[Asterisk-Users] Digium and mailing lists

[Steve Underwood]

Benjamin on Asterisk Mailing Lists wrote:

>[...]
>In respect of the G.729 codec and encumbrance by patents, questionable
>or not, I really don't understand what the problem is. If you don't
>like the fact that, for all practical purposes, there is a tax on
>G.729, then DON'T USE IT.
>
>To me it would seem that finding an *alternative* to G.729 is the real
>issue here, not how realistic it is to shoot down patents. In my
>opinion it is far easier and much cheaper to replace G.729 altogether.
>
>ILBC is a very good alternative to G.729. Talk to your suppliers and
>demand that they support ILBC. We have been doing just that and phone
>manufacturers are telling us that they will support ILBC as soon as
>their DSP circuitry suppliers support ILBC. DSP chipsets with ILBC are
>becoming available now, ie Texas Instruments TNETV901 or Centrality
>Communications PA168.
>  
>
Its interesting that a number of people have complained about this 
discussion. Obstacles are the most critical issue in the development of 
free telephony systems. Every step I take I am thinking about the IP, 
approvals, and other obstacles that will be in my path. Its a pain, but 
ignoring, or even sidesteping, these things carries huge risk. Nobody 
can run a G.729 codec and not have a physical device.Even the most 
trivial IP court cases would take a lot more resources than I have. 
Approvals are also a big headache. We have SS7 running in Europe now. To 
get it approved would probably cost upwards of $100,000. That's quite a 
hurdle, and makes GPLing it pretty unlikely.

The ETSI database doesn't cover G.729, but it covers the GSM stuff more 
thoroughly than the ITU database covers things like G.729. Look in 
there, and see the minefield that has been created for codec development 
*IN THE EU* as well as elsewhere. You can't patent G.729. You patent 
techniques within G.729. An enormous number of techniques were patented 
in the early 90's when the wireless feeding frenzy was in full swing. 
Some are questionable. Some are as solid as a rock. Trying to create a 
replacement for G.729 is hard. Jean-Marc Valin's Speex is the result of 
a lot of working around the patent obstacles. He is Canadian, so US 
patents are not his greatest concern. He is, however, at Sherbrook U, 
which holds many of the juiciest patents on voice compression methods. 
Speex is good, but it currently consumes more MIPs than G.729. With 
refinement that might be fixable, but it isn't easy when most of the 
good tricks have been taken. A codec requires man years of trialing with 
people, to both refine and evaluate it. Serious users also want to know 
the MOS of a codec. This is something based on human opinion, and 
requires a lot of humans over a long period. Cost estimates for doing 
MOS testing are usually high.

iLBC isn't really head on with G.729. Its bit rate is twice that of 
G.729. It is trying to solve a somewhat different problem, with varying 
success. Speex is currently the best chance to bypass the grip G.729 has 
in the market. The resources aren't really there to drive it quickly, 
though.

Regards,
Steve