Path: sparky!uunet!uunet.ca!canrem!dosgate![roger.ram...@canrem.com]
From: "roger ramsey" <roger.ram...@canrem.com>
Newsgroups: comp.os.os2.misc
Subject: if os/2's
Message-ID: <1992Jul6.1480.20771@dosgate>
Date: 6 Jul 92 17:11:22 EST
Reply-To: "roger ramsey" <roger.ram...@canrem.com>
Distribution: comp
Organization: Canada Remote Systems
Lines: 46
Don . Ext 5955 wrote on 07-02-92
D >What is NT supposed to do that OS/2 2.0 doesn't?
Symmetric multiprocessing. Treat the input queue as separate threads (no
single app monopolozes the OS). Isolate the OS exclusively from
applications (even device driver calls go through APIs). Be fully 32-bit
(unlike the hybrid 16- / 32-bit OS/2 2.0). And last but not least: run
Windows applications efficiently and quickly.
D >When was the last time MicroSoft wrote a Operating System from
D >scratch.
When was the last time IBM wrote a PC OS from scratch?
D >Who is willing to give up 68 Meg of HD space amd 8 Meg Ram for NT.
No one I know: NT's requirements will be in the same range as, if not
lighter than OS/2 2.0 (4Mb of that 8Mb is immediately reserved for
application workspace and the last reports I got indicated 40Mb
including the 10Mb startup swapfile).
D > Can you say Multi-threaded?
Yes: NT.
D > How about Pre-Emptive Multi-tasking?
Yes: NT.
Can you say Portable?
Can you say Alpha?
Can you say Familiar Widespread Interface?
D >Well, I've blown off a little steam for now........
Yup: it did sound like a lot of hot air...
Roj
--- WinQwk 2.0 a#0
--
Canada Remote Systems - Toronto, Ontario/Detroit, MI
World's Largest PCBOARD System - 416-629-7000/629-7044
Newsgroups: comp.os.os2.advocacy,comp.os.os2.misc
Path: sparky!uunet!timbuk.cray.com!hemlock.cray.com!bgm
From: b...@cray.com (Bert Moshier)
Subject: Re: if os/2's
Message-ID: <1992Jul6.212804.7776@hemlock.cray.com>
Lines: 68
Organization: Cray Research, Inc.
References: <1992Jul6.1480.20771@dosgate>
Distribution: comp
Date: 6 Jul 92 21:28:04 CDT
In article <1992Jul6.1480.20771@dosgate> "roger ramsey" <roger.ram...@canrem.com>
writes:
>Don . Ext 5955 wrote on 07-02-92
>
>D >What is NT supposed to do that OS/2 2.0 doesn't?
>
>Symmetric multiprocessing. Treat the input queue as separate threads (no
>single app monopolozes the OS). Isolate the OS exclusively from
>applications (even device driver calls go through APIs). Be fully 32-bit
>(unlike the hybrid 16- / 32-bit OS/2 2.0). And last but not least: run
>Windows applications efficiently and quickly.
>
Maybe someone can explain to me. I thought vaporware portable OS/2 was
suppose to do Symmetric multiporcessing. It is suppose to treat the input
queue as separate threads including PM. Be fully 32-bit, etc.
Of course, NT is really as much vaporware as is portable OS/2. WHY DON'T
PEOPLE COMPARE NT TO PORTABLE OS/2, instead of comparing it to OS/2 2.0 which
is real?
Oh, you say NT is not vaporware while portable OS/2 is vaporware? Sorry, but
both products are vaporware because I can't go out and buy either of them.
NT is not vaporware because MS is showing it off at tradeshows, while IBM
is not showing off portable OS/2. Sorry, but IBM just doesn't do business
that way, and that fact won't change what portable OS/2 does offer.
>D >When was the last time MicroSoft wrote a Operating System from
>D >scratch.
>
>When was the last time IBM wrote a PC OS from scratch?
What does that have to do with anything?
The fact is, today's machines are mainframes on the desktop. Desktop
mainframes **need, no demand** a real operating system.
Also IMHO, DOS is not something to point to and say "Look DOS is a great
operating system. Let us believe in MS for doing it." Fact is, MS did not
do DOS nor OS/2 alone. IBM did several operating system alone which are
good for the purpose they serve.
Fact is, IBM has research, experience and knowledge to call upon which MS
doesn't have available. IBM also has enough programmers to do just what
it takes to create portable OS/2 on time.
>
>D >Who is willing to give up 68 Meg of HD space amd 8 Meg Ram for NT.
