WebTP: A User-Centered Web Transfer Protocol

点击翻译按钮获取中文摘要

The World Wide Web is now a critical communication infrastructure forsociety. However, the foundations on which the Web is built are largely adhoc and haphazard, with a protocol framework consisting of the hypertexttransport protocol (HTTP) layered on top of the Internet's universalreliable delivery protocol-the transmission control protocol (TCP).Consequently, the performance of HTTP over TCP suffers from a number ofdeficiencies that are all rooted in the fact that TCP's byte-streamsemantics are often at odds with user goals and the performancecapabilities of end clients and servers.Several ad hoc remedies have been proposed earlier to compensate for thesedeficiencies, e.g. parallel connections, persistent connections, and ACKpacing. However, all these solutions are overly constrained by thepresumption of backward compatibility. We believe that the time has come toadopt a holistic and integrated approach embodied in a new transportprotocol tailored specifically for the Web. Herein, we propose a set ofresearch tasks to study, design, implement, test and evaluate our new ``WebTransport Protocol" or WebTP.Our approach is based on a conceptual model of user validation, applicationsemantics, network and client states, and a flexible protocol that reflectsthis syntax and semantics. The conceptual model emphasizes the interactionbetween document structure, network state, and user preferences. Ourresearch prototype will leverage the well-established and highly successfulcongestion control framework of TCP and vastly improve upon the existingWebprotocols by bringing user requirements into the design loop. To do so, wewill apply the principles of Application Level Framing (ALF) to the designof WebTP. ALF says that to optimize communication paths within theapplication and across the network, an application's semantics (i.e. itsuser requirements) should be explicitly reflected in the design of itsnetwork protocol. Although ALF has been adopted in modern protocols likethe Real-time transport protocol (RTP), paradoxically it has not beensystematically applied to previously existing applications like the Web.We will study the following questions:* How do we best structure the ALF-based Web protocol in terms ofcongestion control algorithms, loss recovery schemes, packet framingmethodology, header formats, toolkit APIs, and so forth? How do thesedesign decisions impact network performance, user satisfaction, protocolcomplexity, and server load?* How then does the client elicit user preferences in the least obtrusivefashion?* How does the receiver (client) learn of the content structure of aserver's Web data and combine that structure with its estimate of networkstate and user preferences to control selective transmission via WebTP?* How should functionality and state information be distributed among thesource (server or proxy), client, and user and what are the design andperformance tradeoffs? How does WebTP scale with increased speed and fasterclients?Our approach to these questions draws upon network engineering andcommunication theory, signal processing and stochastic control. We willuse existing network emulation infrastructure from the the INDEX Project tocomprehensively test and evaluate WebTP. This project holds the potentialto dramatically impact the way we think about Web transport and willfurther contribute a design methodology, an actual protocol architecture,and performance analysis and results. We believe that our versions of WebTPwill demonstrate dramatic improvements and pave the way for furtherdevelopments.