Studio Online 3.0:
An Internet "Killer Application" for Remote Access to IRCAM Sounds
and Processing tools
Guillaume
Ballet, guballet@club-internet.fr
Riccardo Borghesi, ricborghesi@hotmail.com
Peter Hoffmann, Hoffmann_Peter@hotmail.com
Fabien Lévy, fabien.levy@wanadoo.fr
http://sol.ircam.fr/external/joba
Abstract: Studio Online 3.0 is the final
version of an Internet music application with distributed objects developed at
IRCAM in 1996/8. This application offers high-quality instrumental sound
"online" for music researchers, composers and professional audio
studios. Studio Online is 3-tiered: a client applet runs in a standard Web
browser and connects to a server hosted at IRCAM providing access to IRCAM
sound transformation tools and to a large sound database. The uniqueness of
Studio Online lies in its ambition to serve the needs of scientific music
research, contemporary composition and pedagogical activity. The overall goal
of the project was to provide an efficient and easy-to-use application for
contemporary audio research, composition and music production with the
exclusive use of non-proprietary software tools and open standards.
Keywords: World Wide Web, Audio Databases, Sound Processing, Distributed
Computing, Client/Server Architectures, CORBA.
1. Introduction
IRCAM is a
world renowned institution with specific competencies in acoustics and
psychoacoustics of instrumental sound, sound analysis and transformation
software, and computer aided composition. As a major center of musical research
and production, it has a special mission of archiving/documenting/teaching
contemporary musical use of technology. Additional artistic competence is input
by visiting composers and musicologists. Besides IRCAMs departments for
research and documentation, there is also a large department for music
production and pedagogy.
All of these
departments are in constant need of high quality sound samples of musical
instruments: the research departments need reference material for their various
analyses, and the pedagogy needs specific sound samples for their music
productions. Until recently, research and production had to laboriously produce
their own sound material every time they needed it, as there was no centralized
collection of instrumental sound readily available for their work. With Studio
Online, high-quality instrumental sound can now be instantly downloaded at any
time to the personal computer or workstation on the office desk.
One other
problem tackled by Studio Online is the availability of advanced IRCAM sound
processing software for the different users in the institute (composers,
researchers, engineers, students) who are typically working on different
computer hardware platforms (UNIX workstations, Macintoshes and PCs) and with
different sound file formats (IRCAM's floating point format, Macintosh AIFF and
AIFC, and Microsoft's WAVE format). All these people need a unified, and
somewhat more user-friendly access to the power of IRCAM sound processing
tools, without dwelling too much on their various versions, platform
dependencies, specificities and intricacies of their handling. Not all IRCAM
sound processing tools have graphical frontends like SVP, FTS, Diphone or
Patchwork do, and these are only available on specific platforms (Audiosculpt,
Diphone and Patchwork only on Macintoshes, jMAx only on some UNIXes).
This was
probably part of the motivation for IRCAM to respond to a call by the French
Ministry of Industry for proposals of a 3-years project on the
"Information Highway", in 1996. After successful completion of the
project by the end of 1998, IRCAM has at its disposal a versatile service for
the various needs of the inhouse staff, visiting composers, and external users
(e.g. Forum members). Moreover, Studio Online can be used by any computer
connected to the internet. IRCAM technology can therefore be remotely accessed
and evaluated from anywhere in the world. Studio Online permits an
instantaneous, around-the-clock access to IRCAM sound archiving and processing
power, a remotely configurable and controllable personal music studio for music
professionals and laymen. Not only is it possible to download sound from the
IRCAM database but also to upload own sound files to a private user space and
have them treated by IRCAM processing tools. According to the international
character of the Internet, Studio Online is entirely bilingual (English and
French).
