With V5/V6 C I/O was a lot like what Larry was describing for getopts(3) it
was all over over the map.  There was the portable I/O library which I sort
of think of as the prequel to studio but I don’t remember uSing it much.  I
must have run into most of the different ways people did I/O in some
program(s) but I don’t remember any one off hand.

I think the thing to remember is that at the time system programming
languages such as C and Bliss were noted for not having I/O built into the
language- it was supported externally.  DEC (CMU) with Bliss had rich set
of libraries (often in assembler already available) and force/matched by
them with their users.  Unix and C grew up independently which I think is
part of why it was a tad more random.   BY the time dmr adds stdio, it was
still early enough in the life to displace the randomness for something as
important as I/O, whereas lack of use of something.like getopt would not
become clearly deficient until after widespread success.

On Mon, Jul 9, 2018 at 12:37 PM Random832 <random832@fastmail.com> wrote:

> On Sun, Jul 1, 2018, at 07:34, ron@ronnatalie.com wrote:
> > The bigger issue with the early printf is it just called putchar and
> > putchar only output to stdout or what ever the global fout variable was
> > set to.
> > There was a comment in the manual that the fout concept was kludgy.
>
> V6 'iolib' printf has an interesting approach to fixing this:
>
> If the first argument was 0 through 9, it was taken to be a file
> descriptor, and the second argument was the format string. If it was -1,
> the second argument was the output string (as for later sprintf), and the
> third was the format string. Otherwise, the first argument was the format
> string.
>
> (I'm curious as to how much "iolib" was actually used, since it doesn't
> appear to have been included by default - there was a different printf
> routine in libc)
>
-- 
Sent from a handheld expect more typos than usual