3 ms·

Android is Apache licensed. There's a whole thing about the Play Services absorbing more of the features, and I feel like it's a great indicator of the difficu

by rtpg 9y ago

Android is Apache licensed. There's a whole thing about the Play Services absorbing more of the features, and I feel like it's a great indicator of the difficulties with open source. A lot of people enjoy replacing the lowest level of the system with OSS. But consumers only notice that afterwards they have nothing on the highest level. I think the Raspberry Pi took the right approach here accepting the binary blob GPU but having the software stack otherwise be free. They got adoption because things were useful! And now they "just" have to solve the GPU problem, instead of a "rewrite all software " problem. I would be very excited by people who could offer a "skin " over Android (with its own set of apps) so the higher level stuff could be tackled. Once that part of the stack is nice, swapping out the backend will be an even easier pitch. The counterpoints I can think of: - If you do not trust the backend, hard to have proper security. - People are doing what they want to do! Lots of people like writing kernel-y code more than a dialer UI

ZenoArrow 9y ago
> "There's a whole thing about the Play Services absorbing more of the features, and I feel like it's a great indicator of the difficulties with open source." Let's imagine a hypothetical scenario where Play Services were open source. How would this hold back Google from maintaining/growing their market share? As for developing for Android and swapping out the backend later, by that point your apps would be tied into the Android API calls, unless you intentionally make your apps cross-platform, at which point what do you gain from running Android? If it's just for the drivers then libhybris is an option. If you want the ability to run Android apps whilst you build up your own, this is something you can do with Sailfish OS, which you'll almost certainly be able to run on the Librem smartphone if it's released.
- rtpg 9y ago
  My point about Play Services was that, despite Android being OSS, it is not sufficient to just set up AOSP. People still want to use apps, so you have to get a copy of Play Services. But Play Services is a Google blob, phoning home and the works. The higher-level stuff is what "matters" to people. If Play Services were open source, then someone could release forks of that which patch out the questionable parts of it (just like Chromium/Chrome, in reverse). Efforts could be combined on the open source parts of things, so that we don't spend our lives rewriting battery-efficient geofencing services. the gain from running on Android is that you can concentrate on higher-level value-adds first, to build adoption of your stack. You could build out some cross-platform thing too, it's just that Android tends to make things a bit easier for swapping things out. Sailfish sounds like a good option for this, but what I'm reading is that the people making Librem are spending a lot of effort building out a new OS mostly from Linux?
- s17n 9y ago
  Play services exists to give Google a way to deliver updates that isn't subject to carrier or oem interference. If it was open sourced, that would defeat the purpose. To put it another way, giving the general public access to the source wouldn't make any difference to Google, but giving Samsung that access would make a big difference.
hutzlibu 9y ago
"People are doing what they want to do! Lots of people like writing kernel-y code more than a dialer UI" erm, I think yes, those people who can write kernel-y code, probably want to do that more than UI. But when you take the whole crowd of (oss) developers, then all together there are much more people able and willing to write UI-Code, than kernel code ... The bigger problem with unstructured FOSS development is probably more the boring parts, like maintenance, testing and bug fixing ...

by rtpg 9y ago

ZenoArrow 9y ago

> "There's a whole thing about the Play Services absorbing more of the features, and I feel like it's a great indicator of the difficulties with open source." Let's imagine a hypothetical scenario where Play Services were open source. How would this hold back Google from maintaining/growing their market share? As for developing for Android and swapping out the backend later, by that point your apps would be tied into the Android API calls, unless you intentionally make your apps cross-platform, at which point what do you gain from running Android? If it's just for the drivers then libhybris is an option. If you want the ability to run Android apps whilst you build up your own, this is something you can do with Sailfish OS, which you'll almost certainly be able to run on the Librem smartphone if it's released.

rtpg 9y ago
My point about Play Services was that, despite Android being OSS, it is not sufficient to just set up AOSP. People still want to use apps, so you have to get a copy of Play Services. But Play Services is a Google blob, phoning home and the works. The higher-level stuff is what "matters" to people. If Play Services were open source, then someone could release forks of that which patch out the questionable parts of it (just like Chromium/Chrome, in reverse). Efforts could be combined on the open source parts of things, so that we don't spend our lives rewriting battery-efficient geofencing services. the gain from running on Android is that you can concentrate on higher-level value-adds first, to build adoption of your stack. You could build out some cross-platform thing too, it's just that Android tends to make things a bit easier for swapping things out. Sailfish sounds like a good option for this, but what I'm reading is that the people making Librem are spending a lot of effort building out a new OS mostly from Linux?
s17n 9y ago
Play services exists to give Google a way to deliver updates that isn't subject to carrier or oem interference. If it was open sourced, that would defeat the purpose. To put it another way, giving the general public access to the source wouldn't make any difference to Google, but giving Samsung that access would make a big difference.

hutzlibu 9y ago

"People are doing what they want to do! Lots of people like writing kernel-y code more than a dialer UI" erm, I think yes, those people who can write kernel-y code, probably want to do that more than UI. But when you take the whole crowd of (oss) developers, then all together there are much more people able and willing to write UI-Code, than kernel code ... The bigger problem with unstructured FOSS development is probably more the boring parts, like maintenance, testing and bug fixing ...