25 Years of Internet Radio: Part 1
This year, we're celebrating the 20th anniversary of the Streaming Media website and conferences, the first example of streaming goes back a few years earlier.
At the turn of 1974, Steve Casner produced an early specification for Network Voice Protocol (NVP), and packet voice was being used for early conferencing apps even as IP emerged. ARPANET was turned off in April 1984, marking the formal birth of "the internet." Since NVP was in circulation at that point, it means that the first stream on the internet would have been in April 1984. That should finally settle a few Streaming Media conference bar arguments.
To be fair, up until about 1992, the technology was largely being used for academic conferencing. Still, as you'll see below, some of the first internet radio webcasts happened in 1993, which it has been around almost a quarter century, and we wanted to track down some of the early pioneers.. Now, "internet radio" can be interpreted in a number of ways, so we first need to refine our definition. For our purposes, internet radio is audio (and sometimes video) streaming, over IP, of content that is live, scheduled by the transmitter (not by the receiver), and "produced"—so not just a microphone and a lecture. The programming should be synchronous with the listening (lag/latency allowing). So our definition of internet radio is "a one-to-many scheduled live-linear programmed audio content stream delivered over IP and accessible from the internet."
With that out of the way, let's talk with some of the key players in the rise of internet radio. In part 2 of this article, we'll bring our discussion up to the present day.
Carl Malamud: Guerrilla Radio
In 1993 Carl Malamud—who is now president of Public.Resource.Org, a non-profit dedicated to making government information more accessible—decided to launch a formal "program" on the early multicast IP network that ran across many academic (and some industry) internet sites, called the Mbone (multicast-backbone).
His channel was called internet Talk Radio, and the program was called "Geek of the Week." He gathered a number of technically interested parties who were essentially his peer academics, and starting in March 1993 he programmed interviews of them.
The first really amazing thing is amazing is that they are still available here. The second, even more amazing fact is that many of the folks interviewed are now infamous in the internet engineering community. These are the forefathers of the modern internet, talking about the evolving protocols that they probably, at the time, had little idea were going to become quite so successful in terms of the soon-to-explode internet.
I caught up with Malamud to ask him for a few reminiscences, and just a little insight into how they set the tech up for that very first online audio program. Here's what he had to say, via email, edited lightly.
We did 48 kHz DAT masters and used SoX [Sound eXchange] to turn them into .au files. Guido Van Rossum (who is the creator of Python) was doing SoX in those days, and he wrote the 48 kHz module for me. It took 24 hours to do the compression on a Sparc 2, and the result was a 30MB .au file.
I then pushed the files and a readme to the FTP server at UUNET. I had previously gone around the world to ask/beg others to mirror my stuff on their FTP servers, so the data would spread to Japan, Netherlands, Australia. I also maintained a mailing list and sent out an announcement on each episode.
There were some folks that didn't have FTP, so at least one user used the MCI FTP-mail gateway, which would chop the file up into 200 or more pieces and mail them to him. He'd wait until all the components made it in, and then reassemble the audio file.
We did a number of live things. By the time we got to Al Gore at the National Press Club, we were sending the speeches out live over the MBone [starting in late 1993].
I also got congressional press credentials and ran tie lines (dedicated audio) into the basement of the capitol and had the floor of the House and Senate on the net.
There were also a number of special events:
- A congressional hearing of the Joint Economic Committee that sent out audio live and took questions by email.
- A big "cyber" event at internet (October, 1993?) which was totally live. It included a rock band, live talk shows, and even U.S. Senators calling in from Washington.
- The 50th anniversary of the UN. We ran a T1 line into the hall in San Francisco, sent it out live.
Most of that stuff is in our archives. There is a Washington Poststory by John Schwartz (now at the New York Times) about the congress being live. We mostly did VAT/VIC, etc. but I think we used CU-SeeMe [videoconferencing protocol] for the [National Science Foundation's] Global Schoolhouse [project].
One other thing we had starting in January 1994 was a rock and roll station. At the time, licenses were indeterminate; we filed for a BMI license as a public radio station on the internet. They had never seen that, so we had our license for about a year.
Simon Hackett (the guy who created the internet toaster) came in and programmed a Denon 120-CD jukebox with some Tcl [programming language] code and we set it up with playlists and autorotate so you saw the name of the song playing on your VAT and SD screens. Was quite cool!
This was our internal network: go2sm.com/internetmulticastingservice.
For distribution to the net, that wasn't a huge deal for us because we had a 10Mbps line to [the] MAE-East [Internet Exchange Point]. It was the fastest line in DC at the time, and we were the only end user to be on MAE-East, the first big colocation site (which was set up by Mitch Kapor, Rick Adams, and a few others). The 10Mbps line got us to the edge of Mbone. and then we let multicast take it from there.
Remember, in those days, MBone was a limited net … links were hardcoded at a certain bandwidth. So, you couldn't overload the net itself, you could simply fill up the Mbone. (I used to get grief from folks when I put Congress on the Mbone because if I put my legislature online, then everybody would put their legislature online, and there would be no more bandwidth left for Internet Engineering Taskforce meetings).
One other thing about our 10Mbps line … when [Bill] Clinton took office, he wanted to do a big "I know about the internet" demo, but they couldn't get their routers cleared by secret service quick enough. So, I got a call from [the Defense Advanced Research Projects Agency] asking me if I could see the White House from my office. I couldn't see it from my office, but we could from the roof of the National Press Building. So Motorola came out and ran an infrared link down to the White House lawn, I threw an Ethernet cable over the roof and down into the National Press Club and we patched it into our internal network. Clinton was able to come out on the White House lawn, show a video coming in over the net, and everybody was happy. So, I get credit for first real link into the White House.
