Re: [bess] I-D Action: draft-ietf-bess-srv6-services-13.txt

Joel M. Halpern Mon, 21 Mar 2022 04:16:27 -0700

Top-posting, but copying the rest for cotnext if people need it.

What I was trying to suggest is that the presence in the advertissementof an endpoint behavior other than oxFFFF be defined to be the signalthat the Ingress needs to understand the behavior to use theadvertisement.At one point in our conversation, I had thought the arg length beingnon-zero was that indication, but you indicated otherwise.

If we do not do one or the other, and an ingress receives anadvertisement with an unknown behavior, how does it determine if it canuse the advertised SID? Currently, it seems to be the case that itassumes that it can use it. Which seems odd if it has meaning.


Yours,
Joel

On 3/21/2022 7:10 AM, Ketan Talaulikar wrote:

Hi Joel,

Please check inline below with KT2.

On Mon, Mar 21, 2022 at 4:08 PM Joel Halpern Direct<[email protected] <mailto:[email protected]>> wrote:


    In line.

    On 3/21/2022 6:30 AM, Ketan Talaulikar wrote:
     > Hi Joel,
     >
     > Please check inline below.
     >
     >
     > On Mon, Mar 21, 2022 at 2:19 PM Joel M. Halpern
    <[email protected] <mailto:[email protected]>
     > <mailto:[email protected] <mailto:[email protected]>>> wrote:
     >
     >     Okay, so for all cases where the argument length is non-zero, the
     >     receiver can use the advertisement only if they understand
    the behavior?
     >
     >
     > KT> Yes.
     >
     >
     >     What does it mean for the receiver to "use" the behavior when the
     >     argument length is zero?
     >
     >
     > KT> "Use" would be to put the SRv6 Service SID received from the
    egress
     > PE via BGP signaling as the IPv6 DA on the outer IPv6 encapsulation
     > introduced on the ingress PE and routing this encapsulated packet
     > towards the egress PE.

    that was the use I had understood was intended by the draft.  What I do
    not see is any way that the endpoint behavior (and / or the
    understanding or lack of understanding thereof) makes any difference.

     >
     >     What can the receiver do differently based on
     >     understanding the endpoint behavior?
     >
     >
     > KT> The draft does not specify anything in this regard.

    I was prompted to ask this by wondering,, if the behavior could matter,
    if the receiver did not understand the behavior, how could it know
    there
    was a problem.  If this is strictly tied to the argument length being
    non-zero, then I can at least understand it.

KT2> The use of the arguments is the main thing when it comes tointeroperability. Otherwise, the actual processing of the SRv6 SIDbehavior definition is actually happening only at the egress PE thatowns and has instantiated that SRv6 SID. Besides this, the info aboutthe behavior can be used for monitoring/troubleshooting.



    If there is some other case, then there appears to be a determinancy
    problem.   I suspect his could be resolved by having the sender only
    use
    a behavior other than 0xFFFF if the SID can only be used by an Ingress
    who understand the specified (non 0xFFF) behavior??


KT2> Yes, the use of opaque (0xFFFF) is allowed for such scenarios.

Thanks,
Ketan


    Yours,
    Joe;

