Linux For Suits

January 2002

Open Source Radio:

An Interview with Wild Bill Goldsmith of KPIG and Radio Paradise

KPIG is an anomaly. It's a top-rated commercial radio station programmed entirely by its own DJs. It's a serious radio station with a terrific sense of humor. It's as much a fixture in its community as a statue in front of a courthouse -- and a lot more fun.

The station's format has been described as "mutant cowboy rock & roll," but it's hard to make any label stick on a station that is more real, more fun and more like-it-oughta-be than (by my conservative estimate) all the other commercial music stations in the country put together. It's roots go back to the legendary KFAT, and beyond that to the KRAB Nebula of loosely affiliated noncommercial stations out of which "community radio' and "free form radio" evolved in the Sixties. It's a history that's as old and woolly as UNIX's.

The resemblance doesn't end there, either.

The words "free" and "open" are taken seriously by KPIG and its faithful, who contribute code in the form of original voices, programming ideas and even musical selections. KPIG is fortuitously located in (no kidding) Freedom, CA, and has one of the worst signals in the Salinas-Monterey-Santa Cruz market. If you're in the bowl of mountains that surround Monterey Bay, chances are you can get it. If you're not, your only choice is to punch up and choose among a wide selection of streams, including a 128Mb MP3 spigot that's one of the prettiest-sounding signals that ever poured out of your speakers.

KPIG was the first commercial radio station to broadcast on the Web, and it has blazed trails ever since. This last Spring, when AFTRA decided its commercial talent should be paid as much as 300% extra for commercials going out over the Net, thousands of commercial stations shut down their streams. KPIG just hacked around it. When the national ads come on, webstream listeners get pleasant filler.

To the constant astonishment of everybody but the station and its listeners, KPIG consistently manages sits near the top of the local Arbitron ratings, despite its second-rate signal. It is, by all measures, a success. It is also full of open source technologies that are in a good position to burn down Broadcasting As Usual if the right hackers and serious radio fanatics get their hands on it. That's the idea behind what you're reading right now.

KPIG's hacker-in-chief is "Wild Bill" Goldsmith, a KPIG veteran going back to KFAT. He's now living full-time in (no kidding) Paradise California, where he has steadily improved his own solo effort, Radio Paradise, alongside KPIG. I decided to get in touch with Bill after I received email raves about Radio Paradise almost simultaneously from friends in Seattle, New York and North Carolina. The sources were old radio pros who have since moved on to other professions, but know the Real Thing when they hear it.

It was clear from a quick look at both KPIG and Radio Paradise that the stations were up to something more than radically good radio programming. Technically, both were easily customizable hacks built on the open source software and generic hardware. Specifically, Bill had put together what appeared to be a tightly integrated system that involved the music library, the Web site, the whole audio chain, accounting, operations, and scriptable tie-ins to potentially money-making partnerships with anybody who wanted to cross-promote anything, including stores, artists, labels, or whatever -- in a way that allowed as much live operation or automation as one saw fit.

This appeals to me. I'm a radio freak going back to my childhood, when I was a ham radio operator (even today the only code I know is Morse). My nickname, "Doc," is the fossil remnant of the name I used on the air when I worked at WDBS, a KFAT-like station in North Carolina back in the Seventies. Commercial radio today is a snatched body that bears only superficial resemblance to anything anybody loved in its golden age. Revitalizing radio is also a passion I share with Phil Hughes, Linux Journal's chief geek & publisher, an old radio hand who fondly recalls KRAB's golden days in Seattle.

When I wrote and asked Bill exactly what he was hacking together at the two stations, he wrote this back:

I suggested an interview and he agreed. We talked in late October, right after he finished putting up the latest KPIG website.

Hey, I like the new site. The Webcast links and the current song list are right up front. And I like the webloggy thing going on in the center of the page too. Makes it all very live. Very radio.

It should be more webloggy-looking very soon, after the staff at the station takes it over.

I've thought for a long time that the Linux and open source development community could do something fun to blow commercial radio out of the water. Now it looks to me like you're doing the pioneering work here. Are your tracks in the snow the ones we should follow?

I really think so. What I'm doing with the stations I'm working with is gradually evolving towards a complete open source package for programming, managing and scheduling a radio station, with an integrated Web site to go along with it. Basically everything that makes it possible to do what you see on the screen there at is all done with open source stuff. There is not one lick of proprietary software in there anywhere.

Give me a rundown of what I'm looking at when I visit the KPIG site. How is the content organized and served up here?

Basically what builds the page is Apache running on a Linux box. It pulls a lot of stuff dynamically out of an SQL database in Postgres. And it also pulls in a lot of bits and pieces -- little snippets of HTML that are built by scripts. An example of that is the Now Playing box. What you see there is a snippet of code that is actually written by the automation system that the DJs are using to choose and play back the music. Which is another Linux server running the same kind of software. The DJ is looking at a private Web page, by which they control what goes out on the stream.

