Vlad Dumitrescu (EAW)
Mon Mar 3 08:59:29 CET 2003
> From: Vance Shipley [mailto:]
> On Fri, Feb 28, 2003 at 10:30:48AM +0100, Vlad Dumitrescu wrote:
> } ............... I even mentioned it here, and the answer
> was that this
> } would impede with the Erlang node's stability (just like a
> linked-in driver
> } does, but since the external application is more complex,
> the chances for
> } it to crash are bigger).
> I keep hearing this type of thing and always find it humorous. If I
> incorporate an Erlang node into a project I'm concerned about the
> stability of the resulting system. While Erlang could be a shared
> application environment like, a Unix system, it isn't used that way
> anywhere. In OTP you see this philosophy in the release handler.
> There is no way to add a new application on to a running system, you
> may only upgrade. So if we are always using the Erlang node to do
> just the one job, and that job needs a linked in driver, why would
> I be so concerned about the possibility of a node crashing because
> my code was bad? If my code is bad it doesn't work period.
I agree with you, I just forwarded the answer that I got. However there might be cases when this is not true: for example if the "one thing" the node does is a Web server - if one has an admin interface using a linked-in driver GUI that keeps crashing the server, then that's not good. If the GUI was separate, the rest of the world wouldn't have to know something is broken ;-)
More information about the erlang-questions