See also: last week's Back page page.

Linux links of the week

Tired of silly benchmarks? Then don't head over to the /dev/null benchmark page. This benchmark, also known as "Fhlushstone," measures the speed of the null device; results are available for hundreds of systems.

Section Editor: Jon Corbet

Letters to the editor

Date: Thu, 27 Jan 2000 15:12:54 -0500 (EST)
From: Joe Klemmer <klemmerj@webtrek.com>
To: letters@lwn.net
Subject: Linux Security Issue


	I would like to echo your comments about the need for a more
organized security model for Linux.  To this I would like to offer a
donation of $500.00 towards the establishment of some kind of Linux
Security Body for the purpose of organizing and finding security related
issues in the Linux specific open source world.

	I do not do this lightly, I can not afford to put this kind of
money up, but I am doing this in the hopes that others will follow.  As
mentioned it would be a very good idea if the companies that are making
money on Linux were to put some of that money to use for this.

Thank you,
Joe

---
Pray to God, but keep rowing to shore.
                -- Russian Proverb

Date: Thu, 27 Jan 2000 21:58:52 +1100
From: Mark Lillywhite <mark@plasticsoftware.com.au>
To: letters@lwn.net
Subject: Kernel performance and threading

Hi,

I read with some alarm of your claim of a "challenge" which has been "thrown
to proponents of highly-threaded applications to recode their programs in a
single-threaded mode" in the Kernel section of your Jan 27 issue.

The design of a system will take into account many different factors, and if
the design is significantly simplified by using more than one thread of
execution, then that design should not be unduly penalised by the operating
system.

To suggest that a large multi threaded application should be rewritten using
a single thread, solely on the basis of peak theoretical performance
calculated using the relative probability of low level "cache misses" is just
plain silly. We have so many options to improve the performance of our
applications without sacrificing design ideals - such as waiting 12 months
for the CPU speed to double - that basing the architecture of a large piece
of software around something as esoteric as a low level, architecture
dependent metric such as potential cache misses is just irresponsible.

I imagine that a really large system with many processes communicating via
shared memory will have similar issues to those presented as arguments
against heavily multi-threaded applications. Are we to be expected to rewrite
multi-process systems as a single process too? Before long, we will reduce
our system to just one process - and it will look like an operating system.

Anyway, perhaps I'm missing something, but I thought that if I was executing
the same code in the same address space over and over again, then I would get
*loads* of cache hits?

Kind regards,
Mark Lillywhite

From: "Hyre, Max" <Max.Hyre@cardiopulmonarycorp.com>
To: "'lwn@lwn.net'" <lwn@lwn.net>
Subject: I must take exception to that header!
Date: Thu, 27 Jan 2000 13:43:04 -0500

   Dear Mr. Corbet:

   In the Commerce page of this week's LWN

	http://lwn.net/2000/0127/commerce.phtml

is the announcement that Sigma Designs is building a DVD player for
Linux.  The heading for the item is ``(Legal) DVD playback for
Linux'', and I object to the implication of ``(Legal)'', and claim
it would be much better presented as ``(Licensed) DVD playback for
Linux''.

   The status of DeCSS playback for Linux is in dispute, but most
disinterested parties hold that DeCSS is legal, probably in the US,
and certainly in many foreign countries.  The implied disparagement
of DeCSS's status should be avoided.

   I would appreciate it if you would consider this matter, and would
be thrilled if you'd change the text.

   Thanks for a great news site.



			Best wishes,

					Max Hyre

From: Art_Cancro@uncnsrd.mt-kisco.ny.us (Art Cancro)
To: letters@lwn.net
Date: Thu Jan 27 10:56:25 2000
Subject: Fun Sun Quote of the Day:      "There is no Unix marketplace

Fun Sun Quote of the Day:  
   
 "There is no Unix marketplace anymore.  It's a Solaris Marketplace."  
   
    -- Ed Zander  
   
 This one really floored me.  What Mr. Zander (could he be Sun's 
counterpart to Microsoft's own loudmouth Ed Muth?) is essentially saying 
here is: "Sun is no longer committed to open systems and open standards. 
 Our proprietary software is better than Microsoft's proprietary software. 
 And pay no attention to IBM, HP, and the Linux community; we're so far 
ahead of them they'll never catch up."  
   
