In its 1992 proposal to DARPA entitled "Research on Linking Electronic Libraries," CNRI outlined the case for supporting independent development of digital library systems using an overall architecture to link them into a federated system; and as a result, several universities were funded to digitize and make available their collections of computer science technical reports (CSTRs). This project, organized and run by CNRI, came to be known as the CSTR project.
A presentation on the CNRI Digital Object Architecture was given at a meeting of the CSTR project in October 1993. As a result of certain concerns about the desirability of including or omitting semantic references in the digital object identifiers (also known as handles), two participants in the project, Robert Kahn of CNRI and Robert Wilensky of U.C. Berkeley, agreed to further investigate the relative merits of the two views. This investigation took place during the period 1993-1995 and resulted in their paper "A Framework for Distributed Digital Object Services." This paper described many of the basic concepts in the repository in a way that elaborated on the issue of semantics vs. non-semantics. The notion of a Digital Object consisting, in general, of multiple elements that each contain "type" information was introduced. The Repository Access Protocol (RAP) was described in terms of its primitive components, Deposit and Access, each of which (in turn) involved specification of the Digital Object by its handle, and a service request that identified how the specific object would be disseminated. In the 1995-1997 time frame, David Ely of CNRI and Carl Lagoze of Cornell further collaborated to implement and refine various aspects of the RAP protocol; Christophe Blanchi and Ed Overly also contributed to the CNRI repository software during that period and CNRI funded Cornell to implement a modular version of their Dienst software based on the modular approach taken by the CNRI Digital Object Architecture.
As an outgrowth of the CSTR effort and others, CNRI later enabled and funded a series of interoperability activities with several universities, one of which involved interoperability testing of repositories between CNRI and Cornell University. Specifically, in the 1998-2000 time frame, Christophe Blanchi of CNRI conducted interoperability testing of the CNRI repository software with Sandy Payette of Cornell with assistance from others at CNRI and Cornell. The software made available with this release, primarily developed by Christophe Blanchi, was specifically adapted to achieve early interoperability with Cornell. CNRI and Cornell temporarily agreed to use a common semantics for these tests, despite the fact that CNRI had been and continues to use somewhat different terminology. It remains a longer term goal of the CNRI architectural effort to achieve system interoperability without the need for agreement in advance on semantics.