Dave Phillips - IA de Investigación especializada en el MIT

David Phillips
400 Glessner Avenue
Findlay, OH 45840 EE.UU
dlphilp@bright.net

Computer Music and the
Linux Operating System:
A Report from the Front

The purpose of this article is to introduce readers
to Linux and to describe the factors that make this
operating system attractive to computer musicians.
I have divided this exposition into two main parts.
The ﬁrst describes and assesses Linux as a general
platform for computing. El segundo, longer section
focuses on Linux’s audio and music capabilities. En
that section, a historical overview precedes a neces-
sarily brief account of the Linux audio software
base, followed by a prediction for the near future of
Linux audio development and the acceptance of Li-
nux for music applications.

A General Introduction to Linux

Many readers of this journal will need no introduc-
tion to Linux, but for the rest a brief description is
en orden. I will ﬁrst review Linux in general terms,
outlining its advantages as well as aspects that his-
torically have been of concern. The remainder of
this article will then describe in detail the consid-
erations speciﬁc to audio and music.

Linux is a UNIX-like operating system designed
to provide computer users with a free system com-
parable to traditional (and usually expensive) UNIX
systems such as IBM’s AIX, Sun’s Solaris, or SGI’s
IRIX. ‘‘Free’’ here means that the system software
itself is freely available at no cost, with free access
to the source code, and with the assurance that
anyone’s development of the system beneﬁts every-
one else who develops or uses the system. Linux is
compliant with the Portable Operating System In-
terface (POSIX), a standard of the Institute of Elec-
trical and Electronics Engineers (IEEE).

The system kernel was developed in the early
1990s by Linus Torvalds at the University of Hel-
sinki in Finland, and the system was essentially
completed by making use of components developed
for GNU (a recursive acronym for GNU’s Not
UNIX), which itself is a project of Richard Stall-

Computer Music Journal, 27:4, páginas. 27–42, Invierno 2003
(cid:1) 2003 Instituto de Tecnología de Massachusetts.

man’s Free Software Foundation. Other notable
contributions toward a complete system include
the X Windows System (a client-server windowing
engine) from the XFree86 organization, and net-
working tools and utilities from Berkeley Standard
Distribution (BSD) UNIX. Originally designed for
Intel’s i386 processors, Linux now runs on many
other CPUs and platforms, including the PowerPC,
64-bit Alpha systems, IBM’s S/390, various embed-
ded systems, and even hand-held machines (palm-
tops).

It should be noted that many users (y algunos
distributions) refer to the platform as ‘‘GNU/Li-
nux’’ in recognition of the Free Software Founda-
tion’s signiﬁcant contribution to the system.
Estrictamente hablando, Linux is only the kernel, mientras
most of its surrounding system software has indeed
come from the GNU project’s efforts.

Although the Linux kernel source code is in fact
available at no cost, most new users begin their Li-
nux experience by purchasing a software-plus-
manual-plus-support bundle called a Linux
distribución. A distribution is normally a complete
system including the X11 graphics and windowing
sistema, a robust TCP/IP networking subsystem,
and broad support for hardware extensions. Distri-
butions also typically include a wealth of applica-
ciones, utilities, and programming tools far in excess
of comparable commercial systems. Linux itself is
legally protected by the remarkable ‘‘copyleft’’ pro-
visions of the GNU Public License (GPL), but avail-
able applications for Linux may be covered by any
one of a variety of licenses ranging from freeware
with absolutely no restrictions to commercial offer-
ings protected by the end-user license agreements
(EULA) commonly associated with off-the-shelf
software. The great signiﬁcance of the Linux/GPL
marriage lies in the stipulation that any source-
level modiﬁcations to the system must in turn be
freely available upon redistribution. By establishing
a source-code base that is publicly open and exten-
sible by contributors, Linux has developed rapidly
since its initial public release. Distributions are
now available commercially from a number of

Phillips

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

high-proﬁle vendors such as Red Hat, Mandrake,
and SuSE, and extensive documentation about
Linux is available online and in print.

The Linux applications software base has bene-
ﬁted from a rich inheritance of tools and utilities
designed for UNIX computing environments, en-
cluding sophisticated packages such as the Emacs
text editor, the X11 windowing system, y el
GNU C compiler (gcc). Developers have trans-
formed the Linux user’s environment from one
dominated by an image of the daunting UNIX com-
mand prompt into one now media-rich and appar-
ently inﬁnitely customizable. The Linux graphical
user interface (GUI) can be tailored to available re-
sources and needs; it may consist of a simple win-
dow manager (p.ej., BlackBox, FVWM, or IceWM)
or a complete desktop environment such as the
K Desktop Environment (KDE) or the GNU Net-
work Object Model Environment (GNOME). De
curso, users are free to dispense with a GUI com-
pletely and run the system entirely from a text-
mode console.

Linux development has naturally strived toward

interoperability, supporting a wide variety of ﬁle
sistemas (Windows VFAT, MS-DOS, the Reiser
journaling ﬁlesystem, etc.) and network protocols
(TCP/IP, UDP, Samba, etc.). The system integrates
well with other platforms, promoting the use of
each platform for its particular strengths.