>
>No one I know: NT's requirements will be in the same range as, if not
>lighter than OS/2 2.0 (4Mb of that 8Mb is immediately reserved for
>application workspace and the last reports I got indicated 40Mb
>including the 10Mb startup swapfile).
Of course, with a vaporware product I can't disagree with you. It is
whatever you say. It is also whatever I say. I say it will take what
MS said earlier this year. A 486 or fast 386 with 8MB RAM, 19" monitor
suggested, 1024x768 minimum suggested, 400 MB HD suggested, etc.
>Can you say Portable?
>
>Can you say Alpha?
Can you say portable OS/2 on the SUN Sparc, RS/6000, and Intel x86 line from
IBM?
Bert.
Newsgroups: comp.os.os2.advocacy
Path: sparky!uunet!gatech!ncar!uchinews!ellis!sip1
From: s...@ellis.uchicago.edu (Timothy F. Sipples)
Subject: Re: if os/2's
Message-ID: <1992Jul7.033423.19263@midway.uchicago.edu>
Sender: n...@uchinews.uchicago.edu (News System)
Reply-To: s...@midway.uchicago.edu
Organization: Dept. of Econ., Univ. of Chicago
References: <1992Jul6.1480.20771@dosgate>
Date: Tue, 7 Jul 1992 03:34:23 GMT
Lines: 84
In article <1992Jul6.1480.20771@dosgate> "roger ramsey" <roger.ram...@canrem.com>
writes:
>D >What is NT supposed to do that OS/2 2.0 doesn't?
>Symmetric multiprocessing. Treat the input queue as separate threads (no
>single app monopolozes the OS). Isolate the OS exclusively from
>applications (even device driver calls go through APIs). Be fully 32-bit
>(unlike the hybrid 16- / 32-bit OS/2 2.0). And last but not least: run
>Windows applications efficiently and quickly.
Again, I'll reiterate, "symmetric multiprocessing" is just two words
put together without hardware. There are no hardware standards that I
know of in this area -- everyone does it differently. If NT
implements this, it will be tied to one or a couple specific systems
(e.g. Compaq SystemPro) and that's that. (Also, IBM's new Model 295
will run a multiprocessing version of OS/2 2.0, according to IBM, by
December.)
As for the separation of the input queue, this is a nice feature, but
it would be incorrect to say that applications running under OS/2 2.0
can lock up the input queue. There is a timer to prevent this from
happening -- hit CTRL-ESC and you'll get a "this application is not
responding" message.
I'm not sure what your third point means.
If Microsoft plans to deliver 16-bit Windows, 16-bit DOS, and 16-bit
OS/2 compatibility, NT will be hybrid as well. No pure 32-bit
operating system can provide backward compatibility without some
16-bit code.
Microsoft has publicly stated, previously, that they do not expect
significant (if any) performance gains (in fact, on certain platforms
you might pay a penalty) with 16-bit Windows applications on NT.
>D >When was the last time MicroSoft wrote a Operating System from
>D >scratch.
>When was the last time IBM wrote a PC OS from scratch?
Anyone remember the PC version of VM/CMS?
>D >Who is willing to give up 68 Meg of HD space amd 8 Meg Ram for NT.
>No one I know: NT's requirements will be in the same range as, if not
>lighter than OS/2 2.0 (4Mb of that 8Mb is immediately reserved for
>application workspace and the last reports I got indicated 40Mb
>including the 10Mb startup swapfile).
Microsoft has not stated this, I'm afraid. The NT SDK, sans DOS,
Windows 16 bit, and OS/2 compatibility layers, mind you, has minimum
requirements of CD-ROM, 386DX-33, 12 MB of RAM ("16 MB recommended"),
and 100 MB hard drive ("300 MB recommended"), per Infoworld 7/6/92. I
expect they'll trim these requirements down (and it is a developer's
platform), but do you really expect adding three compatibility layers
is going to have no impact on requirements?
>D > Can you say Multi-threaded?
>Yes: NT.
>D > How about Pre-Emptive Multi-tasking?
>Yes: NT.
>Can you say Portable?
Yes, OS/2. It's a poorly kept secret you'll see this on the Motorola
PowerPC before too long.
>Can you say Alpha?
This month's NT SDK will not include any Alpha support.
>Can you say Familiar Widespread Interface?
I think it is generally acknowledged by even Windows partisans that
the OS/2 Workplace Shell is novel, useful, and conceptually superior
to the Windows interface, which is now, frankly, growing a little long
in the tooth.
Even Microsoft recognizes that (the so-called "Cairo" project way over
the horizon).
