[9fans] So What is P9 good for.....

[9fans] So What is P9 good for.....

[John Stalker]

 From Jeffrey Haun <ziplok@bellsouth.net>:
>I ask this in all seriousness...After visiting the main page at bell
>labs, and reading from the links there, I'm sorta missing the point of
>Plan 9. The interface is nice. But applications are sparce. What is
>the draw of thie OS. I tried to install it. But the hardware I have is
>not supported. So why go to the trouble. Why does the bunny not die???
> Why do YOU use Plan 9 and not Linux/FreeBSD?

The following is just my personal opinion.  I don't really use plan9 to do
productive work.  I use primarily FreeBSD at home and Linux at work,
though I have computers with NetBSD, WinME, and plan9 installed on
them.  The main reason I installed plan9 was to see how it works.  It
is a small and well designed OS.  Those are hard to find.  Early linux
is small, but not well designed.  If you want to write software it is
a relatively easy platform to write for.  There are two reasons I don't
use it seriously and probably never will.

	Hardware: Yes, it is not easy to get plan9 to work with your
	hardware.  A small user and developer means few drivers.  You
	can buy hardware specifically for plan9, and I have bought a
	graphics card specifically for its plan9 support, but hardware
	is definitely a problem.  For this reason I have plan9 only
	on one machine, even though its strength is distributed computing.

	Software: Some applications are very nice.  Acme is a good example.
	There are very few applications though.  Some important categories,
	e.g. graphical web browsing, aren't really covered.  I don't know
	of a single plan9 game.  No, fortune is not a game.

I would say running plan9 is a good idea if you want to learn about
systems programming, do a bit yourself, and maybe borrow some ideas.
If you want a nice stable system with good hardware and software
support then you are much better off with FreeBSD.

The other reason to use plan9 is glenda, the bunny.  She is pretty cute.
--
John Stalker
Department of Mathematics
Princeton University
(609)258-6469

Re: [9fans] So What is P9 good for.....

[Ronald G. Minnich]

- in the long run, an OS which uses integer IDs is totally useless
  for distributed computing. I learned that one the hard way
  11 years ago with a large enough Condor flock. Globus should have
  learned it the hard way by now but for some reason they have not.
- "you can't secure a system with a privileged user" - Spaf
  How many other OSes avoid having root/administrator/you-name-it
  accounts. Bad deal. Plan 9 avoids this mess.
- device numbers are a Bad Thing, but the various Unix derivatives
  still are having trouble doing a devfs that works well. No such
  issues in Plan 9.
- It's nice to be able to (as a user) mount and unmount things, esp.
  CDROMs etc. Doing it without setuid-root kludgery is even nicer
- Graphics that you can program without a 2000-page stack of packs.
  Now there's a concept.

That's just a few things.

It all boils down to Unix being past the point of smelling bad, now it's a
set of decaying bones with no smell. We need to figure out what comes
next. Arguably, that should have been Plan 9 ca. 1991, but that's another
story. It's hard to believe that in 10 years we'll all still be running
Linux or BSD, so it makes sense to try to see what we SHOULD be running.

ron

Re: [9fans] So What is P9 good for.....

[Boyd Roberts]

Ronald G. Minnich wrote:

>It's hard to believe that in 10 years we'll all still be running
>Linux or BSD, ...
>
It's hard enough to believe we're running them _now_ ...

Re: [9fans] So What is P9 good for.....

[Phil White]

On Thu, Feb 13, 2003 at 04:43:40PM +0100, Boyd Roberts wrote:
>
> It's hard enough to believe we're running them _now_ ...
>

You sound like a COBOL programmer around the turn of the century ; )

-Phil/CERisE

Re: [9fans] So What is P9 good for.....

[Jim Choate]

On Thu, 13 Feb 2003, Phil White wrote:

> On Thu, Feb 13, 2003 at 04:43:40PM +0100, Boyd Roberts wrote:
> >
> > It's hard enough to believe we're running them _now_ ...
> >
>
> You sound like a COBOL programmer around the turn of the century ; )

No, he sounds like a modern man being made to use COBOL tools from the
turn of the century.

I fully agree with his point. Commercial and social 'inertia' are -bad-.


 --
    ____________________________________________________________________

      We are all interested in the future for that is where you and I
      are going to spend the rest of our lives.