Scalability

En 1992, Linus Torvalds stated that his evolving
brainchild would likely continue to run only on In-
tel i386-based hardware. But by mid 2003, the sys-
tem has been deployed on a surprising number of
platforms. Linux still runs on x86 machines, pero
now PowerPCs, Alpha hardware, Macintosh com-
puters, and even the Amiga enjoy the beneﬁts of
the system. Ports to unusual hardware are often
undertaken as interesting programming projects,
and some ports have been written for every type of
hardware from palm-tops to supercomputers.

Linux scales well in almost any computing envi-

ambiente, from the single-user desktop to huge
clusters and processor ‘‘farms.’’ The system has al-

ready scored major successes in the professional
3D-rendering industry, with large Linux-based ren-
dering farms proving to be cheaper, faster, y
more reliable than any other solution. In-house
programmers for post-production facilities have
found the Linux source code to be a particular
blessing: they are no longer bound to wait for ﬁxes
and enhancements to arrive at the convenience of a
commercial provider. And thanks to the GPL, ellos
are free to make whatever changes are needed to
achieve the required performance.

Manufacturers requiring an embedded operating
system have also embraced Linux. The availability
of source code enables easier customization, y
the GPL relieves companies from burdensome and
restrictive licenses. NEC, Sony, Panasonic, y
IBM have already employed embedded Linux in de-
vices such as mobile phones and network routers.

Security

Computer security is a primary concern to system
administrators of environments networked to the
Internet or on local intranets. Una vez más, el
open-source nature of Linux proves most amenable
to resolving this concern: the code base is continu-
ally pored over by the many eyes of its develop-
ment community and tested by thousands of users
in demanding real-world situations, including high-
security corporate and military installations. De nuevo,
system administrators are freed from depending on
a vendor’s schedule for security updates and
patches, though it must be stressed that no system
(including Linux) is automatically secure. Sin embargo,
with the proper precautions, Linux can be easily
conﬁgured as a safe, solid, and famously reliable
network server and gateway.

Cost of Ownership

The total cost of ownership (TCO) of Linux has
been variously reported as both higher and lower
than that of other systems, particularly Windows.
How the relevant ﬁgures are computed indicates
the biases necessarily inherent to such reports, pero

Computer Music Journal

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

by and large they all point to a similar conclusion:
if an organization already includes Linux-savvy per-
sonnel, its startup and maintenance costs will be
bajo. Sin embargo, training a staff that is totally unfa-
miliar with the system is likely to result in a larger
initial outlay of time and energy. Still, such an ex-
penditure may be desirable to enjoy the beneﬁts of
an open-source solution and to avoid the increas-
ingly onerous costs and licensing terms from
single-source vendors. Finalmente, it is indeed true that
Linux itself is cost-free. Users can download the
kernel source code for free, a single disc may be
freely redistributed (es decir., with no per-seat licensing),
and many distributors even place their commer-
cially available products online as free downloads.
The high quality of the system and the scale of its
continual development, combined with this aspect
of the Linux TCO, creates an increasingly difﬁcult
and competitive ﬁeld for commercial vendors of
closed-source operating systems.

Problems and Concerns

En 1995, installing Linux was something of a labor
of love. Disk partitioning, ﬁle system deﬁnition,
manual conﬁguration of sound and video cards, y
other preparatory processes guaranteed that only
the intrepid dared to try it. Fortunately, today’s in-
stallation procedures are far simpler. Hardware is
now automatically detected, and system conﬁgura-
tion is often automatic, making the process as near
painless as possible for the beginner, yet the system
remains fully customizable by the more experi-
enced user. Contemporary distributions such as
Mandrake and Red Hat are quite easy to install,
and the GNOME and KDE desktop environments
are now typically the new user’s ﬁrst experience in
Linux. The familiar ‘‘point-and-click’’ GUI creates
a comfortable learning area for users migrating
from other GUI-based systems such as Mac OS or
Microsoft Windows.

Some vendors even sell machines pre-loaded
with a Linux distribution, obviating the installa-
tion phase; sin embargo, most users are likely to go
through the process themselves on their own ma-
chines. By comparison, relatively few users of Win-

dows or the Macintosh ever install those systems
ellos mismos, so ease of installation has been a pri-
mary goal for distribution vendors.

Technical support has also been something of a
thorny issue for Linux. Without a central location
or single company in charge of Linux, there is ap-
parently no one to call when things do not work as
they ought. Sin embargo, most mainstream distribu-
tions now include limited support for the typical
desktop user, with full support contracts available
to enterprise users. Some companies offer Linux
technical support as their primary business model,
but for many users, the most valuable support
comes from the helpful resources on the Internet.
Linux-speciﬁc newsgroups, mail lists, and Internet
Relay Chat (IRC) channels are available for every
level of user and developer, from ‘‘newbie’’ to ‘‘wiz-
ard.’’ The popular Google Internet search engine
(available online at www.google.com) provides ex-
cellent Linux-speciﬁc searches throughout the Web
and Google’s own massive archive of newsgroup
trafﬁc.

