We envision a time when an Earth scientist from any institution will be able to access high-quality data regardless of where it may actually be archived. The scientist will not have to know the data's actual location nor its format. He or she will only have to specify the type of data, and the workstation software will handle finding it and gaining access. Here we describe a method to achieve this vision that is now in use at Lamont and several other institutions. Institutions make “Views” of their databases publicly available to Internet users, employing database-serving software that runs on one of their computers. This software completely automates the process of finding out what kind of data are available and retrieving them. Hence a wide variety of different databases become, from the scientist's perspective, parts of an Internet-wide Earth science database. At the same time, the institution that archives the data is not locked into any particular database management system, so it is free to provide alternative access methods and to exclude some of its data from the system. The self-teaching nature of this concept is especially useful in large cooperative scientific efforts.