No description available.

The Internet consists of thousands of autonomous networks connected together to provide end-to-end reachability. Networks of different sizes, and with different functions and business objectives, interact and co-exist in the evolving "Internet Ecosystem". The Internet ecosystem is highly dynamic, experiencing growth (birth of new networks), rewiring (changes in the connectivity of existing networks), as well as deaths (of existing networks). The dynamics of the Internet ecosystem are determined both by external "environmental" factors (such as the state of the global economy or the popularity of new Internet applications) and the complex incentives and objectives of each network. These dynamics have major implications on how the future Internet will look like. How does the Internet evolve? What is the Internet heading towards, in terms of topological, performance, and economic organization? How do given optimization strategies affect the profitability of different networks? How do these strategies affect the Internet in terms of topology, economics, and performance? In this thesis, we take some steps towards answering the above questions using a combination of measurement and modeling approaches. We first study the evolution of the Autonomous System (AS) topology over the last decade. In particular, we classify ASes and inter-AS links according to their business function, and study separately their evolution over the last 10 years. Next, we focus on enterprise customers and content providers at the edge of the Internet, and propose algorithms for a stub network to choose its upstream providers to maximize its utility (either monetary cost, reliability or performance). Third, we develop a model for interdomain network formation, incorporating the effects of economics, geography, and the provider/peer selections strategies of different types of networks. We use this model to examine the "outcome" of these strategies, in terms of the topology, economics and performance of the resulting internetwork. We also investigate the effect of external factors, such as the nature of the interdomain traffic matrix, customer preferences in provider selection, and pricing/cost structures. Finally, we focus on a recent trend due to the increasing amount of traffic flowing from content providers (who generate content), to access providers (who serve end users). This has led to a tussle between content providers and access providers, who have threatened to prioritize certain types of traffic, or charge content providers directly -- strategies that are viewed as violations of "network neutrality". In our work, we evaluate various pricing and connection strategies that access providers can use to remain profitable without violating network neutrality.

Personal data is increasingly conceived as a tradable asset. Markets for personal information are emerging and new ways of valuating individuals' data are being proposed. At the same time, legal obligations over protection of personal data and individuals' concerns over its privacy persist. This article outlines some of the economic, technical, social, and ethical issues associated with personal data markets, focusing on the privacy challenges they raise.

We consider the problem of designing revenue-optimal mechanisms for selling two heterogeneous items to a single buyer. Designing a revenue-optimal mechanism for selling a single item is simple: Set a threshold price based on the distribution, and sell the item only when the buyer’s valuation exceeds the threshold. However, designing a revenue-optimal mechanism to sell two heterogeneous items is a harder problem. Even the simplest setting with two items and one buyer remains unsolved as yet. The partial characterizations available in the literature have succeeded in solving the problem largely for distributions that are bordered by the coordinate axes. We consider distributions that do not contain (0; 0) in their support sets. Specifically, we consider the buyer’s valuations to be distributed uniformly over arbitrary rectangles in the positive quadrant. We anticipate that the special cases we solve could be a guideline to un-derstand the methods to solve the general problem. We explore two different methods – the duality method and the virtual valuation method – and apply them to solve the problem for distributions that are not bordered by the coordinate axes. The thesis consists of two parts. In the first part, we consider the problem when the buyer has no demand constraints. We assume the buyer’s valuations to be uniformly distributed over an arbitrary rectangle [c1; c1 + b1] [c2; c2 + b2] in the positive quadrant. We first study the duality approach that solves the problem for the (c1; c2) = (0; 0) case. We then nontrivially extend this approach to provide an explicit solution for arbitrary nonnegative values of (c1; c2; b1; b2). We prove that the optimal mechanism is to sell the two items according to one of eight simple menus. The menus indicate that the items must be sold individually for certain values of (c1; c2), the items must be bundled for certain other values, and the auction is an interplay of individual sale and a bundled sale for the remaining values of (c1; c2). We conjecture that our method can be extended to a wider class of distributions. We provide some preliminary results to support the conjecture. In the second part, we consider the problem when the buyer has a unit-demand constraint. We assume the buyer’s valuations (z1; z2) to be uniformly distributed over an arbitrary rectangle [c; c + b1] [c; c + b2] in the positive quadrant, having its south-west corner on the line z1 = z2. We first show that the structure of the dual measure shows significant variations for different values of (c; b1; b2) which makes it hard to discover the correct dual measure, and hence to compute the solution. We then nontrivially extend the virtual valuation method to provide a complete, explicit solution for the problem considered. In particular, we prove that the optimal mechanism is structured into five simple menus. We then conjecture, with promising preliminary results, that the optimal mechanism when the valuations are uniformly distributed in an arbitrary rectangle [c1; c1 + b1] [c2; c2 + b2] is also structured according to similar menus.

A domain name is a unique alphanumeric designation that facilitates reference to sets of numbers which actually locate a particular computer on the Internet. Domain names are a fundamental part of the Internet's user interface. Improving the usability of the Internet depends upon effective domain name policy. This study is intended to contribute to improvement in Internet usability for the end users of domain names. Benefits of more usable domain names include: higher sales, customer satisfaction and productivity, and reduced support costs. Domain name policies worldwide vary considerably. Consequently, end users are inconvenienced by contradictory domain name policies, diminishing the predictability of an entity's domain name, and thus decreasing usability for end users. The study objective was to develop criteria with which policy makers can evaluate their domain name policies, in order to improve the usability of domain names for end users. The main research question posed was: What are the criteria for an effective domain name policy? The research methodology included a literature review, domain name policy examination and an ethnographic narrative. The literature review found existing research examining either domain names or usability in isolation. However, research examining the intersection of the two is scarce. The existing research describes domain names as part of the web user interface. In practical terms, this is about how people use domain names to access web sites, email addresses and other Internet resources. It was established that the predictability (and thus usability) of domain names relies on effective domain name policy. The non-standardised and widely delegated process of domain name policy development leads to unpredictable and inconsistent domain names. The narrative recollection presented the researcher's inside perspective on the domain name industry, with a focus on domain name usability. The researcher provided first-hand insights into the evolution of the industry and policy development process, from Australian and international perspectives. To address the problem of poor domain name usability, a framework for domain name policy evaluation is proposed. The framework extends the current research that treats domain names as a user interface by proposing criteria which address usability concerns. The framework allows policy makers to critically assess domain name policies with end users in mind. Examples of the criteria include: understanding who are its intended and untended users, and whether it's consistent with other domain names. The framework has the potential to set an international standard for the critical evaluation of domain name policy, and become the basis for further research. This study was developed from the researcher's perspective as a participant in the domain name industry. A secondary lens regarding the usability of domain names was then applied. This study has only scraped the surface in terms of how the research fields of domain names and usability may be considered together. The research methodology for this study was primarily qualitative and interpretive. A quantitative study of domain name policies globally could provide further insight into areas including: the differences in second level country code domain names, and language implications of domain names.