Open Modeling in Multi-Stakeholder Distributed Systems
<-- ^
Robert J. Hall
AT&T Labs Research
180 Park Ave, Bldg 103
Florham Park, NJ 07932
http://www.research.att.com/~hall/
Abstract
Multi-stakeholder distributed systems (MSDSs), wherein the constituent
nodes are owned or operated by distinct stakeholders having limited
knowledge and possibly conflicting goals, challenge our traditional
conception of requirements engineering. MSDSs, such as the Internet
email system, networks of web services, and the Internet as a whole,
satisfy no globally consistent high-level requirements and, therefore,
have behavior which is impossible to validate according to the usual
meaning of the term. We can sidestep this issue by changing the
problem from "does the system do the right thing" to "will the system
do the right thing for me (now)?" To solve that simpler problem,
however, we need a way to predict behavior of the system on inputs of
interest to us. The OpenModel idea proposes to solve this by
establishing open standards for behavioral modeling: each node will
provide via http (or through a central registry) a behavioral model
expressed in terms of shared domain-specific function/object
theories. A tool will support validation by assembling these models
and simulating, animating, or formally analyzing the assembled model,
helping the user to detect unfavorable behaviors or feature
interactions in advance of actually using the real system. While
there are clearly scenarios in which the OpenModel approach can't
succeed, I believe there are many clear cases (such as Internet email,
component reuse, and Enterprise Application Integration) in which it can.
This talk is intended to float the OpenModel proposal and discuss the
issues (advantages, challenges, limitations) involved. Please provide
feedback.