From the third ACM conference on History of programming languages: https://dl.acm.org/doi/proceedings/10.1145/1238844
So nice to see Modula-2 get some love! I really enjoyed that language and rather miss working in it.
I see there are still some old implementations floating around, I should grab one and refresh my memory.
I’m a bit surprised that Wirth categorizes variant records as “loopholes”. I guess it’s because in his languages, it’s possible to access fields from the wrong variant? That seems to me like a flaw of the language, not the construct, since sum types have existed in ML since the 80ies and are perfectly safe: you need to destructure/match before accessing the corresponding fields. So, that leaves me a bit puzzled.