Re: [tlug] Peeling onions.

["Stephen J. Turnbull" <stephen@example.com>]

>>>>> "Uva" == Uva Coder <uva.coder@example.com> writes:

On Fri, 04 Mar 2005 16:16:05 +0900, Stephen J. Turnbull
<stephen@example.com> wrote:

    >> So I don't really think that which C compiler is used matters.

    Uva> Well, I've been studying the 7th to 10th edition Unix,
    Uva> Linux, Plan 9, Inferno, and V Kernel. Perhaps I'm seeing
    Uva> patterns where there are none; that is very possible.

I don't doubt that the patterns are there.  This is important for
understanding why the people do what they do.  But there's a clear
modularization between the toolchain and the kernel and the standard
library, etc.  There's no reason why the GNU toolchain can't be used
to bootstrap a new OS.

    Uva> It's amazing that the benefits of each system haven't made
    Uva> its way into a newer operating system.

Well, that's a people pattern thing.  Remember, the Bell Labs guys are
researchers.  Small, simple, understand what you're doing so that when
you try something new you can understand that too.  They have *sucked*
at getting their work to market.  Remember, Unix is older than CP/M!
(OK, OK, getting Unix to run on a Z80 would probably ybe too painful
to even think about in 1980.)

The people I hang out with eventually go back to a different
tradition, the Lisp macghines.

The Linux kernel, as you know, was a platform for a graduate student's
research and work-avoidance mode, that happened to be production
quality from the start.  So people who needed a production kernel
started using it.  I'm not sure why the BSDs didn't get off the ground
until after Linux did; I'm sure I booted 386BSD in '93 or earlier, but
I _had_ to have X for Japanese (there was no kon then).  Why Linux got
X first I can only speculate, but it had it by '94, while the BSDs
didn't work as well.
 
    >> On the other hand, while you clearly can write operating
    >> systems in C++, Modula-3, and Lisp, where are you going to get
    >> the reasonable-quality implementations except from GNU?

    Uva> From Ken (Thompson) C, it's open source.

Plan 9 has a Lisp compiler?

    Uva> Yeah, but I see GRUB and LILO as being train wrecks. I
    Uva> only see them as getting worse.

LILO was always Linux-specific.  It works for me on those rare
occasions I need to boot, and that's all I care about LILO.  GRUB,
maybe there's an issue, but I'm not sure what it would be.  Getting
the bits from the metal into silicon and starting the kernel is always
going to be a dirty task.  It's what happens after the kernel loads
and starts initializing.

    Uva> I don't see GNU as a coherent foundation where all
    Uva> applications can share resources through a system protocol
    Uva> (i.e., P9 protocol).

Ah, now that's a point.  _GNU_ is no such thing.  _GNU_ is about
supporting _individual_ hackers, and freeing them from dependency on
others' resources.  Remember, the GNU system was an OS without a
kernel (in many ways, it still is).  Stallman is a lunatic; the
precise name for the disease is "anarcho-socialism."  He believes that
if you set people free to cooperate, it will just happen.  I agree at
some level, but it won't work the wya he thinks it should.

    Uva> Why not a single protocol that extends across all
    Uva> boundaries to access data the same way, securely and
    Uva> safely across networks?

Well, let me quote Rusty Russell: "Your network is not *secure*.  The
problem of allowing rapid, convenient communication while restricting
its use to good, not evil intents is congruent to other intractable
problems such as allowing free speech while disallowing calls of
'Fire!' in a crowded theater."  Or consider the Xrender extension (at
least as implemented in X11R6.8 it's really slow, ADSL @ ~45Mbps down
is not fast enough by far).  Or the Coda caching distributed file
system, which is designed to allow Unix semantics as much as possible,
but this seems to requiring blocking in open(2) until the entire file
is in the cache.  Technically, "one protocol communicates all" is a
_very_ hard problem.

Note that for most users, pay-per-view seems to be acceptable.  Sure,
they're willing to pay money to shady engineers to circumvent the BS
encryptions, but making phone calls to the gi-in to get the laws
fixed, no.  Socially, it's not obvious that Plan 9-style sharing is
_universally_ desirable.  There will be a big role for an OS that
supports Internet Explorer-style browsing for but not a truly
networked data environment a long time to come, maybe forever.

-- 
Institute of Policy and Planning Sciences     http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba                    Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
               Ask not how you can "do" free software business;
              ask what your business can "do for" free software.