A survey of background literature on Knowledge Theory, a model for state in distributed systems that matches intuition well. This survey discusses common knowledge and variants, along with application to Byzantine Agreement, Atomic Commitment, Routing, and Secrecy (i.e., multi-agent security).

Conventional approaches for delivering information to mobile users are unavailable in the absence of a ubiquitous wireless network. The design of a prototype of a software architecture for solving this problem, 7ds, is described. 7ds running on mobile devices form an ad-hoc network of 7ds peers; these peer systems exchange data to satisfy user requests.

7ds exploits spatial locality of information among mobile users -- i.e., the clustering within physical space of users with related information needs. This thesis presents a study of the prevalence of spatial locality among web requests made by users of the wireless network on the campus of a major United States university.

An environment is characterized by spatial locality of queries and information when it is likely that users in close geographic proximity query for similar data. Information exhibits spatial locality when it is coupled to a real-world place. For example, play reviews are most relevant in a theater; and users in a Dental school may be particularly interested in web sites on related subjects. The prevalence of such spatial locality is related directly to the feasibility of deploying location- dependent services. Intuition suggests that a high degree of spatial locality of information exists because people often gather to exchange information. As such, we expect that appropriate location-dependent services may harness the spatial locality ectively.

The web is not primarily a location-dependent service, but it provides a ready testbed to study the prevalence of spatial locality and mobile users. This paper results from a three-week study of spatial locality phenomena among mobile web users on a major university campus using the 802.11 wireless infrastructure. We show that users are often near other users with similar interests. In addition, we categorize the urls and present a classifcation of the wireless information as a function of the location from which it was accessed. We also model the associations of wireless users to access points. Finally, we discuss the implications on the feasibility of location-dependent services and potential improvements of the wireless access using caching mechanisms.

This paper discusses the idea of the `file' that is so prevalent and common. Specifically, I look at a series of operating systems from Bell Labs, and talk about how the idea seems to have started and blossomed to be a thing of great generality in their work. Multics, Plan 9, and Inferno are discussed.

Some distributed applications function well only when guarantees are made on the maximum latency which the applications' packets will experience. Common packet-switched networks, such as the current public Internet, do not provide straightforward mechanisms for deriving such guarantees. Two approaches to the problem of bounding packet-switched network latency are presented: PWP-Sporadic, and PWP-Rate, each of which corresponds to a well-known model of real-time tasks. Feasibility relations are presented, and the relative benefits of each are discussed. Finally, critical issues in the deployment of PWP-Rate are explored.

The JUnit testing tool is widely used to support the central XP concept of "test first" software development. While JUnit provides Java classes for expressing test cases and test suites, it does not provide or proscribe per se any guidelines for deciding what test cases are good ones for any particular class. We have developed a method for systematically creating complete and consistent test classes for JUnit. Called JAX (for Junit Axioms), the method is based on Guttag's algebraic specification of abstract data types. We demonstrate an informal use of ADT semantics for guiding JUnit test method generation; the programmer uses no formal notation other than Java. Experiments show that informal JAX-based testing finds more errors than an ad hoc form of JUnit testing. We also discuss how JUnit test method generation can be automated for developers who don't mind using partial or full formal specs.

The design of a system for highly-available client/server applications is presented, together with an example application for scheduling video services.