Yes, I'm on 2.6. I checked the Python 3.x tarfile just for this one line in TarFile.next():

self.members.append(tarinfo)

to conclude it would have the same problem.

Reducing 2.5gb memory usage as measured in my particular case by 60%, still leaves 1.5gb ram burned which is too much on a 32-bit 2gb ram machine.

My solution was to comment out that line which worked perfectly for my case but may not be the solution for the module.