*
 * <p>
 * <b>About NetBIOS:</b> The NetBIOS name
 * service is a dynamic distributed service that allows hosts to resolve
 * names by broadcasting a query, directing queries to a server such as
 * Samba or WINS. NetBIOS is currently the primary networking layer for
 * providing name service, datagram service, and session service to the
 * Microsoft Windows platform. A NetBIOS name can be 15 characters long

        Map<String, Lifecycle> phaseToLifecycleMap = new HashMap<>();

        for (Lifecycle lifecycle : getLifeCycles()) {
            for (String phase : lifecycle.getPhases()) {
                // The first definition wins.
                Lifecycle original = phaseToLifecycleMap.put(phase, lifecycle);
                if (original != null && logger.isWarnEnabled()) {
                    logger.warn(

      "FINE: Q10000 starting              : task",
      "FINE: Q10000 finished run in   0 µs: task",
    )
  }

  /** Repeat with a delay of 200 but schedule with a delay of 50. The schedule wins. */
  @Test fun executeScheduledEarlierReplacesRepeatedLater() {
    val task =
      object : Task("task") {
        val schedules = mutableListOf(50.µs)
        val delays = mutableListOf(200.µs, -1)

    // Start by looking for exact matches. We start at the leftmost label. For example, foo.bar.com
    // will look like: [foo, bar, com], [bar, com], [com]. The longest matching rule wins.
    var exactMatch: String? = null
    for (i in domainLabelsUtf8Bytes.indices) {
      val rule = publicSuffixListBytes.binarySearch(domainLabelsUtf8Bytes, i)
      if (rule != null) {
        exactMatch = rule

                }
                else { /* keep using the old list */
                    this.dcListExpiration = now + 1000 * 60 * 15; /* 15 min */
                    log.warn("Failed to retrieve DC list from WINS");
                }
            }

            int max = Math.min(this.dcList.length, this.netbiosLookupRespLimit);
            for ( int j = 0; j < max; j++ ) {
                int i = dcListCounter++ % max;

  /** If a cookie incorrectly defines multiple 'Max-Age' attributes, the last one defined wins.  */
  @Test fun lastMaxAgeWins() {
    assertThat(parseCookie(0L, url, "a=b; Max-Age=2; Max-Age=4; Max-Age=1; Max-Age=3")!!.expiresAt)
      .isEqualTo(3000L)
  }

  /** If a cookie incorrectly defines multiple 'Expires' attributes, the last one defined wins.  */
  @Test fun lastExpiresAtWins() {
    assertThat(
      parseCookie(

                        dc_list_expiration = now + 1000 * 60 * 15; /* 15 min */
                        if (SmbTransport.log.level >= 2) {
                            SmbTransport.log.println( "Failed to retrieve DC list from WINS" );
                        }
                    }
                }

                int max = Math.min( dc_list.length, LOOKUP_RESP_LIMIT );
                for (int j = 0; j < max; j++) {

            List<ArtifactRepository> remoteRepositories,
            ArtifactRepository localRepository)
            throws RepositoryMetadataStoreException {
        // TODO currently this is first wins, but really we should take the latest by comparing either the
        // snapshot timestamp, or some other timestamp later encoded into the metadata.

This enables multiple processes with the same effective UID to bind to the same port (see `man 7 socket` for details).
Linux will then send new connections to one of the processes (first to accept wins, so effectively random).

This still requires delicate sequencing.
Consider a rolling restart of the DaemonSet; the default configuration for Istio's installation.
We will do the following:

   * - 2 requests happen concurrently to a host that can be coalesced onto a single connection.
   * - Both request discover no existing connection. They both make a connection.
   * - The first request "wins the race".
   * - The second request discovers it "lost the race" and closes the connection it just opened.
   * - The second request uses the coalesced connection from request1.