Hector Martin: we care about fixing Linux problems on Asahi | Hacker News Reader

Hector Martin: we care about fixing Linux problems on Asahi

social.treehouse.systems

146 pointssohkamyung57 comments

The amount of shit takes on that thread are astonishing from people wanting to paint Hector's comments in a bad light. Trying to twist his words to extrapolate a meaning that isn't there. Of course, there are tons of people who also have no clue how hardware support comes into being. Some hardware is buggy on linux, has been and may always will be. A lot of the reason is, either the drivers had to be reverse engineer, which takes a lot of time and effort. Or, something that happens a lot, vendors throw together shitty drivers and call it a day. No hate for the SEs at these companies though. Some of these companies invest all their money in the development of hardware and clench up the purse strings once it comes time to write the actual drivers. Then from here, either the driver is open source and no one gives a damn to go fix the driver, or your back at needing to do some reverse engineering.

>The amount of shit takes on that thread are astonishing from people wanting to paint Hector's comments in a bad light. Trying to twist his words to extrapolate a meaning that isn't there. [...]

um isn't there only one comment that matches what you're describing? i actually thought it was pretty neat how that discussion started out in the vein of a twitter thread, with one person misinterpreting another person's tweet and moralizing about it, but then you have that super long comment from alyssa 6 hours later that says "i think you're both right" and goes way into depth about how things are the way they are

on twitter there wouldn't have been the space for that comment and it would've been overrun by a bunch of people making quick moral judgements about a tweet they deliberately misinterpreted

This is actually my first time reading a mastodon thread, and I must say that the extra long comments allow for thoughtfulness, insightfulness and pleasant reading. At first I didn't read any of the replies because I was expecting the standard Twitter experience.

Definitely about Mastodon. I've been wanting to try out Mastodon and I finally took the plunge. So far, it seems a lot better because it allows for thoughtfulness. Will have to see how it goes!

I'm not surprised. Any time you talk about how terrible the Linux desktop experience is, you'll get people saying shit like "Why do you need that feature? Linux works just fine on my machine". Marcan really pissed off that crowd when he acknowledged the feature disparity between Linux and the preinstalled OS on prebuilts:

"I just realized that one of the things that bugs me is people equating stereotypical Linux problems on Asahi with other platforms, because it's not the same, because we actually care.

Suspend not working, audio not working, display tearing issues, power management being bad... all those things are Linux on $random_platform support memes. And they're memes because they never get fixed, because nobody cares."

It's a simple logical fallacy. Yes, the memes are true, but the explanation is false. Many many many people care.

It's evident if one looks at endless bug reports, mailing list discussions. Many users and maintainers care. But even though they do, the bugs are not trivial.

Many strange bugs are a consequence of subpar software architecture. One example that comes in mind is a bug about the xfce(?) lockscreen and laptop lid closing, that only activates when the lid opens. So for a time the desktop is visible. (There's also one bug where the password input field is invisible in some cases, etc.)

It seems security and UX critical functionality is broken.

But the truth of the matter is that it's also broken on many Windows setups too, just in other ways. Eg. complete freeze on restore.

Or look at the GitHub issues for WSL, there are a lot of strange bugs, like DNS stops working and only reboot solves it. And those Microsoft engineers care even less than some open source maintainers. (For example the TypeScript maintainers are very responsive, even though that repo also has a lot of bug reports.)

> Many users and maintainers care. But even though they do, the bugs are not trivial.

Hector's point is also that the ability to fix the bugs is diffuse: to fix a problem with the SSD you ultimately need to get the SSD maker to care. To fix a bug with the sound, you need to get the sound chipset maker to care. And actually to get a broadly good user-experience such that it works good not just on your system but every possible system, you need to get an awful lot of people to care.

As ever, one of the plusses of the Apple platform is the absolutely fixed set of hardware it needs to target. There is only one wireless chipset that Asahi needs to get working, for example. There is only one sound chipset they need to get working. When it's working, it's working for everyone, and if there's bugs they're bugs that affect everyone and aren't a result of some obscure combination of drivers interacting in unexpected ways.

