RTcmix and the Open Source / Free Software Model
David Topper
University of Virginia
McIntire Department of Music
Virginia
Center for Computer Music
Charlottesville, VA 22903
topper@virginia.edu
Abstract
While not entirely by design, Cmix has been an open system since its inception. Sharing code between a core group of developers was more or less the rule of the day for Unix software developed in the 1980s. The rapid expanse of computer music applications under the Nextstep operating system furthered this trend. Similarly, in the early 1990s, the still open nature of the Cmix source code, combined with newer and faster CPU technology, led to the development of a real-time version called RTcmix under the SGI platform. Soon thereafter, a Linux version appeared.
Work now continues on RTcmix in many different areas. The internal scheduling algorithm has been refined. The ability to load instruments dynamically at runtime has been added. Dozens of interfaces have been developed to send control messages to the RTcmix synthesis engine. An API for creating such interfaces has even been designed. Data can be read from MIDI and special control devices. There are also now several mailing lists and online bulletin boards where people exchange ideas and
bug fixes.
This paper will briefly document the history of (RT)cmix and evaluate its development from within the framework of the Open Source software paradigm. A primary contention being that this software paradigm offers unique benefits, particularly to the field of computer music, one that capitalizes on the sense of community and freedom of expression that the Open Source model promotes.
Recent applications developed under SGI and Linux will also be presented from that perspective. That is, each application will be briefly demonstrated and explained from the perspective of the program’s author (paraphrased by the paper’s author), documenting how RTcmix and its Open Source model allowed for a powerful type of artistic freedom both conceptually and technologically. Problems with the Open Source model will also be addressed such as quality and version control. Future direction will be postulated. Question and answer feedback will be solicited.
I. Open Source History
The idea of infinite recursion seems to best explain the virtues of Open Software: an idea taken into many hands, each with the ability to refine and evolve the original notion. The paradigm promotes progress and evolution by design in that it allows anyone to join in and contribute, building upon previous work. An outgrowth of this mass evolutionary characteristic is interaction. Open software projects have a tendency to interact with each other. The idea is in essence very similar to the scientific method, where work is published, reviewed, and often built upon by peers. Imagine if Einstein or Newton placed a copyright on their work. Dr. Mann of Wearcomp8 provides several eloquent discourses.
conceived in 1983, founded by Richard Stallman. A related organization,the Free Software Foundation, works in part to assist the continued development of GNU. The GNU Public License is a commonly used document to make a given piece of software, free software. The GPL basically states three things.A user may or may not pay for GNU software, but once in possession, the user can:
1. freely copy the program and give it to friends and coworkers,
2. freely change the program, having access to the source code,
3. freely redistribute and improved version, and even charge for it.
A familiar phrase within the community is, “The word `free’ above pertains to freedom, not price.”
The Open Source software movement2started in 1998 and was sparked into creation largely by Netscape’s announcementto release its source code. The definition of Open Source is taken from the Debian (a Linux vendor) Free Software Guidelines, modified from the Debian/GNU Linux distribution. So much of the GPL remains here. Open Source attempts to address the business community, to help them formalize business plans using the Open Source model.
II. Open Source Projects and Issues
projects would include: the Linux operating system3, the FreeBSD operating system5, the Apache web server6 and the Perl programming language4. A good deal of “free software” also exists outside the scope of either Open Source or GNU. There are hundreds,if not thousands, of free software programs currently available.
Linux is a particularly good example to look at. In 1998, the Linux user base grew more than any other operating system.It has been continuously refined by programmers around the world, always incorporating the latest technological advancements. It has also provided the community with a robust platform to develop even more open software.
There are some major issues facing Open Source development, which are addressed quite well by the GNU Manifesto and the Open Source project. One key issue is financial: how do programmers make money? It is completely possible to charge money for Open Source software.Many firms generate revenue and employ programmers by providing support and branding, offering a value added service to users.
Another glaring problem with the Open Source paradigm is organization. With so many people working on projects, it seems almost impossible to coordinate programming efforts. Many projects make use of some kind of version control system. A particular system gaining popularity is Concurrent Versions System (CVS)7, which is of course also under the GPL. With CVS, programmers can check in and checkout, via a network connection, various parts of a program. Coordination of team efforts is often directed by a core group, and communicated via mailing lists. Open Source projects illustrate the importance of the Application Programmer Interface (API), the means by which programs communicate with other programs. Many closed source software packages offer APIs, in essence it’s the only part of them that is open.
II. Cmix History
Cmix was derived from the MIX program, a20 track mixer written by Paul Lansky in 1978, which ran on VMS in FORTRANon IBM mainframes. Synthesis abilities were added, and Paul ported Cmix to run on a PDP11/34 in 1983-84 under BSD2.9 Unix. In 1985, he moved it over to Ultrix on a DEC MicroVAX. He and Lars Graf added the MINC parser in 1987, with help from Brad Garton and Dave Madole on various parts.