                              Criswell, "Plan 9 from Outer Space"

      ravage@ssz.com                            jchoate@open-forge.org
      www.ssz.com                               www.open-forge.org
    --------------------------------------------------------------------

Re: [9fans] So What is P9 good for.....

[Dan Cross]

> http://open-forge.org

For a good time, call....

> And only geeks care, and the VAST majority of users are -NOT- geeks.

That's okay, Plan 9 isn't designed for end users.  Neither was Unix.

> I fully agree with his point. Commercial and social 'inertia' are -bad-.

This coming from a guy who seems to have never had an original thought
in his life.

How about this unoriginal cliche: ``put up or shut up.''  I have yet
to see anything come out of, ``Hangar 18'' or ``Open Forge''.  Hey,
maybe you could rip off the IdeaLab! logo: OpenForge!  You're just a
snazzy dot-com office and a zucchini shack away from Chapter 11!

For crying out loud, this guy has had his say, but hasn't produced
a single thing.  Can't we boot him off the mailing list now?  Comic
relief aside, his inanities are stale.  Even Bushnell was better; at
least he had technical knowledge.

	- Dan C.

Re: [9fans] So What is P9 good for.....

[Phil White]

On Fri, Feb 14, 2003 at 02:45:17PM -0500, Dan Cross wrote:
> > http://open-forge.org
>
> For a good time, call....

I'll double the silent part of that.  Nothing from Opera or Lynx under
Linux.

> > I fully agree with his point. Commercial and social 'inertia' are -bad-.

Jim: So much for building to last?  The COBOL bit was in reference to
old COBOL code still in use in 2000 which 'needed' to be updated lest
the sky fall.

> For crying out loud, this guy has had his say, but hasn't produced
> a single thing.  Can't we boot him off the mailing list now?  Comic
> relief aside, his inanities are stale.  Even Bushnell was better; at
> least he had technical knowledge.

Dan:  I don't know what your history is with Jim, suffice it to say
longer than I've been on the list (6 months perhaps?).  In any case,
the last I checked, he is apparently a fan of plan 9.  So the bit
about booting him from the list is kinda self-defeating.

Besides, it's nice to hear that another plan 9 network apart from the
one at the Armory exists.

-Phil/CERisE

Re: [9fans] So What is P9 good for.....

[matt]

My favourite features of the moment in Plan 9 are the plumber and acme

In a few regular expressions I'd built a class browser for PHP

[PHP classes use the -> notation]

If I plumb [right click] something like    $foo->bar($p1, $p2, $p3);

I've got a plumbing rule that executes "grep '^( |\t)?function( |\t)+bar(
|\t)?\(  " on my PHP code [on a seperate FreeBSD machine no less !]  and
lists the matching files

/n/freebsd/phplib/baz.inc:24 function bar($param1, $param2, $param3) {
/n/freebsd/phplib/baz_new.inc:28 function bar  ($param1, $param2, $param3) {

such that clicking on one of the matches will open the source code file for
that definition.

And all that took me about 15 minutes to get working [+ the time it took me
to learn plumbing rules of course - about an hour of fiddling]


And to think that the code I wrote to accomplish this is so trivial it
wasn't even worth sharing it with the world!


Matt

Re: [9fans] So What is P9 good for.....

[Douglas A. Gwyn]

Ronald G. Minnich wrote:
> - "you can't secure a system with a privileged user" - Spaf

Actually you can, but only if the *only* use of privilege
on the system is to implement a suitable security policy.
The use of set-UID-0 *applications* on Unix was extremely
short-sighted.

Re: [9fans] So What is P9 good for.....

[Ronald G. Minnich]

On Fri, 14 Feb 2003, Douglas A. Gwyn wrote:

> Actually you can, but only if the *only* use of privilege
> on the system is to implement a suitable security policy.

well you're an old Unix buzzard so I'll let you and Spaf argue that one.
Security makes me dizzy.

> The use of set-UID-0 *applications* on Unix was extremely
> short-sighted.

Hey, it was worth a patent. And at the time it certainly beat the options
(which on the systems I user were: none)

ron

Re: [9fans] So What is P9 good for.....

[Douglas A. Gwyn]

Ronald G. Minnich wrote:
> On Fri, 14 Feb 2003, Douglas A. Gwyn wrote:
>>The use of set-UID-0 *applications* on Unix was extremely
>>short-sighted.
> Hey, it was worth a patent. ...

