CBC Radio on the Internet: An Experiment in Convergence

Andrew S. Patrick (Communications Research Centre)

Alex Black (Communications Research Centre)

Thomas E. Whalen (Communications Research Centre)

Abstract: An experimental trial was conducted to determine: (1) if there was any demand for regular radio programming distributed as digital audio files over the Internet, (2) if the Canadian Broadcasting Corporation (CBC) was willing to distribute programming in this format, and (3) what implications such a service would have for the corporation. Both sample and regular radio programs were digitized using computer sound equipment and made available via FTP, Gopher, and World Wide Web. Analysis of the traffic logs and a review of the comments submitted by users showed a very high demand (possibly all that was possible with this trial configuration) and a keen interest in the service. CBC has decided to adopt the trial and start a permanent service on the Internet. The trial identified a number of issues that will have to be addressed for this service, including archival storage, copyright, royalties, production changes, and electronic commerce opportunities.

Résumé: Un essai expérimental a été effectué afin de déterminer: (1) s'il existe une demande pour des émissions radiophoniques ordinaires distribuées sous forme de fichiers audionumériques sur l'Internet, (2) si la Société Radio-Canada (SRC) était disposée à distribuer des émissions dans ce format, et (3) les répercussions que pourrait avoir un tel service sur la SRC. Des émissions de radio échantillons et ordinaires ont été numérisées à l'aide de matériel de sonorisation informatisé et ont été offertes via FTP, Gopher et sur le World Wide Web. L'analyse de la densité du trafic et l'examen des observations présentées par les usagers ont indiqué une demande très élevée (probablement maximale, compte tenu de la configuration de l'essai en question) et un grand enthousiasme pour ce service. La SRC a décidé d'avaliser l'essai et d'offrir un service permanent sur l'Internet. L'essai a permis d'identifier un certain nombre de questions-clés sur lesquelles il faudra se pencher pour ce service, dont le stockage des archives, le droit d'auteur, les droits d'exploitation, les modifications à la réalisation d'émissions et les possibilités de commerce électronique.

CBC Radio on the Internet: An experiment in convergence

In the past few years there have been many predictions of a "convergence" of four or more traditionally separate means of communication. Telecommunications, broadcasting, publishing, and computers are predicted to combine in interesting ways to produce a "new media" that combines aspects of all these things. For example, traditional publishing is expected to change so that material will not be printed on paper, but will instead be transmitted via telecommunications links and be displayed on computer screens. Or, material that in the past has been broadcast, such as radio and television programming, may now be distributed via telecommunications links and made available on-demand and interactively. Digital encoding, widespread computer technology, and large networks will mean that many of the traditional media boundaries will no longer be valid.

Perhaps the best evidence that media convergence has already begun can be found on the Internet. More than 13 million people have access to Internet services worldwide, while another 14 million have e-mail connections to the Internet (see "New Data on the Size of the Internet and the Matrix," http://www.tic.com /mids/pressbig.html). A variety of material is now being transmitted over the Internet, including images, audio and video files, and text. Material that would normally be published or broadcast, such as newsletters or advertisements, is now being distributed on the Internet in digital formats. The Internet is becoming increasingly important in the business world, and it is being recognized as a social phenomenon as more and more people gain access to the network.

One of the interesting new media services on the Internet is digitized audio. Audio files have been available over the Internet for some time. One group, the Internet Multicasting Service, has been producing interviews for its Internet audience for over a year. The trial to be described here is the first time that audio programs produced for traditional radio broadcasts have been made available on the Internet on a regular basis.

The Canadian Broadcasting Corporation (CBC) is Canada's national public broadcaster. CBC operates radio and television networks in both French and English, and provides a variety of programming from news to music to entertainment. The CBC has a mandate from the Canadian federal government, its main funding source, to distribute Canadian content and foreign material to the entire nation.

The CBC uses a variety of technologies and distribution systems to achieve its goals. The purpose of this experimental trial was to determine: (1) if there was interest in, and demand for, regular radio programming distributed as digital audio files over the Internet, (2) if CBC was willing to regularly distribute programming in this format, and (3) what implications such a service would have for the corporation. A trial service on the Internet would also demonstrate the convergence that has been long predicted, and give CBC valuable experience in developing programs for this new media.

