The most reliable Linux phone I've ever had was a Nexus 5 running #UbuntuTouch. After giving it a lot of thought, I can't help but think that #Halium is the best path forward for #LinuxMobile. Several projects have tried to make calling and audio routing as reliable on Linux phones without Halium as on Android or iOS, but it just isn't happening. (1)

As I see it, making and receiving calls and feeling confident about the quality and stability of those calls is non-negotiable. I believe this remains one of the largest obstacles to the broader adoption of Linux phones. Projects like #postmarketOS, #Mobian, and #PureOS are simply amazing. Each have clear values and focus; whether to reduce waste or run only free and open source software. (2)

I worry, however, that hardware is developing too quickly. Upstreaming drivers is a short term solution and will never get Linux phones into the hands of millions. (3)

This is why I wish Halium played a larger role in the Linux Mobile world today. Ubuntu Touch and #Droidian have demonstrated how it can be done. Maybe the final answer isn't Ubuntu Touch or Droidian. They have both experienced their own set of challenges. Financial support for these sorts of ventures is always in short supply. (4)

All in, I would venture to guess that the wider adoption of Linux phones will arrive on the shoulders of Halium or some distant fork of Halium. Let us leverage the mobile hardware drivers that the multi-million dollar companies spent years developing, testing, and fine-tuning. Get Linux into the hands of millions. Build security features to protect against the drawbacks of older kernels. This is the way. (5)

@williamtries I don't think it's necessarily Halium, I believe it's something else: The fact that Ubuntu Touch is the OG of being immutable; and their release (testing) model which means you don't have random crap err new releases land on your phone all the time and even the few updates in a LTS distro get additional testing in the release candidate channel to avoid regressions (e.g., by fixing audio configs).

@williamtries

I'll half agree in that I believe that #Hallium will deliver daily driveable Linux phones long before mainline Linux phones do because drivers take (a lot of) time.

That said, from the sounds of it, Hallium is hacky and unlikely to provide the most reliable long term solution.

Linux phones don't need to keep up with the latest hardware to become popular, just to provide a usable phone experience.

Eventually mainline will get there.

#MobileLinux

@williamtries

The way I see it is that the most functional ecosystem provides a number of steps from very pragmatic, minimally ideal through to near ideal. For example once upon a time proprietary drivers were 100% necessary for Linux accelerated graphics. Over time mainline caught up with all major players now having or planning to have FOSS drivers.

We needed proprietary drivers, *and* to keep pushing to get past them.

#Hallium seems similar.

@williamtries No, it is not best path forward. It is good path towards short term usability. It does not work long term.

@williamtries I kinda disagree here. There are definitely big issues to figure out with kernel upstreaming (e.g. all the "downstream drivers" we carry in forks that nobody wants to rewrite), but every year Qualcomm's Android BSP gets closer and closer to upstream, compared to a decade ago pretty much everything "just works" on new SoCs, you just need to fight with the display, touch, and audio... Compared to 6 years ago this is huge.

I think the major things we're missing are stability (kernel regressions affect us, unlike on halium where the kernel is already 3 years old), and we don't do immutable (yet).

I won't pretend call audio isn't a huge glaring issue, unfortunately the audio stack for Qualcomm is not in fantastic shape, though progress is being made.

tl;dr: while halium projects try to deliver something usable NOW, postmarketOS and related projects are working on Linux mobile for the future.

@Blort @williamtries I honestly believe that once we make some progress on de-duplicating the kernel rebasing and version bumping efforts, set up automated testing, and start shipping immutable images the number of regressions we see will really start to drop. It's a losing game right now especially on postmarketOS edge since a random pipewire release could just break stuff and we'd only know about it because someone hits the bug and reports it...

@williamtries There are other issues, some quite annoying, sure, but I haven't had any troubles with calls and call audio on my Librem 5 with PureOS for years now, so I don't think it's as hard as you make it seem. Unless you reach for bleeding edge (I don't see PipeWire stacks as ready for mobile just yet), this is a solved problem; and when you do, there are still some low hanging fruits to reach for in order to solve it.