In cloud native arrangements, we are looking at the database as the ultimate integration tool. These schema are not owned by any specific service. There is a to

In cloud native arrangements, we are looking at the database as the ultimate integration tool. These schema are not owned by any specific service. There is a totally separate universe responsible for how all of that works outside code. Allowing code to drive schema (and permitting the implication that this is the best way) is the biggest frustration I have with ORMs & their higher-order consequences. In my view, if the schema must change so radically that traditional migrations and other 'grow-only' techniques fall apart, you are probably looking at a properly-dead canary and in need of evacuating the entire coal mine. The Quote regarding flowcharts and tables applies here - if you radically alter the foundation, everything built upon it absolutely must adapt. Every flowchart into the trashcan instantly. Don't even think about it. They're as good as a ball & chain now. Allowing parts of the structure to dictate parts of the foundation is where we find ourselves with circular firing squads. Take things to the extreme - There is a reason you will start to find roles like "Schema Owner" in large, legacy org charts. These people cannot see the code or they will become tainted. They only have one class of allegiance - LOB owners. These are who they engage to develop & refine schema over time. The schema owner themselves has a full time job that is entirely dedicated to minimizing the impact of change over time to the org. This person is ideally the most ancient wizard in the org chart and has the prior Fred Brooks quote framed on their wall. You can make schema change a top-down event that touches the entire organization. This happens quite often in banking when the central system is completely swapped for a different vendor & tech stack. Most of a bank is just a SQL database, but every vendor has a different schema that has to be adapted to. This is known as "core conversion" in the industry and is one of the more hellish experiences I have ever seen. If a bank with 4 decades of digital records can pull something like this off with regularity, there aren't many excuses that remain for a hole-in-the-wall SaaS app with 6 months of customer data.