[TUHS] [TUHS} communication files: interprocess IO before pipes

[TUHS] [TUHS} communication files: interprocess IO before pipes

[Doug McIlroy]

It's not really Unix history, but Dartmouth's "communication files"
have so often been cited as pipes before Unix, that you may like
to know what this fascinating facility actually was. See
http://www.cs.dartmouth.edu/~doug/DTSS/commfiles.pdf

[TUHS] [TUHS} communication files: interprocess IO before pipes

[arnold]

Doug McIlroy <doug at cs.dartmouth.edu> wrote:

> It's not really Unix history, but Dartmouth's "communication files"
> have so often been cited as pipes before Unix, that you may like
> to know what this fascinating facility actually was. See
> http://www.cs.dartmouth.edu/~doug/DTSS/commfiles.pdf

Thanks! That was a very interesting read.

As a general remark, the historical discussions here of late have been
very interesting (if voluminous :-).  I think it's increasingly important
to preserve as much of this history as possible.  I've enjoyed reading
about some of the systems I never worked on, such as ITS.

Arnold

[TUHS] [TUHS} communication files: interprocess IO before pipes

[Larry McVoy]

On Sun, Mar 05, 2017 at 10:50:22PM -0500, Doug McIlroy wrote:
> It's not really Unix history, but Dartmouth's "communication files"
> have so often been cited as pipes before Unix, that you may like
> to know what this fascinating facility actually was. See
> http://www.cs.dartmouth.edu/~doug/DTSS/commfiles.pdf

If I'm understanding correctly, an I/O on the slave side would cause
an event on the master side.  So the master was almost like a debugger
with break points at open/read/write/lseek/close?

That would explain a lot of the complexity.  And if the slave does a
read(comm, buf, 1<<20) I suspect that the master does multiple writes?
As in is there a PIPEBUF analogy?
-- 
---
Larry McVoy            	     lm at mcvoy.com             http://www.mcvoy.com/lm