GameDesign: Difference between revisions

and a we have created a PlayerEntry for this player with a unique clientid.

RuntimePlayerEntry is the structure that contains the information about the player while it is online. <br>RuntimePlayerEntry contains:

* <i>clientid</i><br>

Clientid is the field that indexes players in the server. See the documentation about clientid generation to understand what they are and how they are generated.

* <i>source</i><br>

Source is the IPv4 address of the client. This is used to determine if the message is really coming from the client or another person trying to impersonate it.

* <i>timestamp</i><br>

Timestamp is used to determine if a client has timed out in which case it is only wasting resources on the server. As you may already know, UDP is not a delivery-guaranteed protocol, so we need to check for dead clients ourselves. Note that this only indicates that the player timed out and it doesn't apply any kind of measures on them.

* <i>storedTimestamp</i><br>

storeTimestamp is used to determine when the player was last stored in the database. We don't store each time the player info changes as this would obviously be very CPU time consuming. Instead we cached the changes and store them only every 5 minutes.

* <i>username</i><br>

Username is filled in at runtime with a Login event. If we store the username here we are able to use the database from PlayerContainer thus by knowing the clientid we can also now know the username without having to look to the actual database.

* <i>choosenCharacter</i><br>

choosenCharacter is filled in at runtime with a ChooseCharacter event. If we store the information here we are able to use the database from PlayerContainer and hence by knowing the clientid we also know the choosenCharacter without having to refer to the actual database.

* <i>state</i><br>

**<b>Have to login

**Logout</b>

When we create the entity, by default, the state is <b>Have to login</b>. Once you have logged in correctly, the state changes to <b>Login Complete</b> and once the player has chosen a Character it changes to <b>game begin</b>. The logout state is pretty trivial :)

The idea is that some operations are only allowed in certain states, so the state property stores which state they are in to make validating actions easier.

The Perception counter is used to keep an incremental count of the perceptions sent so that the client can see if it gets out of sync.

Perception previous RPObject is the RPObject that was sent on the last perception. Using this we can track changes to a RPObject without disturbing the rest of the system.

This flag indicates to the server if the player has become out of sync. This allows us to re-sync it as soon as possible.

Hence, all we need to operate PlayerDatabase is a username and choosenCharacter. So using PlayerEntryContainer we can fully operate it.

Each client MUST have a session id to prevent another player impersonating it. sessionid must be of short or int size to make guessing the ID much harder.

To make it even more secure, clientids are generated randomly for each player with the only condition that two different players MUST have two different clientids.

Why bother with this? Well, imagine that a player logs out while the perception is being built, it will no longer be accessible by the RP Manager when it expects the object to be there, or if RPManager tries to remove a player which has already been removed, these situations are very serious as they will probably make the server fail.

So we must synchronize the Game and RP Managers.

The idea behind the solution is that the each manger requests access to the PlayerEntryContainer via a central mutex (a mutex is a syncronisation element attached to a resource, which can be owned by one task at any point in time. If the mutex is owned already when a task tries to access the object protected by it then the mutex will inform the task that it doesn't have access at this point in time to the object).

There are two types of accesses, read accesses and write accesses. Note that two readers can access an object in parallel but only one write can happen at the same time.

In GameManager there are only Write actions, as this mahanger only modifies the state of the PlayerContainer. However, in the RP manager we have both Reads, when we build the perceptions, and Writes when the manager removes idle players. Hence here we have two different locks.