After a first
phase of consolidation, the Studio Online Team formed under the management of
Guillaume Ballet and comprised developers Rolf Wöhrmann (1997) and Rodolphe
Bailly, Riccardo Borghesi and Peter Hoffmann (1998), artistic directors Joshua
Fineberg (1997) and Fabien Lévy (1998), sound engineers François Eckert (1997)
and Vérène Gribonval (1998), post production engineers Antoine Mercier, Gérard
d'Elia (1997), and Cécile Lenoir (1998), and psychoacoustic research assistant
Nicolas Misdariis, as well as a couple of interns. One other project, also
supervised by Guillaume Ballet and funded by the Ministry of Culture,
specifically used Studio Online sounds for a popular Web site of pedagogical
vocation called "Web Culture". This site was developed by artistic
director Fabrice Guédy and Web designer Guillaume Dimanche and is accessible at
http://sol.ircam.fr/instruments/.
2. Aims and Scope of the Project
It may be
adequate to recall the ambitions and aims of the Studio Online Project as
conceived by the Studio Online Team led by Guillaume Ballet [Ballet 1998]. One
important technical ambition was the exclusive use of open standards (HTML,
TCP/IP, CGI, CORBA, SQL) and non-proprietary technology (Java and C++
development under UNIX, standard CORBA tools[1], JDBC). An integrated database
client/server development system could probably have facilitated the task, but
only at the expense of provider dependency, hardware/operating system
limitations, or restricted availability by the internet community (additional
client software or even hardware etc.)[2] For using Studio Online, there is
no dedicated hardware necessary, no encryption involved, no specific client
software, no installation of the client, even no version control necessary, and
no dependency on a specific carrier or provider. Everybody can connect: all one
needs is a working internet connection (some of them are even free of charge!)
and an up-to-date internet browser, which is a piece of free and ubiquitous
software. Everybody who has a computer can have that.
One important
artistic ambition was to cover contemporary aspects of instrumental sound of
specific interest for contemporary electroacoustic composers and/or of specific
pedagogical interest. For example, in Studio Online, most instruments (except
those with a homogeneous timbre like the strings) are sampled in quarter tones.
One can interactively search a large systematic collection of wind
multiphonics, as well as some more 200 different 20th century playing modes and
techniques as exotic as e.g. for woodwinds: "jet whistle", "key
clicks", "kiss sounds", "subtones", breathing and singing
through the instrument, and for strings such as rubbing with the finger nail,
knocking on the instrument's body, pressure bowing, artificial harmonics
(strings), to name but a few [Lévy 1998]. Room acoustics are taken into account
by providing 6 channels for each sample: 2 near and 2 far stereo microphone
pairs as well as a near and an internal (built-in or contact) microphone.
3. The Architecture of Studio Online
Studio Online
has been conceived as a 3-tiered internet application. A client applet (the
front tier) runs in a Web browser and connects to a powerful server machine at
IRCAM. Two server programs (the middle tier) connect the applet to a sound
database as well as to a collection of audio processing tools (the back tier).
In spite of being a Web application, on a fast machine and with a good Internet
connection, Studio Online almost feels like a local program. The previewing and
downloading of sounds and other data (even simultaneously) is managed by the
browser so that the user can immediately go on working with the interface while
the browser manages the retrieval of the data (even of several sources
simultaneously) in the background.
For sound
transfers, the applet uses the browser's capabilities of handling various
multimedia data for previewing (by spawning appropriate helper applications) or
downloading to the user's local hard drive for later use. All other
communication is handled by an IIOP CORBA connection through a special proxy
server, Visibroker's "Gatekeeper", which works around some of the
applet's sandbox security restrictions. For example, it permits connecting to a
different server interface than from where the applet was loaded, and it holds
a callback connection allowing the server to recontact the applet on
asynchronous events. (We use the callback feature to notify the user upon
arrival of sounds uploaded to IRCAM and to constantly check the liveness of the
client applet during the session.) For details, see [Hoffmann 1998].
Studio Online
has two main server processes: one server is written in Java and manages all
session oriented aspects. Another server is written in C++ and manages the
coordination of the various audio transactions like sound transformation,
format conversion and downloading. Both servers are of course fully
multithreaded and so handle multiple sessions and requests concurrently. The
session server works between the client applet and the database containing all
informations about user login and preferences and the sound taxinomy of Studio
Online. The transaction server works between the client applet and several
IRCAM sound processing and format conversion tools as well as a number of other
tools for generating archives, checking the user's disk quota, etc.