In other words the disc jockey has an internal web page that works as a control console. It's interactive in some way.

Yes. They use that to pick and schedule the music.

For all practical purposes the screen looks like any other radio station automation system. It doesn't even look like a Web page. Most of them don't even know that they're looking at a Web page. I run it in full screen mode. They use this screen to control the server that does the audio playback, which they can program in advance. They can preview what all the segues are going to sound like, right down to the Nth degree. And then they can walk away. You could -- though we never would -- run a completely automated station this way. The capabilities are there. I use it that way for Radio Paradise and for Smooth Jazz (

Smooth Jazz is yours too?

The technology of it. The back end: the automation and the piece that builds the Web site.

So this back end is writing scripts all over the place...

Okay, when the DJ hits the button to start something, or the system starts something on its own because it's been programmed to do that by the DJ, it not only starts the music playback but also writes a new little Now Playing page. That song in the Now Playing slot moves down to the top the next three. A total of four songs are listed: the song now playing and the most recent three. Each has its own page. The system also writes the page that you see when you click on Playlist, which is a history of what's been playing for the last six hours. Those are actually done on a separate machine and then FTP'd over to the Web server and included in the main page.

Does all the music have to be on a hard drive somewhere?

No. Most of it will be, but there are exceptions. At KPIG if the jock wants to play something not in the database, he can type in the title so the system will display it.

And everything you're hearing on KPIG is MP3?

They're high band rate MP3s. Mostly 256Kb. Some are 192. All played out of a Linux box using a $90 sound card. The whole hardware cost for one of these operations is about $600. It's really cheap. Which is why a business can be built around it. But it would be a specialized business: serious next-generation radio.

I've been amazed at how little the public and noncommercial stations use MP3 as an encoding and transmission system on the Web. Most of them use Real and a few use Windows Media Player. Yet that seems like it would be more expensive to me. Is it just because they don't know better?


It's just branding.

It's probably 80 or 90 percent branding. Setting up a free Real server is easy for somebody to do if they're not interested in serving more than 20 people at a time. But there's nothing serious going on there. Okay, say 95 percent branding. And inertia. A lot of people started with Real back when it was the best choice and have stuck with it because they've had no real reason to switch.But man, if I was paying for a server license for a Real server? I'd drop that for MP3 or QuickTime or something else that was free. It's nearly impossible to recoup even the bandwidth costs downstream.

Is there an advantage to Real at any of the different speeds?

Real's latest codecs are much more optimized for low bit rates than MP3 is. They do have a quality advantage at bit rates below about 32k.That's the 5%. But the difference isn't that extreme.

Is your system in tight enough shape for you to productize it?

I would like to get this out there. I don't want to sell it as a product. I want it to be the nucleus of an open source project. I don't want to be the Linus of the system. Rather I want to use the system. That's why I built it in the first place. I'm not a programmer by trade. I've learned enough to build this -- it's a lot of stuff, and I'm happy I know it -- but it's a little too big for me to handle alone. Others need to come in and run with it.

Have you looked into putting this up on SourceForge?

I'm looking at a number of options. Right now I'd like somebody to come in from the outside and check out what we've got here. It's a really big project. It encompasses all these different activities involved in running a broadcast operation. It's not a tool for somebody who wants to do radio as a hobby. It's a tool for people who are serious about doing high quality radio by using 21st century tools.

How many people know what high quality radio is?

In the insustry, hardly anybody. All the people who would be open to doing what I'm talking about here have long since been driven out of the industry. KPIG is the exception. If I weren't hooked up with KPIG I wouldn't have anything to do with commercial radio. I'd be off doing my own thing.

KPIG breaks the mold in lots of ways.

Our staff is huge by medium-market radio standards. We have live DJs twenty-four hours a day. Nobody has that. Even in San Francisco you don't see that. Not any more.

You stream a wide selection, but you're basically about MP3s. Is that because just about everybody with a computer and a love of music has an MP3 player of some kind?

The penetration isn't quite as high as for Real or Windows Media Player. But Real does a fine job of playing MP3 streams, too. In any case, the number of people out there who have Real, or WinAmp or iTunes is pretty high.

We're reviewing Apple's OS X here, which sits on Darwin, a form of BSD. And the default MP3 player shipping with it is iTunes, which comes with a tuner that demonstrates how hard it is to categorize stations. It's also fed by something called the Kerbango database. (Kerbango is the late embedded Linux radio company that was absorbed by 3Com and killed off early this year.)

Apple hired a friend of mine who was a former Kerbango employee to maintain the iTunes database as a part-time project. He has KPIG under "Americana" and Radio Paradise under "Alt/Modern Rock." It used to be under "Americana," but I told him to change it.

