need papa linus to save us.
I always take the surface level complaints about anything linux related with a grain of salt, when i looked at btrfs i saw them. But i wanted that snapshot granularity + compression. So i went for it.
There are a lot of uAPI things that are "odd" especially with the whole snapper "make the snapshot subvolume then unmount then make it then remount then delete and don't do it wrong or snapper has a stroke" sorcery. I immediately moved to just doing snapshots manually because no way i was going to even touch something that depends on a naming convention to work.
I really liked btrfs, i was able to get my sophisticated layout, i do a lot of NPU related low level projects so i have a flat layout with seperate subvolumes for ssh, certain compilers like v++ i will build from source and keep isolated so i can have them out of the equation for a different one path resolution wise.
The issue arrived however, when i realized it was just the fact i THOUGHT i liked it because of running btrfs send to my network drive with it as RO, running btrfs receive, then checking that it hit the drive. I mean if you look at a directory and see the file than it's there... one would assume.
Suffice to say, one day i checked on the drive to grab a python wheel i knew i had tucked away and had accidentally delete on my test machine. The snapshots just kind of walked away... Like i thought peter dinklage was going to come to me and say the "the file system just... never cared" or some shit. Wasn't an unmounted subvolume, wasn't a userfacing miss. The data was just gone.
This brings me to the main point of this post, PLEASE LET OVERSTREET BACK INTO THE MAINLINE EMPEROR LINUS. It truly has boiled down to my issues with it, and just the objective state of the actual btrfs team that all roads lead back to the overstreet. Read the front page to his website and you see exactly the type of shit that is overlooked in btrfs development but is just wrong. "Never writes bad data", things that are just outlook and introspective. But the type where there's objectively correct and then there's stupid
Overstreet pushed a filesystem commit for an issue that was seriously important for bcachefs users- because i mean having the kernel doesn't really matter when you have critical filesystem issues that need to be constrained for some autisitc nerds unrelated NPU drivers ( that's me btw, not to demean ). Bcachefs is structurally better than btrfs even as a dkms, but pretty unusable for mainline users.
Regardless of it's usage btrfs is still just beating at a dead horse, i do not see why the technically inferior filesystem is being taken seriously when the only reason it's still the COW standard is some maintainers having their feelings hurt. Like idk how anybody could look at that commit past the merge window and not think "oh that'd be a pretty fair exception"
What is the general consensus on this one? my bad for the long post, just wanted to make it clear i didn't just hastily snapshot things and send as RW with no btrfs receive being ran- or just anything that isnt in the docs.
I say petition to bring/invite Kent Overstreet to come back into the kernel cycle. He's got an ego, he's kind of a douche, but in my opinion even if charles manson was on the prison's kindle E book, rooted it, and was pushing something that showed promise -- id go with it.
just overall curious on what the community thinks about this topic, raid 5 coming before the heat death of the universe!
https://redd.it/1tj9h9e@r_linux
