Let's Face It
Malcolm L. Carlock
and Windows-NT is so that IBM and MS, respectively, can provide (to some
extent) services already available under Unix for many years (including DOS
under Unix), but while locking those users into IBM as their OS vendor.
It would appear, then, that any claims of the developers of these three
products that they're doing what they do in order to "do what's best for
the user" are secondary to the above constraint.
If IBM/Microsoft were really interested in providing the best multitasking
environment for their users, they could have provided the same, and probably
better, services by throwing their considerable resources into providing
The Best Desktop Unix Ever (they might even have succeeded.) Any "new"
features that OS/2 or Windows provide could easily have been added to that
Unix, as new features have always been more or less easily added to Unix.
Instead, it seems arguable that both vendors have copped out and followed
the path of trying to lock users into relatively mediocre products provided
by a single vendor, while as much as possible attempting to ignore or obscure
(with the help of "computing" journals that are in far too many cases
IBM/MS-sycophantic to the point of inducing nausea*) the existence of
alternatives. This approach does little to advance the state of the art or
provide the best solutions to users.
Users, businesses and institutions who need real power, flexbility, and the
ability to get real work done will, after a bit of research, usually choose
Unix. Users who continue to think of IBM and MS as their "buddies" in the
desktop world will continue paying more (at the middle and high end) for
a combination of less capable hardware (Intel chips in PC's still slower
than most RISC processors) and OS software (e.g., lack of multiuser support
in OS/2, no real virtual memory support in MS-Windows -- else how can a
memory violation by an application necessitate a reboot??)
Users of OS/2, MS-Windows, Windows-NT may like the products, but in
purchasing them are supporting vendors who provide users of these products
with little, if anything, that is "new" or not available through Unix
(multi-vendor), while at the same time locking those users into a single
OS vendor.
This is nice for the vendor. One hopes, however, that users will have
somewhat higher standards than this, and thereby support a marketplace
that provides the user with considerably more choice and power.
-----------------------------------------------------------------------------
* Does anyone remember "Computer Language" magazine's "Multitasking" issue
from a few years ago? Instead of something useful and informative to
people who use computer languages professionally, like an issue comparing
multitasking programming in a number of environments (Unix, VMS, VM, MVS,
Amiga, etc.), it was essentially ENTIRELY devoted to hacks for doing non-
preemptive (!) multitasking under DOS.
Does this magazine still exist? If so, hopefully they've renamed it to
"Computer Language, but only under MS-DOS"
-----------------------------------------------------------------------------
Malcolm L. Carlock
>under Unix), but while locking those users into IBM as their OS vendor.
Sorry, the above should read "into IBM and Microsoft, respectively, as
their OS vendors."
Mike Dahmus
And your whole posting is full of conspiratorial tripe. The free market works,
buddy - get used to it.
OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
box with X (and I've tried plenty) that could perform anywhere nearly as well
as OS/2 (the only one that comes close is linux, and then all you have to do
is take the Workplace Shell off of OS/2 to bring it down to linux's level of
function). If IBM had just produced Yet Another Slightly Incompatible Version
Of UNIX (like every other company under the sun), the performance advantage
would have been impossible. Likewise, if IBM had built their user interface on
top of X, they would have suffered the same performance and usability problems
as all the other UNIXes.
Why are there no usable applications for any major UNIX platform? (I'm not
calling NeXTStep UNIX).
UNIX is no panacea. Grow up and join the real world.
------
Mike Dahmus Internet: mi...@vnet.ibm.com
MVP Development, IBM PSP IBM: mi...@schleppo.bocaraton.ibm.com, MDAHMUS at BOCA
Disclaimer: Not an official IBM spokesman
Dan Pop
>
>OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
>box with X (and I've tried plenty) that could perform anywhere nearly as well
>as OS/2 (the only one that comes close is linux, and then all you have to do
>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>function). If IBM had just produced Yet Another Slightly Incompatible Version
>Of UNIX (like every other company under the sun), the performance advantage
>would have been impossible. Likewise, if IBM had built their user interface on
>top of X, they would have suffered the same performance and usability problems
>as all the other UNIXes.
>
>Why are there no usable applications for any major UNIX platform? (I'm not
>calling NeXTStep UNIX).
>
>UNIX is no panacea. Grow up and join the real world.
>
According to you, that's one of their greatest mistakes.
Dan
--
Dan Pop
Tel: +41.22.767.2335
Email: dan...@cernapo.cern.ch
Mail: CERN - PPE, Bat. 21 1-023, CH-1211 Geneve 23, Switzerland
Mike Dahmus
>In <CC9Mz...@hawnews.watson.ibm.com> mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>>
>>OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
>>box with X (and I've tried plenty) that could perform anywhere nearly as well
>>as OS/2 (the only one that comes close is linux, and then all you have to do
>>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>>function). If IBM had just produced Yet Another Slightly Incompatible Version
>>Of UNIX (like every other company under the sun), the performance advantage
>>would have been impossible. Likewise, if IBM had built their user interface on
>>top of X, they would have suffered the same performance and usability problems
>>as all the other UNIXes.
>>
>>Why are there no usable applications for any major UNIX platform? (I'm not
>>calling NeXTStep UNIX).
>>
>>UNIX is no panacea. Grow up and join the real world.
>>
>Then, why does IBM sell AIX for their RS 6000 computers, instead of OS/2?
>According to you, that's one of their greatest mistakes.
I hate to quote an entire article, but I think it's important to make a point
here.
"UNIX is no panacea". What does this mean to you? Would you understand the
point better if I also said "OS/2 is no panacea"? There's room for both. My
response was to someone who got the idea that any o/s not built on UNIX had
no place in the computing world. I'd like it if IBM sold OS/2 and AIX for their
RS/6000s. We'll be close to that situation when Workplace OS is written
(allowing one to run some combination of the two on the next generation of IBM
hardware).
Malcolm L. Carlock
>In <1993Aug23.2...@unlv.edu>, ma...@pyramid.unr.edu (Malcolm L. Carlock) writes:
>
>The free market works, buddy - get used to it.
Indeed it does. And luckily, free speech lets us comment on corporations
and other entities when they do things that appear to be attempts at making
the market less free.
>OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
>box with X (and I've tried plenty) that could perform anywhere nearly as well
>as OS/2 (the only one that comes close is linux, and then all you have to do
>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>function).
The Unix machines I deal with (from many vendors) tend to stay up for months
at a time (try this with MS-Windows), despite occasionally full filesystems,
the occasional user who writes and runs an application that violates memory
bounds, etc. Despite such activities, and "heavy" loads due to multiple
simultaneous users, applications and the (X-based) windowing systems on the
consoles still perform well.
For this kind of robustness and flexibility, I'll take a slight performance
hit (if any) if I have to.
>Why are there no usable applications for any major UNIX platform?
I was sure you would have heard of Autocad, ArcInfo, WordPerfect, Lotus, dBase
(originally developed on Unix, BTW), etc. by now. And generally cheaper per-
seat for Unix as well. Or don't these count as "usable"?
>(I'm not calling NeXTStep UNIX).
Why not?
>UNIX is no panacea.
During the past few years I and many of my colleagues have, quite
frustratingly, accumulated man-months of lost time dealing with IBM and MS
"solutions" (or lack thereof) from the mainframe level all the way down to
PC's. What takes a few seconds to set up under Unix (host-controlled
terminal printing, for example) takes months of interminable committee
meetings with the local IBM mainframe rep. Windows, supposedly a virtual-
memory environment, locks up or crashes if a DOS application commits a
"memory violation" (or, sometimes, for no apparent reason at all.)
As someone else has mentioned, open systems are "in", and with good reason.
Given the track record above, efforts by IBM, MS etc. to entice users into
environments essentially controlled entirely by those vendors, and
sanctimonious pronouncements that furthermore these vendors and these
environments hold "all the answers", are just about enough to make me toss
my cookies.
> Grow up and join the real world.
Strangely, local users who chose Unix for large-scale computing or server
tasks have generally been able to start getting their work done almost "out
of the box", while users who've chosen non-Unix solutions to support such
enterprises have often run into many problems, delays, and sometimes remarkable
levels of expense (not to mention lack of features, e.g., no multiuser
support in OS/2, 8+3 uppercase filenames, etc., and the usual monopoly-style
disadvantages of single-source availability.)
If the "real world" involves IBM and/or MS worship (and the huge number of
lost man-hours and lesser results that such an approach has, in my experience,
yielded) vs. getting the job done most effectively, I'll take my "real world"
over yours, thanks.
Followups to alt.flame.
Malcolm L. Carlock
>
>The free market works, buddy - get used to it.
Indeed it does. And luckily, free speech lets us comment on corporations
and other entities when they do things that appear to be attempts at making
the market less free.
>OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
>box with X (and I've tried plenty) that could perform anywhere nearly as well
>as OS/2 (the only one that comes close is linux, and then all you have to do
>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>function).
The Unix machines I deal with (from many vendors) tend to stay up for months
at a time (try this with MS-Windows), despite occasionally full filesystems,
the occasional user who writes and runs an application that violates memory
bounds, etc. Despite such activities, and "heavy" loads due to multiple
simultaneous users, applications and the (X-based) windowing systems on the
consoles still perform well.
For this kind of robustness and flexibility, I'll take a slight performance
hit (if any) if I have to.
>Why are there no usable applications for any major UNIX platform?
Not sure what your definition of "usable" is, but I was sure you would have
heard of Autocad, ArcInfo, WordPerfect, Lotus, dBase (originally developed on
Unix, BTW), etc. by now. And generally cheaper per-seat for Unix as well.
Or don't these count as "usable"?
>(I'm not calling NeXTStep UNIX).
Why not?
>UNIX is no panacea.
During the past few years I and many of my colleagues have, quite
frustratingly, accumulated man-months of lost time dealing with IBM and MS
"solutions" (or lack thereof) from the mainframe level all the way down to
PC's. What takes a few seconds to set up under Unix (host-controlled
terminal printing, for example) takes months of interminable committee
meetings with the local IBM mainframe rep. Windows, supposedly a virtual-
memory environment, locks up or crashes if a DOS application commits a
"memory violation" (or, sometimes, for no apparent reason at all.)
As someone else has mentioned, open systems are "in", and with good reason.
Given the track record above, efforts by IBM, MS etc. to entice users into
environments essentially controlled entirely by those vendors, and
sanctimonious pronouncements that furthermore these vendors and these
environments hold "all the answers", are just about enough to make me toss
my cookies.
> Grow up and join the real world.
Strangely, local users who chose Unix for large-scale computing or server
tasks have generally been able to start getting their work done almost "out
of the box", while users who've chosen non-Unix solutions to support such
enterprises have often run into many problems, delays, and sometimes remarkable
levels of expense (not to mention lack of features, e.g., no multiuser
support in OS/2, 8+3 uppercase filenames, etc., and the usual monopoly-style
disadvantages of single-source availability.)
I stand by my original assertion that OS/2, and Windows in any form, exist,
apparently, mainly as mechanisms for trying to lock users into a single
vendor solution, and that the development paths of these environments are
subject, first and foremost, to that constraint.
If the "real world" involves IBM and/or MS worship (and the huge number of
lost man-hours and lesser results that such an approach has, in my experience,
yielded) vs. getting the job done most effectively, I'll take my "real world"
over yours, thanks.
Followups to alt.flame.
---
Opinions expressed here are my own only.
--
Malcolm L. Carlock Internet: ma...@unr.edu
UUCP: unr!malc
BITNET: malc@equinox
U20...@hnykun11.urc.kun.nl
general, but I have always been fascinated by the fact that UNIX seems to able
to run on an old crate, while the above mentioned systems always need some kind
of upgrading whenever you want to do something very ordinary, like running an
application.
"OS/2 needs at least 8 Mb. to run satisfactorily." Really....
Paul.
Steve Houle
>In <CC9Mz...@hawnews.watson.ibm.com> mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>>
>>
>>UNIX is no panacea. Grow up and join the real world.
>>
>Then, why does IBM sell AIX for their RS 6000 computers, instead of OS/2?
>According to you, that's one of their greatest mistakes.
From what I've heard lately, AIX is so bad that it IS one of IBM's
greatest mistakes.
Wook
--
Steve Houle ...is: woo...@panix.com
Jacqueline U. Robertson
>box with X (and I've tried plenty) that could perform anywhere nearly as well
>as OS/2 (the only one that comes close is linux, and then all you have to do
>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>function). If IBM had just produced Yet Another Slightly Incompatible Version
>Of UNIX (like every other company under the sun), the performance advantage
>would have been impossible. Likewise, if IBM had built their user interface on
>top of X, they would have suffered the same performance and usability problems
>as all the other UNIXes.
>
(IBM PS/2 model 90 486) and a Sun Sparc 10, I had a MUCH easier time on the
Sparc. When I closed my windows, they didn't wander off into a folder
somewhere, they went to the bottom (or side) of the screen - just like
most other windowing environments I've ever used (Mac, Windows).....
Yes, I know that OS/2 is configurable. But I had WORK to do, and did not have
time to spend twiddling an interface. Configurability is NOT what most
users want, they want familiarity. The interface should be close to
what is offered under Mac or Windows simply because that's what most people
are used to.
Now I KNOW that I'm going to get flamed as a luddite for this - but in
my consulting experience the above is what user's want. There's a small
percentage of wonks out there who like to twiddle with the interface, and
then there's everybody else, who just wants to get a spreadsheet or document
out. The major issue I have with OS/2 is the interface - diffrent for the sake
of being different just doesn't cut it.
