> I really don't like this idea because it is so much of a kludge. I > suppose, though, thinking more about it, that since the effect will be > to change a 5xx error into a 4xx error, it might be a tolerable kludge. > A possible problem is that not all Unix-like OS have the same set of > error values, though that can be coped with by the use of #ifdefs, I > suppose.
All three points are entirely correct. The question is, can we handle a partial machine failure in a way that is on balance better than what happens at the moment, given that 99.99% of the time the machine isn't broken. Hence my suggestion of a conservative test and a cautious response. It's a case of working out the circumstances in which exim can return a hard error, asking "can we be pretty sure this is a machine problem?" and if so converting it to a soft error and logging the incident (to detect false positives). Crisis management is never going to be pretty, the only alternative is not to have any at all. John -- ## List details at http://www.exim.org/mailman/listinfo/exim-users ## Exim details at http://www.exim.org/ ## Please use the Wiki with this list - http://www.exim.org/eximwiki/
