[icecast] VBR reencoding @128k problem

Brendan Cully brendan at icecast.org
Fri Sep 7 12:16:02 UTC 2001

On Friday, 07 September 2001 at 12:30, Nough Krapough wrote:
> I did some further investigating into this issue and here's what I found:
> It appears somewhere between ices0.1.0 and ices 0.2.0 is where the "bug" (for lack of a better term) was introduced that i'm noticing.  When I compile ices0.2.0 w/lame 3.89beta, the VBR is *NOT* reencoded, but instead streamed out at full VBR bitrate(s).  However, when using ices0.1.0 w/lame 3.89beta, it properly streams the VBR mp3 @ 128k.  
> I've tried various combinations of ices & lame, but pretty much narrowed it down to changes between ices 0.1.0 and 0.2.0.  I wasn't privy to any ices CVS sources between those two versions, otherwise I would have tried those to further narrow down the changes.
> Jack, Brendan, or others that may know, would that change be intended for some reason, or am I the only person seeing it?  Can anyone else reproduce my findings?  I kinda feel like i'm smockin crack here, or mebbe the only person trying to reencode VBRs? *shrug*

Yeah, I think I know what's happening. Ices identifies the bitrate of
the current song from the FIRST frame. It wasn't designed for VBR, so
that worked ok. The problem is some code I added after 0.1 that tries to
be clever and not reencode a track if that track is already at the
destination bitrate (my personal server is and AMD 450, so I had an
interest in reencoding as little as possible, especially at high
bitrates :) Also I figured it saved a little fidelity).

I would bet that the first frame of these VBRs is 128kbit, so ices is
thinking it doesn't have to reencode.

Right now I'm torn between just ripping out that conditional reencode
stuff (since it makes the code path much uglier and the advantages are
dubious) or making it a config option. Unfortunately I've been pretty
insanely busy in RL these days, so I haven't had a chance to fix it. And
then mutt is in beta for 1.4, which tends to swallow my copious free
time. (By the way jack, I think your mutt bug has been fixed in CVS :)


--- >8 ----
List archives:  http://www.xiph.org/archives/
icecast project homepage: http://www.icecast.org/
To unsubscribe from this list, send a message to 'icecast-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.

More information about the Icecast mailing list