factory_client_access.go

/*
Copyright 2016 The Kubernetes Authors.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/

// this file contains factories with no other dependencies

package util

import (
	"errors"
	"flag"
	"fmt"
	"io"
	"os"
	"path/filepath"
	"regexp"
	"strings"
	"time"

	"github.com/spf13/cobra"
	"github.com/spf13/pflag"

	"k8s.io/apimachinery/pkg/api/meta"
	"k8s.io/apimachinery/pkg/runtime"
	"k8s.io/apimachinery/pkg/runtime/schema"
	utilflag "k8s.io/apiserver/pkg/util/flag"
	"k8s.io/client-go/discovery"
	restclient "k8s.io/client-go/rest"
	"k8s.io/client-go/tools/clientcmd"
	"k8s.io/client-go/util/homedir"
	fedclientset "k8s.io/kubernetes/federation/client/clientset_generated/federation_internalclientset"
	"k8s.io/kubernetes/pkg/api"
	"k8s.io/kubernetes/pkg/api/service"
	"k8s.io/kubernetes/pkg/apis/apps"
	"k8s.io/kubernetes/pkg/apis/batch"
	"k8s.io/kubernetes/pkg/apis/extensions"
	"k8s.io/kubernetes/pkg/client/clientset_generated/internalclientset"
	"k8s.io/kubernetes/pkg/kubectl"
	"k8s.io/kubernetes/pkg/kubectl/resource"
)

type ring0Factory struct {
	flags            *pflag.FlagSet
	clientConfig     clientcmd.ClientConfig
	discoveryFactory DiscoveryClientFactory
	clientCache      *ClientCache
}

func NewClientAccessFactory(optionalClientConfig clientcmd.ClientConfig) ClientAccessFactory {
	flags := pflag.NewFlagSet("", pflag.ContinueOnError)

	clientConfig := optionalClientConfig
	if optionalClientConfig == nil {
		clientConfig = DefaultClientConfig(flags)
	}

	return NewClientAccessFactoryFromDiscovery(flags, clientConfig, &discoveryFactory{clientConfig: clientConfig})
}

// NewClientAccessFactoryFromDiscovery allows an external caller to substitute a different discoveryFactory
// Which allows for the client cache to be built in ring0, but still rely on a custom discovery client
func NewClientAccessFactoryFromDiscovery(flags *pflag.FlagSet, clientConfig clientcmd.ClientConfig, discoveryFactory DiscoveryClientFactory) ClientAccessFactory {
	flags.SetNormalizeFunc(utilflag.WarnWordSepNormalizeFunc) // Warn for "_" flags

	clientCache := NewClientCache(clientConfig, discoveryFactory)

	f := &ring0Factory{
		flags:            flags,
		clientConfig:     clientConfig,
		discoveryFactory: discoveryFactory,
		clientCache:      clientCache,
	}

	return f
}

type discoveryFactory struct {
	clientConfig clientcmd.ClientConfig
}

func (f *discoveryFactory) DiscoveryClient() (discovery.CachedDiscoveryInterface, error) {
	cfg, err := f.clientConfig.ClientConfig()
	if err != nil {
		return nil, err
	}
	discoveryClient, err := discovery.NewDiscoveryClientForConfig(cfg)
	if err != nil {
		return nil, err
	}
	cacheDir := computeDiscoverCacheDir(filepath.Join(homedir.HomeDir(), ".kube", "cache", "discovery"), cfg.Host)
	return NewCachedDiscoveryClient(discoveryClient, cacheDir, time.Duration(10*time.Minute)), nil
}

