Neynar

Celebrating the experimental launch of hubs.neynar.com

Any client of even basic complexity will need joins IMO

For example, the hot/new/top filters on Flink.

Neynar is using a replicator so I consider that to be the same thing as running your own.

In memory joins will work until you have a dataset larger than a few thousand records, which I think you’ll run up on for a feed algo depending on the parameters.

I’m excited about the potential of an in-hub DB from @haardikkk

haardik.dev | unreasonably optimistic, relentlessly curious

likes building things • https://alexgrover.me

In memory joins will work until you have a dataset larger than a few thousand records, which I think you’ll run up on for a feed algo depending on the parameters.

I’m excited about the potential of an in-hub DB from @haardikkk

https://warpcast.com/haardikkk/0x6cd113a2

Depends on the use case IMO! For example “top casts by day/month/year/all time” will require much more than 10k. I think there’s always going to be some edge case feature that’s not possible through the hub APIs

If you don’t need the extra infra by all means keep things simple, but it’s a case by case decision rather than a blanket prescription and I think as clients get more complex the DB will become necessary more of the time