In recent years, collaborative and group-oriented applications and protocols are gaining popularity. These applications typically involve communication over open networks, security thus is naturally an important requirement. Group key management is one of the basic building blocks in securing group communication. Most prior research in group key management focused on minimizing computation overhead, in particular minimizing expensive cryptographic operations. However, the continued advances in computing power have not been matched by a decrease in network communication delay. Thus, communication latency, especially in high-delay long-haul networks, is increasingly dominating the key setup latency, replacing computation delay as the main latency contributor. Hence, there is a need to minimize the size of messages and especially the number of rounds in cryptographic protocols. Since most previously proposed group key management techniques optimize computationalcryptographic overhead, they are particularly impacted by high communication delay. In this work, we discuss and analyze a specific group key agreement technique which supports dynamic group membership and handles network failures, such as group partitions and merges. This technique is very communication-efficient and provably secure against hostile eavesdroppers as well as various other attacks specific to group settings. Furthermore, it is simple, fault-tolerant and well-suited for high-delay networks.