[TUHS] Proper use of TUHS (was Re: Typesetter C compiler)

[Larry McVoy]

On Fri, Feb 03, 2023 at 11:39:38AM -0500, Dan Cross wrote:
> On Fri, Feb 3, 2023 at 9:16 AM Larry McVoy <lm at mcvoy.com> wrote:
> > On Fri, Feb 03, 2023 at 09:11:35AM -0500, Chet Ramey wrote:
> > > On 2/2/23 8:44 PM, Dave Horsfall wrote:
> > > >>So, the question becomes: what _is_ that forum, if such a thing exists
> > > >>at all?
> > > >
> > > >A new list?  Social media is for the birds; Usenet is dead, film at 11.
> > >
> > > A new list, if it serves its purpose of providing interesting content, will
> > > eventually undergo the same thing. It's always going end up being some
> > > variant of Yogi Berra's famous "nobody goes there anymore, it's too
> > > crowded."
> >
> > In my opinion, Warren has been threading the needle nicely.  He lets stuff
> > go into the weeds a bit but has a pretty good sense of when it is annoying
> > people that we all want to keep around.  He's very understated about it
> > all but he keeps this list pretty sane.
> >
> > If you haven't joined COFF, maybe consider it, if that got big enough
> > then the other stuff could be done over there.
> 
> Some folks have given good suggestions, which I appreciate, but
> perhaps it would help to explain exactly what I'm looking for.
> 
> I think that new system designs remain important, but without a
> historical perspective, they run the risk of repeating old mistakes,
> ignoring prior art, etc. Something I appreciate about TUHS is how one
> has access to so many of the key players in Unix as well as other
> systems: the perspective those people bring to the discussion is
> illuminating.
> 
> But TUHS is clearly meant to be a Unix history list, not a "list about
> new systems where we can ask about history because that helps us build
> those new systems". In that sense, I'm not looking for COFF or a retro
> or classic computer list, either, but for something explicitly modern
> yet informed by history.

Well, be careful listening to gray beards too much.  Much of my knowledge
comes from working with rotating disks and all of those instincts are
wrong when you can get a 1TB SSD (Samsun no less) for $104 at Best Buy.

The old guys who held sway in some of the jobs I had, fell into the trap
of "we tried that and it didn't work".  Well, yeah, trying to write AI
programs in Prolog in 1985, yeah, that didn't work, the CPUs were nowhere
near fast enough.  Today we have machine learning that seems to work,
self driving cars that sort of work, ChatGPT that literally scares me,
the old rules don't apply.

Some things will never go away, like keep your fingers off of my L1
cache lines.  I think it's mostly lost because of huge memories, but
one of the things I love about early Unix is how small everything was.
Most people don't care, but if you want to go really fast, there is no
replacement for small.

Personally, I'm fine with some amount of "list about new systems where
we can ask about history because that helps us build those new systems".
Might be just me, I love systems discussions.