I'm assuming that the SIEVE checking ceases when an object is found - if so it feels like it'd overly aggressively clean the front of the list but under aggressively clean the tail of the list potentially leading to some rather frequently accessed tail objects getting stuck behind a load of cruft.
I also dislike simple boolean freshness indicators when TTLs are so computationally simple to maintain but yield much better results for large cache pools.
I'm assuming that the SIEVE checking ceases when an object is found - if so it feels like it'd overly aggressively clean the front of the list but under aggressively clean the tail of the list potentially leading to some rather frequently accessed tail objects getting stuck behind a load of cruft.
I also dislike simple boolean freshness indicators when TTLs are so computationally simple to maintain but yield much better results for large cache pools.