[TriLUG] Information Week Article on Linux and Sound

[Marty Ferguson]

Fred Langa,
CC: Triangle Linux Users Group

I read and enjoyed your efforts to get sound working on Linux with
mainstream Intel hardware.
Particularly, where you could go all the way back to Win95.   Your intrepid
efforts are valiant,
and demonstrate great skill in diagnosis and problem solving.

A possiblity exists that Intel never tested new revisions of this
mainboard/sound processor
combination with Linux, yet still intend to provide (limited) compatiblity.
But, based on
the results of your thorough testing, I doubt that this is the case.

LOADLIN has been used for several years to conquer the specific sound card
problem you've encountered.
Loadlin is a windows based  utility that allows users to boot (or
springboard) Linux once WINxx is up
and running.  This problem boils down to the single issue of closed
interface architectures versus
open interface architectures.

The Intel integrated sound system on your computer is a programmable device,
perhaps (and very likely)
it is "Sound Blaster Compatible"; yet still it is a closed-technology
Intel-proprietary device.  It may well
never be "Linux supported" directly from _within_ the Linux community,
because it would be a violation of
license agreements to do so.  At a minimum, it would be a violation of
Intel's copyright protection over thier
sound card chip(s/set) for a Linux kernel contributor to reverse-engineer
the binary object code that is
downloaded into this device without formal approval from Intel.

Only Intel can provide a solution.  They must contribute an open source
module which downloads the binary
code to the chip.  Clearly, Intel has decided not to follow this path,
prefering to protect their sound
system through mantaining trade secrecy in their technology. Any other
choice on Intel's part could
be a potential compromise of their proprietary technology.  So it is quite
justifiable and well within
Intel's rights to protect their intellectual property.

Here is your key question: "And if the hardware was to blame, how could XP
handle it out of the box,
with no special drivers or setup?" And the key answer is, of course, that
hardware vendors write their
proprietary drivers and provide the code directly to Microsoft.  These
drivers provide the interface between
the OS and the layers of abstraction (Hardware Abstraction Layer) see
http://hal.freedesktop.org/ for a
description of the nascent efforts in this arena)  Based on a quick skim of
your recent article on Microsoft's
Virtual PC product, it comes to me as no surprise that it would not solve
your sound card interface problem.  The
abstraction layer would be no different, would it?


So here is how Loadlin solves the problem:
1 - Boot into Windows.
2 - Windows downloads the proprietary code into the sound device(s)
   a -  e.g., perhaps Digital Signal Processor (DSP) microcode,
   b -  e.g., and some SoundBlaster emulation mode interface routines
C - Run Loadlin. Loadlin starts up Linux
D - Linux probes hardware, and sees what looks like a
sound-blaster/compatible interface.
E - Sound works under Linux as expected.

As an aside, in the SCSI disk adapter world, similar issues used to arise in
Mylex versus Adaptec.
Mylex opened their interfaces many years ago, making it much easier to
design, test, probe and integrate
their products.  At that time, it provided Mylex with a distinct competitive
advantage over Adaptec in
the Linux server world.

In summary:
(1) Caveat Emptor.
(2) Apparently Intel Inside doesn't necessarily indicate open architecture
compatibility.
(3) Look for the Linux friendly Tux the Penguin on your retail packaging.

Respectfully,
Marty Ferguson, RHCE, RHCX, LPI-1 Certified