aws/aws.go (21 lines):
	- line 313: // TODO: Remove when user/admin can configure volume types and thus we don't
	- line 329: // TODO: Should we rename this to AWS (EBS & ELB are not technically part of EC2)
	- line 483: // TODO: Allow other clouds to implement this
	- line 519: // TODO: Allow other clouds to implement this
	- line 590: // TODO: Is there any use for this?  We can get it from the instance metadata service
	- line 1371: // TODO: Clean up double-API query
	- line 1448: // TODO: implement ONLY for external cloud provider
	- line 1524: //TODO: It would be nice to be able to determine the reason for the failure,
	- line 1533: //TODO: It would be nice to be able to determine the reason for the failure,
	- line 1546: //TODO: It would be nice to be able to determine the reason for the failure,
	- line 1611: // TODO: Other IP addresses (multiple ips)?
	- line 1777: // TODO: Caching / expose v1.Nodes to the cloud provider?
	- line 1778: // TODO: We could also query for subnets, I think
	- line 2437: // TODO: Check if the volume was concurrently attached?
	- line 2507: // TODO: Continue?  Tolerate non-attached error from the AWS DetachVolume call?
	- line 3167: // TODO: There is a limit in VPC of 100 rules per security group, so we
	- line 4064: // TODO: Wait for creation?
	- line 4313: // TODO: Wait for creation?
	- line 4352: // TODO: replace DefaultLoadBalancerName to generate more meaningful loadbalancer names.
	- line 4664: // TODO: Check if error was because load balancer was concurrently deleted
	- line 4906: // TODO: Move to instanceCache


gce/gce_loadbalancer_external.go (4 lines):
	- line 101: // TODO: distinguish between unspecified and specified network tiers annotation properly in forwardingrule creation
	- line 762: // TODO: we report loadbalancer IP through status, so we want to verify if
	- line 799: // TODO: If the user modifies their Service's session affinity, it *should*
	- line 1033: // TODO: Make this more generic. Currently this method is only


aws/aws_loadbalancer.go (4 lines):
	- line 177: // TODO: What happens if we have more than one subnet per AZ?
	- line 205: // TODO: Sync internal vs non-internal
	- line 988: // TODO: What happens if we have more than one subnet per AZ?
	- line 1037: // TODO: Sync internal vs non-internal


gce/gce_clusters.go (3 lines):
	- line 43: // TODO: Scoping?  Do we need to qualify the cluster name?
	- line 94: // TODO: Issue/68913 migrate metric to list_location instead of list_zone.
	- line 96: // TODO: use PageToken to list all not just the first 500


gce/gce_instances.go (3 lines):
	- line 434: // TODO: this should be removed from the cloud provider.
	- line 463: // TODO: remove this method.
	- line 725: // TODO: We could store the tags in gceInstance, so we could have already fetched it


openstack/openstack_loadbalancer.go (3 lines):
	- line 496: // TODO: replace DefaultLoadBalancerName to generate more meaningful loadbalancer names.
	- line 660: // TODO: This code currently ignores 'region' and always creates a
	- line 719: // TODO: Convert all error messages to use an event recorder


aws/aws_routes.go (2 lines):
	- line 36: // TODO: All clouds should be cluster-aware by default
	- line 191: // TODO: use ClientToken for idempotency?


gce/gce_interfaces.go (2 lines):
	- line 35: // TODO: Mock Global endpoints
	- line 46: // TODO: Expand the interface to include more methods.


gce/gce.go (2 lines):
	- line 176: // TODO: replace gcfg with json
	- line 888: // TODO: use PageToken to list all not just the first 500


gce/gce_instancegroup.go (2 lines):
	- line 83: // TODO: should cull operation above this layer.
	- line 100: // TODO: should cull operation above this layer.


azure/azure_vmss.go (2 lines):
	- line 542: // TODO (khenidak): This should read all nics, not just the primary
	- line 891: // TODO: is this a bug of azure SDK?


vsphere/nodemanager.go (2 lines):
	- line 53: // TODO: replace map with concurrent map when k8s supports go v1.9
	- line 77: // TODO: Make it configurable in vsphere.conf


vsphere/vsphere.go (2 lines):
	- line 573: // TODO: implement ONLY for external cloud provider
	- line 1777: // FIXME: For now, pick the first zone of datastore as the zone of volume


gce/gce_healthchecks.go (1 line):
	- line 41: // TODO: use a shared constant once ports in pkg/cluster/ports are in a common external repo.


aws/retry_handler.go (1 line):
	- line 166: // TODO: This is pretty arbitrary


aws/instances.go (1 line):
	- line 151: // TODO: Get rid of this field, send all calls through the instanceCache


gce/support.go (1 line):
	- line 51: // TODO: the current cloud provider policy doesn't seem to be correct as it


vsphere/vclib/connection.go (1 line):
	- line 94: // TODO: Add separate fields for certificate and private-key.


gce/gce_routes.go (1 line):
	- line 57: // TODO: Should we lastComponent(target) this?


gce/gce_networkendpointgroup.go (1 line):
	- line 63: // TODO: filter for the region the cluster is in.


openstack/openstack_instances.go (1 line):
	- line 67: // TODO: implement ONLY for external cloud provider


azure/azure.go (1 line):
	- line 690: // TODO: implement ONLY for external cloud provider


azure/azure_standard.go (1 line):
	- line 615: // TODO (khenidak): This should read all nics, not just the primary


gce/gce_loadbalancer.go (1 line):
	- line 124: // TODO: replace DefaultLoadBalancerName to generate more meaningful loadbalancer names.


azure/azure_routes.go (1 line):
	- line 385: //TODO (khenidak) add support for unmanaged nodes when the feature reaches beta