The CORBA
middleware enables a direct communication among objects distributed between the
client applet and the two server programs as if it were just one single
object-oriented program. The object distribution is not only perfectly
transparent to the user but also to some degree to the programmer, a fact which
allows much flexibility in design and implementation of a distributed C/S
application.
The session
server connects to the database via a JDBC bridge. This server program stands
between the database and the client applet and provides a convenient functional
layer of abstraction to the SQL database communication. The applet, for its
part, presents to the user an even more intuitive visual interface for
navigation within the sound taxinomy stored in the database. In an iterative,
interactive process, the user is invited to incrementally refine his/her
choices on a number of aspects of the sound (instrument, playing mode, pitch,
etc.) while the interface constantly updates in order to present to the user
the number and the aspects of the sounds that are still available. Behind the
scenes, every choice of the user in the interface is converted by the applet
into a request over the internet to the session server which sends a
corresponding SQL query to the database, evaluates the result and returns the
information necessary in order to update the applet interface. Thanks to an
efficient implementation of the remote querying process, the interactive sound
selection game almost feels as if one had to do with a local installation
(provided the Internet access is not too bad).
The actual
sound data are not stored in the database itself but on a large RAID disk
array. It was found that it was easier to handle them on a file system, as no
prediction could be obtained at the time of how the database would eventually
behave when loaded with hundreds of Gigabyte of data. In addition, we needed
direct access to the sounds during development time and, last but not least,
the database BLOB primitives were bugged in Oracle 8. So the database only
references the sounds by an identification number which is passed by the applet
to the transaction server. The transaction server dynamically creates Perl
scripts which contain the sequence of command lines for various tools accessing
the sound files on the RAID file system and converting them according to the
user's preferences concerning the microphone configuration (near-mono,
internal, stereo-near, stereo-far and their left and right channels), the
preferred sound file format, sampling rate, quantization and the volume
compression of the sounds. On sound download by the browser, the Perl scripts
generated by the transaction server are executed through a CGI invocation and
transparently convert the desired sound "on the fly" according to the
user's specification. The user's preferences are configurable at any time and
persistent between different sessions.
Studio Online
sounds have not been subjected to any audio compression with possible data
loss. The main concern was 100% uncompronized sound quality, and we did not
want to trade it off against accelerated downloading time. It is also in order
to ease access and to avoid dependency on proprietary compression formats.
Figure 1:
The Studio Online 3 Distributed Architecture
4. Related Work
Other projects
have been and are being undertaken around the world to use the Internet as an
easy and direct way to access large sound collections. EastWest
(www.soundsonline.com) serves 16 bit, 44.1k, WAVE and AIFF sound (over 20,000
instrumental sound samples). Search is done by standard HTML forms on
categories and keywords. The sound can be previewed in Real Audio. Download is
accelerated by E-magic ZIP compression (30% without data loss) after online
payment. However, sound quality is not above standard CD quality. In comparison
to Studio Online's 120,000 sounds, the repertoire is limited.
Sound Dogs
(www.sounddogs.com) serves over 60,000 sounds and special effects for the
cinema (more than 110 GB of data). Search is by categories and keywords, but
there is no sound preview yet. Downloading is not interactive at all, for sound
is sent by e-mail. The sounds are of a high quality (up to 24 bit, 48k sampling
rate) and many different formats are available. These services, however, are
not interactive as Studio Online is in the sense that an HTML request is sent
and then the user waits for the answer. In Studio Online, client-server
interaction is immediate and two-way.
"Studio On
Line" (www.audiosoft.com), not to be confounded with Studio Online, serves
16 bit 44.1k sounds and samples for post production studios and professionals.
