An important task in software development and maintenance is configuration management. A software development project is not a static entity. It evolves and changes over time as new requirements are discovered, new functionality is added, platform architecture changes, new releases are delivered, and bugs are discovered and fixed. Configuration management embodies a formal methodology for managing these inevitable changes in the lifecycle of a software development project. The purpose of this paper is to present the results of a real-world implementation of configuration management for a large-scale AML software development project using the Revision Control System (RCS) found on most UNIX platforms. RCS provides functionality including storing and retrieving multiple revisions of source code files, change history, concurrency management for developers, merging revisions, control for releases and configurations, and automatic identification.
Quantitative spatial data analysis involves the exploration, visualization and modeling of spatial data. Spatial data analysis is an interactive process involving three procedures. First, the analyst looks for relationships in the data using both graphical and analytical methods. This procedure is known as Exploratory Data Analysis (EDA). Second, the analyst builds a quantitative model for the data. Third, the analyst will repeat each of these operations several times as they learn about behavior and relationships in their data. To work efficiently, the geographic data analyst must use an environment integrating data storage and retrieval, cartographic rendering, data visualization and quantitative spatial modeling. The combination of ArcInfo, S-PLUS for ArcInfo and S+SPATIALSTATS provides an integrated environment with the required tools. We demonstrate the methodology of quantitative spatial data analysis on a number of applications taken from the environmental sciences, resource management, the natural sciences and epidemiology. In each case we show how we performed the operations of EDA, modeling and model confirmation. Our examples demonstrate analysis methods for the three classes of spatial data; geostatistics or kriging methods (continuous field models), spatial correlation measures and spatial regression (lattice models) and measures of randomness or clustering for point data (point process models). We demonstrate a common and critical pitfall of using ordinary regression models on spatial data as opposed to spatial regresion models. Our examples use the powerful data visualization methods of hexagonal binning and trellis graphics. These demonstrations use the unique capabilities of our integrated environment consisting of ArcInfo, S-PLUS for ArcInfo and S+SPATIALSTATS.