Also, Apple actually internalized a lot of control in ways that really piss people off when they want to upgrade hardware, but it also means the SOC has a lot of power to just do things right in the first place since it's all software-driven anyway. You don't have to worry about the NVMe controller doing something weird because... the NVMe controller lives on the SOC (might even be fully software-defined?) and just talks to some flash chips sitting on a card. It's completely different from how the rest of the market does it, and it's undoubtedly more work to get it going, but it also (paradoxically) means they don't have to worry as much about hardware quirks, because there's less hardware in the middle to begin with.

It's undoubtedly a reason they've been able to make such great progress - not that the dev team isn't extremely extremely talented anyway. But they don't have to solve the general Linux problem of getting 10 billion possible permutations of hardware to behave together. They just have to solve one platform with one set of peripherals.

Yea, Apple's vertical integration and absolute homogeneity are definitely big factors here. Also size of userbase, and the expectations of said users. (As someone already mentioned folks buy Pine64 SBCs for 20-50 bucks for fun, to tinker, to run their hobby projects, but almost nobody buys an M1/M2 chipped device for thousands of dollars "for this funny little project".)

But what I find very frustrating is that many software projects - especially open source ones - don't really think about the end-to-end user experience and don't have "interfaces" to help manage the UX, as in signal that the device is not respecting some standard/protocol/API/whatever. At best there's some obscure error message. Assertion failed. Blahblah.

I mean ... I know there were latops with Ubuntu stickers, but it's just not a big thing - as far as I know. Sometimes niche manufacturers with some newsworthy feature (FOSS firmware, drivers, etc) get into the news, but that's why it's news, because there's just no run-of-the-mill boring supplier...

A Mac mini is $600, which is pricier than $50, but that's not thousands, plural.

It's the Asahi team using Apple's own cultural proclivities for their benefit. All the single chipsets, and merged functionnality into the SOC are things Apple wants for themselves. I can't begin to imagine how cost-effective it must be to skip the NVMe controller chips in all their hardware.

It isn't always even bugs. There are features I'd GNU+Linux to get rid of most notably the two clipboards. Screw that noise. We only need one clipboard. However, what we get is more code to keep the clipboards in sync.

https://codeyarns.com/tech/2017-01-18-the-two-clipboards-in-...

Don't touch my two clipboards. I use them both often and love it.

(There is actually more than two)

Oh gawd is THAT what that is??

I spend my days in aix and my evenings in windows. Every time I use Linux though, I get frustrated at seemingly completely unpredictable behaviour of clipboard! Now I at least have a basic understanding of the model - thx!

Oh come on, Hector constantly comes off as condescending and insulting to people that don't immediately bow down to his opinion and his immediate viewpoint. It's not like he's a saint, his Mastodon (and previously Twitter) feed is full of drama where he belittles other people and also takes things out of context. He's like that guy in the team you have that's very smart but also thinks everyone else is an incompetent monkey and constantly puts them down over it.

This linked post is pretty clear example of that - after decades of hundreds of people making sure that hundreds of configurations worked very well on Linux, he's now all smug because they're working on a single Apple configuration.

The social interactions he ends up having are of his own making as well.

> after decades of hundreds of people making sure that hundreds of configurations worked very well on Linux, he's now all smug because they're working on a single Apple configuration.

Nowhere in the linked post does he talk negatively about those efforts, more about hardware manufacturers not doing their job. But he's objectively right about the average Linux experience on a random laptop. I have never owned a hardware configuration where everything worked with Linux except for the Steam Deck. My best experience was on a Thinkpad from 2016 and on that one I still had to adapt and compile my own wifi driver.

The opening post concludes with "nobody cares", which is obviously false, and it was a nice gigantic straw man dunked in jet fuel and then set on fire to attract all the flame war moths.

> Oh come on, Hector constantly comes off as condescending and insulting to people that don't immediately bow down to his opinion and his immediate viewpoint

Don't know the guy, but from my experience people who "vehemently contest" the opinions and viewpoints of project leaders can come off arrogant and know-it-all from the POV of the people who are actively working on the project and bootstrapped it from zero to success.

It's one of those "clash of the titans" scenario.

I mean it's to be expected: a popular project gets a lot of praise but also a minority of vocal critiques.

How single project leaders react, unless they are obviously abusive and/or violent on average/randomly, it's not really that important IMO.

> The social interactions he ends up having are of his own making as well.

That doesn't make him wrong, this topic in particular is a real pain point: manufacturers don't care about Linux users at all, it's their fault if Linux is mostly a less smooth than it could and should be experience.

