[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [PWE3] RE: [AVT] Re: Header Compression over MPLS & Allison'sDISCUSS Commenton PWE3 IANA Allocation Draft

> -----Original Message-----
> From: Allison Mankin [mailto:mankin at psg.com]
> Sent: Monday, October 24, 2005 1:26 AM
> To: Yaakov Stein
> Cc: Stewart Bryant; Ash, Gerald R (Jerry), ALABS; pwe3; mankin at psg.com
> Subject: Re: [PWE3] RE: [AVT] Re: Header Compression over MPLS &
> Allison'sDISCUSS Commenton PWE3 IANA Allocation Draft 
> 

> 
> May I ask:  why is the troublesome behavior of
> CRTP acceptable and the performance of eCRTP not sought?
> The loss effects mean resetting the compressor at least
> and poor compression, and possibly other defects.
> 

A couple of things.  First, if you're using cRTP for VoIP, then things
are probably engineered so that there is negligible packet loss due to
congestion.  So you're left with packet loss due to corruption/noise.
And in the wireless environments that are most susceptible to packet
corruption, I suspect that they may be using RoHC, or possibly other
schemes (I am not working in that environment, so I can't say for sure).

In the evironments in which I've had to do voice quality
troubleshooting, while I've sometimes suspected that the cRTP packet
loss multiplication effect had a significant impact on voice quality,
I've never been able to show this conclusively. It's even sometimes
difficult to discover if burst VoIP loss is occuring: it does not show
up in the CDR records of most equipment that I've worked with.  The one
place I've run into in which intermittent packet loss seems to give us
trouble is on terrestrial microwave links, and they are subject to burst
corruption loss anyway, so it's difficult to determine the extent to
which cRTP loss multiplication may be exacerbating the problem.

> 
>  "cRTP is what is presently implemented"
> 
>  "there is a lot of reluctance to go over to ECRTP""

I have not seen reluctance, just lack of demand.  And from what we've
seen so far, even if ECRTP were widely available, and the implementation
was essentially free, we might only use it in very specific environments
unless there was little or no increase in CPU load vice cRTP.
Particularly if we have to do compression/decompression hop-by-hop.

> 
> Allison
> 
> 

Thanks,
Jim Hand
jameshand at att.com
732-420-3017

_______________________________________________
pwe3 mailing list
pwe3 at ietf.org
https://www1.ietf.org/mailman/listinfo/pwe3