Nick Molnar on Warpcast

Content pfp

0 reply

20 recasts

20 reactions

mike rainbow (rainbow mike) ↑ pfp

mike rainbow (rainbow mike) ↑

@mikedemarais.eth

if we started a petition to stop using (boycott) farcaster until they made direct casts an open platform (likely using XMTP) would you sign it? (love u dwr)

29 replies

10 recasts

97 reactions

boscolo.eth pfp

XMTP is not open either…(at least not yet) Also, maybe there already is a silent boycott of builders going on as we speak.

1 reply

0 recast

0 reaction

Chinmay 🕹️🍿 pfp

Chinmay 🕹️🍿

What do you mean by "XMTP is not open"?

1 reply

0 recast

0 reaction

Cassie Heart pfp

it's centralized.

1 reply

0 recast

0 reaction

Chinmay 🕹️🍿 pfp

Chinmay 🕹️🍿

Are we ditching "sufficiently decentralized" when it comes to other protocols?

1 reply

0 recast

0 reaction

Cassie Heart pfp

I assume you're trying to be tongue in cheek here, but hubs are permissionless and running one allows you to send/receive/store messages the same as any other hub. XMTP is not decentralized at all, and their FAQ even clarifies this

2 replies

0 recast

1 reaction

Shane Mac pfp

@chinmay.eth thx Cassie for all of your feedback. Testnet in Q4 with external nodes running. Been working hard on this and excited to work together on it.

2 replies

0 recast

4 reactions

Cassie Heart pfp

What variant of MLS is allowing multiple nodes?

1 reply

0 recast

0 reaction

Nick Molnar pfp

Great Q as always. Previously we talked about using a variant of MLS that didn’t require total ordering. Decided to change gears and just use a smart contract for sequencing commits, so we have total ordering even post-decentralization. Lets us keep vanilla MLS.

1 reply

0 recast

0 reaction

Cassie Heart pfp

interesting approach, brings many more questions to mind: 1. how are epochs assured when smart contracts cannot invoke themselves at a given point in time? is it a master controller for the contract? 2. how are reorgs handled? 3. how is key validity assured? does the contract validate them? what about invalidating/revoking them? I suppose this could be self-answered: is the contract already out or is that for Q4?

1 reply

0 recast

0 reaction

Nick Molnar pfp

We’ll have an XIP up soon that should answer all those questions in more depth. Sneak peek: 1. Causal ordering between on and offchain messages. 2. Hardest part of the thing. Good news is that you can know that a reorg happened and re-create the group in the worst case. 3. All on the client

2 replies

0 recast

2 reactions

Cassie Heart pfp

very cool, who invokes the contract to advance the epoch?

1 reply

0 recast

0 reaction

Nick Molnar pfp

The app typically will do it on behalf of the user, so no one has to pay gas fees themselves or get de-anonymized by the tx.

1 reply

0 recast

0 reaction

Cassie Heart pfp

I'm somewhat confused here – 1. i'm assuming the app has a central intermediary API that pays the gas fee? 2. with that indirection – reorgs have a strong probability of getting clients into a very disjointed state, but worse, will induce a sled of reinits. are the apps themselves supposed to be a buffered channel to the contract state? I can see that working in a single-app case with no issues for the user, but I'm not fully seeing the decentralized case. Even a federated scenario with two distinct apps can quickly cascade, unless message delivery is delayed to some number of block confirmations 3. taking a step back, epoch rotations occur periodically and when users are added/removed from a group, and special scenarios for reinits and new subgroups (if supported). At 1000 groups, the rate of these is probably fine on-chain, but at 1MM groups, this will start to get very costly and congested.

1 reply

0 recast

0 reaction