rohc team mailing list archive

[Didier Barvaux <didier@xxxxxxxxxxx>]

Faraz,

> I noticed that ROHC behaviour is different for different traffic
> pattern. For example
> a/ it was more robust for a periodic traffic pattern. On the other
>    hand, less robust for burst (VOIP) pattern.

I agree with you for the robustness of traffic without important
changes. However, I'm curious to known why do you say that VoIP is not
in this category?


> b/ Mode is O-mode, with out periodic update but compressor send IR
>    header for a queue time out (beyond 30 sec) at compressor end with
>    out recovery request from de-compressor.

The library does not have any timer. There is no mechanism for queue
timeout either. This is a task for the applications using the library,
not the library itself.

With what program do you use the ROHC library? Your problem might come
from that program.


> c/ Notice feedback recovery message at a latency of more than 30 sec
>    for decompression failure.

A new feedback packet is not emitted by the decompressor for every
decompression problem. They are rate-limited.

The rate-limiting algorithm may add latency to the recovery-on-error
mechanism. It is a compromise between quick recovery-on-error and
over-reacting on sporadic problems (= too many transmitted feedbacks +
too many transitions to lower compression states).


> e/ losing N packets in a row cause more damage.

OK. At least one thing seems to work as expected ;-)

Regards,
Didier

Attachment: signature.asc
Description: PGP signature