// DefaultClientConfig creates a clientcmd.ClientConfig with the following hierarchy:
//   1.  Use the kubeconfig builder.  The number of merges and overrides here gets a little crazy.  Stay with me.
//       1.  Merge the kubeconfig itself.  This is done with the following hierarchy rules:
//           1.  CommandLineLocation - this parsed from the command line, so it must be late bound.  If you specify this,
//               then no other kubeconfig files are merged.  This file must exist.
//           2.  If $KUBECONFIG is set, then it is treated as a list of files that should be merged.
//	     3.  HomeDirectoryLocation
//           Empty filenames are ignored.  Files with non-deserializable content produced errors.
//           The first file to set a particular value or map key wins and the value or map key is never changed.
//           This means that the first file to set CurrentContext will have its context preserved.  It also means
//           that if two files specify a "red-user", only values from the first file's red-user are used.  Even
//           non-conflicting entries from the second file's "red-user" are discarded.
//       2.  Determine the context to use based on the first hit in this chain
//           1.  command line argument - again, parsed from the command line, so it must be late bound
//           2.  CurrentContext from the merged kubeconfig file
//           3.  Empty is allowed at this stage
//       3.  Determine the cluster info and auth info to use.  At this point, we may or may not have a context.  They
//           are built based on the first hit in this chain.  (run it twice, once for auth, once for cluster)
//           1.  command line argument
//           2.  If context is present, then use the context value
//           3.  Empty is allowed
//       4.  Determine the actual cluster info to use.  At this point, we may or may not have a cluster info.  Build
//           each piece of the cluster info based on the chain:
//           1.  command line argument
//           2.  If cluster info is present and a value for the attribute is present, use it.
//           3.  If you don't have a server location, bail.
//       5.  Auth info is build using the same rules as cluster info, EXCEPT that you can only have one authentication
//           technique per auth info.  The following conditions result in an error:
//           1.  If there are two conflicting techniques specified from the command line, fail.
//           2.  If the command line does not specify one, and the auth info has conflicting techniques, fail.
//           3.  If the command line specifies one and the auth info specifies another, honor the command line technique.
//   2.  Use default values and potentially prompt for auth information
//
//   However, if it appears that we're running in a kubernetes cluster
//   container environment, then run with the auth info kubernetes mounted for
//   us. Specifically:
//     The env vars KUBERNETES_SERVICE_HOST and KUBERNETES_SERVICE_PORT are
//     set, and the file /var/run/secrets/kubernetes.io/serviceaccount/token
//     exists and is not a directory.
func DefaultClientConfig(flags *pflag.FlagSet) clientcmd.ClientConfig {
	loadingRules := clientcmd.NewDefaultClientConfigLoadingRules()
	// use the standard defaults for this client command
	// DEPRECATED: remove and replace with something more accurate
	loadingRules.DefaultClientConfig = &clientcmd.DefaultClientConfig

	flags.StringVar(&loadingRules.ExplicitPath, "kubeconfig", "", "Path to the kubeconfig file to use for CLI requests.")

	overrides := &clientcmd.ConfigOverrides{ClusterDefaults: clientcmd.ClusterDefaults}

	flagNames := clientcmd.RecommendedConfigOverrideFlags("")
	// short flagnames are disabled by default.  These are here for compatibility with existing scripts
	flagNames.ClusterOverrideFlags.APIServer.ShortName = "s"

	clientcmd.BindOverrideFlags(overrides, flags, flagNames)
	clientConfig := clientcmd.NewInteractiveDeferredLoadingClientConfig(loadingRules, overrides, os.Stdin)

	return clientConfig
}

func (f *ring0Factory) DiscoveryClient() (discovery.CachedDiscoveryInterface, error) {
	return f.discoveryFactory.DiscoveryClient()
}

func (f *ring0Factory) ClientSet() (*internalclientset.Clientset, error) {
	return f.clientCache.ClientSetForVersion(nil)
}

func (f *ring0Factory) ClientSetForVersion(requiredVersion *schema.GroupVersion) (*internalclientset.Clientset, error) {
	return f.clientCache.ClientSetForVersion(requiredVersion)
}

func (f *ring0Factory) ClientConfig() (*restclient.Config, error) {
	return f.clientCache.ClientConfigForVersion(nil)
}
func (f *ring0Factory) BareClientConfig() (*restclient.Config, error) {
	return f.clientConfig.ClientConfig()
}

func (f *ring0Factory) ClientConfigForVersion(requiredVersion *schema.GroupVersion) (*restclient.Config, error) {
	return f.clientCache.ClientConfigForVersion(nil)
}

func (f *ring0Factory) RESTClient() (*restclient.RESTClient, error) {
	clientConfig, err := f.clientCache.ClientConfigForVersion(nil)
	if err != nil {
		return nil, err
	}
	return restclient.RESTClientFor(clientConfig)
}

func (f *ring0Factory) FederationClientSetForVersion(version *schema.GroupVersion) (fedclientset.Interface, error) {
	return f.clientCache.FederationClientSetForVersion(version)
}

