> (L)GPLv3 is out of the question in all circumstances - it scares the phone industry so much that we'd be hurting the entire Android ecosystem if such code made it anywhere into the Android Open-Source Project.
It scares the phone industry... yet... the overwhelming majority (if not all?) of the code in Maemo and Moblin is (L)GPL ? Oh wait, I also forgot the Palm Pre which uses (L)GPL stuff... oh and the Amazon Kindle... and the list goes on. Anyone said FUD ? :)
The biggest problem, having worked on making widely-used open-source middleware (like GStreamer) compile/work/integrate with Android... is that Google (or other partner of OHA) have gone out of their way to rewrite virtually all of the software stack, and what should have been a trivial task ends up being a never-ending nightmare.
What do users want: A shiny slick experience (1)
What do phone application developers want: A consistent environment (2)
What do phone manufacturers want: Be able to (re-)use the experience of their in-house/contracted engineers so they can ship a device as soon as possible (3)
(1) : Android is slick, no problem there
(2) : The Android java layer works accross several systems... well... if you don't try to use features from new phones that required creating "additional" APIs instead of extensions to existing APIs (which were not thought to be extended)
(3) : Total failure. Everything from kernel to the Java layer is unlike any other system. It's not a standard linux kernel, not a standard glibc, not a standard IPC system, not a standard multimedia framework, not a standard I/O system, not a standard ... oh man... I'm lost here. It's just a completely new software stack. So all your engineers/contractors that have got all this knowledge of existing LSB ... have to start again from scratch.
How to solve this ? Somewhere along the lines of :
Keep (1), Improve (2) and keep it backwards compatible, Fix (ditch?) the softare stack below Java by using the existing work done by Maemo, Moblin, etc...