Your Quarterly Linus Torvalds Linux Rant

Every so often it's fun to check in on the Linux kernel to see what Linus Torvalds have got himself into, and a few releases back there was the merger of Bachefs but now there is a bit of drama surrounding it.
==========Support The Channel==========
► Patreon: brodierobertso...
► Paypal: brodierobertso...
► Liberapay: brodierobertso...
► Amazon USA: brodierobertso...
==========Resources==========
Rant Email: lore.kernel.or...
Phoronix 1: www.phoronix.c...
Phoronix 2: www.phoronix.c...
Phoronix 3: lkml.iu.edu/hy...
Status Update: lkml.org/lkml/...
Phoronix 4: www.phoronix.c...
Phoronix 5: www.phoronix.c...
Phoronix 6: www.phoronix.c...
Phoronix 7: www.phoronix.c...
Phoronix 8: www.phoronix.c...
Phoronix 9: www.phoronix.c...
Phoronix 10: www.phoronix.c...
Phoronix 11: www.phoronix.c...
Phoronix 12: www.phoronix.c...
Phoronix 13: www.phoronix.c...
Phoronix 14: www.phoronix.c...
Phoronix 15: www.phoronix.c...
Phoronix 16: www.phoronix.c...
Phoronix 17: www.phoronix.c...
=========Video Platforms==========
🎥 Odysee: brodierobertso...
🎥 Podcast: techovertea.xy...
🎮 Gaming: brodierobertso...
==========Social Media==========
🎤 Discord: brodierobertso...
🐦 Twitter: brodierobertso...
🌐 Mastodon: brodierobertso...
🖥️ GitHub: brodierobertso...
==========Credits==========
🎨 Channel Art:
Profile Picture:
/ supercozman_draws
🎵 Ending music
Track: Debris & Jonth - Game Time [NCS Release]
Music provided by NoCopyrightSounds.
Watch: • Debris & Jonth - Game ...
Free Download / Stream: ncs.io/GameTime
#Linux #OpenSource #FOSS #FileSystem
DISCLOSURE: Wherever possible I use referral links, which means if you click one of the links in this video or description and make a purchase I may receive a small commission or other compensation.

[@xymaryai8283]

Linus never challenged Kent's development skill. he was doubting his collaborative skill. and Linus was right.

[@TheCorruptedHuman]

Exactly why I just made the comment that I did. If you're a solo developer and want to contribute to the kernel, find somebody who has done it before, because there are a lot of hurdles, and more are added every day.

[@tomaszgiba]

I scrolled to comments section to post the same comment. I totally agree.

[@ImNickolicious]

I'm not gonna lie, when Linus steps in he makes his point clear and concise

[@mattkeith530]

And he does it without insulting these days so it is productive

[@matiasm.3124]

And 99% of the time is totally right.. that's why linux still works as a community.. no forks etc.

[@thesneakysnek]

@@matiasm.3124 Honestly, I think that no one has forked Linux yet because it's just too much work. Dealing with upstream can be a nightmare if you wanna submit patches, but it's better to deal with them than to run your own fork that no one will use. Linus has made a few bad decisions regarding how the kernel handles drivers in particular but for the most part, he was able to hold down this project very well. Kudos to Linus!

[@ChaosTheory666]

@@mattkeith530 People who get hung up on insults even though they're merely the spice of a well-formed argument are inherently unproductive, so insults are a good way to weed out these inferiors.

[@GreedoShot]

​@@ChaosTheory666you're an absolute moron

[@stayblueee]

Both positions are understandable, but to be honest, I think Linus Torvalds prevails victorious here. The kernel has existed for decades and these standards have helped it run like a well-oiled machine, and Linus's point that people who are currently using Bcachefs are guinea pigs makes enough sense. Kent's argument to undo the way of things isn't enough to convince me personally.

[@qwesx]

Especially since there's no real reason for him to do this, other than laziness or, even worse, messy software development practices. I suspect both, because that's the only reason I can imagine why the pull requests can't be split up into incremental small patch sets. And if "messy software development practices" are the cause for this, then the grandest set of automated software tests and an army guinea pigs is not enough to write good software - they can only find the smelliest parts of the bad software.

[@andersjjensen]

Only one position is understandable. There are thousands of contributors, and if the release cycle rules are allowed to slide for one others expect the same. And with git it is so screamingly easy to have separate branches. It was really cumbersome with CVS and SVN, so back then sometimes things had to be more on a case-by-case basis. But with git the "If your bug was there in the previous release, don't fix it in the RC window"-rule is trivial to abide by.

[@orbatos]

No they aren't. There's one position, follow procedure. If you have some issue with that, open that discussion. This is just being an ass.

[@Novacification]

@qwesx Yeah, not writing clear explanations about the changes is crazy. I hate when I get anemic PRs/commits and I don't work with a kernel used by millions of users and billions of systems. I'm sure he's a very good developer but not explaining his changes and making incremental commits (not just splitting up a huge PR into 12 patches) after being told to multiple times really is unacceptable.

