|
Existing presence servers, including IBM's SIP Presence Server (ISPS), support the delivery of presence information, but only in a limited sense. Most presence applications to date provide only four basic states of awareness (active, idle, away, do-not-disturb). The next generation of applications will include rich presence, potentially delivering information such as a user's location, activity, physiological state, etc. Another limitation of the current presence server model is the inability to interact with users across peer service providers.
The increased amount of information provided/requested by such applications has implications on the network load, the user's cognitive "bandwidth" (i.e., usability), and privacy. Typically, in applications that provide awareness, such as instant messaging, every time there is a change in a user's state, this information is sent to all watchers (all clients that have the specific user in their roster). The resulting network load can be partially alleviated by having on-demand request for data, as opposed to automatic updates. However, users are accustomed to simply glancing at their buddy-lists and having up-to-date information; hence, on-demand presence data is only feasible for a subset of applications.
Besides issues of network bandwidth, it is necessary to consider the user/recipient of the rich presence (a.k.a. context) information. An application which, for example, makes use of accelerometers to determine a user's activity cannot simply send a watcher updates with acceleration on three axes every few seconds. Not only would the recipient be overwhelmed by the sheer quantity of data, but more importantly, the data in this format is too complex to be easily interpretable. It would be much more meaningful to periodically provide information such as "driving". The translation of raw data to meaningful events will both reduce the amount of traffic on the network and will enable more usable applications.
The interpretation of raw data can in some cases be done on-the-fly, e.g., indicating that a person has a fever only when their temperature is above a certain threshold. However, in other cases, it is necessary to accumulate the data and find patterns within it, for example, to determine that a user will arrive at the office in about 10 minutes, as opposed to providing his exact latitude-longitude coordinates or signal strengths to a particular base-station.
Any system that provides presence information raises privacy issues. The accumulation and storage of this data and the pattern learning required to provide meaningful events amplifies the privacy problem. Hence, there is a need for powerful and flexible privacy management, enabling a user to define which data is disseminated to whom and under what circumstances. Since people generally do not take the effort to manage and maintain profiles, in order to enhance usability, these profiles are made up of dynamic rules based on simple heuristics. The ability to automatically identify meaningful events and to provide flexible privacy management will have a strong impact on usability and adoption of rich presence services.
TAPAS (Telco Advanced Presence Applications and Services) is a framework that addresses these issues.
| 
