======= Review 1 ======= *** Paper Summary: Please summarize the paper in your own words. The authors consider networks with Byzantine faults, and derive algorithms = to achieve fault-tolerant communications over such networks. Two strategies to use node-disjoint paths are considered, i.e., message = replication and coding. *** Strengths: What are the main reasons to accept the paper? You may = comment on the importance of the problems addressed, the novelty of the = proposed solutions, the technical depth, potential impact, and the = writing. Your overall rating should be supported by your review. The paper proposes a set of algorithms to identify the optimal set of paths = to achieve reliable communication. The paper appears to be technically solid and to contain numerous = results. *** Weaknesses: What are the main reasons NOT to accept the paper? Again, = think about the importance of the problems addressed, the novelty of the = proposed solutions, the technical depth, potential impact, and the = writing. Your overall rating should be supported by your review. The introductory section of the paper is not well written. The authors fail = to convey the significance of the results obtained. Maybe even more importantly, analysis and results in the paper are not put = in context. It is not clear what scenario is being considered (Is this a = wired or a wireless network? what specific application scenarios are the = authors trying to capture with their model?). There is basically no comparison with existing literature, and very little = discussion of previous work, existing results, different approaches. The authors mention that conflict graphs have been used before, but they do = not provide much information to assess the significance of their = contribution. *** Additional Comments: Additional comments (if any) that you would like = to provide to the authors. Please do not repeat what you stated above. = If none, leave the following blank. Overall, this paper appears to be technically solid, but the reader is not = given enough information to assess the relevance of the contribution. *** Overall Rating: Your overall rating (Please try avoid giving = borderlines). borderline - top 30% of all INFOCOM submissions, but not top 20% (3) ======= Review 2 ======= *** Paper Summary: Please summarize the paper in your own words. A path selection technique, called iPath, is proposed in the paper to = address Byzantine fault tolerant issues in communication networks. With = the existence of malicious nodes or misbehaving nodes, they may report = conflicting status on paths and links. The approach is to use a so-called = conflict path to reduce the overall number of required node disjoint paths = to achieve fault tolerance. The proposed scheme is presented clearly and with theoretical analysis as = well as some numerical results. *** Strengths: What are the main reasons to accept the paper? You may = comment on the importance of the problems addressed, the novelty of the = proposed solutions, the technical depth, potential impact, and the = writing. Your overall rating should be supported by your review. The problem of reducing the required number of node disjoint multiple paths = is interesting. The use of conflict graph to achieve the goal is unique. = The paper is well written with a good mixture of analysis and discussions. = The numerical results are limited, probably due to the page limit. The discussions on symbol replication as well as MDS encoding on these = multiple paths are interesting. *** Weaknesses: What are the main reasons NOT to accept the paper? Again, = think about the importance of the problems addressed, the novelty of the = proposed solutions, the technical depth, potential impact, and the = writing. Your overall rating should be supported by your review. The number of available node disjoint paths can be limited. Such an issue = can be argued for or against the proposed scheme. There is no discussion = on the actual reduction of required node disjoint paths due to the use of = the iPath scheme. Such results can be really helpful for readers to = understand how well iPath works in practical systems. *** Additional Comments: Additional comments (if any) that you would like = to provide to the authors. Please do not repeat what you stated above. = If none, leave the following blank. The paper is well written with clear presentation of the scheme design as = well as its underlying reasons. The analysis looks interesting as = well. The number of node disjoint paths is sometimes limited. The authors should = find out how many of such paths might be available in any reasonable-density = networks. If such a number of way too small, then any multi path = transmission scheme will be in vain, eliminating the environment for which = iPath can function. Replication can be considered as a special case for MDS codes. There are = many prior work using MDS code for multiple path transmission of symbols. = The authors should cite some of these: Path diversity for enhanced media streaming JG Apostolopoulos, MD Trott - = Communications Magazine, …, 2004 Multipath key establishment for wireless sensor networks using just-enough = redundancy transmission J Deng, YS Han - Dependable and Secure Computing, = IEEE …, 2008 Multipath routing for multiple description video in wireless ad hoc networks = S Mao, YT Hou, X Cheng, HD Sherali… - INFOCOM 2005. As discussed briefly at the end of the paper, computation time of the = selection and encoding overhead can be excessive with all different = scenarios of multiple paths. Is there anyway to compute those offline or = ahead of time? If offline computation is impossible, computational overhead = should be investigated and discussed. *** Overall Rating: Your overall rating (Please try avoid giving = borderlines). weak accept - top 20% of all INFOCOM submissions, but not top 10% (4) ======= Review 3 ======= *** Paper Summary: Please summarize the paper in your own words. The paper studies the problem of reliable communication in the presence = byzantine faults. The paper uses ideas from system level diagnosis and = further expands them. Using a conflict graph constructed whenever a = conflict is identified the paper would first determine a set of paths for = reliable communication. The paper discusses how iPath can be used in = conjunction with error correction and detection to achieve performance = improvement. *** Strengths: What are the main reasons to accept the paper? You may = comment on the importance of the problems addressed, the novelty of the = proposed solutions, the technical depth, potential impact, and the = writing. Your overall rating should be supported by your review. The paper to the best of my knwledge is novel. This is probably the first = the ideas of system level diagnosis developed about fifty years in a = networking application *** Weaknesses: What are the main reasons NOT to accept the paper? Again, = think about the importance of the problems addressed, the novelty of the = proposed solutions, the technical depth, potential impact, and the = writing. Your overall rating should be supported by your review. I believe any algorithm that first identifies faults before routing should = necessarily be distributed and on line.There are such algorithms available = in the literature. The paper does not mention them. For example see: • K. Thulasiraman and Ming-Shan Su, “Diagnosis in a Network = of Processors: Centralized and Distributed Models and Algorithms,” = Journal of Combinatorial Mathematics and Combinatorial Computing, vol. 84, = pp. 167-178, February 2013. The bottlenck is the need to find a vertex cover. But this is a not problem = because there is an n2.5 algorithm for this because of the special nature = of the conflict graph. There also several ideas explored under specific = assumptions. They could be used in this appliation. Fo example see:A. = Das, K. Thulasiraman, V. K. Agarwal and K. B. Lakshmanan, = “Multiprocessor Fault Diagnosis Under Local Constraints”, = IEEE Trans. on Computers, Vol. 42, No. 8, Aug. 1993, pp. = 984-988. My only concern: the overhead involved in doing a fault identification = every time a conflict is noticed.If there ae several faulty nodes one may = have to do this several times before the path is identified Why not do this = periodically so that complete information, Here origianl system level = diagnosis ideas will be very useful. *** Additional Comments: Additional comments (if any) that you would like = to provide to the authors. Please do not repeat what you stated above. = If none, leave the following blank. Seems to pioneer a new direction *** Overall Rating: Your overall rating (Please try avoid giving = borderlines). strong accept - top 10% of all INFOCOM submissions (5)