1. Hardware companies need some software to test their product, so this will be wasting their effort.

2. Good documentation and testing is probably not as common as you think.

3. It takes a huge amount of effort to write high quality drivers by reverse engineering. The thing about Asahi Linux is that the hardware does not have many variants, but there are a lot more different hardware for PCs.

> 2. Good documentation and testing is probably not as common as you think.

Not to mention that documentation != the actual behavior of the card. both nvidia and AMD release a ton of day-one driver fixes for new games being released that provide stability and sometimes double-digit performance gains over a pre-patched driver. The only way this works in a disintegrated ecosystem is if game devs become pros at creating driver bug workarounds.

> The only way this works in a disintegrated ecosystem is if game devs become pros at creating driver bug workarounds.

They already are, they have to because that is the reality on mobile games. IIRC the devs of Real Racing (or another Android car race game) made a blog post on that.

> 1. Hardware companies need some software to test their product, so this will be wasting their effort.

They can also open source their testing bench code. The only ones who have to be afraid are those who don't care about quality. "Sunlight is the best disinfectant" also applies here - I'd really like to be able as a consumer to choose the product that demonstrates the better engineering!

> 2. Good documentation and testing is probably not as common as you think.

Oh, I'm aware of that, and the consequences of that are becoming more and more dire as electronic waste from defunct companies or unsupported and unsecure devices piles up, together with hacked IoT devices causing mayhem on the wide Internet in botnets.

It's high time for that to stop.

> 3. It takes a huge amount of effort to write high quality drivers by reverse engineering. The thing about Asahi Linux is that the hardware does not have many variants, but there are a lot more different hardware for PCs.

So what? Ordinary PC hardware at its core still conforms to decades old standards that can be used as a fallback. Given a sufficient supply of beers, coffee and pizza, every CS student worth their money can throw up a basic graphical OS that runs on any mainstream (i.e. non-embedded) x86 system. Input, output (video, serial, audio, keyboard, mice, accessories) and storage all have standardized interfaces - the only thing that does not is networking.

There is no reason at all to say that the ARM world can't standardize on basic components or at least interfaces. Apple can - their UART implementation, according to rumors, dates back to the early iPod/iPhone days, while Samsung has something like five or six completely different implementation, and Mediatek and all the other SoC vendors run their completely own stuff.

Every year, Apple releases new hardware and software in tandem. How long does it take Android manufacturers and Windows manufacturers to take advantage of the latest OS features?

How are those Android updates working out compared to iPhones?

> How long does it take Android manufacturers

Way way too long, but to be fair a huge part of the issue are carriers and their insistence to pack the phones full of bloatware that needs to be tested against each update.

> and Windows manufacturers to take advantage of the latest OS features?

Windows device manufacturers tend to do what Microsoft demands of them.

> How are those Android updates working out compared to iPhones?

Samsung is fairly decent, although I'd be happier if they'd ... consolidate their lineup a bit and instead offer longer availability for spare parts.

Fairly decent? The iPhone 5s from 2013 just got an update and it’s the carriers is a poor excuse. Apple runs on the same carriers.

Google’s first party phones don’t have third party bloat ware and still have a piss poor history of updates

And guess why? Because ~3-ish years used to be what the SoC vendors provided as BSP support timeframe, and there was no mandate that they open up their specifications. Most don't even care about following the legal minimum aka provide the kernel and bootloader source code, especially not all these fly-by-night gongkai "brands".

So it’s what happens when hardware and software is not integrated…

No, it's what happens when there is no competition in the SoC vendor area. Apple and Samsung don't sell to external customers, NVIDIA doesn't do much with Tegra outside of the Switch and automotive, Broadcom doesn't do much in mobile SoCs (and from what the Raspberry Pi community learned hard, their chips are utter dogshit which is how RPi got started after all - they used surplus crap that Broadcom wasn't able to sell), which leaves Qualcomm as a sole supplier for the "high end" and Mediatek/Rockchip for bottom-of-the-barrel crap. There used to be Annapurna Labs as well, QNAP used their SoCs for a while, but I think they stopped selling to externals as well as they were bought up by Amazon and now only do Graviton which are Amazon exclusives.

And yet Apple uses Qualcomm chips and manages to update 10 year old phones.

It couldn’t possibly be your premise is flawed could it?

> And yet Apple uses Qualcomm chips and manages to update 10 year old phones.

For modems, yes, but only because there's even less competition in that area (which is why they bought Intel's mobile modem business, but failed to produce a viable design in all the years since).

They've been running their own SoC designs since the iPhone 4 era in 2010.

So now it’s the severe lack of competition that allows Apple to support phones for a decade but stops Android manufacturers?

Apple has an interest in keeping their devices alive, since they can make money of the store and services. Android manufactures want you to throw the phone away and buy a new one.

Since manufacturers didn't demand long support, Qualcomm didn't provide updates for their SoCs. This meant Fairphone couldn't support their phones even if they wanted to, which is why they used am automotive chip in their new phone — which is nearly the same as a phone chip except the 13 years of updates.

> Apple has an interest in keeping their devices alive, since they can make money of the store and services

Isn’t that the free market working the way it should without government interference?

