, but NOT ready for primetime
> widespread adoption in enterprise computing environments.
>
> Also, another problem with reiserfs is that reiserfsck is
> horribly broken.  So if someone DOES have a problem with the
> journalling, reiserfsck will more likely make hamburger stew out
> of their data than fix anything.
>
> For these reasons, reiserfs is fine for "experimental" use, but
> definitely not ready for mission critical usage.  Shipping it in
> a widespread Linux distribution as a choice during installation
> would be a major recipe for disaster IMHO, and would not be wise
> at this stage.

 This is all very well Mike but for those of us who *are* testing
ReiserFS already, it's a pain that we can't even do an upgrade on
existing partitions ... it would make sense to at least allow an
upgrade to work, even if it isn't supported as an installation
option.  Perhaps I was wrong to assume this would be the case, but
the support is in the "standard" kernel, experimental or not.

 It would be a shame for Red Hat to be excluded from the range of
distributions that are useful for kernel development and testing
for those of us who were at least sensible enough to leave /boot
on ext2 and use the notails option for everything else :)

> Once the reiserfs codebase has been shaken down a bit, and bugs
> found and fixed, it will mature into a much more robust
> filesystem.  Also, the filesystem tools need severe work to be
> properly useable.  IMHO this is a requirement for this filesystem
> to be used seriously in production environments.  "It works for
> me" is not enough of an argument with which to make such a
> feature available to the wide scale of hardware and software
> combinations out there.

 No, I'm not saying that ... I've been lucky, as a fair number of
folks have had real problems with ReiserFS.  But this choice is
actually preventing me from testing the beta releases adequately.
I realise I'm a drop in the ocean and I have no illusions about
the size of my personal contribution to the QA effort though :)

> As such, the choice is there for those who REALLY want to try it,
> however by doing the extra work necessary to enable it, they take
> responsibility for any problems that may occur using reiserfs.
> Many use it with no problems, but many != all.

 Uh, that's the point ... I made that choice, waited until the code
was in a mainstream kernel release before I tried it, and spent an
interesting night playing Sokoban with the contents of partitions.

> Once a feature has been integrated, that means it must be
> supported.  If such a feature is buggy or carries high risk
> potential, then it could also come with high support costs if and
> when it breaks for a given customer.  It is not wise to include
> high risk code in a kernel meant for such widespread adoption
> IMHO unless that code can pass rigorous testing.

 I would have no problem at all with its only being supported purely
in an upgrade where the filesystems already exist.  I was under the
impression you did something like this with ext3 support ?

> ----------------------------------------------------------------------
>     Mike A. Harris  -  Linux advocate  -  Free Software advocate
>           This message is copyright 2001, all rights reserved.
>   Views expressed are my own, not necessarily shared by my employer.
> ----------------------------------------------------------------------

-- 
Bill Crawford, Unix Systems Developer, GTS Netcom
work: bill ops netcom net uk, home: billc netcomuk co uk




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]