Perhaps you missed the point I was making.  The
capability of executing a process with enhanced
privilege was fine, but should have been used
only to implement an access control layer or
service, not to elevate every operation in a
high-level application to superuser privilege.
At BRL we spent many man-months fixing security
holes in Research Unix but even more for BSD,
where evidently the quickest implementation was
usually the one chosen, without much regard for
security ramifications.  That would have been
adequate for a single trusted error-free user,
but not in a networked timesharing environment.
CERT still receives security problem reports for
bind, sendmail, etc., and many of them can be
directly attributed to a set-UID process having
at some point during execution more privilege
than it needs to perform its intended function.

It's experiences like that that make me a big
fan of capability-based systems architecture.

Re: [9fans] So What is P9 good for.....

[Geoff Collyer]

> CERT still receives security problem reports for bind, sendmail, etc.,

As I commented to friends recently,

> I've just been looking at the security bugs fixed in the pre-release
> of freebsd 5.0 and it's dozens (maybe hundreds) of nasty bugs.  Worse,
> it's the usual suspects: buffer overflows, bugs in BIND and the DNS
> resolver, bugs in sendmail (inconceivable!).

It's twenty years on, and lunix distributions still include sendmail
and BIND, despite their size, slowness, complexity, buginess, painful
configuration files and long history of security bugs, usually due to
buffer overruns.  There are certainly plausible alternatives to
sendmail available.  I'm not sure about BIND, but I'd rather port dns
and whatever machinery it needs to Unix than deal with BIND again.
The stupidity of permitting (nay, encouraging) forward and reverse
maps to differ is bad enough, but it nibbles at you, with
manually-maintained zone serial numbers and trailings dots and just
endless crap.  (Of course, the DNS itself won't win any beauty
contests, but that's a harder problem.)

The first security measures I take when installing lunix are:

	- find any sendmail processes and kill them all.
	- find all sendmail binaries, set permissions to 0 and then remove them.
	- find any sendmail queue directories or configuration files and remove them.

What's wrong with these lunix people, are they stupid or something? ☺

Re: [9fans] So What is P9 good for.....

[Lucio De Re]

On Mon, Feb 17, 2003 at 03:32:49AM -0800, Geoff Collyer wrote:
>
> It's twenty years on, and lunix distributions still include sendmail
> and BIND, despite their size, slowness, complexity, buginess, painful
> configuration files and long history of security bugs, usually due to
> buffer overruns.  There are certainly plausible alternatives to

... and bash.  What are you suggesting?  I run sendmail by choice,
even direct requests to move on to qmail or postfix I meet with a
"as long as someone else does the administration and maintenance".
Do I really need to move out of my comfort zone?

Well, I _am_ here to learn, but I don't have the confidence to move
away yet.  Incidental, perhaps, but I'm a lot more fearful of MS
Exchange.  I seldom look at Sendmail sources, but there's definitely
some value in their availability.  I suspect that applies to others in
the lunix fold.

> sendmail available.  I'm not sure about BIND, but I'd rather port dns
> and whatever machinery it needs to Unix than deal with BIND again.
> The stupidity of permitting (nay, encouraging) forward and reverse
> maps to differ is bad enough, but it nibbles at you, with
> manually-maintained zone serial numbers and trailings dots and just
> endless crap.  (Of course, the DNS itself won't win any beauty
> contests, but that's a harder problem.)
>
That's DNS, not BIND.  If the RFC demanded one-to-one relationships
between IPs and names, BIND would have been different indeed.  Or
are you suggesting that BIND's implementation of zones predates
the RFCs?

> What's wrong with these lunix people, are they stupid or something? ?

I guess so.  My take is that the _real_ choices are not as obvious
as they may be to an outsider.  For one, legacy _does_ exist and
has to be dealt with.  I have NetBSD installations, all of them to
be subsumed into a single server "real soon now" varying from 1.1
through 1.2.1, 1.3, 1.4.2, 1.5.2 to 1.6.  Upgrading is a real pain.
And Plan 9 has no upgrade path that is superior where it really
matters, namely in the configuration files.

Now imagine if I had to roll my own mail exchanger, DNS server and
other useful tools (web server, proxy, you name it).