The problem with an unregulated free market is that eventually one or two companies get so dominant that it is virtually impossible to challenge them as a competition, especially in fields that require immense amounts of money to break into. One might even argue that the end game of capitalism is to acquire a monopoly, which one can then use to extract rents as one feels free to do, or just keep milking locked-in customers (like IBM and Oracle are).

It took Apple, flush with cash from iPhones and iPods, a decade worth of work to create a SoC able to throw punches at eye level with Intel and AMD, which themselves grew to the unholy duopoly over decades. Chipmakers have it even worse: TSMC has all the cards, Samsung and Intel have completely fallen behind with no chance in sight that they'll match TSMC any time soon. All the other competition has gone bankrupt or stopped at lesser nodes because they couldn't keep up the pace.

There's no disrupting that, not even if you are an actual nation state.

So guess what happened when the EU forced browser choice in Windows? It had no long term affect on the market share.

Geeks just always hate when normal people make their own choices using their own free will.

Every failure that you cited were well funded companies who failed to execute. Amazon, Facebook and Microsoft failed trying to introduce phones.

The market worked as it should - poor execution led to failure.

Lame excuses, Google could do the same as Microsoft with PC standards and OEM legal obligations.

They have chosen not to.

Agreed. Google has gone down the drains just as well on all fronts... the thing is, once again, there is no viable competition left since Microsoft gave up. The various Linux phone projects are extremely niche stuff, Blackberry and Symbian are long gone, and Google seems to be happy coasting along on mediocrity and billions of dollars in Play Store revenue.

That wouldn't change at all. The newly separated software company would get appropriate hardware specs and prototypes during the same timeline they do today (as a vertically integrated branch of Apple).

So instead of working together , the hardware team would work in isolation and then when they are finished they deliver to the hardware company and keep going back and forth? Yeah that’s going to work out real well. Let’s just look how well it work for Android.

> the software team would work in isolation

Only if the hardware company refuses to share info with them, which - as you clearly laid out - is not in their interest.

You really don’t see how that would slow things down compared to the hardware and software company working together?

I bet you also think that a company should set up a DevOps “department” to increase efficiency

Are you telling me that Apple doesn't have a separate hardware division? I sincerely doubt that.

I am saying that Apple has a software and hardware division that work in concert.

Chris Lattner, the inventor of the Swift programming language and the open source LLVM was once interviewed on the Accidental Tech Podcast and he was asked whether bytecode could be used to allow Apple to switch architectures without forcing them to require developers to recompile and still have native compilation. He made up some half truth about whether that would be possible.

Years later when the S chips in the Apple Watch became 64 bit, he admitted that at the time of the interview, he knew that the 64 bit S chip was on the roadmap internally and he designed the LLVM with that in mind.

That’s what happens when you have software and hardware teams working together. The same is true with Nvidia and the their software developers who design Cuda and Intel and the compiler makers.

There were also reports that the software team responsible for the JS engine in WebKit worked with the processor team and the processor team did some design work around Apple’s ARM variants to make JS pre compilation faster.

This is also the whole idea of VLIW architectures where the processor manufacture depends on a team of compiler makers to ring out efficiencies and they work closely together.

Heck even back in the Apple // days, Woz a developer, designed hardware and software together.

Have you wrote drivers? Writing a driver off documentation is: a) very hard, how do you debug anything without JTAG ports and simulators? and b) not correct on principle because the documentation is not the hardware, there are errata in the hardware so what documentation says is not always what the hardware does and there are mistakes in the documentation for the same outcome.

>The only ones in trouble would be the vendors of crap SoCs who ship extremely buggy hardware and make up for the lack of QA in software quirks instead.

I.e. the every single vendor currently on the market...

> Writing a driver off documentation is: a) very hard, how do you debug anything without JTAG ports and simulators? and b) not correct on principle because the documentation is not the hardware, there are errata in the hardware so what documentation says is not always what the hardware does and there are mistakes in the documentation for the same outcome.

I agree with you, no system (particularly not something as complex as an entire SoC) is free of bugs.

> I.e. the every single vendor currently on the market...

And yet, we have Apple, whose hardware quality alone is good enough for a dedicated team to reverse engineer and write drivers for, with almost zero documentation available.

The problem, as I expanded below, is the complete lack of competition in the SoC space. Everyone gets away with mediocrity because the largest players in town (Apple and Samsung) refuse to sell their chips to externals. The result is that the mobile ecosystem as a whole suffers, because that is what happens in markets where entities get too large.

>And yet, we have Apple, whose hardware quality alone is good enough for a dedicated team to reverse engineer and write drivers for, with almost zero documentation available.

Apple own software is pretty janky, my current MacBook cannot bring up GPU after sleep about once a month and needs a reboot to begin drawing GUI correctly again, used to be dying completely waking up before it was fixed couple OS updates back. My previous MacBook could not stay on WiFi for more than a couple hours straight, never had been fixed.

Also, reversing an existing working driver is different from writing a driver from scratch. In one case you just need to repeat the functionality, produce the same outputs for the same inputs, the quality of hardware is orthogonal here since it's already worked around in the existing driver. It's orders of magnitude easier than writing the thing from scratch.

Documenting this in detail is neigh impossible. Ashai Linux has the benefit of having working code to RE - this wouldn't work for silicon or board bring up.