reiserfs reliability

[Matt Graham]

From: Michael Butash 
> Most of my issue with fses is having to layer mdadm, dmcrypt/luks, and 
> lvm still to accomplish "enterprise-y" function.

If you're going to do something complex, then this sort of thing is
unavoidable.  Layers of abstraction all over the place, oh joy.  At least the
way these things work mean that the setup is partially hardware-independent.

> Btrfs and zfs somewhat replace mdadm and lvm, but still absolutely
> require luks wedged in there for encryption on personal systems,
> and both are still somewhat work in progress.  I'd love to replace
> all three with one fs finally with native performance gains.

Encryption will always be a thing you have to wedge in with another layer of
abstraction because there are so many filesystems that don't need it. 
Encrypting /home may make sense.  Encrypting /usr doesn't for most things I
can think of.  Moving softRAID and logical volume functions into a filesystem
also seems like it'd make that filesystem hardware-dependent and/or very
special purpose as well as duplicating already-existing mostly-working code. 
This is *possible*, it just seems like something the kernel FS developers
would proceed very carefully on.  IANAkernel developer though[0], and for my
FS needs, stability >> performance, so I'm using ext3.


[0] I did manage to fix the bad md/tuxonice interaction in Sep. 2011 a couple
of days before the main tuxonice developer did, but his patch was more elegant
and shorter than mine.

-- 
Matt G / Dances With Crows
The Crow202 Blog:  http://crow202.org/wordpress/
There is no Darkness in Eternity/But only Light too dim for us to see