 This sentiment sounds more like a spectre from the dark ages of 
fragmented proprietary Unix than like something from the modern world, 
which is finally starting to realize the value of a unified Unix (which is 
what Linux now represents).  It's hard to believe that any Unix vendor 
would still have the audacity to make such a comment.  
   
 It's also somewhat hypocritical, considering that open systems played a 
crucial role in making Sun a successful company in the first place.  They 
can sit on their high horse now, because they truly do have a world-class 
operating system that is extremely robust and scalable ... but Linux is 
rapidly closing the gap.  For that matter, even Microsoft is "shrinking" 
the gap somewhat.  If Sun continues to pooh-pooh the standards-based 
world, they won't be the Unix leader for much longer.  
   
    Art Cancro                           Visit UNCENSORED! BBS at  
    ajc@uncnsrd.mt-kisco.ny.us           http://uncnsrd.mt-kisco.ny.us  

Date: Thu, 27 Jan 2000 10:30:09 -0700
From: Bruce Ide <nride@uswest.net>
To: editor@lwn.net
Subject: Sun, Java and the Meaning of Suck

Sun's antics are very obvious. They want to be the next Microsoft pure
and simple, and I don't see any reason why we should replace one
gigantic monopoly with crappy software and proprietary "standards" with
another.

Sun probably realizes (As I hope do the rest of the readership here)
that Linux everywhere pretty much eliminates the need for Java
everywhere. Java was originally written on the concept that the same
code could be run on any platform without the need to worry about
platform dependent API's. Well if all the platforms have the same API,
it really doesn't matter if you do it with Java or C, C++ or whatever
other language you happen to like. And I personally have always found
Java's supposed cross platform capabilities lacking for any program more
complex than "Hello World." Moreover, Java has always been a huge memory
hog and horribly slow.

I think we should just jettison Java and hack Perl, Python and Scheme
(Oh my!) into Mozilla.

--
Bruce Ide               nride@uswest.net

SOMEONE had to put all that chaos there!

Date: Thu, 27 Jan 2000 10:43:45 -0500
From: John Klar <j.klar@xpedite.com>
To: Bruce Ide <nride@us.ibm.com>, lwn@lwn.net
Subject: Re: Why do you Bother With the Gartner Group?


Bruce,

I absolute agree with your assessment of Gartner.  Their opinions over
the years have definitely qualified for Petreley's "Beverage Through the
Nose" award.


I must respond to some of your points and prognostications.

> So Linux makes sense to most of those people, except SCO

I think Sun Microsystems is far more deserving of this comment
especially in light of their recent moves.  If I remember correctly SCO
recently added a Linux binary compatibility layer to their OS.

> I'm expecting the Merced to be a flop. It wouldn't surpise me if
> AMD doesn't step up to the plate within plus-or-minus a few weeks
> of the Merced release with an ultra-fast 64 bit chip of their own.

Intel has hedged their bets with Merced.  They've had Cygnus Solutions
(with the help of HP?) adapting gcc for its use.  In addition, I believe
they've also been porting Linux.  Similarly, Sun made statements that
they wanted to be THE definitive Merced OS.  I've not heard anything
within the last year, but take that with a grain of salt since I've let
most of my trade mag subscriptions lapse.

> Despite Wintel protestations to the contrary, PC's are still toys
> compared to the big blue iron.

Really, anything compared to a mainframe is a toy.  OTOH, if the
"pathetic little quad Xeon" can saturate your uplink, what's the point? 
Maybe a better measure would be cost/transaction.

> ... those chips are 64 bit TODAY, as are the MIPS chips but we
> don't see many of those anymore AFAIK.

The various flavors of MIPS are very much alive and well.  A significant
number of WinCE machines run MIPS derived chips as well as at least one
of the current crop of game consoles and (I think) a digital camera.


