hot take: running Linux Mobile with a proprietary Android compatibility layer is better than running (any downstream) Android.

https://mastodon.social/@verdre/111612437079067480

@cas

I have very mixed feelings.

On one hand it might help get out of Android+iOS duopoly and push for a FLOSS alternative.

On the other, we might be killing any chance of a community led app platform.

Why use GTK/Qt/libadwaita whatever if I can make my app with the Android stuff and ship it to hundred of million of users?

@sonny

(E)mbrace - Android app support
(E)xtend - better integration and UX for native apps (we need proper runtime power management!!)
(E)xtinguish - "ew why would you make your app on that old stack"

ok I'm not entirely serious, but I really feel like this is a logical step to take, if you can replace the core system without breaking UX, you make the learning curve for moving to better native apps a lot more comfortable. It's less of a scary cliff dive to make the switch to Linux Mobile.

@cas even better: running Linux Mobile with a fully open-source Android compatibility layer.

Even even better: not relying on any proprietary Android app (I know, I'm daydreaming, I soo wish the world wasn't as f***ed up as it is)

@cas grapheneos is looking to transition to this model eventually although i believe they actually have their own kernel planned

@hipsterelectron that makes sense. I think embedded Linux+Android container is more production ready for embedded than native Android

@cas i don't really consider android an open source project like most code from google so (1) i'm not surprised that experts don't want to rely on it (2) i think it would benefit society to extricate itself from that restricted base system. i'm very hopeful that technology to emulate these corporate systems continues to improve rapidly

@awai @cas Hint hint: implementing things like MPRIS and notification proxy in Waydroid isn't hard at all.

@hipsterelectron Android is very much "source available" rather than OSS/FOSS I agree.

@cas imo, Linux mobile actually manages to look more comfy as well (Tho, tis be just personal thing)

And very cute weee

@cas @sonny kinda wondering, if this would also allow for some apps like signal (without a desktop connection) to work

@razze @sonny pretty sure it would yeah

@sonny @cas To add to that, none of the companies that tried this so far managed to migrate developers or gained any marketshare from it or even stayed around

@alatiera @sonny @cas We're not a company though (i.e. there's no single entity that can just pull funding from one day to the next), and we have the advantage of not being limited to phones. A big part of the success of Libadwaita is that you get mobile basically for free with desktop apps :)

@tbernard @alatiera @sonny libadwaita is really really cool. I'm really frustrated with the lack of consistent fractional coordinate support in GTK, it seems to be the source of a whole host of UX issues I didn't even consider - like the cursor handle droplet thing being really hard to use.

a super offtopic and maybe naive question/grip: will we begin to move past the whole "scale factor" thing with gtk5? like why aren't we just using units based on physical dimensions rather than pixels? A "scale factor" is just nonsensical....

@cas the problem with "ew why would you make your app on that old stack"

Is that Android will always have an infinite amount of money to pour into Android developer experience, advocacy, documentation etc

So it's most likely always going to be better for most people.

Perhaps we just have to wait for the enshitifcation there, like Windows. But we'll have to double our effort on DX.

Would be pretty funny if the community managed to pull a EEE :)

@cas so far EEE in FLOSS feels more like

Users embrace
Contributors extend
Industry extinguish

@sonny @cas It probably is the same dangerous slope as with #IBM #OS2 back then which could run #Win31 and #DOS apps perfectly - why should anybody invest into native OS/2 applications?

On the other hand I would love to switch to a #Linux phone but without all the proprietary banking apps etc that is simply impossible.

So I would say: #Android compatibility is key for #MobileLinux to succeed at least somewhat.

@hnygd @sonny @cas Hmm, I think that those are 2 different scenarios. OS/2 was a proprietary OS courting proprietary commercial app development. IE, it was all about profit motive.

I don't think that really matches the reason people get into open source software, nor has the availability of Wine stopped the Linux application ecosystem.

@dperson @hnygd @sonny its a chicken and egg problem. And Android app support is a cloning lab.

the key is just educating new users, sure it's great that your banking app works, but wouldn't it be better if your bank supported the open banking protocols and you could use a native app with the same features?

We need to actually have a big enough userbase to exercise collective power over corporations, and that's the point where we start getting native Linux Mobile apps.

the users wont come without the apps, and the apps won't come without the users. Apple get this, that's why we got Rosetta (and rosetta 2), Valve get this, hence Proton.

but hey, imho none of this is even relevant because we're yet to provide a remotely production ready base OS, probably pureOS on the L5 is the closest to this right now...

@cas @tbernard @alatiera @sonny I don't agree with basing UI off of physical size personally. Scaling should always be a user-facing setting. The default scaling on most mobile user experiences is set way too high IMO. This applies to Android and mobile Linux. I have long tweaked build.prop to lower DPI on my old Android devices but newer Android devices make it a setting usually. Same with pmOS Phosh on my OnePlus 6T. 300% default is way too high. 200% is pretty good IMO.

@CalcProgrammer1 @tbernard @alatiera @sonny yah right i didn't phrase that super well. UI should use dimensions which correspond to physical dimensions, you need additional ways to tweak this based on form factor and user preferences of course. Android let's you adjust the DPI and font size both of these are sensible controls for sure

@cas @tbernard @alatiera

For an app toolkit you want scale factor relative unit for most primitives. That's what everybody does too.

