V8 was the first of a series of refinements that unified things nicely, allowing programs to interact more smoothly. Nothing too dramatic, really: things like a shell that could export its environment, including functions; tweaks to how $PATH worked so we could have binaries with names like n/m1 n/m2 etc. to connect to machines m1 and m2; a push for output from programs that worked as input to the same programs (a huge deal for the shell); and so on. Lots of cleanups (db really worked, and worked well; stuff like that). Not to mention clean networking and graphics APIs that showed how easy it was to incorporate them into Unix.

What is a socket for, anyway? Why do you need them when you have file descriptors? (Rhetorical question, because the answer is, you don't. But the earliest sockets didn't even implement read and write!)

And so on.

But we did Plan 9 after v10, so it's clear we didn't think it was perfect, yet.


On Tue, Aug 27, 2019 at 12:53 PM Larry McVoy <lm@mcvoy.com> wrote:
Hey Rob,

I followed Bell Labs through the papers, the Lions doc, but I didn't get
any insight into Research after v7 or so.

Can you tell us what you liked about the later versions?

I don't want to be a total suck up but I've been a fan of your insight
ever since you said something like "if you think you need threads your
processes are too fat".  I've had long discussions with Linus about how
to make that statement 100% true (partial page table sharing across
processes, how do you make that work in general).  We didn't come to
an answer but we both agreed that processes should be as cheap as
threads and mmap is the way to share data.

On Tue, Aug 27, 2019 at 10:58:54AM +1000, Rob Pike wrote:
> I always thought Research 10th Edition was fantastic. Even the 8th edition
> was an improvement on most of its successors. But things flowed another
> way, with muddy streams mixing in.
> -rob
> On Tue, Aug 27, 2019 at 10:30 AM Larry McVoy <lm@mcvoy.com> wrote:
> > On Mon, Aug 26, 2019 at 08:19:45PM -0400, Arthur Krewat wrote:
> > > On 8/26/2019 7:56 PM, William Pechter wrote:
> > > >ZFS
> > >
> > > Here, here!
> >
> > I really don't understand the love for ZFS.  I hired Bonwick and I
> > hired Moore, I had high expectations but they were all dashed when I
> > realized ZFS doesn't use the page cache.  That's so crazy busted I lost
> > all interest in ZFS.  ZFS took us back to HP-UX mmap semantics.
> >

Larry McVoy                  lm at mcvoy.com             http://www.mcvoy.com/lm