[Detnet-dp-dt] One more comment - Re: Draft update in github

[Loa Andersson <[e-mail]>]

Jouni,

You say:

    Depending on the network topology the "overlay label" may be part of
    the label stack.  This "L-label" has actually nothing specific to do
    with DetNet but can be used to build overlay topologies over the
    provider network in a same way virtual private networks (VPN) are
    built.

What you sday about overlay topoöogies is correct, butthere is one more
characteristic of the L-label it guarantee that the PW-label is
unchanged from DA-x-PE to DA-x-PE, a tunnel that does not
expose the PW-label other than to the DA-S-PE or the DA-T-PE is not
optional, it must be there.

/Loa


On 2017-03-10 13:58, Loa Andersson wrote:
> Jouni,
>
> I was working on reviewing the previous version, I see that
> you captured most of (very close to sll<9 of the comments I had.
>
> Questions:
>
> 1. IETF is mostly doing "on the wire" specifications, what is in the
> box is mostly viewed as implementation specific. Against this background
> why do we need "local-ID", isn't that implementation specific?
>
> 2. There are two sentences "In the context of this document DA-T-PE is
> referred as T-PE." and "In the context of this document DA-S-PE is
> referred as S-PE." Wouldn't it be better to actually use the new
> abbreviations, DA-T-PE and DA-S-PE?
>
> 3. Then I wonder if you got what is optional in the label stack
> and what is not; what needs to be there is one single tunnel, we have
> called that L-labels (PW architecture call it PSN Tunnel) all the rest
> of the T-Label tunnels are  optional.
>
> I wrote it down like this:
>
>    +-------------------------------+
>    |                               |
>    |          DetNet Flow          |
>    |            Payload            |  n octets
>    |                               |
>    +-------------------------------+
>    |      DetNet Flow Id           |  4 octets
>    +-------------------------------+
>    |      DetNet Control Word      |  4 octets
>    +-------------------------------+
>    |          MS-PW Label          |  4 octets
>    +-------------------------------+
>    |            L-Label            |  4 octets
>    +-------------------------------+
>    |   (optional) MPLS T-Label(s)  |  n*4 octets (four octets per label)
>    +-------------------------------+
>
>
> DetNet Flow Payload - n octets
> DetNet Flow Id      - 4 octets, part of the encapsualtion header,
>                       i.e. not in the label stack
> DetNet Control Word - 4 octets, the 16 least significant but are a
>                       a sequence number.
> MS-PW Label         - 4 octets, this label is unchanged between two
>                       DA-x-PEs, and at PW set up it is decided if
>                       the Native Service Processing includes DetNet
>                       FRER or not, the MS-PW Label is swapped at
>                       DA-S-PE.
> L-Label             - carries the MS-PW Label unchanged from one
>                       DA-x-PE to the next
> T-Label(s)          - are optional, and strictly not part of the
>                       DetNet encapsulation.
>
> I don't want you to change but maybe capture a few bits and pieces
> from this.
>
> Then I have one ridiculous concern, the DA-S-PE does not need to
> interface a CE, and does not necessarily sit on a domain border, and is
> tthus not necessary a "real" PE. If we ant to keep calling it a PE
> (I think we should), we should have some words around this.
>
>
> /Loa
>
>
> On 2017-03-10 12:55, Jouni Korhonen wrote:
>> Folks,
>>
>> I did quite a bit of updates to the draft. All in GitHub. I’ll
>> continue writing over the weekend etc..
>>
>> - Jouni
>>
>>
>

-- 


Loa Andersson                        email: [e-mail]
Senior MPLS Expert                          [e-mail]
Huawei Technologies (consultant)     phone: [phone]