Today I am closing out the week with a comment from Tim:
Just to add an alternative spin on this argument…
Firstly, we work quite often in an agile development methodology, which practically means that the customer’s requirements are "a bit fluid". That often means that we cannot establish a definitive schema or data model in either the code or the database early in the project. In such cases, we find that using a "vanilla" ID column is very useful as it can act as a proxy for a better primary key. It does mean that we sometimes need to add extra indexes in the database later to ensure consistency etc…
Secondly, we often work with people whose data is sensitive. But very often we are able to take away copies of subsets of the data for off-site analysis as long as the data has been suitably obfuscated (typically names etc need to be removed). Having bland vanilla ID columns gives us a very easy way to get a consistent obfuscated view of the data, and does allow traceability back to the original data if necessary. That is a surprisingly useful aspect in practice.
I really enjoy getting feedback from folks. It’s very nice to get an opinion other than my own, and I’m sure it provides a lot more value to you as well. So, please keep the conversation going with questions about software development processes, tools, techniques, etc. You can be sure that at least one of our readers will have experienced the same issue…and maybe even have a solution. Send your comments or questions to btaylor@sswug.org.
Cheers
Ben