MongoDB Is a Document Store With a Marketing Department

MongoDB stores BSON documents in collections. That is the database. The brochure sold “flexible schema” and a long stretch of slides about leaving rows behind. I have queried it as a user on other people’s stacks. I have never been a MongoDB operator for Dark Heart Labs. mystic-bytes is markdown in git. The Railway scheduler uses Postgres. There is no Atlas cluster on the studio invoice.

A document is a nested structure with a primary key, usually _id. Collections are bags of those structures. Indexes are still indexes. Uniqueness still has to be declared. The schemaless pitch was that the engine would not yell when you added a field.¹ The application still had a schema. It lived in the driver, in the implicit contract between writers, and in the ticket when two shapes of user coexisted because someone shipped a rename without a migration. Rows make that fight visible. Documents hide it until the aggregation pipeline gets ugly.

Atlas is how the company meters the store: backups, replica sets, a dashboard that looks like ops. The marketing did its job. The product is still a document store. Use it when the record is a blob you want by id and occasionally slice with an aggregation. Skip it because a tutorial said SQL was the past. I read the manual so I can argue with the pitch. I do not owe MongoDB Inc a cluster for a Jekyll site.

— JV · Dark Heart Labs.

References

MongoDB Manual, “Documents,” https://www.mongodb.com/docs/manual/core/document/. BSON documents, field-value pairs, and _id — the store under the schema story. ↩