[@markhaus]

Yeah I’m sorry but as someone who still suffers from the same bad habit of large software patches it is not a good engineering practice. For too many reasons to list. File systems are hard, period. But I haven’t encountered a situation where more controlled releases leads to better design overall. I have to think taking Linus’s advice here is actually better for both parties involved

[@GSBarlev]

"It's a bugfix! Why would you ever withhold a bugfix?" is the type of attitude you see from the kind of person who pushes to prod on a Friday.

[@aeghohloechu5022]

crowdstrike moment

[@ai-spacedestructor]

you push to production on friday if you hate your colleagues or are masochistic in some kind of way but to be fair it can work, if you can guarantee that you can give it the attention it needs.

[@marcogenovesi8570]

@@ai-spacedestructor No it cannot, you don't push nothing on friday unless the servers are literally on fire

[@ai-spacedestructor]

@@marcogenovesi8570 except your obviously wrong because people do it all the time and only 10% of the time it goes wrong. there is nothing particularly wrong with pushing on friday if your competent and dont set the entire world economy on fire.

[@Person1873]

And this exactly the kind of "it won't happen to me" mentality that Linus was trying to stamp out, the very mentality that caused the cloudstrike outage

[@the_kovic]

Linus is absolutely correct. You don't get to push a thousand changes this late in the release cycle because you *feel* it's solid.

[@federicomedinauy]

Exactly. This Kent guy just didn’t want to listen to what he was being told, always replying with something that was not the point at all.

[@tarjeidavidsen7611]

And you especially don't randomly fuck with other peoples code just to make yours work because you couldn't make yours work with other peoples code, that's the exact wrong order to do the work in

[@NotAFanMan88]

Linus' gatekeeping is what keeps the kernel alive. You can't fuck around with the heart of all Linux-based OSes.

[@mcdermottpa]

Literally billions of dollars are at steak, and the bcachefs dev seems completely unconcerned. It's honestly baffling. The guy should be focused on convincing people he, and by extension his filesystem, would be a reliable choice in large scale deployment. Instead he's having fights because he wants to ignore the processes and procedures everyone relies on to ensure stable and predictable releases. I don't know why anyone in a major distribution or organization would trust him at this point.

[@ai-spacedestructor]

@@mcdermottpa he will have a lot of work ahead to gain trust after that behavior for sure, even if he sees why and changes his way of doing things.

[@generallyunimportant]

@@mcdermottpa billions of dollars on a steak? this is worse than gold dust on food ;-;

[@widget5963]

@@mcdermottpa The "I trust me, so why don't you?" type of person that can't leave their own head to get a proper sense of what others are thinking.

[@LieseFury]

people in TLE's comments pointed out that this is how the xz exploit was smuggled in. Linus is right to not want to ship this when the dev is acting very fishy about the release cycle.

[@thingsiplay]

It's always good to show context and development before the point of interest. This way, it's totally understandable why Linus is reacting like that, instead just showing his reaction.

[@hellowill]

I am just worried WTF will happen when Linus is gone.

[@ChrispyNut]

There will be others who'll be able to take up the role themselves, how long it'll take for them to get to that role and earn the respect of the appropriate people, is another matter, whilst the damage that occurs in that time-frame is the scary part.

[@ai-spacedestructor]

people immediately will try to throw all the rules out of the window for whatever reasons that may be and just see what actually happens if you do the things that currently are getting blocked.

[@chri-k]

⁠@@ChrispyNutideally linus should take care of this beforehand

[@ChrispyNut]

@@chri-k No. Linux should take care of this beforehand. I know some work has been done, but I'm not nearly deep enough into that to know how solid that is.

[@sprinklednights]

BSD.

Bcache developer's position basically amounts to "nah I'd win"

[@MadsterV]

he's just a single big mistake behind an attitude adjustment, and you can't have a big mistake until you're contributing to a big project.

[@WMan37]

I don't know why he feels like it's a good idea to have to overlook thousands of lines of code every update. If every developer did this, we'd have an xz backdoor/crowdstrike situation every month. I'm with Linus on this one. Bcachefs seems pretty cool, I'm not poopooing the project, but like, big merges = big bugfixing, kent might be confident because "It works on my machine™" but _that's not how stuff works._

[@dvhh]

There is already the documented attempt to introduce faulty patches from the University of Minnesota

[@user-yw1nm4je8o]

@@dvhhwdym

[@user-yw1nm4je8o]

@@dvhhlike intentionally?

[@dvhh]

@@user-yw1nm4je8o yes there was a paper written "On the Feasibility of Stealthily Introducing Vulnerabilities in Open-Source Software via Hypocrite Commits"

[@Dipj01]

I think he's compromised and actually intends to plant exploits like this. His behaviour is very suspicious. He continues to repeat this despite warnings.

[@simo47768]

Respect for both Being such a good Linux developer is awesome. Being able to tell enough is enough is also awesome

[@OhhCrapGuy]

You know the biggest issue with the developer? It seems like he fundamentally doesn't understand what a major.minor.patch version system is. He's been toxic at Debian maintainers because he thinks Debian Stable should be shipping the absolute newest version of bcachefs-tools. That's antithetical to the entire point of a point release distro, and that's been explained to him before, and he simply ignores everyone who explains that to him and keeps fighting with the definition of a Stable Release.

