Open Source Video Streaming Needed

Mike McCune writes "This article discusses how streaming video is controlled by three companies: Real, Microsoft and Apple. It discusses how open sourced video streaming software is needed. I looked around and found the start of some OS streaming software. There is a video streamer based on Darwin here and there is the start of several players here." But what about the codec patent problems? I have been told that they're the big holdup. [sigh]

Comments

Codecs are the sticking point

[Anonymous Coward]

While Microsoft and Real use proprietary streaming protocols, the Apple QuickTime software uses non-proprietary documented streaming protocols (RTP, RTSP). So while it is more difficult to create software that will interact with Microsoft and Real servers, it is less so to create one that would interact with a QuickTime server or create a new streaming server based around the same protocol.

The real problem are the video codecs. Most of them are not even owned by the respective companies (e.g. most of the high-compression, high-quality codecs that Apple/MS use are licensed from other companies - Sorenson, Cinepak, Indeo). Note that for some of these it is possible to create freely-distributable binaries (a la Xanim) but it is highly unlikely that these companies will release the algorithms or source code any time in the near future.

What would probably be the best bet is to scour the academic literature on video compression and/or bug people in the field for info and create an alternative codec that is freely distributable. You could then produce a plugin for QuickTime, VFW, etc. to encourage the general adoption of the codec. If a content-based company can run a server and deliver content without technology licensing fees, I'm sure they'd be interested.

Re:Codecs are the sticking point

[phutureboy]

RTP/RTSP is an open IETF standard, and is used by both RealNetworks and Apple's Darwin Streaming Server. Real *used* to use a proprietary protocol called PNM, but no longer.

The codecs are indeed the problem. Most are patented, and licensed only in binary format.

From what I understand about it so far, it seems like MPEG4 is our best bet for an open codec. If I recall correctly:

MPEG1 = Up to VHS quality. File-based format not at all suitable for streaming.

MPEG2 = broadcast-quality video, very compute-intensive algorithms require expensive coding/decoding hardware on either end.

MPEG3 = was abandoned during development, I forget why

MPEG4 = open spec for encoding AV content, a new revision specifically designed for Internet streaming & multimedia is currently under development

More info about MPEG can be found at http://drogo.cselt.stet.it/u fv/leonardo/mpeg/index.htm.

Also, the best introductory resource I've found is at http://www.nyquist-media.co .uk/streaming/streaming.html. It's an excellent overall introduction to the technical standards and mechanisms of streaming video.

On a related note, is there a consensus on the license for Apple's Darwin Streaming Server? Is it open enough?

--

Bandwidth, patents

[Anonymous Coward]

MPEG-2 video looks reasonable at 3-4mbit/s. H263, which is what netmeeting uses, can be tweaked to be TV quality at 1mbit/s. MPEG-4 will probably look OK at less. "Talking head" video, for videoconf, is easier to encode and looks ok at 64kbit/s

There are three things a video codec requires:

• spatial transform, eg idct, which I think is patented. There are many wavelet types, some of which may not be spatial transformed are used to find the most significant information; the rest is chopped.
• Coding scheme: huffman is unpatented, syntax-arith encoding is (but when does it expire?) Both require that lots of analysis be done on symbol frequency. This is tedious, not difficult.
• motion compensatio: in oreder to make use of duplication between frames - again, not hard.

. The open source community should not write off the idea of doing its own video codec. The most difficult part is the transform; if a decent unpatented one can be found, the rest is within the capabilities of many coders. Video codecs are EASIER to write than audio codecs, as the proportion of information that the eye actually percieves is lower. In order to do decent compression in audio, very complex models of the ear are required; this is not true of video. Oh, and the eyes colour resolution is lower than its luminance resolution

The Standard is MPG

[Effugas]

Lots of companies with large amounts of money invested in overly complicated streaming systems will complain, but there's a real bottom line:

The standard video format is MPG, because MPG Just Works. Everywhere.

AVI has failed. The general perception of an AVI file is one that might play, might not, might suddenly install a new codec, might not, who knows. No predictability.

MOV? All the pain of an AVI, with free delays while you deny Apple their cut. From the guys who killed Firewire...

RM. Realmedia ain't bad, but it just doesn't scale up too well. There's this common delusion that only people with broadband links should be able to view high quality video--in this paradigm, RealMedia can do OK, since relatively few people have consistently extreme high bandwidth links to the Net. But, ya know what? This paradigm leaves millions of people unable to view high quality video, except on television.

Presume people can download clips and watch them later, and suddenly the stream-biased, bandwidth-capped format that is RealMedia suddenly looks stale and chunky.

The bottom line, beyond quality issues, is that MPG has won for the same reason MP3 did: It Works. All the various copyright protection systems are obsessed with creating situations where the consumer tries to do something and It Doesn't Work. As I'm sure the consumer trials are showing, when Things Don't Work, consumers simply refuse to buy in. And that's the key--the investors may fund, the studios may create, but it's the consumer that pays for it all.

MPG may not be a low bandwidth streaming format, by any means, but the general obsession of streaming--and streaming only--is short sighted at best, and suicidal at worst. It will be interesting to see how this pans out over the next few months.

See y'all at the DVD trial...

Yours Truly,

Dan Kaminsky
DoxPara Research
http://www.doxpara.com

Quicktime Streaming Server is open

[Mononoke]

Yes, it's Apple's version of "open source", but it is available:

Quicktime streaming source

Can't help ya with on the client side. Maybe Apple doesn't know you want it.

--