>Why are there no usable applications for any major UNIX platform? (I'm not
>calling NeXTStep UNIX).
Actually, there are, they are just priced WAY to high (in general). The unix
market has been more of a vertical (read high priced) market than has the PC
market. Thus fewer 'mainstream' apps and higher prices. This may (or may
not) change.
>
>UNIX is no panacea. Grow up and join the real world.
I could say the same.... OS/2 is a dead end - not technologically, but as a
market force. Thus far in the game, IBM has shown all the marketing savvy of
a lost four year old - marketing and money will ALWAYS beat technology. Which
is why Windows (with all of it's faults, which are legion) will continue
to dominate. I don't revel in this, I merely accept it. I've made sure
over the past year to stay current with the Windows market to ensure my future
billability. The OS/2 market just isn't big enough yet, and I doubt that it
will grow out of it's niche.
James A. Robertson
----------
Disclaimer: The views expressed above are mine, and do not necessarily
reflect those of my employer or any affiliated companies
Hugh JE Davies
[snip]
>
>Why are there no usable applications for any major UNIX platform? (I'm not
>calling NeXTStep UNIX).
>
Interleaf. Framemaker/builder. (I'm sitting in front of a Sun running both, right now.)
>
>UNIX is no panacea. Grow up and join the real world.
>
Sigh. Your bosses must be really proud of you. You have "IBM" through you like a
stick of rock.
Regards,
Hugh.
------------------------------------------------------------------
Huge...@rx.xerox.com Rank Xerox Technical Centre, WGC, UK.
I don't speak for Xerox, nor they for me.
The road to Paradise is through Intercourse.
Mike Dahmus
>In article <CC9Mz...@hawnews.watson.ibm.com> mi...@vnet.ibm.com (Mike Dahmus) writes:
>>
>>OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
>>box with X (and I've tried plenty) that could perform anywhere nearly as well
>>as OS/2 (the only one that comes close is linux, and then all you have to do
>>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>>function). If IBM had just produced Yet Another Slightly Incompatible Version
>>Of UNIX (like every other company under the sun), the performance advantage
>>would have been impossible. Likewise, if IBM had built their user interface on
>>top of X, they would have suffered the same performance and usability problems
>>as all the other UNIXes.
>>
>Gee, I don't know about that. Having been working between an OS/2 2.1 box
>(IBM PS/2 model 90 486) and a Sun Sparc 10, I had a MUCH easier time on the
>Sparc. When I closed my windows, they didn't wander off into a folder
>somewhere, they went to the bottom (or side) of the screen - just like
>most other windowing environments I've ever used (Mac, Windows).....
Gee, some consultant you are, mixing up "minimize" and "close". (If you "close"
a window on the Mac or Windows, it goes away, just like in OS/2).
>Yes, I know that OS/2 is configurable. But I had WORK to do, and did not have
>time to spend twiddling an interface. Configurability is NOT what most
>users want, they want familiarity. The interface should be close to
>what is offered under Mac or Windows simply because that's what most people
>are used to.
The OS/2 interface closer to the Macintosh than anything else. The difference
is that if you want to make it act like Windows (with abstract program
references) or like twm (with a big desktop menu), you can.
>The OS/2 market just isn't big enough yet, and I doubt that it
>will grow out of it's niche.
3 million is a niche? Care to explain that?
------
Mike Dahmus Internet: mi...@vnet.ibm.com
Pen for OS/2 Development, IBM PSP IBM: mi...@schleppo.bocaraton.ibm.com
Disclaimer: Not an official IBM spokesman IBM Vnet: MDAHMUS at BOCA
William F. Pemberton
Malcolm L. Carlock <ma...@pyramid.unr.edu> wrote:
>If IBM/Microsoft were really interested in providing the best multitasking
>environment for their users, they could have provided the same, and probably
>better, services by throwing their considerable resources into providing
>The Best Desktop Unix Ever (they might even have succeeded.) Any "new"
>features that OS/2 or Windows provide could easily have been added to that
>Unix, as new features have always been more or less easily added to Unix.
>
One of the major advantages of OS/2 (and probably headaches in developing
it) is that it still runs DOS and Windows programs. I don't think
a Unix for the PC could do that well. "So what?" many people will say.
Well, that's a big deal. Users don't want to buy all new applications
along with the upgrade.
>
>Users, businesses and institutions who need real power, flexbility, and the
>ability to get real work done will, after a bit of research, usually choose
>Unix. Users who continue to think of IBM and MS as their "buddies" in the
>desktop world will continue paying more (at the middle and high end) for
>a combination of less capable hardware (Intel chips in PC's still slower
>than most RISC processors) and OS software (e.g., lack of multiuser support
>in OS/2, no real virtual memory support in MS-Windows -- else how can a
>memory violation by an application necessitate a reboot??)
>
Is multiuser support REALLY that important to most of the PC users? I
don't believe that it is.
And is Unix really *THE* best OS? Is it impossible to produce an OS that
is not a Unix descendant that is good? Obviously, you think not. Unix is
good, but I don't believe that an OS must be a Unix descendant to be good.
--
Bill
Lars R{der Clausen
[...rant & rave about UNIX vs the half OS...]
>>The OS/2 market just isn't big enough yet, and I doubt that it
>>will grow out of it's niche.
>3 million is a niche? Care to explain that?
Ah! So the Amiga is not a niche, either?
>------
>Mike Dahmus Internet: mi...@vnet.ibm.com
>Pen for OS/2 Development, IBM PSP IBM: mi...@schleppo.bocaraton.ibm.com
>Disclaimer: Not an official IBM spokesman IBM Vnet: MDAHMUS at BOCA
-Lars
--
managing security with all the grace of a Brontosaurus trying to type.
..managing security with all the grace of a Brontosaurus trying to type.
-Michel E. Kabay, about security in 'Jurassic Park' (and elsewhere)
Faisal Nameer Jawdat
> In article 1E...@hawnews.watson.ibm.com,
> mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
> >Why are there no usable applications for any major UNIX platform? (I'm not
> >calling NeXTStep UNIX).
>
> Interleaf. Framemaker/builder. (I'm sitting in front of a Sun running
> both, right now.)
WordPerfect, WingZ (well, it's sort of :) ), mail, rn, trn, xrn,
gnu-emacs (what's usable? it does everything *I* want it to do (edits
text, reads mail and news, ftp's, has a directory viewing system, is
extensible, etc...)), TeX (how many books are typeset in Microsoft Word,
I ask you...), etc... You should see the stuff we have running at CMU
under unix.
-- faisal jawdat | "If you can spend a perfectly useless
email: fj...@andrew.cmu.edu | afternoon in a perfectly useless
if mail handler barfs | manner, you have learned how to live."
fj...@andrew.cmu.edu | -Lin Yutang
Dave Brown
>In article <1993Aug23.2...@unlv.edu>,
>Malcolm L. Carlock <ma...@pyramid.unr.edu> wrote:
>>If IBM/Microsoft were really interested in providing the best multitasking
>>environment for their users, they could have provided the same, and probably
>>better, services by throwing their considerable resources into providing
>>The Best Desktop Unix Ever (they might even have succeeded.) Any "new"
>>features that OS/2 or Windows provide could easily have been added to that
>>Unix, as new features have always been more or less easily added to Unix.
>>
>
>One of the major advantages of OS/2 (and probably headaches in developing
>it) is that it still runs DOS and Windows programs. I don't think
>a Unix for the PC could do that well. "So what?" many people will say.
>Well, that's a big deal. Users don't want to buy all new applications
>along with the upgrade.
% Linux-bigot mode on
Haven't you ever heard of WABI for Solaris? Runs Windows apps
*better* than Windows does (but then again, so does OS/2). And
waiting in the wings to go along with Dosemu for Linux is WINE
which'll do the same thing. Even *more* reasons to switch to
Linux....
>Is multiuser support REALLY that important to most of the PC users? I
>don't believe that it is.
Guess not, but it comes in kinda handy.... :-)
>And is Unix really *THE* best OS? Is it impossible to produce an OS that
>is not a Unix descendant that is good? Obviously, you think not. Unix is
>good, but I don't believe that an OS must be a Unix descendant to be good.
It is of *course* possible to make an OS that is not a Unix descendent
that is good. Look at the Amiga. For regular everyday users, look at
System 7. Or even *gasp!* VMS...(even though VMS uses up so much
CPU on the operating system).
--
Dave Brown -- dagb...@napier.uwaterloo.ca -- (416) 669-5370
"No vision of God and heaven ever experienced by the most exalted prophet
can, in my opinion, match the vision of the universe as seen by Newton
or Einstein" -- Isaac Asimov
Antonio Vasconcelos
:
: It is of *course* possible to make an OS that is not a Unix descendent
: that is good. Look at the Amiga. For regular everyday users, look at
Not unix descendent ? Are you sure ? Then I wander why there are so
many unix C-sources in the Net that have a "make amiga" option.
I'm a 3 days-old Amiga user, and one of the resons (apart from the
games, of course) because I choose it instead of a Sega Megadrive was
that unix lookalikeliness (sort of). I want games, YES, but I'd like
to have at hand something related to my beloved unix when I choose to
program it.
--
__ ____________ Regards, Vasco
/ /\ /\ _________\ ..........................
/ / \ \ \ \___ __ / Antonio Vasconcelos
/ / /\ \ \ \ \ / / / Bolsa de Valores de Lisboa
/ / /\ \ \ \ \ \/ / / [Lisbon Stock Exchange]
/ /_/__\ \ \ \ \/ / / R. dos Fanqueiros, 10
/________\ \ \ \ / / 1100 LISBOA
\___________\/ \/_/va...@bvl.pt PORTUGAL
Jim Frost
>* Does anyone remember "Computer Language" magazine's "Multitasking" issue
> from a few years ago? Instead of something useful and informative to
> people who use computer languages professionally, like an issue comparing
> multitasking programming in a number of environments (Unix, VMS, VM, MVS,
> Amiga, etc.), it was essentially ENTIRELY devoted to hacks for doing non-
> preemptive (!) multitasking under DOS.
> Does this magazine still exist? If so, hopefully they've renamed it to
> "Computer Language, but only under MS-DOS"
>-----------------------------------------------------------------------------
I remember the issue. I believe "Computer Language" has indeed died,
being replaced by "Software Development" (or some such; a freebie
magazine now) with its subscriptions rolled over to "Dr. Dobb's
Journal" subscriptions. This happened two, maybe three months ago.
I bought the magazine because occasionally there are real gems in it,
like the quantization algorithm in their Aug 1990 issue on graphics.
That's been rather rare, though.
jim
Jim Frost
>>The OS/2 market just isn't big enough yet, and I doubt that it
>>will grow out of it's niche.
>3 million is a niche? Care to explain that?
Given the overall PC market of 115 million PC's (according to recent
IDC figures), 3 million is paltry. UNIX has more than ten million of
them (around 10%, generally used as central systems in corporate or
manufacturing environments), and the PC UNIX is considered a "niche"
market (even by you).
Truly.
jim frost
ji...@centerline.com
Jim Frost
>And your whole posting is full of conspiratorial tripe. The free market works,
>buddy - get used to it.
>OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
>box with X (and I've tried plenty) that could perform anywhere nearly as well
>as OS/2 (the only one that comes close is linux, and then all you have to do
>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>function).
Unless you add "on Intel hardware" that's not even close to true. I'd
put an SGI POWER system running X over any OS/2 box you can find. If
I want more UNIX or X performance all it takes is more dollars, all
the way up to the fastest available systems in the world.
If you've seen performance problems relative to competitive operating
systems on the same hardware it's because of the particular UNIX
implementation. I keep telling you that there are other ones -- quite
a few on the Intel platforms, and some are far better than average.
BSDI's implementation is cheap and very, very fast. Linux is *free*
and reasonably fast.
If you want a nice, usable graphical shell for a UNIX box, the only
thing you seem to really be griping about, you can buy one. There are
quite a few other than NeXT's products, believe it or not.
(A pretty shell is what we used to call "fluff": something that looks
pretty and sells well but doesn't provide any additional
functionality. You disagree, I suspect, but there's nothing you can
do with WPS that I can't do from a CLI, and a lot I can do from a CLI
that you can't do at all with WPS so to me it's just fluff.)
>Why are there no usable applications for any major UNIX platform? (I'm not
>calling NeXTStep UNIX).
Ever hear of Frame? Oracle? Interleaf? Wingz? Ingres?
WordPerfect? Lotus 1-2-3? What about manufacturing controllers?
Accounting systems (ALCIE, etc)?
What about built-in functionality like the ability to be a
file-server, network router, etc?
Most of the money-making applications on UNIX aren't targetted towards
a single end-user, so you don't see them advertised in PC-World.
There certainly are very usable applications if you want them.
>UNIX is no panacea. Grow up and join the real world.
Depends on what you want to do. OS/2 is fine for running your little
bitty spreadsheet or wordprocessor. That's great for a lot of people,
but you just can't support a very large business on that. And God
help you if you need a zero-downtime system or a fast WAN. You can't
even buy OS/2 boxes fast enough to make good enterprise file or
database servers. But it does do a fair job of handling one or two
smallish graphical applications, and even as a terminal to connect to
more capable systems (which is what IBM likes to see!).
