Open-ended activities such as CAD/CAM, VLSI layout and software development require consistent concurrent access and fault tolerance associated with database transactions, but their uncertain duration, uncertain developments during execution and long interactions with other concurrent activities break traditional transaction atomicity boundaries. We propose splittransaction as a new database operation that solves the above problems by permitting transactions to commit data that will not change. Thus an open-ended activity can release the committed data and serialize interactions with other concurrent activities through the committed data.

/pdf/split-transactions-for-open-ended-activities-kk7dimnfy7.pdf

Split-Transactions for Open-Ended Activities

A method for controlling versions of software and other hierarchically structured entities is presented. Using the variant structure principle, a particular version of an entire system is formed by combining the most relevant existing versions of the various components of the system. An algebraic version language that allows histories (numbered series), subversions (or variants), and joins is described. It is shown that the join operation is simply the lattice least upper bound and together with the variant structure principle, provides a systematic framework for recombining divergent variants. The utility of this approach is demonstrated using LEMUR, a programming environment for modular C programs, which was developed using itself. The ways in which this notion of versions is related to the possible world semantics of intensional logic are discussed. >

http://www.cse.unsw.edu.au/~plaice/archive/JAP/P-TOSE93-lemur.pdf

A new approach to version control

Artificial intelligence is a growing branch of computer science that studies ways of enabling computers to do tasks that seem to require human intelligence. These tasks include game playing, expert problem solving, natural language understanding, and theorem proving. Many existing AI programs perform such tasks with varying degrees of success. Whereas a program that can understand natural language is still a dream, many chess-playing programs can beat expert human players. More successful AI programs include knowledge-based expert systems that are being applied to a wide spectrum of real-life problems from airline catering to oil exploration. These systems can acquire knowledge about a domain from a human expert and then use it to solve routine problems in that area. For example, expert systems can perform almost as well as human experts in the diagnosis of infectious diseases (MYCIN), finding the structure of chemical compounds (DENDRAL), and performing mathematical symbol manipulations (MACSYMA). These AI systems offer new capabilities to tackle several problems difficult to solve using the conventional algorithmic approach.

Software Development Support for AI Programs

Jasmine is an object-oriented system for programming-in-the-large. Jasmine describes software using system model objects. These objects are persistent (they have lifetimes of days or decades) and immutable (since system models act as historical records). This paper describes JStore, a distributed, replicated repository for system model objects. JStore provides robust, transactional, write-once storage.Designs are presented for the serialization, location, and replication of objects. Description procedures serialize objects for network transmission and permanent storage. An expanding ring multicast search algorithm locates saved objects. JStore replicates objects using a lazy replication algorithm built on top of the location mechanism. Decision procedures determine the replication policy used at each storage site.

A distributed repository for immutable persistent objects

Programs products are quite often families of large and modular programs. Modern programming languages support the formulation of such program families only partially. At the time being it is usually not possible to describe different revisions, variants, and versions of single program building blocks and whole programs. This paper presents a proposal for the formulation of such version information as part of the program text. In a newly introduced CONFIG part of a program building block the programmer can express: (1) to which versions the building block belongs, and (2) which versions of other building blocks it uses. In this new construct a version is defined as a pair (revision, variant), and a variant as a vector of attribute values. With these language constructs the “knowledge” about the program versions can be expressed by facts and rules. The representation of this knowledge is adapted to the structure of the program, and the generation of specific program versions can be done automatically. The configuration algorithm has been implemented in Prolog, and this prototype implementation has been used for some examples.

A. N. Habermann

Papers

Automatic deletion of obsolete information