>in other words, it's up to the user of the memory to do this: >it's implementation specific. > >but you are right: novice programmers should be Warned :) > Ok so now here is my point. Why attempt to treat the two memory systems as one when they are two different kettle of fish. Especially since a novice developer may be very confused by all of this.
I tend to think to make this shared memory stuff work it should be treated as in-memory database minus the indexing and complex searching abilities. If you surf to the website http://www.ispras.ru/~knizhnik/ you will see some databases called gigabase, fastdb, post++, Goods, etc. What I like regarding his work is that it is very interesting from a shared memory point of view. He knows how to deal with inmemory C++ objects, pointer offsets and their serialization. It is C++, but I think it could be applied in C. And the best part of his sources are that a novice could use them without major issues. Christian Gross
