A place for developers to talk about building on Farcaster.

/dev

Ponder

After years of "web3" development, I still can't get over how arbitrarily stupid RPC nodes limitations are. Take eth_getLogs: Alchemy offers 100k, QuickNode — 10k, but these numbers can be different depending on how many results will be returned.

You either have to code against a single RPC provider (lock in) or run something like /ponder or graph (ops complexity++)

FWIW here's our solution from the `@ponder/utils` package (open-source but poorly documented rn). It handles ~20 different eth_getLogs implementations.

Your critique is spot-on, though - the standard RPC is underspecified!

Building ponder.sh. EVM, TypeScript, open-source enjoyer.

Oh main reading this file triggers PTSD from dealing with all these things ad-hoc through dozen of projects lol

FWIW here's our solution from the `@ponder/utils` package (open-source but poorly documented rn). It handles ~20 different eth_getLogs implementations.

Your critique is spot-on, though - the standard RPC is underspecified!

https://github.com/ponder-sh/ponder/blob/main/packages/utils/src/getLogsRetryHelper.ts#L36

software artist | building /outpaint | ex: @coinbase ex: bisontrails | seeker of the flow state

FWIW I love Ponder, well designed and simple. I just wish for simpler cases one didn't need to use it (or alternatives)

all infra leads to lock in. that’s how they make $$. railway.app is an exception. but ux as a moat is hard to maintain.

i think users will choose them for there growing library of templates.

Isn't that why we have cursor based iteration?

decentralist, turtle stacker, /livecoding cycle carpenter -- building /tofui

Well by arbitrary I mean more "non-standard". If Ethereum RPC spec had a clause saying you can only query X blocks (where X can be computed from more fundamental parameters like gas limit, block size, etc.), then I'd have no problem with it — at least all RPC providers would behave the same.

This wouldn't stop Alchemy from supporting, say, 100*X block limit. Great for them! But if I wanted to code a universal client, I'd just stick with X.

Today, however, there's no X and I have to empirically figure out X based on whatever popular RPC providers do. This adds complexity

These aren't really "arbitrary" limits. and Alchemy's limit is 10k as well btw. Logs fundamentally put a heavy load on nodes, and if you go further than that (especially depending on the node client) nodes will start to struggle and fall behind, drop your request, return a timeout/rpc error, and even just never provide a response.
There may actually be limits in certain node clients themselves too, not sure about that though

I want to provide a good UX without too much complexity or depending on third-party services

Do you actually want RPC calls or do you want a lot of data in a certain schema?