Note that you should post your opinions on either product, OS/2 or
Windows, in comp.os.os2.advocacy, not in comp.os.os2.misc.
--
Get the OS/2 FREQ. ASKED QUESTIONS LIST | Timothy F. Sipples
from 128.123.35.151, anonymous ftp, | Internet: s...@ellis.uchicago.edu
directory pub/os2/all/faq, or from | IBM VNET Alias: SIPPLES AT BITNET
LISTS...@BLEKUL11.BITNET (send "HELP"). | Dept. of Econ., U. Chicago, 60637
Newsgroups: comp.os.os2.advocacy,comp.os.os2.misc
Path: sparky!uunet!mcsun!ieunet!wsl!jja
From: j...@wsl.ie (John J. Allen)
Subject: Re: if os/2's
Message-ID: <1992Jul7.145435.6247@wsl.ie>
Organization: Workhorse Systems Limited, Dublin, Ireland
References: <1992Jul6.212804.7776@hemlock.cray.com>
Distribution: comp
Date: Tue, 7 Jul 1992 14:54:35 GMT
Lines: 17
OS/2 2. something >= Windows NT 1. something
Network PM (a la X Windows) Local windowing only
SMP SMP
Workplace Shell Program Manager
Portable Portable
Runs OS/2-16,OS2-32,PM16,PM32,Win16,Win32,WinNT,DOS
As you can see I expect that subsequent releases of both OS/2 and NT
(is that OS/2 NT or Windows/NT) will support applications written for
both environments (with probably some restrictions) and as for features
they will provide much the same.
To be honest (I'd prefer OS/2 to dominate) I'm just really fed up with
pushing OS/2 and getting Win/NT pushed back at me all the time.
I personally doubt the NT will fail or just go away but I also doubt
that it will just dominate.
Newsgroups: comp.os.os2.advocacy,comp.os.os2.misc
Path: sparky!uunet!microsoft!wingnut!philipla
From: phili...@microsoft.com (Phil Lafornara)
Subject: Portable?
Message-ID: <1992Jul08.185421.23445@microsoft.com>
Date: 08 Jul 92 18:54:21 GMT
Organization: Microsoft Corporation
References: <1992Jul6.212804.7776@hemlock.cray.com> <1992Jul7.145435.6247@wsl.ie>
Distribution: comp
Lines: 44
In article <1992Jul7.145435.6...@wsl.ie> j...@wsl.ie (John J. Allen) writes:
>OS/2 2. something >= Windows NT 1. something
>Network PM (a la X Windows) Local windowing only
>SMP SMP
>Workplace Shell Program Manager
>Portable Portable
> Runs OS/2-16,OS2-32,PM16,PM32,Win16,Win32,WinNT,DOS
>
Just a nit - when did OS/2 2.0 become portable? It was my
understanding that the bulk of the code was written in assembly,
which is notoriously non-portable. Portability isn't something
that can be retrofitted onto a product.
>As you can see I expect that subsequent releases of both OS/2 and NT
>(is that OS/2 NT or Windows/NT) will support applications written for
>both environments (with probably some restrictions) and as for features
>they will provide much the same.
I'm not altogether certain that this is true. Some recent
renegotiations of the licensing agreements between IBM and MS
have raised some new questions as to what is going to be
compatible with what.
>To be honest (I'd prefer OS/2 to dominate) I'm just really fed up with
>pushing OS/2 and getting Win/NT pushed back at me all the time.
>I personally doubt the NT will fail or just go away but I also doubt
>that it will just dominate.
This all remains to be seen. For obvious reasons, I hope that
NT will dominate, but I have a vested interest. I've used NT, and
I've used a beta of OS/2 - both seem to be fairly good products.
The market will decide.
-Phil
-------------------------------------------------------------------------
Phil Lafornara 1 Microsoft Way
phili...@microsoft.com Redmond, WA 98052-6399
Note: Microsoft doesn't even _know_ that these are my opinions. So there.
Newsgroups: comp.os.os2.advocacy,comp.os.os2.misc
Path: sparky!uunet!seas.smu.edu!mikek
From: mi...@seas.smu.edu (Michael Kaply)
Subject: Re: Portable?
