Re: [PWE3] Mandatory Control Word

Is that correct?  I'm pretty sure that most Ethernet PW deployments don't
use a CW (none of the Ethernet PWs I configured ever used a CW, but to be
fair that was a few years ago...)

For Ethernet there are no L2 flags, or minimum length issues, so the main
reasons for a CW would seem to be for VCCV (for which there are other
encapsulation options) and ECMP (which is only an issue in some networks).

Oh, from reading the draft you state that FR Port Mode, PPP, and HDLC all
have the same encap.  That's not strictly true.  For PPP mode you remove the
HDLC address/control fields (if present - remember that PPP can run over
non-HDLC media).

Giles

On 27/10/2010 18:28, "Thomas Nadeau" <tnadeau@lucidvision.com> wrote:

> 
> Eric,
> 
>     What you surmise below is PRECISELY the point that Nick/I/others are
> making.
> It is clear that in the large majority of deployments, the use of the CW *is*
> the typical configuration. The only real non-use is with ATM PWs to apparently
> "save bandwidth", but the real issue is to keep old equipment working. There
> is no argument that we do not want to obviate older equipment, but what we
> want is going forward to be doing the right thing for interoperability.
>  
>      The other observation we've made is that the myriad of options
> available to VCCV result in serious interoperability issues. These
> different modes are available largely due to not making the CW
> mandatory (as well as the bailing-wire and bubble gum approach we
> originally took to encapsulating different PW types).
>  
>      The road to interop is simplification.  All we are proposing is:
>  
>      a) any protocols/documents in PWE3 that are not RFC at present, be
> required
>           to use the CW.
>  
>      b) any existing protocols in PWE3 and future ones be required to use the
> CW 
> when going to proposed standard status.
>  
>      c) adoption of the negotiation of the CW status based on the draft that
> we proposed      
> in the interim to aid in more automatic interop for operators.
> 
> --Tom
> 
> 
>> Stewart,
>> 
>> I agree with your suggested limitations to new PW types
>> and suggestion to recommend its use for existing PW types.
>> 
>> However, the observation about deployed systems - when
>> taken together with what appears to be a growing impression
>> that use of the control word with all PW types may be a good
>> idea - seems to argue that we would (eventually) need a BCP
>> that specifies use of the control word, and possibly we may
>> even have to publish a "non-use of PW control word considered
>> dangerous" RFC.
>> 
>> I don't see a way to avoid this, assuming that is where
>> we are headed, but it might be good to be clear on whether or
>> not that _is_ where we are headed.
>> 
>> It also causes me to wonder if there are implementations
>> deployed in the field where use of the control word is not a
>> configurable capability.  Certainly, if deployed equipment is
>> all capable of being configured to always use control words,
>> then it is not unrealistic to suggest that configuring things
>> that way would simplify future implementations.
>> 
>> --
>> Eric
>> 
>> -----Original Message-----
>> From: pwe3-bounces@ietf.org [mailto:pwe3-bounces@ietf.org] On Behalf Of
>> Stewart Bryant
>> Sent: Tuesday, October 26, 2010 1:24 PM
>> To: pwe3
>> Subject: [PWE3] Mandatory Control Word
>> 
>> From
>> 
>> draft-delregno-pwe3-mandatory-control-word/
>> 
>> =======
>> 2.  Mandatory Control Word
>> 
>>    The Control Word SHALL be mandatory for all PWE3 encapsulations.  The
>>    use of the sequence number remains OPTIONAL.
>> 
>>    As a result of the Control Word being Mandatory, all implementations
>>    of the PWE3 encapsulations SHALL follow Section 6.1 of [RFC4447]
>>    wherein the "PWs MUST have c=1".  This requirement SHALL remain until
>>    such time, if ever, RFC4447 is superceded and the support for Control
>>    Word negotiation is removed as a result of this mandate.
>> 
>> ======
>> 
>> Given the reality of network deployments I do not see how we can
>> talk about removing the requirement to negotiate the CW in any
>> realistic time frame.
>> 
>> It's fine to require the CW for all new PW types, and it's
>> probably OK to RECOMMEND operation with the CW, but it's
>> unrealistic to remove the option given the extent of the deployed
>> systems.
>> 
>> - Stewart
>> 
>> 
>> 
>> _______________________________________________
>> pwe3 mailing list
>> pwe3@ietf.org
>> https://www.ietf.org/mailman/listinfo/pwe3
>> _______________________________________________
>> pwe3 mailing list
>> pwe3@ietf.org
>> https://www.ietf.org/mailman/listinfo/pwe3
>> 
> 
> _______________________________________________
> pwe3 mailing list
> pwe3@ietf.org
> https://www.ietf.org/mailman/listinfo/pwe3