T
Thomas Mercer-Hursh
Guest
A couple of comments. Ultimately, a "primary" index is really nothing more than the default index when no other index is specified. However, it is also often used to do things like enforcing uniqueness and such. Historically, many ABL applications used something like a customer number as the primary index. This works, if customer numbers are assigned serially and don't have any other meaning (like type of customer by range of number), but best practice indicates that the primary key for a record should be a meaningless field like a GUID or sequence. That way, if there is ever a need to change one of the meaningful fields in the record, any connections based on that GUID still hold without repair. None of the indices has anything to do with the sequence of physical storage. In a table with all additions and no deletions, the physical ordering *might* correspond to the order of creation, but you shouldn't count on it. A dump and load will physically reorder the records, but that shouldn't be part of your design considerations since dumping and loading should be very rare.
Continue reading...
Continue reading...