Experimental, to me, means "it might eat your data, have backups" not "we might decide to remove this module from the kernel for non-technical reasons, good luck users"

Reiserfs sat in the kernel for years after he went to prison and didn't get removed on such short notice even though it was equally if not more unmaintained.

> Experimental, to me, means "it might eat your data, have backups" not "we might decide to remove this module from the kernel for non-technical reasons, good luck users"

I don't think the kernel devs share that definition.

> Reiserfs sat in the kernel for years after he went to prison and didn't get removed on such short notice even though it was equally if not more unmaintained.

Reiserfs wasn't marked experimental.

Although there are "experimental" labels on features in the kernel, there's no coherent definition for what that means. Historically most distros have enabled CONFIG_EXPERIMENTAL features by default, until the flag was deemed meaningless and removed. Linux has never removed a mainline-merged feature this quickly. Features in staging may be removed quickly, but bcachefs was merged into mainline. The bcachefs removal, not to mention the removal timeline, is unprecedented for a mainline-merged feature. There's never been any expectation that experimental means it may be removed on short notice.

Going forward, I'll agree with you: mainline does not care about users of experimental features. But it's disingenuous to suggest that this has always been the expectation.

The biggest mistake is not having a staging subtree for filesystems like we do for most other drivers.

Oh, that's interesting. Is there any reason it couldn't have just gone under drivers/staging?

I think it's mostly a policy thing? I brought it up a few years ago and the fs developers were not very enthused about the idea.