Running the trial

Digitizing radio programs

Audio digitizing chips and add-in boards are available for most computer systems. The most popular products for IBM-PC-style computers are the "Sound Blaster" boards, and many computers have sound equipment as a standard feature. The equipment used to prepare the audio files for this trial consisted of Sun (or Sun "clone") computers equipped with analog/digital interface chips. Audio digitizing chips convert analog audio signals into a digital approximation by sampling the audio signal at particular intervals. For the best results, the shortest possible sampling interval should be used, resulting in the highest possible accuracy. In commercial Compact Discs, for example, the audio is sampled 44,000 times per second (44 kHz) and two 16-bit numbers are stored for each sample. This produces a very good digital representation of the audio signal, at the expense of storage. Each minute of audio on a CD requires approximately 10 megabytes of storage. Transmission of the uncompressed digital audio signal, in real time, would require a channel data rate of approximately 1,280 kilobits per second (Kbps) or 1.2 megabits per second (Mbps).

Given that typical Internet link rates are often much less than 1.2 Mbps, and users' disk capacity is often limited, an audio format that uses a much lower frequency of sampling (8 kHz) and a much lower precision (8-bits mono) was chosen for this trial. In this format each minute of audio requires approximately 0.5 megabytes of storage, and it can be transmitted over a 64 Kbps channel in "real time." The resulting audio quality is far from that achieved by Compact Discs, but is equivalent or better than a standard telephone service. Moreover, for the type of CBC programming used in this trial, mostly spoken voice, the quality is usually acceptable. (The best results were achieved when the analog audio signals were passed through an equalizer before they were digitized. The audio digitizing chips have a limited dynamic range and using the equalizer to reduce the extreme high and low ranges and increase the mid-ranges produced better results.)

Another reason for choosing this 8 kHz, 8-bit audio format (often called "basic audio" on the Internet) is because it is the format supported by the common Internet tools. Internet navigation tools like Gopher, Mosaic, and Netscape are able to recognize and play these files, and this makes building interesting interfaces to the programs much simpler.

The disadvantage of the "basic" audio format is that even at this reduced quality the files can be rather large for transferring over the Internet. Many links on the Internet are 56 Kbps lines or less and the result is that, at the best of times, transfers of audio files will be slightly slower than "real time," that is, a five-minute audio file will take over five minutes to transfer. However, if someone else is also using the Internet link the transfer times may double, and it is common for the transfers to take three to five times longer than the length of the segment. The problem is even worse if a listener is connected to the Internet at lower speeds, such as the 14.4 Kbps rates used in many modems. Faster Internet links and /or a compression system for the audio files would help. However, standards and tools for audio compression have not been well established on the Internet yet, and upgrading the world-wide Internet is beyond our control.

Installing the server

The trial began in December 1993 when the first CBC Radio programs were prepared on a computer in the laboratory and made available through the Internet. The program files were made available via FTP, Gopher, and World Wide Web (WWW) using standard Internet server software. The trial was conducted on a server that was already well known as a source of Canadian government documents and a test site for a natural language information retrieval system. This site is listed in many network directories as an interesting site to visit.

Announcements of the experimental CBC service were made in various Usenet newsgroups. Also, a press release was issued and this attracted much interest from the traditional media. Stories appeared on the news wire services and in many newspapers, and various CBC radio and television programs provided coverage of the trial.

Initial program offerings

The initial files made available on the server were samples of CBC Radio programs. These included an episode of Quirks & Quarks (a science magazine show), an episode of Basic Black (a variety show), sample segments from Sunday Morning (documentaries), Christmas stories read by Fireside Al, and Ideas of Canada (a documentary about Canada).

These radio programs were made available "on demand" in that users could request them from the server at any time. The larger programs were broken into segments that were described in accompanying text so users could select only the parts of the program that were of interest to them. The result was that users could listen to the programs when they wanted. They also had control over the order of the programs, and they could select material based on the content that interested them. In addition, textual descriptions of CBC product offerings and a searchable catalog of transcripts were included with instructions for ordering products via telephone or postal mail.

Early reactions

The reactions to the trial in the first month were very positive. An e-mail address for sending comments was prominently displayed and people responded by saying this was a wonderful idea and an important innovation, both for the CBC and for the Internet. Users suggested that the service be expanded to include more programs and more current material. The only negative responses were complaints about the speed of transfers over the shared 56 Kbps link to the Internet used for the trial.

Expanding the trial

Based on the early reactions and comments, a decision was made to expand the trial service. An FM radio receiver was installed in the laboratory to constantly monitor the CBC broadcasts. Using a "cron" program, a Sun computer automatically recorded programs and transferred them to the server. Two newscasts (8:00 a.m. International and 5:00 p.m. Domestic, eastern times) were recorded each day and made available on the server immediately after the broadcast. Also, the Quirks & Quarks science magazine show was recorded each week, broken down into its component parts, and made available on the server.

