[jonnytran]

This is simply not true. From 2014 to 2017, I worked at a place where I kept starting new projects in Python 3. But people with less foresight wouldn't let their Python 2 habits die. It was a constant struggle to get anyone to realize that Python 3 was the future, and the majority of the code ended up using Python 2. In 2019 when the end-of-life for Python 2 was finally announced, I knew of companies scrambling to upgrade.

I can't help but contrast this with the upgrade from Ruby 1.8 to 1.9. It was also painful, but no one in the community was holding on to 1.8 years later.

[pansa2]

> I can't help but contrast this with the upgrade from Ruby 1.8 to 1.9.

One difference is that upgrading Ruby from 1.8 to 1.9 brought a significant performance increase, whereas going from Python 2.x to early 3.x, performance actually got worse.

AFAICT this was mostly caused by the removal of the machine-word-sized integer type - in Python 3, even 1+1 is calculated using arbitrary-precision integers.

Python 2 => 3 had lots of other problems as well - ultimately they changed just enough to break everyone’s code, but not enough to make upgrading worthwhile.

[tonyedgecombe]

I've often wondered if part of the problem was supporting Python 2 for so long and hence prolonging the pain.

[rightbyte]

Without that I think many would lose trust in Python and just switch language.

I mean it has only been 11 years since Python 3.1/2.7 and that's probably a common lifespan for maintenance mode code projects? 3.5 is still supported and that one is 5 years old. Why the hurry.

[tonyedgecombe]

Why the hurry.

Because some people will always leave it to the last moment or beyond. Meanwhile the Python team has had the overhead of supporting more code than necessary.

[xiaodai]

What r they gonna switch to?

[newen]

Python is not technically superior to other languages enough that you can rhetorically ask that question. It’s main advantage is the ecosystem and network effects. If a bunch of people, especially the people who work on numpy, scipy, etc., decide to work on developing libraries for other languages like R and Julia, the data science ecosystem would switch over in a few years. Similarly for other fields people might switch to languages like Elixir, Haskell, OCaml, Go, Swift, Scala, Ruby, Kotlin, etc.