Subject: Linux-Misc Digest #586
From: Digestifier <Linux-Misc-Request@senator-bedfellow.MIT.EDU>
To: Linux-Misc@senator-bedfellow.MIT.EDU
Reply-To: Linux-Misc@senator-bedfellow.MIT.EDU
Date:     Sun, 23 Jan 94 00:13:26 EST

Linux-Misc Digest #586, Volume #1                Sun, 23 Jan 94 00:13:26 EST

Contents:
  Re: Slackware by Satanists? (R. Day)
  Re: Linux vs. microkernel based Unix: questions? (Rennie Allen)
  Re: Fortran compilers for linux? (lih-sin the)
  Re: Fortran compilers for linux? (Erik Werner)
  Re: Global Alert For All: Jesus is Coming Soon (mksho@uno.edu)
  Re: Mathematica for Linux (mksho@uno.edu)
  Re: Fortran compilers for linux? (Thomas G. McWilliams)
  Re: Does the Mitsumi 2xSpeed CD-ROM work with Linux? (Nemesis)
  Fortran compilers for linux? (Tim Bandy)
  Re: IP routing problem in pl14 kernel? (Terry Dawson)
  Re: Linux vs. microkernel based Unix: questions? (Brandon S. Allbery)

----------------------------------------------------------------------------

Crossposted-To: sci.skeptic
From: rpjday@cuug.ab.ca (R. Day)
Subject: Re: Slackware by Satanists?
Date: Sat, 22 Jan 1994 14:09:46 GMT

(I have removed the followup to comp.os.linux.misc.)

jj, curmudgeon and all-around grouch (jj@alice.att.com) wrote:
: In article <CJG1wx.2FE@cuug.ab.ca> rpjday@cuug.ab.ca (R. Day) writes:
: >This is a totally nonsensical claim as there is a world of difference
: >between Wicca and Satanism.  Anyone making this kind of claim has only
: >proven how thorougly ignorant and gullible they really are.

: Not that I disagree at all with the above, but it amazes me how it
: is considered possible for someone who does not believe in any
: incarnation of ultimate evil to be a satanist.

There is a very precise characterization of Satanism according to its
founder, Anton LaVey, and its holy book, The Satanic Bible.  Quite
simply, the Church of Satan does not worship Satan.  In fact, the
Church of Satan does not even recognize a literal Satan, instead
using the term "Satan" to refer to a ->metaphorical<- adversary that
one is always struggling against.

In Arthur Lyons's book, "Satan Wants You," Lyons quotes sociologist
Marcello Truzzi,

  "The Church of Satan's philosophical  world view is really more
  accurately designed as an ideological than a religious one.
  The name "satanism" and its other seeming relations to Christianity
  are actually somewhat misleading, for these are mainly used in
  a symbolic sense (thus, Satan is simply the symbol of the Adversary,
  in this case to the dominant belief system of Christianity).
  Thus, the Church of Satan is not really a sect of Christianity
  in the same sense as are most present and past Satanic groups."

In the Satanic Bible, LaVey also makes it clear that the Church of
Satan does not, ->in any way<-, promote or condone the sort of 
distasteful behaviour that Christian propagandists have charged it with.

On p. 70, LaVey writes:

  "If you attempt to impose your sexual desires upon others who do
  not welcome your advances, you are infringing upon their sexual
  freedom.  Therefore, Satanism does not advocate rape, child
  molesting, sexual defilement of animals, or any other form of
  sexual activity which entails the participation of those who are
  unwilling or whose innocence or naivete would allow them to be
  intimidated or misguided into doing something against their wishes."

Once could only wish that numerous Catholic priests had such ethics.

R. Day
Vice-chair, Alberta Skeptics

------------------------------

From: rgallen@muug.mb.ca (Rennie Allen)
Subject: Re: Linux vs. microkernel based Unix: questions?
Date: Sat, 22 Jan 1994 18:25:17 GMT

In <1994Jan21.033046.27685@kf8nh.wariat.org> bsa@kf8nh.wariat.org (Brandon S. Allbery) writes:

>Okay, maybe I have time to reply to this now :-(  It's been tough trying to
>find time to do decent responses to messages this week...