[@YaySyu]

I like thinking of debian sid as like a rolling release version of deb. Dudes tripping, I would be stoked to have my packages in the repo.

[@NJ-wb1cz]

Yep, but so what. His relationship with reality is his deal. We aren't his teachers or parents, we don't need to fix him. You somehow fixing his drives to be less of a primadonna would probably make him realize he doesn't need any of this, doesn't need to develop any file systems for free, and can instead spend time with his family and friends.

[@jongeduard]

I know it from you, because you posted this same comment on SavvyNik's video. No problem though. :P

[@bulletflight]

​@@NJ-wb1czIf he doesn't want to follow the standards, he can leave. The best coders are not as important as the best integrators of other people's code.

[@te-wei]

​@@NJ-wb1czMaybe he should.

[@darukutsu]

missed oportunity for outro: follow the kernel rulez

[@Krmpfpks]

If you ask Linus to merge something he will read it and make a risk assessment. If your changes are contained to some subsystem that is considered experimental (like bcachefs), the risk is quickly evaluated to be low. If the changes are to multiple different areas then Linus will have to review very carefully, which is hard work, or move it to the beginning of next release cycle where there will be much more testing by a lot more hands. Kent is basically saying ‚trust me bro‘ and he probably is one of the few devs good enough to make Linus even think about it. Linus let him get through with this earlier essentially valuing Kent’s fixes above his comfort level. But in his answers Kent is not considering the burden he is putting on Linus alone to review it this late in the cycle. I am sure a younger Linus would have written a response to Kent’s last mail with a few choice words. Maybe current Linus has written it too but decided to not send it 😂

[@knghtbrd]

I want to be a fan of bcachefs. But Kent is his own worst enemy when it comes to making this filesystem successful.

[@mcdermottpa]

As a happy btrfs user of many years, this has cleared away any curiosity I had about using bcachefs any time soon. While some things have been slow in coming with btrfs development, I have confidence in the development process and have few concerns about updates. Everything discussed here, and elsewhere about the interactions between bachefs's dev and the kernel devs, and people at distributions like debian makes me weary. The stability of the codebase is irrelevant if the stability of the project and future development is in question. That's especially true for projects like this one that have not seen widespread adoption.

[@PredatoryQQmber]

Could be worse, like being one of FreeDesktopOrg or Khronos guys who get away with anything thanks to their corpo overlords...

[@orbatos]

@@mcdermottpa I use an SSD bcache in conjunction with btrfs to tier my storage array and soak iops, it works great and had been a stable feature for years. As much as I want to use bcachefs, btrfs is usable now and provides largely the same feature set.

[@mcdermottpa]

@@orbatos Yep. I use BTRFS a lot, and even have used RAID6 for years in my home lab without serious issues. I've replaced multiple drives, had two drive failure events, and havre had no data loss. BTRFS in RAID6 actually has pretty good performance across a 19 drive array. I'm not relying on it for anything critical, and keep multiple backups of course. If they can get the last few issues resolved, I honestly think BTRFS would see much wider adoption.

[@woobilicious.]

I started using bcachefs on a patched v5.9 (and only lost data once due to a bug in untested master), I'm still using the same file system atm, I started writing some patches due to timezone issues and getting in contact with Kent, He's a very competent developer but is an abrasive asshole, hang out in the IRC channel and You'll eventually see him yelling at noobs for not knowing how gdb works or asking "dumb questions". I no longer bother working on the file system because of his inability to "work well with others". So not surprised Kent he's self sabotaged the project lmao.

[@Winnetou17]

While I'm not wishing to work with someone like that, if being abrasive is his only flaw, I'm totally fine with that. Growing a thicker skin is also a very valuable skill in life in general.

[@uncrunch398]

What's the IRC channel for? It should restrict conversations to that purpose; have a mod for it if people won't comply. Busy people get frustrated when asked things that are somewhat easy to find the answer to without bothering anyone, unless answering such questions is what they're there for.

[@NJ-wb1cz]

@@uncrunch398 I think they are often used as some kind of worse Discord clone

[@woobilicious.]

@@uncrunch398 The thing is he'll deliberately try and help but then get frustrated when he decides after the fact that it was a waste of time, He could spend 30mins adding a page to his website describing how to do common debug methods (linux fs kernel debugging isn't exactly something you can easily google for), but instead just repeats the same mistakes over and over and yells at others due to his own inability to preemptively manage that frustration. I had patches waiting in review for like months, that I had been pestering him all the previous week to review and he got pissed at me that I hadn't fixed them within 24hours after doing a Friday review saying I wasn't respecting his time, I called him out on it saying he was a huge hypocrite and he blocked me, kicked me from the IRC channel, and I was the only other dev at the time working on the project, so yeah ofc his IRC channel is run like shit lol.

[@woobilicious.]

@@Winnetou17 I didn't say he was an asshole because he was abrasive, I also didn't say "I left" because of him being abrasive, I left because he's an asshole, look at the way he response to Torvalds telling him to stop trying to get special treatment, there's no "Sorry", at somepoint it's malicious derespect for the procedure and Torvalds authority, It's not about making good justification of why that specific 1000 lines of code should be merged in an RC release despite it violating policy, it's because his fragile ego is hurt and he has to defend himself in some vague sweeping statements about his commitment to stability.

