Newsgroups: comp.os.ms-windows.advocacy,comp.os.linux.misc
Path: gmd.de!newsserver.jvnc.net!howland.reston.ans.net!europa.eng.gtefsd.com!
uunet!cs.utexas.edu!swrinde!menudo.uh.edu!nuchat!frobozz!kevin
From: ke...@frobozz.sccsi.com (Kevin Brown)
Subject: NT versus Linux, the updated and expanded comparison chart, version 2
References: <930818233207.23008@lambada> < CC2pM1.6Lp@frobozz.sccsi.com>
Date: Sun, 22 Aug 1993 08:09:29 GMT
Organization: Frobozzco International
Message-ID: < CC5Int.Ao5@frobozz.sccsi.com>
Lines: 162

Based on the comments I've received in the mail and on this forum, here's the
second version of the NT versus Linux comparison chart:


For Immediate Release: NT versus Linux, a feature comparison
------------------------------------------------------------

Feature                     NT                          Linux
-------                     --                          -----

microkernel                 yes                         no (monolithic)
kernel size                 50k                         200k-500k

kernel+necessary
extensions to run	    ~500k+			200k-500k

loadable device
drivers                     yes                         yes (available)

true dynamic link
libraries (DLL)             yes                         yes (shared)

per-application DLLs        yes                         yes, but unnecessary
							due to paging.

integrated GUI              no (loadable module)        no (X11 executable
							add-on)

networked window support    no				yes (X11)

drag and drop               yes                         depends on GUI and/or
							application

IPC support                 DDE                         sockets for local
							and network IPC, fifos,
							pipes

max chars in file name      depends on filesystem       depends on filesystem.
							Most have 255.

max partition size          limited, but probably large 64 meg -~2 gig,
			    (probably 2 gig or more)	depending on
							filesystem.  Most have
							~2 gig.

C2 securiy                  yes (see comments) 		no (implementable.  You
			    				have the source, right?)

multiuser		    no				yes

process memory model        flat                        flat

multiple platforms          yes                         not currently.  680x0
							under development

runs unix apps              limited (source level,	almost all (source
			    limited POSIX subset	level)
			    only)

runs DOS apps               yes                         some, via emulator

runs Windows apps           yes                         not currently.  Under
							development.

kernel max addressable mem  2048 Gig                    4 Gig
proc max addressable mem    4 Gig (???)			4 Gig

multimedia support          DLLs                        application level only
							(via DLLs possible)

QIC-80 support              yes                         yes

minimum CPU		    386DX-33			386sx-16
min required mem            12 meg                      2 meg (4 meg w/ X)

min required disk space     60 meg                      10-20 meg (base system)
							40 meg (w/ X)

networking                  stable (reportedly     	stable
			    unstable in March beta)

price                       $110 (educational)          $69 (SLS), free via
		 	    ~$300 (normal price)	ftp or BBS.

source available	    no				yes
source price		    not applicable		free
freely redistributable	    no				yes

bug fix response time	    several months, minimum	immediate.  Several
							weeks at most.

application availability    limited to DOS, Windows	DOS, Unix (almost all
							source-available
							applications)

coexists with other OSes    yes				yes


Comments:

- I don't think the Linux kernel is set up to address more than 4 gig
  internally.  You'd have to play with segment registers to access more
  than that.  Not that it matters at this point.  By the time you have
  that much memory (core + swap), you're running more than what the
  fastest Pentium would be able to comfortably handle in most circumstances,
  and you'll probably want a faster bus anyway.

- Linux wasn't designed at the outset to port to every piece of hardware
  available.  If you want that attribute, then perhaps this comparison should 
  be between BSD and NT.  That Linux is being ported to the 680x0 is evidence 
  enough that it's at least somewhat portable.

- DDE is presumably higher level than the sockets, pipes, and other things
  that you use under Unix.  I don't know what this buys you, so I'll wait
  for more comments to come in.