func (f *ring0Factory) FederationClientForVersion(version *schema.GroupVersion) (*restclient.RESTClient, error) {
	return f.clientCache.FederationClientForVersion(version)
}

func (f *ring0Factory) Decoder(toInternal bool) runtime.Decoder {
	var decoder runtime.Decoder
	if toInternal {
		decoder = api.Codecs.UniversalDecoder()
	} else {
		decoder = api.Codecs.UniversalDeserializer()
	}
	return decoder
}

func (f *ring0Factory) JSONEncoder() runtime.Encoder {
	return api.Codecs.LegacyCodec(api.Registry.EnabledVersions()...)
}

func (f *ring0Factory) UpdatePodSpecForObject(obj runtime.Object, fn func(*api.PodSpec) error) (bool, error) {
	// TODO: replace with a swagger schema based approach (identify pod template via schema introspection)
	switch t := obj.(type) {
	case *api.Pod:
		return true, fn(&t.Spec)
	case *api.ReplicationController:
		if t.Spec.Template == nil {
			t.Spec.Template = &api.PodTemplateSpec{}
		}
		return true, fn(&t.Spec.Template.Spec)
	case *extensions.Deployment:
		return true, fn(&t.Spec.Template.Spec)
	case *extensions.DaemonSet:
		return true, fn(&t.Spec.Template.Spec)
	case *extensions.ReplicaSet:
		return true, fn(&t.Spec.Template.Spec)
	case *apps.StatefulSet:
		return true, fn(&t.Spec.Template.Spec)
	case *batch.Job:
		return true, fn(&t.Spec.Template.Spec)
	default:
		return false, fmt.Errorf("the object is not a pod or does not have a pod template")
	}
}

func (f *ring0Factory) MapBasedSelectorForObject(object runtime.Object) (string, error) {
	// TODO: replace with a swagger schema based approach (identify pod selector via schema introspection)
	switch t := object.(type) {
	case *api.ReplicationController:
		return kubectl.MakeLabels(t.Spec.Selector), nil
	case *api.Pod:
		if len(t.Labels) == 0 {
			return "", fmt.Errorf("the pod has no labels and cannot be exposed")
		}
		return kubectl.MakeLabels(t.Labels), nil
	case *api.Service:
		if t.Spec.Selector == nil {
			return "", fmt.Errorf("the service has no pod selector set")
		}
		return kubectl.MakeLabels(t.Spec.Selector), nil
	case *extensions.Deployment:
		// TODO(madhusudancs): Make this smarter by admitting MatchExpressions with Equals
		// operator, DoubleEquals operator and In operator with only one element in the set.
		if len(t.Spec.Selector.MatchExpressions) > 0 {
			return "", fmt.Errorf("couldn't convert expressions - \"%+v\" to map-based selector format", t.Spec.Selector.MatchExpressions)
		}
		return kubectl.MakeLabels(t.Spec.Selector.MatchLabels), nil
	case *extensions.ReplicaSet:
		// TODO(madhusudancs): Make this smarter by admitting MatchExpressions with Equals
		// operator, DoubleEquals operator and In operator with only one element in the set.
		if len(t.Spec.Selector.MatchExpressions) > 0 {
			return "", fmt.Errorf("couldn't convert expressions - \"%+v\" to map-based selector format", t.Spec.Selector.MatchExpressions)
		}
		return kubectl.MakeLabels(t.Spec.Selector.MatchLabels), nil
	default:
		gvks, _, err := api.Scheme.ObjectKinds(object)
		if err != nil {
			return "", err
		}
		return "", fmt.Errorf("cannot extract pod selector from %v", gvks[0])
	}
}

func (f *ring0Factory) PortsForObject(object runtime.Object) ([]string, error) {
	// TODO: replace with a swagger schema based approach (identify pod selector via schema introspection)
	switch t := object.(type) {
	case *api.ReplicationController:
		return getPorts(t.Spec.Template.Spec), nil
	case *api.Pod:
		return getPorts(t.Spec), nil
	case *api.Service:
		return getServicePorts(t.Spec), nil
	case *extensions.Deployment:
		return getPorts(t.Spec.Template.Spec), nil
	case *extensions.ReplicaSet:
		return getPorts(t.Spec.Template.Spec), nil
	default:
		gvks, _, err := api.Scheme.ObjectKinds(object)
		if err != nil {
			return nil, err
		}
		return nil, fmt.Errorf("cannot extract ports from %v", gvks[0])
	}
}