[@thingsiplay]

18:38 That's how Clownstrike happened.

[@gljames24]

true

[@dunk7073]

indeed

[@thephoenix215-po2it]

Well someone predicted this video back on the Linux Experiment Channel..... lo and behold it came true!

[@muizzsiddique]

:D

[@mritunjaymusale]

Can't wait for the rust haters video as well

[@Dipj01]

It seems like he's intentionally doing this to insert some malicious code into the kernel. A big patch with a tiny exploit hidden in it is much more difficult to spot by others than if he had put it in a small patch. When he keeps trying to push these huge changes into the stable environments (which comprises majority of users) despite repeated warnings, it starts to look intentional and malicious.

[@CattoRayTube]

I'm not trusting my data to a filesystem which is developed by someone who refuses to follow a very clear, basic, well-established testing process. Bit of a Kent about it, too.

[@muizzsiddique]

@@CattoRayTube Bravo

[@dashcharger24]

Today I learned you need to say ButterFS. I always said BetterFS. Not because I thought it's better, just thought this was the actual full name?

[@kuhluhOG]

there are multiple valid ways of pronouncing it

[@exotericidymnic3530]

btrfs stands for B-tree filesystem, because it uses B-trees, but people call it butterfs because it's funny.

[@Nina-cd2eh]

I've always said it as either each letter separately "BTRFS", or just "B Tree FS". First time I'm hearing someone call it "butterFS"

[@SirWolf2018]

I am missing "Zoo FS" from the list of puns. :)

[@NJ-wb1cz]

@@Nina-cd2eh some people think it's better than others, others think it's butter than others

[@Tb0n3]

I can't not read BCA Chefs whenever I see this thing.

[@Linuxdirk]

I am not alone! 😂

[@lula4260]

Cook Book Available

[@GSBarlev]

Having never heard of this project before, I legitimately thought it was related to Chef, the DevOps tool.

[@angeldude101]

Pronounced like "Baka Chefs"

[@iatheman]

One of us!

[@midplanewanderer9507]

This was fascinating. How the sausage gets made. I don't envy meatpackers or kernel maintainers...

[@DryPaperHammerBro]

...I agree with Mr. Torvalds on this, he shouldn't have added bcachefs

[@a_maxed_out_handle_of_30_chars]

was waiting for your take on this

[@maezrfulldive2770]

Best vid in a while, 10/10

[@Bvngee]

I don’t understand why he can’t just… wait?

[@necuz]

Right? There's literally zero people relying on bcachefs working.

[@RustIsWinning]

Because he wrote 1000 lines bugfree + had automated tests. Just believe it's stable 😂

[@jello3456543]

He's got tunnel vision. He's so focused on bcachefs that he's ignoring the rest of the ecosystem.

[@DaVince21]

Ooo, story time! There might be drama involved in them, but I always appreciate you going through the entire history to find the highlights and create something easy to follow along with.

[@enkaskal]

DJ Ware has bcachefs included in his latest benchmarks and it was not impressive. The more interesting thing to me is the connection between this, the maintainer of the Rust for Linux project leaving, and the COSMIC desktop. EDIT: capitalization

[@dvhh]

And the bcachefs-tools package being orphaned in debian

[@LTPottenger]

Makes you wonder if there is some plot here. Like why is it so vital everyone take your code in the kernel itself?

[@marcogenovesi8570]

@@LTPottenger because it's a badge of honor, and because it's how Linux works in general. If you stay out of tree you need to maintain your drivers every time the kernel changes the driver interface (which happens often)

[@CandyCaneChris]

The algorithm really wanted me to watch this. I've scrolled past it all week saving it for my Friday shift, but it keeps showing up under unrelated videos.

[@nezu_cc]

I would love to love BTRFS but I have my own horror stories with it (yes, multiple). ZFS is nice but not really suitable outside of storage servers built with ZFS in mind. I kinda get the guy, he's a "modern" developer. "Move fast, break things" is also my mentality but unfortunately, it is not compatible with the pragmatic approach of the Linux kernel. If ZFS can succeed living out-of-tree then maybe bcachefs also can. I wouldn't mind the little inconvenience in exchange for a development speed 10x that of most other filesystems. I feel like he'll waste a lot of time trying to upstream it that could have been spent making it better.

[@formbi]

ZFS can live out of tree because it's from BSDs

[@marcogenovesi8570]

Not sure what "storage servers built with ZFS in mind" even means. A raspberry 5 is more powerful and has more ram than the original hardware ZFS was designed to run on

[@lazyh0rse]

I have a weak 15 year old pc been running openmediavault with zfs rocksolid with 4 gigs of memory. You don't need a server pc to run zfs.

[@Mankepanke]

​@@marcogenovesi8570 would you say ZFS is suitable for a laptop or phone?

[@marcogenovesi8570]

@@Mankepanke people have been using zfs on laptops and NASes using laptop CPUs for many years, like for example the HP Microserver Gen7. Modern phones can theoretically run it fine, (see what I said about people running ZFS on raspberry 4 and 5). The "storage servers built with ZFS in mind" of the old times, are orders of magnitude less powerful than a laptop or a phone nowadays.

