Dynamic self-stake means, rather than a fixed amount of stakes you vote for yourself such as 1.2M IOTX we currently use, you change this amount of stake dynamically to adjust your commitment to the protocol, which makes sense in two ways at least:
The more commitment you have to the protocol, the more support you could be getting from others such as voters. Tezos realizes this concept by instantiating the “support” using something called “capacity” - The votes one can receive is a function of his self-stakes.
The slash of self-stakes becomes more reasonable if such a delegate decides to attack the network, e.g., double sign blocks at the same height. For now, the protocol could only slash up to 1.2M IOTX (for many delegates) in such a circumstance. Ideally, the protocol should bias (in terms of responsibility and reward) to the delegates who have more commitment (and who should have more votes as well), to squeeze out less-committed delegates who could turn to an attacker without too much cost.
I am almost going to propose such a change to IoTeX network. Before formalizing this scheme, what early feedback can I get from the community?