Hacker News Reader

Comment by avodonosov | Hacker News Reader

  1. Bring everything down?
  2. Try to capture the error and recover?
  3. Kill the crashed process and launch another one in its place?
  For example, C uses approach 1. Most modern languages
  with Exceptions like Java and Python use 2.

No, in Java and C the "let it crash" (aka the "failstop principle") is also used.

View on HN

In C the OS process takes place of the Erlang "process". C is designed for multi-process systems (Unix), where a task is often implemented by a number of processes. Upon error the process is terminated, but the system (OS) is never "brought down" by a single failure, remains fully operational and you can retry your command again.

While similar, it's not the same. Erlang/OTP/Elixir provides baked in tools to watch, restart, etc. processes. In C you'd need to do this yourself.

The tools to watch, restart, etc are bash, ps, top, etc

Not really. Those are equivalent to monitors and links in Erlang/BEAM, if we go this line of thought. OTP is a collection of architectural patterns coded down in Erlang (what means OTP is a framework).

We are talking about error handling. My point: "let it crash" aka "failstop principle" is well known for decades, and practiced in all major languages.

> We are talking about error handling.

I thought we switched to a little more down-in-the-trenches discussion with you saying:

> The tools to watch, restart, etc are bash, ps, top, etc

Carried away a little. Not to say it's not interesting to compare Unix/C programs with OTP/Erlang, but it's a different subject.

Reply on news.ycombinator.com