Re: [9fans] design issues in operating systems

Re: [9fans] design issues in operating systems

[forsyth]

>>The plan is to put a fileserver on the outside for everyone to access.

we were going to do that here once but our bandwidth is a bit limited
for general file server access by more than a small group of people.

Re: [9fans] design issues in operating systems

[Fco.J.Ballesteros]

[-- Attachment #1: Type: text/plain, Size: 107 bytes --]

Although our network is not too reliable, I'm willing
to put such a file server here at escet.urjc.es.


[-- Attachment #2: Type: message/rfc822, Size: 1372 bytes --]

From: forsyth@vitanuova.com
To: 9fans@cse.psu.edu
Subject: Re: [9fans] design issues in operating systems
Date: Tue, 4 Dec 2001 10:36:22 0000
Message-ID: <20011204103537.373C71998A@mail.cse.psu.edu>

>>The plan is to put a fileserver on the outside for everyone to access.

we were going to do that here once but our bandwidth is a bit limited
for general file server access by more than a small group of people.

Re: [9fans] design issues in operating systems

[erik quanstrom]

> On Mon Dec  3 15:52:18 EST 2001, mike@ducky.net wrote:
> > I suspect one important factor in the popularity of Linux
> > is the relative openness of the development process of Linux.
> > There are regular releases of the latest version of the
> > code, even when there are works in progress that aren't
> > finished yet.
> >
> > By contrast the Plan 9 developers are constantly dropping
> > all these hints about how great the next thing will be,
> > but the rest of us never get to see any code until it's
> > "done" -- and big changes take a long time.
> >
> > I think Plan 9 would have a more enthusiastic following
> > if the ongoing development tree were out there for everyone
> > to see (and contribute to).
>
> Do we hint that the next thing will be great? If so, we should stop.
> It may or may not be great but it will be different, I think that's
> about as far as I'd be willing to go.

in so many words, i think so.

9p2000 has been talked about on this list for quite some time now.
(i'm waiting with baited breath to see it.) and rather than somebody
/saying/ that the next thing will be great, i think that folks have
gotten excited to see the 9p2000 without any hype. i know i have.

don't stop dropping tidbits. they're interesting. but i think mike's
point is that having something akin to (for lack of a better analogy)
the linux development series (e.g. x.y.z where y is odd) would be a
Good Thing.

of course, doing this isn't free. it takes time & effort to release.
and releasing early & often take time early & often.

Re: [9fans] design issues in operating systems

[jmk]

On Mon Dec  3 15:52:18 EST 2001, mike@ducky.net wrote:
On Mon Dec  3 15:52:18 EST 2001, mike@ducky.net wrote:
> I suspect one important factor in the popularity of Linux
> is the relative openness of the development process of Linux.
> There are regular releases of the latest version of the
> code, even when there are works in progress that aren't
> finished yet.
>
> By contrast the Plan 9 developers are constantly dropping
> all these hints about how great the next thing will be,
> but the rest of us never get to see any code until it's
> "done" -- and big changes take a long time.
>
> I think Plan 9 would have a more enthusiastic following
> if the ongoing development tree were out there for everyone
> to see (and contribute to).

Do we hint that the next thing will be great? If so, we should stop.
It may or may not be great but it will be different, I think that's
about as far as I'd be willing to go.

The plan is to put a fileserver on the outside for everyone to access.
The details and time-line are still foggy as they rely on 2 of the 'big
changes' - new authentication and a reworking of the fileserver code.

Re: [9fans] design issues in operating systems

[rob pike]

> I think Plan 9 would have a more enthusiastic following
> if the ongoing development tree were out there for everyone
> to see (and contribute to).

I agree.

-rob

Re: [9fans] design issues in operating systems

[Alexander Viro]

On Mon, 3 Dec 2001, rob pike wrote:

> > I think Plan 9 would have a more enthusiastic following
> > if the ongoing development tree were out there for everyone
> > to see (and contribute to).
>
> I agree.

Hmm...  I doubt that "for everyone to contribute" is a good thing,
but seeing is a different story...

Visible /lib and /src would be Real Nice(tm) - subset you can legally
distribute rsync'ed with external well-connected box to avoid eating
bandwidth on your side...  Finding such box wouldn't be a problem,
the question is whether you want/can do that.  Up to you, indeed...

