On Jan 23, 2014, at 7:15 AM, Bob Briscoe <[email protected]> wrote:

> Fred, Gorry,
> 
> In case you've tuned out of the thread below, it's worth reading the (very 
> short) section on ECN in 3GPP TS 36.300 to understand how RFCs could be 
> misinterpreted (link in email below).
> 
> In this case, it's unclear but they seem to be assuming that there could be a 
> special ECN marking behaviour on control signalling. It's ambiguous whether 
> the 3GPP spec below intends to mean that ECN marking of a stream of packets 
> is either all-on or all-off, rather than signalled by frequency/probability 
> of marking.
> 
> It's perhaps another example to be brought out in the AQM Guidance, probably 
> under:
> "4.5. AQM algorithms SHOULD NOT be dependent on specific transport protocol 
> behaviours"
> which currenty has only one example: "don't assume TCP".

That seems reasonable. I'm not sure, though, what specific change you're 
suggesting. Is it the addition of an example, or rewording of the 
recommendation? Can you give us a specific proposed change?

> Bob
> 
> 
>> Date: Thu, 23 Jan 2014 14:51:25 +0000
>> To: Ingemar Johansson S <[email protected]>
>> From: Bob Briscoe <[email protected]>
>> Subject: RE: [aqm] [tsvwg] Who supports tsvwg adoption of adding ECN to  L2 
>> or tunnel protocols?
>> Cc: "[email protected]" <[email protected]>, "[email protected]" 
>> <[email protected]>, "[email protected]" <[email protected]>
>> 
>> Ingemar, inline...
>> 
>> At 06:36 22/01/2014, Ingemar Johansson S wrote:
>>> Hi
>>> 
>>> Please find answers inline
>>> 
>>> /Ingemar
>>> 
>>> > -----Original Message-----
>>> > From: Bob Briscoe [mailto:[email protected]]
>>> > Sent: den 21 januari 2014 23:50
>>> > To: Ingemar Johansson S; [email protected]
>>> > Cc: [email protected]; [email protected]
>>> > Subject: Re: [aqm] [tsvwg] Who supports tsvwg adoption of adding ECN to L2
>>> > or tunnel protocols?
>>> >
>>> > Ingemar,
> 
> [snip]
> 
>>> > 3) 3GPP TS 36.300 is ambiguous whether ECN marking is applied to all 
>>> > packets
>>> > to "indicate congestion", or whether it is applied with a frequency or
>>> > probability that depends on an AQM (it doesn't mention AQM, altho
>>> > obviously it refers to RFC3168 that is based on AQM). Do you know whether
>>> > it was meant to imply use of AQM?
>>> Yes, it is very ambiguous, I advocate for a frequency or probability as it 
>>> makes most sense, but this is not specified at all. The ECN marking can 
>>> imply the use of AQM but that is again not specified. In short, much of the 
>>> actual implementation is vendor specific.
>> 
>> OK. In general vendor-specific is good, and congestion control can be robust 
>> to different behaviours.
>> 
>> Our draft states that algorithm guidance is outside its scope, but...
>> 
>> One of the purposes of the IETF's AQM guidance is to set bounds on how much 
>> variety is feasible without losing interworking. There's a message in 3GPP 
>> TS 36.300 on how the IETF's RFCs could be (mis)interpreted, if we don't make 
>> them clear.
>> 
>> Bob
>> 
>>> > At 20:29 21/01/2014, Ingemar Johansson S wrote:
>>> > >Hi
>>> > >
>>> > >Please note that ECN over LTE radio access is already standardized in
>>> > >3GPP TS 36.300 (see 11.6 in
>>> > >http://www.3gpp.org/ftp/specs/archive/36_series/36.300/36300-c00.zip )
>>> > >
>>> > >/Ingemar
> 
> ________________________________________________________________
> Bob Briscoe,                                                  BT 

------------------------------------------------------
8 issues in virtual infrastructure
http://dcrocker.net/#fallacies

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

_______________________________________________
aqm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/aqm

Reply via email to