I'm currently creating 2.7/3.3+ compatible code in whatever I create, but I'm still deploying to 2.7 hosts. For most of my projects this is due to just one package which has some Python 3 related bugs, so hopefully I can just flick the switch soon and use 3.4.
It's only easy when your project depends on a few popular packages (eg. the top 200 where 86% are currently Python 3 compatible [1]), it's much more difficult if you've got a long list of dependencies.
If deploying a .com website, you can control your dependencies completely, and it's much easier than if you are deploying a system that must interact with existing systems the user/customer might have (for instance, management or monitoring apps).
Thankfully that's not a huuuge class of systems, but it's what I typically work with :)
As you say, it's not a problem for most of us as we get to dictate the platform.
See kaleidoscope of examples:
- https://www.google.com/search?q=site%3Agithub.com+compat.py+...
- https://github.com/search?utf8=✓&q=extension%3Apy+"if+PY2"&t...
[1]: http://lucumr.pocoo.org/2013/5/21/porting-to-python-3-redux/
As someone writing applications using Python 3, I found the switch to be easy and I haven't had any problems with libraries. I've switched from some of the old ones I was using to more modern alternatives that are being actively developed.
It annoys me that there are people on Python 2 that haven't really tried to make the switch and stand on the sideline shouting insults at Python 3. It's a tremendously self destructive thing to do for all involved.
To anyone looking to get started with Python, there's no real reason to start with Python 2 these days.