Re: [9fans] design issues in operating systems

[Dan Cross]

In article <Pine.GSO.4.21.0112031647410.17686-100000@binet.math.psu.edu> you write:
>Hmm...  I doubt that "for everyone to contribute" is a good thing,
>but seeing is a different story...

Nonsense.  Everyong is free to contribute.  But, contribution is not
synonymous with acceptance.

The labs might run into another problem if development source is there
for all to pick at.  Surely they will receive a ton of patches for things
that they're already aware of, have already fixed but not synced to the
outside fileserver, or are still experimental and changing rapidly.  The
cost in terms of time and resources to sift through what comes in from
the outside could be non-trivial.

	- Dan C.

Re: [9fans] design issues in operating systems

[Alexander Viro]

On Tue, 4 Dec 2001, Dan Cross wrote:

> In article <Pine.GSO.4.21.0112031647410.17686-100000@binet.math.psu.edu> you write:
> >Hmm...  I doubt that "for everyone to contribute" is a good thing,
> >but seeing is a different story...
>
> Nonsense.  Everyong is free to contribute.  But, contribution is not
> synonymous with acceptance.

<shrug> well, provided that your .procmailrc is well-maintained - sure.
You _really_ don't want to see the... contributions from Qlogics SCSI
folks.  Or Intel folks who do NIC drivers.  There are less harmful ways
to take a second look at your last meal...

> The labs might run into another problem if development source is there
> for all to pick at.  Surely they will receive a ton of patches for things
> that they're already aware of, have already fixed but not synced to the
> outside fileserver, or are still experimental and changing rapidly.  The
> cost in terms of time and resources to sift through what comes in from
> the outside could be non-trivial.

_That_ is easy to deal with -
	/* XXX: Yes, I know.  It's crap.  Working on that. */
is usually enough for stuff that stays for a week or so and for very
recent breakage you have not too may places you've just touched, so
sorting is not a big deal...

Re: [9fans] design issues in operating systems

[Barry]

 From Alexaner:

_That_ is easy to deal with -
	/* XXX: Yes, I know.  It's crap.  Working on that. */

HEY!!! That's not the way a REAL PROGRAMMER writes code. If it was hard to
write, it should be twice as hard to understand. ;)

(somewhere I have the ancient Usenet (humor) posting about "Real Programmers".
Hmmmm ... also "If OS's were Beer" !)
==================================================================
  Once a proud programmer of Apple II computers, he now spends his days
     and nights in cheap dives fraternizing with exotic dancers.

Re: [9fans] design issues in operating systems

[rob pike]

> It's interesting that LT thinks that Linux wasn't designed, but is just
> a product of random evolution.  I always suspected a million monkeys were
> involved.

By this criterion alone, with our without the jokey extra sentence, it's hard to
distinguish Linux from Windows.

-rob

Re: [9fans] design issues in operating systems

[presotto]

Anothy stole my thunder.  Plan 9 is also a result of
years of churn.  No original code or interface has
survived though some general principles have.

- name spaces should be subjective
- most objects should live in those name spaces
- simplicity is preferable to bells and whistles

If anything, we regularly go through and rewrite
the kernel and every command if we think something can
be done better.  It is a research system and we're not
very interested in backward compatability and only mildly
in a large user community.  We steal equally from everyone we
can, modulo an NIH attitude that's always hard to shake
in a big corp.

Linus started with an incredibly detailed design that
took years to make usable.  He started with the system,
library, and user interfaces of a well used and mature
system.  That's a hell of a lot more design than we stared
with in Plan 9 and it's remained a lot more immutable
than Plan 9's.

I agree with Linus that you can't design and then
walk away, but then again, I don't know anyone
who would agree with that.  It may just seem that
way because mature systems eventually bog down
under the weight of their own backward compatibility.

Re: [9fans] design issues in operating systems

[anothy]

// There's merit in evolution, which adjusts to changing
// conditions, whereas design needs to be predictive.

why are people talking about design and evolution as though
they're mutually exclusive?

the plan 9 kernel was, i believe, designed, but it has certainly
evolved since the first edition (witness the rift between the
file server kernel and the cpu/terminal kernel). we've seen a
number of different graphics models in Plan 9, and i believe
each of them was designed, but it may also make sense to
talk about them as evolutions of each other, based on what
practical application in differing environments shows to be
better or worse - like natural selection. note the movement
of image memory between kernel and userland and the
movement of the IP stack (if i'm remembering these right).

a good design is one which is simple enough to comfortably
allow for reasonable evolution. whereas i agree design _is_
predictave, the results need not be static.
ア

Re: [9fans] design issues in operating systems

[Lucio De Re]

On Mon, Dec 03, 2001 at 11:07:49AM -0500, anothy@cosym.net wrote:
>
> why are people talking about design and evolution as though
> they're mutually exclusive?
>
To a degree, that's the case, though.  Design enforces rigidity and
wherever there's flexibility there's also room for error, which is
what design attempts to factor out.  Flexibility is where adaptation
occurs, with fatal mutations more the norm than the exception.

