The liquid architecture
Coşkun Bayrak, Chad Davis · ACM SIGSOFT Software Engineering Notes · 2003
In terms of benefiting from the potential to be gained from full distribution, today's most common implementations of distributed systems follow only limited linear versions of distribution such as client-server or n-tier models. Even many "peer to peer" systems still rely on centralized servers to provide the message passing connectivity between the peers. While these systems do provide increased robustness and computational speedup, they fail to realize the full measure of what fully distributed architectures offer.With many of the techno-historical reasons behind these linear distributed systems becoming obsolete, we should try to think of new, more truly distributed models. Fully distributed architectures should demonstrate the ability to move functionality to a completely contingent set of machines determined at or just prior to run time. Under such an architecture -- a" liquid" architecture, functionality and data are completely freed from any fixed locations or functional paths and may flow at will. Especially in the domain of shared space applications such as chat and pseudochat (i.e. instant messaging) services, and virtual collaboration, the flexibility and security to be gained from the full distribution of a liquid architecture represent not only premium benefits but possibly core requisites to their essential purpose and functionality.This paper examines the notion of a liquid architecture and explores a case study implementation of such an architecture via the Virtual Collaboration Tool introduced below.