Robert

It seems like having a per-class, per-virtual file of fixed sized blockjects might ease the abstraction(s) somewhat.
It does tar archives too so compression is optional, not required.

Thanks

I am from Germany too. If you search in google for neodatis google ask you, if you meant neonazis. Maybe you should indeed rename it.

What about "jdb.net"

Example: if you had 4 byte blocks, a read at pos 3 would be in the 1st block. The next read at pos 5 would be in the 2nd. An update of bytes read from pos 3 to pos 5 would 1st re-write 1st block (read all 1-4 bytes, update 3-4 bytes and then write all 1-4 back) and then write the 2nd block (read all 5-8 bytes, update byte 5 and then write all 5-8 back).

I would expect this overhead to be slow and sure its complex, but I don't see a major problem.

The next file format version of ODB will have fixed size blocks.