Given the international character of the early audience, information and sample programs from Radio Canada International, CBC's shortwave broadcasting division, were added. This included sample radio programs in various languages and complete frequency and program guides.

To improve the transfer times an upgrade to the Internet link, from 56 Kbps to 256 Kbps, was arranged to take place in late April. "Mirror" sites that would offer a copy of the news programs were also recruited, and users were encouraged to select the sites closest to them.

Evaluating the trial

To evaluate the trial the log files of the FTP, Gopher, and WWW servers from December 15, 1993, through October 31, 1994 (321 days), were examined. The log files contained a record of all the server accesses, but did not identify individual users. Thus, the number of times a file was accessed or transferred was measured, rather than the number of people using the service. A single person may access a few or many files during a session, so it is difficult to translate the number of accesses to the number of people. All transactions originating from within the local domain (doc.ca) were excluded from the analysis, as was a small "logo" sound file associated with the WWW greeting page.

There were some fundamental limits to this trial that must be considered when interpreting the results. The trial was primarily a demonstration rather than a test of marketability and no attempt was made to sample a potential audience, or to advertise the trial widely. Further, because of the limited speed of the Internet connection being used, it is quite likely that possible users were discouraged by the slow response times. Thus, any estimates of usage probably represent a lower limit to the potential audience for this service on the Internet. Moreover, because some of the most popular programs were mirrored at other sites where log files were not kept, and some users made arrangements to share programs that they transferred, it is not possible to accurately count a total audience. Nevertheless, the traffic statistics do provide an approximate measure of the interest in this service.

Total traffic

The number of accesses per month for the CBC material, for each of the three server protocols, is shown in Figure 1. Some data was lost for the FTP server records due to hard-disk space limitations, and this is represented by the non-continuous line for the FTP data.

1 Total CBC Server Traffic for FTP, Gopher, and WWW Protocols

The traffic on the server was very heavy, with the total traffic in the last month (October) reaching a maximum of 59,857 accesses, or an average of 1,930 accesses per day. Assuming equal traffic throughout the day (which is probably not a safe assumption given that most of the Internet sites are located in the North American time zones), this is an average of 80 accesses per hour or approximately 1.3 accesses every minute. Moreover, this only represents the CBC traffic on the server. Overall, the server was accessed once every 15 to 20 seconds to provide one service or another.

The traffic for the trial service increased dramatically over the course of the experiment. There were some slight reductions in traffic in the summer months (July and August), but an overall steady increase. This indicates that interest in this material may not dissipate with time and either new users accessed the material each month and /or the same users came back again and again. The effect of the bandwidth upgrade in late April may also be evident as the WWW traffic increased in May and the following months.

Overall, the most common access method was the HTTP protocol used by the World Wide Web (WWW). It is interesting to note the growth of the WWW traffic over the course of the trial, as this parallels the rapid deployment of WWW software on the Internet (and the growth of the Internet itself ). The Gopher protocol was also used frequently, and Gopher accesses continued to increase throughout the trial, although not as dramatically as the WWW traffic. Although the WWW may be developing into the dominant Internet access service, the data shows that the Gopher service is still being used and Gopher traffic is still growing.

The FTP service was least used in this trial, but its traffic also increased somewhat during the period. FTP is the oldest and most primitive access service, but it is popular because various script and mirror programs can be used to automatically schedule FTP transfers. Much of the traffic seen with the FTP protocol may be mirror sites and other automatic downloads that users have scheduled.

Audio file traffic

More important than the total number of accesses for the entire trial service are the transfers of audio files. It is the radio program material that makes this service unique. Moreover, in the previous analysis the WWW traffic may be inflated since each Web page can cause a number of file accesses when "in-line" images are displayed along with the text and menus. In-line images are graphics files that are embedded in WWW documents and used to decorate or augment the textual information.

The news audio files (the twice-daily newscasts) were very popular and it was necessary to limit the number of simultaneous downloads from the server in order to maintain a functional bandwidth for the research campus. To apply these limits the news audio files were made available only via the FTP protocol and the FTP server parameters were set to limit the traffic to five simultaneous downloads (the mirror sites had unrestricted transfer ability). WWW and Gopher users could still access the news programs, but the menus and links pointed to the FTP service. Therefore, it is necessary to analyze the news audio file traffic separately from the traffic for the other audio files.