You could make the argument that Malamud is the first podcaster, since his early shows were delivered some time after they were recorded and listened to on demand. He also carried out some of the most important early live streams, using the Mbone, and produced audio content that would absolutely be considered internet radio programs. However while looking at the "scheduled live-linear programmed content stream" from our internet radio definition, I couldn't be sure if Malamud's Gore or Clinton streams were actually the first. The dates of those specific live transmissions are unclear, but It seems all this activity was at the end of 1993 and early 1994.
But then I tracked down something that happened earlier.
Jon Crowcroft: MBone and the Rolling Stones
Picking up on Malamud's mention of VAT/VIC and the MBone, I reached out to an old contact of mine, Jon Crowcroft, currently Marconi Professor of Communications Systems in the Computer Laboratory of the University of Cambridge and the chair of the programme committee at the Alan Turing Institute. I first met Crowcroft at UCL about 15 or 20 years ago, when I became aware of a project he was close too called MICE, which was a group focused on video and audio streaming tools for use on the Mbone that was running between many of the early academic and commercial internet-connected institutions. It was a perfect environment for those pioneering live linear streams, and I asked him if he could recall any very early transmissions on the MBone that were programmed (rather than conferences). Again, this was an email interview that has been lightly edited.
Certainly the first audio/video on the net was in the very early 1980s using a thing called the voice funnel and some specialised hardware connected to imps I think, but it wasn't scheduled content. Some folks at BBN staged a string trio playing in three far-apart U.S. Locations, only to discover that you can't follow someone else's tempo changes when the round trip time is more than a few 10s of milliseconds. Later folks did dance music and just ran a click track. [My] next memory is about 1986 when Van Jacobson first shipped vat, and a bit later vic (and Henning Schulzrinne [now co-chair of the Internet Technical Committee of the IEEE Communications Society] had his own IP multicast-based video/audio tool and Thierry Turletti at INRIA [the French Institute for Research in Computer Science and Automation] did ivs i think, which had an H.261 video codec he wrote). But [there was] nothing live until we set up the Mbone (the first tunnel was from ucl to berkeley in about 1987) with Steve Deering's code. I think we had BBC content on that for a while by just attaching an analog radio receiver to some PC at UCL, but can't recall when (we claimed it was ok for copyright be cause universities in the UK at the time streamed BBC content on private analogue radios round halls of residence anyhow).
There was a big event (I can't put a date on it), when a Rolling Stones concert was streamed over the Mbone, and Sky News came and filmed us watching it at UCL. At the same time, an awful in-house band at Xerox PARC (called Severe Tire Damage) tried to hijack things by streaming their gig too on the same multicast session. Giiven how the audio/video tools worked, we "tuned them out!"
I can't give priority date to the Global Schoolhouse's use of CU-SeeMe versus the MBone; as far as I know, CU-SeeMe as a tool predates the MBone tools, but had very limited use (it was Apple Mac only, and few sites had the bandwidth), and the big Global Schoolhouse lashup involved the kids all coming to university host sites to get fast enough links .We hosted one group [that] was very cool; the kids were all doing ecology projects and did demos/talks about their work to kids 3000 miles away.
It was all very cool, and I think we were all happy at all the different initiatives and cooperated rather than competed!
The Rolling Stones concert Crowcroft refers to was in November 1994; you can read more about it at go2sm.com/stones.
Some of the material that Crowcroft mentioned could only be found on the Internet Archive, and while I was searching quite deeply, I noted an early screen shot of the Mbone Session Directory tool, which listed something called Radio Free VAT.
Dave Hayes: Radio Free VAT
The geeks among you will love this: I then found an old multicast Session Description Protocol (.sdp) file (the MBone/multicast metadata) for Radio Free VAT, and, on opening that in my text editor in there I found the originator's email from back in the early 90s. Tentatively, I emailed, and amazingly it worked! I was "OMGing' and jumping around like a teenage girl getting a Snapchat from Justin Beiber when Dave Hayes replied!
Here's what Hayes, the man behind Radio Free VAT, said about the stream in an email that has been lightly edited.
It was originally an accident, and it happened in 1991 or 1992 (I can't quite remember).
I used to be the key internet techie for the NASA Jet Propulsion Laboratory (JPL). I was computer security, news admin, mail admin, the system administrator's last line of defense, etc. etc. so I was pretty busy. I also was (and still am) a musician and a DJ.
At the time the Mbone was in its infancy, using multicast IP for voice and video (and I still maintain IP multicast is the proper way to do streaming...if only ISPs would agree). I was assigned the task of paying attention to the Mbone, getting it to work, and assuring JPL/NASA that they could have arbitrary video/audio conferences whenever they wanted, which (at the time) was pretty nifty.
Back when online video was the size of a postage stamp, audio companies were changing the way the world heard radio. Here are reminiscences from RealNetworks and other pioneers.
Building on previous DIY articles, this installment will walk you through capturing audio, encoding and packaging with FFmpeg, then playing out through the Icecast server you set up in the last article
In the latest installment of our do-it-yourself series, we'll look at how to set up an Icecast server on an AWS cloud instance.
For the second installment in our do-it-yourself series, we look at how to set up a very basic audio stream between two computers on the same LAN. While the coding is simple, the concepts underneath are crucial to an understanding of how streaming works.