Embracing legacy research code: Should we? If so, how?

Computer programs have been an essential part of scientific research for decades, including complex models, data pipelines and real-time experimental control software among others. As the platforms and frameworks for software development have changed dramatically during that time period, older software is often incompatible with modern systems. Preserving the thought and work that went into legacy code and make it usable today can be a difficult task. We here report on lessons we learned bringing more than 60,000 lines of C++98 code originally written for Windows 2000 onto modern systems. Drawing from these experiences we discuss the reasons why and how scientific code should be modernized.