<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
I've updated my OggIndex spec to Skeleton 3.3, which now no longer
includes the checksum for keypoints, and is delta and variable byte
encoded. This reduces the index size to about a quarter of its previous
size, or about 6.2KB per track per hour. I've updated my OggIndex and
OggIndexValid tools to output and verify this new format respectively,
and I will send J^ a patch later on today for ffmpeg2theora. I've
uploaded new skeleton 3.3 index capable Firefox builds to <a
href="http://pearce.org.nz/video/firefox-indexed-seek-linux.tar.bz2"
class="external free"
title="http://pearce.org.nz/video/firefox-indexed-seek-linux.tar.bz2">http://pearce.org.nz/video/</a>
for testing. <br>
<br>
I don't yet have a reliable way for ffmpeg2theora to estimate how much
space to reserve for the (now variable length) indexes. I'm planning on
measuring a bunch of indexes of various files to try to figure out a
reasonable approximation. I am adding command line options to
ffmpeg2theora so that at least you can specify on the command line how
much space to reserve for the index, and ffmpeg2theora will output the
optimal size. When running in two-pass mode this may be less of an
issue...<br>
<br>
I decided not to shift or offset the variable-byte encoded values, as
the index needs to be easily verifiable.<br>
<br>
Chris P.<br>
</body>
</html>