On Thu, Aug 31, 2017 at 10:38 PM, Dan Cross <crossd@gmail.com> wrote:
​....​

To be fair, I think that Knuth originally wrote both TeX and METAFONT in the SAIL language for the PDP-10. He switched to Pascal (again on the PDP-10) later.

​Dan, True enough but .... having learned Pascal on the 10 (well 20 at that point) and SAIL on the 10s, the differences between them were minor in this case.   I suspect the conversion was done in a few days in emacs.   SAIL basically was the Python of its day - with a load of libraries that people used.  But it was just an extended Algol with a number of PDP-10isms.  I have a the manual if you are curious (and I even had ??have?? a working compiler for the Masscomp [68000] at some point).   

​
​...​

Anyway, I wonder if Knuth would have produced the same TeX and METAFONT had he started in Pascal; perhaps he would have, but maybe he would have given up on some of the more ambitious parts of both because the language made it much less convenient (not knowing SAIL, I'm speculating). Perhaps having gone through the exercise of producing TeX in SAIL gave him insight that inspired him to work around Pascal's expressive limitations. Or perhaps the opposite is true.
​An interesting thought, but in this case, I'm not so sure.   I think the more interesting question is if we had developed it on a 16/32 bit system instead of a 18/36 bit system.​   If I think back about moving things from the 10s (or the CDC systems) of that day, the problem was almost always an issue of 16/32 not being 'enough' when you move the code to the PDP-11 or Vax later.   It was less the pure language, and more the processor representation that caused me trouble.

Funny, I remember in those days, we obtain a really nice Pascal compiler that generated code for the 68000 that had been written on the CDC systems somewhere in Europe ??CERN maybe??.   But moving it was damned near impossible because it was assuming the word size of the CDC system.  It was written in Pascal but because it how it was written, it could not even compile itself.

We switched to Modula-II shortly there after as the C alternative, in Tek Labs.

Object Pascal, Delphi and Free Pascal (which seeks to implement the best of both) seem to have definitely fallen into that category.

        - Dan C.
​+1 Indeed - this why I still use it as a teaching tool.  I start them with a pure Pascal variant with all the bumpers and other shielding in place that a beginner should have.   If they are more interested in systems stuff, I'll let them learn C, if they want to start to do more applications, then I continue with Delphi or simple Java (PBH's subset) so they learn OO before I throw C++ at them. 

Ches' comment about using Rust or Go is possible, but I have not found a book I trust.  If PBH or Clancy would would redo one of their text, I'ld feel a lot more comfortable.

Clem   ​