Rationale: with the current code most revision ids start with 1, 2, 3 or 4, and there are far more collisions with 2 initial digits than necessary (cumbersome to check a file by revision on command line and such). So instead of revision ids that are 1-12 digits long with ~45 bits of information, lets just have them 12 digits long and contain full 48 bits of info.
def rev_id():
- val = int(uuid.uuid4()) % 100000000000000
- return hex(val)[2:-1]
+ return uuid.uuid4().hex[-12:]
def to_list(x, default=None):