Message-ID: <1992Jul8.231509.21642@seas.smu.edu>
Sender: n...@seas.smu.edu (USENET News System)
Nntp-Posting-Host: express.seas.smu.edu
Organization: School of Engineering and applied science; S.M.U.; Dallas, Tx
References: <1992Jul6.212804.7776@hemlock.cray.com> <1992Jul7.145435.6247@wsl.ie>
<1992Jul08.185421.23445@microsoft.com>
Distribution: comp
Date: Wed, 8 Jul 1992 23:15:09 GMT
Lines: 19
In article <1992Jul08.185421.23...@microsoft.com> phili...@microsoft.com
(Phil Lafornara) writes:
>In article <1992Jul7.145435.6...@wsl.ie> j...@wsl.ie (John J. Allen) writes:
>>OS/2 2. something >= Windows NT 1. something
>>Network PM (a la X Windows) Local windowing only
>>SMP SMP
>>Workplace Shell Program Manager
>>Portable Portable
>> Runs OS/2-16,OS2-32,PM16,PM32,Win16,Win32,WinNT,DOS
>>
>
> Just a nit - when did OS/2 2.0 become portable? It was my
>understanding that the bulk of the code was written in assembly,
>which is notoriously non-portable. Portability isn't something
>that can be retrofitted onto a product.
>
OS/2 2..0 is written in mainly C. Where did you get the idea it was written in
assembly? This has been hashed and rehashed many times.
Path: sparky!uunet!usc!zaphod.mps.ohio-state.edu!swrinde!ringer!mlevis
From: mle...@ringer.cs.utsa.edu (Mike Levis)
Newsgroups: comp.os.os2.advocacy,comp.os.os2.misc
Subject: Re: Portable?
Message-ID: <1992Jul8.233508.26285@ringer.cs.utsa.edu>
Date: 8 Jul 92 23:35:08 GMT
References: <1992Jul6.212804.7776@hemlock.cray.com> <1992Jul7.145435.6247@wsl.ie>
<1992Jul08.185421.23445@microsoft.com>
Distribution: comp
Organization: University of Texas at San Antonio
Lines: 32
In article <1992Jul08.185421.23...@microsoft.com> phili...@microsoft.com
(Phil Lafornara) writes:
>In article <1992Jul7.145435.6...@wsl.ie> j...@wsl.ie (John J. Allen) writes:
>>OS/2 2. something >= Windows NT 1. something
^^^^^^^^^
>>Network PM (a la X Windows) Local windowing only
>>SMP SMP
>>Workplace Shell Program Manager
>>Portable Portable
>> Runs OS/2-16,OS2-32,PM16,PM32,Win16,Win32,WinNT,DOS
>>
>
> Just a nit - when did OS/2 2.0 become portable? It was my
>understanding that the bulk of the code was written in assembly,
>which is notoriously non-portable. Portability isn't something
>that can be retrofitted onto a product.
John Allen said ``2. something'', not ``2.0''. I don't think anyone
is saying 2.0 is portable now, although a future version is planned
to be.
Also, 2.0 is mostly written in C (source: _The Design of OS/2_ by
Deitel and Kogan, from Addison-Wesley).
>Phil Lafornara 1 Microsoft Way
--
======= Mike Levis mle...@ringer.cs.utsa.edu =======
:: ftp ftp-os2.nmsu.edu for OS/2 software and info :: ./\. .--.
:: ftp syrinx.umd.edu for Rush lyrics and info :: <OS/2> (OS/2)
:::: Reader of the OS/2 Freq. Asked Questions List :::: "\/" "--"
Newsgroups: comp.os.os2.advocacy,comp.os.os2.misc
Path: sparky!uunet!microsoft!wingnut!philipla
From: phili...@microsoft.com (Phil Lafornara)
Subject: Re: Portable?
Message-ID: <1992Jul09.182426.20717@microsoft.com>
Date: 09 Jul 92 18:24:26 GMT
Organization: Microsoft Corporation
References: <1992Jul7.145435.6247@wsl.ie> <1992Jul08.185421.23445@microsoft.com>
<1992Jul8.231509.21642@seas.smu.edu>
Distribution: comp
Lines: 19
In article <1992Jul8.231509.21...@seas.smu.edu> mi...@seas.smu.edu (Michael Kaply)
writes:
>>
>OS/2 2..0 is written in mainly C. Where did you get the idea it was written in
>assembly? This has been hashed and rehashed many times.
>
From people that work with the source code daily. Where did you
get your information?
-Phil
-------------------------------------------------------------------------
Phil Lafornara 1 Microsoft Way
phili...@microsoft.com Redmond, WA 98052-6399
Note: Microsoft doesn't even _know_ that these are my opinions. So there.
Path: sparky!uunet!zaphod.mps.ohio-state.edu!swrinde!ringer!mlevis
From: mle...@ringer.cs.utsa.edu (Mike Levis)
Newsgroups: comp.os.os2.advocacy,comp.os.os2.misc
Subject: Re: Portable?
