Jesus Cea  wrote:
> In current Durus, packing is needed only to clean up:
>
> * Old instances, replaced with updated ones.
> * Garbage instances, unreachable from "root".
>
> The first is already solved with BerkeleyDB "per se". Now I'm thinking
> about the second one.

A FileStorage pack is equivalent to a simple two-space copying
garbage collector.  As I mentioned in my other message, you will
need some sort of garbage collection algorithm.  I'm pretty sure
there is lots of current gc research that tries to minimize virtual
memory paging.  I suspect those algorithms could be adapted to work
for your storage backend.  Maybe applicable:

    Garbage Collection Without Paging
    Matthew Hertz. Yi Feng. Emery D. Berger.

The "train" algorithm may also work:

    http://www.daimi.au.dk/~beta/Papers/Train/train.html

Coming up with a good gc for your file storage is probably a
research project in itself. :-)

  Neil