Comment by whateveracct
1 year ago
I've come to hate "Postgres is all you need." Or at least, "a single Postgres database is all you need."
1 year ago
I've come to hate "Postgres is all you need." Or at least, "a single Postgres database is all you need."
I won´t call it "hate", but I've ran into quite some situations where the Postgres version caused a lot of pain.
- When it wasn't as easy as a dedicated solution: where installing and managing a focused service is overall easier than shoehorning it into PG.
- when it didn't perform anywhere close to a dedicated solution: overhead from the guarantees that PG makes (acid and all that) when you don't need them. Or where the relational architecture isn't suited for this type of data: e.g. hierarchical, time-series, etc.
- when it's not as feature complete as a dedicated service: for example I am quite sure one can build (parts of) an ActiveDirectory or Kafka Bus, entirely in PG. But it will lack features that in future you'll likely need - they are built into these dedicated solutions because they are often needed after all.
Why?
Just to share an anecdote, we've been able to get to market faster than ever before by just using Postgres (Supabase) for basically everything. We're leveraging RLS, and it's saved us from building an API (just using PostgREST). We've knocked months off our project timeline by doing this.