Specific hardware is needed to use this service: a dedicated Client Computer of
a specific brand is preconfigured with a Digital Video Broadcast Card, which
decrypts audio data from a dedicated satellite connection (Astra Net) in real
time. The immediate access (faster than Studio Online, which is not real time)
must be paid, in addition to the renting of the service, by proprietary
technology and dependency on a specific connection service.
None of the
mentioned sites are really musical sites with an artistic vocation but
commercial delivery services of some big players in the audio and multimedia
market. There are no sound transformations offered as in Studio Online, no
interactive navigation on the content of the sound databases or graphical
control of parameter configurations as there are in Studio Online like Break
Point Functions, compression graphs, and the like, which are only possible
through Java programming.
5. A Typical Session With Studio
Online
One should not
try to connect to Studio Online with an outmoded computer and obsolete browser
software. Required is a fast Pentium or comparable processing power, sufficient
memory on top of what is already consumed by the operating system and browser
software (which is much), Microsoft Explorer 4.01 (build no. 4.72.xxxx.xx) or
Netscape Communicator 4.5.
5.1. Startup
On visiting the
Studio Online Web page containing the client applet
(http://sol.ircam.fr/external/joba) a new browser window opens, the Java Virtual
Machine starts up and the compressed Java archive (ca. 3 MB) is loaded. It
takes additional time until the classes (a couple of hundreds) are verified by
the Java security system and instantiated. This can take a while, and both
Netscape and Explorer do not really indicate the progress of this procedure, so
one must be patient. (Netscape seems almost frozen during this period, while
with Explorer one can easily go on surfing in another browser window.) If the
user has configured a sufficiently large Browser cache (10 MB, say) to keep all
the loaded Java classes on the local machine, the applet will start up within
seconds the next time. This is because the browser then just checks if the
local classes are up to date and if so, verification and download of code is
skipped.
5.2. Selecting Sounds
The applet
first presents the user a login screen where he/she can type a user name and a
password. (On a first visit, the user just enters a user name and a password of
his/her own choice in order to identify him/herself on later logins.) On
pressing the login button, the applet instantiates the ORB classes, connects to
the IRCAM server and opens a session. The user preferences of the last session
are retrieved (default is AIFF/16 bit/44.1 kHz as sound format and English as
language). The user is first presented with a "sound selector" screen
which actually is a graphical frontend to the database containing the sound
taxinomy. The user selects the attributes of the desired sounds among a number of
categories (instrument, playing mode, dynamic, pitch(es), octave(s), channel
configuration, etc.) and sees how the selector interface updates in order to
show the choices that remain, until the number of hits is reduced to less than
or equal to 24 sounds (e.g. a quarter tone octave).
5.3. Managing Sounds: Downloading,
Uploading, and Transforming
The user can
then load the set of these sounds into the "sound manager". This is a
directory view on the user's workspace and the central part of the application.
From here the user can transform selected sounds and recursively create
subdirectories containing the results of these transformations (which we call
"productions": the resulting sound file(s), some analysis files, the
parameters of the transformation and a log file). The directory structure thus
reflects the transformation "history" of an original sound. The
transformation result is automatically previewed by the audio helper
application the user has configured in the browser's preferences as soon as the
transformation is done.
An interesting
chain of transformations, for example, is to split off the noise part of a
noisy instrumental sound (e.g. a sul ponticello on a double bass or a flutter
tongue on a trombone) with additive resynthesis, to transpose it two octaves
higher and to time stretch it by a factor of four. The result is a most
interesting sound which has lost almost all similarity to an instrumental
sound, while it still benefits from the complexity and richness of a natural
acoustic phenomenon. The user can also be interested in downloading the
spectral analysis data, and use them on his/her own computer by displaying them
graphically with standard software or even inputting them into his/her own
software.
The user can
upload his/her own sounds up to 20 MB per file to a private directory on the
IRCAM server. WAVE, AIFF, AIFC (uncompressed) and IRCAM floating point/short
sample sound file formats are recognized. These sounds can be transformed in
the same way as the database sounds and downloaded again, or left on the server
for the next session with Studio Online. Up to 300 MB of disk space can be
claimed by the user. The transformation results are stored on the server in
IRCAM floating point format in order to preserve a maximum of sound quality, so
this quota might be reached after some transformations of lengthy sound files.