>In article <1994Jan17.235809.16128@muug.mb.ca>, rgallen@muug.mb.ca (Rennie Allen) says:
>+---------------
>| In <1994Jan16.191756.1229@kf8nh.wariat.org> bsa@kf8nh.wariat.org (Brandon S. Allbery) writes:
>| >In article <1994Jan16.070815.1375@muug.mb.ca>, rgallen@muug.mb.ca (Rennie Allen) says:
>| >+---------------
>| >| In <1994Jan15.012134.4874@kf8nh.wariat.org> bsa@kf8nh.wariat.org (Brandon S. Allbery) writes:
>| >| >The *attributes* he was discussing are real-time architecture.  Nothing he
>| >| >mentioned requires a microkernel architecture; that QNX happens to *be* one is
>| >| 
>| >| Just what is a *realtime* architecture ?  Why would a *microkernel* realtime
>| >| O/S be magically different from a non-realtime O/S ?  Funny, you know, I can
>| 
>| >A *real-time* OS is different from a non-realtime OS.  "Real-time" does not
>| >automatically mean "microkernel".  There are non-real-time microkernels and
>| 
>| Please answer my question:  what is a (in your words) "realtime architecture" ?
>+------------->8

>Do I need to explain this?  I am talking about a kernel (of any kind) which is
>designed to be able to cope with external devices which require response from
>the host within a fixed (short) time frame.  This requires that the kernel be
>preemptable and reentrant, among other things.  A microkernel architecture can
>(and usually does) encompass most of the attributes which are required, but is
>not required to; it is possible to construct a microkernel which itself cannot
>be preempted and which consumes a significant percentage of CPU time itself
>instead of making it available for its server processes, even though the
>servers themselves are preemptable.  It may even be reasonable in some cases.

>I say "architecture" instead of "attributes" because it is a design decision
>which must be accounted for in all stages of kernel development and
>implementation; to my mind, "attributes" implies something which can be added
>or adjusted after the fact, but true real-time response cannot be achieved in
>that way.  The kernel must be designed from the ground up with the intent of
>providing either a microkernel architecture or a real-time architecture or
>both.  ---I grant that this may be more of a terminology problem than a real
>dispute.

OK.  I have never thought of realtime features as a part of the architecture
(I still don't really), but I understand your point (i.e. that these consider
ations must be made at design time).  My point is that you can procede from the
design point of realtime characteristics like pre-emption etc., and develop
either a monolithic or microkernel architecture; and that whichever route you
choose, the quality of the implementation will have a far greater impact on
performance than the choice of architecture.  Performance figures vary widely
between both monolithic, and microkernel implementations, just because *most*
microkernels demonstrate worse performance than a mono kernel on the same
hardware does not mean that this is due to the architecture; more likely it
is due to the fact that (in general) more effort has been put into the
implementation of mono kernels, and that people know "how to do" mono kernels.

>architectures were equivalent.  They clearly are not; the fact that you can
>compare a single attribute (context switch time) in a manner which is
>uniformly favorable to QNX indicates the difference between the requirements
>of a real-time architecture and a non-real-time architecture, and would be
>expected to favor the real-time system.  How does raw disk I/O compare?
>Filesystem performance?  (Linux without the cluster mods will probably come
>out deficient here...)

In this thread I have posted the figures for QNX file I/O, and network I/O and
have asked for postings of similar figures for Linux on the same hardware.  I
got one response, which showed that Linux compared about equal with QNX in the
file I/O category, as long as the cluster mods were applied.  No performance
figures for network I/O were posted.  Also I didn't post a figure for QNX
serial I/O processing, but here it is:

33 Mhz 486 16450 (yes 450 *not* the buffered 550) on both ports.

115 kbps continuous on 2 serial ports.

This test was done by writing a small program which received on a port using
read() and wrote on the other port using write(), therefore both input and
output processing were being measured.  An external computer wrote the serial
data at 115 Kbps into the receiving port on the machine being tested, while
a second external computer received the data.  No interrupts were lost,  and
the resulting data file on the second externel computer matched the original
file on the first external computer.  No flow control was enabled.

>In the single benchmark statistic (context switch time), yes.  How about a
>benchmark which includes other attributes?  How, for example, does the same
>hardware perform under QNX vs. a monolithic architecture running TPC-A with
>the same (aside from OS API considerations) database manager?