My predictions:

I belive the "embedded linux" market will explode.  The "always-on"
internet connection market desparately needs a cheap, easy to
configure/administer device by the average consumer.  I also see
computers increasingly used in home entertainment roles (streaming
audio, MP3, DVD) that were the traditional baliwick of consumer
electronics companies.  Home automation is another potential market. 
These products all need to be _fanless_, low power, capable of 24x7
operation and absolutely require a bulletproof user and service
interfaces.

Consequently the low power properties of Crusoe really excite me.


Opinions my own and not representative of my employer.

John Klar

Software Engineer
Premiere Document Distribution

Date: Sat, 29 Jan 2000 11:24:26 GMT
From: Duncan Simpson <dps@io.stargate.co.uk>
To: letters@lwn.net
Subject: Open sourec fixes --- is Red Hat the problem?

I read this with and agree modulo quite a lot. A large portion of the
problem seems to be getting the fixes into RPM format and not the
speed of actual fixes. Freely avialable fixed lpr was avialable a long
while back and it is sad that RH failed to put it in a RPM until
recently.

A far more recent example is last time I looked the latest bind 8
package is one apparently featuring the NXT buffer overrun bug. Bind
8.2.2p5, which does not have this bug, has been avialable for over a
month but last time I looked there was no sign of an update to this
version (plently of ISP name server boxen have been upgraded anyway).

Taking over a week to produce an update despite advisories with the
relevent code illistrated and a patch to fix the bug in them is a
problem. It is a little hard for the linux community in general to fix
this problem. Having said this the slow speed of fixes for a major
linux distribution, given that fixes on bugtraq are really fast, is a
problem that needs attention---how about a community site that RPMises
the bug fixed versions and signs them itself?

Disclaimer: I have neither RPM nor a RH system, so can do not this
site msyelf.

From: Jan Gruber <jgruber@fbw.htwk-leipzig.de>
To: letters@lwn.net
Subject: Linux-Pavillon @ cebit
Date: Sun, 30 Jan 2000 22:49:31 +0100

Hi,

it looks like Linux is gonna have its own pavillon
at this years Cebit.

Check 

http://www.heise.de/newsticker/data/jk-30.01.00-002/

for more details.
It's in german, but Babelfish should do the job.

Rgds.
Jan
-- 
######################################
Earth is a beta site.
######################################
cat /dev/world | perl -pe "(while (<>)\
{(/(*.?\?) 42 \!/) && (print $1)}"
- 

From: Richard Simpson <rsimpson@ewrcsdra.demon.co.uk>
To: letters@lwn.net
Subject: IBM's Java for Linux?
Date: Tue, 1 Feb 2000 15:18:14 +0000

Sir,

Last week you said, "IBM's Java implementation may not be truly free, but it
looks like it may now be free enough to become the de facto Java implementation
for Linux systems".  Are you sure that this is what you meant?  I spent some
time hunting on IBM's web site and could find their Java only as an i386 binary
with precisely no indication that it would ever be ported to the other 8(?)
platforms which Linux currently supports.

Please do not fall into the common trap of assuming that Linux == i386 Linux.
As an Alpha user I am constantly annoyed by reading that some wonderous piece
of software is now available for Linux, and then discovering that it is only
ever likely to be available for i386.

I appreciate that many companies who do not wish to release their products as
open source are unable to support the likes of Alpha and ARM simply because
they do not have the hardware to compile on.  A solid Java implementation would
allow them (as I understand it) to ship a Java bytecode (effectively binary)
version and have it run on every Linux platform.  This would be great news for
those of us with the "other" platforms, but ONLY if we have, at least, the
runtime environment.

If a solid Java implementation becomes available only for the most popular
platforms then those left out will be at a positive disadvantage.  Some open
source developers whose software can currently be compiled by (for example)
m68k users will switch to Java and said users will then not have access to new
versions.

I know that I am always banging on about other platforms, but surely as Linux
users we believe in hardware choice just as much as we believe in operating
system choice?

Thank you,
	Richard Simpson

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

Richard Simpson
Farnborough, Hants, Uk                 Fax: 01252 392118
rsimpson@ewrcsdra.demon.co.uk