[VoiceOps] SIP trunk providers emitting RTP packets containing wrong number of samples

[Tim Bray]

On 13/04/2022 21:25, Liudvikas Bukys wrote:
> My company (a SIP equipment manufacturer) has noticed a sudden flurry 
> of problems related to RTP streams containing some packets with the 
> wrong number of samples (different from what was negotiated). Our 
> equipment negotiates 20ms packets, we're seeing the emission of a few 
> 5ms packets with MARK bit set, before the 20ms packets start flowing.  
>   I theorized that some vendor's SBC or Media Gateway has released 
> software that causes this.
>
> It has erupted at multiple customers.  I thought there was no 
> commonality of carrier or trunk provider, but today I heard a report 
> that T-Mobile is definitely implicated and reproducible.  Other 
> sources are possible too.
>
> Has anyone else noticed this?
>
> Does anyone have a good contact for SIP-related matters at T-Mobile?

I've seen something similar to this before, this year.  I can't remember 
exactly when or what.  (I'm usually looking at  PCAPs over somebody's 
shoulder when they need a second opinion.)


It isn't usually a problem, I don't think.   Worst that happens is the 
first few RTP packets are dropped at the receiver.  I've always presumed 
just a bit of a race condition between the SIP part and the RTP part.    
5ms unusual.

-- 
Tim Bray
Huddersfield, GB
tim at kooky.org