In its earlier years, Linux was justly criticized for
its rather difﬁcult learning path, but that criticism
is largely no longer justiﬁed. An enormous amount
of material, online and in print, is now available to
new users and developers. Monthly magazines ded-
icated to Linux are available, and new Linux-
speciﬁc books appear regularly on major
booksellers’s shelves. Certiﬁcation of Linux techni-
cal proﬁciency is available from companies such as
Sair and Red Hat, guaranteeing competence at vari-
ous levels of deployment.

Support from hardware manufacturers remains a

problematic issue. The classic ‘‘chicken-and-egg’’
syndrome is hard at work here, notably with regard
to audio equipment. Manufacturers commonly
write and distribute drivers for Windows and the
Macintosh but not for Linux. Depending on the
willingness of the manufacturer, one of three typi-
cal scenarios results: the manufacturer decides to
write and distribute Linux drivers; the manufac-
turer hands the needed speciﬁcations to Linux de-
velopers for them to develop the driver; or the
manufacturer has no interest in either of the other
scenarios and provides no support for Linux. Unfor-
tunately, this last scenario has been most common

Phillips

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

with regard to audio and MIDI hardware, but grow-
ing media coverage of the system may widen its
user base, sufﬁciently increasing demand to the
point of recognition by major audio hardware man-
ufacturers. Given the rate and quality of the sys-
tem’s development as an audio platform, es
certainly becoming more difﬁcult to ignore.

Linux as a Platform for Audio and Music

Now that we have reviewed Linux as a general
computing platform, we will examine its suitabil-
ity for computer music in particular. This section
traces the evolution of Linux’s audio capabilities
and describes the spectrum of available software,
from low-level tools through commercial music
applications.

General Factors Favoring Linux for Music

Availability of an operating system’s source code
is not usually regarded as an important factor to
aspiring computer musicians, but it has proved
critical to developers working on low-latency per-
formance enhancements at the Linux kernel level.
Developers interested in improving Linux multi-
media performance have been very active modify-
ing the kernel sources, a process utterly impossible
to implement in closed-source commercial operat-
ing systems (es decir., impossible without paying high
fees) such as Windows or the Macintosh OS. Given
this free and open-source availability of code, es
theoretically possible for an ambitious computer
music student to design a modiﬁed version of Li-
nux optimized to do nothing but process audio and
MIDI.

Other aspects of Linux’s UNIX heritage can be
found in its support for various mechanisms sup-
porting inter-process communication. Pipes and
other redirectors can be used at the command line
to create complex processing chains; Por ejemplo,
Csound supports line-events, a mechanism for feed-
ing score events to the program directly from the
command prompt in real time, dispensing with the
traditional Csound score. Reliable piping and redi-

rection is characteristic of Linux and other UNIX
sistemas.

History of Linux Audio

Thanks to the initial labors of Hannu Savolainen in
1992, the Linux audio applications programming
interface (API) provided support for the basic
SoundBlaster-compatible sound devices. At that
tiempo, those devices included a pulse-code modula-
ción (PCM) recording and playback device (/dev/
dsp), a raw MIDI input/output (I/O) device (/dev/
midi), and an audio device mixer (/dev/mixer). Este
kernel audio API, known as the Open Sound Sys-
tem interface (OSS/Free), has been included with
the kernel sources since 1992. Sin embargo, en 2002,
Linus Torvalds announced the inclusion of the Ad-
vanced Linux Sound Architecture (ALSA) API into
the 2.5.x development series of the Linux kernel,
setting the stage for the adoption of the ALSA API
and drivers into the stable kernel releases (2.6.x and
onwards). (See www.alsa-project.org for informa-
tion on ALSA.) This event is particularly important
to Linux audio development: it signiﬁes the arrival
of a freely available, open-source API for sound
and MIDI programming in Linux with fully
professional-level capabilities.

Nota: the Linux kernel is available in an even-

numbered stable series (p.ej., 2.4.X) and an odd-
numbered development track (p.ej., 2.5.X).
Separating the stable and development series in
this way ensures reliability for the normal user
without the risks associated with incomplete or ex-
perimental features under development and testing.
Hannu Savolainen and other Linux kernel hack-
ers had expanded the capabilities of the OSS/Free
API primarily to accommodate features found on
popular sound cards. Thanks to the willingness of
some companies to donate the necessary speciﬁca-
ciones, OSS/Free could provide support for some
relatively advanced features, such as on-board syn-
thesizers and full-duplex recording. Desafortunadamente,
most manufacturers of professional digital audio
hardware refused to provide free access to relevant
specs and documentation, resulting in a classic
loop: Linux applications developers could not write

Computer Music Journal

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

high-performance sound software without support
for more powerful hardware, whereas the hard-
ware manufacturers cited the absence of high-
performance software as evidence for insufﬁcient
demand. One way out of this impasse was taken by
Hannu Savolainen: he formed a company that
would sign non-disclosure agreements to provide
Linux audio developers and users with a greater se-
lection of sound drivers, albeit in the form of com-
mercial software with some binary-only
closed-source components. This package, called
OSS/Linux, has itself evolved from the OSS/Free
API and now supports a number of professional au-
dio boards and chipsets.

