[LWN Logo]
[LWN.net]

Sections:
 Main page
 Security
 Kernel
 Distributions
 Development
 Commerce
 Linux in the news
 Announcements
 Back page
All in one big page

See also: last week's Back page page.

Linux links of the week


Appwatch.com has just entered its "beta" stage. Appwatch is another database of free software; their angle is that they list only free software, and they keep a watch on their database and clean out entries for dead projects. Appwatch is built on Zope and PostgreSQL.

Learn how the other side thinks. Here's an anti-Linux page that may be worth a look - successful advocacy depends on an ability to counter the opposition's arguments. See also the Sawman's Consortium, complete with its "Linux Lies" document. (Please do not flame the maintainers of these sites, that will not do any good for anybody).

Section Editor: Jon Corbet


September 23, 1999

   

 

Letters to the editor


Letters to the editor should be sent to letters@lwn.net. Preference will be given to letters which are short, to the point, and well written. If you want your email address "anti-spammed" in some way please be sure to let us know. We do not have a policy against anonymous letters, but we will be reluctant to include them.
 
   
Date: Wed, 15 Sep 1999 22:35:17 -0700
From: Raph Levien <raph@acm.org>
To: editor@lwn.net, raph@acm.org
Subject: Some clarifications on the tp600

Hi lwn'ers,

   Very nice writeup of the tp600 compatibility issue. As a tp600 Linux
user and maintainer of a tp600 page (http://www.levien.com/tp600.html),
I have been tracking the modem issue quite carefully, and have had some
interesting conversations recently with people about what it means to be
"compatible" with Linux.

   First, a small factual clarification. The modem in the tp600 may be
considered a WinModem, but the processing is done by an on-board DSP
chip (the MWave), not by the host processor. Further, your statement
that the lack of interface specs is what's holding back WinModem
compatiblity is also not true. All important specs on the MWave are in
fact public (Linux driver writers have worked with a lot less). What's
missing is the important software and/or firmware that makes a modem a
_modem_, rather than just a pile of telephone interface circuitry, A/D
and D/A converters, and digital logic. The free software community has
not yet come up with such a piece of software. If it had, the WinModem
issue would be moot. Incidentally, Russ Nelson's
http://www.linmodems.org page is an interesting step in this direction.

   From what I can tell, it would be fairly easy for IBM to make the
modem work under Linux. This is based on both familiarity with the
technical issues and some informal conversations with IBM'ers. The
software is already written and ships with Windows 9x. If IBM were
unwilling to release it as free software, I don't think anyone would
fault them for releasing it as binary-only. Indeed, one other vendor of
WinModems already has:
http://www.linuxworld.com/linuxworld/lw-1999-08/lw-08-linmodem.html

   So what I think galls me about this is that IBM is reaping the
publicity goodies of being the first certified Linux Compatible laptop
without having actually _done_ anything. For what surely must be less
than the marketing expenses associated with the certification, they
could have made the modem work.

   That said, "Linux compatibility" is a fuzzy concept (a friend pointed
out to me that Linux itself is a fuzzy concept, when you get down to
it). On any given computer, you'll find that some things work perfectly
right out of the box, other things work pretty well with little hassle,
still other things can be made to work if you have infinite patience
(currently, IrDA and USB seem to fall into this category), and lastly
you've got the things that just flat out don't work at all. Hopefully,
all the really important stuff falls into the first two categories. But
beyond that, where exactly do you draw the line on what's considered
compatible and what's not. If the TP600 were sold with an epoxy plug
over the RJ11, would it be more compatible? What about those servers
with neat new features for diagnostics and failover that are not
supported by Linux yet? As with most things in reality, it's
complicated. A blanket certification of the TP600 as "Linux Compatible"
conveys very little real information to me. If you really want to know
how well a laptop runs Linux, the best resource by far is the Linux
laptop page: http://www.cs.utexas.edu/users/kharker/linux-laptop/

   Thanks again for a good article.

Raph
   
From: Nathan Myers <ncm@linuxlaptops.com>
To: letters@lwn.net
Subject: Winmodems and Thinkpads

In the Sept. 16 edition of LWN, you wrote:

  [The IBM Thinkpad's built-in modem] is a "WinModem" - a modem that
  requires most of the signal processing to be done by the central
  processor. There is no reason why such modems shouldn't work under
  Linux, except one: the interface information for these modems has 
  not been made available by their manufacturers. This information 
  is all protected under non-disclosure agreements; thus, no Linux 
  driver can be written.