func (f *ring0Factory) ProtocolsForObject(object runtime.Object) (map[string]string, error) {
	// TODO: replace with a swagger schema based approach (identify pod selector via schema introspection)
	switch t := object.(type) {
	case *api.ReplicationController:
		return getProtocols(t.Spec.Template.Spec), nil
	case *api.Pod:
		return getProtocols(t.Spec), nil
	case *api.Service:
		return getServiceProtocols(t.Spec), nil
	case *extensions.Deployment:
		return getProtocols(t.Spec.Template.Spec), nil
	case *extensions.ReplicaSet:
		return getProtocols(t.Spec.Template.Spec), nil
	default:
		gvks, _, err := api.Scheme.ObjectKinds(object)
		if err != nil {
			return nil, err
		}
		return nil, fmt.Errorf("cannot extract protocols from %v", gvks[0])
	}
}

func (f *ring0Factory) LabelsForObject(object runtime.Object) (map[string]string, error) {
	return meta.NewAccessor().Labels(object)
}

func (f *ring0Factory) FlagSet() *pflag.FlagSet {
	return f.flags
}

// TODO: We need to filter out stuff like secrets.
func (f *ring0Factory) Command() string {
	if len(os.Args) == 0 {
		return ""
	}
	base := filepath.Base(os.Args[0])
	args := append([]string{base}, os.Args[1:]...)
	return strings.Join(args, " ")
}

func (f *ring0Factory) BindFlags(flags *pflag.FlagSet) {
	// Merge factory's flags
	flags.AddFlagSet(f.flags)

	// Globally persistent flags across all subcommands.
	// TODO Change flag names to consts to allow safer lookup from subcommands.
	// TODO Add a verbose flag that turns on glog logging. Probably need a way
	// to do that automatically for every subcommand.
	flags.BoolVar(&f.clientCache.matchVersion, FlagMatchBinaryVersion, false, "Require server version to match client version")

	// Normalize all flags that are coming from other packages or pre-configurations
	// a.k.a. change all "_" to "-". e.g. glog package
	flags.SetNormalizeFunc(utilflag.WordSepNormalizeFunc)
}

func (f *ring0Factory) BindExternalFlags(flags *pflag.FlagSet) {
	// any flags defined by external projects (not part of pflags)
	flags.AddGoFlagSet(flag.CommandLine)
}

func (f *ring0Factory) DefaultResourceFilterOptions(cmd *cobra.Command, withNamespace bool) *kubectl.PrintOptions {
	columnLabel, err := cmd.Flags().GetStringSlice("label-columns")
	if err != nil {
		columnLabel = []string{}
	}
	opts := &kubectl.PrintOptions{
		NoHeaders:          GetFlagBool(cmd, "no-headers"),
		WithNamespace:      withNamespace,
		Wide:               GetWideFlag(cmd),
		ShowAll:            GetFlagBool(cmd, "show-all"),
		ShowLabels:         GetFlagBool(cmd, "show-labels"),
		AbsoluteTimestamps: isWatch(cmd),
		ColumnLabels:       columnLabel,
	}

	return opts
}

func (f *ring0Factory) DefaultResourceFilterFunc() kubectl.Filters {
	return kubectl.NewResourceFilter()
}

func (f *ring0Factory) SuggestedPodTemplateResources() []schema.GroupResource {
	return []schema.GroupResource{
		{Resource: "replicationcontroller"},
		{Resource: "deployment"},
		{Resource: "daemonset"},
		{Resource: "job"},
		{Resource: "replicaset"},
	}
}

func (f *ring0Factory) Printer(mapping *meta.RESTMapping, options kubectl.PrintOptions) (kubectl.ResourcePrinter, error) {
	return kubectl.NewHumanReadablePrinter(options), nil
}

func (f *ring0Factory) Pauser(info *resource.Info) ([]byte, error) {
	switch obj := info.Object.(type) {
	case *extensions.Deployment:
		if obj.Spec.Paused {
			return nil, errors.New("is already paused")
		}
		obj.Spec.Paused = true
		return runtime.Encode(f.JSONEncoder(), info.Object)
	default:
		return nil, fmt.Errorf("pausing is not supported")
	}
}

func (f *ring0Factory) ResolveImage(name string) (string, error) {
	return name, nil
}