[@orbatos]

I like bcachefs, and have been using bvache itself for several years. That said, Linus gave a too much leniency with these people and it's had effects. bcachefs-tools are being developed even more haphardly to the point that the Debian maintainer has thrown in the towel. They are simply unwilling to consider the people they work with or effects on end users.

[@collectorguy3919]

Developers tend to get so close to their work they lose sight of many risks. It's as though your mind is secretly trading away wisdom to stay sane enough to finish, while gaining over confidence in the quality. It gets worse when you actually do quality work. Most of the problems are just outside of your imagination, but others can see them. It's such a mind-fuck that it's hard to believe that problems others describe are even plausible.

[@veronikawinters6435]

the music at the end is boppin, I put it on my speaker and was dancing around

[@NFvidoJagg2]

DJware has been using bcachefs for his filesystem showdowns. It still has a long way to go performance wise

[@n5ifi]

Linus is right as usual. Taking care of business.

[@smorrow]

0:42 Even if they were a full copy, if it's to the same drive it isn't "a backup", it's a copy

[@liquidsnake6879]

This is crazy 32 patches lol that would be rejected immediately without even looking at my company 😂 which is good because in others it might be approved without looking in either case nobody has time to read all that crap. Also agree with Linus, from my experience most bugs come from bug fixes that overlook dependencies it's painfully common. You'd fix this by making Kent part of the reviewing team on the other side of this so he feels the pain himself lol that's how i learned at least

[@saucepewpewpew1147]

is this a Honey B. Lovely reference ???

[@Tobi-ci3ns]

I can't fathom why a new filesystem would ever need to change code in other parts of the kernel. If the existing kernel structure works fine for btrfs, it's unlikely there's anything that would need changing to support bcachefs, and on that off-chance there really is something that absolutely has to change, the developer ought to be able to articulate to other people what that is and why.

[@formbi]

BTRFS doesn't always work fine, it destroys data on RAID 5/6

[@ALinuxEnthusiast]

@@formbi Because they have yet to figure out how to make RAID 5/6 not destroy data without adding a journal or killing performance until recently.. And still btrfs works fine outside of that one specific use case.

[@somenameidk5278]

@@formbi That's why they label it as an experimental feature and tell you not to use it outside of testing

[@marcogenovesi8570]

@@formbi RAID 5/6 is a specific feature marked as experimental. The code/logic used for this feature is not used for other RAID modes

[@breadmoth6443]

well I remember Reiser stating how superior his FS was..... i do hope this developer doesn't go down that route.

[@NomenNescio99]

Well bcachefs do have a killer performance.

[@bookle5829]

@@NomenNescio99 LMAO

[@andersjjensen]

@@NomenNescio99 Ouch! :P

[@mcdermottpa]

"Known for ReiserFS, murder" Yikes.

[@ahettinger525]

@@NomenNescio99 more like performance killer, if DJWare's shootout is to be believed.

[@wonderings8973]

20+ years ago, I read the LKML every day for a month. All i can say is - holy hell, flamewar central

[@muizzsiddique]

I was looking forward to this!

[@nitrogenez]

days without rust drama: 0

[@sentinel2199]

Thanks for doing a Bcachefs video. I have high hopes for it, but Kent does seem to have serious problems understanding things from other's point of view (or just heavily undervaluing them), and I do worry he's doing to get kicked out of the kernel at this rate.

[@michaelkreitzer1369]

What's missing from the discussion is the actual nature of the bug and why the author felt it important to push into an RC. Is it an actual corruption bug? I'm sure Linus could be convinced if a filesystem was about to be released that will for sure eat peoples data and this fix had to be done in the way it was, but is that the case?

[@miguelagueda3928]

Some of them might be, and some of them might not, but that's not the issue in question. The issue is he was sending a patch during development cycle and basically saying: "I'll just add a few bug fixes while I'm at it", even though that's not how kernel development is supposed to work

[@michaelkreitzer1369]

@@miguelagueda3928 Yes I'm aware I did watch the video.

[@ArtificialDjDAGX]

I'mma be real, the dev for bcachefs seems like an all around bad actor, and not someone that should ever be unblocked from contributing to Linux, since if they are they'll just try to push backdoors, and other malware into the codebase, or do some other heinous bs to the codebase. It's literally impossible to not understand something as simple as "dev->test->prod", especially after it's been pointed out multiple times.

[@hygri]

We're spoilt for choice these days - we have zfs, bachefs is coming along nicely, btrfs is "stable", xfs is near-on perfect, hans reiser is still locked up...

[@gengHAr15]

Chuckled on that last part.

[@ggsap]

Ill keep using ext4 for regular desktop use

[@the-answer-is-42]

I think Linus is correct. Regardless of the quality of your work, you need to follow the proper procedures for a project like Linux because your pet project isn't the only thing that matters. But that's not to say this filesystem doesn't sound interesting, it absolutely does. The guy behind it is just a bit eager to push his stuff through and I think he needs to learn the value of patience.

[@autohmae]

Always very well done with your context, we know from the earlier days he was a problem in following process. The same things can happen in companies. If you can't follow the rules/can't communicate/etc., you are fired. It doesn't matter how good of a developer you are.

