[erlang-questions] Handling Crash Reports at scale

[Peer Stritzinger]

On 2013-06-04 14:27:25 +0000, Max Lapshin said:

> It is important to remember that remote user is usually waiting for 
> something like http 500 or 403, not just closing socket.
> 
> So, just "let-it-crash" is not always suitable way to write software.

Well "let it crash" is for unexpected errors.

In your eample HTTP 500 error seems to be the right thing that your 
webserver should handle right on the top level when any handler does 
crash.  HTTP 403 and other errors are a different story.

Besides that I agree with Jespers way of adding error handling sparsely 
while development.  The nice thing is if you have the crashes happening 
you no longer need to write error handling code that you can't trigger 
in a real situation.

-- Peer