Farcaster Devs

it's not going to be viable for most farcaster clients to index the entire network up front in the future - quite soon probably

we're going to see a handful of key infrastructure providers like neynar emerge that provide data for the vast majority of clients

the alternative is that clients will have to index "missing parts" of the network on-demand when new users sign up for their service. this is what opencast does to keep costs reasonable

this begs the question: if everything is going through a handful of infra providers anyway, does the level of data redundancy that farcaster hubs provide *and everyone is paying for* make sense?

tbh i think we'd be lucky if we avoid total consolidation and end up with even a handful of popular infra providers

if you are trying to build a fully featured Farcaster client and run the data infrastructure yourself: good luck! it’s a slog.

after indexing, you will have to build recommendations to beat churn. algofeed. then clustering. and AI. then hourly incremental DAG tasks and data pipelines. 

all under 200ms latency 👍

why would someone need (or want) to run their own data infra?

You do whatever you want with your own infra

easiest way to build on farcaster @ neynar.com | /neynar

Building @recaster-fc and other farcaster tools

@neynar doesn't get you there? 

i'm curious to know what the gaps are

Oh neynar indexer as a service may help with that, but i think you still need maintain the database

A usecase someone ask for: get notification when someone unfollowed me.

1. a lot of folks prefer to be as decentralized as possible, so this is following those efforts

2. you run your own infrastructure to keep data as close to your application as possible (you can collocate a hub and indexer within the same VM)

3. you want to own your ecosystem if Warpcast/Neynar is down

gotcha, right, maybe i'm confused about the initial cast