Bcachefs Linux File-System Sent Out For Review With Exciting Feature Progress

[jacob]

I asked this before a couple years ago, and got a constructive reponse, so time to ask again;

With ANY filesystem, do you need to reformat in order to gain new feature?

My use cases; I could really use some of the new XFS feature's for my media storage, but currently don't have the fund's for more massive storage drives to 'fiddle'. Can I simply update the kernel? Or, my second use case being an older BTRFS array, simply do a scrub (or whatever the integrity checker is called) to get newer/improved feature's?

Depends. Sometimes the filesystem is designed from the start with extensibility in mind so that new features can be added without reformatting. For example the original Ext2 didn't support journaling, H-tree indexing, xattrs, extents, 64 bit addressing etc but people could upgrade all the way to Ext4 and gain those features without reformatting because Remy Card designed it to be upward compatible. BTRFS has room for extensions too and many have been and are being added to it, but of course then a volume that uses them becomes impossible to mount with an older kernel that doesn't support those features.

I'm not that familiar with XFS but AFAIK some of the new features that are being worked on will require reformatting.

[pal666]

It doesn't put an end to the "BTRFS is not read/unreliable/etc" idea.

nobody takes your idiotic ideas seriously

[jacob]

i don't see how it correlates. bcachefs had same amount of momentum when fedora defaulted to xfs
yeah, whole world is conspiring against imbeciles like you

It does correlate because XFS and Bcachefs are different types of filesystems. In some ways and all things being equal, an XFS user who would like to get data CoW, checksumming etc may hypothetically decide whether to move to BTRFS or Bcachefs. On the other hand, if the choice is use a CoW checksumming filesystem right now (BTRFS), or pass on it and instead support us to deliver a CoW checksumming filesystem a couple years from now (Bcachefs), what do you reckon people will tend to do?

[jacob]

nothing would justify migrating to inferior fs, just like nothing would justify funding some empty claims. but some people do it nevertheless. regarding your first point, it don't think fedora default by itself means "now you have to reformat your ext4". but it means that indeed more polish will get into btrfs, its rate of improvement will accelerate and in closer future people will be more likely to reformat ext4

Of course a huge majority of existing Fedora users who already have ext4 partitions with important data on them will just stay with ext4. But when people install Fedora from scratch they will now get btrfs by default. I think the main point is that that will entrench btrfs within the user base and will make alternative projects like bcachefs much less appealing, like there is virtually zero chance that another traditional, extend-based journaling FS would now get traction within the Linux community against Ext4 and XFS. You could say that Ext4 and XFS belong to the same class and offer similar features, but that's a historical accident and an exception to the rule - XFS came as a reliable, full-fledged technology at a time when market share was up for grabbing and had time to gain traction, while Ext4 came later as a smooth upgrade path for everyone else (read: Ext2/3 users) who couldn't or wouldn't move to XFS. So unless bcachefs somehow provides first class, in-place migration from Ext4 and/or XFS, I don't see it becoming much more than another Tux3 or xiafs.

[darkbasic]

This thread became a pal666 monologue

[birdie]

This must be a god damn joke, right? You're excited about an overhyped patch that contains zero code WRT the actual filesystem? I would be worried as hell, if I was expecting this project to ever progress from being a loneman's quest in wonderland.

Linux kernel developers normally reject tons of code thrown at them so he did exactly right. Not sure what you're concerned about.

[pal666]

It does correlate because XFS and Bcachefs are different types of filesystems. In some ways and all things being equal, an XFS user who would like to get data CoW, checksumming etc may hypothetically decide whether to move to BTRFS or Bcachefs.

he can only have such choice when bcachefs exists in real world rather than in advertisements on its website.

On the other hand, if the choice is use a CoW checksumming filesystem right now (BTRFS)

fedora supported btrfs for many years. anyone eager to get cow fs did not have to wait for default change.

, or pass on it and instead support us to deliver a CoW checksumming filesystem a couple years from now (Bcachefs), what do you reckon people will tend to do?

i can't say for all people, most people are not very smart. but i can say that i will not support someone who has nothing but ridiculous claims contradicting reality. it's very stupid to switch to waiting for one man because dozen of people are not fast enough

[gbcox]

This thread became a pal666 monologue

He is just a troll. I added him to the ignore list long ago.

[lyamc]

he can only have such choice when bcachefs exists in real world rather than in advertisements on its website.

???

I’m using it right now. But who knows, maybe I imagined it all.

[Awesomeness]

ok, let's rephrase it "defaulted to ext4 or xfs"

Past tense is wrong. XFS is still default for Fedora Server.