What's the best relational database model for Nostr? Just event and tag tables or tags as JSON objects inside the events table? Or maybe best to use a native array column for tags?
Does the randomness of the hex keys (event id, pubkey) affect the indexing performance?
One table per kind and denormalized to the bones.
Drawback is that this can't store out of spec events.
Please Login to reply.
It also requires searching all tables for the ID you have.
Bloom filter per kind?
Don't know, havent build a relay. So don't know the specific queries that are needed.
Is it possible to trace the requests on a relay? That would be really cool for a benchmark set.
