* requests, responding to them with {@link #notifyStarted()} and {@link #notifyStopped()}
 * callbacks. Its subclasses must manage threads manually; consider {@link
 * AbstractExecutionThreadService} if you need only a single execution thread.
 *
 * @author Jesse Wilson
 * @author Luke Sandberg
 * @since 1.0
 */
@GwtIncompatible
@J2ktIncompatible

        SavedModel.
    *   The user object metadata field in the SavedModel proto has been
        deprecated as part of the updates to Keras SavedModel. Keras was the
        only consumer of this field prior to the update.
*   GPU
    *   TF 2.3 includes PTX kernels only for
        [compute capability](https://developer.nvidia.com/cuda-gpus) 7.0 to

 *
 * @author Louis Wasserman
 * @since 14.0
 */
@GwtCompatible
public abstract class BaseEncoding {
  // TODO(lowasser): consider making encodeTo(Appendable, byte[], int, int) public.

  BaseEncoding() {}

  /**
   * Exception indicating invalid base-encoded input encountered while decoding.
   *
   * @author Louis Wasserman

  // 'inputFuture'. That field is non-final and non-volatile. There are 2 places where the
  // 'inputFuture' field is read and where we will have to consider visibility of the write
  // operation in the constructor.
  //
  // 1. In the listener that performs the callback. In this case it is fine since inputFuture is

    }
  }

  /** Wait for the given thread to reach the {@link State#TIMED_WAITING} thread state. */
  @SuppressWarnings("ThreadPriorityCheck") // TODO: b/175898629 - Consider onSpinWait.
  void awaitTimedWaiting(Thread thread) {
    while (true) {
      switch (thread.getState()) {
        case BLOCKED:
        case NEW:
        case RUNNABLE:
        case WAITING:

     * linear-time operation, and it can even have side effects. Thus, we limit the special case to
     * List, which is _even more likely_ to have size() implemented to be fast and side-effect-free.
     *
     * We could consider recognizing specific other collections as safe (like ImmutableCollection,
     * except ContiguousSet!) or as not worth this optimization (CopyOnWriteArrayList?).
     */
    if (parts instanceof List) {

    }
  }

  /** Wait for the given thread to reach the {@link State#TIMED_WAITING} thread state. */
  @SuppressWarnings("ThreadPriorityCheck") // TODO: b/175898629 - Consider onSpinWait.
  void awaitTimedWaiting(Thread thread) {
    while (true) {
      switch (thread.getState()) {
        case BLOCKED:
        case NEW:
        case RUNNABLE:
        case WAITING:

 * is not guaranteed to traverse the elements of the MonitorBasedPriorityBlockingQueue in
 * any particular order. If you need ordered traversal, consider using {@code
 * Arrays.sort(pq.toArray())}. Also, method {@code drainTo} can be used to remove some or
 * all elements in priority order and place them in another collection.
 *

   * line, then it means that the area (time) is increased, thus storedPermits are more costly than
   * fresh permits, thus the RateLimiter becomes /slower/ after a period of underutilization.
   *
   * Last, but not least: consider a RateLimiter with rate of 1 permit per second, currently
   * completely unused, and an expensive acquire(100) request comes. It would be nonsensical to just

Then APIStar stopped to exist as a server and Starlette was created, and was a new better foundation for such a system. That was the final inspiration to build **FastAPI**.

I consider **FastAPI** a "spiritual successor" to APIStar, while improving and increasing the features, typing system, and other parts, based on the learnings from all these previous tools.

///