A long time ago, I was once told that too much reliance on databases in programing can create slow programs. Databases provide a nice way to store and organize information but they may not retrieve and write information quickly. Clearly ram is the fastest way to access meomry but the size of ram is limited and their is overhead associated with transferring information between ram and the hard drive. One way to store information is to serialize it. For instance you can serialize objects. You often have the option to select between several formats and I presume that one format represents the way the information is stored in Ram. So my first question is with regards to pointers. For programing languages that can use pointers, does the pointer care if the pointer points to ram or non volatile memory or in order to treat non volatile memory as ram is it necessary to deal with virtual memory? This could be handy if you have a large object that would take considerable time to load (unserialize) into memory. I'm thinking if you have a data source, that is a binary file and you want to access that information though several sources (process, threads, separate programing languages).......a client server model might be good (not I'm not sure if this would be slow though), the server decides to keep the object on the hard disk or load into ram depending on the size of the object, the amount of free memory available and the demand for that object. I've scene some bridge programs between languages that use client sever models (com connect for instance) and they seem to pass strings around to tell the object what method to use. This strikes me as not the most efficient way to do things. I'm wondering about dynamic link libraries, are these accessible from multiple sources? I need to do more research, but I'd be interested in any comments people might have for managing large amounts of data between programs. Perhaps a good idea is to use a model like is use in com connect for programing languages that don't support pointers but give the option to get a pointer for programing languages that do. So for instance, on the java side of com connect, you can use the client server model but on the com side (visual basic, C, etc....) you can get a pointer and treat it as a native object.