Given these constraints, we cannot always follow best practices.  This ADR proposes the way we should handle the use of Provider APIs in the gradle/gradle codebase.

## Decision

### Types of properties that should not use lazy types

	public abstract fun getRequestBody ()Lokio/BufferedSource;
	public abstract fun getResponseBody ()Lokio/BufferedSink;
}

public abstract interface class mockwebserver3/StreamHandler {
	public abstract fun handle (Lmockwebserver3/Stream;)V

          <description>
            How to handle downloading of releases from this repository
          </description>
          <association>
            <type>RepositoryPolicy</type>
          </association>
        </field>
        <field>
          <name>snapshots</name>
          <version>1.0.0+</version>
          <description>
            How to handle downloading of snapshots from this repository

			// pointers to this type will not be considered pointers in Go. They won't
			// get writebarrier-ed or adjusted during a stack copy. This should handle
			// all the cases badPointerTypedef used to handle, but hopefully will
			// continue to work going forward without any more need for cgo changes.
			tt.Go = c.Ident(incomplete)
			typedef[name.Name] = &tt
			break
		}

Try to be as specific and objective as possible, avoid generalizations.

For conversations that are more difficult, for example to reject a PR, you can ask me (@tiangolo) to handle it directly.

## Edit PR Titles

* Edit the PR title to start with an emoji from <a href="https://gitmoji.dev/" class="external-link" target="_blank">gitmoji</a>.

   * of the function in [7.0, 10.0] is equivalent to the sum of the integrals of [7.0, 8.0], [8.0,
   * 9.0], [9.0, 10.0] (and so on), no matter what the function is. This guarantees that we handle
   * correctly requests of varying weight (permits), /no matter/ what the actual function is - so we
   * can tweak the latter freely. (The only requirement, obviously, is that we can compute its
   * integrals).
   *

			return
		case <-t.C:
			buckets, err := sys.objAPI.ListBuckets(ctx, BucketOptions{NoMetadata: true})
			if err != nil {
				internalLogIf(ctx, err, logger.WarningKind)
				break
			}

			// Handle if we have some buckets in-memory those are stale.
			// first delete them and then replace the newer state()
			// from disk.
			diskBuckets := set.CreateStringSet()
			for _, bucket := range buckets {

)

// Recycler will override the internal reuse in typed handlers.
// When this is supported, the handler will not do internal pooling of objects,
// call Recycle() when the object is no longer needed.
// The recycler should handle nil pointers.
type Recycler interface {
	Recycle()
}

// MSS is a map[string]string that can be serialized.
// It is not very efficient, but it is only used for easy parameter passing.
type MSS map[string]string

This "lazy loading" optimizes pod startup time, for the common case where an application is not sending outbound traffic instantly.
For applications that do, they will see a slight latency increase during startup; most applications can handle this far better than failed connections.

## Pod Shutdown

### Pod Shutdown Requirements

namespace)||d&&d!==k.selector&&("**"!==d||!k.selector)||(m.splice(f,1),k.selector&&m.delegateCount--,l.remove&&l.remove.call(a,k));g&&!m.length&&(l.teardown&&l.teardown.call(a,p,r.handle)!==!1||n.removeEvent(a,o,r.handle),delete i[o])}else for(o in i)n.event.remove(a,o+b[j],c,d,!0);n.isEmptyObject(i)&&(delete r.handle,L.remove(a,"events"))}},trigger:function(b,c,d,e){var f,g,h,i,k,m,o,p=[d||l],q=j.call(b,"type")?b.type:b,r=j.call(b,"namespace")?b.namespace.split("."):[];if(g=h=d=d||l,3!==d.nodeT...