ALSA began as Jaroslav Kysela’s project to pro-
vide better support for his Gravis Ultrasound sound
card. Eventualmente, he found himself at the center of
a team of talented developers dedicated to the goal
of a completely free open-source sound driver and
API that would provide professional-grade services
to Linux audio programmers and users. Thanks to
the efforts of Mr. Kysela’s team and the foresight of
a few enlightened manufacturers, the project is
well on its way to meeting and surpassing its origi-
nal goals. Most popular sound-card chipsets are al-
ready supported, and I have already mentioned the
adoption of ALSA into the kernel sources. Otro
signiﬁcant ALSA achievements include the drivers
for the M-Audio Delta series of multichannel audio
boards, Takashi Iwai’s support for the Creative
Labs SBLive and Audigy sound cards, and Paul
Davis’s drivers for the RME Hammerfall series of
professional digital audio boards. Documentation
to assist in getting started with ALSA is available
online at www.suse.de/(cid:1)mana/alsa090_howto
.html.

The Modern Linux Sound System

Linux audio developers have been evolving a pro-
fessionally capable sound system based entirely on
free software. The system’s current status presents
a layered approach to solving the problems inher-
ent to the increased system performance demands
made by audio and video applications. Thus we can
view the modern Linux sound system at three lev-

los: the kernel level (p.ej., basic system hooks and
driver modules); a middle layer (p.ej., ALSA, JACK,
LADSPA, libsndﬁle, esd, and artsd); and user space
(programs such as Ardour, Sweep, and MusE).

Development on each of these levels has been
deep and extensive. The rewards of ALSA develop-
ment at the kernel level have already been men-
cionado. ALSA also has a signiﬁcant presence in
the middle layer with its exceptional utilities and
tools such as aconnect (a mechanism for routing
MIDI I/O between ALSA-aware programs) and alsa-
mixer (a text-mode system audio mixer). Other no-
table middle-level software includes the JACK
Audio Connection Kit (JACK), sound servers such
as KDE’s Analog Real-Time Synthesizer daemon
(aRts) and GNOME’s Enlightened Sound Daemon
(esd), libraries such as libsndﬁle and libaudioﬁle,
and plug-ins based on the Linux Audio Developer’s
Simple Plug-in API (LADSPA). (See www.ladspa.org
for more information on LADSPA.) Note that there
is no explicitly prescribed middle level, but I ﬁnd it
convenient to refer to it when describing essential
software that is itself neither kernel-space property
nor a direct concern of the normal user.

Paul Davis’s JACK, available from jackit

.sourceforge.net, deserves special description. Este
software provides a mechanism for connecting the
audio I/O of applications without affecting perfor-
mance latency. Por ejemplo, it is possible to route
the audio output from a MIDI sequencer running a
software synthesizer to the recording channel input
of a hard-disk recorder that itself may be recording
an audio stream from yet another concurrently run-
ning application. JACK ensures low-latency and
sample-synchronous execution for all streams
without loss of data, and it has been designed to
provide a method of transport control between
JACK-aware applications. Audio multiplexing is
also characteristic of the aRts and esd system
sound servers, but although those solutions are
suitable for common desktop audio, they are not
robust enough to handle the demands of profes-
sional audio applications (particularly hard-disk
recording systems).

Erik de Castro Lopo’s libsndﬁle is representative
of essential middle layer audio software. A collec-
tion of such software has been growing steadily, a

Phillips

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Mesa 1. Some Web resources for Linux audio

4Front Technologies
‘‘Achieving Low-Latency Response Times Under Linux’’

www.opensound.com
linux.oreillynet.com/pub/a/linux/2000/11/17

/low_latency.html

www.agnula.org

A GNU/Linux Audio Distribution (AGNULA)
The Advanced Linux Sound Architecture (ALSA) Project www.alsa-project.org
‘‘ALSA 0.9.0 Programming HOWTO’’
Debian Multimedia Distribution (Demudi)
‘‘Introduction to MIDI Programming in Linux’’
JACK Audio Connection Kit
Linux Audio Developer’s Simple Plug-in API (LADSPA)
The Linux Audio Developers Mailing List
‘‘The Linux Audio Quality HOWTO’’
‘‘The Linux Audio Users Guide’’
The Linux Audio Users Mailing List
Planet CCRMA
Turn-Key Linux Audio

www.suse.de/(cid:1)mana/alsa090_howto.html
www.demudi.org
ccrma-ww.stanford.edu/(cid:1)craig/articles/linuxmidi
jackit.sourceforge.net
www.ladspa.org
www.linuxdj.com/audio/lad
www.linuxdj.com/audio/quality
www.boosthardware.com/LAU/Linux_Audio_Users_Guide
www.linuxdj.com/audio/lad/user.php3
www-ccrma.stanford.edu/planetccrma/software
lulu.esm.rochester.edu/kevine/turnkey

the point where Linux audio developers now have
excellent open-source libraries for sound ﬁle I/O
(libsndﬁle, libaudioﬁle), sample rate conversion
(libsamplerate), and time/pitch stretching (el
SoundTouch library). This type of software works
in both directions, aiding developers by providing
commonly available high-quality audio libraries
and giving users a set of libraries shared by various
applications.

