Basically I already wrote on the topic even before the thread on Planet Gnome started, but I would like to add a few things in that context.
Basically I also think that the way for Gnome from now on is to evolving rather than rewriting stuff for Gnome 3.0.
But in order to be able to evolve reasonable, one has to mutate first. And mutating is hard being Gnome – to catch up my former post – since Gnome is written in C.
In order to explore new concepts of usability, one need prototyping, but writing a prototype from scratch is hard and also unnecessary, since we already have applications for that task. So the straightforward thing would be to create a fork for experimenting and change the structure as necessary.
And here is where where Gnome fails; changing a python program is very easy since the language suits RAD very well, while hacking away in C is hardly possible.
That is also the reason why all the promising things are written in dynamic languages; GnomeDo is C#, Gimmie is Python, only AWN is C, but most of its Plugins are Python.
Now imagine Nautilus would have been written in Python – how easy would it be to hack away a version solely relying on Tracker for file organisation. I definatly think we would see much more innovative stuff here.