Google will kill Python 3, and it might be a good thing

We know that there are issues with the Python echosystem at large [1]. Google is heavily invested in Python 2, and thus created Grumpy [2], which is a new Python 2 runtime, in Golang. A company like Google, stepping in, will inevitable lead to the sorting out of many of the issues [1], and introduce new developments for Python 2. The Golang backend of Python will enable the development of many benefits we see in Golang (built in concurrency?, speed?, easy package management?) This would mean that the community will gravitate towards python 2. What do you think?

[1] * Fragmentation of package managers: Pip, Conda, etc.. * Fragmentation of runtime: Pypy, CPython, Jython, Cython, etc.. * Golang (and probably Grumpy in the future) doesn't suffer from these issues, because a central entity (Google) is leading it.

[2] https://github.com/google/grumpy

How do you explain the fact TensorFlow targets Python 3 for speed, features, etc.?

Google isn't going to "kill" Python 3. This is a relatively niche use case (grumpy only converts Python 2.7 code into Go code, and it's not 1:1 as it's missing features like decorators), and there doesn't seem to be any intention on adding new Python features into grumpy. My guess is that this was made more for Google's purposes of migrating some of their legacy Python 2.7 code to Go, in an effort to move towards Go internally. It also doesn't run Python code directly, it trans-compiles Python to Go code, which wouldn't work for some workflows. For this to have any possibility of becoming another Python runtime, it would at the very minimum have to support every feature or have a real reason to not support the specific feature in Python 2.7.

The other thing, arguably more important, is that there isn't really a community around grumpy, whereas there is around Python 3 (the latest nontrivial commit was 3 weeks ago for grumpy, whereas the latest nontrivial commit was 6 hours ago for Python). Even if Google did try to build a community, many people worry that Google will just abandon software when they want to, so people would be hesitant to work with this software long term, which means developers wouldn't want to build any software for it; the majority of developers developing new Python code are doing it in Python 3, and those that aren't are most likely using six rather than only 2.7 unless they're supporting a legacy codebase that either they don't want to or cannot upgrade to Python 3 or to upgrade using the six library.

Python 3 may, however unlikely, "die", but this and Google won't be the death of it.

I'm with you. Also, I don't see that Google is going to make a difference with packaging, they have very particular use cases.

Grumpy will primarily make a difference for people who think they're Google, I can't imagine it'll be widely adopted.

Google culture is all about arrogance and elitism. Their coder/snobs are definitely not pro python any version. They have their own internally developed ivory tower languages with horrible tooling and features to champion.