En 1996, I began listing the available user-space
audio applications for Linux. En ese tiempo, the list
totaled fewer than 50 elementos, ranging from com-
pletely developed programs such as UNIX audio
stalwarts MiXViews and Csound to barely usable
prototypes and alpha releases. In mid 2003, the list
includes more than 1,000 hyperlinks to a wide vari-
ety of software that still ranges from the newly
born to the fully developed. This situation is not
very different from the Windows and Macintosh
worlds with their various sites and repositories for
freeware and shareware applications, pero hay
one tremendous difference: nearly all Linux audio
and MIDI software includes the source code, hacer-
ing it possible for any knowledgeable user to make
any necessary or desired changes to the software at
any time.

The maturity of Linux sound applications has it-
self evolved along with the various Linux develop-
ment toolkits. En particular, GUI toolkits such as

GTK, FLTK, and QT now provide attractive inter-
face components that lend a more sophisticated ap-
pearance to applications. Multimedia toolkits such
as SDL (Simple DirectMedia Layer), CSL (Common
Sound Layer), and OpenAL have also brought more
capabilities to audio and video performance under
Linux while providing programmers with tools for
simpler access to those capabilities.

The variety of user-level applications has also
grown steadily since 1995. Nearly all the gaps in
the early lists have been ﬁlled, and Linux audio and
MIDI software users now enjoy full-featured,
professional-grade hard-disk recording , sound-ﬁle
editores, MIDI sequencers, software sound synthe-
sizers, effects plug-ins, network broadcasting, y
audio software in many other categories, far too
many to enumerate here. A subsequent section of
this article details some speciﬁc applications from
the categories mentioned, and I urge the interested
reader to consult the latest edition of the Linux
Sound and Music Applications Web site (linux-
sound.org) for a complete overview of the catego-
ries and varieties of Linux audio software. A
sampling of resources on the Web for Linux and
audio is provided in Table 1. Readers might also be
interested in my publication The Book of Linux
Music and Sound, published in 2000 by No Starch
Press in San Francisco.

Computer Music Journal

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Emulation of Other Operating Environments

Some excellent operating system emulators are
available for Linux, but the performance of multi-
media applications under an emulated operating
system remains problematic. Some MS-DOS MIDI
and sound applications will run very well under the
DOSemu emulator, and some non-real-time audio
applications for the Macintosh will run well under
the ARDI Executor Mac emulation environment,
but applications such as Cubase and Cakewalk are
still beyond the capabilities of the available Win-
dows emulators. Sin embargo, using program code
from the Windows Emulator (WINE), the Code-
Weavers development team has created software
(the Crossover Plug-in) that successfully invokes
the QuickTime 5 player and the Windows Media
Player under Linux. The VMWare program provides
yet another level of interoperability: this software
opens a virtual machine that can run Windows ap-
plications as though they were opened in their na-
tive environment. De nuevo, performance of real-time
audio and visual applications is compromised to
the extent of being unusable in anything like a pro-
fessional or semi-professional studio. The STEEM
and UAE programs emulate the Atari ST and
Amiga computers, and many MIDI and sound ap-
plications for those machines will run quite well in
the emulated environments.

Audio Latency

The Linux kernel can be patched to reduce audio
latency to professionally acceptable levels, down to
3 msec or less. The kernel provides access to a real-
time clock for ﬁne-grained timing for sequencers
and other real-time applications, and hard-disk
tuning utilities can dramatically optimize disk per-
rendimiento. Combining these factors with the sys-
tem’s exceptional stability, Linux becomes a
natural choice for the design of a low-cost high-
performance digital audio workstation (DAW).

An un-patched, un-tuned Linux 2.4 kernel can
create latencies of up to 300 mseg, far beyond any
acceptable limit for professional audio applications.
Fortunately, patches from Ingo Molnar and Andrew

Morton (see linux.oreillynet.com/pub/a/linux/
2000/11/17/low_latency.html) have proven that the
Linux kernel can be tuned to eliminate perfor-
mance bottlenecks, particularly when scheduled
processes unnecessarily remain active, blocking ac-
cess to other system activity and creating high-
latency conditions. These low-latency patches are
now available for anyone to apply, and they are
normally coupled with the use of the hdparm util-
ity for ﬁne-tuning disk performance. The hdparm
program gives users control over data transfer rates
and sizes, allowing drives to be tuned for maxi-
mum performance with maximum security.

Other research has also brought other solutions
to the latency problem. One such solution is Victor
Yodaiken’s RTLinux, a Linux system optimized for
real-time work where latencies must be measured
in microseconds, not milliseconds. Desafortunadamente,
RTLinux is not an option for Linux audio develop-
ers, because sound card drivers must be rewritten
to take advantage of RTLinux timing, and the gains
are not generally regarded as worth the effort. El
MontaVista Linux distribution (www.mvista.com)
is another example of a version of Linux highly op-
timized for real-time performance, though audio la-
tency is not a direct concern of the developers of
MontaVista. The performance of MontaVista Linux
certainly impresses the Yamaha Corporation:
MontaVista has been chosen to be the embedded
operating system for a wide variety of future Ya-
maha products. It may not be too unbelievable that
MontaVista could show up in some of Yamaha’s
next-generation line of digital synthesizers.

