[erlang-questions] Dependencies, included applications and supervision

[Felix Gallo]

On Wed, Jun 25, 2014 at 2:28 PM, Loïc Hoguin <essen@REDACTED> wrote:
>
>
> But why would you make the application crash if you need it to stay alive?
> If your DB driver application crashes because the DB is gone, something is
> *very* wrong with your DB driver application. I personally would stay the
> hell away from it.
>

This is one of those things that as an erlang newbie I don't have a great
intuition for yet.

On one hand I can imagine a properly implemented database application that,
when the database goes away, crashes, is restarted appropriately, restores
the connection or picks a new slave or whatever, etc., etc. -- and that
design could be quite clean, free of ugly error handling code, and so on.

On the other hand, I can also envision a scenario under which that is a
giant pain in the ass for clients of the database application, especially
if there's any asynchronous messaging involved.

I know there's one subschool of the 'let it crash' mentality that appends
'...until the error shows up in the log, and then update your code to
handle it'.  And there's a subschool also which seems to draw a distinction
between infrastructure applications (e.g., db drivers) and
application-applications and applies 'let it crash' to each slightly
differently.

In the end I know it's a pragmatic architectural decision which is heavily
dependent on circumstances, but it would be interesting to hear from the
old erlang dogs about, e.g., db drivers and other core underpinnings and
what the battle-tested best practice is.

F.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20140626/a491b048/attachment.htm>