[SchemaLoad]

I hate to complain, but I can't help but feel this is kind of impossible with the resources available to the people working on it. Reverse engineering a modern phone would take years and years of work from many people, and by the time you have it worked out, the phone is obsolete and very few people still use it.

The Apple Silicon macbooks seem a good example. The M1 came out about 5 years ago now and with a whole project and a lot of work later there is still limited hardware support. Having to put this effort in for all the models of phones seems massive.

[torben-friis]

One would hope that enough things stay similar between devices that replacing, say, the galaxy s25 paves the way for a far easier implementation of the s26, particularly now that the market is stagnating a bit.

And I’m not knowledgeable about this at all, but intuitively I’d expect apple stuff to be much more customized than the average android phone - they’re famous for vertical integration and owning the end to end process.

[pjmlp]

Phones aren't x86, each is own snowflake, and on Android the nature of being a managed userspace, means there is a certain freedom regarding which ARM designs that Samsung, Qualcomm, Mediatek, and whatever else is out there comes up with.

Then there is everything else that happens to be on the motherboard.

[mcny]

The camera is also surprisingly software dependent. Using the pixel camera vs using default configuration with an app called open camera, the difference is so clear. The camera is basically all software — the images are pure trash before processing.

[torben-friis]

>Using the pixel camera vs using default configuration with an app called open camera, the difference is so clear.

Is this hardware dependent though?

I ask because I’ve seen custom android mods that port the pixel camera, presumably if that works for other devices it’s a good sign that postprocessing can be decoupled from specific hardware.

[VGHN7XDuOXPAzol]

I, too, have noticed this. Would be curious if anyone has a more in-depth article to read about why the difference is so stark!

[fsflover]

puri.sm/posts/cameras-its-complicated/

[qingcharles]

Can you access the camera, though? I thought full access to the camera sensor was locked down only to awful Samsung apps?

[SchemaLoad]

It's been many years since I did android rom stuff but how it was back then was that yes you can access the camera, but you lose the proprietary image processing pipeline which exists just as software. So your photos come out looking much worse since phone cameras rely heavily on software to look good.

[MrDrMcCoy]

On the other hand, you can take DNG RAW images and do the software processing later, except exactly the way you want. I rather like that flexibility, myself.

[sznio]

>by the time you have it worked out, the phone is obsolete and very few people still use it.

If you work out a phone from 5 years ago, you're not that far off from a phone of today. Nobody designs it all from scratch, you mostly modify the old one. Getting the foundations going will take years - adapting the foundation to a different phone of the same series will only take months.

[wrasee]

Hardware is hard. It doesn’t always have the transparent composability that software has because you hit physics and the real world.

The example already given makes the point. Work has been great on the M1 but my understanding is that this has not translated all that well to e.g. M2, M3 and M4.

[NetMageSCW]

1) The article states they are focusing on the phone model that they guess will require the least work to become totally free. This may make the project useless, but it does give it some hope of finishing.

2) The hope is that the M2-M5 won’t be that different from the M1 models - after all, Apple doesn’t want to spend their money reinventing the wheel without compelling reason. I think that is less likely with phones from different manufacturers, though Android phones typically share a lot of single source components.

[fsflover]

> The hope is that the M2-M5 won’t be that different from the M1 models - after all, Apple doesn’t want to spend their money reinventing the wheel

From the Asahi Linux website, M2 is sufficiently similar to M1, while M3 and M4 won't likely be supported soon due to significant differences.

[zozbot234]

They're aiming to perfect their support for M1/M2 prior to working on the M3 and later models. Seems like a sensible choice, given that even a baseline M1 or M2 Mac is still a highly compelling device for a vast majority of uses. And Asahi will become more relevant as these devices cease to be supported by newer releases of macOS.

[panny]

>a baseline M1 or M2 Mac is still a highly compelling device for a vast majority of uses.

Maybe in 2020. Lenovo released an ARM chromebook this summer which has benchmark performance of M1/M2 chips and is perfectly supported by Linux (ChromeOS) out of the box.

[gradeless]

Yes it will take many years. This whole thing has already played out with FSF and Replicant. They ended up stuck working on a couple of ever aging devices as many new generations of devices were launched and all the technologies in smartphones evolved.

If people want open devices they should maybe better explore open hardware. Im not talking about devices, like Librem where the schematics are open but the chips, which are the parts which do all the work, are all closed, but rather devices with open silicon.

[fsflover]

Can you name such devices? Precursor?

[AnthonyMouse]

> The Apple Silicon macbooks seem a good example. The M1 came out about 5 years ago now and with a whole project and a lot of work later there is still limited hardware support. Having to put this effort in for all the models of phones seems massive.

Apple Silicon isn't a single SoC. They have support for the M1 and M2 lines across both Macbooks, the Mini, iMac and Studio.

It's the same thing with phones, isn't it? You're not starting over for each individual phone. First you get one device working which uses some popular chip, which is a large undertaking. But then all the other devices that use the same chip are nearly working and it's not a matter of starting over, it's a matter of mapping the couple of changes they made to the base design.

Meanwhile the chip designs themselves are incremental, so the 2026 device is really the 2023 device with a die shrink and a new feature.

[wqaatwt]

How is that going for e.g. postmarketOS? Doesn’t seem to be that simple

[j45]

I thought Asahi Linux had made some pretty sizeable progress, it was a very impressive project and labour of love and talent.