Kimball usually leaves implementation details up to the implementer. I would simply say you've demonstrated one implementation approach to accomplish the same end goal. I've never implemented a DW without encountering issues with source data. I don't think Kimball meant to imply that you should blindly import dates from the source system without cleaning them. In fact you often don't have explicit from/to dates in the source system at all, and instead must use a change tracking mechanism or temporal tables in order to derive the start/end dates, in which case you can usually be guaranteed continuity if setup properly. Even if you take your approach, if your legacy source database has inconsistencies and gaps, you need to address that and the answer may not be to default to trusting the start dates. Not at all a criticism of your article, I just wouldn't categorize it as "debunking".
↧