func (f *ring0Factory) Resumer(info *resource.Info) ([]byte, error) {
	switch obj := info.Object.(type) {
	case *extensions.Deployment:
		if !obj.Spec.Paused {
			return nil, errors.New("is not paused")
		}
		obj.Spec.Paused = false
		return runtime.Encode(f.JSONEncoder(), info.Object)
	default:
		return nil, fmt.Errorf("resuming is not supported")
	}
}

func (f *ring0Factory) DefaultNamespace() (string, bool, error) {
	return f.clientConfig.Namespace()
}

const (
	RunV1GeneratorName                          = "run/v1"
	RunPodV1GeneratorName                       = "run-pod/v1"
	ServiceV1GeneratorName                      = "service/v1"
	ServiceV2GeneratorName                      = "service/v2"
	ServiceNodePortGeneratorV1Name              = "service-nodeport/v1"
	ServiceClusterIPGeneratorV1Name             = "service-clusterip/v1"
	ServiceLoadBalancerGeneratorV1Name          = "service-loadbalancer/v1"
	ServiceExternalNameGeneratorV1Name          = "service-externalname/v1"
	ServiceAccountV1GeneratorName               = "serviceaccount/v1"
	HorizontalPodAutoscalerV1Beta1GeneratorName = "horizontalpodautoscaler/v1beta1"
	HorizontalPodAutoscalerV1GeneratorName      = "horizontalpodautoscaler/v1"
	DeploymentV1Beta1GeneratorName              = "deployment/v1beta1"
	DeploymentBasicV1Beta1GeneratorName         = "deployment-basic/v1beta1"
	JobV1GeneratorName                          = "job/v1"
	CronJobV2Alpha1GeneratorName                = "cronjob/v2alpha1"
	ScheduledJobV2Alpha1GeneratorName           = "scheduledjob/v2alpha1"
	NamespaceV1GeneratorName                    = "namespace/v1"
	ResourceQuotaV1GeneratorName                = "resourcequotas/v1"
	SecretV1GeneratorName                       = "secret/v1"
	SecretForDockerRegistryV1GeneratorName      = "secret-for-docker-registry/v1"
	SecretForTLSV1GeneratorName                 = "secret-for-tls/v1"
	ConfigMapV1GeneratorName                    = "configmap/v1"
	ClusterRoleBindingV1GeneratorName           = "clusterrolebinding.rbac.authorization.k8s.io/v1alpha1"
	RoleBindingV1GeneratorName                  = "rolebinding.rbac.authorization.k8s.io/v1alpha1"
	ClusterV1Beta1GeneratorName                 = "cluster/v1beta1"
	PodDisruptionBudgetV1GeneratorName          = "poddisruptionbudget/v1beta1"
)

// DefaultGenerators returns the set of default generators for use in Factory instances
func DefaultGenerators(cmdName string) map[string]kubectl.Generator {
	var generator map[string]kubectl.Generator
	switch cmdName {
	case "expose":
		generator = map[string]kubectl.Generator{
			ServiceV1GeneratorName: kubectl.ServiceGeneratorV1{},
			ServiceV2GeneratorName: kubectl.ServiceGeneratorV2{},
		}
	case "service-clusterip":
		generator = map[string]kubectl.Generator{
			ServiceClusterIPGeneratorV1Name: kubectl.ServiceClusterIPGeneratorV1{},
		}
	case "service-nodeport":
		generator = map[string]kubectl.Generator{
			ServiceNodePortGeneratorV1Name: kubectl.ServiceNodePortGeneratorV1{},
		}
	case "service-loadbalancer":
		generator = map[string]kubectl.Generator{
			ServiceLoadBalancerGeneratorV1Name: kubectl.ServiceLoadBalancerGeneratorV1{},
		}
	case "deployment":
		generator = map[string]kubectl.Generator{
			DeploymentBasicV1Beta1GeneratorName: kubectl.DeploymentBasicGeneratorV1{},
		}
	case "run":
		generator = map[string]kubectl.Generator{
			RunV1GeneratorName:                kubectl.BasicReplicationController{},
			RunPodV1GeneratorName:             kubectl.BasicPod{},
			DeploymentV1Beta1GeneratorName:    kubectl.DeploymentV1Beta1{},
			JobV1GeneratorName:                kubectl.JobV1{},
			ScheduledJobV2Alpha1GeneratorName: kubectl.CronJobV2Alpha1{},
			CronJobV2Alpha1GeneratorName:      kubectl.CronJobV2Alpha1{},
		}
	case "autoscale":
		generator = map[string]kubectl.Generator{
			HorizontalPodAutoscalerV1Beta1GeneratorName: kubectl.HorizontalPodAutoscalerV1Beta1{},
			HorizontalPodAutoscalerV1GeneratorName:      kubectl.HorizontalPodAutoscalerV1{},
		}
	case "namespace":
		generator = map[string]kubectl.Generator{
			NamespaceV1GeneratorName: kubectl.NamespaceGeneratorV1{},
		}
	case "quota":
		generator = map[string]kubectl.Generator{
			ResourceQuotaV1GeneratorName: kubectl.ResourceQuotaGeneratorV1{},
		}
	case "secret":
		generator = map[string]kubectl.Generator{
			SecretV1GeneratorName: kubectl.SecretGeneratorV1{},
		}
	case "secret-for-docker-registry":
		generator = map[string]kubectl.Generator{
			SecretForDockerRegistryV1GeneratorName: kubectl.SecretForDockerRegistryGeneratorV1{},
		}
	case "secret-for-tls":
		generator = map[string]kubectl.Generator{
			SecretForTLSV1GeneratorName: kubectl.SecretForTLSGeneratorV1{},
		}
	}

