It uses a proprietary stream protocol called RTMP.

http://developer.apple.com/library/mac/documentation/Java/Reference/JavaSE6_AppleExtensionsRef/api/com/apple/concurrent/Dispatch.html

The MPEG-2 patent license typically costs *more* than the H.264 license for most situations, so I doubt you could be in any more trouble than you already are. Of course, I haven't studied it in detail, and nothing will ever happen to you specifically no matter what you do anyway.

Matroska is free and "okay". It has some technical annoyances (the authors liked a few codecs so much they made their own special complicated setups for them, the EBML parser is a lot of code) but does have generic codec wrappers and better seeking. It also has commercial support and several extremely practical tools.
MOV is the best one I know of and is generic (MP4 might not be, the spec 14496-14 isn't public so I'm not sure). The only thing I don't like about it is that the index comes at the end and is required to play the file, so you either have to write the whole thing out twice or you can't play partial files.
Something like this will probably be added to the article soon, I think he's researching it better than I just did.

> MPlayer would stop playing altogether if the CPU couldn't keep up.
This is really a problem with the GUI defaults. Assuming you're using MPlayer OS X Extended, set Preferences -> Video -> Framedrop "Soft" and "Use multithreaded ffmpeg build", and it will play faster than any other OS X player. (excluding CorePlayer of course)

x264 defaults to fast mode; rerunning it with more analysis on that file (basically -A all -m 9 should cover most of it) gains another 2db. Also, that test, like some other Theora tests, is on a 176x144 video. Exactly how often do you watch those?

SIMD actually helps a lot with motion compensation; it's just a bunch of parallel multiply-adds, after all. But Theora has bitstream reading problems too; you have to decode the whole frame's bitstream before displaying any of it. And there is absolutely no question of it supporting HD (sometimes even SD) because the MV length is broken.

Youtube uses completely stock mencoder from 2005 (+ a handwritten h264 encoder by Skal for fmt=18, which probably isn't faster than x264). They just have way more CPUs than you think they do. I'm actually not sure a hardware Theora decoder could be really efficient - the memory requirements are quite bad for a codec with no multiple reference frames.

Man, I suck at formatting.

> Theora just scores higher on a scoring algorithm when compared ot a single h264 encoder, the open-source x264. It doesn't even do that; it only scored higher when using Xiph's PSNR tool, because it respected a buggy colorspace header written by ffmpeg that didn't match the video. x264 won rather heavily when that was fixed, but /. never retracted the story.

ffmpeg has no idea what a broadcast flag is; your file is probably encrypted, which it won't deal with.

FFmpeg can open images too, although I don't remember the supported formats. It's also much faster at it.

Of course, it is. And sometimes people get mad about that, and then they go off and invent gstreamer and its 1000 plugins wrapping different libs of various quality, all of which break at different places. I don't think this is better.

When was the last snapshot that didn't build or work?

NSA outsources data recovery to Kroll Ontrack, who cannot recover data from a zeroed disk anymore than anyone else can. And why would you bother physically destroying a disk platter? This isn't homeopathy or anything, you can actually get rid of magnetism.