New ask Hacker News story: Breaking Changes in Python-3.14
Breaking Changes in Python-3.14
4 by wef | 1 comments on Hacker News.
One of the reasons Linux is so successful is Torvald's insistence on never breaking user space. The python people don't adhere to this principle and maybe I need to give up on it, I'm just sick of this crap. Those of us of a certain vintage will recall the tumult of python-2 to -3. Suddenly, production code needed significant re-writing. We got an arguably better python out of it, but oh! the pain. In 3.14, (among many other things) the python developers decided to make 'forkserver' the default instead of 'fork' for the Process() method (this is for starting a child process - https://ift.tt/EocrFtJ). Why on earth break our code in such a wanton way? Why not leave the default alone - there was always the option to use 'forkserver' if one wanted it. Or maybe they could have created a new entrypoint with the new behaviour Process_fastserver() or some such? Oh no! Just break it and make their customers patch furiously! When we adopt a language, we like to think that what runs today will run tomorrow - C and bash programs that I wrote 30 years ago still run. Not with python - if you use it, buckle up and make sure your regression tests are thorough, it'll be a rough ride. Move slow and break things, perhaps?
4 by wef | 1 comments on Hacker News.
One of the reasons Linux is so successful is Torvald's insistence on never breaking user space. The python people don't adhere to this principle and maybe I need to give up on it, I'm just sick of this crap. Those of us of a certain vintage will recall the tumult of python-2 to -3. Suddenly, production code needed significant re-writing. We got an arguably better python out of it, but oh! the pain. In 3.14, (among many other things) the python developers decided to make 'forkserver' the default instead of 'fork' for the Process() method (this is for starting a child process - https://ift.tt/EocrFtJ). Why on earth break our code in such a wanton way? Why not leave the default alone - there was always the option to use 'forkserver' if one wanted it. Or maybe they could have created a new entrypoint with the new behaviour Process_fastserver() or some such? Oh no! Just break it and make their customers patch furiously! When we adopt a language, we like to think that what runs today will run tomorrow - C and bash programs that I wrote 30 years ago still run. Not with python - if you use it, buckle up and make sure your regression tests are thorough, it'll be a rough ride. Move slow and break things, perhaps?
Comments
Post a Comment