UNIX has its ups and downs, no doubt about it, but it's terribly
flexible and scales amazingly well. Certainly it's not the
be-all-to-end-all of desktop systems, but it's not bad even in *that*
environment. OS/2, on the other hand, is an end-user-point product
that's almost totally unscalable. There's usefulness in that, but
it's not for everyone. Even as a end-user product it's not very good;
supporting a cluster of them is a royal pain in the ass.
jim frost
ji...@centerline.com
Dave Brown
>dagb...@undergrad.math.uwaterloo.ca (Dave Brown) writes:
>:
>: It is of *course* possible to make an OS that is not a Unix descendent
>: that is good. Look at the Amiga. For regular everyday users, look at
>
>Not unix descendent ? Are you sure ? Then I wander why there are so
>many unix C-sources in the Net that have a "make amiga" option.
Because it's a lot easier to port things to the Amiga than it is to
port stuff to DOS....DOS is more Unix-descended than the Amiga
(question marks and stars in wildcards) and most programs that just
use the standard C library without doing any funky Unix-dependent
stuff are trivial[*] to port anywhere.
>I'm a 3 days-old Amiga user, and one of the resons (apart from the
>games, of course) because I choose it instead of a Sega Megadrive was
>that unix lookalikeliness (sort of). I want games, YES, but I'd like
>to have at hand something related to my beloved unix when I choose to
>program it.
Best o' luck to you. The Amiga's the most fun I've ever had
programming a box in C, but it ain't Unix. main(){while(;;)fork();} don't
work for example (unless some fool[**] who wants the Amiga to look exactly
like a unix box made it).
When I use my Amiga, I celebrate the fact that it *isn't* a Unix box.
I get so bored when every new computer looks exactly the same--it's so
monotonous.
But enjoy your Amiga though...it's a fantastic little box. I wouldn't
give mine up for anything. Except, perhaps, another Amiga.
[*] *guffaw*
[**] The guys who ported GCC to the Amiga, for example
[gigantic .sig deleted]
--
Dave Brown -- dagb...@napier.uwaterloo.ca -- (416) 669-5370
"I am quoted in Dave Brown's .signature twice."
--"Polar Bear" Morse
Tom Lane
> ma...@pyramid.unr.edu (Malcolm L. Carlock) writes:
>> Does this magazine still exist? If so, hopefully they've renamed it to
>> "Computer Language, but only under MS-DOS"
>
> being replaced by "Software Development" (or some such; a freebie
> magazine now)
Yeah, CL is dead. I had a freebie sub to it, now replaced by a freebie sub
to SD. When that runs out, I won't bother to renew it :-(
CL had occasionally decent articles; the main thing that made it worth
reading was P.J. Plauger's monthly column. SD has dropped Plauger.
SD's target customer appears to be an MBA who knows less than zero about
software but has been appointed to run a software development group.
Looking for another source of Plauger columns...
regards, tom lane
Lars R{der Clausen
>dagb...@undergrad.math.uwaterloo.ca (Dave Brown) writes:
>:
>: It is of *course* possible to make an OS that is not a Unix descendent
>: that is good. Look at the Amiga. For regular everyday users, look at
>Not unix descendent ? Are you sure ? Then I wander why there are so
>many unix C-sources in the Net that have a "make amiga" option.
Not UNIX descendent. UNIX-inspired, maybe, esp. in its CLI. But the OS in
itself is far more compact than UNIX, and much different. No memory
protection, no virtual memory, but easy messages passing between tasks
and a built-in GUI. And small. I still think the possibility of protecting
memory should have been there, at least from the A3000.
The 'make amiga' option is probably there because the Amiga is
sufficiently UNIX-like to allow standard C, even with file accesses,
to compile easily. Or an Amiga user has #def'd it to be Amiga
compilable.
>I'm a 3 days-old Amiga user, and one of the resons (apart from the
>games, of course) because I choose it instead of a Sega Megadrive was
>that unix lookalikeliness (sort of). I want games, YES, but I'd like
>to have at hand something related to my beloved unix when I choose to
>program it.
Then I bet you want csh, gcc, emacs...and a large harddisk!
>--
> __ ____________ Regards, Vasco
> / /\ /\ _________\ ..........................
> / / \ \ \ \___ __ / Antonio Vasconcelos
> / / /\ \ \ \ \ / / / Bolsa de Valores de Lisboa
> / / /\ \ \ \ \ \/ / / [Lisbon Stock Exchange]
> / /_/__\ \ \ \ \/ / / R. dos Fanqueiros, 10
> /________\ \ \ \ / / 1100 LISBOA
> \___________\/ \/_/va...@bvl.pt PORTUGAL
And a smaller signature.
Antonio Vasconcelos
:
: Because it's a lot easier to port things to the Amiga than it is to
: port stuff to DOS....DOS is more Unix-descended than the Amiga
: (question marks and stars in wildcards) and most programs that just
Gosh... I was *shocked* when yesterday I found out that * in an Amiga
does not mean all files but keyboard/console.
And ':' instead of '/' and "//" instead ".." is not the way God want
it to be :-) I still have no idea about the Amiga wildcards (out of
time to read, I've been very busy plaing Arkanoid II :-)
: Best o' luck to you. The Amiga's the most fun I've ever had
: programming a box in C, but it ain't Unix. main(){while(;;)fork();} don't
: work for example (unless some fool[**] who wants the Amiga to look exactly
: like a unix box made it).
Have to got GCC and find out what happen... Amiga UNIX crash in 2
seconds...
: When I use my Amiga, I celebrate the fact that it *isn't* a Unix box.
: I get so bored when every new computer looks exactly the same--it's so
: monotonous.
There you may be right.
: But enjoy your Amiga though...it's a fantastic little box. I wouldn't
: give mine up for anything. Except, perhaps, another Amiga.
Only if the games get to gigantic and slow, like now in the PC.
However I'd like that they stop thinking that no one have an hard
disk. I spend money buying an HD just to find out that unlike the PC
most of the games must be booted from floppy...
: [gigantic .sig deleted]
You are right, please enjoy my new .Sig :-)
--
Regards
Vasco
---------------------------------------------------------------
Antonio Vasconcelos at The Lisbon Stock Exchange - va...@bvl.pt
---------------------------------------------------------------
I DO NOT speak for BVL TANSTAAFL
Mike Dahmus
>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>>And your whole posting is full of conspiratorial tripe. The free market works,
>>buddy - get used to it.
>
>>OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
>>box with X (and I've tried plenty) that could perform anywhere nearly as well
>>as OS/2 (the only one that comes close is linux, and then all you have to do
>>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>>function).
>
>Unless you add "on Intel hardware" that's not even close to true. I'd
>put an SGI POWER system running X over any OS/2 box you can find. If
>I want more UNIX or X performance all it takes is more dollars, all
>the way up to the fastest available systems in the world.
>
>If you've seen performance problems relative to competitive operating
>systems on the same hardware it's because of the particular UNIX
>implementation. I keep telling you that there are other ones -- quite
>a few on the Intel platforms, and some are far better than average.
>BSDI's implementation is cheap and very, very fast. Linux is *free*
>and reasonably fast.
Jim, I should have said "on the same hardware". Sorry - and you're right
about OS/2 not being scalable past a Pentium at this point.
But I *was* comparing OS/2 to linux, by the way. And I've only heard you say
that BSD is quicker than linux - most people seem to think the opposite is
true.
>If you want a nice, usable graphical shell for a UNIX box, the only
>thing you seem to really be griping about, you can buy one. There are
>quite a few other than NeXT's products, believe it or not.
Untrue. The only one worth consideration for someone who requires an OO shell
is NeXTStep.
>(A pretty shell is what we used to call "fluff": something that looks
>pretty and sells well but doesn't provide any additional
>functionality. You disagree, I suspect, but there's nothing you can
>do with WPS that I can't do from a CLI, and a lot I can do from a CLI
>that you can't do at all with WPS so to me it's just fluff.)
This is also untrue - you can't start an application from a document in a CLI.
You can't determine what application(s) should/can be run on a particular
document in a CLI.
>>Why are there no usable applications for any major UNIX platform? (I'm not
>>calling NeXTStep UNIX).
>
>Ever hear of Frame? Oracle? Interleaf? Wingz? Ingres?
>WordPerfect? Lotus 1-2-3? What about manufacturing controllers?
>Accounting systems (ALCIE, etc)?
Usable, usable, usable. There was a really big, important reason I put that
word in front of the word "applications", Jim. Even the few major commercial
apps on UNIX platforms are typically behind in function from their Windows
counterparts, and often have disjointed interfaces to boot. I think my main
beef here is that there isn't any standardization at ALL on interfaces. Most
Windows apps act a certain way; most PM apps act a certain way; most Mac apps
act a certain way; most UNIX apps act totally differently.
>>UNIX is no panacea. Grow up and join the real world.
>
>Depends on what you want to do. OS/2 is fine for running your little
>bitty spreadsheet or wordprocessor. That's great for a lot of people,
>but you just can't support a very large business on that. And God
>help you if you need a zero-downtime system or a fast WAN. You can't
>even buy OS/2 boxes fast enough to make good enterprise file or
>database servers. But it does do a fair job of handling one or two
>smallish graphical applications, and even as a terminal to connect to
>more capable systems (which is what IBM likes to see!).
I think you need to start reading some non-UNIX trade press. OS/2 was already
in a lot of big businesses with the 1.x versions. OS/2 1.3 qualifies very
easily as a "zero-downtime system" (I'm not going to make this claim for 2.x,
at least not yet). There are hundreds, if not thousands of apps written by
businesses that run on OS/2 1.x, including a lot of airline systems, banking
systems, and other large operations.
------
Mike Dahmus Internet: mi...@vnet.ibm.com
Mike Dahmus
Jim, recent figures indicate that OS/2 is now selling more copies than all
varieties of Intel UNIX put together. I suspect that some of your figures
above are inflated, but even if not, OS/2 is certainly becoming a bigger
"niche" than PC UNIX.
Vadim Antonov
: I think you need to start reading some non-UNIX trade press.
I think you need to stop reading *any* trade press. They *never*
printed anything resembling the truth. Only naive newbies believe
to the sweet words of people trying to sell them complete shit
and make losers to believe they're in heaven.
[UNIX trade press is as full of shit as any other, and normal
people do not read it. Avoiding trade press helps to save trees,
too].
--vadim
E M Bergersjo
In article <1993Aug27.1...@bvl.pt>, va...@bvl.pt (Antonio Vasconcelos) writes:
|> Gosh... I was *shocked* when yesterday I found out that * in an Amiga
|> does not mean all files but keyboard/console.
|> And ':' instead of '/' and "//" instead ".." is not the way God want
|> it to be :-) I still have no idea about the Amiga wildcards (out of
|> time to read, I've been very busy plaing Arkanoid II :-)
Just like with Unix you can use different shells on the Amiga. One of them, cshell, works exactly as its Unix equivalent. Moreover, you _can_ use * as a wildcard with the normal Amiga commands if you want to, but you have to select it when you start the system. The normal wildcards are quite powerful. # means "any number" and ? "any character", hence #? corresponds to *. You can also use brackets and many other things. Not many systems allow you to write things like #a?b#?(c|d) and end up with something me
aningful! :-)
-------------------------------------------------------------------------------
Erik Bergersjo "I don't see much sense in that," said Rabbit.
"No," said Pooh humbly, "there isn't."
e...@doc.ic.ac.uk "But there was _going_ to be when I began it."
d9e...@dtek.chalmers.se "It's just that something happened to it on the way."
-------------------------------------------------------------------------------
DoN. Nichols
[ ... ]
>Yeah, CL is dead. I had a freebie sub to it, now replaced by a freebie sub
>to SD. When that runs out, I won't bother to renew it :-(
>
>CL had occasionally decent articles; the main thing that made it worth
>reading was P.J. Plauger's monthly column. SD has dropped Plauger.
>SD's target customer appears to be an MBA who knows less than zero about
>software but has been appointed to run a software development group.
>
>Looking for another source of Plauger columns...
I also miss the colums by Stan Kelly-Bootle, which were dropped
earlier, and their absence was the reason why I didn't bother to renew. I
thought that I was getting an unusual level of "please renew" literature and
calls. This explains it.
I, also, was dissapointed by the concentration on MS-DOS.
--
Email: <dnic...@d-and-d.com> | ...!uunet!ceilidh!dnichols
<dnic...@ceilidh.beartrack.com>
Donald Nichols (DoN.) | Voice (Days): (703) 704-2280 (Eves): (703) 938-4564
--- Black Holes are where God is dividing by zero ---
Thor Lancelot Simon
Antonio Vasconcelos <va...@bvl.pt> wrote:
>dagb...@undergrad.math.uwaterloo.ca (Dave Brown) writes:
>:
>: It is of *course* possible to make an OS that is not a Unix descendent
>: that is good. Look at the Amiga. For regular everyday users, look at
>
>Not unix descendent ? Are you sure ? Then I wander why there are so
Yes, he's quite justifiably sure. AmigaDOS has no unix in its heritage.
>many unix C-sources in the Net that have a "make amiga" option.
Because Commodore sells SVR4 Unix for the top-end Amiga models, and because
AmigaDOS is flexible enough in any case that lots of Unix tools are reasonably
easy to port.
>I'm a 3 days-old Amiga user, and one of the resons (apart from the
>games, of course) because I choose it instead of a Sega Megadrive was
>that unix lookalikeliness (sort of). I want games, YES, but I'd like
>to have at hand something related to my beloved unix when I choose to
>program it.
Well, you don't. AmigaDOS has some nifty features, though.
--
Thor Lancelot Simon t...@panix.COM
When you're not feeling holy, your loneliness says that you've sinned...
