See also: last week's Back page page.

Linux links of the week

The next Linus need not start from scratch to make a new operating system; the OSKit now exists to provide a set of 31 component libraries which can be used to piece together a new kernel. This is a serious project, with a lot of (open source) code and documentation; worth a look.

Letters to the editor

Date: Thu, 28 Jan 1999 23:34:46 -0500 (EST)
From: Conrad Sanderson <conrad@hive.me.gu.edu.au>
To: lwn@lwn.net
Subject: correction to LWN 28 Jan edition, kernel section


You mention in the kernel section of LWN:

> Allocation of large (4MB) physically contiguous memory areas in the
> kernel.  The sad fact is that there are increasing
> numbers of peripherals, such as sound cards and frame grabbers, which
> require this sort of allocation.  Solutions to this problem seem to
> involve some sort of complicated shuffling of data structures in memory,
> not for the faint of heart. An attempt to make it work in 2.3 is likely. 

Actually, there's already a fairly elegant and simple solution.
It may not be the best possible solution, but it works very well.
It's called bigphysarea patch, available from:

http://www.uni-paderborn.de/fachbereich/AG/heiss/linux/bigphysarea.html

As far as I know work is being done to port this patch to 2.2 kernels.

The patch reserves, at boot time, a user defined contiguous block of
memory, out of the reach of the rest of the kernel and normal user apps.
Specifically enabled apps can then allocate contigious blocks of ram with
minimum fuss.  There's even a proc interface.

I have been using a Matrox Meteor frame grabber with this patch,
without problems.  

In my opinion this patch should become part of the kernel proper.


Conrad Sanderson - Microelectronic Signal Processing Laboratory
Griffith University, Queensland, Australia

Date: Sat, 30 Jan 1999 03:39:42 +0100
From: Ulf Carlsson <ulfc@bun.falkenberg.se>
To: editor@lwn.net
Subject: Structure vs purism

Hi,

I think it was stupid to post 'Structure vs purism' message in lwn
without including some of the replies, telling *why* gotos and strange
syntaxes are used in the kernel. It gives an incorrect view of the
kernel as an unprofessional piece of work, produced by programmmers
who don't know "the first thing that is being taught on pretty much
every programming course", which certainly isn't true.

- Ulf

To: editor@lwn.net
Subject: sigh...
From: Gordon Matzigkeit <gord@trick.fig.org>
Date: 28 Jan 1999 07:11:00 -0600

Please don't mince words.

If you had said ``sex games'' instead of ``adult games'', the
AdultLinux news item is a lot clearer.  As it stands, your description
of this site makes it sound more interesting than it actually is.

OTOH, maybe I'm just naive to have thought that ``adult games'' meant
something different than ``sex games''.

-- 
 Gordon Matzigkeit <gord@fig.org> //\ I'm a FIG (http://www.fig.org/)
    Lovers of freedom, unite!     \// I use GNU (http://www.gnu.org/)
[Unfortunately, www.fig.org is broken.  Please stay tuned for details.]

Date: Thu, 28 Jan 1999 01:46:57 -0600 (EST)
From: Dave Finton <surazal@nerp.net>
To: editor@lwn.net
Subject: Is there life for Linux after business?


A number of comments have been thrown around about the future of Linux,
now that Linux is becomeing Big Business(TM).  One reporter said that
linux would "lose its soul" as corporate interests dominated the scene.
I've also made a comment in an earlier letter to the editor (I can't
remember when and I'm too lazy to go look for it :^) about how business
are going to increase their role in not only using Linux, but in also
controlling its future development, much in the same way it's been done
with the World Wide Web.

Well, it's happening.  Businesses like Sun, Dell, Oracle, Netscape, and so
on are all jumping on the open source and Linux bandwagons, although each
are doing it their own ways.  Linux is no longer perceived as the
"hacker's toy", but rather as a professional tool.  Open Source and Free
Software are no longer regarded as flukes, but as a direction software
development is heading towards.