[@Dany-z9s]

Speed is not that important and only shows in benchmarks. Stability is extremely important.

[@fengxii]

This is great context because I thought he was bugging out at first, but what Linus is saying makes a lot of sense

[@iruatlocalhost]

I use encrypted raid1 bcachefs on root. Has a split brain problem once but never lost data and the speed is awesome.

[@bronkolie]

Wake up babe, new Brodie Robertson video on a Linus Torvalds rant just dropped

[@Omnifarious0]

I'm interested in bcachefs when it gets 'send' and 'receive' support. I just finished writing a backup utility to make use of this for btrfs.

[@bloepje]

Kent was right into not putting it in the kernel right away. Because once you go into the kernel, you have to make your patches reviewable. I hope Kent sees the error in his ways, because I am wanting this filesystem for ages, because I am indeed severely disappointed by btrfs.

[@iatheman]

Linus' rough language is unnecessary, but it comes with his personality, and it comes with the firmness and clarity that every high-stakes big project needs. This is priceless if you want to catch mistakes quick, produce high quality work, and learn from mistakes fast.

[@marcogenovesi8570]

This is one of the cases where even the language is probably ok, the bcachefs developer has been toxic and annoying for a while on both kernel list and debian packaging of his tools

[@iatheman]

@@marcogenovesi8570 I usually don't mind the language if I can catch an underlying good intention, glad he improved a lot over the years. I hope the bcachefs author follows the Linux kernel process and stops thinking his project is so good he gets a pass.

[@Karurosagu]

Rough language is not necessary, but rough language is necessary

[@ahettinger525]

Honestly, Linus was pretty restrained here. The bcachefs dev was the only one using any language that could be considered harsh.

[@iatheman]

@@Karurosagu Pretty much. I wonder if the weight of Linus' responses would be lost if it was any less "rough". I could have phrased it better.

[@MatthewWilliamsX]

Yes! Scary new file systems are exciting. Keen to break a few development machines 😂

[@catcatcatcatcatcatcatcatcatca]

Having a working thing and a thing multiple people agree will work when put into use are two very different things. The catch is that there is no reliable way to determine something works without being able to convince people that it actually works.

[@shirro5]

I would like bcachefs to have chance. The rules aren't going to be changed for an individual. The pragmatic choice for Kent is to bend instead of break and keep the maintainers happy.

[@pepepepe796]

We are truly blessed we HAVE Linus Torvalds managing the whole process. He is fair, and just. He truly IS the soul of the machine in the kernel development process.

[@Scoopta]

He complains that BTRFS eats data...I have run BTRFS in production and have never seen it eat data...I use it on my personal systems too. Hell even raid 5/6 SHOULDN'T eat your data as long as you use raid1 or raid1c3 for metadata. I'm sure the FS did that when it was new but there's a reason facebook and SUSE use it in production...

[@dashcharger24]

@@Scoopta it did happen to me, but this was because of faulty memory. ZFS feels really bloated to me, don't understand me wrong, I really like it. But it isn't in the kernel (not that I'm aware of), and it does require quite a lot of memory (for good reason). I like both, but you shouldn't run ZFS/Brtfs and don't expect any problems. Always create offside backups.

[@Winnetou17]

@@dashcharger24 From what I know ZFS is not in the kernel because it has an incompatible license.

[@Scoopta]

@@dashcharger24 yeah, agreed, I backup all my important data. Additionally I used to be ballsy and run BTRFS in linear mode but after a dying SSD ate some of my data(just a few files thankfully which were not important enough to be backed up) I switched to a raid1 setup because while I technically can only lose 1 drive I like the fact that it's like linear raid 10 and with my 4 drive setup I can maybe lose 2. raid6 is too expensive with mixed sized drives which is why I picked raid1

[@qwesx]

@@Winnetou17 Correct. @Scoopta Btrfs did eat data on a regular basis up to a few years ago and still seems to be a bit vulnerable with some features, like using a quota wrecking the filesystem of a user just in March of this year. It's RAID5+ modes still don't cover the write-hole. Also, it's still a good idea to regularly scrub a btrfs filesystem, even if it's constrained to only a single disk, in order to detect potential problems early. For its reputation it didn't really help that during all of its development time zfs had none of those problems (any more, who knows what kind of wild things happened in its early iterations at Sun). But yes, as long as you run btrfs on a single disk per filesystem or use only RAID 1/10, do regular scrubs and don't use the less-often used features you'll be fine.

[@squishy-tomato]

my btrfs laptop can't find a boot device after a forced power cycle. I'm procrastinating to see exactly what happened and if my data is gone (I rarely used it), but I believe the problems and I think they are not mentioned enough when people recommend it to others. On the other hand, I've been running F2FS smoothly on desktop and laptop, and I very rarely see that one recommended.

[@middyjohn]

tldr the chef fs need more time in the kitchen and wait for the next serving

[@Bob-of-Zoid]

I love Linux developers!!!!🥰 They make my computing life fantastical!!!🥳🥳🥳

[@cheako91155]

Back in the r600 days there was a loop that was waiting on a frame counter... I patched it to panic the kernel after 600ms, because when the GPU hung it shouldn't cause me to reach for the reset button.

