Context awareness is essential for having a rich computing experience. Current applications provide simple contexts like time, user identity to provide customized experience. But advancement in sensors and wireless technologies are giving to new class of context aware applications.

This paper is a mash-up of other projects on context aware systems, with a focus on having a infrastructure providing context aware services. There is a difference between libraries, framework, toolkits and infrastructure. Libraries, frameworks and toolkits are built on top of one another (from libraries->framework->toolkits). Infrastructure is a pervasive, public and reliable systems that support other systems.

Advantages of infrastructure approach

  1. Independent of hardware and software. Use standard data and network formats
  2. Easy to maintain. Sensors can be changed independently and dynamically
  3. Share sensors across devices. Store data in the infrastructure. Data like zip code, stock prices are too large hence can’t be stored in a single device

Challenges to context aware infrastructure

  1. Absence of standard data formats. XML/RPC is promising. Context data is inherently ambiguous. Need to represent context data probabilistilisticaly – applications can have confidence about the data before acting on it. Example provide location with higher confidence while using GPS data than while using radio trianulation. Data unavailability should always be a valid data. Data may not be available for reasons like hardware failure, privacy, etc.
  2. Designing the services is next challenge. Some services will be application specific.
    • Automatic path creation: Chaining services like UNIX pipes to provide higher level applications. To create automatic path creation, the following problems need to addressed: critical mass of operators, standard data, ability to choose between path and context query representation.
    • Proximity based discovery. Problems: getting location of the sensors, representing and storing location of the sensors.
  3. Splitting responsibilities between devices, application and infrastructure. Smart devices can make decisions locally, but infrastructure can also make smart decisions.
  4. Access control for the data. Restricting unauthorized access and handle privacy concerns.
  5. Infrastructure needs to scale to handle large number of sensors, minimal infrastructure

This papers highlights some of the known challenges in collecting and handling context data. While arguing for a infrastructure based services the authors haven’t provided any directions for the actual implementation.