When loading untrusted serialized computation graphs (in form of a `GraphDef`,
`SavedModel`, or equivalent on-disk format), the set of computation primitives
available to TensorFlow is powerful enough that you should assume that the
TensorFlow process effectively executes arbitrary code.

The risk of loading untrusted checkpoints depends on the code or graph that you

    img: https://fastapi.tiangolo.com/img/sponsors/stainless.png
bronze:
  - url: https://www.exoflare.com/open-source/?utm_source=FastAPI&utm_campaign=open_source
    title: Biosecurity risk assessments made easy.
    img: https://fastapi.tiangolo.com/img/sponsors/exoflare.png
  - url: https://testdriven.io/courses/tdd-fastapi/
    title: Learn to build high-quality web apps with best practices

as a Kubernetes DaemonSet.

At a high level, our goal is to provide complete connectivity to a workload throughout its entire lifetime.
Failing to do so can be an availability risk (if we deny traffic that should succeed) or a security risk (if we allow traffic that should be denied).

## High level overview

At a high level, the relevant components look as such:

```mermaid
flowchart TD
    CNIP["CNI Plugin"]

	var wg sync.WaitGroup
	wg.Add(1)

	// Start the callback in a new green thread so that if we later fail
	// to switch the namespace back to the original one, we can safely
	// leave the thread locked to die without a risk of the current thread
	// left lingering with incorrect namespace.
	var innerError error
	go func() {
		defer wg.Done()
		runtime.LockOSThread()
		innerError = containedCall()
	}()
	wg.Wait()

			if ch, ok := val.(chan struct{}); ok && ch != nil {
				// Close previous non-nil channel.
				xioutil.SafeClose(ch)
			}
		}
		globalServiceFreezeCnt = 0 // Don't risk going negative.
	}
	globalServiceFreezeMu.Unlock()

   * - Make checkOpen return the non-null `seq`. Then callers can assign that to a local variable or
   *   even back to `this.seq`. However, that may suggest that we're defending against concurrent
   *   mutation, which is not an actual risk because we use `synchronized`.
   * - Make `remaining` require a non-null `seq` argument. But this is a bit weird because the
   *   method, while it would avoid the instance field `seq` would still access the instance field

   * - Make checkOpen return the non-null `seq`. Then callers can assign that to a local variable or
   *   even back to `this.seq`. However, that may suggest that we're defending against concurrent
   *   mutation, which is not an actual risk because we use `synchronized`.
   * - Make `remaining` require a non-null `seq` argument. But this is a bit weird because the
   *   method, while it would avoid the instance field `seq` would still access the instance field

  final void interruptTask() {
    // Since the Thread is replaced by DONE before run() invokes listeners or returns, if we succeed
    // in this CAS, there's no risk of interrupting the wrong thread or interrupting a thread that
    // isn't currently executing this task.
    Runnable currentRunner = get();
    if (currentRunner instanceof Thread) {

* For large API changes, 2 members of the TOC must approve the PR before release manager approval in the release branch
  of the istio/api repository. This does not have to wait for the weekly TOC meeting.
    * Risk should be assessed in the PR.
        * Have installs and upgrades affected by this feature?
        * Is the feature still being worked on?
        * Is the default behavior altered?
        * Is this turned on by default?

IMG="${IMG:-${TOOLS_REGISTRY_PROVIDER}/${PROJECT_ID}/${IMAGE_NAME}:${IMAGE_VERSION}}"

CONTAINER_CLI="${CONTAINER_CLI:-docker}"

# Try to use the latest cached image we have. Use at your own risk, may have incompatibly-old versions
if [[ "${LATEST_CACHED_IMAGE:-}" != "" ]]; then
  prefix="$(<<<"$IMAGE_VERSION" cut -d- -f1)"
  query="${TOOLS_REGISTRY_PROVIDER}/${PROJECT_ID}/${IMAGE_NAME}:${prefix}-*"