Message-ID: <1992Jul9.222215.19296@ringer.cs.utsa.edu>
Date: 9 Jul 92 22:22:15 GMT
References: <1992Jul08.185421.23445@microsoft.com>
<1992Jul8.231509.21642@seas.smu.edu> <1992Jul09.182426.20717@microsoft.com>
Followup-To: comp.os.os2.advocacy
Distribution: comp
Organization: University of Texas at San Antonio
Lines: 31
In article <1992Jul09.182426.20...@microsoft.com> phili...@microsoft.com
(Phil Lafornara) writes:
>In article <1992Jul8.231509.21...@seas.smu.edu> mi...@seas.smu.edu (Michael Kaply)
writes:
>>>
>>OS/2 2..0 is written in mainly C. Where did you get the idea it was written in
>>assembly? This has been hashed and rehashed many times.
>>
>
> From people that work with the source code daily. Where did you
>get your information?
>
Look on page 118 in _The Design of OS/2_:
"The 32-bit kernel is written in C for portability, although it
contains portions that must be written in assembler on any
architecture."
Who are these ``people that work with the source code daily''?
Are these MS guys or IBM guys? Unless NT will have 32-bit OS/2
code in it, why would MS guys need to work with this code every
day?
>
>Phil Lafornara 1 Microsoft Way
Follow up to comp.os.os2.advocacy!
--
======= Mike Levis mle...@ringer.cs.utsa.edu =======
:: ftp ftp-os2.nmsu.edu for OS/2 software and info :: ./\. .--.
:: ftp syrinx.umd.edu for Rush lyrics and info :: <OS/2> (OS/2)
:::: Reader of the OS/2 Freq. Asked Questions List :::: "\/" "--"
Newsgroups: comp.os.os2.advocacy,comp.os.os2.misc
Path: sparky!uunet!microsoft!hexnut!gordonl
From: gord...@microsoft.com (Gordon Letwin)
Subject: Re: Portable?
Message-ID: <1992Jul10.235753.14885@microsoft.com>
Date: 10 Jul 92 23:57:53 GMT
Organization: Microsoft Corporation
References: <1992Jul7.145435.6247@wsl.ie> <1992Jul08.185421.23445@microsoft.com>
<1992Jul8.233508.26285@ringer.cs.utsa.edu>
Distribution: comp
Lines: 23
In article <1992Jul8.233508.26...@ringer.cs.utsa.edu> mle...@ringer.cs.utsa.edu
(Mike Levis) writes:
>Also, 2.0 is mostly written in C (source: _The Design of OS/2_ by
>Deitel and Kogan, from Addison-Wesley).
Well, the word "mostly" appears to be getting quite a work out. OS/2 2.0
has over 1.3 million lines of assembly language code in it. The source
for this statement is my personally counting them (via "wc") from the
actual shipped sources.
Maybe they mean that "most" of the bytes of RAM used are being used by
code written in C. This would be a reflection on the quality of their
C programming and the C compiler.
From a portability viewpoint, which was in fact the root basis for this
discussion, the 1.3 million lines of ASM make the "mostly" quoted above
a rather misleading term...
gordon letwin
a spokesman for LWPI
but not for Microsoft
Newsgroups: comp.os.os2.advocacy
Path: sparky!uunet!microsoft!hexnut!gordonl
From: gord...@microsoft.com (Gordon Letwin)
Subject: Re: Portable?
Message-ID: <1992Jul11.001318.15210@microsoft.com>
Date: 11 Jul 92 00:13:18 GMT
Organization: Microsoft Corporation
References: <1992Jul8.231509.21642@seas.smu.edu>
<1992Jul09.182426.20717@microsoft.com> <1992Jul9.222215.19296@ringer.cs.utsa.edu>
Distribution: comp
Lines: 34
In article <1992Jul9.222215.19...@ringer.cs.utsa.edu> mle...@ringer.cs.utsa.edu
(Mike Levis) writes:
>
>Look on page 118 in _The Design of OS/2_:
> "The 32-bit kernel is written in C for portability, although it
> contains portions that must be written in assembler on any
> architecture."
Ah ha! *This* explains the confusion!
They say that the "32-bit kernel is written in C". This may well be
true, I haven't checked. The kicker is, though, much of the kernel
ISN'T 32-bit. The FAT file system, for example, is virtually unchanged
from the 286 versions of OS/2. I mention that specifically because I'm
doing file system work now and wanted to see IBM's changes. There weren't
any. I haven't bothered to take the time to do a detailed survey of
which parts of the kernel are still in 16-bit, but there apparently
are many other major 16-bit parts unless the authors of _The Design of OS/2_
are fibbing.... (i.e., if all 32-bit is in C, and there's 1.3 million
of ASM, then there must be a lot of 16-bit stuff remaining...)