[@PremiumGerman]

IT'S NOT BUTTERFS IT'S "BEE-TREEFS" Standing for Binary-tree fs

[@squishy-tomato]

"butter-fs" and "f-stab" are more fun

[@ahettinger525]

it is, and always will be "butterfs." Farthest you can get from that, based on what they named it is "Bee-Tee-Ar-Eff-Ess" but noone's actually going to say that. If they wanted people to call it "BTreeFS" they should have named it that, not BTRFS

[@georgehelyar]

All these procedure issues like missing commit messages and unsigned commits should just be caught by tools before they get to a person.

[@MadsterV]

Linux is on point here and with exactly the right amount of force required.

[@cannaroe1213]

Kent's FS might be stable but emotionally i'm not so sure. The shared btree Lib thing had me rolling, i can see myself doing that. Honestly I should try bcachefs before having an opinion. Sounds a lot like the art and zen of thinking your hot shit though.

[@rhekman]

I've read some of these LKML posts and and the Phoronix articles over the last few years, and I can't help but think Kent Overstreet has a serious case of Main Character Syndrome. And sometimes I can't tell if he's just a typical Phoronix forum troll that takes every opportunity to trash btrfs.

[@5fr4ewq]

Amount of tabs opened in this video is scaring me

[@marcogenovesi8570]

puny human

[@hellowill]

This guy should be happy Linus even bothers to provide feedback and help him lol.

[@Sypaka]

From reading the kernel threads... BcacheFS is a development nightmare. This should (currently) not be in the kernel at all. There are even plans to make it in Rust (lol) - yet the driver itself has not even left experimental. I'd even say it's still in "early Proof of Concept" stage. Well, we will see in 2-3 years when this is actually ready for daily use, not just testing - if it's still developed by then.

[@sprinklednights]

Contributing sounds fun.

[@entelin]

I just wish Oracle would publish ZFS using a GPL compatible license. Their license is very similar to the gpl anyway... It's just this bizarre situation where both the CDL and GPL require projects using the source to publish under the same license. So despite both licenses being copy-left they can't be mixed XD

[@SussyBaka-nx4ge]

Sun wrote CDDL to be deliberately GPL-incompatible. They didn't want Linux to be able to just take parts of Solaris. Oracle is even worse, they shut down OpenSolaris and brought it closed source. That said, even if Oracle GPLed Solaris ZFS tomorrow, OpenZFS is a fork from over a decade ago and they have no interest in re-licencing their decade of enhancements.

[@marcogenovesi8570]

Oracle would never ever do that, and even if they do, their ZFS codebase is very different from the opensurce ZFS project, that was forked from the last code drop from Solaris when it was still under Sun

[@ahettinger525]

Actually, the CDDL does _not_ require projects to be under the same license, because it's not scoped to the project, only the file. It also requires the developer to grant usage permission on any patents they might hold. It actually combines, quite easily, with all non-viral licenses. That's why the BSDs support it, without making swaths of their system something other then BSD. The GPL _does_ require entire projects to be GPL, by design. So, any license that cannot be fully supplanted by the GPL is incompatible. That's always been a design goal of the GPL. In short, the GPL is viral, the CDDL is not.

[@marcogenovesi8570]

@@ahettinger525 There are permissive licenses that are compatible with GPL. CDDL was created with the specific intent of being not compatible with GPL, which is not hard due to the GPL's requirements as you stated.

[@ahettinger525]

@@marcogenovesi8570 ONE person has claimed that, everyone else involved has said it's bunk. The goal was to have a license for code that can be used in commercial works, but still require the released code (and improvements) to be made available. That cannot be achieved in a GPL compatible way, period. Realistically, the only permissive licenses that are GPL compatible are ones that are so permissive they can be entirely supplanted by the GPL, and usually whatever other license a user wants (such as the MIT or BSD licenses).

[@JohnMiller-mmuldoor]

Benevolent dictator/philosopher-king

[@James2210]

bigegofs

[@MiguelGonzález-k5p]

Linus has gotten really good at this. He barely sweared this time.

[@MasterHigure]

And here I was wondering what a bca-chef is, and why Torvalds would care.

[@Treviath]

I do love bcachefs, but fucking hell could Kent please behave.

[@elmariachi5133]

'Reliability of btrfs'? Well, this doesn't sound like a quality one would like to have .. This sh*t is the only LinuxFS that ever made me lose data.. and I am by far not the only one. Even worse than NTFS/FAT32 from my experience ..

[@foobarf8766]

btrfs wrecked a customer's data I tried telling the guy he 'was on his own' with that but do people listen

[@hockdudu]

Wait, is the thumbnail a joke on "The Curious Case of Benjamin Button"? 😂

[@jameskelly892]

Hey Brodie, are you planning on attending everything open in Adelaide next year?

[@NJ-wb1cz]

Is this some kind of nudist thing?

[@jameskelly892]

@@NJ-wb1cz lmfao no, though I see how arrived at that conclusion. Everything Open as in all open source technologies. It’s an Australian Open Source convention

[@jameskelly892]

@@NJ-wb1cz Unless it’s both - an Open Source Nudist convention and I’m painfully unaware. What a shock turning up to that would be!

