Quote from: Evil-Knievel on February 04, 2017, 10:30:25 PM

1. Since earlier everyone had to verify work, we were very limited: we had very strict constraints regarding the complexity and memory requirement of work packages, simply due to the reason that the weakest node still has to be able to verify everything in a reasonable amount of time. Otherwise, netsplits or node shootdowns would be possible with complex long running tasks. With super nodes we are more flexible: we can set strict requirements that a super node's hardware has to meet and so can allow programs that run longer and use more memory.
Supernodes are basically nodes that do the verification and have no incentive to cheat due to a huge deposit. Contrary to other "well hyped solutions" we do not rely on trust at all ... all calculations and decisions that the super nodes perform can be publicly verified by anyone. For example by the so called guard nodes. These are nodes that randomly cross-check a small fraction of jobs. If they detect that a super node lied, its deposit gets forfeited.

3. Also, and this is a PR benefit, the earlier design effectively executed foreign code on the own hardware. While it's super safe due to Elastic PL's nature (our own programming language that does not allow to program anything dangerous), managers who want to use Elastic in their commercial setting might get scared. Now, nobody but the supernodes has to execute anything.

Three short questions on above:
1. Who and how will control the guard nodes then, so they don't cheat?
2. How the supernode deposit will be forfeited, and where the money will go to?
3. If nobody but the supernodes has to execute anything, what the other miners will do? Aren't they supposed to execute job assignments?