Who is going to be doing serious radio, and is there a business in it?

I think there is a nice small-scale business for onen person who has a wide-ranging skill set -- or for a small group of people If it's done right, an online radio station at this point in time could probably support a staff of two or three people working full-time.

What's the revenue model?

At this point the major revenue source for all the successful operations I know about is listener donations.

The public radio model.

Yes. And it's much more applicable to this situation than the commercial radio model, which any number of people have tried to come in and make work online and failed at. As long as there is commercial-free competition, there is no reason anyone will want to listen to something that has advertising in it, unless it was spectacularly better than the alternative. And nobody's coming in and doing things that are spectacularly better.

Seems to me the best online stations, like the best over-the-air stations -- few as there are -- sound like somebody who knows their stuff, playing their record collection for you.

Which is the old underground FM radio system.

Not many people have observed that the fundamental flaw with commercial radio is that its customers and consumers are different populations. That's a disadvantage noncommercial broadcasting do not have. The listeners are the the customers. And on the Net it's easy to put out a pay jar so people can easily compensate the broadcaster for the service. Can't do that with a car radio.

That payment system seems to work remarkably well at Radio Paradise.


Here's my experience. For about a year or so people would occasionally write me and say "Hey, I'd like to pay for this. Do you accept donations?" And I'd always write back and say "No, no. Keep your money." Until one day my wife and I were sitting around trying to figure how the hell we're going to pay for all this. And we said, "Y'know, people have been offering to give us money. Wonder what would happen if we made that possible?" So we put up the link and did a very, very low-key promotion -- maybe once or twice a day on the stream -- and a little blurb in the news section of the Web site, and a little link down at the bottom of the page. In the first month we got about $2000 in donations. Which is about $2000 more than we made the monnth before. Now we've got a target of $3000 -- and we get about $3000 to $3500 every month, plus another $300-$400 in affilliate program stuff, like with CD Now.

Since when?


And that covers you?

That's enough so I can drop off a number of consulting projects I had been doing.

Not bad for being this early in the game.

Yeah. The station is more or less playing for itself.

Will your expenses go up?

We've got a free bandwidth deal going right now. After that ends the expenses will probably double, but I think the revenue will double as well. The audience is growing.

Do you collect data on listeners?

We're starting at KPIG. It's all voluntary for the listeners. The sense that about 80% or better are at work. A.lot of them are home workers with cable or DSL accounts. At work is also when people are looking for some music in the background and they've got a computer sitting there. They're also mostly in the U.S. About 80% or so. Quite a few in Europe. Age range is all over the place, but mostly in the 30s and 40s.

What do you make of Live365, which hosts about 35,000 MP3 streams? It's like this vast mutant thing thatt manages to crash for me on three different platforms.

They require that you use this little pop-up player of theirs. A little javascript thing.

But do they demonstrate a business in brokering individual MP3 streams?

I think there is definitely a market in doing infrastructue for online radio. That's basically what Live365 does. I question their revenue model, which is advertising. They seem to do almost random in-stream ad insertions. Their original model, like everyone else's, was to get rich selling banner ads on their Web site once it had a gazillion visitors. We all know how well that worked.

We all could have saved ourselves a lot of trouble if we read the MUTE buttons on our remote controls.

(Laughter.) Live365 does have new people paying to stream through the service. But they also have thousands and thousands grandfathered in for free. Radio Paradise sneaked in under the wire. We're on there too. We probably grab another 200 listeners at a time that way.

How many simultaneous streams do you run?

It tops out at about 800-900. KPIG does about the same. Maybe slightly higher.

At what bit rate are most people listening?

On Radio Paradise, about 600 out of 800 will be listening at 128k.

Due mostly to cable and DSL, I would guess.

Yes. 128k streams work just great on cable and DSL.

Isn't there a fundamental inefficiency involved as that number goes up?

Yes, it's incredibly inefficient, but I don't think that's going to change. And we're at the point where bandwidth and hardware are so cheap, or headed that way, that there just isn't that much of a return on reinventing the whole infrastructure of the Internet to make it more efficient.

Are you worried about what Disney and RIAA and these other creeps are trying to do with legislation right now?

Yeah. If they get their way, they will drive people like me out of business. There would be no way to cost-effectively do what I want to do here.

It would be endless digital rights management everywhere.

I can't believe that's going to work -- that they are going to shoehorn that thing through. There is just too much standing up against it. The biggest is reluctance on the part of consumers. They've had too much of a taste of how it oughta be, with the free exchange of MP3 files, for them to trade that for anything else.

Yet when Napster died, not that many people cried. They just kind of moved on.

Well, there are any number of alternatives. There's Gnutella, which has a good Linux client. Mac too. True: none are as good as Napster was at its peak.

So you're not worried.

No, I'm really optimistic.