Robert Love has introduced a set of patches for
giving certain processes preemptive status at the
kernel level. A popular approach to optimal latency
reduction is to apply these patches along with An-
drew Morton’s low-latency set. This combination
has been shown to maintain very low-latency over
very long periods of time. De hecho, the 2.5.x develop-
ment kernel series has introduced Robert Love’s
preemptive patch as a kernel conﬁguration option.
Además, the 2.5.x series ofﬁcially dispenses
with the aging OSS/Free API and introduces ALSA
as the new kernel sound system. The real-time
clock (RTC) is still a conﬁguration option, y un
ﬁner-grained POSIX clock has also been added to
the most recent development releases.

Phillips

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Audio Drivers and Libraries

To reiterate, the three primary sources for sound
card and audio board drivers are the OSS/Free ker-
nel modules, the ALSA drivers and library, y el
OSS/Linux commercial driver package. Each source
has its strengths and weaknesses, but ALSA is un-
questionably becoming the standard set of audio
and MIDI drivers for Linux. Stand-out features of
ALSA include support for audio interfaces from
consumer-grade sound cards to professional digital
audio boards; multi-processor, thread-safe capabil-
idad; fully modularized drivers; a user-space library
to simplify audio applications programming and
provide high-level functionality; and compatibility
with the older OSS/Free API. Además, ALSA
supports serial port and Universal Serial Bus (USB)
interfaces, and it is completely free and open-
source, with its code base licensed under the GPL
or the GNU Lesser General Public License (LGPL).
The OSS/Free API should now be regarded as ob-
solete, and authors of new Linux audio software are
well advised to learn and utilize the ALSA API.
OSS/Linux is a ﬁne package, and it may be the only
solution where speciﬁc hardware may be unsup-
ported by either OSS/Free or ALSA, but it is deﬁ-
nitely not an open-source solution. A few other
independent driver projects exist, but they are usu-
ally for speciﬁc cards and do not propose a general-
ized API.

Model Linux Audio Setups

ALSA support for professional audio hardware de-
serves further comment. Multichannel boards such
as the RME Hammerfall and the M-Audio Delta se-
ries have excellent ALSA drivers, so it is now possi-
ble to work with professional-grade hardware under
Linux. I shall describe some of the software that
works with those boards, but ﬁrst I want to de-
scribe two typical Linux studio systems, one for ad-
vanced desktop work and another for professional
duty.

Both systems should include a fast computer,

large and fast hard-disk(s), al menos 256 MB of
random-access memory (RAM), a good video card

and monitor (preferably with support for acceler-
ated OpenGL), and optionally a MIDI keyboard. Ba-
sic software should include a recent Linux kernel
(2.4.x series and up) with real-time clock support
and patched for low-latency; XFree86 4.0 or higher;
ALSA (0.9.x or higher) drivers, library, and utilities;
JACK; and the LADSPA plug-ins.

Beyond these speciﬁcations, the main difference
between the advanced desktop system and the pro-
fessional system will be in the choice of sound card
or audio board. Some consumer-grade sound cards
can be quite effective even for semi-professional
trabajar. Signal-to-noise ratios can be bearable, y
S/PDIF digital audio interfaces show up on a few
cards (p.ej., Creative’s SBLive! and Audigy), haciendo
possible a digitally-connected Linux home studio.
Sin embargo, for work at professional standards, otro
digital audio boards are necessary (p.ej., the RME
Hammerfall or the M-Audio Delta cards). A
professional-level studio is also likely to include
considerably more expensive and powerful out-
board equipment not within the scope of this
artículo.

The question often arises as to which Linux dis-
tribution is best suited for audio work. The truth is
that any of the available distributions can be cus-
tomized to the extent needed by the user. Happily,
the trend is to supply the user with complete turn-
key solutions, relieving the user from the burden of
searching for and installing the necessary compo-
nents for a totally Linux-based sound system. El
Debian Multimedia Distribution (Demudi) proyecto
led the way (see www.demudi.org), and there are
now at least three other turnkey solutions available
for users.

Noncommercial Audio and Music Applications

Given the large number of Linux audio applica-
ciones, it is not possible for this article to give more
than a glimpse of the variety of such software
disponible. I refer the interested reader to the Linux
Sound and Music Applications Web site for a better
overview of the available software. Sin embargo, es
worthwhile to consider some of the larger catego-
ries and take a peek at their contents.

