Re: In praise of Plan 9

Message ID
DKIM signature
Download raw message
So, time ago I discovered the Plan 9 operating system, and I was quite suprised,
since that until then I just getting into the history of UNIX (Research UNIX,
BSD, System V, UNIX wars, MINIX and Linux).

It made me think of the limits of UNIX, because we're still using systems that
stem from it and retain much of its "crust", although the hardware radically
changed over the years: minicomputers and dumb terminals are no more, and today
we take for granted things like graphics, networks, sound, and (to a certain
extent) parallel computing.
Before Plan 9, networks (Berkeley Sockets, Sys V STREAMS, Datakit, ecc) and
graphics (X Window System, CMU Andrew, Sun NeWS, HP Windows 9000, ecc) were
"retrofitted" on UNIX, and we see the effects: their relatively crusty C API,
the adoption of some standards over the others (even if they weren't as good as
others) because of the UNIX wars. This likely backfired, since that M$ Windows
then took over on the market: no more fancy RISC boxes with interesting
hardware, just boring x86 boxes (if you really wanted, hybrid PC-{Alpha;MIPS;
Power} machines were an option) with the same operating system, getting rid of
the portability problem.

I see why Plan 9 was made: to start fresh and try to get a system similar to
UNIX, but suitable for a world made of workstations (at the time they got
Sun, NeXT and SGI workstations and the Gnot terminal) and distributed
computing: you could retain your workstation as a "thin client", but if you
needed more power you would just replace the backend (originally they used an
SGI Power IRIS with 4 MIPS R3000 at 25MHz each, then adopted Alpha, PowerPC and
x86 CPU servers), and this was convenient for research departments.
It was mainly a research vehicle, it wouldn't have made sense complying to
ANSI/POSIX standard (and they made that clear), because standards would cut the
potential of innovation most of the time, and to explore new territories you
inhevitably end to break the standards. It also demonstrated that an operating
system could be flexible, portable yet small.

Unfortunately, due to a lack of commercialization (you could get a license with
the book from AT&T for 350$, and that was pretty expensive for a student) and
the fact that Linux was taking over, it wasn't adopted. Quite a shame, but its
legacy still lives on somehow: apart from the Plan 9 Ports, some of its ideas
made into the Linux kernel (union directories, namespaces, a port of 9P,
proc_fs, utf-8).

I remember a quote (who knows if Dennis Ritchie really said it): "UNIX has
retarded OS research by 10 years, Linux by 20". So, 50 years passed since the
first UNIX version, and I think it's really time to explore new territories,
the priority nowadays is fighting the software complexity and "rediscovering"
the hardware: a good operating system is portable and provides a uniform and
flexible programming environment, but should leverage the hardware whenever
possible. Other forgotten/orphaned systems (like Oberon, BeOS, Symbolics
Genera, AmigaOS, and many others) deserve to be rediscovered, since that old
systems might have good hidden ideas that for some reason were forgotten.

Reply to thread Export thread (mbox)