We characterize CDAT as “open system” because the software subsystems are independent and the object-oriented nature of python allows CDAT to be “delay bound” or that the actual tool is built at run time, i.e., is not fixed. Further, the software subsystem python modules are open source and freely available for community-wide development. Thus, CDAT is easily extended and represents a different approach to the technical problem of implementing a model diagnosis infrastructure.
In this paper, we compare and contrast the CDAT approach with more traditional tools built from system-level software (e.g., C and X windows), such as GrADS and ferret, and show how CDAT complements and offers an alternative interface to data accessible by these popular tools. We also demonstrate CDAT-hosted data service applications using the popular Live Access Server (LAS) and Distributed Oceanographic Data System (DODS) systems and using the metadata search capabilities of CDMS.