So now what?

I think one of the dangers of all this is that the mainstream press and a
few of the open source leaders have over-emphasized the fact that "Linux
must be accepted by the business world to be important".  I absolutely
disagree.  While I'm glad to see Linux being used more and more (as it
means I'll be that much more likely to use it for my job when I get out of
college), I think some people have gotten this mentality of
all-or-nothing.  Either business accepts Linux, or it will die and fade
away into history.

The real power that feeds Linux doesn't come from Oracle or Dell or Compaq
or Sun.  That's like saying the true power of the web comes from MSN or
Netcenter or Yahoo.  Yahoo or Netcenter wouldn't even exist if it weren't
for the thousands of web pages they index that are created by individuals
or small organizations.  The same is true for Linux.  The real "action"
comes from individual programmers, Linux User Groups, small business, and
so on.  True, they don't make the big headlines like the big boys do, but
that doesn't make them less important.

Linux has always been important.  Big business is now just along for the
ride.  I just hope they don't crowd everyone else out of the bus.  :^)

      - Dave "suffering from long-winded Jon Katz-ism" Finton

---------------------------------------------------------
| If an infinite number of monkeys typed randomly at    |
|   an infinite number of typewriters for an infinite   |
|   amount of time, they would eventually type out      |
|   this sentencdfjg sd84wUUlksaWQE~kd ::.              |
| ----------------------------------------------------- |
|      Name:      Dave Finton                           |
|      E-mail:    surazal@nerp.net                      |
|      Web Page:  http://surazal.nerp.net/              |
---------------------------------------------------------

Date: Tue, 02 Feb 1999 11:11:02 -0500
From: Dan Ginsberg <dan@hibernal.com>
To: editor@lwn.net
Subject: Reactions to poll data

I just read Nicholas  Petreley's bit 
( http://www.linuxworld.com/linuxworld/lw-1999-01/lw-01-penguin.html )
about RedHat and it struck a nerve, not because of what NP had to say
(as usual he made a lot of sense) but because of what motivated the
article:  

	"LinuxWorld ran an interesting poll recently...."  

Two things struck me upon reading this.  The first is that I've seen a
lot more Linux related news based on poll data of late.  This is fine.
If your self-image hangs on Linux's stature, then all the more reason
to strut; you, qua Linux user, have become a target-able demographic
group.  (Woohoo, and stuff.)  If you care about the future of Linux,
then here is the promise of useful data (e.g., as the demographics of
the Linux community change, what new concerns about or demands on the
OS and its community arise?  What does that mean for developers and
distributors?).

Unfortunately, the second thought that struck me upon reading NP's
piece was that most of the polls said nothing and that it was
unfortunate that anyone took them seriously enough to react to them.
Let me unpack that a bit.  Consider the poll that inspired NP's
article, or the bit of demographic polling that LWN pointed us to a
few weeks back (lost that URL, sad to say).  In both cases they polled
a lot of people.  LinuxWorld talked to 895 people and 73.75%
identified Linux with RedHat.  That sure *seems* to say something; but
as anyone who's done a bit of sampling can tell you, the numbers mean
nothing.  Given the number of Linux users out there, you can sample
385 folks and be 95% confident that data gathered from the sample
projects to the universe of Linux users with 5% error.  And that aint
bad.

However, that only holds true if your sample was randomly selected.
If your sample wasn't randomly selected then you could sample a few
thousand people and you would have no idea if data collected from that
sample projects to the universe.  Worse yet, you can be pretty sure
that your selection method has skewed your data to exclude or
underrepresent some groups.

For example, neither NP's piece nor the demographic survey captured
users who could not participate because they couldn't get network
services up.  In fact there are two problems with this poll data.  The
first is that part of the universe is excluded from the poll (they
aren't aware of it, can't get to it, don't read that language and so
on).  The second problem is that those who are not excluded from the
poll determine on their own whether or not they will be a part of the
sample.  It turns out that folks who choose to participate in polls
differ significantly from folks who choose not to.  Oops.