Transaction processing is one of QNX's forte's, Visa uses QNX to do on-line
credit card transaction processing.

>My second post in this thread clarified that:  by "equivalent" I meant
>"equivalent levels of optimization".  Both real-time and microkernel virtually
>require extreme optimization, in my experience; Linux, on the other hand,
>hasn't needed it that much for what it's used for.

If your saying that a micrkernel requires more effort to achieve reasonable
levels of performance, I'll buy it; but if your saying that a mono kernel has
an inherent ability for performance which is unattainable by I microkernel, I
disagree strongly.
 
>You're comparing an attribute which is of paramount importance in real-time
>systems:  context switch time.  (If the context switch time is high, you
>cannot respond quickly to a real-time device; you need to switch to the
>driver's context...)  It is also important to microkernels, but is not quite
>as important to them as it is to real-time response.

I compared an attribute which is of paramount importance in realtime between
two realtime operating systems (HP-RT & QNX), one a mono kernel and one a
microkernel, with the mono kernel running on a faster processor, how much more
conservative do you want me to be....

>I was also responding to the fact that "real-time" and "microkernel" were
>being used virtually interchangeably in your discussion.  This most definitely
>colored my perception of what you were saying, as it appeared to be intended
>to do so.

I think you misunderstood.  I was not using realtime and microkernel
"interchangeably", I was not referencing the realtime capability of QNX at all;
since I felt it was not relevent to the discussion of microkernel architecture.
I gather that you interpreted this as meaning that I considered QNX's realtime
abilities to be a function of the microkernel architecture;  I can see how you
drew this conclusion, but that was definately not my intent.  My original point
(way, way, back) was that given the quality of the implementation of Linux,
that if it were re-written as a microkernel, that it could be made to perform
better than it currently does.  

I never really got to state, *why* I think that it would perform better.
The reanet 
supported product, and microkernel architectures are inherently modular, the
modularity would allow faster development of the "O/S" (i.e. the services
which make up the operating system), while the kernel would remain small
enough (~ 10K), so that it could easily be optimized and maintained by 1 person
(Linus).  Due to the fact that filesystems would simply be user level
processes, improvements (such as the cluster patches which you reference),
would not have to go through the laborious process of incorporation into the
kernel, but could simply be obtained as just another MK-Linux executable.  This
would add stability to Linux, and allow performance optimizations to be easily
incorporated, therefore Linux would obtain higher levels of performance much
more rapidly.  The concept that microkernels have some inherent architectural
flaw which would limit the maximum obtainable performance, would obviously 
negate my argument, and therefore I felt I had to address this issue fully 
before I could explain the rest of my point.  I used QNX (as a superb example
of a microkernel implementation - and since I know it), to "prove" that there
is no inherent limit (theoretical discussions about architectures aside) to
performance within the architecture (assuming that things like memory
protection are going to be required in a mono kernel anyway).

email: rgallen@muug.mb.ca              mail:  Expert Technology Corporation   
QUICS: rgallen (613) 591-0934                 34 Riverstone Rd. 
Voice: (204) 339-8005                         Winnipeg, Manitoba, Canada R2V 4B2
Fax:   (204) 488-5943

------------------------------

From: tlihsin@hubcap.clemson.edu (lih-sin the)
Subject: Re: Fortran compilers for linux?
Date: 22 Jan 1994 15:23:13 -0500

ramos@engr.latech.edu (Alex Ramos) writes:

>Tim Bandy (timn8r@eelpout.micro.umn.edu) wrote:
>> I have searched high and low, yet found no trace of a fortran
>> compiler for linux.  If any of you out there in the great 
>> unknown know of where I can find one, that would be
>> simply Maaaahhhvelous.  
>> email to timn8r@mermaid.micro.umn.edu

   I use linux slackware, to compile a fortran program I just type:
   f77 prog.f. 
   Actually I think f77 is a shell script that automatically run f2c
   then gcc.

   lihsin

>A related question.... does anybody have binaries for matlab and pspice?
>(they're distributed in Fortran source code).


