1) size of checkpoint records (128 now as a place holder) 
2) random log reading during commit intentions ??
3) restart file duplication
4) crash recovery as a single pass (after short ckpt to eof scan)
5) redo only log --> redo prefered but undo supported