Hi everybody:
Here is this week's Ogg Traffic with all new updates on recent
activities in the world of Xiph.org. The HTML version is up at
http://www.vorbis.com/ot/20030415.html.
Enjoy!
-Carsten
<p>Ogg Traffic for Tuesday, April 15, 2003
[1]Carsten "Purple" Haese
April 15, 2003
_________________________________________________________________
Table of Contents
1. Status Updates
1.1. Monty
1.2. Michael Smith
1.3. Karl Heyes
1.4. Oddsock
1.5. Brendan Cully
1.6. Ralph Giles
2. Interesting Discussions
2.1. Bitrate Peeling
3. Recent Developments
3.1. Xiph.org Wiki
3.2. French Radio offers Ogg Vorbis streams
3.3. More Tremor Improvements
[2]Previous Issues of Ogg Traffic
1. Status Updates
1.1. Monty
Monty has worked a lot on drastically reducing Tremor's memory usage.
It doesn't take much to guess that this work is directly linked to his
work on [3]Neurosetta. For all the technical goodies, read more under
Recent Developments.
1.2. Michael Smith
Mike made improvements and fixed various bugs all over the map:
* In libshout, he added code to calculate the ogg granulepos
directly, which provides accurate streaming in situations where
there are granulepos gaps.
* In icecast, he fixed a typo in the socket handling code, and he
repaired the util_url_escape() function, which Oddsock pointed out
to be chronically broken.
* In ogginfo, he made improvements to ogginfo to make it more robust
against files with minor errors in the stream.
1.3. Karl Heyes
Karl fixed a typo in libshout's connection handling code, and he
applied a fix to ices to correctly handle when PCM based metadata
updates occur and resampling is in use.
1.4. Oddsock
Oddsock sent in lots of icecast fixes and improvements:
* Fixed alot of yp logic. Timeouts now work properly so the
tolerance of the unavailability of yp servers is much much better
now.
* New icecast config option yp-url-timeout to specify the timeout.
* Url encoding is now fixed so that the yp data is formatted much
nicer and is correct.
* Added url encoding for some fields that were not url-encoded.
* modified util_dict_urlencode() to not url-encode the key (still
does the value)
* New curl option CURLOPT_NOSIGNAL which prevents curl from using
signals when timeouts are hit. This new option needs curl 7.10 at
least.
1.5. Brendan Cully
Brendan applied some libshout code cleanups.
1.6. Ralph Giles
Ralph applied a patch from Rodolphe Ortalo to allow building Theora in
a separate directory.
2. Interesting Discussions
2.1. Bitrate Peeling
Segher's experimental peeler (which was an April Fool's joke, by the
way) has rekindled interest in a working bitrate peeler, as evidenced
by [4]this thread. The bottom line is that Version 1.0 files don't
peel very well due to how the packets are organized.
Segher has plans to write a tool that will rearrange (losslessly, i.e.
without additional re-quantization loss) a Vorbis file into a form
that peels better. He is convinced that this can be done, but that
takes time, which is something he doesn't have in excess. Of course,
if somebody offered him a contract to implement bitrate peeling for
money, the process could be accelerated a lot. Oh, and did I mention
that [5]donations to Xiph.org are now tax-deductible?
3. Recent Developments
3.1. Xiph.org Wiki
In order to provide a central hub for information about Xiph.org's
past, present, and future, we have installed a Wiki at
[6]http://wiki.xiph.org/. Due to its decentralized and collaborative
nature, this Wiki will be an excellent source of up-to-date
information about all Xiph.org projects. Check it out!
3.2. French Radio offers Ogg Vorbis streams
[7]An alert reader points out that [8]Radio France is now offering
Ogg streams of their radio programs. This will certainly help increase
public awareness of Ogg Vorbis, and that is definitely a Good Thing!
3.3. More Tremor Improvements
After completing the zerocopy rewrite of the Ogg memory management
code, Monty opened a new branch of Tremor with the promising name
lowmem-branch. While current Tremor is suitable for devices with
plenty of memory (for portable standards) and a relatively slow CPU,
this new branch aims at devices with little memory and more CPU power.
The goal is to come as close to the theoretical limits of heap and
stack usage as possible without incurring too much of a complexity
penalty.
Monty started with the code that handles codebook decoding, and
proceeded with floor decode and residue decode, leaving no stone
unturned. The code for codebook decode is now mature enough for
heavy-duty assembly optimizations, according to [9]this message. The
memory requirements for codebook decode have gone down from about 70kB
to a mere 9kB for Vorbis 1.0 files and from 300kB to 40kB for early
betas (which used considerably larger codebooks than the 1.0 encoder.)
Comparisons for floor and residue decode are not available yet, since
these parts are still under construction, but so far, things are
looking really promising.
References
1. mailto:carsten@xiph.org
2. http://www.vorbis.com/ot/
3. http://www.neurosaudio.com/
4. http://www.xiph.org/archives/vorbis/200304/0042.html
5. http://www.xiph.org/ogg/vorbis/donate.html
6. http://wiki.xiph.org/
7. http://www.xiph.org/archives/icecast/4724.html
8. http://www.radiofrance.fr/
9. http://www.xiph.org/archives/tremor/200304/0007.html
--- >8 ----
List archives: http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to
'vorbis-request@xiph.org'
containing only the word 'unsubscribe' in the body. No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.