I have read some documents about datawarehouse time dimension implementation. Some documents about temporal implementation speak about a "Date Begin" and "Date End" attributes, other documents speak about a unique date timestamp and a "master" attribute that link a record with previous (time list of records)...
What is the best method? What method is commonly used in datawarehouse implementations?
In my opinion:
I think that the first method (2 timestamps) is more difficult to implement because I have to create triggers in order to maintain the "temporal referential integrity" but, once implemented, the table population phase is easier. The second one is more simple to implement but in the population phase I have to seach the previous record ID in order to insert it in the new record.
I found the first method implementation in:
developing time-oriented database applications in SQL (Richard T. Snodgrass)
What is the best implementation in order to use kettle to populate the tables?
Thanks in advance.