GVR said to expect a five year migration. We are four years in. Major projects like NumPy have had support for a while. Pyramid supports it. Django has good enough support that library authors can start migrating their changes. Python 3.3 took out a major source of 2/3 conflict by putting u"" strings back. It would not surprise me to see a lot of newly updated code be 2.7/3.3 or greater as a result, but it will hasten the conversions.
All in all, with a year left in the migration path, I'd say we are doing OK. The VM is in good shape for production use and the libraries are coming along well.
I think the biggest danger is all the hand-wringing about how everything has gone wrong, who didn't read the fine-print to start.
A question I have is how the duck can a language like C# introduce TONS of changes from 1.0 to 5.0, including major syntactic, semantic and library features, and still be widely adopted, whereas Python has to have a "five year migration" for what are essentially some minor language changes.
They were not minor changes, though they might appear to be so from the outside looking in. The unicode changes alone are significant enough to require a significant refactoring of existing code - due to the monkeying around that was required to properly deal with Unicode in 2.x.
All in all, with a year left in the migration path, I'd say we are doing OK. The VM is in good shape for production use and the libraries are coming along well.
I think the biggest danger is all the hand-wringing about how everything has gone wrong, who didn't read the fine-print to start.