> They astounded the attendees – who had made estimates of up to 100 man-years to implement the language, using up to 7 pass compilers – when they described ho

> They astounded the attendees – who had made estimates of up to 100 man-years to implement the language, using up to 7 pass compilers – when they described how they had already implemented a one-pass compiler which was in production use in engineering and scientific applications...The ALGOL 68-R compiler was initially written using the RRE ALGOL 60 compiler including extensions for address manipulation and list processing but stripped out of features such as real number handling which made it more suitable and more focused on compiling a compiler. Dropping any kind of real or float support sounds like a major issue for scientific use; did the users simply have to express everything as fixed-point integers? What else did they simply drop to get their subset/prototype working?