Rillian was right, the match byte function was not only hard to read, but also flawed. ;P<div><br></div><div>The concept of simply splitting on cluster boundaries is as sound as it was last week, but it is none the less, </div>
<div>highly coupled to actually having your code split on cluster boundaries more than just most of the time.</div><div>I've reworked my patch, and I have been able to stream for over 30 hours continuously with gstreamer, and also with krad link.</div>
<div><br></div><div>In the last hour, I cleaned up the white-space and style issues, hopefully without accidentally changing how the code works.</div><div>I'm not going to make any claims, other than this is version 2, until further testing is done.</div>
<div><br></div><div>Our intention is to move to full parsing, in order to enable a rich metadata experience. Currently essentially no-metadata is even possible,</div><div>as including icy- http headers is likely to confuse client software. Tags are a top level element in WebM and are arbitrary as Vorbis tags are,</div>
<div>with a few conventions there are lots of possibilities here.</div><div><br></div><div>Cheers,</div><div><br></div><div>David</div><div><br></div>