I agree Marcan instigates a lot of twitter drama, but "hundreds of configurations worked very well on Linux" is kind of a stretch unless you're counting headless servers and really low end machines. There's a reason why Linux desktop has 1% market share (mostly chromebooks) and it's not because of Microsoft. I appreciate the work those devs did, but they really didn't move the needle. Neither has Marcan, but hey at least he is acting like a product guy and prioritizing features that are deal breakers on pretty much every other Linux system.

He's building out same features that already work on many other devices. Yes, there are hundreds of laptop models with fully working hardware (including sleep, bluetooth, speakers, power saving and whatnot), so there have already been plenty of developers that were "product" guys.

And that's all even if we ignore Android and ChromeOS devices exist.

It doesn't make Hectors contribution any less important for all of the Apple fans, but putting down other Linux developers like this is disgusting.

Exactly. These days there are several vendors preinstalling GNU/Linux on their machines where everything works, and it's really not hard to find machines where it's not preinstalled but where everything works anyway. These memes are memes because it's also not hard to stumble upon a machine where it doesn't if you just take a random one, but that's hardly surprising - just take a look at how much effort it takes to run Windows on Steam Deck and have everything work well there. Someone needs to put that effort first when the vendor does not.

I love those "I had issues so everyone else must have had them too" takes. I can give some anecdata too: the last integration issue I had on machines I've been running GNU/Linux on was a non-working SoundBlaster Audigy card somewhere around 2006 - and I've been mostly using laptops afterwards, some of them being from vendors that didn't provide any GNU/Linux support at all.

(that sound card eventually gained support in ALSA, by the way)

He does make a point of wanting things to work well. I have been using Linux on a 2013 laptop for five years now and I can't say I have ever run into a problem. I do remember having to install some obscure wifi driver (that was in the repo!) for some other older machine but that is all.

But I am not sure it works well. Hibernation never worked. Was performance, sound optimal ? Battery life ? I love that they are working on thing like that.

Hardware support is not the reason why Linux has 1% market share. Most users get a fully licensed copy of Windows bundled with their laptop, and it works, and it does everything they want, so why change it? I would argue that the software side is more important, with Linux still sucking at HiDPI setups (as seen in ~every non-ultra-cheap laptop) and with GNOME being GNOME.

For non-OEM desktop PCs, I completely agree that it's all about userspace. Last time I tried daily driving Linux, the terrible HiDPI support (after 4K panels have been widely available for a decade) was one of my biggest issues along with the weird frame pacing that made some apps feel as smooth as sandpaper even on relatively fast 120Hz VRR hardware.

For laptops however, hardware support is still terrible or nonexistent. If you have a modern wifi 6/6E chipset, a fingerprint sensor, or a webcam that uses the chipset ISP, there's a very good chance it's unusable. Battery life is still a decade behind state of the art.

My Dell laptop’s got a WiFi 6 card and a fingerprint sensor. The WiFi card works (although I don’t have a WiFi 6 network). The fingerprint lets me register fingerprints, but I can’t log in with it, which is probably another case of crappy software.

No twisting required.

> we actually care … they never get fixed, because nobody cares

He's pretty explicit. He thinks he and his care more than other distributions.

I'm sure his observations have nothing to do with only supporting two related SoCs vs thousands of disparate chipsets, and the ability to repro from there.

It's a terrible comment.

No. Nobody cares if my cheap white label Walmart laptop crashes on standby. I can open an issue on every distro on existence, nothing will happen, no kernel dev will suddenly go out, buy that same laptop, and fix standby.

You've mudged care and ability to action.

Plenty of hardware gets attention for exactly the same reason M1/M2 do. It's nothing to do with care, everything to do with access.

For me, the best example of the "we care" attitude from Asahi was in the installer itself - even in alpha it was well polished, and usable. Clear defaults, documented behaviour, help messages. And this is an installer that needs to go the extra mile to show up in Apple's bootloader.

It was a breeze to install (M2 Air).

I’m very grateful for the Asahi project, Hector’s leadership and everyone who has contributed to or supported the project. This care for the nitty gritty details, in my view, makes a big difference and is a true service to the world. It is a heartfelt gift to the community.

Thank you for caring.

