DEV Community

Serguey Shinder
Serguey Shinder

Posted on

Keeping Every Row Forever Is a Decision, Not a Default

Nobody at that company had ever decided to keep every record forever. It simply happened, the way most data policies happen, which is to say by nobody making one. Storage was cheap, deletion felt risky, and no single person had the authority or the appetite to say that a nine-year-old event log could go. So it all stayed, and by the time I arrived we had a warehouse holding the complete history of every action any user had ever taken, most of it never queried, all of it faithfully backed up, replicated, and paid for every month.

The cost was not really the storage bill, though that was not trivial. It was everything that scaled with the volume. Backups took long enough that the restore window had quietly grown beyond what the business believed it was. Migrations became multi-weekend operations because moving that much history was the slow part. Queries that had been instant when the tables were young now needed careful indexing to survive, and every new engineer had to learn which parts of the history were meaningful and which were sediment.

Then there is the part that turns a storage question into a legal one. Data you hold is data you are responsible for, and a decade of personal records you never intended to keep is not a neutral asset. It is an obligation, and in the event of a breach it is the difference between an embarrassing incident and a reportable one, because the attacker gets everything you kept, including the years you had no business retaining.

What changed my thinking was realising that not deciding is itself a decision, just one made by inertia rather than judgement. Now I want a retention answer at the point a system is designed, when the question is cheap: how long is this genuinely useful, what has to be kept for compliance, and what happens to the rest? Even a rough answer beats the default, because the default is forever, and forever compounds quietly in every direction at once.

Data does not become harmless with age. It becomes heavier, and someone eventually has to carry it.

– Serguey Shinder

Top comments (0)