I do note that on the NetBSD mailing list the sendmail vs postscript
debate is veering to taking both of those off the distribution,
replacing them with mini-sendmail (by Jef Poskanzer - if memory
serves and my spelling neurons are firing right).  So there is an
awareness that bigger is not better and that alternatives must be
sought.  But I suspect it is a groundswell movement: it needs the
background noise to increase, rather than have some shrill complaints
in the foreground.

I call it democracy, but of course in a techno forum that is usually
taken to mean sheepocracy.  The result is the same :-)

++L

Re: [9fans] So What is P9 good for.....

[Russ Cox]

> I do note that on the NetBSD mailing list the sendmail vs postscript
> debate is veering to taking both of those off the distribution,

postscript, hands down, though pdf is pretty good too.

Re: [9fans] So What is P9 good for.....

[Lucio De Re]

On Mon, Feb 17, 2003 at 08:36:47AM -0500, Russ Cox wrote:
>
> > I do note that on the NetBSD mailing list the sendmail vs postscript
> > debate is veering to taking both of those off the distribution,
>
> postscript, hands down, though pdf is pretty good too.

Oops.  I do check spelling, but I don't re-read what I wrote very
carefully :-(  "postfix".

++L

Re: [9fans] So What is P9 good for.....

[Andrew]

>
> It's twenty years on, and lunix distributions still include sendmail
> and BIND, despite their size, slowness, complexity, buginess, painful
> configuration files and long history of security bugs, usually due to
> buffer overruns.  There are certainly plausible alternatives to
> sendmail available.  I'm not sure about BIND, but I'd rather port dns

well you obviously have something of an almost religous and uninformed
view of linux, theres two things i want to point out, first not all linux
distributions come with sendmail by default, mine certainly doesnt, and
actively disuades anyone from running it. Next FreeBSD (OS/X included,
i dont know about net or open bsd's),  by default, come installed with
sendmail, in the base system. A make world on the last freebsd (4.5)
system wanted to build sendmail for me.

Now before you go ragging on us linux people, out of everyone i've talked
to who runs sendmail for religous reasons, every single one of them was
a BSD person. Come to think of it, the only time ive spoken with a linux
person who admitted to using sendmail did it many years ago when they
needed some obscure feature only sendmail implemented.


> The first security measures I take when installing lunix are:
>
> 	- find any sendmail processes and kill them all.
> 	- find all sendmail binaries, set permissions to 0 and then remove them.
> 	- find any sendmail queue directories or configuration files and remove them.
>

funny, ive never had to do that on a linux system. Maybe you shouldnt
be using such an inferiour distribution.

> What's wrong with these lunix people, are they stupid or something? ???

at least we can spell linux correctly.

>

Re: [9fans] So What is P9 good for.....

[Geoff Collyer]

Sorry, as I said earlier, I use "lunix" as shorthand for "unix and
linux".  I wasn't specifically criticising any given distribution
(other than freebsd), but Unoid systems in general.

I just find it hard to believe that after 20 years, some people are
*still* shipping known security hazards like sendmail.

Re: [9fans] So What is P9 good for.....

[Russ Cox]

oh come on.  sendmail doesn't have any bugs left.
this time for sure!

Re: [9fans] So What is P9 good for.....

[rob pike, esq.]

> I just find it hard to believe that after 20 years, some people are
> *still* shipping known security hazards like sendmail.

Oh yeah? Well, I find it hard to believe that after twenty years
you can't ls /dev to see how large your disks are, or that ps
still thinks a controlling tty is a dominant idea in choosing
which processes to display, or that ...  Or that the system is
still just like it was 20 years go, except for the crap that was
added on, never cleaned up, and never made to fit together,
let alone to be Unix-like in any way.

Sockets? X? They're so, well, not Unix-like.

It's sad.

-rob

Re: [9fans] So What is P9 good for.....

[northern snowfall]

I so love this list.

Re: [9fans] So What is P9 good for.....

[Russ Cox]

> Sockets? X? They're so, well, not Unix-like.

It's also sad that there are so many people
who believe sockets and X are exemplars
of the Unix style.

Re: [9fans] So What is P9 good for.....

[George Michaelson]

STREAMS. so... upper-case-y.

Perhaps if RFS had come along in a different year, we'd have felt different
about making an RFS vs NFS decision.

X was there, SunView was very proprietary. It could have been worse, we might
have all been running smalltalk/LISP workstations.

