We have a policy where I work that all web services must log exceptions at the service boundary. If an exception occurs due to invalid inputs, should this be logged as an error?
I think that any time a service can't complete what it was asked to do, that is by default an error, no matter what the reason is. Others think that I should be considering the bigger picture of the system as a whole, and log those exceptions as warnings, because the consumer of the service could handle the exception and correct the inputs. Isn't the fact that the inputs might be corrected in a subsequent request irrelevant to whether or not an error has occurred?