The number of transfers per month for the audio files that were not newscasts, again for each of the access protocols, is shown in Figure 2. The peak audio file traffic for the last month of the trial is 1,046 transfers per month, for an average of 45 transfers per day. The same pattern is evident here as was seen in the total traffic analysis: traffic increases throughout the trial period, and the WWW traffic is more frequent than Gopher or FTP. The reduction in traffic during the summer months is also seen in the data for the WWW service.

2 Audio File Transfer Traffic for Non-news Programs
for FTP, Gopher, WWW Protocols

The number of transfers per month for the news audio files, which were only available via the FTP server, is shown in Figure 3. News audio file traffic represented 56% of all audio file traffic during the trial. The news audio file traffic for the last month of the trial was 2,940 transfers, for an average of 95 transfers per day.

3 Audio File Transfer Traffic for News Programs
for the FTP Protocol

Combining the news and regular audio traffic data, the total for the last month of the trial was 3,986 transfers, or approximately 129 audio files per day. This represents a large amount of traffic on the server. The FTP transfer logs also recorded the time needed to do each of the transfers and an analysis of this data showed that the average audio file size was 3.9 MB (about 8 minutes of audio) and the average time to download an audio file was 17.39 minutes. This is a significant waiting time for Internet users and suggests that this trial was limited by the bandwidth available to transfer the files. An average of 129 audio files per day represents about 37.4 hours of file transfer traffic per day, just for the audio files.

Other audio programs

Quirks & Quarks, a science magazine show, was also regularly updated on the server. The hour-long show was automatically recorded each week and then manually broken into five- to ten-minute segments at the natural boundaries. The content of each segment was described in enough detail so that users could select those portions of the show that interested them and download the appropriate audio file.

The number of accesses per month for the Quirks & Quarks section of the trial is shown in Figure 4. This data represents all the transfer protocols (FTP, Gopher, & WWW) and all the file types (menus, text, images, and audio). The same pattern is evident here as was seen in the total server traffic and audio file traffic: a general increase in accesses throughout the trial and a quiet period in July and August. In fact, there were no new Quirks & Quarks files on the server during the summer because the show was in hiatus, so this section of the trial was not updated during this time.

4 Traffic for the Quirks & Quarks Program (All Protocols)

Quirks & Quarks audio file traffic represented 15% of all audio file traffic during the trial. The other programs available on the server represented only a small portion of the total traffic: Basic Black (light-hearted variety), 4%; Sunday Morning (documentaries), 3%; Ideas of Canada (audio composition about Canada), 3%; Front Porch Al (popular classic stories), 3%; Radio Canada International (multilingual shortwave), 2%; Brand X ("Generation X" variety), 1%; Cross Country Checkup (sample program on the Information Highway), less than 1%. These were sample programs that did not change during the course of the trial.

Supplemental materials

The server also contained supplemental information about CBC programs and the trial Internet Service. This supplemental material generated a very small portion of the server traffic: audio software tools (5%); transcript catalog (4%); program listings (4%); product lists (less than 1%).

The scope of interest

The server log files were also analyzed to determine the locations from which users were accessing the server. The log files record the Internet host name used for each connection to the server. This provides information about the Internet domain from which the connection originated. It is not always possible to map Internet domains onto geographic locations because countries like Canada and the U.S. can have hosts with a variety of domains (for example, U.S. hosts can use the domains "edu," "com," "net," "org," "gov," "mil," or "us"). However, the domain information does give a rough approximation of the location, especially outside of North America, and the Internet domain statistics do provide an indication of the scope of the interest in this trial service.

The most common Internet domain was the Canadian ("ca") domain, representing 44% of the total server traffic (147,594 accesses in total). The next most common domains were "edu" (16%), "com" (9%), "net" (4%), and "org" (2%). These are predominantly U.S. domains, although some Canadian sites are included here (e.g., the University of Toronto has an "edu" domain, and some of the new Canadian Internet providers offer service via the "com" and "net" domains). The most common domains associated with non-North American countries were (in order of traffic volume): uk (United Kingdom), au (Australia), de (Germany), jp (Japan), se (Sweden), ch (Switzerland), nl (Netherlands), fr (France), no (Norway), fi (Finland), it (Italy), dk (Denmark), at (Austria), be (Belgium), es (Spain), il (Israel). Each of these countries represented less than 2% of the traffic, and combined they accounted for 23% of the total traffic. In total, the CBC material was accessed from 58 different countries.

