James Carlson wrote: >> In either event, I don't think the goal is to assign blame for the failure. > > The goal of sending a termination signal to some subset of all of the > processes on the signal is to impute "blame." Those processes are the > ones "damaged" by the failure and that are sharing fate.
Er, I'll just have to disagree there. I certainly don't infer any assignment of blame. The rest of the service may (or may not) be a completely innocent victim. I think the goal is to try to get the service healthy again, and to call attention to the failure. > It's the fact that this is a _new_ behavior and that it's really not > well-understood that's at issue. Absolutely. It may also be overkill. > However, it's not the _fault_ of ifconfig or its invoker if anything > goes wrong with these other processes. Probably not. However, something going wrong with those other processes probably _does_ indicate a problem with the service _as a whole_. Perhaps it's an unimportant problem that does not affect anything the user can see, or perhaps it's absolutely critical to the usefulness of the service.