Lenovo will release firmware updates for Linux users even on systems that don't officially support Linux (though, it can take a while). But I guess that's one of the exceptions that proves the rule

Devices actively supported by vendors on LVFS tend to be Linux-friendly. The vendors with the most firmware updates pushed to LVFS in the last 6 months are Dell (796), Lenovo (364 combined), HP (35), Star Labs (29), and the Pine64 community (3).

- LVFS vendor list: https://fwupd.org/lvfs/vendors/

- LVFS supported devices: https://fwupd.org/lvfs/devices/

Good for dell and lenovo but good lord it’s depressing that 5th place takes a grand total of 3 pushes

Surprising to see tiny Star Labs in there

This makes me feel better about their laptop offerings, which seem very good.

Awesome, is this basically the "what should I buy for running Linux on it" list?

Yes, but it's not complete. I'd also include manufacturers that maintain Linux distros designed to support their devices, such as System76 (Pop!_OS), Purism (PureOS), and Tuxedo Computers (Tuxedo OS), if you're willing to use these distros. These 3 manufacturers have their own firmware update methods outside of LVFS.

Is it?

Last time I tried to use t480, which is listed as officially supports Ubuntu, I had to use rather ugly hack to fix CPU throttling:

https://github.com/erpalma/throttled

It affects other models as well and not fixed up to this day (at least not by vendor)

Lenovo had a post in their blog acknowledging it, but after some time it mysteriously disappeared

EDIT: Lol, there are HN post about this "fix" happened few hours ago:

https://news.ycombinator.com/item?id=35025838

By quickly checking comments seems like better solution is already available:

https://mjg59.dreamwidth.org/54923.html

Will they ever fix their newer thinkpads throttling like hell on Linux (and even windows, actually)?

That happened to me back when I bought my T490. Later they released a firmware update that fixed it. So the only case where I saw such issue, Lenovo fixed it quickly.

There are plenty of forum posts like this ranging over years: https://forums.lenovo.com/t5/ThinkPad-T400-T500-and-newer-T-...

Anecdotal, but the T14 was also prone to throttling all around my previous company.

Yes.

That’s a pretty baller story if it works out. I use Apple gear because all that shit works perfectly, but would I rather have a hackable tiling WM instead of divvy, obviously.

Great dream if it happens.

Have you tried yabai as your WM on macos?

+1 for yabai! It's awesome, such a timesaver. I do miss dwm, but i think yabai is the best of the bunch when it comes to macOS tiling WM's.

I understand the project goals and everything but isn’t the primary weakness of it the same as every other laptop? If Apple decides to change something (which they will) or lock down the laptop more (which they will) then all of this work goes to nothing.

Isn’t there a better case to be made to strive towards open hardware and driving a community around common hardware and goals? Something like system76, the pine hardware, etc.

In a theological sense you're not wrong, but what are we supposed to do for the people who want to run Linux on a Mac? Have them pound sand? What about the people who want to run Linux on their Microsoft Surface?

What seems to be the issue is that you have a group of people (Mac buyers) who are keenly inconvenienced by Linux's shortcomings and want to solve them. Let them, everybody will ultimately benefit.

> If Apple decides to change something (which they will)

sure but there's only so much they can change - everything past the bootloader is open, non-apple software, by design. hardware changes? sure, could happen. Asahi Linux has already identified some likely ways that future uarchs might change. There could be supplier changes, etc. But there are still only N different hardware configurations to support, where N is like <10 right now.

> lock down the laptop more (which they will) then all of this work goes to nothing.

no probably not and they've actually gone out of their way to open it up even more - they actually have already changed things, and they changed them in a way that opened them up for the Asahi Linux people. And internally people have stated to Martin that the functionality exists is explicitly created to allow third-party OSs to work, that it is tested to work, because Apple has a hard policy for it to work. They're not going to make your OS for you, but they have a hard corporate policy of leaving the necessary pieces there for you to do your own work yourself.

https://web.archive.org/web/20220802091506/twitter.com/marca...

This is where you and philistine below are wrong and are injecting your own personal biases and preconceptions into the issue - Apple is not "ideologically opposed to open software".

