It's a matter of use case - for simple web page or even a lightweight client-side app, maybe too big. But for tool with heavy data processing, I could see it being quite useful.
Youre right, real-time is generally achieved by streaming CDC data with logical replication via a connector server to Kafka (Debezium does this) or a different queue to be picked up by consumers downstream.
Supabase does a great job of abstracting that part away with their real-time feature set
That's an excellent point. My statement was unnecessarily confusing. I've changed it to simply reference another benefit of the WAL - optimizing I/O operations. That would make a good blog post in and of itself :D
Supabase realtime (especially if you want a managed backend) or other streaming CDC setups (like Decodable, which is Flink/Debezium under the hood) are also great choices for logical replication. Streaming tech will continue to get more cost-effective and simpler to implement in the coming year(s).
I should note: I haven't used Decodable in production yet, I'm just a fan of Flink :)
Should be relatively affordable for modest deployments - I would categorize Neon and Airbyte as "bang-for-your-buck" products (Airbyte vs Fivetran, + Neon's solid free tier).
This is a guide to logical replication in Postgres where I break down some of the internal components of the database to explain CDC. If you've ever wondered how WAL buffers work, or what happens when a transaction is executed, check it out!