Offloading the page load process to the cloud is not new to the mobile web. This paper proposes a measurement-based framework for offloading portions of the process, but unlike other browsers such as Opera Mini that offload fixed parts of page loads, this framework offers the ability to offload any portion of the process. The paper describes the opportunities and challenges involved in offloading, and covers the background of the page load process in comprehensive detail.

The paper has several strengths. It considers an interesting problem and is mostly well written. The authors provide a clear statement of the problem and why it is important. However, beyond that I found the rest of the paper disappointing, especially section 5 and beyond. Although the authors only propose their framework, it is very dependent on the measurement results described in section 5; however, there are no preliminary results, which makes the justification of the design choices questionable. Some of the tests could have been done even without the framework being ready. For example, the dependencies in the page load process are described as a limiting factor to offload work, but the authors could have attempted to measure the percent of work available to be offloaded to the cloud, and the dependencies in those pages, using the top 200 Alexa (http://www.alexa.com/) web pages reported in section 1. The actual location of cloud servers is mentioned as an important factor to reduce round-trip latencies, but again there is no empirical support for the feasibility of covering certain ranges and the effect of distance from the actual mobile device to the cloud servers. Lastly, there is not enough evidence to back up the idea of offloading any portion of the page load process to the cloud rather than offloading fixed parts as other mentioned browsers do. The paper does not make it clear in which metrics and scenarios they are slow.

A few issues require clarification or correction. For example, the next-to-last paragraph on the first page needs to better explain the “three metrics,” and the first paragraph of section 2.3 needs a citation for the “three key constraints” described. Finally, some abbreviations are used without giving the full name anywhere in the paper.

For the above reasons, I have reservations about recommending this paper, mainly because it is hard for me to pinpoint its technical research contributions and novelty.