The downvoting is because you can have tables with single jsonb columns without schemas - call them "collections" and pretty much end up with a faster mongo up to having to fiddle a bit more with replication/sharding.

No one is making you enforce schemas with PostgreSQL if you work this way.

As for enforcements - there are things that are _impossible_ to enforce at an application level for atomicity reasons - for example - if you have a `users` table and you can't have duplicate email addresses - you'd need the operation "check if a user with the said email exists, if it does - return it, if it doesn't create it and return it" - PostgreSQL can do this easily, which prevents duplicate entries - Mongo - not so easily (see the "two phase commit" docs for how to simulate transactions).