In this case, the user is invited to delete some unwanted results before he/she
can proceed producing new sound data or uploading more sounds.
5.4. Configuring User Preferences:
Language, Sound File Format, and Compression of Sound Dynamics
At any time,
the user can reconfigure his/her preferences in order to adapt to his/her
specific local environment (typically PC/MAC/UNIX), linguistic background
(English or French speaking), and audio system (does it support Studio Online's
extreme 24 bit dynamic resolution?). Pianissimo sounds from Studio Online can
be extremely weak when played back on a 16 bit audio system, so we devised a
graphical static compressor which permits to directly define a compression
curve (linear compression and an optional constant offset) to adapt 24 bit
dynamics (about 144 dB) to, say, the 16 bit range (about 96 dB). All of these
configurations are immediately taken into account. For example, as soon as the
user switches e.g. from English to French, the configuration interface itself
becomes French at once (with all accents, of course), as does the rest of the
client applet. The same holds for sound transactions: if the user changes from
AIFF to WAVE, the next sound downloaded comes as WAVE with Mime Type
"audio/wav", and spawns the corresponding application configured in
the browser's preferences.
5.5. Search by Psychoacoustic Similarity
This interface
complements the systematic choice of the sound selector interface. It is a very
powerful search engine through the whole database, across instruments, playing
modes, pitches, etc. by preprocessed comparison of spectrally analyzed content
only. Surprising results can be obtained and sounds detected that one would not
have found by looking up the sound taxinomy [Hoffmann/Misdariis 1998].
This interface
has been made much more intuitively by the introduction of evocative terms and
the concentration on 3 major perceptual categories of sound: brilliance,
richness, and attack. Additional constraints can be requested by the user on
spectral energy and/or pitch, restricting the hits to a certain distance in
these parameters, and a distinction between percussive and non-percussive
sounds can be enforced. Up to 50 found sounds can be compared to the original
sound and added to the sound manager for further treatment.
6. Technical Data of Studio Online 3.0
Studio Online
3.0 is what could be called an "Internet Killer Application"
[Orfali/Harkey 1998]. It makes use of the latest achievements in distributed
computing to connect Internet clients to a sound server and to provide a
sophisticated graphical user interface for the remote query process, the sound
transformations, the remote managing of the sound files, and the retrieval of
the sound data. After trying standard Internet technology like HTML pages and
CGI scripts, it was found that only a distributed Client/Server architecture
could satisfactorily respond to the needs of interactively navigating in a
large sound taxinomy, providing a session oriented workspace, supporting
intuitive graphical controls and diagrams, and allowing the user to configure
his/her own persistent preferences of sound format, language, compression
level, etc. [cf., e.g. Weisbecker/Bauer 1998].
The version of
Java used is Java 1.1.3, with the JFC/Swing 1.03 layer on the client side. The
ORB software used is Visibroker for Java 3.2 and Visibroker for C++ 3.1. The
database used is the Oracle 8.0.3.0 database accessed by JDBC 8.0.4.0.6 Level 2
drivers.
The Web server
is Apache 1.2.4, and the dynamical sound conversion scripts are launched by CGI
calls with Perl 5.0.3. The conversion programs are IRCAM's STtools toolkit
(updated and completed by the Studio Online Team).
The server
machine is a Sun Microsystems Enterprise UltraSparc 3000 running Solaris 2.5.1.
The disk array
is a Sun Microsystems Raid RSM 2000, configured Raid 5, with 31 disks à 9 GB
each, plus two hot spare and a cold spare disk. This gives us, after
formatting, a usable disk space of 190 GB. The database disk (1 HD)is fully
mirrored with RAID 0+1.
The sound total
is 113.823 sounds with 6 tracks each, sampled with 48k, 24 bit (ca. 130 GB of
data, which correspond ca. 12 days of continuous listening).
