Question from the Node.js test

I use two callbacks, the first for the error, the second for the result.

if (err) { 
    error_callback(err); 
}
else{ 
    success_callback(data); 
}

The result as the first argument. If there is an error, I throw an exception.

if (err) { 
    throw new Error(err); 
}
else { 
    callback(data); 
}

If no error, data as first argument, and error as second argument(or null if no error).

if (err) { 
    callback(null, err); 
}
else { 
    callback(data, null); 
}

The error as first argument(or null if no error), and if no error, the result as second argument.

if (err) { 
    callback(err); 
}
else { 
    callback(null, data); 
}