It is important that we explore the demographics, beliefs, values, and
quirkinesses of Linux users.  (Humor value aside, the whole
multi-colored iMacs thing bespeaks something powerful.)  It is also
wonderful that folks covering Linux are sensitive to poll data when.
However, it is important that we get it right, especially at the
outset.  Are there virtually no women running Linux?  Do 75% of people
think "RedHat" when they think "Linux?"  I don't think that we know;
and I suspect that it is a particularly bad idea to act as if we do.

dan ginsberg

Date: Tue, 02 Feb 1999 09:33:44 -0700
From: Alan Robertson <alanr@henge.com>
To: pankaj_chowdhry@zd.com
Subject: Re: Labs'-eye view

I have some trouble understanding what was meant by a comment in your
article on Linux found at:
	http://www.zdnet.com/pcweek/stories/jumps/0,4270,387772,00.html

> The community developing for Linux moves at a snail's pace

I assume this refers to something like this...

> With the exception of Novell--and we know how its market share is
> doing--operating system vendors have been providing symmetric
> multiprocessing support for I/O-intensive applications a while now.

Linux has had SMP capabilities which outperformed NT for about 2 years,
since the beginning of the 2.0 series of kernels. The new kernels
provide better performance in 4-way and larger SMP clusters (causing it
to outperform NT by a wider margin).

Release 1.0 of Linux came out in 1994.  Release 2.0 (with SMP) came out
in about 1996.  To make a fair comparison of Linux to NT in terms of
development speed would be to compare Linux 2.0 features and performance
(with it's superior SMP) to NT after 2 years (in 1993).  Although NT has
been slow since it's inception, it is better than it was in 1993.  In
release 2, Linux supported more hardware than NT did in 1993.  NT 4.0 is
now about on a
par with Linux in terms of hardware supported.

Surely everyone wants for the new kid on the block to be as far along in
every respect as the best of every other operating system.  It will be
within a year or so.  After that, it will be the leader in most
respects.  If you have read the Halloween documents, the Open Source
development model's incredible speed was one Microsoft's big concerns.

The size of the development community increases each time the user
community grows.  This may not sound like much, after all probably only
1% of the users develop and test code for it.  With 10 million users,
that's only 100,000 developers or so.  If Linux growth continues at the
current pace, expect to see a quarter of a million developers and
testers soon.

This is not to discount the efforts of the "big guys".  The dozens of
people they may eventually contribute to the effort are very important
-- they have access to detailed and often undocumented information on
how their proprietary hardware works.  Many of them are dedicated and
experienced specialists whose key contributions will be out of
proportion to their numbers - like the core linux developers are now.

To contribute to proprietary OSes, you pretty much have to
live in Redmond, WA, or Silicon Valley, and those two locales are
getting close to maximum capacity.  Although opportunities like these
are disproportionately available in a few places, smart people are
more-or-less uniformly distributed across the globe.  The opportunities
to contribute to Linux are restricted only by having internet access.
Linux development is moving faster than NT today, and will pick up the
pace more and more over the next several years as its development
community continues to grow at > 40%/year.

If you want to say "When compared to XYZ, Linux has the following holes
in it today", I have no problem with that.  If you want to point to a
particular perceived deficiency and say "It shouldn't have this specific
deficiency!", I also perfectly understand.  However, I believe vague
statements like "Linux development moves at a snail's pace" (or "Linux
is slow to fill it's [unnamed] holes") are badly mistaken.  When you
start from ground zero, by definition, you start behind.   When you
compare its recent rate of filling holes to any other OS, it is filling
them very rapidly, with greater performance, reliability, and on many
more hardware platforms than NT and most other OSes.

	-- Alan Robertson
	   alanr@henge.com