Computer Music Journal

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Cifra 1. A screen image
showing Common Music
and Snd.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

As expected of a worthy heir to the lineage of

UNIX software, Linux shines in its support for
sound synthesis packages derived from the seminal
Music V. Csound, Cmix, cmusic, Music4C, y
Common Lisp Music are all freely available in ver-
sions for Linux, all packaged with source code
(though not all are licensed under the GPL). El
extent of Linux-speciﬁc support varies: Csound for
Linux has received much attention from its world-
wide development community, resulting in support
for full-duplex, real-time audio processing, MIDI I/
oh, X11 graphics, and many of the extensions from
Gabriel Maldonado’s CsoundAV program (an en-

hanced version of Csound for Windows). Cmix is
available in a specialized version (RTCmix), writ-
ten by Dave Topper, optimized for exceptional real-
time performance. The entire ‘‘Common’’ family of
music software from Stanford University (Common
Lisp Music, Common Music, Common Music No-
tation, and the Snd sound-ﬁle editor) is available
with full support for Linux, providing a complete
and powerful environment for music composition
and sound design. (See Figure 1 for a screenshot.)
Musicians and researchers looking for Max-
inspired software will be pleased to ﬁnd Miller
Puckette’s Pd (ver figura 2) and IRCAM’s jMax (ver

Phillips

Cifra 2. An audio/visual
Pd patch.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Computer Music Journal

Cifra 3). Both systems enjoy very active develop-
ment and user communities, and both seem to be
turning to design philosophies of ever-increasing
integrative possibilities. Objects have been de-
signed for real-time video capture and processing,
for interfacing with Csound, for animating and con-
trolling OpenGL graphics, and for using software
synthesizers and external plug-in architectures
such as Steinberg’s Virtual Studio Technology
(VST) and LADSPA.

Linux sound-ﬁle editors deserve special mention.

Users migrating from the Windows or Macintosh
environments may be disappointed by the apparent
lack of an editor in the same class as Cool Edit Pro
or Peak, but Bill Schottstaedt’s Snd is arguably one
of the most powerful editors available for any plat-
forma. Its power is largely reserved for users willing
to learn the Guile/Scheme programming language,
but there seems little it cannot do that is done in
Cool Edit Pro (at least at the editing level). Newer
sound editors include Conrad Parker’s Sweep, el
excellent ReZound, Dominic Mazzoni’s outstand-
ing Audacity, and the Cinelerra audiovisual editor.
Older but still powerful stalwarts include Doug
Scott’s venerable MiXViews and Richard Kent’s
Digital Audio Processor (DAP).

At this point a divide exists. The software men-

tioned so far represents a development tradition
based in universities and research centers, pero
many popular audio software categories (p.ej., MIDI
sequencers, MIDI patch editor/librarians, and hard-
disk recording systems) have been deﬁned primar-
ily by commercial development houses. Sin embargo,
manufacturers of large-scale integrated audio-and-
MIDI sequencers and hard-disk recorders (HDRs)
such as Cubase or Logic Audio do not provide
source code to the end user, nor do highly evolved
MIDI sequencers such as Cakewalk. Open-source
programmers who wish to write such applications
must learn to do so without beneﬁt of an inherited
code base, yet that is exactly what has occurred.
Paul Davis’s open-source application Ardour (ver
Cifra 4) is aimed squarely at users familiar with
the ProTools model software for digital audio
workstations. Such a task might daunt most
programmers, particularly when almost every

aspect of the project must be coded from scratch.
Sin embargo, Ardour has been designed to reach
professional levels of capability and performance,
and by mid 2003, it is well on its way to achieving
its goals.

The nature of Ardour’s development is itself a
model of successful Linux audio software develop-
mento. Although Ardour is essentially the vision
and creation of an individual, Paul Davis has ac-
quired a team of talented volunteers to help push
the project forward. Contemporary development
herramientas (particularly those provided by the online
SourceForge project management environment) son
fully employed, and while a high standard of skill
is necessary at the core coding level, the project re-
mains open to all who wish to be involved. en un
manner similar to Linus Torvald’s style, Señor. davis
acts as ‘‘benign dictator’’ and ﬁnal arbiter, ensuring
the integrity of his original vision while welcoming
new ideas and suggestions from his team members
and ordinary users.

Notable current MIDI software for Linux in-
cludes the Anthem, MusE, and Rosegarden se-
quencers. MusE is especially noteworthy for its
support of advanced features such as audio se-
quencing, LADSPA plug-ins, software synthesizers,
and other ALSA clients. Rosegarden claims a simi-
lar set of features, with a special interest to Csound
users: it can save its output as a Csound score
(among many other formats). Both MusE and Rose-
garden perform MIDI and audio sequencing as well
as basic standard music notation, and both are al-
ready JACK-aware. Other worthy Linux MIDI soft-
ware includes the ALSA MIDI patch bay and the
excellent JSynthLib universal synthesizer editor/
librarian.

Perhaps the easiest way to begin learning about

Linux audio projects is to install one of the new
turnkey systems optimized for sound and music
producción: Demudi, Planet CCRMA (from Stan-
ford University’s Center for Computer Research
in Music and Acoustics, available online at
www.ccrma.stanford.edu/planetccrma/software),
Turn-Key Linux Audio (lulu.esm.rochester.edu/
kevine/turnkey), or AGNULA (A GNU/Linux Au-
dio distribution available from www.agnula.org).