- Alan Cox reports that his March beta of NT had unstable networking.  I
  don't know if it's been fixed.  So I've marked it as tentatively stable
  until I find out otherwise (thus giving NT the benefit of the doubt).

- In general, I've been conservative about the NT entries since I've never
  messed with it myself.  I've preserved the original values where there
  hasn't been comment.

- Linux partition sizes are not limited to the maximum size of a partition in
  the partition table.  A filesystem may reside on a device directly, thus
  bypassing the partition mechanism.  However, if you wish to partition a
  device, you're going to be limited by the DOS partition table mechanism.  I
  don't know if NT has its own partitioning mechanism.

- If NT doesn't have networked window support (i.e., client/server model
  windowing), then I don't see what the big hubbub about networking in NT
  is all about.  NFS, FTP, telnet, etc., are all nice and everything, but
  it's old technology.  So's client/server windowing, for that matter, but
  that kind of windowing technology is a huge win in a windowed networking
  environment, e.g. where you're running a bunch of workstations or, better
  yet, a fast machine with lots of X terminals hanging off it.  Of course,
  someone might be able to implement the X protocol for NT.  :-)

- NT has C2 security, but probably only because it's a single-user system!
  That simplifies security *a lot*.  As I don't know what's involved in
  implementing C2 security, I removed the "easily implementable" from the
  Linux entry.  Unless there are requirements that by definition eliminate
  multiuser systems, C2 security should be implementable under Linux, but
  you might have to do some headstanding.

- Being a microkernel is a performance loss on a single-processor standalone
  machine.  It's a win if you have multiple processors or if you have a
  network of machines that can share execution.  Since I don't know the
  details of the NT kernel, I don't know if network processor sharing
  is something that NT can take advantage of.  Networking would have to
  happen on a pretty low level to make that possible.  Linux was designed
  explicitly to run on PC clone hardware, so these considerations weren't, er,
  considered.  And since most of us happen to be running PC clone hardware,
  guess which OS gets the performance win?  Hint: it isn't NT.


-- 
Kevin Brown					ke...@frobozz.sccsi.com
This is your .signature virus: < begin 644 .signature (9V]T8VAA(0K0z end >
	    This is your .signature virus on drugs: <>
			Any questions?

Newsgroups: comp.os.linux.misc,comp.os.ms-windows.advocacy
Path: gmd.de!newsserver.jvnc.net!howland.reston.ans.net!agate!doc.ic.ac.uk!
uknet!cf-cm!cybaswan!iiitac
From: iii...@swan.pyr (Alan Cox)
Subject: Re: NT versus Linux
Message-ID: <1993Aug21.152742.20951@swan.pyr>
Organization: Swansea University College
References: <1993Aug19.224831.30686@kf8nh.wariat.org> <930819201354.23155@world> 
<252bq8$s5u@nz12.rz.uni-karlsruhe.de>
Date: Sat, 21 Aug 1993 15:27:42 GMT
Lines: 16

The whole NT v Linux thing is silly. Firstly they are very different 
products. NT is an expensive corporate system with (I'm told reliably)
very good paid for support. Linux is a free system for the masses.

I did find the reliability issue funny. On our NT march BETA (now destroyed)
setting the terminal program to telnet mode, and typing ^J (control J then
return) crashed NT. Granted it's a beta - but so is Linux.

I don't like microkernels and after comparing Linux v NT speeds I was even
more convinced of this. When the development tools exist that allow you
to write and compile a microkernel based system and get the same output
assembly code as compiling its macrokernel equivalent, then the microkernel
will be a good general idea.

