Subj : Re: My Retro Computing
To   : Oli
From : tenser
Date : Fri Aug 13 2021 01:56:33

On 12 Aug 2021 at 09:01a, Oli pondered and said...
 
 Ol> That's true. Linux is still very old-school (apart from systemd and
 Ol> wayland and sound servers and ZFS (which they don't like)).

It's a matter of focus.  Linux is basically a clone of Unix, and
has stayed true to those roots.  As such, they haven't pushed the
envelope in the same ways that plan9 did.

The flip side was that plan9 was a research system, and never
interested in filling the niche that Linux occupies.  I got into
this tytso a few months ago when he asserted that plan9 "failed"
as an open source project: that was a strawman, because plan9
wasn't trying to _succeed_ as an open source project.  But they
wrote a lot of cool papers and many of the ideas have made it
into mainstream systems (namespaces in Linux, albeit in a
diminished form): in that sense, it was a wild success.

 Ol> Using Plan 9 is kind of retrofuturistic ...

I gave a demonstration of plan9 a few years ago to my team at
Google, as we'd taken a lot of the namespace code into Akaros.
A senior Linux engineer was there (the author of epoll).  His
only comment was that the UI looked like it was from 1991.

Sigh.

 Ol>  t> I wrote about this a few years ago here:
 Ol>  t> http://pub.gajendra.net/2016/05/plan9part1
 Ol> 
 Ol> Nice. Best introduction to Plan 9 I've ever read.

Thank you!

 Ol>  t> There's not a lot to do with it unless you're a C programmer.
 Ol> 
 Ol> Have you tried Go on Plan 9?

Yes, but it's not a great fit.  Plan 9 doesn't have a lot of
tools for a process to manipulate it's own virtual address
space (no `mmap`), and Go really wants to be able to manipulate
the virtual address space of processes running Go programs.

I'm good friends with many of the Go developers, and they keep
making noises about removing support.  This is one of the things
we want to address in Harvey.

 Ol> Until recently I thought Plan 9 development has stopped. Then there was 
 Ol> Inferno. And now we have Harvey OS. It's kind of confusing for anyone
 Ol> who is interested in Plan 9, but hasn't followed the development over
 Ol> the years (decades). But I guess this is part of retro computing.

It basically did stop.  9front has been carrying the torch,
but there are big personality conflicts in the community.

Inferno was a separate thing: it was an attempt to create a
system commercializing the ideas in plan9; Lucent created a
whole business unit to support it, and it nominally competed
with Java for e.g., set-top boxes and the like, but it didn't
succeed in the market.

Harvey is trying to stretch plan9 in new directions: a major
goal is to replace large chunks of the userspace with Go
programs, and I'll probably start integrating Rust into the
kernel.  I had done a major overhaul of the MMU support and
the way we lay out the kernel in virtual memory, but we just
rebased back to 4th Ed + 9legacy and I haven't ported the
changes over yet; I probably have to modify the compilers to
make it all work. Ugh.

Right now, there's work happening in private trees, but the
p9f is debating how to get that out in the open.  For
someone wanting to use plan9, the best way forward is to
run Richard's 9pi distribution on a Raspberry Pi.  If you want
to install on x86, either 9legacy or 9front.  9front is
probably more reliable, but the community is kinda hostile.

--- Mystic BBS v1.12 A46 2020/08/26 (Linux/64)
 * Origin: Agency BBS | Dunedin, New Zealand | agency.bbs.nz (21:1/101)

.