I'd forgotten that. I'm sure, back in the day, I'd thought it was cute, but now I am far less sure.

Anyway, it is an interesting discussion; I'm not sure that there much of an efficiency argument for it, given that you can use the preprocessor to re-base the indexing, and the overhead in doing that isn't much and it is preprocessor overhead rather than run-time overhead. Feels wrong, though.

Multics had a smaller fraction assembly than Unix. It was written almost entirely in a higher level language than C (abstracted further from the machine behavior and less efficient).

Multics being "prior" to Unix is questionable. Multics was under development and not yet at the level of a usable OS from before the idea of Unix was even conceived until after Unix was a usable OS. So the Multics project was certainly underway long before Unix, but Unix was a usable OS before Multics was.

Both borrowed most of their key concepts from ITS, much of which was written in B. But both Multics and Unix took the idea of limiting the assembler portion of the OS much further than ITS did.

1 members found this post helpful.

As far as why the first array index is zero, it is in how the C language was defined.

I have an ancient "The C Programming Language" by K&R and right in the chapter about arrays it distinctly says "Array subscripts always start at zero in C, so the elements are ndigit[0], ndigit[1], ..."

Also as said by johnsfine, arrays are constructed to be related to pointers. This is highlighted in what used to be chapter 5, 5.3 Pointers and Arrays where they discuss [0] is related to the first pointer offset in memory, with respect to the array.

It is a convenience that the creators of the language chose, which also has a parallelism with pointer offsets and therefore is an extremely helpful, as well as natural, convenience.

As you can see, the definitions of some other languages use initial array indexes of 1 versus 0.

True. My recollection of history was much poorer than I thought.

Enumerated types count-up from zero, but even if they did not, the cost is negligible.

Actually, the "prior" is, I think, conceptually unquestionable: The Multics development at M.I.T. was a joint project of several companies and the university. AT&T was one of those companies, and Kernighan and Ritchie were, I believe, part of the Multics development team that was removed when AT&T decided to withdraw from the project. They (K&R) were then assigned to the group developing text processing tools for AT&T. (Have you ever wondered why the UNIX base tool kit is so heavy on string and text tools?) To make their life easier, they upgraded the B (or, maybe, it was BCPL) macro set to include variable types, etc., and called the resulting macro set C. Then, based on what they'd learned from the Multics project, they put together a "stripped-down" OS that they called (in an obvious pun) UNIX.

because it's an offset from a starting memory address, not an index.