You are literally being the precise "hackernews peanut gallery" that Marcan is calling out in the second tweet, on the exact issue he has called out in these tweets. Like seriously, the PG thing about "haters" being the mirror image of fanboys is so absolutely correct - some people are parasocially attached to brands in negative ways and they're just as blind and trite as the people who are parasocially attached in positive ways. Apple and NVIDIA and a few others have always generated these negative parasocial attachments.

http://www.paulgraham.com/fh.html

The macos platform is not the same thing as the iOS platform, and even then there have always been solid technical reasons behind the restrictions that Apple does - they take some security and policy arguments to the uncomfortable conclusions that nobody else will (or even can). But it's not coming from nowhere, or pure avarice, either: iOS is one valid way to build a platform with permissioning that actually matters, that can't be sidestepped by social attacks from powerful actors like Facebook demanding you bypass permissions for them.

Same for nvidia… even when they open up adaptive sync, or opening their kernel land (while providing a userland that is far more usable than ROCm under Linux), or start adding Wayland support, they get nothing but shit. Celebrate the good changes when they happen, instead of nothing ever being good enough. Criticize the bad moves too, but understand there are often very significant technical or business reasons behind them. Just like there are good business reasons why ROCm is a mess for AMD. Companies can support different things and have different business priorities and that’s fine.

But linus doing the middle finger is the end of it for a lot of people - it's the end of any consideration or conscious thought on the issue for a lot of people. Linus said "fuck nvidia" that makes me laugh out loud guys!!! It's always been more complicated than that, and Linus is just a (self-acknowledged) toxic asshole who (self-acknowledged) needs to tone it down, and that phase lasted for about a week before he was back to being a toxic asshole. Stop idolizing him, and realize he's not always right even on the issues in his own domain (see: his comments around AVX-512).

How much are they going to care in 7 years when the sheen of M series wears off, Apple's benefits in performance per watt are no longer as impressive and they just made some massive design change breaking loads of things? Will they really care then?

I personally think the odds are pretty good that Asahi ends up supporting the M1 series longer than macOS.

Apple isn't shy about cutting off support for old hardware; the current version of macOS won't run on any machine more than five years old (according to https://support.apple.com/en-us/HT213264).

Does cuda etc work out of the box on Asahi? I would be minded to check it out if so.

Dose Asahi even run on anything with an Nvidia GPU?

Presumably you could put an nvidia gpu in a thunderbolt enclosure, and apple silicon Mac Pro shouldn’t be too far off, so the answer is “yes, if you connect one”.

Nvidia actually already has armv8 drivers for the jetson platform, no idea how comprehensive they are though. Might only be the things they need for various generations of Jetson and maybe the odd Fujitsu thing.

And because of the miracle of arm ISA versioning, there’s probably some fun gotchas there too.

Well the first step would be adding thunderbolt support to Ashai

AFAIK no eGPUs work on M1 and no CUDA support either.

Computing everywhere seriosuly lack a coallition of the willing. There's so much regular industrial/capitalistic dialing it in, ultra-stratified systems where those who can are a thousand miles apart from those with any power.

This post is remarkably on point, not just for the Linux kernel, but for such a huge majority of computing in general. The people close to the machine are not respected & heard. Those with the most refined tastes are ignored, distrusted aliens speaking their esoteric alien whayever, put far away from power & decision making.

The system is set up to exercise only the priorities that can fight their way to the top, to clear visibility. Many of the "what have you done this quarter" problems are ultra real & compound this all.

The structure is a problem, but to return back, computing's best are just not understandable by most. The deep concerns matter so much, computing is so much about maintaining & caring for complex technical ecosystems, but there's so few well travelled enough to weigh in & understand & make worthwhile assessments; the techies live a too peer-less existence, disconnected from so many non-technicals with so-so motivatiom, capability, & interest to care, & separated by other techies who have their own separate stakes & motivations.

Just being around other people who care & want to make things widely better is a rare gem of a place to be in computing. It shouldnt be so, but it is, and it's awful, and we have such brighter better ways to be that are always so glistening great & close, but so damnably just out of reach. The best intents have so little backing. So much unknowing unseeing resistance impedes the good, purposelessly, ignorantly, from so many assailing places. Woe.

I use Linux since ~25years, and he's totally right.

I cant stand the amateurish "it works here" or "wait your laptop is 8 years old? For sure the new kernel cannot load your nvidia driver" or "yeah that intel wireless card never worked good, buy a atheros card" or "you cant use XFS with that game because of inode64, just use wine or use another filesystem" and and and.

It's a real shitshow, i would never recommend any Linux-distro to anyone if i cant hold that users hand.

Reply on news.ycombinator.com