Communication system for maintaining data consistency among replicated shared objects

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT The present invention will now be described, by way of example.
The standard implementation of a client-server approach is to employ small pieces of code (sometimes termed message stubs) in the client and server processes which interface with some form of inter-process communications (IPC).
The purpose of the code in the client process is purely to relay any requests on the object in the server process as messages to the server process using the IPC mechanism.
The IPC mechanism will queue messages at the server process, with the small amount of code in the server process reading this queue and directing the requests to the object to service them.
The whole process can be thought of as a pipe which takes in requests at one end from the client process(es) and delivers them at the other end of the pipe to the appropriate object in the server process.
All activity required to service the requests (rather than merely transport it) is handled in the server process.
The implementation of the shared object approach replaces the client-server `pipe` with a mechanism which deals with requests locally within the process using the shared object.
When such requests modify the internal data of the shared object, only then will inter-process communications be utilised to ensure other copies of the object are brought into alignment with the changes.
Objects that are to be shared within a system must be identified as such at compile time.
This is achieved by defining a shared object class from which objects wishing to exhibit this characterstic may inherit.
Whenever a shared object is created, a copy of that object is created within the requesting process.
If this is the first time that object is created then two additional items are created; a copy of the object's data is created in share memory along with a used-by table into which an identifier for the requesting process is written.
Subsequent creations of this shared object by other processes will result in a copy of the object being created within the requesting process whose internal data will be an exact copy of the data stored in shared memory, and the used-by table is updated with the requesting processes' ID