Very good organization of the paper. Grammar is clear and concise. Some analysis may be a bit too general; not too much technical content...most of the analysis and examples are informal and high- level. But overall it is good, since the problem and solution and clearly defined. Very easy to follow. -------- On page 4 you define partial connection as "a connection that does not have the high bandwidth or quality of service of a full connection." This implies that a partial connection may be related to the load on the network. Is this the desired definition? On page 7 (bottom) with "If the client is taken off the fully connected network and moved to the wireless connection, then the application only has a partial connection," you are offering another alternative definition of a partial connection that is not necessarily consistent with the original one. Some wireless connections can offer excellent bandwidth and quality of service whereas some network connections may offer the opposite. This could have been solved by precisely defining definitions of fully connected and partially connected at the start of the paper. By combining the concepts that you have presented, you were able to arrive at the hybrid solution. One of the reasons that this works (and one that I felt needed to at least be stated if not explored further) was that the concepts are somewhat orthogonal. By this I mean that the assumptions that each technique operates under are different enough that they can resolved into an adaptive policy. I guess what I am getting at is that "why the hybrid solution works" needs to be addressed. This approach also addresses limitations of the hybrid solution which was not developed in the paper at all. Grammar was overall good, but some proof reading could have caught more inconsistencies (i.e. synergistic vs. synergetic) Overall the presentation and logical structuring was very good. By the way, I've never heard of a networked digital camera before. If there is such a thing, I would imagine that the camera supplies the information (i.e. uploading just taken pictures to a picture database) instead of receiving it thus making it a server. -------- - Many systems/techniques are analyzed. Some parts need more explanations. A Figure describing the properties of techniques would be nice for understandibiliy. - The hybrid approaches need justification. Some evaluations or references that use hybrid approaches would be nice. - It is not clear how the hybrid approaches used in some sections. e.g. section 3.3; building updates at the server. Is it a hybrid approach or totally a different approach? - There are several restatements of ideas. (both within a section and among sections) -------- I found this paper to be extremely well written and clear. The paper was well-organized and the different approaches to dealing with partial and total disconnection were clear. In addition, the comparisons of the different approaches helped clarify the differences between the approaches. The analyses were helpful, as they brought up the advantages and disadvantages of the approaches. The analyses were more in depth than just bringing up the advantages and disadvantages of the approaches, as they also described how the approaches could be used in conjunction. -------- What authors try suggesting is that by combining the approaches that are used for disconnected and partial connected operation and update propagation, it is possible to solve all problems which are used to be dealt separately by different approaches. The structure of paper is good in that it provides two or three approaches in three areas (disconnected, partially connected, and update), that explanation of those approaches are somewhat brief but well summarized enough to understand, and that they finally come up with an hybrid approach. I guess the hardest part of making this paper would be to find complement properties from each other. In section 3, they present the reasoning how it works in three area. It is obvious to see that their solutions make sense because their observations, complementary parts, seems OK to me. But the problem can be raised in real implementation environment, in that we only know there are complement approaches both of which can be integrated ideally, but we don't realize any possible confliction or resource overhead. -------- Overstated introduction, would rather hear about mobile data access than why users would use a mobile device. -------- Great examples for working set and good explination of the solution(i.e. access trees). The information is sound but tables graphs and especially alogoriths(especially algoriths!!) need to be used. On page 7 for example the authors refer to accessing file1.txt, I think some sort of psudo code for access or at least give a more hand on example. I'm not saying we need kernal code but a better graphical and hands on aproach is needed for this paper. Still the logic was well formed with very valuable information. -------- Mobile computing is a hot topic and an important issue is how to support mobile data access. In this paper, the authors described various techniques to solve data accessibility while disconnected or partially connected, and techniques to solve update propagation and conflict resolution upon reconnection. Generally the authors understood the materials well and the paper is well organized. But I think the title is not very precise. I choose the scores for the following reasons: Import: 6 Mobile computing is hot and will continue to be hot in the future. But their topics on disconnected and partially connected environment is not the most important issues in mobile computation now. Novelty: 4 Basically we have seen all the approaches before such as preparing for disconnection and replaying upon reconnection Quality: 6 I like the organization of this paper. First introduce pure approaches and then come up with hybrid systems. Each approach is described clearly and explained why such an approach is chosen. But these approaches can not solve all the problems in mobile data access. But the title suggest that disconnected and partially connected approaches are capable of solving all the problems. The format is not very formal Overall: 6 -------- The paper was entertaining to read, and provided a clear picture of the challanges inherent in mobile data access design. However, the ordering of the paper was a little ackward with its single long section that loses the reader under layers of subsections. Maybe it would be clearer if this section was divided into a theory and an application section. It seemed as though the research was well done, and all relevant articles cited. The paper did not present new ideas, it simply focused on existing ideas. Overall, a very good paper on a relevant topic. -------- This paper is about methods of providing data availability to a client in the absence of a link or connection by a slow/unreliable link to a server. The paper explains and cites usage of two disconnected techniques: predictive caching and hoarding of files based on some database of known desirable files. Next, it explains the partially connected techniques of caching individual records in a file, logging changes to files for later replay, both for reduce required bandwidth. Another technique for working in a partially connected network is allowing the user to specify what quality of service is acceptable and the operating system can signal user programs when the quality of service cannot be met. Third, a technique is described for update propagation that uses an intermediary (surrogate client) that is well connected to a server to perform the update operations. Last a hybrid approach is defined to combine the presented techniques into "best-of-both-worlds" techniques. This paper does a good job explaining the problems that the different data availability approaches address. Topics in general are explained very clearly. I liked the section on the hybrid approach. In addition to explaining the benefits of joining earlier defined approaches, it provides a recap on some details of the previous section which makes it easier to read. One problem I had was with section 2.2.2. The "fetch-only approach" should be introduced in section 2.2 or that subsection should be removed. Currently only the PFS and Odyssey are mentioned and only those sections' purposes are defined. Also, the section headings don't exactly reflect what the parent parent sections' introductions say, making it harder for me to follow the discussion. For example, take "2.1 Preparing for Disconnection" and "2.2.3 Agile Application-Aware Adaptation with Odyssey". While these titles accurately reflect the information that is in the subsection, it is not explicitly tied to the introduction. I think an example of a better section heading is "2.3 Update Propagation and Conflict Resolution" -- which is explicitly mentioned in the intro. -------- 1) This paper is highly inspired from CODA paper which we read in class. It served well to refresh our memories. 2) One of the main concerns in Mobile Data access is security. There is no insight into that aspect. 3) The main problem in the hybrid approach for partial connectivity is that it won't support application concurrency unless a single point of resource control is done. Since each application would react separately, these applications would perform poorly in concurrent execution. For example if there are 2 Video players running at the same time and both react to increase in B/W by increasing fidelity, even though B/W has increased sufficiently only for one client. The paper fails to highlight this important point. 4) The paper provides no information on agility when comparing the different approaches. Agility gives the measure of time required to arrive at a decsion in adaptive system. To operate in a turbulent environment a mobile system must be highly agile to change in demand for and supply of resources making up the environment. 5) The paper provides no insight into Laissez-Faire Adaption mechanism. 6) The file pre-fetching (described in the paper) is very CPU intensive and Mobile devices genrally have a low power cpu and would also want to conserve CPU power as much as possible. This point was not highligted in the paper. 7) It was not clear to me as how Fetch-only approach is actually done. -------- Perhaps the hybrid approach could use more detail. In suggesting the two systems that could be combined, the possible matches or conflicts could be addressed.