Thomas Spencer IV
mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes in
<CC9Mz...@hawnews.watson.ibm.com> on 24 Aug 93 13:33:26 GMT
<Stuff about which vendor is locking who into there product deleted>
>OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
>box with X (and I've tried plenty) that could perform anywhere nearly as well
>as OS/2 (the only one that comes close is linux, and then all you have to do
>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>function).
HA! HA! HA! HA! HA! WOOOOOOOOO!!!!
Please excuse me while I get up off of the floor. OS/2 running X versus
a Unix box running X? OS/2 is faster? What clock have you been watching?
I've been running almost identical systems, one with OS/2 with X and one
Unix with X, and the OS/2 platform is so slow that I can't use it as an X
station. If you are curious, the Unix box has both a slower hard drive
and a slower video card. (OS/2 - a 340 meg 12ms IDE WD drive, ET4000 video;
Unix- 380meg 16ms Maxtor ESDI, OEM paradise SVGA card)
> If IBM had just produced Yet Another Slightly Incompatible Version
>Of UNIX (like every other company under the sun), the performance advantage
>would have been impossible. Likewise, if IBM had built their user interface on
>top of X, they would have suffered the same performance and usability problems
>as all the other UNIXes.
So now you can only process the graphics output of the local system
and only this system. Granted, OS/2 PM is faster than X, but then it doesn't
do what X does. Start comparing products that do the same thing and
you will find OS/2 is a resource hog and much slower.
>Why are there no usable applications for any major UNIX platform? (I'm not
>calling NeXTStep UNIX).
Gee!!! I've been using MS-Word for Unix for five years with all of
the same functionality as the DOS version, I've got foxpro running and
a Lotus 1-2-3 look alike. I also deal with _REAL_ apps like multi-user
accounting and Order Processing on Unix much more easily than bolting on
an ugly bag of networking to a DOS or OS/2 system. At about half the
startup cost.
>UNIX is no panacea. Grow up and join the real world.
Why don't you take your own advice. Go take a good look where the
information systems money is being spent. Not on OS/2. Unix is winning
on your big data systems. NT _might_ become a threat in future releases,
but your DP people are all scrambling around to learn Unix because that is
the OS that the VM and MVS sites are moving to if they are moving at all.
>------
>Mike Dahmus Internet: mi...@vnet.ibm.com
>MVP Development, IBM PSP IBM: mi...@schleppo.bocaraton.ibm.com, MDAHMUS at BOCA
>Disclaimer: Not an official IBM spokesman
PS- go look at the RS/6000.
tom
Thomas Spencer IV
U20...@HNYKUN11.URC.KUN.NL writes in <16C357BC...@HNYKUN11.URC.KUN.NL>
^^^^^
That is if the OS is going to run, to say nothing of any applications.
Open your first app in 8meg and the hard drive starts to do the tango.
You need 16meg to get anything more than trivial applications to work
without very long execution delays due to paging.
>Paul.
tom
Giles Todd
> You can't determine what application(s) should/can be run on a particular
> document in a CLI.
<*Against his better judgement, Giles joins (very briefly) in an OS wars
thread.*>
Check out UNIX document files which start "#! <something>". The
"<something>" specifies the application which the kernel will invoke to
run on the document. All you have to do at the CLI is type the
document's name. The kernel will invoke the appropriate application.
The main difference between this and a proprietary GUI shell such as
those under discussion is, of course, that the mechanism is explicit and
there for all to see. This explicitness may interfere with some
people's illusions of how clever their (to be defended at all costs) OS
is.
Giles.
--
Giles Todd g...@rundart.demon.co.uk (Internet)
Rundart Ltd Demon Internet
Voice: +44 925 414696 PGP 2.x public key available.
Thomas Spencer IV
wf...@holmes.acc.Virginia.EDU (William F. Pemberton) writes in
<CCBK7...@murdoch.acc.Virginia.EDU>
<stuff deleted that doesn't need to be repeated yet again>
>One of the major advantages of OS/2 (and probably headaches in developing
>it) is that it still runs DOS and Windows programs. I don't think
>a Unix for the PC could do that well. "So what?" many people will say.
>Well, that's a big deal. Users don't want to buy all new applications
>along with the upgrade.
It is very obvious that you haven't played with DOS merge or any of the other
DOS under UNIX packages. On an intel box you can actually get better than
95% of the original performance of a DOS/windows box. Using a non-Intel
platform you do suffer from having to emulate an entire CPU instead of just
OS calls, but it is still somewhat usable.
<stuff on a subject covered below deleted>
>Is multiuser support REALLY that important to most of the PC users? I
>don't believe that it is.
Then why are all of these networks being put in? Because they are
a good thing? No, because they deliver services that businesses grew
to rely on that multi-user systems have had since multi-user systems
were first created. They
a) Share data easily.
b) Share resources easily. (one and the same thing really)
c) Provide a greater level of system security.
d) Are more reliable.
e) Get more work done than single user/stand alone systems.
While a kid at home who wants to play Wing Commander may not care about
these things, anyone who uses computers to help make money sure would.
>And is Unix really *THE* best OS? Is it impossible to produce an OS that
>is not a Unix descendant that is good? Obviously, you think not. Unix is
>good, but I don't believe that an OS must be a Unix descendant to be good.
While I can't speak for the person this was addressed to, I can say that
there are other non-Unix derived systems that are both very useable and
very good at what they do. I may not like them, but they do command
my gruding respect. VMS, MVS and VM are all very good when used for the
right reasons.
>--
>Bill
tom
Andrew Bray
>In <25j7t6$p...@wcap.centerline.com>, ji...@centerline.com (Jim Frost) writes:
>>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>>>>The OS/2 market just isn't big enough yet, and I doubt that it
>>>>will grow out of it's niche.
>>
>>>3 million is a niche? Care to explain that?
>>
>>Given the overall PC market of 115 million PC's (according to recent
>>IDC figures), 3 million is paltry. UNIX has more than ten million of
>>them (around 10%, generally used as central systems in corporate or
>>manufacturing environments), and the PC UNIX is considered a "niche"
>>market (even by you).
>
>Jim, recent figures indicate that OS/2 is now selling more copies than all
>varieties of Intel UNIX put together. I suspect that some of your figures
>above are inflated, but even if not, OS/2 is certainly becoming a bigger
>"niche" than PC UNIX.
You must not forget the recent phenomenon of free UNIX for PC's, these
will never appear in sales figures, but there are a lot of such systems
around.
How many free UNIX systems there are around (probably nowhere near 10
million yet :-) is probably impossible to gauge, but must not be
ignored when engaging in OS popularity flame wars.
Regards,
Andy
Antonio Vasconcelos
:
: Not UNIX descendent. UNIX-inspired, maybe, esp. in its CLI. But the OS in
: itself is far more compact than UNIX, and much different. No memory
: protection, no virtual memory, but easy messages passing between tasks
That part about no memory protection is a bit stupid for a
multitasking OS. About virtual memory, well, we can't have evrything,
and most of the Amigas out there don't have an HDD anyway.
: and a built-in GUI. And small. I still think the possibility of protecting
: memory should have been there, at least from the A3000.
Should be there even for the A500.
: Then I bet you want csh, gcc, emacs...and a large harddisk!
Nope, I want ksh, vi and gcc.
I can't do nothing about the hdd, I have a 120M and that's all that
will have for that machine (that's me talking now :-)
: > /________\ \ \ \ / / 1100 LISBOA
: > \___________\/ \/_/va...@bvl.pt PORTUGAL
:
: And a smaller signature.
Done, sorry...
--
Regards,
Antonio Vasconcelos
:
: >many unix C-sources in the Net that have a "make amiga" option.
:
: Because Commodore sells SVR4 Unix for the top-end Amiga models, and because
: AmigaDOS is flexible enough in any case that lots of Unix tools are reasonably
: easy to port.
Gosh. Never thought that the "make amiga" could be for the Unix OS.
And I think most of GNU have been ported. (elvis ???)
: >that unix lookalikeliness (sort of). I want games, YES, but I'd like
: >to have at hand something related to my beloved unix when I choose to
: >program it.
:
: Well, you don't. AmigaDOS has some nifty features, though.
And some sad bugs/features. Like no memory protection.
Anyway, I've written hello.c allready, not an easy thing to do with
the Amiga connected to a PAL TV via RF, (hugly, hugly, hugly!!!)
--
Regards,
====================================================================
Antonio Vasconcelos at The Lisbon Stock Exchange - va...@bvl.pt
"I do NOT speak for BVL" NeXTmail - ro...@bvl.pt
Andrew Bulhak
:
: I remember the issue. I believe "Computer Language" has indeed died,
: being replaced by "Software Development" (or some such; a freebie
: magazine now) with its subscriptions rolled over to "Dr. Dobb's
: Journal" subscriptions. This happened two, maybe three months ago.
Who's going to do the Jolt Cola best product competition now?
-- A.
Signature has been
retroactively moderated (tm).
Andrew Bulhak
: In <25j7t6$p...@wcap.centerline.com>, ji...@centerline.com (Jim Frost) writes:
: >mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
: >>>The OS/2 market just isn't big enough yet, and I doubt that it
: >>>will grow out of it's niche.
: >
: >>3 million is a niche? Care to explain that?
: >
: >Given the overall PC market of 115 million PC's (according to recent
: >IDC figures), 3 million is paltry. UNIX has more than ten million of
: >them (around 10%, generally used as central systems in corporate or
: >manufacturing environments), and the PC UNIX is considered a "niche"
: >market (even by you).
:
: Jim, recent figures indicate that OS/2 is now selling more copies than all
: varieties of Intel UNIX put together. I suspect that some of your figures
: above are inflated, but even if not, OS/2 is certainly becoming a bigger
: "niche" than PC UNIX.
Which does not mean that OS/2 is burying UNIX. Only a minority of UNIX
systems run on PCs, as UNIX is portable and UNIX users have a choice of
much better platforms. OS/2 (in its present incarnation) is bound to the
Intel platform, with its crufty 8MHz bus and backward-compatible architecture.
Some day in the not-too-distant future, when Intel have stretched the laws
of physics to their limit in order to get their latest CISC elephant to fly,
MS-DOS has gone the way of CP/M and the Alpha (or some other RISC chip)
is widely accepted, this architecture will join the S-100 bus and the
Apple II in the museum; it is hoped that we will all be running NeXTstep
on our 250MHz Alpha machines...
iv...@cc.usu.edu
> Instead, it seems arguable that both vendors have copped out and followed
> the path of trying to lock users into relatively mediocre products provided
> by a single vendor, while as much as possible attempting to ignore or obscure
> (with the help of "computing" journals that are in far too many cases
> IBM/MS-sycophantic to the point of inducing nausea*) the existence of
> alternatives. This approach does little to advance the state of the art or
> provide the best solutions to users.
Many of us thing that Unix is mediocre. Ask me about the >100 millisecond
interrupt latencies I'm getting under OSF/1 on the hot new Alpha boxes as
opposed to the interrupt latencies of the same box under VMS.
> Does this magazine still exist? If so, hopefully they've renamed it to
> "Computer Language, but only under MS-DOS"
Yes, the magazine does still exist. As far as I know, they haven't renamed it,
although the last time I read it, it seemed more like "Computer Language, as
long as that language is C under MS-DOS." Ah, the heady days of the Obscure
Language of the Month...
Roger Ivie
iv...@cc.usu.edu
Benjamin Ketcham
[...]
>main(){while(;;)fork();} don't
>work for example
Shouldn't that be "don't work while example"?
[...]
--ben
Here, Life. Nice boy. C'mere. Goood Life. Niiiice Life. Here, boy.
P-Jay Nield
> Which does not mean that OS/2 is burying UNIX. Only a minority of UNIX
> systems run on PCs, as UNIX is portable and UNIX users have a choice of
> much better platforms. OS/2 (in its present incarnation) is bound to the
> Intel platform, with its crufty 8MHz bus and backward-compatible architecture.
>
Hey, Have you heard of _MICRO_CHANNEL_??
This eats EISA!!!
You should sit down in front of a REAL IBM sometime. PS/2 variety.
(not that I have done, but the microchannel machines simply _don't_ have the
8MHz bottle neck - the bus goes at the processor's speed, whatever that is for
a particular machine)
(Well, thats what I'm told by an OS/2 V2 user friend of mine)
/_______________________________
"Well, the way I see it, logic is just a way of being ignorant by numbers."
Didactylus, from Small Gods, by Terry Pratchett
Peter-John Nield nie...@elec.canterbury.ac.nz ZL2...@ZL3AC.CHC.NZL.OC
_________________________________
/
Geoff Mccaughan
>That part about no memory protection is a bit stupid for a
>multitasking OS.
Depends. If you're running applications which are well behaved, it's not a
problem. If you want to run buggy software well... The primary limitation is
that it doesn't pay to develop software on a production machine, but an
Amiga is cheap enough that it's not a problem to buy one for development
anyway.
> About virtual memory, well, we can't have evrything,
>and most of the Amigas out there don't have an HDD anyway.
You can have virtual memory. I think you might find there are more machines
out there with hard drives than you think.
--
Geoff, Sysop Equinox (equinox.gen.nz) +64 (3) 3854406 [6 Lines]
"If you have to run heating in winter, you don't own enough computers."
Vote SPQR Ski Nix Olympica Freedom for Axolotls
Johan Danielsson
[lots of stuff deleted]
> Apple II in the museum; it is hoped that we will all be running NeXTstep
> on our 250MHz Alpha machines...
