Historically, Istio had to really be consumed all-or-nothing for things to work as expected.
In particular, an easy answer to "I just want to have mTLS everywhere, then I can think about adopting the rest of service mesh" was desired.

## Goals

Ztunnel should:
* **Not break users**. This means that deploying Ztunnel should retain all existing Kubernetes behavior.

# Body - Nested Models

With **FastAPI**, you can define, validate, document, and use arbitrarily deeply nested models (thanks to Pydantic).

## List fields

You can define an attribute to be a subtype. For example, a Python `list`:

=== "Python 3.10+"

    ```Python hl_lines="12"
    {!> ../../../docs_src/body_nested_models/tutorial001_py310.py!}
    ```

=== "Python 3.8+"

    ```Python hl_lines="14"

labels: [question]
body:
  - type: markdown
    attributes:
      value: |
        Thanks for your interest in FastAPI! 🚀

        Please follow these instructions, fill every question, and do every step. 🙏

        I'm asking this because answering questions and solving problems in GitHub is what consumes most of the time.

      seenExceptionsLocal = newConcurrentHashSet();
      /*
       * Other handleException() callers may see this as soon as we publish it. We need to populate
       * it with the initial failure before we do, or else they may think that the initial failure
       * has never been seen before.
       */
      addInitialException(seenExceptionsLocal);

      ATOMIC_HELPER.compareAndSetSeenExceptions(this, null, seenExceptionsLocal);
      /*

       *
       * - ReschedulableCallable has a reference back to its enclosing CustomScheduler. (It needs it
       *   so that it can call getNextSchedule).
       *
       * Maybe there is a way to avoid this cycle. But we think the cycle is safe enough to ignore:
       * Each task is retained for only as long as it is running -- so it's retained only as long as
       * it would already be retained by the underlying executor.
       *

      boolean isPossibleChainingCall = interfaceType.isAssignableFrom(method.getReturnType());
      try {
        Object actualReturnValue = method.invoke(wrapper, passedArgs);
        // If we think this might be a 'chaining' call then we allow the return value to either
        // be the wrapper or the returnValue.
        if (!isPossibleChainingCall || wrapper != actualReturnValue) {
          assertEquals(

      boolean isPossibleChainingCall = interfaceType.isAssignableFrom(method.getReturnType());
      try {
        Object actualReturnValue = method.invoke(wrapper, passedArgs);
        // If we think this might be a 'chaining' call then we allow the return value to either
        // be the wrapper or the returnValue.
        if (!isPossibleChainingCall || wrapper != actualReturnValue) {
          assertEquals(

      if (collection.size() <= 2L * k) {
        // In this case, just dumping the collection to an array and sorting is
        // faster than using the implementation for Iterator, which is
        // specialized for k much smaller than n.

        @SuppressWarnings("unchecked") // c only contains E's and doesn't escape
        E[] array = (E[]) collection.toArray();
        Arrays.sort(array, this);

## Version 4.6.0

_2020-04-28_

 *  Fix: Follow HTTP 307 and 308 redirects on methods other than GET and POST. We're reluctant to
    change OkHttp's behavior in handling common HTTP status codes, but this fix is overdue! The new
    behavior is now consistent with [RFC 7231][rfc_7231_647], which is newer than OkHttp itself.
    If you want this update with the old behavior use [this interceptor][legacy_interceptor].

  //    it's the same lock being acquired underneath. Always using
  //    monitor.enterXXX()/monitor.leave() will make it really clear
  //    which lock is held at any point in the code.
  //
  // 3. I think "enterWhen(notEmpty)" reads better than "notEmpty.enter()".
  //
  // TODO(user): Implement ReentrantLock features:
  //    - toString() method
  //    - getOwner() method
  //    - getQueuedThreads() method