CEP 68: Delayed refunds #68
Conversation
AlexanderLimonov
requested review from
EdHastingsCasperAssociation,
MParlikar,
afck,
darthsiroftardis,
goral09,
mpapierski and
ziqi-qiao
|
|
||
| #### Elements | ||
|
|
||
| *Thresholds* in {p<sub>1</sub>, ..., p<sub>N</sub>} are percentages (fractions ranging from 0 to 1) that define ranges of observed past refunds, which determine a *penalty* in {1, ..., M}. |
There was a problem hiding this comment.
ranges of observed past refunds
I'm not sure I understand correctly; maybe this could be clarified in the text:
The pi are fixed parameters, are they? They are not the actual observed refunds?
E.g. we could configure (in the chainspec) {p1, ..., pN} to be {10%, 20%, 50%}, so that penalty 1 applies if a user reserved 10% too much, 2 if it's 20% too much and 3 if it's 50% too much?
There was a problem hiding this comment.
Yes, it's just a step function laid out in the chainspec, exactly as you describe, will clarify.
|
|
||
| *Thresholds* in {p<sub>1</sub>, ..., p<sub>N</sub>} are percentages (fractions ranging from 0 to 1) that define ranges of observed past refunds, which determine a *penalty* in {1, ..., M}. | ||
|
|
||
| Fix N to be #{1, ..., M}. The thresholds define the penalty in the following way: |
There was a problem hiding this comment.
Awkwardness due to M not necessarily being an integer,
There was a problem hiding this comment.
But then why does it start with 1? Does that mean "a penalty of 1 era", i.e. delaying the refund by one era?
There was a problem hiding this comment.
The penalty is multiplicative, so 1 means no delay at all (everything is refunded at the end of the era). I do see where the problem is now, though, since non-integer multipliers on number of eras do not make sense.
| Fix N to be #{1, ..., M}. The thresholds define the penalty in the following way: | ||
|
|
||
| - Take p<sub>i</sub> and p<sub>i+1</sub> | ||
| - If one minus *aggregate utilization* **P** (1 - **P**) is between p<sub>i</sub> and p<sub>i+1</sub> |
There was a problem hiding this comment.
I read this as "P · (1 - P)", i.e. "P times one minus P" at first. Maybe less ambiguous:
| - If one minus *aggregate utilization* **P** (1 - **P**) is between p<sub>i</sub> and p<sub>i+1</sub> | |
| - If 1 - **P**, i.e. one minus *aggregate utilization* **P**, is between p<sub>i</sub> and p<sub>i+1</sub> |
|
@AlexanderLimonov can this be closed? Merged? |
|
There's no reason to touch it either way right now. This work will resume after finality signature rewards are done. |
There was a problem hiding this comment.
@AlexanderLimonov this needs to be updated to reflect updated thinking. Please also include the updates to RPC and the precise refund mechanism.
|
Closing due to a major rework of refund-related logic proposed in CEP 89 |
4 participants
A new method for issuing refunds, introducing dynamic penalties based on observed past refund percentages
Rendered
NOTE: Revision pending