I've spent just about every single day for the last 6 months doing something with Python 3. Some days it was helping port a library, other days it was helping projects put together their porting strategies, and on others I've written prose on the subject. At this point, I am very very bored of talking about porting, and about the health of our ecosystem.
Most of all, I'm exhausted, particularly from arguing about whether or not the process is going well. So here's what I would like:
I would like to know what the success condition for Python 3 is. If we were writing a test case for this, when would it pass?
And let's do this with objective measures. Here are some ideas I have:
- Percentage of package downloads from PyPI performed with Python 3 clients
- Percentage of packages on PyPI which support Python 3
- Percentage of Python builds on Travis CI which featured a Python 3 builder
I'd like a measurement, and I'd like a schedule: "At present x% of PyPI downloads use Python 3, in 3 months we'd like it to be at y%, in 12 months we'd like it to be at z%". Then we can have some way of judging whether we're on a successful path. And if we miss our goal, we'll know it's time to reevaluate this effort.