Farcaster

To be clear, I'm asking the same question, not stating that it is the case.

If I were to digest what I'm seeing:
1. Paying for storage is too much friction to onboard new users but not enough to slow down spam
2. Hubs aren't going to scale 2 - 3 orders of magnitude without a significant change in architecture.
3. The incentives to run a HUB are misaligned and causing congestion instead of capacity

On Hubs

1. We've said from the beginning, people will run Hubs without direct economic incentives. This has proven correct. 10K+ Hubs on the network.

2. This number is artificially high as there are likely many Hub operators hoping for a future economic reward. There will not be a reward.

3. Rewards don't make sense as there's no good way to prove that a Hub was honest and performant for a long period of time. Any deterministic system tied to an economic reward would create an incentive for fraud to fake running a Hub to get paid.

3. This is why Bitcoin and Ethereum do not pay node operators. Economic rewards are independent from running a node. Coinbase ran BTC / ETH nodes for years (pre-staking).

4. Our original goal was get to ~100 independent, good Hubs. Given network size, activity, number of companies building on Farcaster, seems like that number is achievable between 1) companies with an economic interest in access to the network and 2) altruistic individuals for hobbyist or ideological reasons.

1. Paying for storage is too much friction to onboard new users but not enough to slow down spam

> We experimented with free sign ups in a number of countries and rentention was significantly lower than users who paid the $5 to sign up. 

2. Hubs aren't going to scale 2 - 3 orders of magnitude without a significant change in architecture.

> We've said from the beginning we think Hubs scale without sharding to ~10M daily active users, give or take. Scaling is hard, but the existential problem is getting the actual retained users. :)

3. The incentives to run a HUB are misaligned and causing congestion instead of capacity

> There are no incentives? Adding incentives would increase the number of Hubs. The team is exploring how to reduce congestion. Possibly increasing economic cost of running a Hub (since access to the network is valuable). Nothing definitive yet.

See also my cast below On Hubs:

1. Paying for storage is too much friction to onboard new users but not enough to slow down spam

> We experimented with free sign ups in a number of countries and rentention was significantly lower than users who paid the $5 to sign up. 

2. Hubs aren't going to scale 2 - 3 orders of magnitude without a significant change in architecture.

> We've said from the beginning we think Hubs scale without sharding to ~10M daily active users, give or take. Scaling is hard, but the existential problem is getting the actual retained users. :)

3. The incentives to run a HUB are misaligned and causing congestion instead of capacity

> There are no incentives? Adding incentives would increase the number of Hubs. The team is exploring how to reduce congestion. Possibly increasing economic cost of running a Hub (since access to the network is valuable). Nothing definitive yet.

See also my cast below On Hubs:
https://warpcast.com/dwr.eth/0x2d7823b4

Working on Farcaster and Warpcast. Longer thoughts at https://dwr.email

paying for storage isn't just too much friction, it's a non-starter for most mainstream users and also for apps building on farcaster who want to onboard users to the network

i don't think the retention experiment proves much about the success of the storage model without more detail. when was this, how big was the network, what is the sample size, what were they interested in, why were they joining, were they in crypto or were they outsiders, the list goes on.

i also don't think farcaster will get more than a few millions of users paying annually for storage (optimistically), which i would hope is much less than what farcaster is aiming for

lately i have developed a strong opinion, which i hold loosely, that the cost derives from unnecessary data redundancy. if hubs didn't need to store all messages then storage costs would be a non-issue and we wouldn't even have to speculate on whether it's good or bad for the network

and i say all of this with great respect for you and the merkle team's pragmatism approaching this problem. you guys are a big reason why i believe in the potential of farcaster

1. surely the cozy corners includes mainstream users?

2. surely that is the case for any paid product? it drastically limits the size of your market

3. let's try!

4. imo the developer experience hit is marginal. serious apps run their own indexers already and i suspect they will also start selectively indexing soon because there is too much spam even with storage fees in place. services with lots of resources and paying customers like neynar can index everything and amortize the cost for smaller projects. 

also on centralization of federated model: i'm finding it hard to see how the hub model necessarily leads to less centralization. iirc varun's blogpost mentions email as an example, but imo email is sufficiently decentralized! when hubs become hard to run then there i predict that there will only be a handful of organizations running hubs

5. a possible scenario here is that a read-only no protocol sign up client gains traction, adds closed writes for free and just becomes a centralized social network

Thoughtful critique!

1. We don't have many mainstream users signing up right now (no reason for them to), so doesn't seem like a problem to solve?

2. Experiment was earlier this year. It indicates that lower intentionality users are less likely to retain?

3. Let's get to a few million users first?

4. If you shift to a federated model, then developer experience takes a hit. How do you know what servers to index vs. not? Also increases centralization. This was the original v1 architecture.

5. I suspect when the network is larger, a read only no sign up required lurker client experience will do well. Even Elon is moving to charge for write access. AI bots are only going to get more sophisticated.