Persistant Linda: Linda + Transactions + Query Processing

Since other papers in this book describe Linda in depth, we content ourself with a brief review of the essentials. Linda is a small set of extensions to an otherwise sequential language, like C or Pascal, allowing the construction of programs that perform computations in parallel. Communication between the cooperating processes is accomplished by using a globally accessible tuple space composed of passive and active tuples. A passive tuple is merely a sequence of values of various types similar to a Pascal record or C structure. Each value and type is call a eld of the tuple and the acceptable data types include basic types (integer, oat, etc.) as well as arrays and records of data types. Active tuples are also sequences like passive tuples, but include unevaluated function calls with a statically typed result in addition to values. An active tuple cannot be accessed until it becomes passive, when all functions in tuples elds have terminated and resulted in a value. Tuple space is heterogeneous; two tuples in the space can have di erent types and numbers of elds. Tuple space can contain multiple copies of a tuple, unlike mathematical sets or database relations. Linda allows the programmer to access tuple space using four basic operations: two that place new tuples into tuple space and two that retrieve passive tuples from tuple space.