May the high gods of Silicon help us! Id rather be W/O fingers then be
using NeXTstep!
Alpha seems to be too expensive; it will be interesting to see what
comes out of PowerPC though.
> -- A.
>
> Signature has been
> retroactively moderated (tm).
--
| d91...@nada.kth.se | Drag-n-Drop-dead
| Johan Danielsson |
Rui Pedro Mendes Salgueiro
: Which does not mean that OS/2 is burying UNIX. Only a minority of UNIX
: systems run on PCs, as UNIX is portable and UNIX users have a choice of
: much better platforms. OS/2 (in its present incarnation) is bound to the
: Intel platform, with its crufty 8MHz bus and backward-compatible architecture.
: Some day in the not-too-distant future, when Intel have stretched the laws
: of physics to their limit in order to get their latest CISC elephant to fly,
: MS-DOS has gone the way of CP/M and the Alpha (or some other RISC chip)
: is widely accepted, this architecture will join the S-100 bus and the
: Apple II in the museum; it is hoped that we will all be running NeXTstep
: on our 250MHz Alpha machines...
You do know that the Alpha PC from DEC (and also the R4000 PCs) have
a EISA bus. And what is the speed of the EISA bus ? 8 Mhz !
And it is a 32 bit bus . It seems that PCs bus will never catch up
with the processor.
What is the width of PCI ?
--
Rui Salgueiro | Dpt. de Matematica |"In my life / Why do I smile
r...@mat.uc.pt | Universidade de Coimbra | at people who I'd much rather
r...@inescc.pt | Portugal - Europe | kick in the eye" - Morrissey
Mike Dahmus
>
>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes in
> <CC9Mz...@hawnews.watson.ibm.com> on 24 Aug 93 13:33:26 GMT
>
><Stuff about which vendor is locking who into there product deleted>
>
>>OS/2's main advantage for me over UNIX is performance. I've never run a UNIX
>>box with X (and I've tried plenty) that could perform anywhere nearly as well
>>as OS/2 (the only one that comes close is linux, and then all you have to do
>>is take the Workplace Shell off of OS/2 to bring it down to linux's level of
>>function).
>
> HA! HA! HA! HA! HA! WOOOOOOOOO!!!!
>
>Please excuse me while I get up off of the floor. OS/2 running X versus
>a Unix box running X? OS/2 is faster? What clock have you been watching?
Um, and my question to you is what post have you been reading? I didn't say
I was running X on the OS/2 box! (By the way, the new version of tcp/ip that
just came out is a lot quicker, and might solve some of your speed problems -
it certainly has improved the NFS speed a lot).
------
Mike Dahmus Internet: mi...@vnet.ibm.com
Mike Dahmus
Don't be ridiculous. My home PC (a 486) has 8MB of memory, and the only time it
is slow is when I try to do something else during a build (and I'm using a
really piggy compiler). I have no trouble using Quicken to balance the
checkbook (which also requires starting a copy of Windows, btw) while
downloading in a DOS box with ProComm at 9600 baud.
Mike Dahmus
>: Jim, recent figures indicate that OS/2 is now selling more copies than all
>: varieties of Intel UNIX put together. I suspect that some of your figures
>: above are inflated, but even if not, OS/2 is certainly becoming a bigger
>: "niche" than PC UNIX.
>
>Which does not mean that OS/2 is burying UNIX. Only a minority of UNIX
>systems run on PCs, as UNIX is portable and UNIX users have a choice of
>much better platforms. OS/2 (in its present incarnation) is bound to the
>Intel platform, with its crufty 8MHz bus and backward-compatible architecture.
You need to amend your comment to read the "ISA bus", by the way. Neither
Micro Channel nor EISA share those problems.
And I would be very interested to see the figures on where UNIX is being run
outside of the education market. (I believe you, but I want to see the
information).
Anyways, the fact that OS/2 is outselling UNIX on the PC, coupled with the
fact that most here wouldn't relegate the PC UNIX market to a "dying niche"
would tend to prevent OS/2 from getting that label as well.
P.S. - Soon I will be able to run NeXTStep at home - I'm placing my $750
order for 16MB more memory (taking me up to 20MB total) and will just have
to spend another bunch to replace my graphics card and get another hard disk,
and then I'll be able to! Gosh!
Jim Frost
>One of the major advantages of OS/2 (and probably headaches in developing
>it) is that it still runs DOS and Windows programs. I don't think
>a Unix for the PC could do that well. "So what?" many people will say.
>Well, that's a big deal. Users don't want to buy all new applications
>along with the upgrade.
The Sun386i was running DOS and Windows applications under UNIX in
1988.
Interactive was selling SVR3 with the ability to run DOS applications in
1988, using a product known as VP/ix. I believe several other UNIX vendors
were using the same product. I'm not sure if it ran Windows, but I can't
think of a lot of reasons why it wouldn't since VP/ix essentially
emulated the hardware.
SGI was providing DOS and Windows support in 1989 when the Personal
Iris (SGI 4D/20) came out, using software emulation. When the emulator
was run on more powerful workstations (eg the 4D/70) it outperformed
all PC hardware available at that time. It was pretty close even on the
4D/20.
The ability to run DOS and windows applications is not new to UNIX.
>Is multiuser support REALLY that important to most of the PC users? I
>don't believe that it is.
No, but it's often very important to the corporation as a whole, for two
reasons:
1. Multiuser systems are far easier to manage than single-user systems.
2. Single-user systems do not have the power to run enterprise-wide
databases necessary for many businesses. Many times businesses use
PC's merely as smart terminals to multiuser systems.
My big problem with NT is that it obviously targets the server environment,
yet provides almost no capabilities that would make it particularly usable
in that environment. It's a cool OS design and seems to be well
implemented but the services supplied with the product are not up to
the level they must be for the target audience they want to trap.
OS/2 is a different story; it's targetted as a better-DOS and works
well at that task. Unfortunately for IBM they're always going to be
a follower so long as Microsoft Windows is driving the applications.
That's where UNIX has been for years WRT the desktop environment.
jim frost
ji...@centerline.com
Jim Frost
>Steve Houle (woo...@panix.com) wrote:
>: From what I've heard lately, AIX is so bad that it IS one of IBM's
>: greatest mistakes.
>gibber gibber gibber upgrading from AIX 3.1 to 3.2 gibber gibber hide under
>the table gibber crash crash aaaaaargh pain suffering fear loathing
>aaaaaaaaaaiiieeeeeee.....
That's not so bad. At least the install programs don't ask you a
question every four hours or so under 3.2. You can tell it to start
installing, come back sometime the next day, and it's done. Most of
the 3.1 releases required you to hit "return" once every four to six
hours, and to change tapes a few times, so installations typically
ate three days of your time. I got into the habit of doing minimal
installations and adding options as needed; that way if the system
got hosed I hadn't wasted all that installation time.
Of course a lot of 3.1 applications broke when 3.2 came out because IBM
modified the *semantics* of malloc() in the libc shared library....
3.2 is an improvement in a number of places, so it's not really something
to run scared from. (Did I really say that?)
jim frost
ji...@centerline.com
William F. Pemberton
Andrew Bulhak <a...@yoyo.cc.monash.edu.au> wrote:
>
>Some day in the not-too-distant future, when Intel have stretched the laws
>of physics to their limit in order to get their latest CISC elephant to fly,
>MS-DOS has gone the way of CP/M and the Alpha (or some other RISC chip)
>is widely accepted, this architecture will join the S-100 bus and the
>Apple II in the museum; it is hoped that we will all be running NeXTstep
>on our 250MHz Alpha machines...
>
So, RISC is now the only way to make a fast processor? CISC is dead?
Just because RISC is the current hot item doesn't mean that RISC is now
the only "good" way to make a processor. Intel's line may have problems,
but it is not because they stick with CISC.
--
Bill
Jim Frost
>>a few on the Intel platforms, and some are far better than average.
>>BSDI's implementation is cheap and very, very fast. Linux is *free*
>>and reasonably fast.
>But I *was* comparing OS/2 to linux, by the way. And I've only heard you say
>that BSD is quicker than linux - most people seem to think the opposite is
>true.
BSD is the only version I've used that is quicker (SunOS 4 used to be
pretty quick, but Solaris definitely isn't). I dunno where you heard that
linux is faster, it isn't in my experience and you're the first I've heard
mention otherwise. (Actually it probably is quicker for some operations,
but as soon as you start paging BSD is much better.)
>>If you want a nice, usable graphical shell for a UNIX box, the only
>>thing you seem to really be griping about, you can buy one. There are
>>quite a few other than NeXT's products, believe it or not.
>Untrue. The only one worth consideration for someone who requires an OO shell
>is NeXTStep.
Almost nobody "requires" an OO shell. A few people would like one, but
most people who "require" a GUI shell in the first place don't need anything
really complicated (in fact, most users can't even *handle* anything
complicated).
>>>Why are there no usable applications for any major UNIX platform? (I'm not
>>>calling NeXTStep UNIX).
>>
>>Ever hear of Frame? Oracle? Interleaf? Wingz? Ingres?
>>WordPerfect? Lotus 1-2-3? What about manufacturing controllers?
>>Accounting systems (ALCIE, etc)?
>Usable, usable, usable. There was a really big, important reason I put that
>word in front of the word "applications", Jim. Even the few major commercial
>apps on UNIX platforms are typically behind in function from their Windows
>counterparts, and often have disjointed interfaces to boot.
You've never used Frame if you think it's unusable. Or Wingz. Or the
current version of WordPerfect. I'll give you 1-2-3, but that's it -- all
the other applications I mentioned are *more powerful* under UNIX than
under Windows, and some of those applications *don't even exist* under
Windows.
>>>UNIX is no panacea. Grow up and join the real world.
>>
>>Depends on what you want to do. OS/2 is fine for running your little
>>bitty spreadsheet or wordprocessor. That's great for a lot of people,
>>but you just can't support a very large business on that. And God
>>help you if you need a zero-downtime system or a fast WAN. You can't
>>even buy OS/2 boxes fast enough to make good enterprise file or
>>database servers. But it does do a fair job of handling one or two
>>smallish graphical applications, and even as a terminal to connect to
>>more capable systems (which is what IBM likes to see!).
>I think you need to start reading some non-UNIX trade press. OS/2 was already
>in a lot of big businesses with the 1.x versions. OS/2 1.3 qualifies very
>easily as a "zero-downtime system" (I'm not going to make this claim for 2.x,
>at least not yet). There are hundreds, if not thousands of apps written by
>businesses that run on OS/2 1.x, including a lot of airline systems, banking
>systems, and other large operations.
OS/2 is in a lot of big businesses (although "in" might count as only one
installation), but it's not a central system in very many of them. It can't
be -- the reliability and scalability just aren't there. OS/2 1.3 is not a
zero-downtime system -- where's the disk mirroring? Redundant CPU support?
Etc. It's not even *close*. Airlines, banks, and large corporations do not
put their central databases on OS/2 boxes. They probably use them for small
departmental servers on occasion, and probably for terminals to the larger
systems, but nobody running a large company uses it for central operations.
I do real the PC trade press, BTW. They're still gaga over Windows 3.1,
with nary a mention of OS/2. And they all focus on user-point products,
not enterprise-wide products. That's because PCs are supposed to be
user-point products. Believing that they are anything else is foolish.
Happy hacking,
jim frost
ji...@centerline.com
Jim Frost
>In article <1993Aug30.0...@monu6.cc.monash.edu.au>, a...@yoyo.cc.monash.edu.au (Andrew Bulhak) writes:
>> Which does not mean that OS/2 is burying UNIX. Only a minority of UNIX
>> systems run on PCs, as UNIX is portable and UNIX users have a choice of
>> much better platforms. OS/2 (in its present incarnation) is bound to the
>> Intel platform, with its crufty 8MHz bus and backward-compatible architecture.
>>
>Hey, Have you heard of _MICRO_CHANNEL_??
>This eats EISA!!!
>You should sit down in front of a REAL IBM sometime. PS/2 variety.
Gee, aren't there two versions of the Micro Channel architecture? The
original version, and the newer version needed to achieve acceptable
throughput on really fast machines?
("Real" IBMs are ES/9000s anyway.)
jim frost
ji...@centerline.com
Jim Frost
>In <Aug.28.20.28....@pilot.njin.net>, tspe...@pilot.njin.net (Thomas Spencer IV) writes:
>>That is if the OS is going to run, to say nothing of any applications.
>>Open your first app in 8meg and the hard drive starts to do the tango.
>>You need 16meg to get anything more than trivial applications to work
>>without very long execution delays due to paging.
>Don't be ridiculous. My home PC (a 486) has 8MB of memory, and the only time it
>is slow is when I try to do something else during a build (and I'm using a
>really piggy compiler). I have no trouble using Quicken to balance the
>checkbook (which also requires starting a copy of Windows, btw) while
>downloading in a DOS box with ProComm at 9600 baud.
So OS/2 does a nice job of multitasking two light-load applications,
but as soon as you run something moderately heaviweight it really
slows down?
Hmm. ;-)
jim frost
ji...@centerline.com
Mike Dahmus
I know you were being funny, but I want to make sure no one else gets the
wrong idea - it is slow because it is SWAPPING, because this compiler, like I
said, is a bit of a pig. (The system will swap unless this compiler is pretty
much the only active process). Remember that OS/2 with the Workplace Shell
has a working set around 4MB itself, some of which cannot be swapped out, so
if this compiler decides to use 6MB of memory, and I start something else, it's
definitely swap city.
