Seperate RC delegation from SP

[clayop]

After HF20, small users are experiencing restrictions on their activities. Basic users created by Steemit with 15 SP only post 4 comments per day on average. Regarding that newbies are usually very active and they need enough opportunities to be discovered, this limitation is a big obstacle to expand Steem ecosystem. Surely, reducing requirements of RC is not a good solution because it will more benefit spammers.

Fortunately, Steem has strong communities and third party service providers. They are wiiling to support small-stake users in order to attract them and to add values to their services. Currently, some graceful accounts are distributing some SP and RC (100 is standard in Korean community). But delegating SP often ends up with moral hazards with abuses of free voting power.

An alternative way is letting RC to be delegated separately from SP. Here's a proposal by @therealwolf. In this way, third parties can keep their SP to reward contributors at the same time can help small accounts play on Steem better.

[mvandeberg]

Delegation Pools (#2761) are our solution to this problem and helps solve two important challenges when it comes to on-boarding new users.

Delegation Pools will allow an account to delegate just RCs to a pool and then oversubscribe that pool to users. When not oversubscribed it will act a straight RC delegation. The oversubscription won't allow more RCs to be used than delegated to the pool, but will allow greater ceilings on users. For example, I can delegate 500 SP to the pool and then allows 10 users to use up to 100 SP each. The pool will not allow more than 500 SP of RCs to be consumed, but if the pool has RCs available, each user could spend like they had 100 SP. Currently, a lot of SP is wasted on delegations because the stake is not active. This will allow more effective allocations of resources.

Note: I say SP here as an understood unit. Delegation Pools only confer RC rights. Voting Power is unaffected.

[clayop]

I also thought a pool model but didn't suggest it because it seems difficult to implement. Your solution is the best answer of my suggestion.