Doug Scott and Paul ported an initial real-timeversion to SGI in 1995. While sometime in 1993, I took a first stab at a Linux port which had to use Sox to unswap big-endian files. Then in late1995, Brad Garton and myself created a real-time version with a scheduler.Those enhancements were incorporated into the Linux version. Brad alsoadded the ability to read TCP socket data. That version was presented at ICMC ’979. Shortly thereafter, Luke DuBois added an API forwriting to TCP sockets, which facilitated interface development. Doug Scott set up a mechanism to dynamically load instruments at run time in 1997.The Linux version was brought up to date shortly thereafter. Enhancements have continued, most recently with John Gibson’s interface to Bill Schottstaedt’s sndlib for audio file IO, continued development of real-time pfield control and the setup of a CVS server for the Linux version. Work is under way to put RTcmix under the GPL. None of this development would have been possible if the original Cmix and subsequent RTcmix versions had not been open.
IV. RTcmix Future and Projects
Future RTcmix versions can hope to employ:shared audio buffers, more robust real-time pfield control, interfaces to other parsing languages such as Perl, the integration of audio with graphics and video output, support for SMP, and of course more interfaces.With an Open Source paradigm, anything is possible. Many RTcmix projects are often designed to suit specific performance and/or compositional needs.
There have been many Cmix/RTcmix projects over the years. Some work was lost when development of the Nextstep system ended, but much is currently being resurrected. Recent projects employTCP/IP communication protocols to send control messages to the RTcmix engine.Following is a short description of some projects:
Virtchla: Luke DuBois, Columbia University
Luke is considered by many to be the resident expert on Buchla operation at Columbia’s CMC. An outgrowth of his analog experience has resulted in the creation of the Virtchla, where a sequencer sends control data to RTcmix.
Treembre: Doug Geers, Columbia University
Doug is both a programmer and composer atthe Columbia CMC. The Treembre program allows pitch objects to be organized in a tree-hierarchy, which can be modified at any level, maintaining subtree relationships.
Patchmix: Mara Helmuth, CCM, Cincinnati
Mara is the Director of CCM2, at the Universityof Cincinnati. Patchmix was originally designed under Nextstep for use with Cmix. One can create a flow chart of unit generator icons which are turned into Cmix code, compiled and run within Patchmix.
Pslider: David Topper, VCCM, Virginia
I am the Technical Director at the Virginia Center for Computer Music. The pslider program allows a user to control any RTcmix instrument’s pfield in real time. It hopes to be a general purpose tool to control RTcmix interments. It was designed under Linux, using GTK,a free GUI toolkit.
V. Conclusion
There are very real economic factors to consider when looking at Open Source and computer music. Most notably, many composers and students don’t have a great deal of money. This unfortunately puts many creative tools only in the hands of the select few who can afford them. Open software, and free software in particular, eliminate this problem and open up the use of high end tools to everyone, from the general (operating system) up to the specific (application).
It seems to make particular sense for tools to be open to avid users. Imagine a mechanic able to easily change the design of his wrench, or a doctor her scalpel. The point at which Open Source development meets developer is often the very home of creativity.What would inspire someone to contribute a new feature? Often times it’s a feature they would like to use and is not yet available. Such a development model promotes creativity and innovation by design.
Consider the diversity of individual studios.Few have precisely the same combination of hardware. What an amazing possibility,if everyone were able to have access to each other’s studios and offer improvements which were almost immediately available to everyone. Such a synthesis is already beginning, but hasn’t yet made a clear impact on popular music.
Anyone with a home studio has at one point or another confronted the frustrating prospect of getting everything to run on IBM mainframes. Synthesis abilities were added, and Paul ported Cmixt o run on a PDP11/34 in 1983-84 under BSD2.9 Unix. In 1985, he moved it over to Ultrix on a DEC MicroVAX. He and Lars Graf added the MINC parserin 1987, with help from Brad Garton and Dave Madole on various parts.
Doug Scott and Paul ported an initial real-time version to SGI in 1995. While sometime in 1993, I took a first stab at a Linux port which had to use Sox to unswap big-endian files. Then in late1995, Brad Garton and myself created a real-time version with a scheduler.Those enhancements were incorporated into the Linux version. Brad also added the ability to read TCP socket data. That version was presented at ICMC ’979. Shortly thereafter, Luke DuBois added an API for writing to TCP sockets, which facilitated interface development. Doug Scott set up a mechanism to dynamically load instruments at run time in 1997.The Linux version was brought up to date shortly thereafter. Enhancements have continued, most recently with John Gibson’s interface to Bill Schottstaedt’s sndlib for audio file IO, continued development of real-time pfield control and the setup of a CVS server for the Linux version. Work is under way to put RTcmix under the GPL.