Handling user connections on Android - architecture tips

Question

I am implementing an android based chat. I want to make it almost as low level as possible. The reason for that is simple - i want to gain more knowledge on how things work. I am using sockets to connect to a server. With a single socket everything is working quite well but my question is: Will i need multiple connections when the user using the application opens multiple chat windows. If so - what is the best aproach for making those connections. I was thinking about using something like a Util class that opens a connection when needed but i'm still not quite sure what architecture this class must have. For example would it make sense to make it a singleton class? Will i be able to keep track of all the opened connections and close them when they are no longer needed. Any help would be appreciated.

PS If i missed something feel free to tell me what and i will try to edit the question to be as clear as possible.

Answer 1

This is clearly more of an architectural question, but I'll offer some thoughts. I would say this depends on your setup.

It sounds like you are connecting to your "contacts" directly, instead of using a central server. I assume you are determining and using the IP address directly to initiate a chat session. If this is the case, then yes, you will need to open a connection for each chat session you have going on.

If instead you are using a chat server, then theoretically, you only need one connection to that server. This server, of course, will require a connection to each user. Using a chat server requires more work on your part, but it could allow for a more user friendly experience. For example, registering your user name on the server would allow you to speak with other people via their usernames instead of having to know their IP. You still need to connect to your server via a well known IP address or DNS name, however.

As for class architecture, I highly recommend you check out something called "Dependency Injection". Dependency injection in practice usually means you interact with services and providers through an interface. The actual class that implements the interface is also written by you, and "injected" at run time. This allows you to decouple your application from a specific technology or protocol, which means that someday, you can replace your custom sockets implementation with, say, a web service implementation without having to make changes to the code that uses the service. In addition, most dependency injection frameworks allow you to specify how classes are instantiated and used when they are injected. You can use configuration to specify if one and only one class will ever be instantiated (effectively a singleton), or if a new class will be instantiated each time the service is requested.