Also please note that the 1.3 million lines of ASM includes all of the
"system build" tree; this may well contain sources for DLLs, perhaps
even utilities that aren't strictly part of the kernel. There was too
much of it to bother sorting it out. I also didn't count the WPS; it
was huge and there didn't appear to be much assembly in it.
Remember that the original topic was portability. The issue of kernel
vs non-kernel doesn't matter; if it's ASM in the product and you want
to port to a different processor, you've got to rewrite it. Therefore
I didn't bother to try to define and distinguish kernel vs. non-kernel,
that term doesn't have a precise definition anyway.
gordon letwin
not a microsoft spokesperson
Newsgroups: comp.os.os2.advocacy
Path: sparky!uunet!uchinews!ellis!sip1
From: s...@ellis.uchicago.edu (Timothy F. Sipples)
Subject: Re: Portable?
Message-ID: <1992Jul11.050428.26185@midway.uchicago.edu>
Sender: n...@uchinews.uchicago.edu (News System)
Reply-To: s...@midway.uchicago.edu
Organization: Dept. of Econ., Univ. of Chicago
References: <1992Jul09.182426.20717@microsoft.com>
<1992Jul9.222215.19296@ringer.cs.utsa.edu> <1992Jul11.001318.15210@microsoft.com>
Distribution: comp
Date: Sat, 11 Jul 1992 05:04:28 GMT
Lines: 78
In article <1992Jul11.001318.15...@microsoft.com> gord...@microsoft.com
(Gordon Letwin)
writes:
>They say that the "32-bit kernel is written in C". This may well be
>true, I haven't checked. The kicker is, though, much of the kernel
>ISN'T 32-bit. The FAT file system, for example, is virtually unchanged
>from the 286 versions of OS/2. I mention that specifically because I'm
>doing file system work now and wanted to see IBM's changes. There weren't
>any. I haven't bothered to take the time to do a detailed survey of
>which parts of the kernel are still in 16-bit, but there apparently
>are many other major 16-bit parts unless the authors of _The Design of OS/2_
>are fibbing.... (i.e., if all 32-bit is in C, and there's 1.3 million
>of ASM, then there must be a lot of 16-bit stuff remaining...)
Yes, but why would you spend any time rewriting FAT for 32-bit when
(a) the 16-bit code has been exhaustively tested and is reliable; (b)
this area is probably not the first place you'd look for performance
gains in going from 16- to 32-bit; (c) FAT is meant for backward
compatibility with native DOS -- on a portable operating system why
would you ever run FAT on, say, that PowerPC machine? You're not
about to stick a DOS diskette in Drive A on such a system. (This sort
of turns your argument on its head, namely that rewriting FAT for
32-bit is a _waste_ of time -- the time is better spent elsewhere --
because the DOS file system isn't exactly something that needs to be
portable to other machines.)
Incidently, do yourself a favor with NT: don't bother incorporating
FAT into the MIPS and Alpha incarnations. What's the point?
(Actually, I'm sure someone will come up with an interesting idea as
to why FAT should be supported on NT/Alpha, NT/MIPS. Thoughts?)
>Also please note that the 1.3 million lines of ASM includes all of the
>"system build" tree; this may well contain sources for DLLs, perhaps
>even utilities that aren't strictly part of the kernel. There was too
>much of it to bother sorting it out. I also didn't count the WPS; it
>was huge and there didn't appear to be much assembly in it.
Ah, well, there you go. An awful lot of \OS2\* consists of family
mode stuff which, of necessity, needs to be 16-bit. There's also an
OS/2 1.3 backward compatibility issue with some of that material.
>Remember that the original topic was portability. The issue of kernel
>vs non-kernel doesn't matter; if it's ASM in the product and you want
>to port to a different processor, you've got to rewrite it. Therefore
>I didn't bother to try to define and distinguish kernel vs. non-kernel,
>that term doesn't have a precise definition anyway.
But if most of it is FAT and accessory utilities, it seems they've
done most of the groundwork already, correct?
I have to say, though, that this argument is all but pointless.
Let's put it this way, OS/2 2.0 is the most 32-bit backward
compatible, mass market operating system on the market today. I have
faith that Microsoft will counter with an offering of their own in the
future (in fact, I'm counting on it -- competition is healthy). But
I'm afraid, in the meantime, IBM offers a product which, in technical
terms, is far advanced of anything in the Microsoft stable at present.