> a good design is one which is simple enough to comfortably
> allow for reasonable evolution. whereas i agree design _is_
> predictave, the results need not be static.
>
Simplicity in design (minimalism, in fact) is the ideal, in that its
rigidity is limited to essentials and, hopefully, does not apply to
the growth/evolution areas.  My opinion is that one should formalise
useful adaptations and absorb them into a design, which is the way I
think we humans operate at an intellectual level.

The difficulty is finding the motivation or vision to abandon baggage
whose function is exclusively to provide a familiar environment.
Remembering that in this forum I suggested not long ago that one
should not underestimate the importance of familiarity :-)

In this context, familiarity is just another evolutionary pressure
and, I must point out, evolution has no foresight, its purpose is to
increase the viability of an organism in the present, immediately
adjacent environment.

The mistake of assuming that evolution is progressive is evident even
in Linus Torvald's statement quoted earlier: the implication is that
an evolved Linux is somehow "better" that its predecessors.  It could
be agreed that it is better equipped for survival, but not necessarily
superior in some intellectual sense.

Materialists may well argue that that is all that counts, of course.

++L

Re: [9fans] design issues in operating systems

[bwc]

> The market place is irrational

The market place only LOOKS irrational.  It will always use
options it understands to make choices to protect its scare resources.
Few know about Plan 9 and fewer still understand enough to
see its better merits.

Ely Whitney almost went broke with his new idea of interchangable
parts, but even if he had gone broke the idea of interchangable
parts would have survived.

Re: [9fans] design issues in operating systems

[bwc]

It's interesting that LT thinks that Linux wasn't designed, but is just
a product of random evolution.  I always suspected a million monkeys were
involved.

Ken and Dennis did the designing; everyone else (me included) has just been
contributing to local entropy.

And intelligent people disagree about there not being a supreme creator.

  Brantley

[9fans] design issues in operating systems

[Andrey A Mirtchovski]

just before someone else points it out, i'd like to direct your attention to
this collection of email off the linux kernel list:

http://kerneltrap.org/article.php?sid=398 (slashdot gets the credit for
originating it)..

there you will see linus arguing that he never designed his operating
system, and that it evolved much like evolution. furthermore he claims that
when software is done in such way, it ends up better...

of course there's an obligatory plan9 quote, which is indeed the reason i'm
posting this message :)

the text not preceeded by '>' is linus'

<paste>
> Ok. There was no design, just "less than random mutations".
> Deep.

I'm not claiming to be deep, I'm claiming to do it for fun.

I _am_ claiming that the people who think you "design" software are
seriously simplifying the issue, and don't actually realize how they
themselves work.

> There was a overall architecture, from Dennis and Ken.

Ask them. I'll bet you five bucks they'll agree with me, not with you.
I've talked to both, but not really about this particular issue, so I
might lose, but I think I've got the much better odds.

If you want to see a system that was more thoroughly _designed_, you
should probably point not to Dennis and Ken, but to systems like L4 and
Plan-9, and people like Jochen Liedtk and Rob Pike.

And notice how they aren't all that popular or well known? "Design" is
like a religion - too much of it makes you inflexibly and unpopular.
</paste>

Re: [9fans] design issues in operating systems

[josh d]

  Considering the popularity of Linux vs the popularity of Plan9, he may
have a point.  Think about the issues and don't automatically descriminate
because he said something you dislike.