>Thanks.
>--
>Alex Ramos <ramos@engr.latech.edu> * This message is copyrighted material!
>Louisiana Tech University BSEE/Sr  * All rights reserved. No warranty, etc

------------------------------

From: werner@acsu.buffalo.edu (Erik Werner)
Subject: Re: Fortran compilers for linux?
Date: Sat, 22 Jan 1994 21:02:24 GMT

  On a related subject, I have f2c.  I figured that it would be easier to use
this than to learn C. When I try to execute f77, I get the error message

/usr/bin/f2c: can't load dynamic linker 'lib/ld.so'

I seem to get this error when 'make'ing other downloads.
could someone tell me what this means and maybe how to correct it?
oh, BTW, 0.99p10 and MCC

Erik
-- 
=====================================================================
Erik Werner                            
werner@destiny.eng.buffalo.edu             
=====================================================================

------------------------------

From: mksho@uno.edu
Subject: Re: Global Alert For All: Jesus is Coming Soon
Date: 22 Jan 1994 22:06:04 GMT
Reply-To: mksho@uno.edu



You know, Jesus was fun in the sack, but John the Baptist gave better head...


------------------------------

From: mksho@uno.edu
Crossposted-To: sci.math.symbolic
Subject: Re: Mathematica for Linux
Date: 22 Jan 1994 22:14:12 GMT
Reply-To: mksho@uno.edu

In article <1994Jan20.170315.849@cathy.ijs.si>, Andrej.Bauer@ijs.si (Andrej Bauer) writes:
>> Sorry, but I've done the same thing about a half year before.
>> You see: Nothing really happens ...
>> ( Get MuPaD ! )
>
>I don't want MuPad, Jacal or any other animal of the kind.
>I want Mathematica. I have to work with Mathematica (not that
>I mind). I am not going to port an entire OO package from Mathematica
>to something else.
>
>BTW, CAS developers should fear that a free clone of Mathematica
>for Linux will pop up one day if they just sit and wait :-)
>There seems to be an incredible amount of enthusiasts who
>write programs for free.

There is already some sort of Mathematica parser in LISP, one can find it
on peoplesparc.berkeley.edu, it might be of interest to someone...I thought
I read something about being able to use it as a front end to another CAS..
so maybe this thing as a front end for MAXIMA, w/GNUplot for graphics?
(haven't used GNUplot, dunno if this would be feasible..)


>
>Andrej
>
>-------------------
>Andrej.Bauer@IJS.si

------------------------------

From: tgm@netcom.com (Thomas G. McWilliams)
Subject: Re: Fortran compilers for linux?
Date: Sat, 22 Jan 1994 22:09:31 GMT

Tim Bandy (timn8r@eelpout.micro.umn.edu) wrote:
 > I have searched high and low, yet found no trace of a fortran
 > compiler for linux.  If any of you out there in the great 
 > unknown know of where I can find one, that would be
 > simply Maaaahhhvelous.  
 > email to timn8r@mermaid.micro.umn.edu

There is a Fortran compiler under development by the FSF. It is
due to be released a year or two after GNU Hurd. :-)

Thomas

------------------------------

From: peeper@yar.cs.wisc.edu (Nemesis)
Subject: Re: Does the Mitsumi 2xSpeed CD-ROM work with Linux?
Date: 22 Jan 94 22:57:30 GMT

j. hurwitz (jhurwitz@access.digex.net) wrote:


: I have one of the newer Mitsumi double speed CD-ROM drives, when Linux
: is booting up I see a message saying Mitsumi initialzation failed. I
: remember hearing that Linux was compatable with the Mitsumi drive.

: Will the Mitsumi double speed work with Linux?.

: Is anyone using it now?.

: Any help in the matter would be greatly appreciated.
: -----------------------------------------------------------------------
: - jhurwitz@digex.net ---------- LINUX, the choice of a GNU generation -
: -----------------------------------------------------------------------


If it is the mitsumi 2x drive that I am thinking of...  It is the same
as the Creative Labs cdrom drivers (for which there is a "driver" for
it on sunsite).  If it is the same drive, than the only problem would
really be to find out if the interface board is the SBPro interface or
the LaserMate.  (I found out which one mine is by trial and error)

The driver I use is: sbpcd04.tar.gz
(found on sunsite.unc.edu in /pub/Linux/kernel/cdrom?? or maybe system??)

I also saw that there is now a version 1 of sbpcd, but I have not tried it.

