Rolling out a ROApp - Programming GPS and OpenStreetMap Applications with Java - page 184

client. The client object is acting, while the server object is controlling and
reacting, i.e., the RealCar client is driving, while the ServerCar can assist
and propagate warnings, which is also known as ADAS . 5
The Java Tutorial trail RMI points out another option. A remote ap-
plication can actually transfer an entire object to the server and run it on
the server|without revealing the implementation.
An RO developer could program a (smart) player for a (larger) game
scenario and transfer the entire player to the server. This way the server
processor could replace the client's runtime environment, the developer
could shut down his computer overnight and see how his player did on the
next day. Then, he could get the player back on his local machine, improve
it, recompile it, etc.
This additional implementation makes a lot of sense in the context of
a real-world simulator, since it does not compete with current very fast
computer games. Think of a game on the German Autobahn where cars
try to catch each other, as in the LC game. The distance from north to
south can amount to 1,000 kilometers, which would take a car five to ten
hours to traverse depending on its type and the trac (broadcasted via
TMC). In this case, it would make perfect sense to transfer YourCar to
drive for some hours, then transfer it back to your computer and take over
the steering again.
For LondonChase , the intelligent players are yet to be developed. As
an example, look at LCPlayer.init and think of a server-side player. The
initialization would not serialize the GameSet . It would pass the reference
to the server's game set and use it like any other local object.
Local Server Agents
The Autobahn scenario can be used to introduce another type of server
object. At the opening of a scenario, you can't expect hundreds of clients
to enter at the same time. For the first clients, the scenario can be pretty
boring without other clients to interact with. You can't chase runners
before they actually enter the scenario, while a runner would always like
to observe a scenario before entering. The solution are actors running on
the server side.
If the scenario has a small number of real cars on the Autobahn, the
server can trigger a car factory to ll the gaps according to the car's speed.
These additional cars are smart enough to keep the proper distance and
adjust speed to trac. Population control would leverage the ratio of client
and server actors.
ServerObject s introduce even more variations to create real objects. An
external ROServer should not be able to distinguish a RealObject from a
5 Advanced Driver Assistance System.