I wish more devs would try doing ops things.

Here’s a quick challenge:

1. Set up an application server.
2. If any of the following are true, go back and fix it:
   • Not deployed under its own user
   • application user can ssh in
   • application user can elevate privileges (i.e. user can sudo)
   • application logs are not rotated
   • application logs are not in an easily guessable place (/var/log/$app, /tmp/$app)
   • application will not restart itself on reboot
   • application will re-try starting forever if startup fails
   • application stores config in application directory
3. Now set up the same application on a different distro.
4. Now do it without downloading arbitrary things from the Internet (no rubygems.org, no npmjs.com, no pypi.python.org) (Exception: anything content addressed and cryptographically verifiable is OK (A URL over https does not meet this requirement))
5. Run the server without a C compiler installed.

Think the above is onerous? It’s the bare minimum for repeatable builds and hardening.

It’s amazing how many things make the above hard/impossible. And for some projects, hey I get it. It’s a pet project. But coming from an ops background those are literally the first things I look when comparing software, and they’re much easier to keep in mind during development than go back and fix when an Important User needs them.