"Andrey A Mirtchovski" <aam396@mail.usask.ca> wrote in message
news:Pine.GSO.4.10.10112021058100.6307-100000@ultra5c.usask.ca...
> just before someone else points it out, i'd like to direct your attention
to
> this collection of email off the linux kernel list:
>
> http://kerneltrap.org/article.php?sid=398 (slashdot gets the credit for
> originating it)..
>
> there you will see linus arguing that he never designed his operating
> system, and that it evolved much like evolution. furthermore he claims
that
> when software is done in such way, it ends up better...
>
> of course there's an obligatory plan9 quote, which is indeed the reason
i'm
> posting this message :)
>
> the text not preceeded by '>' is linus'
>
> <paste>
> > Ok. There was no design, just "less than random mutations".
> > Deep.
>
> I'm not claiming to be deep, I'm claiming to do it for fun.
>
> I _am_ claiming that the people who think you "design" software are
> seriously simplifying the issue, and don't actually realize how they
> themselves work.
>
> > There was a overall architecture, from Dennis and Ken.
>
> Ask them. I'll bet you five bucks they'll agree with me, not with you.
> I've talked to both, but not really about this particular issue, so I
> might lose, but I think I've got the much better odds.
>
> If you want to see a system that was more thoroughly _designed_, you
> should probably point not to Dennis and Ken, but to systems like L4 and
> Plan-9, and people like Jochen Liedtk and Rob Pike.
>
> And notice how they aren't all that popular or well known? "Design" is
> like a religion - too much of it makes you inflexibly and unpopular.
> </paste>

Re: [9fans] design issues in operating systems

[Ronald G Minnich]

On Mon, 3 Dec 2001, josh d wrote:

>   Considering the popularity of Linux vs the popularity of Plan9, he may
> have a point.  Think about the issues and don't automatically descriminate
> because he said something you dislike.

I'm not sure it's a fair experiment. Plan 9 made it to "market" ten years
after Linux . Still worse, from 1991-2000 ATT/Lucent were actively trying
to kill it (though I doubt their management saw it that way ...).

I'm kind of hating life with Linux 2.4 series kernels. I keep thinking,
"if only I could just cut over to Plan 9 ...".

It is easy to make a case that Linux was not designed.

ron

Re: [9fans] design issues in operating systems

[Lucio De Re]

On Mon, Dec 03, 2001 at 08:24:29AM -0700, Ronald G Minnich wrote:
>
> It is easy to make a case that Linux was not designed.
>
Minix was designed, I should think.  There's merit in evolution, which
adjusts to changing conditions, whereas design needs to be predictive.

The market place is irrational (my opinion is that it is conditioned by
the media in self-interest, but that's a conspiracy theory) so design
is doomed to failure.  The spin-offs, however will cause mutations in
adaptive (and receptive) organisms.

++L

Re: [9fans] design issues in operating systems

[andrey]

On Mon, 3 Dec 2001, josh d wrote:

>   Considering the popularity of Linux vs the popularity of Plan9, he may
> have a point.

think 'quantity vs quality' of the userbase (me excluded :)

Re: [9fans] design issues in operating systems

[Ozan Yigit]

josh d <blah@microsoft.com> writes: [re: linus commentary on design]

>   Considering the popularity of Linux vs the popularity of Plan9, he may
> have a point.  Think about the issues and don't automatically descriminate
> because he said something you dislike.

i for one classify what linus says about design in the same category
of the "cathedral and the bazaar" proclamations. this is shallow stuff
that happens to have the right sound-bite quotient; ideal for the masses
that want quick, cheap bits of wisdom. i wish someone would do to this
what perlman does to network protocol design.

or perhaps i misunderstand all of it: there may be a taoist
undercurrent here:

	the design (tao) had best be left at rest.
	like little birdies at their nest.

so linus did not "design" linux, but it is emergent.

it makes my head spin.

oz
---
www.cs.yorku.ca/~oz	 | narrowness of imagination leads to
york u. computer science | narrowness of experience. - oz [corollary to rob]

Re: [9fans] design issues in operating systems

[Mike Haertel]

I suspect one important factor in the popularity of Linux
is the relative openness of the development process of Linux.
There are regular releases of the latest version of the
code, even when there are works in progress that aren't
finished yet.

By contrast the Plan 9 developers are constantly dropping
all these hints about how great the next thing will be,
but the rest of us never get to see any code until it's
"done" -- and big changes take a long time.

I think Plan 9 would have a more enthusiastic following
if the ongoing development tree were out there for everyone
to see (and contribute to).