Michael Rogero Brown
issue. It's being replaced by a subscription-only rag called "Software
Development".
Too bad. I got CL back when it first started. Had some of the best covers
and some great articles. It slowly got to C&MS-DOS for me, so I dropped them.
--
<< Michael Rogero Brown | Any opinions expressed are my >>
<< CS Graduate Student-Florida Atlantic Univ | own, and generally unpopular >>
<< Internet: mich...@sol.cse.fau.edu | with others. >>
<< BitNet: m_brown@fauvax | Ask me if I care. >>
Jim Frost
Some would say that Intel has already given up CISC. The Pentium is
more RISC than CISC, except it still has very few registers. They
make up for that by putting in ever-larger caches....
jim frost
ji...@centerline.com
Jim Frost
>In <25t75i$b...@wcap.centerline.com>, ji...@centerline.com (Jim Frost) writes:
>>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>>>Don't be ridiculous. My home PC (a 486) has 8MB of memory, and the only time it
>>>is slow is when I try to do something else during a build (and I'm using a
>>>really piggy compiler). I have no trouble using Quicken to balance the
>>>checkbook (which also requires starting a copy of Windows, btw) while
>>>downloading in a DOS box with ProComm at 9600 baud.
>>
>>So OS/2 does a nice job of multitasking two light-load applications,
>>but as soon as you run something moderately heaviweight it really
>>slows down?
>>
>>Hmm. ;-)
>I know you were being funny, but I want to make sure no one else gets the
>wrong idea - it is slow because it is SWAPPING, because this compiler, like I
>said, is a bit of a pig. (The system will swap unless this compiler is pretty
>much the only active process). Remember that OS/2 with the Workplace Shell
>has a working set around 4MB itself, some of which cannot be swapped out, so
>if this compiler decides to use 6MB of memory, and I start something else, it's
>definitely swap city.
I was only partially joking. Just because the system is swapping
doesn't mean the overall performance should go through the floor;
that's why I keep talking about BSD performance as "very good" -- the
system degrades gracefully under very bad conditions.
Anybody can make a fast operating system if they don't have to swap,
but the really good operating system builders make it fast even if it
*is* swapping by applying good prediction or limitation heuristics.
Unless both the compiler and WPS have nonlocal page useage larger than
the available RSS you shouldn't have really bad paging performance
(bad paging performance is known as "page thrashing"). If one of them
does -- the compiler is a likely candidate -- then the performance of
that application should degrade sharply, but the other active
applications with steady resident-sets shouldn't be too adversely
affected.
In a simplistic global-LRU replacement scheme you'll get thrashing
pretty early because the one big application steals the resident pages
from the other applications and all of them start paging, degrading
overall performance rather badly. But operating systems tuned for
performance, such as VMS, allow limiting RSS per process -- what that
means in real life is that one application using tons of VM will get
very slow, but overall system response is not terrible. I've yet to
see a UNIX system that allows this, but many of them do limit RSS for
large processes to avoid the early thrashing behavior seen with
global-LRU; this approaches the same performance as fixed-maximum-RSS
systems but allows dynamic RSS adjustment based on overall system
load, a fair compromise.
I don't have any OS/2 internals books here, unfortunately. Care to
teach me about the OS/2 page replacement algorithm? I bet it's purely
LRU.
Upon rereading your posting I noticed you said that WPS has a working
set of 4Mb. Is that true, or is that its total usage? 4Mb is a
pretty big total size, but a 4Mb working set is humungous unless your
doing a lot of table-walking. What parts can't be swapped out? I'm
curious, since it's hard to envision needing to lock pages for a shell
program.
jim frost
ji...@centerline.com
Mike Dahmus
>
>I don't have any OS/2 internals books here, unfortunately. Care to
>teach me about the OS/2 page replacement algorithm? I bet it's purely
>LRU.
Unfortunately, I'm not knowledgeable enough on the subject to help you out.
I'll bet some of it is in the book "The Design of OS/2" by Deitel/Kogan, or
you might try the OS/2 "redbooks".
>Upon rereading your posting I noticed you said that WPS has a working
>set of 4Mb. Is that true, or is that its total usage? 4Mb is a
>pretty big total size, but a 4Mb working set is humungous unless your
>doing a lot of table-walking. What parts can't be swapped out? I'm
>curious, since it's hard to envision needing to lock pages for a shell
>program.
I meant that the working set of the O/S + PM + WPS is close to 4MB. (I know
that in 2.0, it was higher, and if you were opening a lot of folders or
something without even running any apps on a 4MB system, it would be swapping
itself). And I didn't mean to imply that any major bits of the Workplace Shell
couldn't be swapped out.
------
Mike Dahmus Internet: mi...@vnet.ibm.com
Pen for OS/2 Development, IBM PSP IBM: mi...@schleppo.bocaraton.ibm.com
Disclaimer: Not an official IBM sp kesman IBM Vnet: MDAHMUS at BOCA
DoN. Nichols
[ ... ]
>And I would be very interested to see the figures on where UNIX is being run
>outside of the education market. (I believe you, but I want to see the
>information).
Just to provide a non-educational data point. (We are a military
organization - and no - I am not posting from there now.)
At work we are running well over 100 Sun SPARC machine and Solbourne
clones thereof, along with a smaller number of networked PCs (the managers
want the hand-holding that windows gives them, rather than the power that
the scientific users want). The managers are just using them for e-mail and
moving document files around between machines. The rest are developing and
testing target-recognition algorithms on large databases of imagry held in
NFS-mounted disks.
OS/2 could probably be used for some of the things that the PCs
don't do well because of memory limitations -- if they had good networking
to allow these databases to be shared around seamlessly, and the access
controls to allow everyone who needs access to data to read it, but only a
select few to change it.
Does OS/2 have adequate VM (or any, for that matter) to handle
really large processes? We have processes needing 40MB or greater of *data*
space alone.
--
Email: <dnic...@d-and-d.com> | ...!uunet!ceilidh!dnichols
<dnic...@ceilidh.beartrack.com>
Donald Nichols (DoN.) | Voice (Days): (703) 704-2280 (Eves): (703) 938-4564
--- Black Holes are where God is dividing by zero ---
Dave Brown
>In article <CCDwH...@undergrad.math.uwaterloo.ca> dagb...@undergrad.math.uwaterloo.ca (Dave Brown) writes:
>
>[...]
>
>>main(){while(;;)fork();} don't
>>work for example
>
>Shouldn't that be "don't work while example"?
Hate it when that happens. You're thinking one thing, you type
another, then the original thought slips in and confuses your fingers.
You'all know what I meant anyways! :-)
--
Dave Brown -- dagb...@napier.uwaterloo.ca -- (416) 669-5370
main(){while(1)fork();}
Andrew Bulhak
:
: P.S. - Soon I will be able to run NeXTStep at home - I'm placing my $750
: order for 16MB more memory (taking me up to 20MB total) and will just have
: to spend another bunch to replace my graphics card and get another hard disk,
: and then I'll be able to! Gosh!
Careful, Mike... Take it slow, or else your head may explode.
-- A.
Terry Kennedy, Operations Mgr.
> And I would be very interested to see the figures on where UNIX is being run
> outside of the education market. (I believe you, but I want to see the
> information).
I suppose it's so obvious that nobody thought about it - but look at the
various phone companies. Here in Bell Atlantic land, service order entry,
customer-reconfigurable Centrex, etc. is all being run on obvious Unix boxes
(where you get the Unix banner and login prompt). Also, the various auxiliary
functions of the phone switches (feature configuration storage, etc.) are
stored on fault-tolerant boxes which run somthing whose ancestor was Unix.
Terry Kennedy Operations Manager, Academic Computing
te...@spcvxa.bitnet St. Peter's College, Jersey City, NJ USA
te...@spcvxa.spc.edu +1 201 915 9381
Antonio Vasconcelos
: In article <CCDwH...@undergrad.math.uwaterloo.ca> dagb...@undergrad.math.uwaterloo.ca (Dave Brown) writes:
:
: [...]
:
: >main(){while(;;)fork();} don't
: >work for example
:
: Shouldn't that be "don't work while example"?
But :g/while/s//for/g will do the trick :-) (-:
--
Regards,
====================================================================
Antonio Vasconcelos at The Lisbon Stock Exchange - va...@bvl.pt
"I do NOT speak for BVL" NeXTmail - ro...@bvl.pt
Mike Dahmus
>
> Does OS/2 have adequate VM (or any, for that matter) to handle
>really large processes? We have processes needing 40MB or greater of *data*
>space alone.
I think your basic limit is 512M. So those processes should in theory work.
Antonio Vasconcelos
:
: That's not so bad. At least the install programs don't ask you a
: question every four hours or so under 3.2. You can tell it to start
: installing, come back sometime the next day, and it's done. Most of
: the 3.1 releases required you to hit "return" once every four to six
: hours, and to change tapes a few times, so installations typically
: ate three days of your time. I got into the habit of doing minimal
: installations and adding options as needed; that way if the system
: got hosed I hadn't wasted all that installation time.
Maybe RS/6000 work slower on the US, but I've installed AIX 3.1.5 in
about 10 RS/6000 320 and 520 and tipically it goes from 15:00 to 20:00
or 21:00, 5 or six hours, not 3 days...
Jim Frost
>ji...@centerline.com (Jim Frost) writes:
>:
>: That's not so bad. At least the install programs don't ask you a
>: question every four hours or so under 3.2. You can tell it to start
>: installing, come back sometime the next day, and it's done. Most of
>: the 3.1 releases required you to hit "return" once every four to six
>: hours, and to change tapes a few times, so installations typically
>: ate three days of your time. I got into the habit of doing minimal
>: installations and adding options as needed; that way if the system
>: got hosed I hadn't wasted all that installation time.
>Maybe RS/6000 work slower on the US, but I've installed AIX 3.1.5 in
>about 10 RS/6000 320 and 520 and tipically it goes from 15:00 to 20:00
>or 21:00, 5 or six hours, not 3 days...
Maybe you're installing from 8mm tape, and/or not installing the full
complement of utilities and languages. Each pass over the install
process with QIC 150 required swapping tapes at least four times
(dunno why, there were only two tapes) -- with each swap separated by
about 45 minutes to an hour. You also couldn't do both the BOS and
X11 installs on the same pass because of some dependency problem,
further drawing out the install process.
Probably you can do an install in a day if you sit there and pay
attention to the process, but I always had other things to do. If
you're using the 8mm tape it is *much* easier since you don't have to
swap tapes all the time -- just set things up, say "go" and come back
a few hours later. I was much happier once we got an 8mm drive.
Anyway 3.1.5 improved the installation process quite a bit over the
earlier releases -- even doing the "update" installs for 3001, 3002,
and 3003 took impressive amounts of time. Back when new releases were
showing up almost weekly that ate up quite a bit of time. 3.2
improved things even further -- I did a couple of 3.2 installs in
about 3 hours (from 8mm tape, that is) without multiple install
passes.
Opinions expressed herein are my own and do not necessarily reflect
those of my employer. (It still annoys me that that's no longer
considered the default case.)
jim frost
ji...@centerline.com
Jens Stark
>So, RISC is now the only way to make a fast processor? CISC is dead?
>Just because RISC is the current hot item doesn't mean that RISC is now
>the only "good" way to make a processor. Intel's line may have problems,
>but it is not because they stick with CISC.
The most RISCy CPU I actually saw working was a single instruction (!) CPU
calculating prime numers and printing "Hello, World !". This does not make
it superior to a CISC machine like my Data General MV15/20, which even has
instructions to handle linked lists. It all depends on the use of it.
Washing machines and toasters are great applications for a RISC CPU,
while some applications ( like actually doing something useful ) need
more instructions. :-)
Intel has problems because they use STATIC microcode.
Greetings,
Jens
Dave Brown
>order for 16MB more memory (taking me up to 20MB total) and will just have
>to spend another bunch to replace my graphics card and get another hard disk,
>and then I'll be able to! Gosh!
<mildly> And I'm running Linux on a 1.6M 386SX with a 60M hard drive.
Whoop-te-piddle.
--
Dave Brown -- dagb...@napier.uwaterloo.ca -- (416) 669-5370
"It takes insects to scare me. And lizards. Not humans."
--Amy Fong
Andrew Bulhak
:
: So, RISC is now the only way to make a fast processor? CISC is dead?
: the only "good" way to make a processor. Intel's line may have problems,
: but it is not because they stick with CISC.
CISC has a major disadvantage; there is a lot of microcode to get through
which adds overheads. RISC is often hardwired and has no such problem.
Basically, to make a machine faster than the 486, you can
a) Build a new upward-compatible chip; put in some optimisations
(such as the pipeline), use a state-of-the-art process to
make the Damned Thing smaller and worry about cooling it.
With each generation, speed upgrade, etc. this becomes more
difficult as the baggage of upward compatibility prevents you
from taking certain short cuts. Eventually the laws of physics
are going to make this unviable.
b) Start from scratch; make a chip of a totally new architecture,
using a highly optimised RISC design. You don't have to worry
whether old MS-DOS software will run on it and are free to
tune it for performance rather than compatibility.
But why not CISC? Well, because CISC is slower. If you knew exactly what
instructions would be used by applications and in what patterns, CISC may
be more viable; however, you do not, and by adding more utility instructions,
you increase overheads and make the chip slower than it need be. With a
hardwired RISC processor, however, the software almost _becomes_ the microcode
(I say almost because it is in slow external memory); thus programs run
faster.