I mean, Microsoft did release Windows 3.1 within a month or so of OS/2
2.0, and we're not exactly talking 32bitness on that update, are we?
Incidently, one might argue that IBM already has a full, POSIX
compliant (better than that, actually -- POSIX does not equal Unix),
32-bit, portable operating system. It's called AIX. It's got Display
Postscript, even (that's genuine Adobe). And you can buy it for the
PS/2, RS/6000, and even a giant IBM mainframe (ES/9000). It runs
windows. X-Windows, that is. Even all those workstation applications
(like FrameMaker). You can even make it DOS compatible (with various
DOS emulators).
It supports symmetric multiprocessing (on RS/6000 and above, albeit at
the process level, I would assume), fault tolerance (if provided with
appropriate hardware), and lots of other nifty features.
--
Get the OS/2 FREQ. ASKED QUESTIONS LIST | Timothy F. Sipples
from 128.123.35.151, anonymous ftp, | Internet: s...@ellis.uchicago.edu
directory pub/os2/all/faq, or from | Dept. of Econ., U. Chicago, 60637
LISTS...@BLEKUL11.BITNET (send "HELP"). | H. Ross Perot in '92!
Newsgroups: comp.os.os2.advocacy
Path: sparky!uunet!zaphod.mps.ohio-state.edu!magnus.acs.ohio-state.edu!smsmith
From: smsm...@magnus.acs.ohio-state.edu (Stephen M Smith)
Subject: Re: Portable?
Message-ID: <1992Jul12.034800.755@magnus.acs.ohio-state.edu>
Sender: n...@magnus.acs.ohio-state.edu
Nntp-Posting-Host: top.magnus.acs.ohio-state.edu
Organization: The Ohio State University
Distribution: comp
Date: Sun, 12 Jul 1992 03:48:00 GMT
Lines: 21
gord...@microsoft.com (Gordon Letwin) writes:
>The FAT file system, for example, is virtually unchanged
>from the 286 versions of OS/2.
According to everything I've read, the FAT file system under OS/2 2.0
has "enhancements" and therefore it is called "super-FAT". I don't
know what that means exactly, but I doubt that it's the same as OS/2 1.3's
FAT.
Here's a lowly example: Running Might and Magic 3 under plain DOS
gives very slow performance because the main file is 3.5 MB in size.
Accessing a character's items takes 3-4 seconds. But running the
same game under OS/2 makes it fly: the items list pops up in less
than a second, and this happens whether the game is loaded on a
FAT drive or HPFS drive.
Steve Smith | __|__ | " #*&<-[89s]*(k#$@-_=//a2$]'+=.(2_&*%>,,@
<smsm...@magnus.acs. | | | {7%*@,..":27g)-=,#*:.#,/6&1*.4-,l@#9:-) "
ohio-state.edu> | | |
BTW, WYSInaWYG | | | --witty.saying.ARC
Newsgroups: comp.os.os2.advocacy
Path: sparky!uunet!ftpbox!motsrd!mothost!white!rtsg.mot.com!lido16!wiegand
From: wieg...@rtsg.mot.com (Robert Wiegand)
Subject: Re: Portable?
Message-ID: <wiegand.711122759@lido16>
Sender: n...@rtsg.mot.com
Nntp-Posting-Host: lido16
Reply-To: motcid!wieg...@uunet.uu.net
Organization: Motorola Inc., Cellular Infrastructure Group
References: <1992Jul8.231509.21642@seas.smu.edu>
<1992Jul09.182426.20717@microsoft.com> <1992Jul9.222215.19296@ringer.cs.utsa.edu>
<1992Jul11.001318.15210@microsoft.com>
Distribution: comp
Date: Tue, 14 Jul 1992 14:05:59 GMT
Lines: 49
gord...@microsoft.com (Gordon Letwin) writes:
>In article <1992Jul9.222215.19...@ringer.cs.utsa.edu> mle...@ringer.cs.utsa.edu
(Mike Levis) writes:
>>
>>Look on page 118 in _The Design of OS/2_:
>> "The 32-bit kernel is written in C for portability, although it
>> contains portions that must be written in assembler on any
>> architecture."
>Ah ha! *This* explains the confusion!
