Once we have declarative deployments available for lattice (#67), we should look into the possibility of using Open Policy Agent as a way of deferring the scheduling logic. Instead of each host having hard-coded logic that decides if it can accept an incoming request for work (deploy actor, deploy capability, etc), it will ask an OPA policy if it can accept that work. This would give us the flexibility of growing and improving our scheduling/deployment algorithms without having to break code interfaces.
Once we have declarative deployments available for lattice (#67), we should look into the possibility of using Open Policy Agent as a way of deferring the scheduling logic. Instead of each host having hard-coded logic that decides if it can accept an incoming request for work (deploy actor, deploy capability, etc), it will ask an OPA policy if it can accept that work. This would give us the flexibility of growing and improving our scheduling/deployment algorithms without having to break code interfaces.