Re: [Softwires] A General Question for Flow Label Usage

[Brian E Carpenter <[e-mail]>]

I believe it is compatible as stated, but with two provisos:

1. There needs to be a complete specification, to check definitively.

2. There's active debate in 6MAN about changing the spec for the flow
label; personally I don't expect any changes that would invalidate
the usage below.

   Brian

On 2010-08-24 23:34, Yiu L. Lee wrote:
> I want to follow up this topic on the list. I exchanged few off-listed
> emails with Brian to ask his opinion of using flow label. It seem the way
> the ER to encapsulate every packet of a given host is a compatible use of
> flow label. Brian, can you confirm that please?
> 
> Thanks,
> Yiu
> 
> 
> On 7/30/10 4:57 PM, "Brian E Carpenter" <[e-mail]> wrote:
>>>> Hi,
>>>>
>>>> In RFC 3697, it says:
>>>>
>>>> "A flow is a sequence of packets sent from a particular source to a
>>>>    particular unicast, anycast, or multicast destination that the source
>>>>    desires to label as a flow."
>>>>
>>>> It later says the usage of a flow:
>>>>
>>>> "The usage of the 3-tuple of the Flow Label and the Source and
>>>>    Destination Address fields enables efficient IPv6 flow
>>>>    classification, where only IPv6 main header fields in fixed positions
>>>>    are used."
>>>>
>>>> I am confused why we can't use a flow label to effectively identify a flow
>>>> from a host behind a CPE in the AFTR. In what part this usage is against the
>>>> specification?
>> The devil is in the details. You'd have to define the use case completely
>> before I (as a co-author of 3697) would try to answer the question.
>>
>> The use case still has to work with hosts that do not set a flow label,
>> which is of course essentially all of them today.
>>
>>    Brian
> 
>