John L. Miller

Avatar movement in World of Warcraft battlegrounds

Evaluating DVE topology management and message propagation schemes requires avatar movement models. Most models are based on reasoned assumptions rather than measured data, potentially biasing evaluation. We measured player movement in World of Warcraft battlegrounds, and compared our observations against common assumptions about player avatar movement and navigation. We found that when modeling a highly interactive DVE such as a battleground, a waypoint model is not sufficient to describe most avatar movement. We were surprised to find that despite game incentives for grouping, the majority of avatar movement between objectives is individual, not grouped. Finally, we found that a hotspot-based model for avatar movement is consistent with our traces.

The near-term feasibility of P2P MMOG's

Broadly deployed distributed virtual environments (DVEs) are based upon client-server architectures. Significant research over the last decade proposes a variety of distributed topologies for message propagation to enhance scalability and performance. We ran simulations using traces from World of Warcraft (WoW) and typical broadband speeds, and found that DVEs such as WoW could not employ P2P message propagation schemes. We found pure P2P pub-sub solutions would occasionally saturate residential broadband connections, and would result in average latency more than lOx greater than client-server solutions. To our surprise, we discovered message aggregation before transmission can not only reduce bandwidth requirements, but also reduce latency in both client-server and P2P message propagation schemes.

Group movement in World of Warcraft Battlegrounds

Distributed Virtual Environment (DVE) topology management and message propagation schemes have been proposed for many years. Evaluating DVE message propagation schemes requires a variety of assumptions whose verity significantly affects results, such as details about avatar movement characteristics. We implemented two schemes for waypoint and hotspot detection, and examined their applicability for characterising avatar movement. We confirmed that waypoint detection does not yield good results for characterising human avatar movement, and gained new insight into why by rendering avatar movement as point clouds. We implemented an existing hotspot detection model, and proposed an enhancement to help overcome one limitation of cell-based hotspot detection. We were able to immediately apply this hotspot detection technique to help analyse group movement. We discovered that although a third of movement time in the battlegrounds is spent in inter-node journeys, less than a quarter of these journeys are made in groups.

The Militia and the Army in the Reign of James II

Two of the most basic functions of any government are the maintenance of order and defence against rebellion or invasion. In the England of Charles II these functions, one police, the other military, were performed by the small standing army and the militia. James II enlarged the army and so was able to use it to maintain order to a greater extent than Charles had done. At the same time he deliberately neglected the militia (except in London) and made sweeping and highly unpopular changes among the lords lieutenant and deputy lieutenants who commanded it. As a result, when William of Orange invaded late in 1688 and James tried to raise the militia, he found it both disorganized and disaffected. Many lieutenancies failed to perform the auxiliary military functions which James expected, and some sections of the militia joined, or were raised by, insurgents against the king. However, the police function of the militia, unlike the military function, did not fail; both properly-appointed lieutenants and insurgents used the forces at their disposal to maintain order, having no desire to encourage or condone violence and looting.

Probabilistic event resolution with the pairwise random protocol

Peer-to-peer distributed virtual environments (DVEs) distribute state tracking and state transitions. Many DVEs - such as online games - require ways to fairly determine the outcome of probabilistic events. While trivial when a trusted third party is involved, resolving these actions fairly between adversaries without a trusted third party is much more difficult. This paper proposes the Pairwise Random Protocol (PRP), which uses secure coin flipping to enable adversaries to fairly determine the result of a probabilistic event without a trusted third party. Three different variations of PRP are presented, and the time impact and network overhead are examined. We conclude that PRP enables DVEs to distribute the work of determining probabilistic events between adversaries without loss of security or fairness, and with acceptable overhead.