Patrick has selected reiserfs to be the default file system for Slackware. If it's good enough for him, it's good enough for me

Every Slackware boot disk since version 8.0 (June 2001) has had support for reiserfs.
That was Linux-2.4.11 (which was released in October 2001). I didn't find out until a week later, and so I was running 2.4.11 for a whole week. No damage occured.
How does 6 years worth of rock solid stability sound? I've been using Reiser here for at least that long. I've not had a single problem with it. Even with the kernel that was supposed to kill it.
Isn't that always the case when moving between different filesystems? Reiser4 is a completely different filesystem to previous versions. Anyhow, I'd prefer to be using something which was designed from the ground up as a journalling filesystem rather than something which has a journal tacked on.

IMO, your concerns are outdated and unfounded.
Maybe, but the 2.4 kernels (which are still the default) support reiserfs out of the box. When Pat decides to bring 2.6 in as the mainstream kernel, I'm sure that he'll include reiser support.

You seem to have taken offence at a simple personal opinion. No-one's casting aspersions on Reiser and more than enough people were backing Reiser to make representation fair. I merely brought up some interesting points that someone might want to consider if they are considering moving their data to an entirely new FS (a decision not to be made lightly) from one that would already be perfectly adequate for the task.

(BTW: What is this fascination with people constantly moving away from known stuff that works fine and is still supported to unknown stuff on their own personal PC's? Why on earth would someone want to run -rc, -mm, pre-release, beta, alpha, software on a machine you use everyday to get stuff done?)

Correct. However every Slackware boot disk since 2001 has also supported ext2/ext3 and for a LOT longer (back to 1993 possibly?)if you go by ext2 support. And for most of that time ext2/3 was the PRIMARY and sometimes only viable choice for a Linux filesystem. Others can read it differently - it's not a competition and there's room for both.

And the other more serious problem I was actually describing is http://www.uwsg.iu.edu/hypermail/lin...06.3/0219.html (On Fri, Jun 24, 2005, under 2.6) which is the one where Reiser3 partitions cannot be used to store images of Reiser3 partitions safely (whoops - that means that it's not safe to store things like PXE boot images, dd backups, virtual machine images, even stuff like source code or documentation that describe Reiser internals? etc). That's a bit more recent and a bit more worrying (although whether it's now fixed I don't know, I haven't checked) and it's also purely an *example*. If nothing else, it's an example that:

is not actually true at all. Let's wiki for a minute:

"Some file operations (including unlink(2)) are not synchronous on ReiserFS, which can cause some subtle breakage in applications relying heavily on file-based locks."

"Early implementations of ReiserFS (prior to that in Linux 2.6.2) were also susceptible to out-of-order write hazards (files being appended to during a crash, for example, would gain a tail of garbage upon next mount). The current journaling implementation in ReiserFS, however, is now on par with that of ext3's "ordered" journaling level."

Ignoring an obvious bias in the article and the chances that the article is outdated, we're talking a FS that's been in development for much longer than ext3 but is still, in some aspects, playing catchup.

A single user does not a world make. I'm not talking specifics. If I wanted to get into that, I can categorically and truthfully state that I have FAT partitions that are nearly 20 years old lying around on their original disks and have never, ever lost any data from the FS (unused, obsolete, non-vital data, I'd like to point out)- it does NOT mean that FAT is a good filesystem design or is stable. It doesn't mean that if tomorrow I crash a PC while accessing those disks that the FS will still be intact.

If the world were to end tomorrow, or I had to ship my PC out to Outer Mongolia and power it by pedal power, and I had to keep the data that was on the PC on there for as long as possible without external support, ext3 would be the format I would use. Other's opinions may vary. My data isn't "critical", the world won't end if I lose it (although I'll be very upset and the taxman will probably not like me very much for having to sift through thousands of printed pages) but for a personal user, you expect a FS to "just work".

I can't afford data recovery services yet have some extremely important (to me) data on my hard disk. If my backups fail, I want to know that I have stuff in a format **I** stand half a chance of fixing myself. That's a concern of my own that Reiser cannot address, others (like the original poster) may or may not share it. If I DO ever need to recover stuff from that ten-year-old tape/disk/whatever in the cupboard, ext3 is much more likely to be easily readable now and in the future (Reiser3 is already being abandoned except for security and critical updates in favour of Reiser4... how long until it's merely there for backward compatibility or even removed from the kernel for not being adequately supported?).

