Back: History
Forward: First Configure Programs
 
FastBack: History
Up: History
FastForward: Invoking configure
Top: Autoconf, Automake, and Libtool
Contents: Table of Contents
Index: Index
About: About this document

2.1 The Diversity of Unix Systems

Of the programs discussed in this book, the first to be developed was Autoconf. Its development was determined by the history of the Unix operating system.

The first version of Unix was written by Dennis Ritchie and Ken Thompson at Bell Labs in 1969. During the 1970s, Bell Labs was not permitted to sell Unix commercially, but did distribute Unix to universities at relatively low cost. The University of California at Berkeley added their own improvements to the Unix sources; the result was known as the BSD version of Unix.

In the early 1980s, AT&T signed an agreement permitting them to sell Unix commercially. The first AT&T version of Unix was known as System III.

As the popularity of Unix increased during the 1980s, several other companies modified the Unix sources to create their own variants. Examples include SunOS from Sun Microsystems, Ultrix from Digital Equipment Corporation, and HP-UX from Hewlett Packard.

Although all of the Unix variants were fundamentally similar, there were various differences between them. They had slightly different sets of header files and slightly different lists of functions in the system libraries, as well as more significant differences in areas such as terminal handling and job control.

The emerging POSIX standards helped to eliminate some of these differences. However, in some areas POSIX introduced new features, leading to more variants. Also, different systems adopted the POSIX standard at different times, leading to further disparities.

All of these variations caused problems for programs distributed as source code. Even a function as straightforward as memcpy was not available everywhere; the BSD system library provided the similar function bcopy instead, but the order of arguments was reversed.

Program authors who wanted their programs to run on a wide variety of Unix variants had to be familiar with the detailed differences between the variants. They also had to worry about the ways in which the variants changed from one version to another, as variants on the one hand converged on the POSIX standard and on the other continued to introduce new and different features.

While it was generally possible to use #ifdef to identify particular systems and versions, it became increasingly difficult to know which versions had which features. It became clear that some more organized approach was needed to handle the differences between Unix variants.


This document was generated by Gary V. Vaughan on May, 24 2001 using texi2html