Decompressor NACK feedback

Asked by Arne on 2012-02-13

Hi,

first of all thanks for the library. I try to run some tests with RoHC and would like to modify the k and n values for the NACK feedback from the decompressor (section 5.3.2).
For my understanding you currently create a ratio of the decompression failure (okval, curval, maxval) but independent from the state of the decompressor, right?

Is it possible to convert the 'k_1 out of n_1' values to your ratio calculation?

Thanks and regards,
Arne

Question information

Language:
English Edit question
Status:
Answered
For:
rohc Edit question
Assignee:
Didier Barvaux Edit question
Last query:
2012-02-13
Last reply:
2012-02-13

Arne,

> first of all thanks for the library.

You're welcome :)

> I try to run some tests with RoHC and would like to modify the k and n values for the
> NACK feedback from the decompressor (section 5.3.2).
> For my understanding you currently create a ratio of the decompression failure (okval,
> curval, maxval) but independent from the state of the decompressor, right?

You're right. I gave an explanation of the algorithm in response to question #181462:
https://answers.launchpad.net/rohc/+question/181462

I also opened a bug to better describe the algorithm. See bug #902465.

> Is it possible to convert the 'k_1 out of n_1' values to your ratio calculation?

I think it is not possible with the current code. It does not conform to RFC.
I added a note in bug #902465 to mention that the current algo is wrong.

The nearest solution with the current algorithm is to set okval=errval=1 and
maxval=k_1. It won't give you "k_1 out of n_1", but an ratio of sucesses/failures
since the beginning of the stream.

You can however change the code to better match RFC 3095. It should be quite
straight (I may provide some support if you need). Contributions for better
conformance with RFCs are welcome!

Regards,
Didier

Can you help with this problem?

Provide an answer of your own, or ask Arne for more information if necessary.

To post a message you must log in.