ALan

Newsgroups: comp.os.ms-windows.advocacy,comp.os.linux.misc
Path: gmd.de!xlink.net!howland.reston.ans.net!agate!doc.ic.ac.uk!uknet!cf-cm!
cybaswan!iiitac
From: iii...@swan.pyr (Alan Cox)
Subject: Re: NT versus Linux, the updated and expanded comparison chart
Message-ID: <1993Aug21.155607.21462@swan.pyr>
Organization: Swansea University College
References: <930818233207.23008@lambada> < CC2pM1.6Lp@frobozz.sccsi.com>
Date: Sat, 21 Aug 1993 15:56:07 GMT
Lines: 111

In article < CC2pM1....@frobozz.sccsi.com> ke...@frobozz.sccsi.com (Kevin Brown) 
writes:
>For Immediate Release: NT versus Linux, a feature comparison
>------------------------------------------------------------
Oh no here we go again. This is fun
>
>Feature                     NT                          Linux
>-------                     --                          -----
>
>microkernel                 yes                         no (monolithic)
 ^-- Microkernels are a performance misfeature.
>kernel size                 50k                         350k-500k
 ^-- + about 0.5Mb on our March beta to get up and going. ^- 230K my home kernel
>loadable device
>drivers                     yes                         yes (available)
The NT ones go further than Linux in what can be loaded/not loaded
>true dynamic link
>libraries (DLL)             yes                         yes (shared)
>
>per-application DLLs        yes                         yes, if you're root :-)
No you don't need to be root these days. Anyone can build per application
shared libraries. However you don't need to. The Linux memory manager does it
all correctly anyway.
>integrated GUI              yes                         no (add-on, X)
 ^-- the NT GUI aint integrated - its just an optional loadable
 module that everyone uses.
>
>drag and drop               yes                         depends on GUI and/or
>							application
>
>DDE support                 yes                         sockets, local and
							network interprocess
							 communications
>
>max chars in file name      unlimited                   14-255, depending on
>							filesystem.  Most have
>							255.
							I can't see anything in
							the code giving a 255
						        limit just the current FS
>max partition size          unlimited                   64 meg -~2 gig,
 This is wrong: The NT structs ^^^
 have a max size they can return
 off system calls.
>							depending on
>							filesystem.  Most have
>							~2 gig.
>C2 securiy                  yes                         no (easily implemented)
                                                         ^-dunno how easy...
>memory model                flat                        flat
>
>multiple platforms          yes                         not currently.  680x0
>							under development
>
>runs unix apps              yes (source level,		yes (source level)
>			    limited POSIX subset
>			    only)
>
>runs DOS apps               yes                         yes, via emulator
>
>runs Windows apps           yes                         not currently.  Under
>							development.
>
>max addressable mem         2048 Gig                    4 Gig
   ^- But not per process.
>multimedia support          yes                         application level only
 ^- you are having MAJOR concept problems aren't you
 the multimedia support is just DLLs (shared libs).
>QIC-80 support              yes                         yes
>minimum CPU		    486				386sx
   ^-- NT is OK on a 386DX33
>min required mem            12 meg                      2 meg (4 meg w/ X)
>
>min required disk space     60 meg                      10-20 meg (base system)
>							40 meg (w/ X)
>networking                  stable                      stable
              ^-- our march beta wasn't!
>
>price                       $110 (educational)          $69 (SLS), free via
>		 	    ~$300 (normal price)	ftp or BBS.
>
>source available	    no				yes
>source price		    not applicable		free
>freely redistributable	    no				yes
>
>bug fix response time	    several months, minimum	immediate.  Several
>							weeks at most.
>
>application availability    limited to DOS, Windows	DOS, Unix (almost all
>							source-available
>							applications)
>
>coexists with other OSes    unknown(probably)		yes
                              ^- Fairly well
                              
Now how about adding

Networked Window Support	no			yes (X11)

>A couple of comments:  I don't know what DDE is, so I can't say if Linux
>has it or not.  Probably not.
DDE is interprocess comms. Linux uses sockets/pipes for the same purpose.
>Multimedia programs for Linux may exist.  I haven't heard of them, so I'll
>assume that they don't.  But why in the world would you want multimedia
>support in the KERNEL???  If the applications can run efficiently enough,
>you don't need it.
SLS comes with an MPEG player. The berkeley MPEG encoder compiles under
Linux. (Haven't run it yet). It has sound card support.
There is _NO_ multimedia support in the NT kernel, its shared library (DLL)
support.