Member
I name my deploy scripts after the incident they caused. `friday_afternoon_2026_08_panic.sql` is very descriptive. The rollback is in the same folder, named `friday_afternoon_2026_08_panic_UNDO.sql`. Self-documenting.
42U cabinet update: added a second UPS, now drawing 13.1A peak. I have started naming my photos with the exact model number. `eaton_9px_3000_2026-08_rear_c13.jpg`. The filename is the spec sheet.
Member
X86 is dead but x86 documentation conventions live on. ARM servers have the same problem. I have seen `final.s` and `final_v2.s` on Ampere machines. The ISA does not fix human nature.
Descriptive paths break down when the year turns over and you are still using the same playbook. `2026/04/ssl-renewal/playbook.yml` ran in August because nobody moved it. The path lied. Sam
Member OP
We solved the year problem by dropping the year from active paths. `current/ssl-renewal/playbook.yml` and a symlink. Archive the old ones. The filename stays honest.
Member
I have a client whose entire network documentation is in filenames. `switch-core-01_48p_gigabit_2019_firmware-bug.cfg`. The bug was fixed in 2021. The filename still warns you. I left it.
Watts are the real currency and so is my time. I name my Ansible playbooks for the power state they achieve. `idle_12w_ubuntu2404.yml` versus `stress_87w_ubuntu2404.yml`. The numbers are from my meter at the wall.
This is why I just use git lah. The commit message is the documentation. `final_final_USETHIS` becomes a joke in the blame view.
Git does not help when you are staring at a directory listing at 3am because production is down and you need the right file now. WHOIS never lies. A domain expires or it does not. Binary state.
I worked at a place that enforced a filename linter. Jira ticket number, short description, date, author initials. `PROJ-1842_nginx-ssl_20260828_ec.txt`. It worked. People hated it. It worked.
