| I really don't understand this stance. I try to keep current on Wayland implementation work and was around when the Mir debacle was originally going down (and prior to it, when Canonical stated they would contribute to and use Wayland!). Writing a competing implementation when you could simply (and announced you would!) contribute to an existing open-source project, then publishing articles with misleading and vague FUD (about the Wayland input stack, which is really just the upstream Linux input stack), when the only reason not to collaborate is to shipping your phone spin quicker while reusing closed-source drivers... is plenty to deserve the response Mir got. It reeks of a get-rich-quick scheme that will divide the graphics infrastructure work on Linux in half while introducing unmaintainable code stuck to an old, frozen, unfixable design. Letting closed-source blobs dictate the design decisions of open-source products results in crippled products and unresolvable technical debt. If Canonical did not have any intent to profit from deploying Mir, they could have taken the time to work out with the Wayland developers exactly how to make Wayland flexible enough to suit everyone's needs. I'm still not convinced that it isn't, but I'm not an expert on Android graphics infrastructure and there may be some technical hitches. In particular, see the following (in rough reverse chronological order): This statement of technical reasons it isn't possible to correctly (every pixel in every frame on screen is deterministic) implement a Wayland environment on top of libhybris:
http://web.archive.org/web/20141203070549/http://www.jlekstr... Articles and demonstrations about Wayland compositors running atop libhybris:
https://mer-project.blogspot.com/2013/05/wayland-utilizing-a...
http://ppaalanen.blogspot.com/search/label/android I guess Mir just gives up on the goal that "every frame is perfect". The right thing for Canonical to do would have been to pour some of its significant capital into funding reverse-engineering efforts to develop open-source drivers for Android hardware. This would benefit the GNU/Linux ecosystem, the Android ecosystem, and device vendors (who wouldn't have to act as middle-men between GPU vendors and consumers for software updates). And we'd get features that are considered impossible today. But the huge cost of GPU reverse-engineering wouldn't have helped Canonical's bottom line. |