+---------------------------------------------------------------------------+
| Andrew Bulhak |"Laundry is the fifth dimension!! the |
| a...@yoyo.cc.monash.edu.au | washing machine is a black hole and the |
| Monash Uni, Clayton, | pink socks are bus drivers who just fell in!"|
| Victoria, Australia | -- Zippy the Pinhead |
+---------------------------------------------------------------------------+
Payson
>William F. Pemberton (wf...@holmes.acc.Virginia.EDU) wrote:
>:
>: So, RISC is now the only way to make a fast processor? CISC is dead?
>: Just because RISC is the current hot item doesn't mean that RISC is now
>: the only "good" way to make a processor. Intel's line may have problems,
>: but it is not because they stick with CISC.
>
>instructions would be used by applications and in what patterns, CISC may
>be more viable; however, you do not, and by adding more utility instructions,
>you increase overheads and make the chip slower than it need be. With a
>hardwired RISC processor, however, the software almost _becomes_ the microcode
>(I say almost because it is in slow external memory); thus programs run
>faster.
>
Suppose a program on a CISC spends 90% of its time executing instructions
which are found on a RISC [with a few exceptions this is almost always the
case]. Even if a RISC would take twice as long to "emulate" those CISC
instructions which it did not directly support, this would only slow the
machine down by 10%. If going to RISC makes all the instructions go faster
by 10% [which it often does], then there is a net win.
Note that RISC's advantages are further underscored by the fact that many
CISC instructions aren't all that great [STOSW on a 486, for example, is
slower than MOV AX,[DI]; INC DI].
There are occasions when CISC has an advantage. Particularly in graphics
co-processor chips where much of one's time is spent performing specific
operations, CISC chips can work well when their microcode is a good fit for
the application. What I would like to see, however, would be a machine with
a general-purpose CPU *PLUS* a very fast VLIW Harvard-architecture RISC with
a small program store [any idea whether this description fits the Atari
Jaguar? Would be really neat if so] Since 90% of most programs' time is
spent in a tiny portion of the code, hyper-optimizing that portion of the
code for a VLIW RISC could have a major payoff.
--
-------------------------------------------------------------------------------
pay...@cs.wisc.edu | "Je crois que je ne vais jamais voir... | J\_/L
John Payson | Un animal si beau qu'un chat." | ( o o )
da...@gilly.uucp
> I suppose it's so obvious that nobody thought about it - but look at the
>various phone companies. Here in Bell Atlantic land, service order entry,
>customer-reconfigurable Centrex, etc. is all being run on obvious Unix boxes
>(where you get the Unix banner and login prompt). Also, the various auxiliary
>functions of the phone switches (feature configuration storage, etc.) are
>stored on fault-tolerant boxes which run somthing whose ancestor was Unix.
If you've ever been called up by a system that then told you all of the
operators were busy, please hold... it was probably a unix system too.
(I ran such a beast for the Death Star a while back.)
------------------------ uunet!quack!gilly!dave ------------------------
================= Dave Fischer - Nature's Perfect Food =================
----------------------- dave%gi...@speedway.net ------------------------
Lars R{der Clausen
>wf...@holmes.acc.Virginia.EDU (William F. Pemberton) writes:
>>So, RISC is now the only way to make a fast processor? CISC is dead?
>>Just because RISC is the current hot item doesn't mean that RISC is now
>>the only "good" way to make a processor. Intel's line may have problems,
>>but it is not because they stick with CISC.
>The most RISCy CPU I actually saw working was a single instruction (!) CPU
>calculating prime numers and printing "Hello, World !". This does not make
...a full computer, or what? I wondered how many commands you actually need
to make a reasonable computer? You need some kind of test, some kind of
arithmetic and what else? One book stated that a RICSC processor had 3
commands, but I didn't quite believe that.
>Greetings,
> Jens
-Lars
--
..managing security with all the grace of a Brontosaurus trying to type.
-Michel E. Kabay, about security in 'Jurassic Park' (and elsewhere)
Jim Frost
>It all depends on the use of it.
>Washing machines and toasters are great applications for a RISC CPU,
>while some applications ( like actually doing something useful ) need
>more instructions. :-)
I can't tell how serious you're being, but Turing says otherwise. The
only real difference between the two design philosophies from the
point of view of a toaster is that RISC generally requires more code
space than CISC, making your toaster more expensive. CISC is
wonderful for smallish embedded systems (although RISC is getting used
more and more often for such things, particularly things like engine
controllers which have major real-time requirements).
jim frost
ji...@centerline.com
Richard F. Drushel
In a previous article, ji...@centerline.com (Jim Frost) says:
> ... The
>only real difference between the two design philosophies from the
>point of view of a toaster is that RISC generally requires more code
>space than CISC, making your toaster more expensive. CISC is
>wonderful for smallish embedded systems (although RISC is getting used
>more and more often for such things, particularly things like engine
>controllers which have major real-time requirements).
I thought that code compactness was the original driving force
behind the development of CISC instruction sets, especially in 8-bit
microprocessors with 64K address space, and with RAM being extremely
expensive. Why write an entire 16-bit multiply or divide subroutine
when you could do it with a single, 1- or 2-byte instruction? Each
byte of RAM or ROM saved could, in aggregate, allow room for extra
features in software, or else reduce hardware costs (less memory needed).
But, of course, nobody optimizes for size any more :(
Dr. Rich Drushel
Diehard Z80 assembly programmer for the Coleco ADAM
Dept. of Biology, Slug Division :)
Case Western Reserve University, Cleveland, Ohio 44106 U.S.A.
Andrew Bulhak
:
: I don't have any OS/2 internals books here, unfortunately. Care to
: teach me about the OS/2 page replacement algorithm? I bet it's purely
: LRU.
And while you have the book at hand, could you answer a question that has
been troubling me for some time? Does the OS/2 scheduler really use the
bogo-sort to sort its process table, or does it just use a FCFS
scheduling algorithm?
Andrew Bulhak |
a...@yoyo.cc.monash.edu.au | OpenLook may be gone,
Monash Uni, Clayton, | but it will live on in our hearts forever!
Victoria, Australia |
DoN. Nichols
[ ... ]
>If you've ever been called up by a system that then told you all of the
>operators were busy, please hold... it was probably a unix system too.
>(I ran such a beast for the Death Star a while back.)
Arrrgh! You!!! Death threats! Mail bombs! May a plague of Phone
solicitors descend upon you.
Now you'll probably tell me that it was run my my favorite little
AT&T machine, the 3B1/7300/UNIX-PC. How low can we stoop?
Perhaps you can do pennance by telling us how best to confuse the
machines, perhaps so the operators will talk to each other.
Michael L. Kaufman
>It's being replaced by a subscription-only rag called "Software Development".
Actually, I just saw _SD_ in a local Barnes and Noble bookstore.
Michael
Bernd Meyer
>I think your basic limit is 512M. So those processes should in theory work.
Everybody, sing on:
"Who'll ever need more than 640k? Who'll ever need more than 512M?"
Bernie
--
We both know that the earth is round | Bernd Meyer, EE-student
So we can't see the way before us to its end | "Nobody is a failure who has
We walk on this way, hand in hand, | friends" (from: isn't it a
And I hope you are still with me behind the horizon| wonderful life?"
Mike Dahmus
>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>
>>I think your basic limit is 512M. So those processes should in theory work.
>
>Everybody, sing on:
>"Who'll ever need more than 640k? Who'll ever need more than 512M?"
Part of the reason for this limit is that OS/2 runs in the flat memory model
of the 386. Would you rather have segments?
Helge Hafting
> In article <1993Sep1.0...@monu6.cc.monash.edu.au> a...@yoyo.cc.monash.edu.au (Andrew Bulhak) writes:
> >William F. Pemberton (wf...@holmes.acc.Virginia.EDU) wrote:
> >:
> >: So, RISC is now the only way to make a fast processor? CISC is dead?
> >: Just because RISC is the current hot item doesn't mean that RISC is now
> >: the only "good" way to make a processor. Intel's line may have problems,
> >: but it is not because they stick with CISC.
> >
> >But why not CISC? Well, because CISC is slower.
that execute in 1 clock. The instructions using many clocks tends to be
little used ones.
William Chesters
R{der Clausen) writes:
# Thus spake ro...@foobar.hanse.de (Jens Stark):
#
# >The most RISCy CPU I actually saw working was a single instruction (!) CPU
# >calculating prime numers and printing "Hello, World !". This does not make
#
# ...a full computer, or what? I wondered how many commands you actually need
# to make a reasonable computer? You need some kind of test, some kind of
# arithmetic and what else? One book stated that a RICSC processor had 3
# commands, but I didn't quite believe that.
I'm sure I remember being told (in a computer-language semantics
course I did in a previous life) that you can get all the lambda-
calculus out of a single "combinator". On the other hand I can't
remember which one. Call it C. Then you can compile any lambda-
calc function (LISP program, if you like) into something that looks
like
(C C (C C) (C (C C) (((C C) C) C (C ......
---in other words, you can build a graph-reduction machine with
only one "instruction". This isn't quite the same as a one-
instruction trad. processor, because (obviously!) the brackets
are significant. But it's pretty neat. I believe people have
actually made such machines based on a slightly more generous
"instruction set" of three combinators.
--
William Chesters (will...@aifh.ed.ac.uk)
`/usr/games/fortune`
Andrew Bray
>outside of the education market. (I believe you, but I want to see the
>information).
I used to work for a company that produced SNA and X25 s/w for Unix machines,
it was a US company, but I only worked for the European bit. Most sales
were for PC Unix, but a resonable proportion were for larger machines.
Places taking over 100 licences included (1-2 licences per Unix box generally,
most machines running >1 user):
Banks.
Motor manufacturers.
UK government.
Austrian police force.
Norwegian Telecom.
Petrol vendors.
Some sites were ordering 1000-10000 licences. We usually sold in the
region of 10000 licences a year, with two sales staff and no distribution
other than IBM. Other smaller sites included lawyers, retail, insurance
brokers, etc.
This software was only really usable on sites with leased lines, and
was almost exclusively used to communicate to IBM mainframes and AS/400s;
there were very few pure X.25 applications. Do you want to hazard a guess
about what proportion of commercial Unix machines are directly hooked up
to a leased line?
There are definately a lot of Unix boxes in a commercial environment, at
least here in Europe.
Regards,
Andy
Alan Bowler
>
>...a full computer, or what? I wondered how many commands you actually need
>to make a reasonable computer? You need some kind of test, some kind of
>arithmetic and what else? One book stated that a RICSC processor had 3
>commands, but I didn't quite believe that.
>
machines. Both were quite "reasonable computers". The RISC concept of
a few simple easily implemented instructions has been reinvented
several times. These days when something that has a floating point
square root instruction is still called RISC, I wonder what the concept
will be called next time it is rediscovered.
Part of the problem of course is that when you go to build the next
version of a machine for any architecture, you need to preserve the
idiosyncrasies of the previous version. The can become major
bottlenecks as future versions are developed. When someone starts a
new architecture those particular bottlenecks can be avoided and some
impressive gains may be made. The problem is that this is at the cost
of not being able to run the old software. In the set of articles on
the Alpha in CACM, DEC pointed out that the "execute 1 instruction
placed after a branch" feature of some RISC architectures had become a
problem in later implementations of the same architecture with longer
pipelines.
Alan Bowler
>
> I thought that code compactness was the original driving force
>behind the development of CISC instruction sets, especially in 8-bit
>microprocessors with 64K address space, and with RAM being extremely
>expensive. Why write an entire 16-bit multiply or divide subroutine
>when you could do it with a single, 1- or 2-byte instruction? Each
>byte of RAM or ROM saved could, in aggregate, allow room for extra
>features in software, or else reduce hardware costs (less memory needed).
>But, of course, nobody optimizes for size any more :(
Reducing the code size also allows more memory bandwidth for data
transfer. A memory cycle used to fetch an instruction is one that is
not used to fetch data. However, if instead you use the time for
multiple microcode memory cycles you have not gained anything. Like
everything else there are design tradeoffs, and as chip building
technology changes, the optimum point moves.
Jim Frost
>In <1993Sep1.1...@umibox.hanse.de>, ro...@umibox.hanse.de (Bernd Meyer) writes:
>>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>>
>>>I think your basic limit is 512M. So those processes should in theory work.
>Part of the reason for this limit is that OS/2 runs in the flat memory model
>of the 386. Would you rather have segments?
Where did the rest of the bits in the 32-bit address space go? I
thought the limit was 4Gb virtual per segment (ie a full 32 bits), but
that the physical constraints are significantly smaller. I can't find
my #%^$^ 386 reference to check, though.
jim frost
ji...@centerline.com
Mike Dahmus
>In article <CCKsn...@hawnews.watson.ibm.com> mi...@vnet.ibm.com (Mike Dahmus) writes:
>>And I would be very interested to see the figures on where UNIX is being run
>>outside of the education market. (I believe you, but I want to see the
>>information).
And, after six or seven posts...
These anectdotes are both heartwarming and informative. However, what I'm
really looking for is not "X site runs Y licenses", but rather, "Y licenses
of X system were sold in 1993 outside of the educational market".
Antonio Vasconcelos
: Antonio Vasconcelos (va...@bvl.pt) wrote:
:
: >That part about no memory protection is a bit stupid for a
: >multitasking OS.
:
: Depends. If you're running applications which are well behaved, it's not a
: problem. If you want to run buggy software well... The primary limitation is
: that it doesn't pay to develop software on a production machine, but an
: Amiga is cheap enough that it's not a problem to buy one for development
: anyway.
