// should never reach here!
      throw new UnsupportedOperationException();
    }

    @Override
    public Entry<K, V> belowSamplesGreater() {
      // should never reach here!
      throw new UnsupportedOperationException();
    }

    @Override
    public Entry<K, V> aboveSamplesLesser() {
      // should never reach here!
      throw new UnsupportedOperationException();

      // should never reach here!
      throw new UnsupportedOperationException();
    }

    @Override
    public Entry<K, V> belowSamplesGreater() {
      // should never reach here!
      throw new UnsupportedOperationException();
    }

    @Override
    public Entry<K, V> aboveSamplesLesser() {
      // should never reach here!
      throw new UnsupportedOperationException();

   * reduced the available size of the buffer each time a call to read didn't fill up the available
   * space in the buffer completely. In general this is a performance problem since the buffer size
   * is permanently reduced, but with certain Reader implementations it could also cause the buffer
   * size to reach 0, causing an infinite loop.
   */

   * reduced the available size of the buffer each time a call to read didn't fill up the available
   * space in the buffer completely. In general this is a performance problem since the buffer size
   * is permanently reduced, but with certain Reader implementations it could also cause the buffer
   * size to reach 0, causing an infinite loop.
   */

   * free slot.
   *
   * The table size is capped because, when there are more threads
   * than CPUs, supposing that each thread were bound to a CPU,
   * there would exist a perfect hash function mapping threads to
   * slots that eliminates collisions. When we reach capacity, we
   * search for this mapping by randomly varying the hash codes of
   * colliding threads.  Because search is random, and collisions

		ChecksumAlgo: userDefined[hash.MinIOMultipartChecksum],
	}, nil
}

// NewMultipartUpload - initialize a new multipart upload, returns a
// unique id. The unique id returned here is of UUID form, for each
// subsequent request each UUID is unique.
//
// Implements S3 compatible initiate multipart API.
func (er erasureObjects) NewMultipartUpload(ctx context.Context, bucket, object string, opts ObjectOptions) (*NewMultipartUploadResult, error) {

   * free slot.
   *
   * The table size is capped because, when there are more threads
   * than CPUs, supposing that each thread were bound to a CPU,
   * there would exist a perfect hash function mapping threads to
   * slots that eliminates collisions. When we reach capacity, we
   * search for this mapping by randomly varying the hash codes of
   * colliding threads.  Because search is random, and collisions

   * free slot.
   *
   * The table size is capped because, when there are more threads
   * than CPUs, supposing that each thread were bound to a CPU,
   * there would exist a perfect hash function mapping threads to
   * slots that eliminates collisions. When we reach capacity, we
   * search for this mapping by randomly varying the hash codes of
   * colliding threads.  Because search is random, and collisions

If it is allowed, we will forward to the target destination.

#### Hairpin

In the case that the destination has a waypoint, that waypoint must have been bypassed to reach the inbound passthrough codepath.
How we handle this is [under discussion](https://docs.google.com/document/d/1uM1c3zzoehiijh1ZpZuJ1-SzuVVupenv8r5yuCaFshs/edit#heading=h.dwbqvwmg6ud3).

### Inbound

 *
 *  * [TCP handshake][connectSocket]
 *  * Optional [CONNECT tunnels][connectTunnel]. When using an HTTP proxy to reach an HTTPS server
 *    we must send a `CONNECT` request, and handle authorization challenges from the proxy.
 *  * Optional [TLS handshake][connectTls].
 *
 * Each step may fail. If a retry is possible, a new instance is created with the next plan, which
 * will be configured differently.
 */