User comments

Users were encouraged to send comments via electronic mail, and correspondence from a total of 256 different people was received during the trial period. Overall the comments were very positive. An informal categorization of the responses was conducted and 135 responses (53%) could be described as "very enthusiastic" about the trial service. The most enthusiastic responses were from 46 Canadians (18%) who were living outside of the country. For example, one person wrote: "I recently moved to the San Francisco Bay Area (Cupertino to be specific) and I am blown away by your service. Every day I'm able to listen to my beloved CBC Radio News (I was a CBC listener before moving) and preserve my Canadian identity."

In addition, 121 responses (46%) could be described as "supportive" of the trial, while mentioning some problem with the service, suggesting other services that should be added, or requesting more information. Only five responses (1%) expressed negative comments about the trial--four said that it was pointless to distribute audio files over the Internet and another said that the service was redundant because people could get the same information from television.

The most common complaint concerned the time that it took to download the audio files. For example, 69 people (27%) commented that the audio files were too large, the transfers took too long, or a transfer of an audio file was interrupted. The amount of time mentioned by the users varied with the size of the file and the network distance from Ottawa, and one person in Sweden reported that it took five to six hours to transfer the news files. Fourteen people (5%) suggested that the files should be compressed to reduce transfer time. Nine people (3%) suggested alternative file formats that would be compatible with their personal computer (i.e., DOS and Macintosh users), even though tools to handle the audio files were provided on the server.

A total of 181 people (71%) made suggestions for more services: 48 people (19%) requested program transcripts; 33 people (13%) suggested more audio programming be included or some specific segment of a CBC show be added; 35 people (14%) requested a method for contacting people who work for the CBC; 28 people (11%) recommended adding programming schedules for CBC and RCI (which was done); 13 people (5%) requested more audio file conversion tools; 9 people (3%) requested the addition of French language programming; and 7 people (2%) said they would like to order products from the CBC via the Internet.


Interest in the service

The first research question for this trial was to determine the level of interest for receiving regularly broadcast radio programs in digital format over the Internet. The traffic that was logged and the comments that were received suggest that the demand for such a service is very high.

In fact, there was ample evidence to suggest that demand for this service was higher than the configuration was able to support. The increase in traffic starting in May when the bandwidth was upgraded suggests that bandwidth restrictions may have been limiting the traffic prior to May. Even after the link upgrade, the most frequent complaint from the users was that the transfers were too slow and this made the service somewhat frustrating. Also, the analysis of download times shows that the audio file transfers were quite slow.

The results demonstrate that there is a demand for the audio files in particular and that the supplemental text and software material constitute only a small part of the service. In fact, it was necessary to limit the transfer of news audio files because the shared network bandwidth available was not able to support the demand. This suggests that enough Internet users are capable of playing the audio to sustain this level of traffic.

Even with these limitations, the traffic increased during the duration of the trial with the peak month coming at the end. This suggests that such a service may have some sustained appeal and may not just be a novelty. It was also evident that the material that changed frequently, the newscasts and the Quirks & Quarks show, were also the most popular. This may be evidence that people were using the service regularly, and one might expect sustained traffic over a much longer period if timely material is provided.

The most frequent access method for the programs was the World Wide Web. The data from the trial is a good illustration of the growing popularity of the WWW. However, it is interesting to note that Gopher traffic continues to increase, and there is non-trivial use of the FTP service. It is clear that a complete service package should still provide support for all of these protocols.

Another interesting result is the finding that people from within Canada made heavy use of the server. CBC Radio is broadcast to most parts of the country and yet Canadians used this service frequently. This suggests that the service tested here fulfills some demands that are not met by traditional broadcasting. The on-demand nature of the service allows people to listen without being constrained by the broadcaster's schedule, and this may account for this finding.

There was also a large international interest in this service. The frequency of access from various countries appears to reflect the prevalence of the Internet outside of North America--most common in the United Kingdom and Europe, with many other sites in Australia and Japan. Many users commented that this service allowed them to get Canadian material that was unavailable by any other means. This Internet service is international in scope and audience, and it is difficult to offer a similar service with traditional broadcasting technologies.

Adoption by CBC

The second research question was whether CBC, a very large and traditional national broadcaster, was willing to pursue the road to convergence and participate in such a new service delivery. The results of this trial and the interest in the service have been very convincing. CBC has decided to adopt the service and make it a permanent offering. CBC has purchased a connection to the Internet, installed a new server computer, and is retraining staff who used to work in traditional radio production and corporate computing on the development and operation of this new Internet service (the CBC server can be reached at ftp://www.radio.cbc.ca or http://www.radio.cbc.ca/ ). This represents a significant investment by CBC, and a commitment to the long-term future of this new service. CBC has already gone beyond the service offerings available during the trial and they have included a broader range of information and programs.

