Working principles of Android Push NotificationSys

I have two questions related to Android Push Notification System:

What is the working principle of this system? The client sends its own IP to the Google Cloud Messaging Service (for example when it switches its own IP)? So it a sort of pooling?.
How do you know that Google Cloud Messaging Service "looks into" the content of the notification message (created in the server and dispatched to the client)?

回答1:

Answering the question about how GCM service contacts the client, the GCM client contacts GCM to create the connection. You are correct that device addresses change as the device disconnects and reconnects so GCM cannot initiate the connection from the server side.

This connection is maintained as much as possible and is not created for specific messages.

The registration id identifies the device and app and allows GCM to route the message to the device, if it is connected. If it is not connected, GCM needs to wait until the device reconnects.

回答2:

What is the working principle of this system? The client sends its own IP to the Google Cloud Messaging Service (for example when it switches its own IP)? So it a sort of pooling?

I'm adding an image here describing how GCM works.

This is a step by step presentation. You need to get the push registration ID first when your application launches. So if you've a backend server to send some push notification in your application, you need to pass the registration ID to your backend server. So when you need to send a push notification, your backend server will send the push directly to GCM with the targeted registration ID. GCM manages to push the notification in your device when your device comes online.

So this is not any kind of pooling. The only thing GCM needs to know is the registration ID of your device when it comes online and tries to communicate with GCM. Once your device is registered, GCM sends the push notification using that registration ID.

How do you know that Google Cloud Messaging Service "looks into" the content of the notification message (created in the server and dispatched to the client)?

This question is not very clear to me. As far as I have understood, you wanted to know how GCM understands to whom it needs to send the push notification when the notification is coming from your backend server. If this is your question, then I think I have answered it already in the previous section of my answer.

GCM doesn't need to look into your notification content to know the destination of the push notification. As I said earlier, when your application launches, it requests for a push registration ID from GCM and when it receives an registration ID, you might have to pass the registration ID by calling a service of your backend server. The server then knows to whom it might send some notification.

So, when its time to send a notification to your client application, the backend server sends the notification to GCM with the registration ID you sent to your backend server earlier. GCM then handles sending the push notification to the client when the application comes online.

Hope that helps!