The traditional TCP sender-driven approach to data communication in transport protocols can lead to ambiguity between the sender and receiver regarding packet delivery status. This issue stems primarily from the sender relying on explicit feedback from the receiver in the form of cumulative acknowledgments. While optimizations such as SACK can mitigate this issue to some extent, ambiguity may still arise due to receiver reneging or under high-loss environments, including retransmission loss. Recent research in transport protocols and new architectures has highlighted the advantages of using a receiver-driven approach over a sender-driven one for Internet communication. This shifts the traditional push-based data retrieval paradigm to a pull-based paradigm, allowing the creation of new transport services such as transparent caching and multicasting. This paper builds on these efforts to abstract and formalize a receiver-driven ARQ (RDQ) that follows established end-to-end principles in transport protocols, providing in-order reliability from the perspective of the receiver. We present the design of RDQ, layered on top of UDP, leveraging sender-driven and receiver-driven protocol elements. RDQ is implemented in the ns-3 simulator and evaluated against TCP- and SACK-style sender-driven ARQ under high-loss conditions. The preliminary results indicate the feasibility of incorporating a receiver-driven ARQ with a classic retransmission strategy in transport protocols, offering positive gains in reduced recovery delay and transmission efficiency relative to TCP/SACK-style sender-driven ARQ.
Abdulazaz Albalawi (Fri,) studied this question.