In that instance, even FAT fares as well as Reiser, if not better - the FAT format is well documented, easily readable (can even do plain-text recovery with simple text tools) and there is a vast library of code for many devices that use FAT (from BRAND NEW digital cameras and embedded projects, to the original FAT12 floppy code).

Erm... no... the example I just posited, converting from ext2 (a 13 year old FS) to ext3 (a 5 year old FS) required new userspace tools, new kernel and a single tune2fs command. That's a bit different to all the above AND enough disk/tape space to move all your partitions over, check all files transferred correctly with MD5 or something similar especially if you are talking LOTS of data.

I personally have 250Gb of data across my two main drives (mirrored in countless ways to other machines/media), about 10Gb of which is completely and utterly irreplaceable. 250Gb is nothing by corporate standards but asking me to find another 250Gb somewhere spare inside a machine just to upgrade my filesystem is asking a lot when you don't have money (not everyone can afford a new drive /controller card whenever they run out of space).

There's only about a hundred megs of that data that I would gladly pay someone some money to recover were I to lose all my backups of it. However, current prices run at least £1000/Gb in my area, smallest undertaking 1 Gb. It'd be better for me as an individual to a) be able to do the job myself, b) struggle along without the data or c) recreate the data from scratch. Personally, I'd rather keep optional A open for as long as practical. Some people may share that opinion.

And Reiser3 has now been abandoned by it's creator. How long until Reiser4 has the same problem? I don't like this "let's abandon stuff that works and people are still using" idea.

A crass simplification and a personal opinion stated as if the world should agree. Journalling is not that important for personal use, it really isn't. Crashes that take a filesystem past the point where a simple fsck/scandisk cannot recover are few and far between.

You might lose a few minutes work, you may get some harmless stray FS artifacts but it's very, very rare they can't be fixed simply by running a single util at bootup. Journalling saves you in those few rare instances at the cost of massive amounts of I/O and CPU time all the time you use the computer (in comparison to a non-journalling FS).

The original question was (I assume) concerning personal use (talk of mp3's and such). A journalling filesystem is convenient there if only because a dirty FS is immediately obvious at boot (unfinished transactions in the log etc.) and thus a scan every X days/mounts is unnecessary. It's hardly vital for such a purpose. The user was already on ext3 which is PERFECTLY adequate for personal use in it's default configuration (for perfect journalling, for a "serious" system, you'd want to enable the full ext3 options that would utterly KILL performance and there Reiser would probably win).

This is the problem - if ext3 is perfectly adequate and you're NOT noticing any performance barriers (on a desktop system, benchmarks mean little whereas actually NOTICING without measurement that file copies take longer than you think they should is a differnet matter), upgrading to Reiser is not going to bring many improvements, but merely a few possible problems (depending on how important you see your data etc.)

A filesystem to me is a LOT more than just journalling. It's a precious safe into which I place all my critical data every single day and expect it to keep it safe transparently. Tradeoffs appear at all points because I can't afford a 20-disk RAID6 array with daily tape backup. Those tradeoffs to me edge away from maximum performance and towards keeping that data of mine intact and recoverable on a filesystem I know. Others may disagree.

To me, Reiser wins on small file handling, loses on data transparency, wins on journalling (by a close margin) but loses on lack of forward/backward compatibility, wins on technique but ultimately loses out to what I already know, use, trust, understand and can manually repair. There's so much else to take account of other than just a good journalling design.

***I'd*** much rather use a filesystem that's actually DESIGNED with Linux in mind from the get-go, can do journalling just as well for most tasks, is based on an even older and proven FS AND doesn't require enormous upheaval everytime someone decides to update the format a bit. ext3, for example, has incorporated larger limits for filesizes, number of files, FS size etc. constantly over the course of its history and yet, doesn't break compatibility, even with ext2 on which it's based.

My desktop is where I work and play - at work, someone can happily pay me to sit for an hour watching a status bar notch up one bit at a time - at home I cannot be arsed to spend literally hours/days/weeks of my time constantly upgrading major critical parts of my computer design for no good reason. What you use your computer for greatly determines how you make it work.

Unfounded - possibly some minor details that anyone can confirm for themselves if they are interested. Outdated - possibly very slightly (I don't use Reiser as a matter of personal choice so I don't bother keeping track of the very latest stuff except what I pick up from reading mailing lists, announcements, the occasional commit etc.), but I never wanted to enter into an argument about technicalities. Opinions were sought, mine was given. If you believe it casts Reiser in a bad light, so be it.

