Re: Determining p value for WLSB of RTP SN.

Carl Knutsson wrote:
> You are correct, the interpretation interval for k=5 doesn't cover
> the whole interpretation interval for k=4.

Thank you very much.

> 
> This have been discussed before on the rohc list. Please search the
> mailing list archive for more information about this issue.

I tried to search through the mailing list before posting here, but 
couldn't find any relevant entry. Guess I didn't use the right keyword 
to search.

> 
> Colorful language by the way..
> 

Eh... sorry about that :)

> /Calle
> 
> On Sat, 2 May 2009, Ang Way Chuang wrote:
> 
>> Dear ROHC list subscribers,
>>
>> According to section 5.7 of RFC 3095:
>>     SN: The compressed RTP Sequence Number.
>>
>>         Compressed with W-LSB.  The interpretation intervals, see section
>>         4.5.1, are defined as follows:
>>
>>              p = 1                   if bits(SN) <= 4
>>              p = 2^(bits(SN)-5) - 1  if bits(SN) >  4
>>
>>     where bits(x) is the number of bits in the compressed header.
>>
>> At first, when I read that the change in the value of p depends on
>> bits(SN), I was very perplexed. If the algorithm determines k <= 4 can
>> be used, but due to other factors packet type that is to be used only
>> have bits(SN) > 4, then we may have problem with the following situation:
>>
>> Say, compressor receives reordered RTP, then:
>> vref_c = 32
>> val = 31
>> then k=1 is more than enough to cover val because the interpretation
>> interval is
>> [32 - 1, 32 + (2 - 1) - 1 ] = [31, 32]
>>
>> But if the packet type only have bits(SN) = 5, then we are screwed:
>> [32 - (2^0 - 1), 32 + 2^5 - 1 - (2^0 - 1)] = [32, 63]
>>
>> 31 is not covered!!
>> Thus, the logic for getting k for SN becomes very hairy because it has
>> to take care of this situation. k > 5 are okay because whatever the
>> interpretation interval for these k bits are superset of interpretation
>> interval for k <= 4.
>>
>> But when I look at all packet types for RTP, none of them has bits(SN)
>> == 5. Thus, I assume that whatever calculated k bit of SN are generally
>> safe to be expanded to the available bits(SN). Is my understanding on
>> this matter correct? Do tell me if I get it wrong or miss out something
>> important.
>>
>> Thank you in advance.
>>
>> Regards,
>> Ang Way Chuang
>>
>>
>> _______________________________________________
>> Rohc mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/rohc
>>
> _______________________________________________
> Rohc mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/rohc
>