> http://download.microsoft.com/download/a/0/8/a08f801c-ce61-43ae-b31b-27df34

Some ideas I got watching that video:

Most of their design is similar to those of P/S/G project.

Doing a web interface can be interesting
I  specially liked that part of XML/XSLT , etc. Technology but this is a long time request for P/S/G IIRC. I think we can start looking at it very seriously, I'm willing to help programming it.

I also liked the similar to writelog /readlog idea just for commands. So it's possible to execute a txt file with those commands or changing them a bit and execute it, without compiling. I'll work on that soon (Ideas of how to best approach the problem are accepted).

They are very proud they have events notifiers instead of polling, It is very possible to do it with P/S/G. I guess they just have something like this:

install_notifier() //the client data will go ALSO to the notifier
reserve_new_notifier(proxy, event, state, function)
reserve_delete_notifier(proxy, event, state, function)

  for all the proxies we are monitoring
    if the proxy have the event 'event' and we are in the state 'state' call function

I'm not very fond of Microsoft but I agree with Mirko that we must take a close look to whatever idea seems interesting and implement it in P/S/G.

About their physical simulation the only differences with Gazebo are:
- I don't know how their models are done, if they are very realistic, they surely are difficult to make
- They have a pretty interface, a lot of eye candy (compare how they render laser and how is rendered in Gazebo).
- They use Agiea SDK instead of ODE so they have hardware acceleration. As Xavier said, this generation of Agiea cards aren't  that good anyway.

And finally I envy they have so many interfaces for robots (mindstorm and kondo por instance)  :P