Maybe you'd rather be using PNX on a Perq? A system with an editor for the
animated cursors? A system which let you see memory usage by stealing graphic
memory for compilaton?

The best thing about subjective criticism is usually the pint of beer which
goes with it!

P9 needs to write a better porn distribution protocol. Presuming at least 7
others have been tried before, how about its called PDP-8?

-George

> > Sockets? X? They're so, well, not Unix-like.
>
> It's also sad that there are so many people
> who believe sockets and X are exemplars
> of the Unix style.

Re: [9fans] So What is P9 good for.....

[Skip Tavakkolian]

> Perhaps if RFS had come along in a different year, we'd have felt different
> about making an RFS vs NFS decision.

Enough time has passed for somebody to have come up with a better idea.

> X was there, SunView was very proprietary. It could have been worse, we might
> have all been running smalltalk/LISP workstations.

Blit had been around since '82, and it was a better solution.

Re: [9fans] So What is P9 good for.....

[Mike Haertel]

>Blit had been around since '82, and it was a better solution.

X was free and ran on multiple hardware platforms.

Re: [9fans] So What is P9 good for.....

[Douglas A. Gwyn]

Skip Tavakkolian wrote:
> Blit had been around since '82, and it was a better solution.

Well, not quite; to really displace X-Windows it would
have needed (in addition to *much* better marketing):
- more channels (7 not enough)
- some way to support multiple display platforms (not
   just 5620/630/730, and even for those one needed
   duplicate device-dependent downloadable images).
- color

[9fans] Re: acd compile problem

[Conor Williams]

hi
tried compiling acd again last night -
o what do I change doprint to?
o strconv isnt in any library - is there a new one?
o fmtinstall was being used already but
the function being passed was
msfconv(char *vaargs, Fconv *fp)
{
    processing of vaags...

    strconv(buf, fp);
    return 0;
}
(or similar)

fmtinstall needs a function of the form
msfconv(Fmt *fp)...
but how do I get the vaargs into the function?
do I fill Fmt before I send the fp?
or is that already rewritten
If someone could send me the function msfconv or a url
to the new version of acd that would be great
tx
will551

Re: [9fans] Re: acd compile problem

[nigel]

[-- Attachment #1: Type: text/plain, Size: 535 bytes --]

As Russ said, the version on /n/sources is a 4th edition version.
You must have the 3rd edition version off his website.

The real source is in /acme/bin/source/acd.

If it isn't there, then run replica to update your source.

If you can't do that, then read the manual page for fmtinstall(1)
which gives a very good example of how this stuff works.

If you can't do that then

int
msfconv(Fmt *fp)
{
	Msf m;

	m = va_arg(fp->args, Msf);
	fmtprint(fp, "%d.%d.%d", m.m, m.s, m.f);
	return 0;
}

is what you need.

[-- Attachment #2: Type: message/rfc822, Size: 2471 bytes --]

From: "Conor Williams" <connorwilliamsie@aol.com>
To: <9fans@cse.psu.edu>
Subject: [9fans] Re: acd compile problem
Date: Tue, 18 Feb 2003 09:49:49 -0000
Message-ID: <0a1801c2d733$142fb510$9949800a@will551>

hi
tried compiling acd again last night -
o what do I change doprint to?
o strconv isnt in any library - is there a new one?
o fmtinstall was being used already but
the function being passed was
msfconv(char *vaargs, Fconv *fp)
{
    processing of vaags...

    strconv(buf, fp);
    return 0;
}
(or similar)

fmtinstall needs a function of the form
msfconv(Fmt *fp)...
but how do I get the vaargs into the function?
do I fill Fmt before I send the fp?
or is that already rewritten
If someone could send me the function msfconv or a url
to the new version of acd that would be great
tx
will551

Re: [9fans] So What is P9 good for.....

[Skip Tavakkolian]

> Well, not quite; to really displace X-Windows it would
> have needed (in addition to *much* better marketing):

I agree on the marketing.

> - more channels (7 not enough)
> - some way to support multiple display platforms (not
>    just 5620/630/730, and even for those one needed
>    duplicate device-dependent downloadable images).

I recall one of the universities ported the DMD software to an Atari
ST (68000) about the time we were selling an IDRIS+X10R3 on STs
in '87.  DMD for ST worked well.  They couldn't release the sources
then, of course.

> - color

Re: [9fans] So What is P9 good for.....