	return generator
}

func (f *ring0Factory) Generators(cmdName string) map[string]kubectl.Generator {
	return DefaultGenerators(cmdName)
}

func (f *ring0Factory) CanBeExposed(kind schema.GroupKind) error {
	switch kind {
	case api.Kind("ReplicationController"), api.Kind("Service"), api.Kind("Pod"), extensions.Kind("Deployment"), extensions.Kind("ReplicaSet"):
		// nothing to do here
	default:
		return fmt.Errorf("cannot expose a %s", kind)
	}
	return nil
}

func (f *ring0Factory) CanBeAutoscaled(kind schema.GroupKind) error {
	switch kind {
	case api.Kind("ReplicationController"), extensions.Kind("Deployment"), extensions.Kind("ReplicaSet"):
		// nothing to do here
	default:
		return fmt.Errorf("cannot autoscale a %v", kind)
	}
	return nil
}

func (f *ring0Factory) EditorEnvs() []string {
	return []string{"KUBE_EDITOR", "EDITOR"}
}

func (f *ring0Factory) PrintObjectSpecificMessage(obj runtime.Object, out io.Writer) {
	switch obj := obj.(type) {
	case *api.Service:
		if obj.Spec.Type == api.ServiceTypeNodePort {
			msg := fmt.Sprintf(
				`You have exposed your service on an external port on all nodes in your
cluster.  If you want to expose this service to the external internet, you may
need to set up firewall rules for the service port(s) (%s) to serve traffic.

See http://kubernetes.io/docs/user-guide/services-firewalls for more details.
`,
				makePortsString(obj.Spec.Ports, true))
			out.Write([]byte(msg))
		}

		if _, ok := obj.Annotations[service.AnnotationLoadBalancerSourceRangesKey]; ok {
			msg := fmt.Sprintf(
				`You are using service annotation [service.beta.kubernetes.io/load-balancer-source-ranges].
It has been promoted to field [loadBalancerSourceRanges] in service spec. This annotation will be deprecated in the future.
Please use the loadBalancerSourceRanges field instead.

See http://kubernetes.io/docs/user-guide/services-firewalls for more details.
`)
			out.Write([]byte(msg))
		}
	}
}

// overlyCautiousIllegalFileCharacters matches characters that *might* not be supported.  Windows is really restrictive, so this is really restrictive
var overlyCautiousIllegalFileCharacters = regexp.MustCompile(`[^(\w/\.)]`)

// computeDiscoverCacheDir takes the parentDir and the host and comes up with a "usually non-colliding" name.
func computeDiscoverCacheDir(parentDir, host string) string {
	// strip the optional scheme from host if its there:
	schemelessHost := strings.Replace(strings.Replace(host, "https://", "", 1), "http://", "", 1)
	// now do a simple collapse of non-AZ09 characters.  Collisions are possible but unlikely.  Even if we do collide the problem is short lived
	safeHost := overlyCautiousIllegalFileCharacters.ReplaceAllString(schemelessHost, "_")

	return filepath.Join(parentDir, safeHost)
}