* supersource for this class contains a field of type {@code C}.
 *
 * <p>For details about this hack, see {@code GwtSerializationDependencies}, which takes the same
 * approach but with a subclass rather than a superclass.
 *
 * <p>TODO(cpovirk): Consider applying this subclass approach to our other types.
 */
@SuppressWarnings("rawtypes") // https://github.com/google/guava/issues/989
@GwtCompatible(emulated = true)

For now we just have this global repository of ADRs.
If we see fit, we can break these out to per-platform ones, or keep a hybrid approach to having global and platform-specific ADSs.

Our aim is to keep the process lightweight and approachable.

Now, MinIO encrypts IAM / configuration and S3 objects with a KMS, if present. This
change unified the key-management aspect within MinIO.

The unified KMS-based approach has several advantages:

- Key management is now centralized. There is one way to change or rotate encryption keys.
   There used to be two different mechanisms - one for regular S3 objects and one for IAM data.

		// the 'for=', which we ignore, subsequently we proceed to look for
		// 'proto=' which should precede right after `for=` if not
		// we simply ignore the values and return empty. This is in line
		// with the approach we took for returning first ip from multiple
		// params.
		if match := forRegex.FindStringSubmatch(proto); len(match) > 1 {
			if match = protoRegex.FindStringSubmatch(match[2]); len(match) > 1 {

  val request: LauncherDiscoveryRequest =
    LauncherDiscoveryRequestBuilder.request()
      // TODO replace junit.jupiter.extensions.autodetection.enabled with API approach.
//    .enableImplicitConfigurationParameters(false)
      .selectors(selectors)
      .build()
  return request
}

/**

              ?, ? extends OneSizeTestContainerGenerator<BiMap<K, V>, Entry<K, V>>>
          parentBuilder) {
    List<TestSuite> derived = super.createDerivedSuites(parentBuilder);
    // TODO(cpovirk): consider using this approach (derived suites instead of extension) in
    // ListTestSuiteBuilder, etc.?
    derived.add(
        MapTestSuiteBuilder.using(new MapGenerator<K, V>(parentBuilder.getSubjectGenerator()))

 * variety of sequences of the {@link Iterator#next}, {@link Iterator#hasNext} and {@link
 * Iterator#remove} operations. This utility takes the brute-force approach of trying <i>all</i>
 * possible sequences of these operations, up to a given number of steps. So, if the caller
 * specifies to use <i>n</i> steps, a total of <i>3^n</i> tests are actually performed.
 *

 * variety of sequences of the {@link Iterator#next}, {@link Iterator#hasNext} and {@link
 * Iterator#remove} operations. This utility takes the brute-force approach of trying <i>all</i>
 * possible sequences of these operations, up to a given number of steps. So, if the caller
 * specifies to use <i>n</i> steps, a total of <i>3^n</i> tests are actually performed.
 *

    podAnnotations: {}

    # Deploy the config files as plugin chain (value "true") or as standalone files in the conf dir (value "false")?
    # Some k8s flavors (e.g. OpenShift) do not support the chain approach, set to false if this is the case
    chained: true

    # Custom configuration happens based on the CNI provider.
    # Possible values: "default", "multus"
    provider: "default"

    # Configure ambient settings

For example, it was clear that ideally it should be based on standard Python type hints.

Also, the best approach was to use already existing standards.

So, before even starting to code **FastAPI**, I spent several months studying the specs for OpenAPI, JSON Schema, OAuth2, etc. Understanding their relationship, overlap, and differences.