| 7 | 7 | Overall I think the extra complication is not justified, and we should just be mindful of not breaking reasonable existing usage when deciding whether a new diagnostic is a warning or an error. In some cases that existing usage is broken and we were just lacking a check, so sometimes a new Survex version may give an error on existing data, but such cases should be very much the exception and when there's a reasonable fallback we should issue a warning and do that. Typically that will probably be doing what existing versions do, as in the case of `*date` where if the format is wrong we issue a warning and ignore the `*date` command). |