What is the cursor handle droplet problem?

@sonny @tbernard @alatiera the widget sizing is calculated with fractional coordinates but then it's rounded towards zero (cast to int). The pointer coordinates are also handled only as integers, and maybe some other calculations are wonky due to this. It makes them annoying to grab and use on the op6.

I hacked some of this up in gtk3 and it seemed to help, I think some improvements can be made similar to

https://gitlab.gnome.org/GNOME/gtk/-/merge_requests/6668

@sonny @tbernard @alatiera this introduces some new bugs but as a reference for some of the other accuracy issues i found .

https://gitlab.gnome.org/calebccff/gtk/-/commit/ff17e66d5ba797a17a33a02e858d3a8c6669d5a5

@cas @tbernard @alatiera @sonny Also, one of the biggest draws of mobile Linux to me is the seamless blending of desktop and mobile applications on one device. I actually had set up shortcuts to set the scaling from 125% to 200% in my previous pmOS install so I could easily scale down the UI to run desktop applications that needed more real estate. I honestly would love this added to the quick settings in the phosh drop down top bar.

@cas @tbernard @alatiera @sonny Also, that said, I don' think the onscreen keyboard should be tied to the UI scale factor especially on mobile Linux. I want to be able to scale down the UI but keep the keyboard the same size (preferably a separate scale setting). I use a keyboard layout with arrow keys as I've always hated selection droppers no matter the OS. I wish shift+arrow select worked in Phosh. Hacker's Keyboard on Android has its own size settings.

@CalcProgrammer1 @tbernard @alatiera @sonny yeah i plan to patch this in phosh-osk-stub. it should be based on the physical display size.

@cas @dperson @hnygd @sonny

I think there should be some plan for mobile linux. Something like (for a start):

1. Let's pick couple of devices.
2. Make basic things work flawlessly (calls, sms, mms, mobile data).
3. Continue relying on these above.

What I see now is - hey, we support new devicea. And when you dive into details - it means you can boot them. But they lack basic functionalities.

@as400 @dperson @hnygd @sonny yeah it's just really easy to fall into bad patterns here. The postmarketOS core team (for example) can't dictate what some contributor works on (obviously) and we don't dictate what folks in the team work on either. We work on what we care about.

We would all burn out if we went for this approach flat out, it's not what the project is about.

I can't speak for the entire community, but i can say this: postmarketOS is for everybody, our primary goal is to bring freedom to as many people (=devices) as we can. Anything we do to "bridge the gap" to the rest of society, such as trying to build some more polished experience by focusing on one or two devices cannot impede this core tenet, otherwise we're no better than any other alternative OS.

@as400 @cas @hnygd @sonny That has more or less taken place Librem5, PinePhone(Pro), & OnePlus 6(t) all have really good support and the remaining pieces are being worked on.

Above listed in level of maturity of support.

@cas @tbernard @alatiera

I see. I understand the frustration but

* GTK3 is pretty much deprecated
* GTK4 was released years ago
* GTK is not a shell toolkit

I think it's a bit unfair to hold GTK responsible for Phosh being stuck with GTK3.

Not to say that there aren't improvements to be made with GTK 3/4/5 👍

@sonny @tbernard @alatiera scroll slowly in the GNOME settings app on your OnePlus 6

@dsfgs yeah not only are we stuck with proprietary firmware, unless we get OEMs to release their private keys we can never replace it on most phones.

@dsfgs lol, ok lets try and do good advocacy without stepping into deranged conspiracy shite. have a good one

@as400 @cas @dperson @hnygd @sonny There's plan and we have such devices. "PinePhone" is safest one. Then there's PinePhone Pro, Librem 5, OnePlus 6, AFIACT. That's rather short list. Rest are ... "here be dragons". Sounds like you want PinePhone :-).

@pavel @sonny @dperson @hnygd @cas

No Pavel, my pinephone journey has ended long time ago :) Librem 5 price is mind boggling.
So I have OP6 because of that. I managed to buy a new one some time ago. Really nice piece of hardware.

@as400 @pavel @sonny @dperson @cas

What about @postmarketOS on the newest @Fairphone? Does anybody have any experience with it?

I own a #PinephonePro but the hardware is not very powerful to use it every day in my opinion (even though I love the idea behind all the @PINE64 products)

#Fairphone5 #Fairphone

@as400 @sonny @dperson @hnygd @cas You would not be happy with Librem 5, I'm afraid. So you moved from working PinePhone to more experimental OP6. Well, of course it is broken :-). Fix it...

@hnygd @as400 @sonny @dperson @cas @postmarketOS @Fairphone @PINE64 FP5 may be the future as vendor is putting resources into it. But I'm afraid it is "future" at this point.

@hnygd @as400 @pavel @sonny @dperson @postmarketOS @Fairphone @PINE64 lots still missing, but most of the core functionality is available - including display out over type-c!

@as400
I don't trust @purism
I bought a Librem laptop. Hinges fell apart in 2 years, and there was no support.
Not enough a discount on next model.
Hoping form@calyx@calyxos@fosstodon.org on my #Fairphone5
@pavel @sonny @dperson @hnygd @cas