[gentoo-amd64] Re: reiser4 fs

[Duncan]

Andrea Chiavelli posted <43B00205.9000903@×××××××.it>, excerpted below, 
on Mon, 26 Dec 2005 15:45:25 +0100:

> I'm going to install gentoo 2005.1-r1 for amd64 on a turion-powered
> notebook and an awful dilemma caught me: should i use reiser4 for root
> partition or not? I've read tons of docs available on the topic but i
> found it a bit contradictory. Do you know anything about the actual
> state of support of the reiser4 fs? Does the default gentoo-kernel
> support it ? have you experienced reiser4 on a amd64? Is it really so
> fast or current software make it buggy and useless?

I'm running reiserfs (reiser3) on everything, here, and have been somewhat
impatiently waiting for reiser4.  It *IS* possible to run it, but neither
Gentoo in general nor Gentoo amd64 "support" it.  That is, you can do it,
but don't expect a lot of help or documentation, and expect a few things
not to work without some additional effort on your part.

Among other things, reiser4 will require grub patches not included in the
Gentoo grub or grub-static ebuild, if you want to run reiser4 on /boot. 

Additionally, it's not in mainline kernel yet, and it'll likely be some
time before it gets there. This is due in large part to Hans Reiser's
lack of people skills.

Unfortunately, while he's incredibly bright, as with many bright people,
he's used to always being right, because few others have the intelligence
to follow what he's talking about.  Thus, when he gets in with a group of
other very bright people that know the subject matter, such as the kernel
hackers on LKML, and treats them like he's used to treating everybody,
like only he has the intelligence to understand the territory, sparks fly.
The result is that he has created a number of enemies on the LKML, folks
that are normally quite helpful, but which he has repeatedly insulted when
they try to explain what's needed, so they don't bother any longer -- they
just continue to vote it down each time it comes up for inclusion, because
it hasn't met the standards yet, and Hans seems more interested in
insulting them when they try to explain things than in fixing the problem,
so it continues...  Like I said, he has a serious lack of people skills. 
Read some of the threads and see if YOU'D still be interested in working
with him, after the insults he spits.  I know *I* wouldn't be.

Additionally, there's the small matter of his behavior on reiserfs
(reiser3), after it was added to the kernel.  As the kernel progressed,
things such as extended attributes were added to most kernel file systems.
Hans Reiser wasn't interested in doing the same for reiserfs.  In fact, he
OPPOSED the work when someone ELSE did it.  His view was that reiserfs is
in stable mode now, and as such, shouldn't get new features, only bug
fixes.  New features belong in the next version -- Reiser4.  That works
great when your product is a standalone product, but when it's part of the
Linux kernel, it's no longer a standalone product.  When that feature was
added to the kernel, it was added to the kernel's file systems (where it
made sense, obviously it doesn't with things like FAT, which are there
primarily for MS compatibility).  Reiser4 was nowhere /near/ ready, so
reiser3, as part of the kernel, needed the update, but he refused to
support the fs he'd contributed, so the work fell to others, and he
opposed them every step of the way.

The result of /that/ fiasco is that the kernel hackers are rather more
demanding about the code meeting normal kernel code standards for
readability and maintainability, because they expect they'll be the ones
left doing the updates, when Hans Reiser is off doing the next big thing,
and actively opposing updates that keep his contribution in step with the
rest of the kernel.

These two factors combined mean that every time reiser4 comes up for
inclusion in the mainline kernel, there remain problems with it, problems
Hans doesn't seem all that interested in fixing, as he's too busy
insulting everyone trying to be helpful and point out what needs changed
in ordered to fit the way the /rest/ of the kernel works.  Fortunately,
a couple of his employees aren't quite so hard headed, and while their
boss is busy creating fireworks and more political resistance, these guys
quietly make a list of the changes needed, do something about them, and
eventually the package gets resubmitted for the next round, some progress
having been made no thanks to Hans himself.  Only this round, the kernel
hackers are more antagonized than they were /last/ round, and the thing
doesn't get much further, because after several rounds, Hans is correctly
calling quite a bit of  what they list (those still willing to work with
him at all and not just turn it down without giving reason and therefore
becoming a target for more of his insults) nit-picking.  Well, that it may
be, as compared to other submitters, but other submitters don't tend to be
as antagonizing, and /do/ tend to be a bit more cooperative, so, what
might be glossed over for them, ends up getting examined with a fine tooth
comb, when it comes to anything Hans Reiser submits.

So much for the political commentary...  Anyway, to my knowledge, the only
Gentoo packaged kernel that includes reiser4 is Andrew Morton's mm series,
which has become what amounts to the development series, so is generally
considered too unstable for most to run -- and certainly not normally
recommended by Gentoo's devs, altho it's there, for those that /choose/ to
run it anyway.

You can of course do your own kernel and grub patching, if desired.  A
number of folks do so, or simply run the mm kernel and don't run reiser4
on /boot, so don't need those patches.

The stability of both reiserfs and reiser4 continues to be called into
question by many.  As I said, I run reiserfs on everything, here, so I
obviously think it's upto the task.  My opinion is that while disaster
recovery in the case of reiserfs can be more difficult and less reliable
than with, say, ext3, the filesystem is in general stable enough that
disaster recovery is rarely needed.  Keep good backups, as you should be
doing anyway, because disasters /do/ happen, and reiserfs works great!

I'm not yet running reiser4, tho I've considered it.  The main reason I'm
not doing so here is only that I don't wish to complicate things, when I'm
running reiserfs on everything, and running a straight mainline kernel
(from kernel.org, not thru the portage tree, and often running the -rcs),
now. If I ran reiser4, I'd definitely have to either do my own kernel
patching, sometimes waiting to upgrade because the reiser4 patch won't
apply to new versions, or switch off of mainline kernel.  Additionally,
I'd either have to decide reiser4 was stable enough to run on
/everything/, or split my system between reiserfs and reiser4.  Further,
if I ran reiser4 on everything, I'd have to figure out what I was going to
do with grub, in terms of patching it for reiser4.

Thus, while I'd like to try reiser4, as long as it isn't in the mainline
kernel, I'm really not likely to do so.  Even then, I'd probably only run
it on part of my system to begin with, likely /tmp and other non-critical
stuff like the portage tree, which can always be redownloaded if it gets
screwed up.  For root, and my backup root-snapshot as well, along with
/home and its snapshot, my media partitions and their backup, and the
like, I'd stay with reiserfs, at least until I was decently comfortable
with reiser4.

So, what that all means is...  you *CAN* run reiser4 if you decide you
want to, but it's not going to be recommended any time soon either here or
by Gentoo, and be prepared for some significant additional hassle if you
/do/ decide to run it.

That said, looking toward the future, when asked about reiser4 support,
the response I've seen from the devs, is that they do plan on supporting
it, when/if it gets added to the mainline kernel.  (It would be when, if
Hans Reiser wasn't as poor a people person as he seems to be, and most do
seem to think it'll be added /eventually/, but given the situation, it's
not the almost sure thing it likely would be otherwise, thus the "if".)

Don't think I'm anti-Hans Reiser, or his filesystems, by any means! 
As I said, I run reiserfs everywhere, and I certainly wouldn't be
doing so if I distrusted him or his solutions.  It's just an unfortunate
fact that he tends to make thing harder for himself than they'd otherwise
be. Such facts... well, it's best to recognize them for what they are,
instead of trying to deny that they exist.

JMO, FWIW...

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html


-- 
gentoo-amd64@g.o mailing list