Bcachefs Submits Lots Of Fixes For "Extreme Filesystem Damage" With Linux 6.9

[cynic]

You are being sensationalist, I can barely remember any such comments.

What people are saying is the biggest difference between bcachefs and btrfs is that bcachefs has a better design foundation compared to btfrs which should lead to a better filesystem. btrfs design philosophy appeared to be rushed and haphazard, which is why for example RAID 5/6 isn't properly supported because it would require changes to on disk format and even worse they didn't warn users about this when creating a partition up until somewhat recently.

Don't get me wrong, btrfs seems to work great for the usecases that the big companies use it for (i.e. RAID 10) but the developers appeared to both throw untested extra stuff onto it and not test things properly before being submitted into tree (and unlike bcachefs this happened in cases when massive companies where using it, not when it was newly added into the kernel half a year ago)

this would be a reasonable critics to move to btrfs, except this is not what happened.

the big "The COW filesystem for Linux that won't eat your data" is still there, on bcache site.
This is the emblem of how much arrogant the approach of bcache has been so far.

We warned everyone with this mindset that a modern fs is an hell of a project, and that you have to stay humble until you have decades of usage in the real world.

This is the moment when we proved right.

[Quackdoc]

You are being sensationalist, I can barely remember any such comments.

What people are saying is the biggest difference between bcachefs and btrfs is that bcachefs has a better design foundation compared to btfrs which should lead to a better filesystem. btrfs design philosophy appeared to be rushed and haphazard, which is why for example RAID 5/6 isn't properly supported because it would require changes to on disk format and even worse they didn't warn users about this when creating a partition up until somewhat recently.

Don't get me wrong, btrfs seems to work great for the usecases that the big companies use it for (i.e. RAID 10) but the developers appeared to both throw untested extra stuff onto it and not test things properly before being submitted into tree (and unlike bcachefs this happened in cases when massive companies where using it, not when it was newly added into the kernel half a year ago)

sensationalist is putting it lightly, at the very least, me and my circle have already found bcachefs to be more reliable then btrfs.

[Old Grouch]

If Bcachefs was ran on LUKS, is that 2 file systems, one CPU?

LUKS is not a file system. It's an encrypting/decrypting device mapper.

[Old Grouch]

Filesystem's only been around for a few months. Big fat nothing-burger.

Nobody sane should be using a filesystem until it's been around for at least 5 years. If it's >5 years old and still loosing people's data, then maybe it'll be worth criticism.

You mean losing.

Spelling mistakes are coding errors. Try telling a compiler..."but you know what I meant!".

[Britoid]

there are *hundreds* comments on Phoronix bashing btrfs and depiciting bcachefs as the ultimate-absolutely-perfect fs from the gods.

I'd say its probably in the 1000s, usually from those who have never used btrfs or bcachefs, or constantly point out that because of btrfs raid 5/6 problem that its a terrible FS, without realising btrfs have raid1cX which can get you something pretty similar anyway.

[Old Grouch]

I don't care one way or the other until the filesystem is several years old.

I think you mean "I don't care one way or the other until the filesystem has had several years of testing.". Documents do not improve from being left untouched in a drawer for several years. The same is true for filesystems or any other code: age does not improve its quality - it needs to be used and any problems found and remedied. This is what is happening.

[Old Grouch]

DYOR, thanks

Claims require evidence. Those making the claim need to provide the evidence. If you cannot, it makes the claim less credible.

[cynic]

Claims require evidence. Those making the claim need to provide the evidence. If you cannot, it makes the claim less credible.

if you are doubting my claims, feel free to prove me wrong.

[marios]

You mean losing.

Spelling mistakes are coding errors. Try telling a compiler..."but you know what I meant!".

Spelling mistakes are coding errors only in keywords and when you are inconsistent with your spelling. Using loosing instead of losing everywhere will compile just fine, unless you are coding in a language that has either of these words as a keyword.

[mdedetrich]

the big "The COW filesystem for Linux that won't eat your data" is still there, on bcache site.
This is the emblem of how much arrogant the approach of bcache has been so far.

We warned everyone with this mindset that a modern fs is an hell of a project, and that you have to stay humble until you have decades of usage in the real world.

This is the moment when we proved right.

You do realize that technically that statement is still correct? As long as you can recover your data it hasn't been eaten. I mean this is also a basic property of CoW filesystems, something that btrfs has actually broken in the past (albeit sparingly)