[nigel]

> I recall one of the universities ported the DMD software to an Atari
> ST (68000) about the time we were selling an IDRIS+X10R3 on STs
> in '87.  DMD for ST worked well.  They couldn't release the sources
> then, of course.
>

More than one did.

Re: [9fans] So What is P9 good for.....

[Jack Johnson]

George Michaelson wrote:
> X was there, SunView was very proprietary. It could have been worse, we might
> have all been running smalltalk/LISP workstations.

Some of us still are:

	http://www.squeak.org

But on yet another light note, is there anything decent from the
freenices that Plan 9 could benefit from, other than Mozilla's renderer?
  ReiserFS4?

	http://www.namesys.com/v4/v4.html

-J

Re: [9fans] So What is P9 good for.....

[M Heath]

> But on yet another light note, is there anything decent from the
> freenices that Plan 9 could benefit from, other than Mozilla's renderer?
>   ReiserFS4?

users

oh and Tux Racer

Re: [9fans] So What is P9 good for.....

[Martin C.Atkins]

On Tue, 18 Feb 2003 09:23:15 +1000 George Michaelson <ggm@apnic.net> wrote:
>...
> Maybe you'd rather be using PNX on a Perq? A system with an editor for the
> animated cursors? A system which let you see memory usage by stealing graphic
> memory for compilaton?

Ah yes - the system with one of the most bizarre bugs I've ever come
across: If your user name was 6 or 7 letters long, and ended in an
'n', then after logging in, your environment was corrupt! Guess why I
changed my username from 'martin' to 'mca' on our PNX machine, after
one very frustrating afternoon!

Martin
--
Martin C. Atkins				martin@mca-ltd.com
Mission Critical Applications Ltd, U.K.		http://www.mca-ltd.com

Re: [9fans] So What is P9 good for.....

[matt]

The new architects of Lunix [I like that word - I pronounce it Loonix]
consider an increase in complexity progress :

 http://primates.helixcode.com/~miguel/bongo-bong.html

"Bonobo is our proposed solution to create reusable software components in
Unix and keeping the spirit of "Small is beautiful".
Each component will focus on implementing a complete and correct set of
interfaces and features.
We have chosen CORBA as the foundation for our component architecture. "


I particularly like the section on how much simpler system administration
will be using CORBA components rather than text files.



m

Re: [9fans] So What is P9 good for.....

[George Michaelson]

To be fair, an awful lot of mass hysteria accompanied the decision to go
Corba. We see a lot of this kind of Millinarial 'at *last* the software
mechanism which will deliver true <x>' chanting around the place. Dare I say
that in my OSI days, I probably chanted quite a lot of it too.

Most interesting to me is the drop in student registrations for I.T. degrees
here in Australia, and a consequent drop in the entrance requirements to get
into an I.T. degree. It never ceased to amaze me I passed my course. To see
people achieve the same paper worth with even less effort is truly
distressing.  As I believe the Bible says, As you sow, you shall reap. Dumbing
down is clearly delivering the inevitable outcome.

Still, there are moments of humour. I just re-wrote some code here to shrink
data using a classic tape-sort alg, iterating over the datastream. It runs
like a greased whippet, its tiny code, and I had lots of fun doing it, and the
advice to try such a simple method came from my erstwhile research director,
a man of similarly jaded views of our times. Meanwhile the small army of
code-orcs I manage beat out the Perl, in true OO fashion.

Actually, there were moments of pain in this last bout of coding. inet_pton()
is the newish function to map IPv6 and IPv4 addresses from presentation to
network byte order. As you can imagine, managing 128-bit constructs winds up
using a set of unions to see this as 8, 16 or 32 bit quantities, 128 bit
registers being thin on the ground (oh why did we reject VLIW computers?).

And predictably, NetBSD forces at least two of the useful #defines to be
#ifdef KERNEL since they are 'non-standard' while RedHate promotes them into
userspace, useful if alas tending towards the fork-in-the-code solution.

Oh well. ever was it so.


-George

Re: [9fans] So What is P9 good for.....

[Geoff Collyer]

Rob makes a good point; I guess it's going to be 1983 forever in
(l)unixland.  (Quite a trick for Linux.)

As for Bonobo, a private fortune file contains this comment:

	People admire complexity. - Rob Pike

Re: [9fans] So What is P9 good for.....

