Bcachefs Hopes To Remove "EXPERIMENTAL" Flag In The Next Year

[intelfx]

ReiserFS?

It was far from a one-man project. And it got dropped primarily because it was outdated (leading to a lack of general interest from users and developers alike), not because it was unmaintained.

[siyia]

Maybe next decade is more probable! :P

[waxhead]

I think it’s more likely to be kicked out the kernel within a year, personally.

Aye, I think that too - but (honestly) I hope it will stay. For the foreseeable future my data stays on btrfs, but at the same time I hope that bcachefs will stay in the kernel and bring both inspiration and competition to other filesystems with it's planned feature set.

I do however fear that bcachefs has done the same mistake as btrfs did e.g. being merged too early. For btrfs that means that fixing and improving things has been a very slow process to avoid breaking stuff or causing performance problems. The early merge also caused a lot of more or less idiotic rumors and "half-truths" regarding its quality. For bcachefs on the other hand it seems Kent is "in a panic" to rush new things to bcachefs before it gets too "grown-in". The same idiotic rumors and half-truths are starting to emerge and I hope Kent does not screw up as that would really put fuel on the fire. I am crossing my fingers, but would not be surprised one bit if this becomes a political shitshow more than it has already. It has already earned it's share of popcorn so it will be very interesting indeed to follow this.

[pWe00Iri3e7Z9lHOX2Qx]

I do however fear that bcachefs has done the same mistake as btrfs did e.g. being merged too early.

The problem is that you get a miniscule fraction of users beating on it as a new(ish) out-of-tree filesystem. There's a ton of stuff being found because far more people are playing with it now that it's in kernel.

I hope it survives, gets more maintainers, and eventually competes with OpenZFS.

[flower]

Good thing no one is saying anything to the contrary, then.

Kent wrote that he hoped to be able to do this within a year, not that he wanted to do it now (which no one disputes is early).

i just like to see erasure coding first.

[JMB9]

Some quotes for getting out of experimental stage may be worth commenting:

"Maybe next decade is more probable! :P"

"I think it’s more likely to be kicked out the kernel within a year, personally."

I understand the tension here - but there are some notable big differences:
Competiveness:
1) bcachefs is unstable as a lot is made robust and tuned ... normal in this stage;
nevertheless was regarded very interesting from the beginning and this notion
has not ceased ...
And I like to read the comments about changes ... and like the direction therein, too.
2) Maybe 2025 is too short - but 2026 is reachable - as a lot of progress is made,
if there is no fundamental design decision which mut be revoked, of cause.
There are testing methods to get reliable data early ... so not needing a decade
when a stable version is reached to get some proof - big companies can create
a lot of expertise very fast.
3) ZFS was developed at Sun in 2001 - it was said to be trusted without an fsck
program, so it was NOT USED professionally for a very long time as I experienced
(largest EU site without using ZFS!);
and it got several years till they recognized that an fsck program is NEEDED.
ZFS is used by Oracle on Solaris, not on Linux AFAIK.
4) Btrfs was introduced by Oracle on 12. June 2007 ... it is ridiculous that it is so slow
and still has problems in corner cases interesting for professional storage.
I was even sceptically in its early years - too much shining - no proof/hint
(RedHat abandoned it quite fast).
5) ext4 is still competitive - and from my point of view the most robust FS available
on Linux systems (reasonable if none of the shining new features is required);
ext was implemented in Linux 0.96c in Apr. 1992, ext2 started 1993, ext3 with
enhanced by journaling feature in Linux 2.4.15 in Nov. 2001, ext4 and with
Linux 2.6.28 no longer named ext4dev but ext4.
Even ext has got ugly bugs - fixed extremely fast - and some ext4 bugs showed
problems in layers other FSs also used - which show both employment and
quality of ext4.
But relying on ext4 even in 2030-ies may be reasonable, still ... the current FSs
would not be able to attract users of ext4 - as its speed is still competitive!
6) Currently professionally favoured are:
- Stratis (Red Hat: Rust + ZFS + LVM)
- Btrfs (OpenSUSE {and Oracle})
- OpenZFS (Canonical/Ubuntu)
Startis is a mixture, Btrfs has its technical problems, OpenZFS is no 1st class citizen
due to license incompatibilities ... so if ext4 has not all features,
a new kid on the block would be very welcome for big employment.

As long as quality is key - which is my impression with bcachefs (especially when
it seems to be no smooth development) - I would not fear it will get kicked out.
Linus is a technical guy and excellent in (nearly) all of his decisions.
Linux is a big project - and it is not easy to be part of it ... so some problems
will occur naturally ...

Bcachefs would be a save #1 Linux FS if it is the most robust FS on Linux
and is more than competive in benchmarks.
Currently, it looks as this could happen ... but too ealy to say this yet.
But for me it is the only FS which could get number 1 form various
perspectives (my is to currently use ext4 - and no other FS has anything
for me to even think of switching ... this may change in future - but
never for the designated kings in the point 6 above ...).

But eveyone has the right to select criteria and make the decision from those.

[skeevy420]

It was far from a one-man project. And it got dropped primarily because it was outdated (leading to a lack of general interest from users and developers alike), not because it was unmaintained.

It might as well have been a one-man project. Not to denigrate other contributors, but once Hans murdered his wife it was basically Edward doing his damnedest to keep the various Reisers working. Pre-murder, Reiser was one of the best ones file systems to use on a desktop. Post-murder, no one wanted to touch it with a 10 foot pole regardless of any of its merits. Reiser killing his wife also killed the file system. ReiserFS/4/5 should have been renamed after the murder happened because it made Reiser become a dirty word and a bad joke in MurderFS.

[skeevy420]

6) Currently professionally favoured are:
- Stratis (Red Hat: Rust + ZFS + LVM)
- Btrfs (OpenSUSE {and Oracle})
- OpenZFS (Canonical/Ubuntu)
Startis is a mixture, Btrfs has its technical problems, OpenZFS is no 1st class citizen
due to license incompatibilities ... so if ext4 has not all features,
a new kid on the block would be very welcome for big employment.

You meant XFS in the Stratis part. Red Hat fucking hates OpenZFS because it makes Stratis look like a bad joke.

Bcachefs still has a lot of work to do to be anywhere near the robustness and usefulness of OpenZFS.

[botipua22]

I am excited to try bcachefs in 2029.

[intelfx]

Red Hat fucking hates OpenZFS because it makes Stratis look like a bad joke.

Bcachefs still has a lot of work to do to be anywhere near the robustness and usefulness of OpenZFS.

Yeah, yeah, we get it that you're simping hard for muh ZFS, but dial it down — not all of us share this level of fanboiness