Re: exception handling

On Wed, 6 Mar 2002, Cezary Zielinski wrote:

> The classification of errors is a good idea. In MRROC++ we distinguish 3
> types of errors:
> - non-fatal errors (computational errors, e.g. associated with kinematic
> computations - violation of joint limits)
> - fatal errors (malfunction of a robot, e.g. someone cuts off power)
> - system errors (degradation of distributed control system capabilities,
> e.g. someone turns off one of the nodes of the computer network)
> Perhaps a finer granularity is needed.

I am wondering whether one should try to catalogue errors in a
classification of ``importance''... In my view, each component in the
system should have a means (i) to detect whatever errors it can
detect, (ii) to react to them as far as it can react to them, and
(iii) report about them to the ``outside'' world. The architecture of
how one constructs a system out of components then defines implicitly
what is the importance of an error.

This view of mine has not matured very much, so don't take it without
a big grain of salt :-)

Herman