Update MediaKeySession method names to latest EME spec.

Source/modules/encryptedmedia/MediaKeys.cpp

Index: Source/modules/encryptedmedia/MediaKeys.cpp
diff --git a/Source/modules/encryptedmedia/MediaKeys.cpp b/Source/modules/encryptedmedia/MediaKeys.cpp
index 285c58ac43759816a82d9feee78e1ae6a207b0e1..5daedaeadb64c378fa8570938640f619f688e7d0 100644
--- a/Source/modules/encryptedmedia/MediaKeys.cpp
+++ b/Source/modules/encryptedmedia/MediaKeys.cpp
@@ -78,10 +78,11 @@ MediaKeys::MediaKeys(const String& keySystem, PassOwnPtr<ContentDecryptionModule
 
 MediaKeys::~MediaKeys()
 {
+    // FIXME: This reference is outdated.

[ddorwin, 2014/01/07 02:41:55]

Yeah, we should not call release() here. WebCDM should be responsible for any
deinitialization.
There are also lifetime issues we'll need to make sure we get right.

For now, I think we should just remove this code. Maybe add a FIXME to make sure
things are torn down correctly.

[xhwang, 2014/01/09 01:04:41]

Done.

     // From <http://dvcs.w3.org/hg/html-media/raw-file/default/encrypted-media/encrypted-media.html#dom-media-keys-constructor>:
-    // When destroying a MediaKeys object, follow the steps in close().
+    // When destroying a MediaKeys object, follow the steps in release().
     for (size_t i = 0; i < m_sessions.size(); ++i)
-        m_sessions[i]->close();
+        m_sessions[i]->release();
 }
 
 PassRefPtr<MediaKeySession> MediaKeys::createSession(ExecutionContext* context, const String& type, Uint8Array* initData, ExceptionState& exceptionState)
@@ -90,35 +91,34 @@ PassRefPtr<MediaKeySession> MediaKeys::createSession(ExecutionContext* context,
     // The createSession(type, initData) method must run the following steps:
     // Note: The contents of initData are container-specific Initialization Data.
 
-    // 1. If type is null or an empty string and initData is not null or an empty string, throw an
+    // If type is null or an empty string and initData is not null or an empty string, throw an

[ddorwin, 2014/01/07 02:41:55]

Can type be null?
The parameters are no longer optional in the spec IDL. Please update the IDL and
test it. Then you can probably remove this step.

That leaves the empty string, which is not explicitly enforced by the spec. We
should probably just let supportsMIMEType() handle that (but make sure we have
tests!).

initData is not optional either (can't be NULL I hope), but it could
theoretically be empty. (It was never a "string". :) )

[xhwang, 2014/01/09 01:04:41]

This change will be non-tribial. Added FIXME.

[ddorwin, 2014/01/25 01:49:36]

This is now enforced by steps in the spec:
https://dvcs.w3.org/hg/html-media/rev/2e9c25eede16

     // InvalidAccessError exception and abort these steps.
     if ((type.isEmpty()) && (!initData || initData->length())) {
         exceptionState.throwUninformativeAndGenericDOMException(InvalidAccessError);
         return 0;
     }
 
-    // 2. If type contains a MIME type that is not supported or is not supported by the keySystem, throw
-    // a NotSupportedError exception and abort these steps.
+    // 1. If type contains a MIME type that is not supported or is not supported by the keySystem,
+    // throw a NOT_SUPPORTED_ERR exception and abort these steps.
     ASSERT(!type.isEmpty());
     if (type.isEmpty() || !m_cdm->supportsMIMEType(type)) {
         exceptionState.throwUninformativeAndGenericDOMException(NotSupportedError);
         return 0;
     }
 
-    // 3. Create a new MediaKeySession object.
+    // 2. Create a new MediaKeySession object.
     RefPtr<MediaKeySession> session = MediaKeySession::create(context, m_cdm.get(), this);
-    // 3.1 Let the keySystem attribute be keySystem.
+    // 2.1 Let the keySystem attribute be keySystem.
     ASSERT(!session->keySystem().isEmpty());
-    // 3.2 Let the sessionId attribute be a unique Session ID string. It may be generated by cdm.
-    // This is handled by m_cdm and may happen asynchronously.
+    // FIXME: 2.2 Let the state of the session be CREATED.
 
-    // 4. Add the new object to an internal list of session objects.
+    // 3. Add the new object to an internal list of session objects.
     m_sessions.append(session);
 
-    // 5. Schedule a task to generate a key request, providing type, initData, and the new object.
-    session->generateKeyRequest(type, initData);
+    // 4. Schedule a task to initialize the session, providing type, initData, and the new object.

[ddorwin, 2014/01/07 02:41:55]

Hmm, so I even used "initialize" in the spec... (I was struggling with wording,
and that was before loadSession() became a possibility.) At least in the spec,
this is used within the createSession() algorithm, so there is little ambiguity.

[xhwang, 2014/01/09 01:04:41]

Yeah, this is how I come up with this word in the first place :) The new name is
initializeNewSession.

+    session->initialize(type, initData);

[ddorwin, 2014/01/07 02:41:55]

If we are going to keep the asynchronicity in Blink, shouldn't the timer be
here? (Though that's not very appealing since we want to keep the deferred
parameters in the session class.)
OR, the MKS method could actually be called scheduleTask... or schedule...Task
and actually handle (or delegate) the scheduling as it currently does.

Basically, I feel like the comment and the method name don't match at all.

(Another argument for scheduling the task here is that the task is supposed to
be provided the new object. That's a minor issue, though.)

[xhwang, 2014/01/09 01:04:41]

Added FIXME.

 
-    // 6. Return the new object to the caller.
+    // 5. Return the new object to the caller.
     return session;
 }