Re: [9fans] design issues in operating systems

[north_]

Ok, normally I do not take the time to write responses regarding
issues like these. However, I believe Linus is a bit misguided
and I would like to quickly postulate why. This isn't just a
Linux problem. Its a lifestyle problem. You will see why I
believe so in just one second or so ;)
> --- bla bla bla snip ---
> If you want to see a system that was more thoroughly _designed_, you
> should probably point not to Dennis and Ken, but to systems like L4 and
> Plan-9, and people like Jochen Liedtk and Rob Pike.
>
> And notice how they aren't all that popular or well known? "Design" is
> like a religion - too much of it makes you inflexibly and unpopular.
*G*.
Design is in actuality nothing more than a personal concept. Design
itself is a facade. Especially when you believe design is a laid out
perfected ideal that you have generated previous to an action. Action
itself is a flowing movement. It is not bound by time since action
can never be initiated at an exact time and can never be terminated
at one other precise moment in time. We only have approximations.
Thus, it can never be 100% prepared for. As you move the world around
you moves. The universe moves. We are constantly adjusting our theories
and reasoning based on new evidence and new ideas. This is the same
with both Linux and Plan 9. Things evolve over time in their own way.
Design in this case is just few people sharing mostly private ideas
as opposed to a global conglomeration of minds constructing solutions
to globally perceived problems. The latter gets quite sloppy as we
have all seen with the constant volatile state in which the Linux
kernel exists. Regarding flexibility? Who gives you a better develop-
ment system? Linux or Plan 9? If you have read anything about Plan 9's
development scheme or used it what-so-ever you will immediately see
important advances in debugging with imported namespaces cross-platform,
a ready-to-write cross compilation suite over many platforms and the
list just goes on. Bravo! As for the _lifestyle_ choice? This comes down
to one exact statement: The largest group in any society will always choose
the shortest path to its goal. The easiest and seemingly most 'robust'
road. This is why Linux is currently so popular and Plan9 is not.
A Linux user is more likely to install with aptget and the like. Bah!
Give me source! Give me challenge! Give me ideas and theory and a
fresh way at looking at our life! For people who crave knowledge and
really want to understand our world there will always be smart
projects like Plan 9. If you ever looked at an operating system, a
woman, a man, a house, a goal .. and you thought to yourself "That
is beyond me.." ask yourself "Is it really that I am not capable of
this or is it that I believe I am not." Society changes you. You
change society. Its a lifestyle choice. We make it every day.
Decide.

north_

Re: [9fans] design issues in operating systems

[John S. Dyson]

north_ <northern_snow78@yahoo.com> wrote in message news:<48effcda.0112022010.d4a2c9f@posting.google.com>...
> Ok, normally I do not take the time to write responses regarding
> issues like these. However, I believe Linus is a bit misguided
> and I would like to quickly postulate why. This isn't just a
> Linux problem. Its a lifestyle problem. You will see why I
> believe so in just one second or so ;)
> > --- bla bla bla snip ---
> > If you want to see a system that was more thoroughly _designed_, you
> > should probably point not to Dennis and Ken, but to systems like L4 and
> > Plan-9, and people like Jochen Liedtk and Rob Pike.
> >
> > And notice how they aren't all that popular or well known? "Design" is
> > like a religion - too much of it makes you inflexibly and unpopular.
> *G*.
> Design is in actuality nothing more than a personal concept. Design
> itself is a facade.
>

Well, most of the Linux (and *BSD) design wasn't done by those
who are working on it today.  Linux was a copy of a previous
design work done by someone else.

There might be some superior re-implementation in Linux or *BSD,
but claiming that Linux was designed by those who code it would
strongly imply that the concepts were also authored.  Without
the template provided by previous UNIX implementations, there would
not have been a template for Linux or *BSD.

There is some common source heritage between BSD and SVR4, but
the seed design certainly came from a couple of Bell Labs guys.

In an informal sense, the Linux developers might claim that they
designed the code.  However, making the proclaimation that Linux
was designed starting with the first line of code written for
it would be rather overstated.  Linux was essentially DESIGNED
(but not IMPLEMENTED) before the first line of code was written.

John

Re: [9fans] design issues in operating systems

[Douglas A. Gwyn]

north_ wrote:
> Design is in actuality nothing more than a personal concept. Design
> itself is a facade. ...

?  Design is a process by which requirements are combined with
resources, knowledge, and experience to produce a sufficiently
good solution.  The hardest part is getting a handle on the
*real* requirements, as experienced systems analysts can attest.
For example, one Plan 9 requirement seems to be that the people
working on its development have fun doing that, so any design
that doesn't take that into account would be a poor one.