[update: Mir page removed most of the reasons Wayland wasn't suitable, so why did they not use wayland again?]
[update: still my opinion, really, nobody is making me say shit, lwn commenters really like to believe I've got a hand up my ass]
Okay I'm going to write a short piece on why I believe Mir isn't a good idea. If you don't know what mir is then don't bother reading the rest of this until you do.
So lets take a look at Mir from a cynical pov (I'm good at that): Say this is nothing more than a shallow power play by Canonical to try and control parts of the graphics infrastructure on Linux. It must be really frustrating to have poured so much money into a company and not have 100% control over all the code the company produces and have the upstream community continually ignore your "leadership". This would leave you wanting to exert control where you can and making decisions on what spaces you can do that in internally.
So in order to justify the requirement that Mir is required by the community at large above the current project in the space, Wayland, it is necessary to bash wayland in order that your community can learn the lines so they can repeat them right or wrong across the Internet. So you post a page like this
and a section called "Why Not Wayland / Weston?".
Now I've been reliably informed by people who know, that nothing in that section makes any sense for anyone who studied wayland for longer than 5 mins a year or two ago, especially the main points about the input handling. Nobody from Canonical has ever posted any questions to wayland mailing lists or contacted Wayland developers asking to support a different direction.
So maybe I'm being too cynical and Hanlon's razor applies, "Never attribute to malice that which is adequately explained by stupidity".
Now the question becomes do you want the display server that you are going to base the future of the Linux desktop and possible mobile spaces on a server written by people too stupid to understand the current open source project in the space?
The thing is putting stuff on the screen really isn't the hard part of display servers, getting input to where it needs to go is, and making it secure. Input methods are hard, input is hard, guess what they haven't even contemplated implementing yet?
Valve? NVIDIA? AMD? I'd be treading carefully :-)
(all my own opinion, not speaking for my employer or anyone really). Probably should comment on the g+ threads or lwn or somewhere cool.