Craig Peeper
peeper@cs.wisc.edu

------------------------------

From: timn8r@eelpout.micro.umn.edu (Tim Bandy)
Subject: Fortran compilers for linux?
Date: 20 Jan 94 14:58:07 GMT

I have searched high and low, yet found no trace of a fortran
compiler for linux.  If any of you out there in the great 
unknown know of where I can find one, that would be
simply Maaaahhhvelous.  
email to timn8r@mermaid.micro.umn.edu

------------------------------

From: terryd@extro.ucc.su.OZ.AU (Terry Dawson)
Subject: Re: IP routing problem in pl14 kernel?
Date: Fri, 21 Jan 1994 01:58:52 GMT

bos@surfnet.nl (Erik-Jan Bos) writes:

>Dear c.o.l.m.,

>The IP number I use on my testnet is a class C network number from the
>beginning of a range, e.g. x.y.0.0. As soon as I populate the IP routing

What are x and y ?

Terry


-- 
--- Terry Dawson, terryd@extro.ucc.su.oz.au

------------------------------

From: bsa@kf8nh.wariat.org (Brandon S. Allbery)
Subject: Re: Linux vs. microkernel based Unix: questions?
Date: Sun, 23 Jan 1994 00:04:55 GMT

In article <1994Jan22.182517.435@muug.mb.ca>, rgallen@muug.mb.ca (Rennie Allen) says:
+---------------
| either a monolithic or microkernel architecture; and that whichever route you
| choose, the quality of the implementation will have a far greater impact on
| performance than the choice of architecture.  Performance figures vary widely
+---------------

But what if they have similar/"equivalent" quality?

| file I/O category, as long as the cluster mods were applied.  No performance
| figures for network I/O were posted.  Also I didn't post a figure for QNX
+------------->8

Linux's network I/O speeds are a known problem point:  net-2D is not
efficient.  I don't know about net-2E.

| serial I/O processing, but here it is:
| 33 Mhz 486 16450 (yes 450 *not* the buffered 550) on both ports.
| 115 kbps continuous on 2 serial ports.
+------------->8

Not surprising.  A real-time kernel can guarantee that the driver can service
the serial interrupt in a timely manner.  ---For that matter, Linux is no
slouch itself in this area:  although it can't *guarantee* that it will
respond in time, I ran two 16450s simultaneously at 9600 baud and one at 19200
for close to a year and had very few problems.

| If your saying that a micrkernel requires more effort to achieve reasonable
| levels of performance, I'll buy it; but if your saying that a mono kernel has
+------------->8

That's *exactly* what I've been saying.

| an inherent ability for performance which is unattainable by I microkernel, I
| disagree strongly.
+------------->8

Give me credit for some intelligence.  My point is that you need to put less
effort into a monolithic kernel to get a given performance level than you do
for a microkernel, barring special-purpose applications (e.g. as I mentioned
before, microkernels intrinsically have a leg up on monolithic kernels for
real-time processing because they already address many of the important
attributes).

| I gather that you interpreted this as meaning that I considered QNX's realtime
| abilities to be a function of the microkernel architecture;  I can see how you
| drew this conclusion, but that was definately not my intent.  My original point
+------------->8

Okay.  Then I think this thread can now be drawn to a close.  :-)

| modularity would allow faster development of the "O/S" (i.e. the services
| which make up the operating system), while the kernel would remain small
| enough (~ 10K), so that it could easily be optimized and maintained by 1 person
+------------->8

But maybe Linus would like to maintain more than just the microkernel itself?
:-) :-) :-)

++Brandon
-- 
Brandon S. Allbery         kf8nh@kf8nh.ampr.org          bsa@kf8nh.wariat.org
"MSDOS didn't get as bad as it is overnight -- it took over ten years
of careful development."  ---dmeggins@aix1.uottawa.ca

------------------------------


** FOR YOUR REFERENCE **

The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:

    Internet: Linux-Misc-Request@NEWS-DIGESTS.MIT.EDU

You can send mail to the entire list (and comp.os.linux.misc) via:

    Internet: Linux-Misc@NEWS-DIGESTS.MIT.EDU

Linux may be obtained via one of these FTP sites:
    nic.funet.fi				pub/OS/Linux
    tsx-11.mit.edu				pub/linux
    sunsite.unc.edu				pub/Linux

End of Linux-Misc Digest
******************************
