This very interesting paper must be read by every member of the Ada community. It describes the techniques used to produce distributed Ada programs for the DIADEM project. The DIADEM researchers discovered that while Ada provides support for concurrent execution, the language’s main structural units do not lend themselves to distribution. The principal problem involves the use of shared resources such as global variables, which Ada easily allows even though this causes severe performance difficulties on distributed target computers.

The DIADEM researchers solved this problem by defining a virtual node, a unit of allocation that is an abstraction of a network node. The fundamental properties of a virtual node include (1) a totally hidden internal state; (2) a lack of reference to globally shared resources; (3) well-defined communication interfaces and protocols; and (4) an independent activity, or thread of control.

A library of Ada components requires configuration flexibility to support the definition of virtual nodes in a manner that makes dynamic reconfiguration and fault tolerance possible. The communication mechanisms used by virtual nodes must not limit the range of distribution techniques employed by target systems. With this in mind the DIADEM researchers developed the concept of a remote rendezvous protocol.

Finally, the DIADEM researchers made an interesting decision to use source code transformation tools to achieve the configuration flexibility required for virtual nodes. As a result, the DIADEM virtual node capability is independent of the Ada compiler and run-time executive used to create a distributed application.

The paper itself is clearly written at a level that every Ada programmer can easily understand. It is not long, and it refers to several other authoritative sources of information. It should be at the top of the Ada community’s required reading list.