Getting Real

RealNetworks makes some dramatic open source moves. How real are they?

Yesterday morning Real Networks announced Helix, an "initiative" the company calls the "first comprehensive open standard for digital media delivery". It consists of three "elements":

1. Helix DNA Platform, the "the first open, comprehensive platform of digital media products and applications for any format, operating system or device". It is "composed of the software engines that have powered RealNetworks' popular and proven server and player products for several years, as well as a set of more than 1,000 application programming interfaces for building media applications."
2. Helix Community, a "collaborative effort between RealNetworks, independent developers, and leading companies to create and extend" the Helix DNA Platform".
3. Helix Universal Server, a breakthrough product from RealNetworks that for the first time streams all major media types."

The Helix Universal Server and Helix Universal Gateway were both released yesterday. Together they will stream and cache "all major media types," including RealAudio/Real/Video, QuickTime, MPEG-2, MPEG-4 and Windows Media. (No mention of Ogg Vorbis.) The idea is to allow a media source to stream any format to any client, on demand. The server will run on eleven different operating systems, so far, including Linux 2.4.28 with glibc 2.2.4. RealNetworks CEO Rob Glaser claimed that the server, running on Linux, delivered 4x the performance of Windows Media Server 8 while delivering Windows Media and Real files simultaneously. The Windows Media codec was apparently reverse-engineered, although Glaser used more euphemistic language to describe it.

When I asked Eric S. Raymond, who was at the announcement, about the Real codecs, he confirmed their continued closed-source status. "One fight at a time," he added, optimistically.

The Helix Community, which has its own .org site, houses the means by which developers gain access to the source code of Helix DNA Platform software engines. These include source code, object code and APIs for the Helix DNA Client, the Helix DNA Server, and the Helix DNA Encoder. The company's codecs are still only through a commercial license, according the Encoder page.

RealNetworks is working with CollabNet on this whole process.

At this point there are two licenses:

• RealNetworks Community Source License (RCSL), which is "structured to ensure that all products built under the RCSL remain compatible with the Helix interfaces". It's a basic commercial license.
• RealNetworks Public Source License (RPSL), which "is structured to provide developers greater flexibility in their use of the source code". This one is described as "similar" to the GPL in respect to certain copyleft provisions, but different in respect to "patent issues" and restrictions about contributing changes back to the community."

Rob Glaser emphasized RealNetworks interest in soliciting community input both licenses before they submit the RPSL to the Open Source Initiative for certification.

Client source code should be available within 90 days, and server source code in "subsequent months."

The company also says it will license several patents and pending patent applications to the Helix community. The company adds, "Participants under the RPSL will enjoy royalty-free patent licenses, and participants under the RCSL will gain patent rights under the royalty schedule to the RCSL. In addition, RealNetworks will be making available under separate licenses its extensive portfolio of patents and applications that it believes are required for best of class streaming media applications."

The reaction so far has been positive, but guarded. John Markoff in the New York Times calls the announcement "a significant challenge to Microsoft." Dan Gillmor in the San Jose Mercury News calls it "intriguing but incomplete." Andrew Orlowski in The Register says, "Under pressure from Microsoft, and completely open formats, (Real Networks have) decided to meet the open source community halfway".

Both Gillmor and Orlowski obtained at least some of their caution from Bruce Perens' published statement about the announcement. He says, "While RealNetworks is making a
significant contribution to Open Source, today's release does not include the "crown jewels" -- their "codecs", the encoding and decoding software for their proprietary RealAudio and RealVideo formats." He adds this about licensing:

RealNetworks "client engine", the thing that lives in the desktop or the web browser and drives the client half of their codec, will be available under a license that is derived from the Apple Public Source License, but with goals much closer to the GNU General Public License. The license text includes a patent grant. Like the APSL and the NPL licenses, it grants RealNetworks a right to relicense your code under any license of their choice. So it is unlike the GPL in that it gives one party more rights than all others. This license has yet to be approved by the Open Source Initiative board, or accepted by the Free Software Foundation, or even fully reviewed by yours truly. It may have to be modified before it is worthy of acceptance by the community.

He goes on to pick apart what will and won't be released, and under what conditions, as maters stand right now. A short list:

• "On the list of (client) functions that most likely won't be included, besides the codecs, there's a lot of utility and user-interface code.
• "...we're getting some network protocols that go on top of IP and UDP, and do their best to provide continuous playback despite the fact that the Internet doesn't guarantee throughput or latency. On top of that are file formats and data objects, and other pieces necessary to make an Open Source player for some already-open file formats.
• "RealNetworks server and "encoder engine", without the actual codecs, will be under a "community source" license. This means that source code will be disclosed to people who sign an agreement, and those people will get a lot less than the full set of rights that come with Open Source licensing. Since other streaming servers and encoders are already fully Open Source, we can't expect the Open Source community to have much to do with this part of RealNetworks code."
• "The RealAudio and RealVideo codecs will be available in compiled form, as proprietary software that can be linked into a larger product. Again, no joy in the Free Software camp. However, these codecs will be available for use along with various Open Source pieces that Real is releasing, and thus it will be easier to for third parties to produce a half-proprietary Real-format player under Linux and on other operating systems where one is not supported today."

He also adds,

"Perhaps the greatest beneficiary of RealNetworks contribution could be the Ogg Vorbis audio format. Ogg is a fully Open Source codec, unencumbered by patents or royalty payment requirements, which offers audio quality comparable to, or better then, its proprietary competition. The Ogg encoder and servers, not just the client, are available as Open Source. The addition of RealNetworks network protocols and other utilities might make Ogg even better, and might facilitate the inclusion of Ogg as an option in RealNetworks proprietary products.

It's not clear whether the exclusion of Ogg from the published list of supported formats was an oversight due to Ogg's still-miniscule market share, or to whatever threat Ogg might present to the Real codecs. I'll be asking Rob Glaser about that when I meet with him tomorrow at the O'Reilly Open Source Convention in San Diego.

It's also not clear why there hasn't been more attention paid to the open source work Apple has done with QuickTime, which has had an open source server for years (it just hit version 4.1, with full MPEG-4 Support). The Quicktime 5 API reference is a whopping 3118 pages documenting fuction, struct, data type, etc., all cross-linked. And QuickTime is pretty far along in editing as well. Still, like RealNetworks, Apple is keeping its codecs proprietary. And as long as that's the case, those codecsp never become the makings of true Internet infrastructure (as MP3 has, in spite of its patent encumberances, and as Ogg Vorbis may well become).

For now I want to give RealNetworks the full benefit of doubt here, for two serious reasons:

1. They've come a long way in a short time toward embracing open source, both as a value system and as a competitive strategy -- and they seem very open to community input. Their position does not give the impression of being highly fixed, while their movement is clearly in a positive and inclusive direction.
2. As we've seen before -- with IBM, Hewlett-Packard, Apple and other large companies -- these high level strategies are often expressions of grass roots movements inside the corporate firewall. I have heard over the years from a number of people inside RealNetworks, and they've all reported a strong sense of developer affliation with the free software, open source and Linux movements. They've also told me that there has been quite a bit of Linux-based development at RealNetworks, in spite of the company's public focus on the Windows platform. In fact, I'm quite sure this new announcement has as much, if not more, to do with RealNetworks own developers than it does with its top management's competitive moves against Microsoft and other players in the marketplace.

I'll find out more tomorrow when I talk with Rob Glaser. Meanwhile, I welcome your input as well.