Out of Memory a bridge too far

[Liam Routt]

Personally, the place I'd start (if I had time to work on it) would be 
in trying to change the way the data is loaded in the first place, so 
that there doesn't need to be enough memory available to load all of any 
pack file at one time. That would be a fundamental change which needn't 
be focused on any particular type of versioned file at all.

Really, bzr shouldn't need all the memory that 64-bit python can let you 
access in order to deal with these files - it should be able to do it in 
a predictable, almost constant, amount of memory, if the file access was 
designed with that in mind. At the very least I would have thought that 
if your files could each individually be loaded into memory you should 
not have a problem, but that is definitely not the case (because of the 
way changes are packed, I believe).

Take care,

Liam