This document is an introduction to the DODS data delivery system. It presents an architectural view of the system and provides rationale for some of the important decisions that lead up to this design. This document is not a detailed description of the design; that can be found in DODS--Data Delivery Design and related documents.
The data delivery system uses the client-server paradigm to provide access to remote data sets. On each computer which hosts a data set, a data server must be installed. Client programs which communicate with these servers can be built either from scratch or from existing code using reimplementations of commonly used data access Application Programmer Interfaces (API)1. The API libraries are reimplemented so that existing code can be re-linked without modification. Thus the data delivery system provides a pathway for existing code to migrate from local access to remote access without requiring modification of that code.
DODS addresses access to on-line data only. It would be nice to provide access to data that are not already on-line, but the problems associated with that are outside the scope of DODS (e.g., some data collections are not suited to on-line access because of their size or content). Thus, the data delivery architecture addresses the problem of network access to data sets that are already on-line.
Before describing the data delivery system, some important computer networking and data management terms and concepts are discussed. Each of these is first defined in general terms and then particular aspects of those ideas that are important to DODS are discussed. These include networking technologies (remote procedure calls and the hypertext transfer protocol) and data access tools and concepts (application programmer interface, data model and data format).