[Dan Cross]

> well you obviously have something of an almost religous and uninformed
> view of linux,

Those in glass houses....

	- Dan C.

Re: [9fans] So What is P9 good for.....

[Geoff Collyer]

Just a reminder that CPU and file servers can use any old VGA cards
since they only use CGA mode anyway.  For personal use, you probably
only need one terminal.  So limited choices in VGA hardware aren't a
great hardship.

As for the lack of web browsers and other Applications™, the beauty
of distributed computing is that you don't have do everything on one
machine with one operating system.  Once I get a bug or two related to
PCI cards in old Macs fixed in Darwin, I'll be running a web browser
on an OS-X Mac and using vnc over a private Ethernet to reach the Mac
from my terminal.  (Why a Mac?  It's the only platform I can find with
competent browsers [able to deal with banks' web sites, for example]
that don't blow up and take all their windows with them [and usually
pretty frequently].  You'd think they could fork a process per window,
to provide some insulation against core dumps, but noooooo.....  At
least Opera remembers which windows you had open, unlike Netscrape and
its ilk.)

As for why I use Plan 9, it's partly personal preference and partly
that Plan 9 is a very elegant and competent system.  Reading the early
Plan 9 papers, I got the same shiver down my spine as from reading the
CACM Unix paper, and in both cases thought `they've really figured it
out'.  (I was further amazed when I got access to a Unix system to
discover that it worked as described, without arbitrary limits or
hidden gotchas, in marked contrast to virtually all other operating
systems at the time.)  Only Tenex ever came close to that, and that was
probably in part due to my background working on PDP-10s.

Re: [9fans] So What is P9 good for.....

[Richard Miller]

> (I was further amazed when I got access to a Unix system to
> discover that it worked as described, without arbitrary limits or
> hidden gotchas, in marked contrast to virtually all other operating
> systems at the time.)

6th edition Unix was not altogether free of arbitrary limits.  For
example, userids were 8-bit integers ...

Re: [9fans] So What is P9 good for.....

[Geoff Collyer]

Compared to other systems of the day, even 6th edition seemed
marvellously free of arbitrary limits, particularly ones visible to
the user.  I could regale the list with horror stories of other
systems, but that's probably best done elsewhere.

Re: [9fans] So What is P9 good for.....

[Ronald G. Minnich]

On Fri, 14 Feb 2003, Geoff Collyer wrote:

>  I could regale the list with horror stories of other
> systems, but that's probably best done elsewhere.

no need to, we've all used DOS I bet.

ron

Re: [9fans] So What is P9 good for.....

[David Presotto]

[-- Attachment #1: Type: text/plain, Size: 152 bytes --]

not too mention having to recompile the compiler every now and
then with a bigger symbol table, 14 character file names, fixed
size process table, ...

[-- Attachment #2: Type: message/rfc822, Size: 1626 bytes --]

From: Richard Miller <miller@hamnavoe.demon.co.uk>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] So What is P9 good for.....
Date: Fri, 14 Feb 2003 09:31:23 0000
Message-ID: <9e468919efc946c025cece890686df57@hamnavoe.demon.co.uk>

> (I was further amazed when I got access to a Unix system to
> discover that it worked as described, without arbitrary limits or
> hidden gotchas, in marked contrast to virtually all other operating
> systems at the time.)

6th edition Unix was not altogether free of arbitrary limits.  For
example, userids were 8-bit integers ...

Re: [9fans] So What is P9 good for.....

[rob pike, esq.]

> not too mention having to recompile the compiler every now and
> then with a bigger symbol table, 14 character file names, fixed
> size process table, ...

512 bytes of arguments to commands.  yes, echo * was pretty
limited and a typical modern gcc compile line wouldn't even get
past the -Ws.

-rob

Re: [9fans] So What is P9 good for.....

[Ronald G. Minnich]

On Fri, 14 Feb 2003, rob pike, esq. wrote:

> 512 bytes of arguments to commands.  yes, echo * was pretty
> limited and a typical modern gcc compile line wouldn't even get
> past the -Ws.

this is bad, it looks like you're speaking from experience ...

ron

Re: [9fans] So What is P9 good for.....

[Geoff Collyer]

Yes, one can cite individual complaints about 6th edition, but my
point was that in 1975 or 1976, compared to the other systems of the
day, Unix *seemed* *to me*, as a user, virtually free of arbitrary
limits.

