I read the article you linked. It's too bad they didn't decide to pass a fat pointer with the array dimensions.

But when the article implies that null terminated strings and the problems associated with them are caused by arrays decaying to pointers, I get more skeptical.

I mean it's not as if arrays in C are bounds checked anyway, so how would that have helped implement strings any better?

I guess you could have bounds checked them in the stdlib code, but you certainly wouldn't have gotten it for free just by knowing the array dimensions.

> I mean it's not as if arrays in C are bounds checked anyway, so how would that have helped implement strings any better?

I've written an awful lot of C code, and I have much experience with the D way which uses arrays instead of null termination.

When I review a section of code consisting of strlen/strcat/strcpy/etc. I routinely find bugs in it, always centered around a mistake with the 0 termination.

When the array bounds are available, the compiler can optionally insert array bounds checks.

As for getting it for free, avoiding doing the strlen()'s is a big time saver.

Arrays are for FORTRAN is I think the reason.