[@Winnetou17]

LoL, you guys... :))

[@NJ-wb1cz]

@@jameskelly892 that's why you should turn up naked regardless. Either you will fit in or it's the others that will be shocked instead you - it's a win in any case

[@harambeduck4110]

I already am afraid of the time when the Kernel gets more and more enshittified as soon as Linus throws the towel someday in the future...

[@privacyvalued4134]

I use a few tools I think should be deployed with every distro. You know why I haven't pushed the issue with distro maintainers? Because I don't want to rock the boat and cause a ton of drama. The tools I use are deterministically superior to the existing stuff that ships by default with all distros but there are plenty of people who will vehemently disagree and are set in their ways. I personally like stable software and rely on LTS distros wherever possible. If the meaning of "stable" ever changes, I will re-evaluate my use of Linux. Linus understands this while this Kent guy doesn't.

[@TavishMcEwen]

I'm gonna reinstall on BCacheFS accross two SSDs, two HDDs, and an intel optane! *once i finish repairing the ZFS pool that shit itself*

[@sjoervanderploeg4340]

Heh, I stumbled upon bcachefs in 2016 orso when I was looking for a replacement to the shortly lived Intel SSD caching... which worked great mind you... for NTFS partitions...

[@sjoervanderploeg4340]

but took the ZFS route :D

[@tech34756]

If you want peace of mind when it comes to data, you really should be making backups and if possible verifying them somehow e.g. checksums.

[@carbon13]

The moment your argument for why everyone should bend over backwards to incorporate your feature in the next release is that your feature is "just better" and "you'll all want to be running this in a year or two, just watch", you have lost. I honestly don't know why this needed to be an in-tree module at this point, except to try to skip the line and gain testers for a clearly unstable feature for minimal effort. Even if you end up being right, and your feature is the best thing since sliced-bread, people interested in early access to it will have no problem either building it themselves, or working with you to add your feature to their distro/package manager.

[@hellowill]

Yeah every project needs someone like Linus to push back. This IMO is the biggest problem with software engineering atm; devs are too afraid to push back. They let management get their way too damn often. Game dev in particular is fucked because of this. It is on devs to hold their ground when it comes to tech debt, quality, feature estimation, etc.

[@rashidisw]

Even if its really is too many to list, LIST 'em ALL. Yeah, I'm belong to this camp.

[@dansanger5340]

Leaving the drama aside, I'm kind of interested in this file system I never heard of before. ZFS is reliable, but very memory hungry. Btrfs seems to be a perpetual work in progress.. Maybe Bcachefs will be what Btrfs was supposed to be.

[@marcogenovesi8570]

ZFS isn't memory hungry, it just decides to use half the RAM as cache by default but you can change that setting. Btrfs is fine as long as you are not using RAID5/6. Bcachefs now is like Btrfs 10 years ago, but with a single developer

[@stockk_]

the person who said bee my honey sucks is objectively and factually WRONG 🐝

[@LTPottenger]

It's kind of nutty to put a FILE SYSTEM into an OS that is not 100%

[@thisguy6559]

Hey, at least it's not Hans Reiser!

[@denodoko]

I'm using bcachefs to store 8TB of data in RAID 1... And even I think Linus is in the right here. That's why I have backups going back almost a year and scrub the disks once a month to check for errors. It was the only fs supporting raid 1 with asymmetric disk sizes so I took the plunge. But I knew what I was getting myself into. Thankfully, apart from not wanting to mount one time and eating all of my CPU the other time, there haven't been any major problems. And I've been using it since before it got merged. EDIT: As I've been reminded, btfs has also supported asymmetric disk sizes since quite a while. See my later comments.

[@frechjo]

I would have thought ZFS was a better option?

[@deckard5pegasus673]

if you want to live dangerously and want adrenaline rushes, it's probably better to try bungie jumping.

[@denodoko]

@@frechjo Forgot to mention that I also wanted an SSD write-back cache in front of all this so I can work on the fs "directly" despite it being a bunch of slow spinning metal. So I chose it at the time, because: - it could create raid 1-like redundancy on a pool of different sized disks (8TB + 3x4TB + 20TB = 40TB, so 20TB usable) and - it supported easy disk-tiering. From my research back then, it didn't look like ZFS supported all that.

[@SussyBaka-nx4ge]

@@frechjo ZFS doesn't support asymmetric arrays.

[@frechjo]

@@SussyBaka-nx4ge I see, I assumed it did (because btrfs does), thank you!

[@immoloism]

Don't forget XFS has CoW in reflinks.

[@marcogenovesi8570]

reflinks don't make it a CoW filesystem, it's a journaled filesystem with some CoW features

[@punkrockllama]

Linus is basically Linux's Gordon Ramsay. Yeah. He gets heated but I've never seen it towards anyone who shouldn't just know better.

[@timeobserver8220]

I liked that you didnt rant as much in this video. I like your videos and you but sometimes when you have a lot of energy it feels like i need to slog through 20mins of your ranting to get the 10mins of linux dev or it facts. I like your neiche here in this channel, no one else does core linux dev news and social drama like you, and i do want to hear your opinions on the topic. Thanks for all you do but please also respect my time. Love ya mate.