I could create directories beneath my home directory to a depth
greater than 6 (unlike, say, TOPS-10 and perhaps, later, VMS), indeed
to arbitrary depth (yes, I'd run the file system out of blocks or
i-nodes eventually, but at least the system didn't pick a
pseudo-random constant like `6').  I could create files with names
that weren't 6+3 upper-case alphanumerics, unlike most, if not all,
DEC operating systems and others influenced by them.  14 (ASCII)
characters per component seemed most generous and I can't recall
needing more.  The exec argument limit was an implementation artifact
of a small machine, and has been increased repeatedly over the years,
unlike the limits mentioned above, which couldn't be changed without
harming binary compatibility with existing programs.  8-bit userids
were a result of parsimonious choice of the `char' data type on a
small machine, not somebody saying `Let's limit the number of users to
169 just to be arbitrary and a pain', and the size of userids has been
increased over the years.

Reading through section II of the 6th edition manual (the system call
interface), about the only arbitrary limits (perhaps I should say
capricious; those that are not just a consequence of the size of some
data type) that I can see are the maximum number of open files per
process and, arguably, various output delays in STTY(II).  This was
not true of the system call interfaces of most other systems of the
day (not to mention programming language implementations), where it
was normal to find a general mechanism described, followed by a
limitation to some capricious number, usually small enough to
seriously limit the utility of the mechanism, and often further
reducible by the system administrator during system configuration, so
you couldn't even count on the miserly nominal limit in the
documentation.  Quoting Armando Stettner of DEC (from memory), ``Unix
isn't a police-state operating system.''.

And (returning to the question of why I use Plan 9) the feeling I got
reading the CACM Unix paper appeared again when reading about Plan 9
circa 1991.

Re: [9fans] So What is P9 good for.....

[Ronald G. Minnich]

On Fri, 14 Feb 2003, Geoff Collyer wrote:

> Yes, one can cite individual complaints about 6th edition, but my
> point was that in 1975 or 1976, compared to the other systems of the
> day, Unix *seemed* *to me*, as a user, virtually free of arbitrary
> limits.

exactly. One funny thing I remember is how controversial the idea of a
tree-structured file system was. I couldn't find anyone who likd it, as it
was obviously an inefficient fad.

And how about the whole obviously ridiculous idea of a user-mode command
interpreter,  and the bizarre concept of forking a new process for things
such as date and ls! outrageous! Inefficient! unworkable on "real"
systems!

Unix was such a breath of fresh air back then. Kind of the way Plan 9
feels.

ron

Re: [9fans] So What is P9 good for.....

[Digby Tarvin]

> Yes, one can cite individual complaints about 6th edition, but my
> point was that in 1975 or 1976, compared to the other systems of the
> day, Unix *seemed* *to me*, as a user, virtually free of arbitrary
> limits.

I don't recall any arbitrary limits in 6th edition. At UNSW we had well
over 256 users - achieved (in our case) by sacrificing group ownership
of files.

In my opinion, having access to well written and maintainable source
is the ultimate in freedom from arbitrary (or otherwise) limits.

Source availability is more common these days, but Plan9's is the only
one that is a pleasure to delve into, the way Edition 6 was back in its
day.

Regards,
DigbyT
--
Digby R. S. Tarvin                                              digbyt@acm.org
http://www.cthulhu.dircon.co.uk

Re: [9fans] So What is P9 good for.....

[Douglas A. Gwyn]

David Presotto wrote:
> not too mention having to recompile the compiler every now and
> then with a bigger symbol table, 14 character file names, fixed
> size process table, ...

Mostly those constraints were inherited from the PDP-11
platform, which supported only 16 bits of process
address space.  Consider that other PDP-11 OSes had
even more restrictive file name limits.

Re: [9fans] So What is P9 good for.....

[Douglas A. Gwyn]

Richard Miller wrote:
> 6th edition Unix was not altogether free of arbitrary limits.  For
> example, userids were 8-bit integers ...

As I recall the inode link count was 8 bits and link(2)
didn't check for overflow, resulting in a black hole
where a file ued to be..

I have this dream where *all* integers on a system are
open-ended, so one can never overflow their capability
to represent (except by running out of RAM or disk,
etc.).  We know how to represent integers efficiently;
what is lacking is direct hardware arithmetic support.