First, "WinModem" is a trademark of 3Com, Inc.  A "winmodem", generically, 
is one that can only be used in Windows, either because the manufacturer 
won't release the specs, or because no self-respecting Linux hacker would 
bother to write code for it.  However, not all non-traditional modems 
should be called "winmodems", and not all winmodems lack signal-processing 
oomph.

The quoted paragraph conflates two very different devices.  A dumb 
winmodem is little more than a mono sound card with a phone jack.  It 
is the driver, in a sense, that _is_ the modem.  Given a fast CPU and 
a real-time kernel, a sophisticated driver could, in principle, perform 
well; in effect, your whole computer would become a modem that can do 
other things too.  In practice, Linux is not a real-time kernel (neither 
is Windows), and sophisticated drivers for these devices are, on any OS,
rare -- perhaps nonexistent.

A DSP winmodem has its own CPU, and is almost the same as a traditional 
modem, but lacks ROM.  It depends on the driver to download its program 
to it, just as some very popular SCSI controllers need to be initialized 
with a program image provided by the vendor.  (We don't call these 
"winSCSIs" because vendors do provide the program image.)  Normally, as 
with ISDN modems, the driver fields the "AT" commands itself and talks to
the DSP program at a lower level.

The program needed to make a raw DSP device act as a reliable modem 
is easily as large and complicated as an IP stack, and as subject to 
improvements and bugs.  IP Networking has a tradition of Free "reference 
code", yet it has taken many years and several rewrites to get Linux's 
network stack into a respectable state.  I know of no reference V.90
implementations.  Much of the value of 3Com's modem division is in its 
ownership of its modem code.  

While the PCTel modem that is common in laptops is a dumb winmodem, 
devices based on Lucent's product, like the Thinkpad's built-in modem, 
have true DSP capabilities, and lack only a driver to operate them.  In 
fact, the specs for the Lucent device are already in developers' hands.  

Is the Lucent device really a winmodem?  No driver is available, but the 
specs are not secret.  No one has taken on the job of writing a driver yet. 
Probably none will exist until an enlightened company manages to Free up
its private implementation, according to some yet-undiscovered business
plan.  (Still, see http://www.linmodem.org/.)

IBM's Thinkpad presents a more complicated case: most of the devices in 
the machine, including parallel, serial, and sound, are operated by the 
same DSP coprocessor that "does" the modem.  Programming it and interfacing 
with it are correspondingly messier -- indeed, most problems with Linux on 
Thinkpads involve the DSP device.  (Curiously, the same problems manifest
in Windows.  IBM calls these bugs "Considerations".)  IBM has shown some 
desire to get their modem supported on Linux, but (if I understand 
correctly) they seem to be having a hard time finding somebody for their 
lawyers to talk to.

Generally, though, built-in modems are not such a big problem as is often 
suggested.  Excellent PC-Card (PCMCIA) modems are available, and a good 
PC-Card modem may work far better than the built-in one ever can.  A much 
bigger problem for laptop users is the continuing instability in sound, 
USB, and IrDA hardware, because those can't be bypassed the way a built-in 
modem can.  

Nathan Myers
ncm@linuxlaptops.com   http://www.linuxlaptops.com

   
Date: Thu, 16 Sep 1999 11:36:34 +0200
From: Rodolphe Ortalo <Rodolphe.Ortalo@cert.fr>
To: lwn@lwn.net
Subject: A comment on DRM (LWN, Sep. 16)

Hello,

in Sep 16 issue of the LWN, under the 'Kernel development'
section, concerning the new Direct Renderimg Manager,
you say that:
"... The result should be secure access to low-level hardware
and screamingly fast 3D graphics. More information on DRM can
be found on the DRM design document;..."

In fact this is a little inaccurate. The DRI should provide
something that is (only) as secure as the X server. (Even the
DRM design document acknowledges this, as I understood it.)
Unfortunately, X11 servers security is still extremely poor. (I
hope I do not need to explain this statement... Ask if you want
but I'd rather keep that email short.:-)

Non-withstanding security issues, safety issues are no more
adressed. The DRM does not provide any arbitration or access
control with respect to broken hardware etc. For example,
some good old S3 cards simply lock the PCI bus when a MMIO
access is done while the graphic accelerator is running.
This is a hardware bug of course. But the DRM does not protect
the machine against this. Neither does the X server (which only
mean of control would be to deny to applications the access to
DRM-related features - thus denying DRM interest on these
cards). Graphic hardware dependability is currently improving,
but it has a long way to go...

