![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Из комментов реддита
Jan. 10th, 2024 12:08 pmThe dirty truth about any integration (including CI, CD, packaging) is that this is the place where mutually unsound contradicting abstractions are get glued together with bash and some pinky promises never go into contradicting areas.
You can be a vigilante of strict typing, but at the end you still agree that foo can be either `foo: true` or `foo: "False"`, and there are greater reasons for it to be this way.
Just yesterday I argued about the code where 'raid_types' was set to 'perc', and it worked, because in some other place it was `if "perc" in raid_types`, and with Python duck-magic it was true if it was 'perc' and not ["perc"].
Moreover, if you say that this is wrong it all should be strictly typed and invalid states should be non-representative, oh... you will end with sed patching your non-representative code to present invalid sate because it's the single way to make this goddamn thing work.
As soon as someone finds a island of stability, it become a fortress and a tool, and everyone start to use it, but around it a new quagmire of opportunistic typing appears gluing together ungluable, and that where the most work is happens, because sound things are nice, they just work, and they don't need human attention. All human attention should be on the dirty unsolvable problems, and the more dirty unsolvable problem it is, the more human attention it gets, so 99% of the devops work (by time) is dirty unsolvable problems. Also, 99% of devops self-esteem comes from 1% of well solvable beautiful constructions which you admire for a second. And then you go to debug for next 4 hours why there is a disappearing end '\n' in the certificate key file causing changes on convergence.
You can be a vigilante of strict typing, but at the end you still agree that foo can be either `foo: true` or `foo: "False"`, and there are greater reasons for it to be this way.
Just yesterday I argued about the code where 'raid_types' was set to 'perc', and it worked, because in some other place it was `if "perc" in raid_types`, and with Python duck-magic it was true if it was 'perc' and not ["perc"].
Moreover, if you say that this is wrong it all should be strictly typed and invalid states should be non-representative, oh... you will end with sed patching your non-representative code to present invalid sate because it's the single way to make this goddamn thing work.
As soon as someone finds a island of stability, it become a fortress and a tool, and everyone start to use it, but around it a new quagmire of opportunistic typing appears gluing together ungluable, and that where the most work is happens, because sound things are nice, they just work, and they don't need human attention. All human attention should be on the dirty unsolvable problems, and the more dirty unsolvable problem it is, the more human attention it gets, so 99% of the devops work (by time) is dirty unsolvable problems. Also, 99% of devops self-esteem comes from 1% of well solvable beautiful constructions which you admire for a second. And then you go to debug for next 4 hours why there is a disappearing end '\n' in the certificate key file causing changes on convergence.