This paper presents a new technique for efficiently controlling replicas in distributed systems. conventional structured coterie protocols are efficient but incur a penalty of reduced availability in exchange for the performance gain. further, the performance advantage can only be fully realized when write operations always replace the old data item with new value instead of updating a portion of the data item. our new approach significantly improves availability while allowing partial write operations. after presenting our general approach, we apply it to an existing structured coterie protocol and analyze the availability of the resulting protocol. we also show that other classes of protocols can make use of our approach.