Phillips

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Cifra 3. A simple jMax
patch.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Computer Music Journal

Cifra 4. The Ardour
editor window.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

They provide a low-latency kernel, ALSA drivers,
JACK and LADSPA, and a host of sound and music
applications. It is important to recognize that these
either are full-ﬂedged distributions or require only
a simple installation to upgrade an existing system.
(Por ejemplo, Planet CCRMA installs on top of an
existing Red Hat installation, and Turn-Key Linux
installs on top of a Mandrake installation.) Excelente
efforts have been made to make entry into the
world of Linux audio as painless as possible with-
out crippling the power of the underlying system,
and these packages are certainly an excellent intro-

duction to Linux in general and its audio software
base.

As indicated, the scope of Linux audio software
is too broad to delineate in detail here. Other nota-
ble categories include applications for working
with CD technologies (rippers, burners, jugadores), te-
lephony systems (see especially the Bayonne pro-
ject), digital disk-jockey (DJ) herramientas, virtual drum
máquinas (p.ej., the Hydrogen drum pattern editor;
ver figura 5), MP3 and OGG audio compression
software, MOD trackers, software synthesizers
(p.ej., ZynAddSubFX; ver figura 6), and various net-

Phillips

Cifra 5. The Hydrogen
drum pattern editor.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

work audio solutions (network sound servers and
streaming audio delivery systems). Interested read-
ers are again advised to consult one of the Linux
Sound and Music Applications sites for a complete
survey.

Commercial Audio and Music Applications

En este momento, only one company (4Front Technolo-
gies; see www.opensound.com) has survived as a
commercially viable entity selling Linux audio

software. Sin embargo, 4Front provides drivers for a
wide variety of UNIX platforms, not only Linux.
For some platforms, 4Front is the only source for
audio drivers, so it must be noted that they do not
depend solely upon Linux users for the company’s
revenue ﬂow. Other notable attempts at commer-
cializing Linux audio software include the Ceres
sound ﬁle editor (not to be confused with the spec-
tral domain processor of the same name from the
Norsk Nettverk for Akustikk, Teknologi, og Mu-
sik, or NoTAM), the Ultramaster RS101 rhythm
synthesizer, and the Jazz MIDI sequencer. Of those

Computer Music Journal

Cifra 6. The ZynAddSubFX
software synthesizer.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
/
C
oh
metro

j
/

a
r
t
i
C
mi
–
pag
d

F
/

2
7
4
2
7
1
8
5
4
8
8
7
0
1
4
8
9
2
6
0
3
3
2
2
7
3
0
4
8
8
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

attempts, only Jazz has survived (as Jazz(cid:2)(cid:2), un
open-source audio-and-MIDI sequencer project
available under the GPL). Ultramaster and Ceres
have seemingly disappeared, but commercial Linux
audio software may yet occupy a viable market po-
posición. Companies such as Steinberg and Twelve-
Tone Systems have started to show interest in the
Linux audio community, and as more users shift to
Linux from Windows or the Macintosh, they will
likely demand more of their familiar software for
the new platform.

Conclusión: Prospects for Linux Audio

In March 2003, I attended the ﬁrst conference dedi-
cated solely to Linux audio development, hosted by
the Zentrum fur Kunst und Medientechnologie
(ZKM) in Karlsruhe, Alemania. For three days, li-
nux audio developers from Germany, Francia, Italia,
Bélgica, Inglaterra, and the USA met to discuss a
wide variety of issues related to audio software de-
velopment at all levels. The conference was a great

Phillips

success, clarifying and resolving some current is-
sues, and deﬁning many other future plans and di-
rections. Outstanding presentations included
demonstrations of Ardour, Pd, the Bedeviled Audio
Sistema (BEAST), gAlan, Common Lisp Music, y
jMax, as well as technical discussions regarding the
ALSA system, LADSPA plug-in programming, y
the evolution of the AGNULA and Planet CCRMA
entornos. The high levels of skill and coopera-
tion served to underscore my belief that although
there is indeed still a long path ahead, it is getting
shorter every day, thanks to the efforts and talents
of Linux audio software developers.

Linux audio and MIDI software development
moves steadily forward. The Ardour HDR/DAW
is already quite sophisticated and is likely to
become the ‘‘killer app’’ for Linux audio, Kjetil
Matheussen’s VSTserver has made it possible to

use VST plug-ins via LADSPA or Pd, Istvan Varga’s
CsoundFLTK is a state-of-the-art version of Csound
for Linux, and the JSynthLib project supports an
ever-increasing number of MIDI synthesizers. Like
Linux itself, this software is under constant devel-
opment by its programmers and continual review
by its users. The Linux audio community is stead-
ily expanding, as is the amount and quality of
documentation and other tutorial material. Linux
itself has become easier to install, with greatly
improved hardware recognition and support, y
its community of users provides a vast reference
resource for ‘‘newbies’’ and gurus alike. These fac-
tores, in conjunction with the increasing sophistica-
tion of its audio components and the availability of
turnkey systems, make it clear that Linux is evolv-
ing into a platform worthy of serious consideration
by sound researchers, músicos, and multimedia
artists.