In fact, the level of interest within CBC has been so strong that many production and promotion groups have asked to be included in the service before the transfer of technologies and expertise from the laboratory has been completed. Production groups want to use the Internet to distribute their programs, to provide schedules and advanced information about the shows, to give background information, and to communicate with their audience. It is clear that CBC sees the Internet as a valuable method for expanding and enhancing their existing services, and they are more than willing to adopt this new technology.

Implications for CBC

The final question of interest was the possible impact such a service would have on CBC. Although the full implications of this new service will not become apparent until it has been operational for some time, this trial did identify a number of areas that will require attention. Future research should examine the long-term impacts of the service for CBC and its listeners.

The distribution of programming in a digital format can be different than traditional broadcasting. Material that is broadcast is very short-lived: if it is not consumed at the time of broadcast it will likely never be consumed. The material can be recorded for later use, but this is rare, especially for radio programs, and the copies may not be as good as the original.

On the other hand, material that is offered digitally on a server can have a much longer life. Listeners can transfer the files at any time, and material may be preserved for long periods in the original condition. Initially, this led to a fear that the material may not be timely. News and current affairs shows pride themselves on being current and up to date, and there was some concern about users transferring older material without realizing its age. In practice, it was necessary to remove files from the server regularly due to their large size so this was not a problem. There was also the opposite concern--some material should be archived and made available permanently, but this requires a large investment in storage hardware and useful search and retrieval technology.

The issue of copyright and digital duplication was also raised. The CBC material is freely broadcast and anyone is now able to record it. The concern was whether the digital format might make unauthorized mass duplications or alterations more likely. CBC must rely on the same copyright laws that it uses to protect any of its materials, but it is not clear whether the new digital format makes violations of those laws more likely. CBC will have to monitor the use of the digital products they offer on the Internet to determine if copyright problems are more common than those experienced with traditional broadcasting methods.

Another issue was royalty payments to artists for use of their material in these "new media" products. When commercial music is used in traditional radio programs there is a system for paying royalties to the appropriate artists based on the frequency of use. When material is offered digitally on-demand it is not clear how royalty payments should be calculated, especially since the material may be easily shared and passed on (this was done deliberately with the mirror sites). The artist's unions have no experience in this field so it was difficult to communicate the issues and reach an understanding. Commercial music was avoided during this trial by removing those segments from the programs that were offered. However, a long-term arrangement for the use of this material will have to be put in place in the near future to avoid the extra work of removing this material.

CBC also offers a number of value-added products, such as audio tapes, CDs, T-shirts, books, transcripts, and so forth. It may be possible to increase sales in this area simply by making a catalog of these products available on-line. There is a keen interest within CBC in making these products available over the Internet, and methods for electronic ordering and payments are being explored.

Producing material for distribution on-demand also led to some interesting issues for the production process. CBC production offices spend time "packaging" each show into a complete and coherent unit before it is broadcast. For the on-demand service, however, the first step is to break-up the package into its component parts and offer each segment on its own. Thus, material was requested earlier in the production process than the producers were accustomed. Moreover, the on-demand service could accommodate more background material than was normally included in the limited broadcasts. The Internet service could also distribute visual material that is not normally used in radio productions. The production process will likely change to accommodate these factors as on-demand services become more common, and this may be an interesting topic for future research.

Also, it was very helpful to have the program logs and scripts available when installing a show on the server. Each show has a menu attached to it to describe the contents of the various parts, and understanding the structure of the show and describing the contents is much easier with the production notes. These notes are also necessary to get names and attributions correct.

One area where a weakness was found was the quality of software tools available for editing the digital audio files. Sun computers come with an "audiotool" program for recording and manipulating sound files. This tool represents the audio in a file as sections of sound and silence so that users can select portions of the audio by moving a pointer over the graphic representation. The value of this tool was limited, especially when used to break long recordings into short segments. The resolution of the representation for files of more than 10 minutes is too low and better audio editing tools are being investigated.

In conclusion, this trial showed that there is a demand for regular radio programs distributed over the Internet, and such a service can be valuable for a traditional broadcaster such as CBC. CBC has accepted the results of this research and they are building such a permanent service on the Internet.