The Context Fabric: An Infrastructure for Context-Aware Computing Jason I. Hong Group for User Interface Research, Computer Science Division University of California, Berkeley Seungseok Kang

Introduction Context –The circumstance in which an event occurs The goals for context-aware computing –Increasing the number of input channels into computers –Pushing towards more implicit acquisition of data –Creating better models that can take advantage of this increased input –Using the increased input and improved models in new and useful ways It is still extremely difficult to build context-aware applications

Context Fabric Context-aware infrastructure –Distribution, modeling, privacy, robustness Design keys –A flexible and distributed data store to make it easy to model, store, and disseminate context data –A context specification language for declaratively stating and processing context needed –Reasonable and customizable privacy mechanisms to help protect context data about end-users

Context Data Store Logical context data model –A way of representing entities People, places, things –Entities, attributes, relationships, aggregates Physical data store –Where the context data is actually stored –Distribute the data Context can exist in multiple places Advantage –Decouples context acquisition from context modeling and context usage –Multiple applications can request and use the context data

Context Specification Language CSL –Declarative way of stating context needs at a high level –Providing a clean programming abstraction to the context data SQL does for relational databases Context service processes CSL statements –Queries: “How many people are in the room right now?”

Privacy Privacy issues –The most debated issue with respect to ubiquitous computing –Finding the right balance The needs of individuals and the The needs of governments and societies Privacy in Context Fabric –Currently planning on implementing privacy mechanisms directly into infrastructure Restricting context queries Garbage collection Intentional ambiguous answers

Evaluation Five dimensions for evaluations –To see if the data model is expressive –To learn if the C 니 is powerful –To ensure that the overall system is robust –To discover if we have enough useful mechanisms for privacy –To find out if the infrastructure makes it easier to develop context-aware applications Method for evaluations –Iterative design process Working out designs, implementing, building Refining before going to the next iteration

Related Work Schilit’s ParcTab system –First context-aware system infrastructure Interactive Workspaces EventHeap –Connecting devices in a local room Context Toolkit –“operating systems” approach –Hardware is primary, data formats and modeling is secondary