*/
    FAILED,
  }

  /**
   * A listener for the various state changes that a {@link Service} goes through in its lifecycle.
   *
   * <p>All methods are no-ops by default, implementors should override the ones they care about.
   *
   * @author Luke Sandberg
   * @since 15.0 (present as an interface in 13.0)
   */
  abstract class Listener {
    /** Constructor for use by subclasses. */
    public Listener() {}

`ModelServer` collates all computation graphs exposed to it (from multiple
`SavedModel`) and executes them in parallel on available executors. Running
TensorFlow in a multitenant design mixes the risks described above with the
inherent ones from multitenant configurations. The primary areas of concern are
tenant isolation, resource allocation, model sharing and hardware attacks.

### Tenant isolation

So we are going to use that same knowledge to document how the *external API* should look like... by creating the *path operation(s)* that the external API should implement (the ones your API will call).

/// tip

When writing the code to document a callback, it might be useful to imagine that you are that *external developer*. And that you are currently implementing the *external API*, not *your API*.

* Response payloads.

You would also have **inline errors** for everything.

And whenever you update the backend code, and **regenerate** the frontend, it would have any new *path operations* available as methods, the old ones removed, and any other change would be reflected on the generated code. 🤓

   *
   * <p>NOTE: ISATAP addresses are explicitly excluded from this method due to their trivial
   * spoofability. With other transition addresses spoofing involves (at least) infection of one's
   * BGP routing table.
   *
   * @param ip {@link Inet6Address} to be examined for embedded IPv4 client address
   * @return {@code true} if there is an embedded IPv4 client address
   * @since 7.0
   */

        // spec says that we may repeat the request without modifications. Modern browsers also
        // repeat the request (even non-idempotent ones.)
        if (!client.retryOnConnectionFailure) {
          // The application layer has directed us not to retry the request.
          return null
        }

        val requestBody = userResponse.request.body

I'll tell you a bit more about these **concepts** here, and that would hopefully give you the **intuition** you would need to decide how to deploy your API in very different environments, possibly even in **future** ones that don't exist yet.

By considering these concepts, you will be able to **evaluate and design** the best way to deploy **your own APIs**.

    // have this Entry. The value in the map becomes false when this first entry has been copied, so
    // we know not to copy the remaining ones.
    IdentityHashMap<Entry<K, V>, Boolean> duplicates = null;
    int dupCount = 0;
    for (int entryIndex = n - 1; entryIndex >= 0; entryIndex--) {

- [AEAD](#aead): AES-256-GCM if the CPU supports AES-NI, ChaCha20-Poly1305 otherwise. More specifically AES-256-GCM is only selected for X86-64 CPUs with AES-NI extension.

Further any secret key (apart from the KMS-generated ones) is 256 bits long. The KMS-generated keys may be 256 bits but this depends on the KMS capabilities and configuration.

        request.setPath("third\\path\\");
        assertEquals("\\third\\path", request.getPath());

        // Multiple leading backslashes - only first one gets stripped, then trailing ones get stripped
        // But getPath() adds a backslash at the beginning, so \\fourth\path\\ becomes \fourth\path\
        request.setPath("\\\\fourth\\path\\\\");
        assertEquals("\\\\fourth\\path\\", request.getPath());