A decade of CGAL

The Computational Geometry Algorithms Library CGAL is the largest and most influential collection of algorithms and data structures for geometric computing that is available today. It started off in 1996 as a European research project with a small number of partners. Over the last ten years, it has grown tremendously, now being a major open source project with a lot of infrastructure (see www.cgal.org). Contributors come from the original partner sites, from sites involved in subsequent research projects, but increasingly also from many other sites worldwide. In these ten years, CGAL has fostered a new computational geometry research culture, and it has substantially shortened the time that it takes to turn academic results into industrial-strength software. For the editors of this special issue (both being with CGAL from day one), it is a great pleasure to see how the computational geometry community has widely adopted CGAL as a platform that helps in teaching, testing ideas, writing research papers, and building complex software. The papers of this special issue are a living proof for this adoption. The earned reward for the community is that people in many application domains can now successfully use the clever and ingenious techniques developed by computational geometers. This was not always the case. Back in the early nineties, the field of computational geometry was going through an adolescence crisis. Since the early eighties—when the field had come into existence as a distinguished subfield of algorithms design—many of the classical problems had been resolved, with much enthusiasm and many brilliant new ideas. This resulted in a computational geometry theory, a necessary body of fundamental results. In building this theory, its potential relevance in applications was never in doubt, but actual claims of applicability were scarcely substantiated. Computational geometry algorithms were designed for an idealized model of computation (the real RAM), and they often disregarded certain special situations as " degenerate " and irrelevant for the asymptotic results. On the one hand, only these abstractions had enabled the field to advance so quickly, but on the other hand, they now proved to be a serious obstacle toward practical implementations: computers simply did not work with infinite precision numbers, and real world problem instances were full of exactly those special situations that one never saw in testing with randomly generated synthetic data. In order not to become irrelevant in a world of vastly growing application challenges, the computational geometry …