7. New Features in Studio Online 3.0
During 1997,
brilliant developer Rolf Wöhrmann created, in close collaboration with
Guillaume Ballet, versions 1 through 2.6, gradually fixing the architecture as
described above [Wöhrmann 1997, 1999]. This version was a fully working, albeit
somewhat restricted prototype which already served the needs of a constantly
growing user group. During 1998, the sound database was completed to 16
instruments by the recording and editing team, and the software system was
enhanced, completed, and tuned by the new development team. Startup of the
client applet was optimized by loading all classes "on demand". Full
support of the Windows platform was added (WAVE sound file format, support of
the Explorer way to handle multimedia data, etc.), sound transformations became
stereo and could be indefinitely chained. The client interface was completely
redesigned, became fully resizable and uniform on all platforms (Swing/JFC look
and feel), graphically much enhanced by the hierarchical directory view in the
sound manager with indication of the download size, graphical break point
function editors for time-dependent transformations, menus and tab controls
instead of buttons, etc. Many minor improvements were done to the client, the
servers, and the various administration tools. Some small but useful features
were added, like e.g. the progress messages during transformation processes and
the possibility to abort them, the possibility to download multiple sounds in
one archive file or to automatically order sounds on CD, and automatic
dithering of 8 bit sound. More information is to be found in the release notes
[SOL].
8. Some Problems Encountered
We had much
trouble when upgrading from the Java 1.0 AWT to the Swing/JFC library. All Swing
GUI action is done in the user thread which blocks on synchronous remote method
invocations. Since we cannot predict the response time of such remote calls we
had to program our own threading policy in order to refresh the user interface
during a remote call. We finally decided to launch a thread for each remote
connection. But then we had the difficulty of synchronizing these threads,
especially in order to prevent a user nervously clicking on the interface and
flooding the server with newly opened threads (we realized this problem when
during testing, our JDBC drivers simply blocked the whole system without any
error message after we hectically clicked some hundred times in the interface).
This considerably complicated the programming of the selector interface. We
would have wished the Swing/JFC layer to implement the same threading policy as
the old AWT.
We also had
much trouble with a bug in the Gatekeeper which blocked without any error
indication after a couple of days. I shall not mention the dozens of minor bugs
in the various Swing/JFC betas and other used software (especially our Java
development tool). But the most annoying bugs and deficiencies were found in
the browser software: no resizing of the applet in Netscape, no upload of
sounds possible with the Explorer on the Mac, to name only the most flagrant
ones. Microsoft did not correctly port their current Java Virtual Machine to
the MAC, and the Apple MRJ is so sloppily implemented that one has to invent
dozens of workarounds in order to be able to use it at all. See [SOL] for
details.
9. Future Work
Studio Online
is actually a project designed for the future. We used the latest technology
available at the time, much to the disenchantment of some of our users which
were not able or ready to upgrade to the latest Web Browser versions with a
Java Virtual Machines fully compliant to the Java 1.1 API. We needed the new
Java version as well as Sunsoft's Swing/JFC classes for the advanced graphics
in the applet's user interface. We even would have preferred to migrate to Java
1.2 if it had been available at the time, for it integrates the ca. 2 MB of JFC
class code that have to be downloaded on the first use of Studio Online. When
using Nescape Communicator which already integrates Visibroker's CORBA classes
(albeit in an older version), the applet could even shrink again to its old
size of some hundred KB (which at the time gave it the nickname
"JORBA" for "just one really big applet"[3]). Java 1.2 would also invite to
replace the current CGI driven downloading and playing of audio data by an
integrated solution using (an improved form of) the Java Media Framework, with
its support of many more file formats, etc.
Despite its
success, Studio Online should be considered only the beginning of a much more
ambitious project of internetworked sound processing, transparently distributed
sound storage and access, and the design of component software for music
purposes. It has been shown that Studio Online technology can solve some of the
problems of a large and historically grown research institution as IRCAM, but
there is still a long way to go.
