considering kent is probably going to manage to get bcachefs yeeted out of the kernel, i guess i should bother moving my storage back to zfs sometime soon

@quad

something something breaking userspace?

@simon_m i don't exactly keep up with linux mailing list drama, but afaik it's mostly that kent constantly ignores development guidelines and seems to have a severe case of main character syndrome

@quad @simon_m
yeah, basically there's a rule:

only bugfixes after -rc1
features wait for merge window

and Kent notoriously submits features after -rc2 or even later in the release cycle

@wolf480pl @quad

has bcachefs been added to mainline already?

afaik the most sacred rule to Linus is to never break userspace. So I doubt that the feature will be removed entirely.

@simon_m @quad
it has been

but also

Linus has threatened to remove bcachefs from mainline on at least 2 occasions

@wolf480pl @quad

hell is apparently freezing over

@simon_m @quad

more like the arctic is melting

@simon_m @wolf480pl I've been running bcachefs since it first got mainlined in March 2024 lmao, because i do not fear god

@quad @wolf480pl

living on the edge

@wolf480pl @simon_m @quad And "don't break userspace" is public APIs, not filesystems I'm as pretty sure some have been removed.

@lanodan @wolf480pl @quad

good point.

@simon_m @quad
Not just living there, but ruling over it.
You could say he's a lord of it.

@simon_m @wolf480pl @quad it never dropped the experimental label so no one was supposed to be using it for production anyway.

@loke @wolf480pl @quad

Oh I guess then there is nothing stopping Linus to yeet it out of mainline

@simon_m @loke @wolf480pl yeah.

bit sad though. having a multi-device filesystem in the kernel that won't obliterate my data like btrfs was refreshing.

can even interact with it via sysfs like a normal filesystem instead of a bunch of weird and module-specific endpoints

If it leaves, I'll miss it.

@quad @loke @simon_m
is compiling it as an out-of-tree module not an option for you?

@wolf480pl @loke @simon_m of course it is, just a pain in the ass i won't bother doing unless my package manager automates it
:^)

@quad @loke @simon_m
I'd think most distros would have a DKMS package in no time

@quad @loke @simon_m
well ok, half the distros

@wolf480pl @loke @simon_m Who knows, bcachefs is not as new as people thing, it's been in development for ages and /decently/ stable for quite some time. I've occasionally heard of people compiling and using it since like pre-pandemic.

Yet as far as I know, pretty much no distro ever bothered to ship a DKMS module.

I do hope they ship a DKMS if it gets removed from mainline though.

@quad @loke @simon_m
merge window DKMS package street-under-Kent

@wolf480pl @loke @simon_m okay that's a bit of a stretch, the "-under-" is literally the only hint that let me get the joke

@quad @loke @simon_m
ok, I'll let you know when I come up with a better one

@wolf480pl @loke @simon_m (Considering how ass Kent is being about collaborating with kernel development, the project seems to obviously be better off outside the kernel anyways)

@quad @wolf480pl @loke

yea great throw him out of mainline linux so he can annoy multiple distro maintainers at the same time instead

@simon_m @quad @loke
maybe if multiple distros tell him he's wrong he'll get a clue (unlikely)

@wolf480pl @simon_m @loke he seems entirely unwilling to actually internalize any feedback he gets

@wolf480pl @loke @simon_m at least from my understanding, i haven't actually read the mailing lists much, depends on how much phoronix is blowing shit out of proportions

@quad @loke @simon_m
Brodie has been reading the mailing lists and seems to have come to a similar conclusion.

I've read a few threads, and it seems to Kent, making things easy for bcachefs users is a holy end that justifies all means.

It kinda looks like:
- I need this to help my user recover from filesystem corruption
- we have a system, it says you have to wait for the next release
- but your system is stupid, in this case it doesn't have any benefit

@quad @loke @simon_m
I suspect he's so intelligent, that he's rarely in a situation where he can't keep track of all the moving parts, which means he doesn't have appreciation for rules, which are useful exaclty so that you don't need to keep track of all the moving parts

@wolf480pl @quad @simon_m when I was reading some of the threads, I saw him playing a lot about how stable his filesystem is, and on the other hand he's also talking about how he helps his users to recover from filesystem corruption. Those two things should not happen at the same time.

@loke @quad @simon_m
Depends how many users he has, and how many of them run hardware RAID with broken controller's battery, and no UPS

@wolf480pl @loke @simon_m Frankly it's extra stupid when bcachefs doesn't actually corrupt itself much, if at all.

Most reports of bcachefs corruption seem to be from users who added "--force" to some command without knowing what they were doing.

The only common corruption issue I know of that didn't require manual initiation by a user was actually more of systemd's fault.

If your system lost power, systemd would run one fsck per drive or something isolated from each other so they weren't aware of each other running in parallel. Causing corruption.

Even then it was just metadata corruption from my understanding (and I got hit with it before switching to OpenRC). I just had to run one long task after where it re-scanned the btree (took hours though).

@wolf480pl @loke @simon_m I would understand him going "hey this is critical because it keeps blowing up filesystems"

But most of the people he's so intent on fixing seem to have blown up their filesystems by their own actions. My understanding is that this is also why so many are pissed at him, anything that might possibly cause the filesystem to break is a "critical fix" and must make it through immediately, and he keeps sending commits which contain features in addition to the fixes.

He just seems to not understand which fixes are actually important, and he keeps trying to shove in features during the bugfix window.

@quad @simon_m @wolf480pl I am going to be even less apologetic here. I hope I'm not rubbing people the wrong way, at least not too much. 🙂

I'd argue that if there is one software component that doesn't need a 'rock star' maintainer, it's a filesystem.

They're supposed to be boring, and just work. If the rock star needs to do emergency open heart surgery on your data, then something more fundamental is wrong. And if he then goes and brags about his success, he's fundamentally misunderstood his job.

Like security, this is a an area where if your work gets noticed, something went wrong.

@loke @quad @simon_m unfortunately, the standard has been set by btrfs

Unrelated to the particular context this came up, is there a name for this? I've seen this behavior a bunch of times, esp. in IT!

RE: https://mstdn.io/@wolf480pl/115113655884602210

@buherator do you mean people seeing arseholes being arseholes and ascribing them some kind of super-genius status to handwave it away instead of just acknowledging that they’re arseholes?

@itgrrl No, this is not the topic of my question, and I think you are needlessly working yourself up on comments on the Internet.

@buherator y’know, you could’ve just stopped after “question” 🫠