Full threadhantusk·Since writes to object storage are going to be slow anyway, why not double down on read optimized B-trees rather than write optimized LSM's?View on HN