This paper describes the eyesweb open source system and recent improvements. The paper makes a few general points in addition to describing specific eyesweb issues.


General points:


1.  Working with sensing data requires that ability to work with different sensing types, but more importantly, the ability to work with different abstraction levels.


2.  Additions to eyesweb:


a.  Subpatching

b.  Synchornization support. (audio & video). Use of a single sync clock by default (multi clock operation is supported).

c.  multi-processor support.

d.  separate user interface from kernel.  Kernel doesn't rely on the UI, but the UI relies on the kernel.  Addition of a command line interface.

e.  interoperability between eyesweb and other programs.          

f.  addition of device drivers.

g.  catalog of available functions.

h.  plugins.

i.   addition of "kernel" objects that have access to privileged actions.

k.  pins (same as max send-receive).

l.  collections.


3.  UI improvements:

a.  patch zooming

b.  catalog view in addition to tree view.


This system follows a similar paradigm to eyes and Aria by creating a tool for processing sensed data using a "bag of tricks" kind of mechanism.  Theses systems contain nodes that can be pipelined in a data flow like way (as in Max/MSP).  Each node consists of an operation that is performed on the sensor data to refine it into useable information or knowledge. 


All of these systems are highly dependent upon the context of the sensing environment, and because of this, require a some knowledge about how computer systems can use assumptions about the environment to compute/extract knowledge about what is happening in that environment.


This brings up the issue of how to create flexibility with theses systems and at the same time allow for intuitive understandings, simplicity, and ease of use.  The BigEye systsem provided a more simple system but with a limited (but useful) range of capabilities.  The system allowed the user to create pipelines of sensor activity and to explore through experimentation different sequences and thresholds for extracting data.  This applroach of providing a fixed pipeline and threshold management allows for exploration and accidental discovery.   The keys to setting up an exploratory process is for the objects being connected together to be clear about their function and relationship to other objects.