One could assume that "There aint no such thing as a bug-free
application"...
A decent MMU and some software support whouldn't be so expensive.
: You can have virtual memory. I think you might find there are more machines
: out there with hard drives than you think.
Go tell that to C=.
I've been trying to read an ADOS manual and in every eg. I have to
ignore about 10 lines witch are telling me to insert the correct disks
in the "internal drive"... Argh!
Virtual memory ? How ? Use Amiga UNIX ?
--
Regards,
====================================================================
Antonio Vasconcelos at The Lisbon Stock Exchange - va...@bvl.pt
Disclaimer: "I do NOT speak for BVL" * NeXTmail * ro...@bvl.pt
Antonio Vasconcelos
: >Maybe RS/6000 work slower on the US, but I've installed AIX 3.1.5 in
: >about 10 RS/6000 320 and 520 and tipically it goes from 15:00 to 20:00
: >or 21:00, 5 or six hours, not 3 days...
:
: Maybe you're installing from 8mm tape, and/or not installing the full
: complement of utilities and languages. Each pass over the install
: process with QIC 150 required swapping tapes at least four times
: (dunno why, there were only two tapes) -- with each swap separated by
: about 45 minutes to an hour. You also couldn't do both the BOS and
: X11 installs on the same pass because of some dependency problem,
: further drawing out the install process.
No, I was installing from QIC150, but not the FULL installation, no X,
only C and Info only in two of them. I can't remember anymore how many
times I have to switch tapes, but I think that it was only once.
Steve McDowell
> In <1993Sep1.1...@umibox.hanse.de>, ro...@umibox.hanse.de (Bernd Meyer) writes:
> >
> >Everybody, sing on:
> >"Who'll ever need more than 640k? Who'll ever need more than 512M?"
>
> Part of the reason for this limit is that OS/2 runs in the flat memory model
> of the 386. Would you rather have segments?
No, I'd rather have a 32-bit virtual address space.
Give me the same limitations that are imposed by the hardware, and I may accept
them; not taking advantage of the capabilities of the hardware is just plain
sloppy.
--
Opinions are mine, not my employers
Mike Dahmus
X-Newsreader: IBM NewsReader/2 v1.00
Lines: 23
In <265g46$g...@wcap.centerline.com>, ji...@centerline.com (Jim Frost) writes:
>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>>In <1993Sep1.1...@umibox.hanse.de>, ro...@umibox.hanse.de (Bernd Meyer) writes:
>>>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>>>
>>>>I think your basic limit is 512M. So those processes should in theory work.
>[...]
>>Part of the reason for this limit is that OS/2 runs in the flat memory model
>>of the 386. Would you rather have segments?
>
>Where did the rest of the bits in the 32-bit address space go? I
>thought the limit was 4Gb virtual per segment (ie a full 32 bits), but
>that the physical constraints are significantly smaller. I can't find
>my #%^$^ 386 reference to check, though.
You're right, which is why I used the words "part of". I've heard the other
reasons, but don't remember them off-hand.
Jim Frost
>In <CCnJ0...@madhouse.demon.co.uk>, an...@madhouse.demon.co.uk (Andrew Bray) writes:
>>In article <CCKsn...@hawnews.watson.ibm.com> mi...@vnet.ibm.com (Mike Dahmus) writes:
>>>And I would be very interested to see the figures on where UNIX is being run
>>>outside of the education market. (I believe you, but I want to see the
>>>information).
>And, after six or seven posts...
>These anectdotes are both heartwarming and informative. However, what I'm
>really looking for is not "X site runs Y licenses", but rather, "Y licenses
>of X system were sold in 1993 outside of the educational market".
Do you have any idea how many vendors that would cross? If you want
to know how many Windows licenses were sold, ask Microsoft. If you
want to know how many UNIX or UNIX work-alike licenses were sold, you
can get some of them by asking:
IBM (AIX)
Ahmdahl (UTS, I think)
DG (DG/UX)
DEC (Ultrix, OSF/1)
HP (HP/UX)
SGI (IRIX)
Sun (SunOS, Solaris)
Solbourne (SunOS variant)
SCO (Xenix)
Interactive (SVR4)
ESIX (SVR4)
AT&T (SVR4)
MIPS (RISCOS)
NCR (SVR4)
Motorola (SVR4)
Unisys
Bull
Cray
Thinking Machines
Sumitomo
Fuji
There are more, particularly in the international market, but that
probably hits all the reasonably large vendors and a few of the
smallish ones.
The educational market accounts for a very small percentage of UNIX
sales. Educational organizations are notoriously short on cash; they
tend to buy a few largish machines and set them up as multiuser
machines. A few institutions, notably MIT and CMU, have a lot of
workstations available but they are the exception.
You hear about workstation sales in the trade rags, but most UNIX
system sales are not workstations -- they're turnkey DP systems or
embedded controllers, not stuff that's glitzy enough for the news.
Big-business buys the stuff by the truckload, though, and tons of
little systems (often on PC hardware) are sold to businesses like
small medical practices.
I wish I could tell you a concrete number, but that's just not
possible. A very conservative figure is twenty-five million machines,
but a realistic figure is closer to twice that many (PCs and
workstations alone account for most of the smaller figure).
Unfortunately the market is heavily segmented with few vendors
catching more than a couple of million.
Food for thought.
jim frost
ji...@centerline.com
Paul Tomblin
>Do you have any idea how many vendors that would cross? If you want
>to know how many Windows licenses were sold, ask Microsoft. If you
>want to know how many UNIX or UNIX work-alike licenses were sold, you
>can get some of them by asking:
> Sun (SunOS, Solaris)
> Solbourne (SunOS variant)
One correction: Solbourne sells SPARC hardware only - you still have to buy
your OS from Solaris, I believe. (Sun (or some semi-open organization) does
certify it as "SPARC compliant" or some such thing)
> SCO (Xenix)
Does this still exist? I thought it had been supplanted by SCO System V.4
Besides: couldn't you find out how many "real" unix systems there were by
talking to USL (who license the code that all the other unix vendors use)?
(And that includes the _real_ real unix - BSD 4.3, but possibly not later
BSD versions)
--
Paul Tomblin - formerly {pt{omblin},news}@{geovision.}gvc.com
"Ok dear, Want me to call the bike shop and see if they'll sponsor your
mid-life crisis?" "Yeah. Ask 'em if they'll upgrade my shifters, too"
- Calvin's mom and dad
da...@gilly.uucp
> Arrrgh! You!!! Death threats! Mail bombs! May a plague of Phone
>solicitors descend upon you.
Heh heh heh. You don't seriously think *I'd* wait on the line for
one of those things now do you!?!? (Assuming they'd get past my modem,
log in, and send email from the guest account asking me to hold for
next available rep????)
> Now you'll probably tell me that it was run my my favorite little
>AT&T machine, the 3B1/7300/UNIX-PC. How low can we stoop?
Network of 3B2/400s, "Conversant"s (68K-based Unix systems with lots
of special phone hardware like voice synth, etc), and SNA link to
mainframes far away that dealt with the actual billing data.
(Oh - and Amdahl network concentrators/routers or something. I
was really surprised to see *that* name on a little relatively
unimportant box...)
> Perhaps you can do pennance by telling us how best to confuse the
>machines, perhaps so the operators will talk to each other.
Heh heh. Um... just keep them on the line with inane confusion or
questions. It's their long distance call...
Also - give out your local always-busy-test-number to all agencies
that might want to bill you for anything.
Jim Frost
>>Do you have any idea how many vendors that would cross? If you want
>>to know how many Windows licenses were sold, ask Microsoft. If you
>>want to know how many UNIX or UNIX work-alike licenses were sold, you
>>can get some of them by asking:
>> Sun (SunOS, Solaris)
>> Solbourne (SunOS variant)
>One correction: Solbourne sells SPARC hardware only - you still have to buy
>your OS from Solaris, I believe. (Sun (or some semi-open organization) does
>certify it as "SPARC compliant" or some such thing)
Solbourne's UNIX is based on SunOS. They have no Solaris offering
today. Since Solaris sells a true SMP version based on SunOS, and Sun
does not, one assumes that they did some real work on it. (I use it
regularly and it is blindingly fast -- not something you can say for
any Sun multiprocessor product right now.) I don't know how their
licensing works WRT Sun, so I don't know if Sun would count
Solbourne's licenses too. If Solbourne bought "right to sell" rather
than paying royalties Sun wouldn't be able to tell you.
>> SCO (Xenix)
>Does this still exist? I thought it had been supplanted by SCO System V.4
They sold them both last I knew. Doesn't matter; SCO is still another
vendor.
>Besides: couldn't you find out how many "real" unix systems there were by
>talking to USL (who license the code that all the other unix vendors use)?
Perhaps, but it would depend on the OEM deal that got cut between
vendors. Many systems derived from "real" UNIX are derived from early
versions that had far less restrictive licensing. I expect that
anything derived prior to SysV wouldn't have very good right-to-run
license accounting, and a lot of stuff derived from V7/SysIII.
You ought to be able to find how many SysV licenses are out there,
excluding any one-shot deals, though, and that'd be a nice starting
figure. Unfortunately the bigger companies are most likely to have
paid up-front for the right to sell rather than stringing it along on
a per-license basis.
>(And that includes the _real_ real unix - BSD 4.3, but possibly not later
>BSD versions)
Nope; USL doesn't get anything for BSD 4.3 right-to-run licenses.
They could tell you how many source licenses are out there, but that's
not a good indicator of the number of actual BSD 4.3 machines out
there. One source license can equate to a whole pile of actual
machines, and you can get the binaries even if you don't have a
license at all.
jim frost
ji...@centerline.com
Jim Frost
>today. Since Solaris sells a true SMP version based on SunOS, and Sun
Whoops, too many S words. I meant "Solbourne".
jim frost
ji...@centerline.com
Linus Torvalds
>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>
>>In <1993Sep1.1...@umibox.hanse.de>, ro...@umibox.hanse.de (Bernd Meyer) writes:
>>>Everybody, sing on:
>>>"Who'll ever need more than 640k? Who'll ever need more than 512M?"
>
>>Part of the reason for this limit is that OS/2 runs in the flat memory model
>>of the 386. Would you rather have segments?
>
>possible (guess what I have at this moment?).
I hope the "3M" is a typo, and should be "3G", and that the answer is
linux.. The 386 architecture allows 4GB flat segments, but some of the
VM space is usually reserved for the kernel (1GB in the case of linux,
giving you a 3GB user address space). And yes, it's useful to have that
large an address space, as you don't generally need to worry about
sparse memory addressing (used by the linux shared libs for example).
On the other hand, the 512MB limit of OS/2 (if this is true: I haven't
used OS/2) may be perfectly reasonable: there are times when it can be
useful to limit the normal data space of a process to enble sharing of
process memory (having the data spaces of several processes available at
the same time for easier copying etc). Or it may be due to bad design..
Linus
Mike Dahmus
X-Newsreader: IBM NewsReader/2 v1.00
Lines: 50
In <1993Sep3.0...@umibox.hanse.de>, ro...@umibox.hanse.de (Bernd Meyer) writes:
>mi...@schleppo.bocaraton.ibm.com (Mike Dahmus) writes:
>
>>In <1993Sep1.1...@umibox.hanse.de>, ro...@umibox.hanse.de (Bernd Meyer) writes:
>>>Everybody, sing on:
>>>"Who'll ever need more than 640k? Who'll ever need more than 512M?"
>
>>Part of the reason for this limit is that OS/2 runs in the flat memory model
>>of the 386. Would you rather have segments?
>
>No. Nice, flat 3M. This would at least last a few years longer. It's
>possible (guess what I have at this moment?).
OK, here's the scoop, from the OS/2 redbooks:
.. OS/2 Version 2.0 sets this limit to 512MB, reserving the linear address
range above this point for operating system use. The space above 512MB is
known as the *system region*.
This limitation on the size of the process address space is used by the
operating system to ensure protection of the system region from access by
applications...
The 16:16 addressing scheme [OS/2 1.x programs] allows access of up to 512MB
per process, since the local descriptor tables used in this model contain up
to 8192 entries, each of which can point to a segment of up to 64KB in size.
In order to ensure that there is no problem in coexisting 16-bit and 32-bit
applications under OS/2 V2.0, the maximum size of the process address space
has been set at 512MB. This means that all memory in the process address space
can be addressed using either the 16:16 or 0:32 [OS/2 2.x programs] addressing
scheme. This capability is important since it allows applications to be
composed of mixed 16-bit and 32-bit code, allows 32-bit applications to make
function calls to 16-bit service layers, and permits 16-bit applications
written for OS/2 Version 1.x to run unmodified, effectively allowing a "hybrid"
memory management environment.
-end-of-quote-
So, the memory available is <= 4GB because of the flat memory model, and
<= 512MB because of compatibility with 16-bit code. (I know some UNIX freaks
are going to whine and moan about that, but compatibility with existing
applications is an important concern in any operating system that hopes to
make money in the market, and a lot of big companies have a lot of 16-bit
OS/2 apps).
Bernd Meyer
>In <1993Sep1.1...@umibox.hanse.de>, ro...@umibox.hanse.de (Bernd Meyer) writes:
>>Everybody, sing on:
>>"Who'll ever need more than 640k? Who'll ever need more than 512M?"
>Part of the reason for this limit is that OS/2 runs in the flat memory model
>of the 386. Would you rather have segments?
No. Nice, flat 3M. This would at least last a few years longer. It's
possible (guess what I have at this moment?).
Bernie