It’s a weird article. It claims that one only needs Postgres but then presents a half-baked solution: the takeaway I got is “it’s not HA or horizontally scalabl

It’s a weird article. It claims that one only needs Postgres but then presents a half-baked solution: the takeaway I got is “it’s not HA or horizontally scalable, and it’s probably slow” so it sounds like I _do_ need a dedicated cache service? They just handwave benchmarking against cache databases as “out of scope”, that is very much in scope to claim I can use PG rather than redis. Also a cache would impact the rest of the application (reading/writing to this cache consumes cpu and IO). Also, missing an index on the inserted_at and there’s no uniqueness constraint on the key. Overall, it reads like someone read about UNLOGGED and thought it could be useful for caching use cases. I don’t think anyone doubted that PG can be used a cache really, the trade offs would be a more interesting read than the POC