This paper is a general survey about some techniques for improving the performance of web caches. First the reasons for HTTP caching are mentioned, then both the cost and benefits of using caches are analyzed. Then details are given about the limits of caching and several new ideas of exploiting more locality from cache. Finally, a report is given about the idea of duplicate response suppression and its performance on a trace.

The strength of this paper:
1. It is a good survey in telling people about the cache, its benifit and cost, and what are the possible methods to improve the caching performance and their limits. So we can get a good view of HTTP caching after reading this paper, we can know its basic mechanism and the state-of-art of this field.

2. The structure of this paper is quite clear. Usually the authors give several reasons or methods for an issue and they are parallel in form. So it is quite easy for readers to understand and grasp the main idea of the paper and most of its sections. For the reasons of HTTP caching and the possible reasons of the limit of HR and BHR, several items are given in a clear way.

THe weakness of this paper:
1. For most part of the paper except for the last part of introducing duplicate suppression, no figures or pictures are given. So this is not good, the readers must have strong back ground to understand them quickly. And fewer examples are given, so the readers will feel too abstract for most of time. For example, if some simple figures or example can be given when introducing HTML macro or introducing partical transfer.

2. The comparisons are not enough as a survey like this paper. For example, when telling the methods of prefetching, two basic ideas are mentioned, one is the history based method, another is the link based method. THe paper only tells us how these method are, but doesn't say more about the advantages and disadvantages of these two methods. If several sentences are given for this, the readers may get a better view of these two ideas.

3. In the last part, the authors mentioned a new idea: alias discovery and duplicate suppression, no significant benifit can be convinced for the readers. One aspect is that the result is not very promising, another aspect is that no many experiments are done for this method, at least for this paper. So we are not sure if this is a good method. Even the authors themselves did not give high praise for their own idea, how could other people use it?

4. Almost no future work or prediction are mentioned. Although it is hard to predict the future, the authors read so many papers in this field, they should give some ideas about the trend of this filed. Sometimes, this information is useful for other researchers.

5. The author did not give the effect of combination of several ideas in the paper. Or the possibilities of this kind of combination. For example, can we use the prefetching to get the partial content, or for the delta transfer? Can their duplicate suppression be used in the delta encoding or prefetching.

Generally speaking, this paper is a quite good survey. It covers a lot in the area of HTTP caching. Both beginners and special researchers can get useful information from this paper.