they put a lot of work into it but were willing to sacrifice it once they realized the rest of the community was going in a different direction because they can get the same functionality from systemd
So what's different with Mir? They put a lot of work into it (check), the rest of the community is going in a different direction (check), they can get the same functionality from Wayland (check).
It's not. The protocol is extensible, there's nothing they can do with Mir that they couldn't do with Wayland. Protocol not support some Unity-specific feature they need? Write an extension. Simple as that. It's a great system, because if it turns out the extension is something that would benefit other DE's, it can be absorbed into the main protocol. If the extension is something that no one else has use for, they can just keep using it as Unity-specific extension, and since no one else has use for it, it won't cause any compatibility problems either.The main reason they don't want to adopt Wayland is not because there's anything wrong with it, they just don't like the fact that shell behavior is defined by the protocol.
Wayland is as flexible as you can get. You don't get more flexibility by reimplementing everything yourself, making yourself incompatible with standards everyone else uses. All you get is isolation, reinvention of the wheel, reimplementation of things that shouldn't have to be reimplemented in the first place. A lot of pointless repetition of work for no real gain. In other words: wasted resources. Which is something a small company like Canonical definitely should not do. If they were as big as Google or Apple, I could understand how Mir would make sense for them. I still wouldn't like it, but I'd understand it. As it is, they're just shooting themselves in the foot.They want something that is more flexible; My assumption is because they fear being limited in some way.
I feel like this is a good time to quote again my friend, Past Mark Shuttleworth:I'm not going to pretend to know exactly what they are concerned about, and they could certainly be wrong, but since they clearly have competent people working there (just look at the Debian mailing list debate, everyone who commented on Upstart praised the code quality even when voting for SystemD) I give them the benefit of the doubt and assume they know what is best for their platform. I think the init system situation demonstrates that they are willing to work with the community but not if they think it will hinder what they are trying to accomplish.
Those were good reasons. That actually made sense. So what's changed now?Originally Posted by Shuttleworth 2010