> And that's the thing. It's half-assery at best. It's easy to design "minimalistic" languages, after all, all Turing machine models are interchangeable. Especially when you care more about ideological purity.
This is why I called this view grossly ignorant. No, it's not easy when you have the added constraint of producing a language that is practical enough to write an OS and applications with. Have you tried? How many such languages have you designed? Can we see the grammar?
Producing a small language that is Turing complete is easy, yes. And it is also entirely uninteresting if that is all you do.
Oberon was the distillation of many years, and several iterations, of work to find what worked effectively while remaining simple, and proving that it was practical to write a OS, and applications in it.
> Making good minimalistic, _practical_, and pleasant-to-use tools is an art. That's the difference between Oberon and Go.
Go might be practical, but calling it minimalistic is comical, and pleasant-to-use is highly subjective - I find it horribly unergonomic.
> Well, yeah. It's either that or be failed. So there's that. It's also why there exists pretty much no software written in Oberon outside of the ETH.
Entirely missing the point, which is that implementations existed and that including them in the official version of the language was a conscious design choice, rather than any indication it was "half-assed".
> ??? The standard Oberon-2 uses single-class inheritance, without any interfaces/protocols/... It was implemented using regular vtables.
Yes. And? I wrote what I wrote, and what I wrote is true. Nothing I wrote talked about "the standard Oberon-2". The entire point of the Oberon work at ETHZ was that it was a teaching language, and a lot of the projects involved experimental changes to the language and compiler.
To your mention of vtables, I was specifically thinking of Michael Franz' paper on Protocol Extension, that introduced a two-layer set of indirections to allow runtime extension with interfaces.
Again, the point is that not including these in the standard version of Oberon was a conscious choice.
A choice you're free to dislike, of course, but another example of why it is not "half-assed", but the culmination of many years of research by a lot of people.
Personally, I loved Franz' work on Protocol Extension to the point that my (buggy, temporarily abandoned, unfinished) prototype Ruby compiler uses his propagation method for vtable updates to allow for vtable-based dynamic method redefinition, and think it'd have been a great addition to standard Oberon, so it's not that I disagree that it could have used a more flexible way of doing OO.
But I do respect Wirth's extremely focused design, even where I disagree with it.
So I stand by what I wrote: To call it half-assed is grossly ignorant.
> Uhh... What? Package name prefix neatly sidesteps problems like "geo.Coords" vs. "screen.Coords" name clashes. It also serves as a guide to the reader of the source code. The classic packageless names are also available with dot imports.
I did not argue against package name prefixes, so this is entirely a strawman with no relevance whatsoever to my objection.