>They say that the "32-bit kernel is written in C". This may well be
>true, I haven't checked. The kicker is, though, much of the kernel
>ISN'T 32-bit. The FAT file system, for example, is virtually unchanged
>from the 286 versions of OS/2. I mention that specifically because I'm
>doing file system work now and wanted to see IBM's changes. There weren't
>any. I haven't bothered to take the time to do a detailed survey of
>which parts of the kernel are still in 16-bit, but there apparently
>are many other major 16-bit parts unless the authors of _The Design of OS/2_
>are fibbing.... (i.e., if all 32-bit is in C, and there's 1.3 million
>of ASM, then there must be a lot of 16-bit stuff remaining...)
>Also please note that the 1.3 million lines of ASM includes all of the
>"system build" tree; this may well contain sources for DLLs, perhaps
>even utilities that aren't strictly part of the kernel. There was too
>much of it to bother sorting it out. I also didn't count the WPS; it
>was huge and there didn't appear to be much assembly in it.
>Remember that the original topic was portability. The issue of kernel
>vs non-kernel doesn't matter; if it's ASM in the product and you want
>to port to a different processor, you've got to rewrite it. Therefore
>I didn't bother to try to define and distinguish kernel vs. non-kernel,
>that term doesn't have a precise definition anyway.
> gordon letwin
> not a microsoft spokesperson
Asm. code thats in anything used for 80X86 specific stuff shouldn't be
counted. You wouldn't use this stuff when you port to another processor
anywhy. The same goes for the FAT file system - why would anyone want to
port that to another machine? The important part is the kernal, which
you agreed is mostly in C.
--
------------------------------------------------------------------------------
Robert Wiegand - Motorola Inc.
uunet!motcid!wiegand
Disclamer: I didn't do it - I was somewhere else at the time.
Newsgroups: comp.os.os2.advocacy
Path: sparky!uunet!microsoft!hexnut!gordonl
From: gord...@microsoft.com (Gordon Letwin)
Subject: Re: Portable?
Message-ID: <1992Jul14.184455.1788@microsoft.com>
Date: 14 Jul 92 18:44:55 GMT
Organization: Microsoft Corporation
References: <1992Jul12.034800.755@magnus.acs.ohio-state.edu>
Distribution: comp
Lines: 25
In article <1992Jul12.034800....@magnus.acs.ohio-state.edu>
smsm...@magnus.acs.ohio-state.edu (Stephen M Smith) writes:
>gord...@microsoft.com (Gordon Letwin) writes:
>
>>The FAT file system, for example, is virtually unchanged
>>from the 286 versions of OS/2.
>
>According to everything I've read, the FAT file system under OS/2 2.0
>has "enhancements" and therefore it is called "super-FAT". I don't
>know what that means exactly, but I doubt that it's the same as OS/2 1.3's
>FAT.
The FAT file system *is* virtually unchanged. The "Super-Fat" thing refers
to the replacement of the old FAT cache code with - essentially - the
cache code taken from HPFS. I posted about this extensively earlier;
the HPFS design was used virtually unchanged (having FAT as a client instead
of HPFS required one meaningful change which IBM screwed up :-) )
I won't repeat those posts; the bottom line is that the performance of
a FAT disk is improved due to the use of the HPFS cache design (did they
use the code? I dunno... they should have, no need to recode the wheel)
but the file system itself - the code that handles open/close/read/write -
is virtually unchanged 16 bit assembly.
gordon letwin
not a spokesperson for anyone
Newsgroups: comp.os.os2.advocacy
Path: sparky!uunet!microsoft!hexnut!gordonl
From: gord...@microsoft.com (Gordon Letwin)
Subject: Re: Portable?
Message-ID: <1992Jul21.225048.4454@microsoft.com>
Date: 21 Jul 92 22:50:48 GMT
Organization: Microsoft Corporation
References: <1992Jul9.222215.19296@ringer.cs.utsa.edu>
<1992Jul11.001318.15210@microsoft.com> <wiegand.711122759@lido16>
Distribution: comp
Lines: 19
In article <wiegand.711122759@lido16> motcid!wieg...@uunet.uu.net writes:
>
>Asm. code thats in anything used for 80X86 specific stuff shouldn't be
>counted. You wouldn't use this stuff when you port to another processor
>anywhy.
True, in general. But most of the code that you get to discard has
it's equivalent for the other processor. So you don't have to manage
the 386's page tables, but you do have to manage someone else's page tables...
So the amount of ASM is still a reasonable measure of porting effort.
In fact, it's faster to port stuff that's just in ASM because of speed
issues then it is to rewrite stuff for a new environment; rewriting introduces
all new bugs, whereas just transcoding stuff avoids many of the bugs you'd
get from all new code.
Gordon Letwin
not a spokesperson for anyone