In-memory management and state sync #932
Replies: 1 comment
-
I've been thinking a lot about this flow lately. I think the most robust option (to avoid potential misuse) would be to remove this capability and require such mechanisms to be implemented over the reliable stream exported by the VM (#961).
Anything that is required for state transition computation must be stored in state. You should not be using anything updated in I've created an issue that should make this easier to reason about (will update |
Beta Was this translation helpful? Give feedback.
-
Any logic that sits inside the Controller can modify the memory of a node (tokenvm example)
As far as i understand the transaction lifecycle, the logic inside the Controller is executed only after a transaction is marked as
Accepted
:I think any in-memory changes should be persisted on the state, such that all the nodes on the network can sync their memory after a block has been produced.
Maybe the commands (inputs) and the events (outputs) logging could be managed by the sdk so that we have a standard on state synchronisation
Beta Was this translation helpful? Give feedback.
All reactions