Well, anyway, I simply think that security or safety is not
among the design objectives of a thing like the DRI. (If it
were they would not simply assert the fact that the X server
needs root privileges as a security features...:-)
The objective was/is "screamingly fast 3D graphics".

IMHO, this is a perfectly respectable objective. :-)
But suggesting that this kind of architecture can bring also
security or safety improvement is, in my opinion, a giving
a false impression.

Anyway, that's just a personal comments...

Regards,

Rodolphe Ortalo
   
From: "Michael Callahan" <michael@ask.com>
To: <lwn@lwn.net>
Subject: direct rendering explanation in 9/16 LWN
Date: Wed, 22 Sep 1999 15:44:10 -0700

Hi,

Thanks again for the wonderful Linux Weekly News!

In last week's issue, in the Kernel Development section, the following
explanation appears:

>The Direct Rendering Manager (DRM) has also been put into the kernel.
>DRM is the kernel-level support needed for the Direct Rendering
>Infrastructure (DRI), which will be part of XFree86 4.0. Essentially,
>DRI pushes the handling of three-dimensional rendering into the X
>server. The server, in turn, can take advantage of the rendering
>capabilities provided by the video hardware. The result should be
>secure access to low-level hardware and screamingly fast 3D graphics.

In fact, direct rendering is basically the opposite of what is described
here.  When the X server does 3d rendering, that's called indirect
rendering--because the 3d data passes from the application, through the X
server, and into the 3d video card.

Direct rendering allows applications themselves to send their data directly
to the video card.  For 3d graphics, where there can be very high volume
streams of data, this can be vital.

Michael

   
Date: Fri, 17 Sep 1999 08:31:26 +0200 (CEST)
From: chimbis@skjoldebrand.org
Subject: A couple thing off my chest
To: letters@lwn.net

Just got to get a couple of things off my chest, so to speak.

1/ Standards. Linux badly needs standards. Not perhaps strictly to
satisfy a real need within the community but to better it's reputation.
I recently saw (I think it was) an add for Veritas Backup Exec for
Linux, which is available for Red Hat, SuSE, and Open Linux with a
version for TurboLinux soon to follow. Another product was available
for Red Hat with other distros to follow soon.

This is ridiculous, considering that there isn't really a compatibility
issue to talk about. People are just stuffing files in different places
which in fact is making it hard for companies in porting software that
will work across all Linuxes. This is stupid.

2/ One wonders what impact on Linux badly tested, originally Windows
software is going to have on our favourite OS. I have in mind a piece
of ported software that is quite buggy and even crashes at irregular
instances. It is however, quite feature rich so many people would
probably like to try to use it. These software packages could damage
Linuxes reputation for quality. Just a thought.

Cheers,

Martin S.
-- 
Martin Skjöldebrand, Chimbis Design
Sys admin, web design
Hungry? Visit The Olde Cookery Book at
http://www.bahnhof.se/~chimbis/tocb/

   
Date: Wed, 22 Sep 1999 12:42:37 +0530
From: ANAND <anand@nmi.stpn.soft.net>
To: dvorak@zdnet.com
CC: letters@lwn.net
Subject: The Linux Myth

Hi,

I find your contention, that Linux will not make it in the server arena
because it is not very stable, laughable. How is it that NT that can't
stand on it feets for more than a week is still considered as a strong
contender in the server arena if stability and reliability were that
important. Linux is obviously much better in these respects. Linux may
not be the most stable OS, but there are some other things that make it
ideal for the server arena, lack of licensing restrictions. Its also
probably the most efficient of all OSs. The only two places were linux
lacks is in scalability, and User Interfaces, but there too not by much.
Low scalability will prevent it from entering the arena of the super
large systems, which cannot be handled by a single Quad Processor
system. But then NT also lacks those things. UI problems are more of a
perception problem, and as you say won't matter for the ultra cheap
systems. I would like to point you to the latest SAP R/3 benchmarks for
Linux, which claimed that they were the best reported for any Quad
Processor systems. That would definitely include NT.

Anyway the biggest thing that you forgot was that Linux has the biggest
developers resource. No Company can compete on that ground, and anything
proprietory, will become less and less competitive will Linux as time
progresses, because as internet expands, Linux's talent pool will keep
on increasing.

-anand
 

 

 
Eklektix, Inc. Linux powered! Copyright © 1999 Eklektix, Inc., all rights reserved
Linux ® is a registered trademark of Linus Torvalds