Some people have different priorities other than "is it newer", "does it do cool stuff". I call it Windows Upgrade Syndrome. Windows 95 was perrfectly adequate for myself for years (as well as about 90% of the people I work with/for). Upgrading to 2000/XP/etc. didn't give them any perceivable benefits at enormous cost (money, time, frustration). Also, how many people spent WEEKS of their computer time waiting for defrag's to finish and how much time did that actually save them on their file accesses? I know at least a dozen people who can't use a flash drive because it's too complex and yet they will sit and wait very impatiently for their computers to defrag every monday morning because they think it's saving them time because "someone told them it would".

Ext3 vs Reiser in a personal use scenario seems, to me, to be much the same thing.

Ext3 works, it journals enough and it's as well-established as Reiser, if not more. Some people may not want to bother "upgrading" if there's little or no return on their investment of time. I don't and this person may not want to.

i used to use ext3 before (when i got started with linux), but i switched to reiserfs a few years ago and never looked back... it feels zippier and i don't have to wait so long for an fsck upon startup after a power failure or whatever... i won't pretend to have any deep technical knowledge of filesystems and stuff, but i do understand that all of the top filesystems each have their advantages and disadvantages... that is why there are several, because not everyone's needs and likes are the same... personally, i love reiserfs and i have found that it works best for me...

I also used ext3 and had to drop it and switched to reiserfs a few years ago and ... Never looked back... for the same reason win32sux carried out

No offence taken here mate. Its all healthy debate. Nothing wrong with that if you ask me.
Some people like to live on the edge. Usually for some new feature. For me, the biggest incentive to use reiserfs was not having to wait half an hour for an fsck every 10th reboot.
I was referring to my own situation. ReiserFS has not let me down in 6 years' use. And I'm talking about a home PC here, nothing mission critical. Admittedly, I don't keep images of reiserfs partitions. All my backups are on DVD. That said, I have several iso files on my reiserfs partitions with no ill effect.
That's a lot of porn.
Wouldn't it be wise to put that on DVD? For the price of 2 discs (about 40 cents here) you can have peace of mind.

Sorry, I didn't read the rest of your very verbose post. You're happy with ext3. Great. Keep using it. I know what works for me, and have no hesitation recommending it. If that offends you, I'm sorry. This is just a debate, so don't take it personally.

Edit: While we're at it, why limit this discussion to these 2? What about JFS or XFS? They're both supported by the Linux kernel, and have been for at least as long as reiserfs...

If we are going into that territory, then I would put XFS above Reiser.

Granted Reiser has better performance with smaller files (which is what a lot of people probably want, documents, MP3s, etc), XFS is really great for large files. I have a 300 GB file server populated with mainly movies and ISOs, and XFS made a very noticeable improvement over Reiser on it.

How does XFS and ReiserFS (3,4) stands when there's a power failure? With Ext3 I screwed up my system once, since it couldn't repair it. I read that ReiserFS could have problems with recovery from power failures. Is it true? No benchmark take this in consideration.

for me (with reiserfs) it's almost unnoticable... the startup check after a power failure takes about one second on my box... two seconds if there were a lot of transactions happening when the power failed... i've never had a single power-failure related problem with reiserfs...

i don't remember exactly how ext3 was but i do remember it would take an extremely long time to do whatever it is that it does after a power failure... maybe it's different nowadays...

Yes, it does take a long time for ext3 to check, and when it found a problem, then you have to reboot or manually fix the problem. Sometimes (1 time for me) the problem was unfixable.

That's the benefit of a journalling file system. Recovery takes seconds, not minutes or hours.

as far as i rember 2.6 supports reiserfs out of the box.

i personally prefer reiser, it seems faster than ext3 - and the benchmarks on reiser's web site show it is.

In my experience, neither have every had any errors that weren't automatically fixed after a power outage or system lock.

I always go for ReiserFS, just because ext3 checks itself every twenty-something mounts (24 mounts if I remember properly); I just freakin' hate that

Anyway, ReiserFS has always been kind to me, no surpresive corruptions or any other oddities

just my

Hello,

Yes, that is no laughing matter. It is every 24 mounts. It decided to check my ext3 filesystem and the whole filesystem got destroyed. And just when I got everything just the way I liked it. I had to reinstall the whole system.