TI Detecting Conflicts in Multi--Agent Rule-Based Development
Environments 

LT CUCS-005-90 

OR COLUM 

YR 1990 

AU Naser S. Barghouti 

AU Gail E. Kaiser 

AV ftp cs.columbia.edu:pubreportscancelledcucs-005-90.ps.Z 

AB Researchers have been exploring the possibility of applying
production systems to the domain of software development.  One
resulting approach is rule-based development environments
(RBDEs), which provide expert assistance to developers working on
large-scale projects.  RBDEs model the development process in
terms of rules, and then enact this model by automatically firing
rules at the appropriate time to carry out chores that the
developer would have done manually. In order to realistically
model the domain of software development, RBDEs must support the
cooperation among multiple developers, each of whom selects
commands, causing the firing of multiple rules (either directly
or via chaining) that concurrently access shared information. In
this domain, rule parallelism is {\em intrinsic} and not {\em
opportunistic} (i.e., not for the purpose of speeding up the
production system).  One of the problems resulting from executing
rule chains concurrently causes is how to detect conflicts that
occur between the ains. This paper presents the multi-agent RBDE
model xplores the conflict detection problem, and suggests an
approach to 

solving it. The problem of resolving conflicts without
unnecessarily obstructing cooperation among developers is only
briefly discussed.