Re: [mpls] 答复: I send my second question on draft-farrel-mpls-sfc-02.txt according to co-chair's request

So, for you is there a driver to add complexity in MPLS data plane other than supporting NSH?
I am not advocating we add complexity for adding NSH either, but I am advocating to support it.

On 14/11/2017, 16:35, "Adrian Farrel" <adrian@olddog.co.uk> wrote:

    Agree with you about "NSH in MPLS" encapsulation. Needs specification.

    The control plane discussion is (I think) and SR vs swapping discussion. The CP can be used for NSH or MPLS swapping (noting that NSH is a swapping paradigm).

    The CP can (and should) be used to describe what happens at each hop (i.e., NSH, MPLS, foo). IIRC that is in the BESS draft.

    And you may note that I support the BESS draft :-)

    A

    > -----Original Message-----
    > From: Henderickx, Wim (Nokia - BE/Antwerp)
    > [mailto:wim.henderickx@nokia.com]
    > Sent: 14 November 2017 07:30
    > To: Xuxiaohu; adrian@olddog.co.uk; mpls@ietf.org
    > Subject: Re: [mpls] 答复: I send my second question on draft-farrel-mpls-sfc-
    > 02.txt according to co-chair's request
    > 
    > In general, there is a 3rd approach which should be added to the discussion here.
    > The 3rd approach is doing all in the control plane and use MPLS/SPRING as a way
    > to get from one SFF to another. Example: https://tools.ietf.org/html/draft-ietf-
    > bess-service-chaining-03
    > 
    > Can we discuss why complicating the data plane for this is a good idea and what
    > value it brings over doing it in control plane.
    > 
    > Irrespective of this we will need a way to encapsulate NSH with MPLS transport
    > and how we signal, this irrespective of the above
    > 
    > On 14/11/2017, 15:12, "mpls on behalf of Xuxiaohu" <mpls-bounces@ietf.org on
    > behalf of xuxiaohu@huawei.com> wrote:
    > 
    > 
    > 
    >     ________________________________________
    >     发件人: Adrian Farrel [adrian@olddog.co.uk]
    >     发送时间: 2017年11月14日 15:07
    >     收件人: Xuxiaohu; mpls@ietf.org
    >     主题: RE: [mpls] I send my second question on draft-farrel-mpls-sfc-02.txt
    > according to co-chair's request
    > 
    >     Hi,
    > 
    >     Yes, good question.
    > 
    >     Both drafts describe a label stacking approach where two labels are used at
    > each
    >     hop.
    >     The major difference is in how metadata is handled.
    >     draft-xu-mpls-service-chaining uses an NSH at the bottom of the stack to
    >     encapsulate the metadata. This has the benefit of re-using an existing
    >     mechanism/encoding and also the benefit of placing the metadata in the same
    >     packet as the user data. There are also some drawbacks as mentioned in the
    > SFC
    >     meeting: the particular concerns are about packet size and disruption,
    >     encapsulation rules (is a control word used to make it clear that the NSH is
    >     there and prevent accidental inspection of the NSH?), and interpretation of the
    >     other NSH fields (e.g., what does the SI field in the NSH mean? what happens if
    >     the NSH TTL is zero).
    > 
    >     [Xiaohu] Thanks a lot for your clarification about the difference.
    > 
    >     Do you see any other differences?
    > 
    >     [Xiaohu] Another difference I noticed that you rename the node segment label
    > as context label :)
    > 
    >     Best regards,
    >     Xiaohu
    > 
    >     Adrian
    > 
    >     > -----Original Message-----
    >     > From: mpls [mailto:mpls-bounces@ietf.org] On Behalf Of Xuxiaohu
    >     > Sent: 14 November 2017 06:58
    >     > To: mpls@ietf.org
    >     > Subject: [mpls] I send my second question on draft-farrel-mpls-sfc-02.txt
    >     > according to co-chair's request
    >     >
    >     > Hi all,
    >     >
    >     > My second question is what's the difference between the second MPLS-
    > based
    >     > SFC approach (draft-farrel-mpls-sfc-02.txt) and the MPLS-SR-based SFC
    > approach
    >     > (draft-xu-mpls-service-chaining) from the technology perspective.
    >     >
    >     > Best regards,
    >     > Xiaohu
    >     > _______________________________________________
    >     > mpls mailing list
    >     > mpls@ietf.org
    >     > https://www.ietf.org/mailman/listinfo/mpls
    >     _______________________________________________
    >     mpls mailing list
    >     mpls@ietf.org
    >     https://www.ietf.org/mailman/listinfo/mpls
    > 