Codecs & Ogg
------------ Present: James, Ralph, Conrad, Lennert, Erik, Jean-Marc, Mikko, Mike, Bill, Shane, Andre, Tim [re screencast codec] Shane: OggSpots? is a timed image codec. Silvia: It's been defined and Zen's implemented it on Windows. Ralph: There's an MNG profile that you can use for that instead of inventing OggSpots? (that is identical to OggSpots?). Silvia: OK, we'll discuss it later. JM: Is MNG good for e.g. moving the cursor across the screen, for doing something like a screencast? Ralph: If you use the complex profile, yes. If you want lower decoder complexity, you can use the OggSpots? simple profile. Ralph: The code's somewhere in http://svn.xiph.org/experimental/giles/ Ralph: wiki.xiph.org/oggmng Ralph: One problem is that the library was too large to be included in Mozilla. Erik: The library's 460k. Ralph: You can build it much smaller than that if you only want decoding. Tim Ansell: The library wasn't used for thread-safety reasons, not for size reasons. It also didn't work the way that previous GIF animation formats worked. Tim T: It seemed to be more of a political issue rather than a technical one. [re the codec landscape] Mike: I don't think there's any hassle with the actual audio codec landscape -- if you want a free audio codec, there's Vorbis, FLAC, Speex, so lossy, loseless and highly compressed voice audio are all covered. Mike: For video, there's Theora; Dirac's on the horizon (and we haven't released 1.0 of Theora yet!) [re text subtitles] Ralph: There's the ogmtools package that you can use to put subtitles into Ogg files. Tim: I think it supports SRT, and it might support SSA. Ralph: I vote we should just pick one text one, and one graphical overlay one. Andre: I vote for it being extensible so you can add whatever properties to it. Tim: I vote you _do not_ make it extensible so it's really simple! Use the graphical overlay version if you want anything fancy. Mike: There's a couple of minimums: Unicode, CR/LF/space characters. Tim: Anything involving XML and XML-like characters is going to be bloated. Shane: Not necessarily. You can have a basic decoder in very little space. I'd really advocate we had XHTML. Conrad: How about a long-term goal of flexibility in an XML-based format, but for a short-term goal, we just use a very simple text format so we can implement it ASAP. Silvia: If you display CMML on the screen, it already just works. This is the way it already works in VLC and xine. Tim Ansell: Can you annotate it with other types apart from text? Bill: We need to specify a way to indicate that a track is a caption. Ralph: There's a lot of Ogg packetising overhead involved if you are sending a lot of small text strings, since each one in it's own packet. You could put the text for every language into a single packet; mod_annodex can strip it out if needed. [re free DVD spec] Mike: We need to handle DVD menus. Ralph: Does Matroska handle that yet? Silvia: What about using an HTML page? Ralph: What about SMIL? Ralph: DVD menus are intended to some things that are configuration issues. Mike: Is there someone who's actually looking to implement this? I just don't think there's any realistic intent for anyone to implement this. Shane: I agree. We're more about streaming over the 'net, embedding videos in the browser... Mikko: What about formalising a spec so it can be supported in hardware players though? Tim: Well, Vorbis has a spec, Theora has a spec, Dirac has a spec... James: We should jsut make sure that we have enough metadata to present menus to the user if required (e.g. subtitles). [high level Ogg library] Shane: liboggz's about as high-level as you can go if you want to talk about Ogg; liboggplay is higher-level than that. Mike: liboggplay is a bit too high-level for a lot of media players, frameworks, etc. Conrad: liboggz is layered so that those applications can still use it and still have enough low-layer control. Tim: Why don't we rip out libogg from all the other media players etc and put in liboggz instead? Shane: liboggz wasn't around when those applications were written. Shane: xiph.org advertising the presence of liboggz would probably go a long way to fixing that. [lots of dicussion about how you need to know about the codec to properly demux the stream] Mike: One thing we should do is define a standard way of embedding the Xiph codec information in other file formats. [general agreement]