> Can anyone point me in the direction of a reference (or references)
> to the LFS?  It seems to me that there was a Usenix paper a couple of
> years ago that talked about it, anybody know the specific reference?

The USENIX one is:

"An Implementation of a Log-Structured File System for UNIX"
	Margo Seltzer, Keith Bostic, Marshall Kirk McKusick, Carl Staelin
	In: USENIX Winter 1993 (San Diego) Conference Proceedings

and is available in those proceedings and in:
s2k-ftp.cs.berkeley.edu:pub/personal/margo/usenix.1.93.ps.Z

It describes the implementation of LFS found in 4.4BSD (and therefore
4.4-Lite, and NetBSD), and i believe it deals a bit with the
performance of the beast, too.  It turns out that the performance of
LFS compares unfavorably to that of FFS with clustering, like that
which is found in 4.4BSD (and derivatives) or SunOS.

There is also at least one paper about Sprite LFS (probably more) but
i'm not sure of the citation(s).

> I'd like to argue (in a proposal I'm writing) that the LFS is a fairly
> cool thing for supporting debugging of things like distributed
> databases or networked filesystems.  You'd have to add some funky
> hooks into the LFS to allow a debugger to manipulate the log, but it
> might work.  Anyway, I'm hoping to convince some people to let me try.

LFS itself is not stable and needs a heck of a lot of work before it
can be considered for "reasonable" use.  Additionally, the only
user-land utils for it which have been implemented are newlfs and
dumplfs (not a "dump" for lfs, the equivalent to "dumpfs").




chris