For example,
the current integration of IRCAM sound processing tools works by command line
wrapping which is a tedious, limited and suboptimal way to expose them to
remote use. There are 3 to 8 separate processes launched per transformation:
the various sound conversion tools to feed the source sound into and to
retrieve the processed sound from a command pipeline, the splitting of stereo
sound into mono channels for the Additive engine which itself is but a script
executing command lines. Under these conditions, it was already an achievement
to implement such basic remote controls as a gauge to monitor the progress of a
transformation and a button to remotely abort the transformation process!
Instead of
launching shells executing assembled command line strings, the sound processing
engines should be encapsulated as multithreaded CORBA servers. Once there will
be stable versions and well-defined control interfaces of these engines, it
will be much easier to develop intuitive and fault-tolerant integrated systems
using them. It would then be feasible to develop unified graphic interfaces for
expert configuration of these sound processing engines which have remarkable
possibilities if used in advanced mode.
Studio Online
should develop into a testbed for advanced tool integration, sound
internetworking, and audio research with the help of distributed objects. There
is no lack of ideas: projects have been conceived of treating the integration
of distributed databases, advanced content search in audio documents (instead
of the current table lookup), integration of more IRCAM software, etc.
10. Conclusion
Studio Online
has been one of the few realizations answering the call for projects on the
"Information Highway" that have been successfully completed in time.
This fact makes it all the more deplorable that there has not been any kind of
immediate follow-up project. Aside from the fact that much more functionality
could have been added to and on top of Studio Online, it is with the expertise
gained during this project that even more interesting inter/intranet
applications of distributed processing for musical purposes could have been
envisaged.
References
[Ballet
1998] Guillaume Ballet, Vincent Puig, Hugues Vinet, Projet Studio en ligne, Rapport final, IRCAM, Nov. 1998,
unpublished.
[Ballet
1998] Guillaume Ballet, Rapport
d'avancement du projet Studio en ligne I, II, III,
http://sol.ircam.fr/~ballet/rapports/* and http://sol.ircam.fr/docs/sol3.0/
[Hoffmann
1998] Peter Hoffmann, Studio Online III: Distribution and Interaction. Detailed
Space-Time Diagrams of main SOL Functionalities,
http://sol.ircam.fr/docs/sol3.0/distribution/*.
[Hoffmann/Misdariis
1998] Peter Hoffmann, Nicolas Misdariis, "Studio Online 3: Access to IRCAM
Sound Database and Sound Processing Tools from the Inter/Intranet",
Presentation at the 2nd CUIDAD meeting, IRCAM, Dec. 1998,
http://sol.ircam.fr/docs/sol3.0/slides/SOL.html
[Lévy 1998]
Fabien Lévy, Rapports d'enregistrements
de Studio en ligne, IRCAM 1998, unpublished.
[Lévy 1998]
Fabien Lévy, Rapport sur la taxonomie de
Studio en ligne, IRCAM 1998, unpublished.
[Orfali/Harkey
1998] Robert Orfali, Dan Harkey, Client/Server Programming with Java and CORBA,
2nd ed. Feb. 1998, John Wiley & Sons.
[SOL]
Online documentation and help at the StudioOnline site:
http://sol.ircam.fr/help.html, http://sol.ircam.fr/docs/sol3.0/*
[Weisbecker/Bauer
1998] A.Weisbecker, S. Bauer, "Vom Spaghetti-Code zur Komponenten-Architektur:
Die Genesis der Anwendungsentwicklung im Intranet-Umfeld",
Focus/Computerwoche Aug. 1998, pp. 22-24.
[Wöhrmann
1997] Rolf Wöhrmann, "Das Studio Online-Projekt am IRCAM",
Mitteilungen 26 der Deutschen Gesellschaft für Elektroakustische Musik (DEGEM),
3.9.1997, p. 16-21, htp://www.kgw.tu-berlin.de/pub/DegeM/
[Wöhrmann
1999] Rolf Wöhrmann, Guillaume Ballet, "Design and Architecture of
Distributed Sound Processing and Database Systems for Web Based Computer Music
Applications", Computer Music Journal, forthcoming.