     >
     >
     >     Yours,
     >     Joel
     >
     >     PS: By "transport" I was (sloppily, sorry) referring to the
     >     transposition mechanism, which seemed to be defined by the
    argument
     >     length, transposition offset, and the route type.  From what
    you say,
     >     that does not suffice to fill in the argument field.  If so, the
     >     exposition in the draft should be improved.
     >
     >
     > KT> The transposition scheme is limited to the BGP encoding alone
    and
     > not related to the rest of the functionality. Perhaps I am still not
     > getting your point here?
     >
     > Thanks,
     > Ketan
     >
     >
     >     On 3/21/2022 4:41 AM, Ketan Talaulikar wrote:
     >      > Hi Joel,
     >      >
     >      > Please check inline below.
     >      >
     >      >
     >      > On Sun, Mar 20, 2022 at 11:13 PM Joel Halpern Direct
     >      > <[email protected]
    <mailto:[email protected]>
    <mailto:[email protected] <mailto:[email protected]>>
     >     <mailto:[email protected]
    <mailto:[email protected]>
     >     <mailto:[email protected]
    <mailto:[email protected]>>>> wrote:
     >      >
     >      >     I have reread the draft.  let me try asking the
    quesiton the
     >      >     opposite way.
     >      >
     >      >     1) If the argument length is zero, then an Ingress PE
    will always
     >      >     ignore
     >      >     the SRv6 Endpoint behavior, as it will not do anything
     >     differently
     >      >     if it
     >      >     understands or does not understand the behavior.
     >      >
     >      >
     >      > KT> The draft does not say that the behavior is to be
    ignored. It
     >     says
     >      > that ingress PE can still use such behaviors even if they are
     >      > unknown/unsupported.
     >      >
     >      >
     >      >     2) If the argument length is non-zero, but it is being
    filled
     >     in by the
     >      >     transportion mechanism, then again the Ingress PE might as
     >     well ignore
     >      >     the SRv6 Endpoint Behavior Information
     >      >
     >      >
     >      > KT> RFC8986 does not talk about any behavior where the
    argument is
     >      > filled by the transport mechanism. Neither does this draft.
     >      >
     >      >
     >      >     3) If other information such as the use of the MPLS ESI or
     >     label (EVPN
     >      >     Auto-Discovery) is used to fill in the argument, this is
     >     determined
     >      >     from
     >      >     the type information and not from the SRv6 Endpoint
    Behavior?
     >      >
     >      >
     >      > KT> Just the route type information is not sufficient and
     >     understanding
     >      > of the behavior is also required before it can be used.
     >      >
     >      >
     >      >     4) If none of those cases apply, and the SRv6 Endpoint
     >     Behavior is
     >      >     known
     >      >     by the Ingress PE, and that behavior reuries other
     >     manipulation of the
     >      >     argument field of the resulting SID, then the Ingress
    PE acts
     >     on that
     >      >     information?  And the advertiser has to somehow ensure
    that all
     >      >     receivers will correctly understand the necessary
    manipulation?
     >      >
     >      >
     >      >     It seems that carrying the SRv6 Endpoint Behavior, and
    trying to
     >      >     describe when it needs to be understood, is for a use case
     >     that is not
     >      >     even covered in the document?
     >      >
     >      >
     >      > KT> I am not sure that I understand the above two
    paragraphs very
     >     well.
     >      > For any behavior where arguments are used, the
     >     understanding/support for
     >      > the behavior is required at the ingress PE - to set the
    argument
     >     part of
     >      > the SID before sending the packets towards the egress PE.
    Where
     >      > arguments are not used, the SID can be used, as signaled,
    by the
     >     ingress
     >      > PE even if it does not understand the behavior.
     >      >
     >      > Thanks,
     >      > Ketan
     >      >
     >      >
     >      >     Yours,
     >      >     Joel
     >      >
     >      >     On 3/20/2022 2:54 AM, Ketan Talaulikar wrote:
     >      >      > Hi Joel,
     >      >      >
     >      >      > Please see inline below.
     >      >      >
     >      >      > On Sun, Mar 20, 2022 at 11:34 AM Joel M. Halpern
     >      >     <[email protected] <mailto:[email protected]>
    <mailto:[email protected] <mailto:[email protected]>>
     >     <mailto:[email protected] <mailto:[email protected]>
    <mailto:[email protected] <mailto:[email protected]>>>
     >      >      > <mailto:[email protected]
    <mailto:[email protected]> <mailto:[email protected]
    <mailto:[email protected]>>
     >     <mailto:[email protected] <mailto:[email protected]>
    <mailto:[email protected] <mailto:[email protected]>>>>> wrote:
     >      >      >
     >      >      >     I seem to be missing something.
     >      >      >
     >      >      >     The ingress PE (domain edge) applies the
    destination SID
     >      >     (possibly as
     >      >      >     part of a SID list).  Either it is deciding to
    use the
     >      >     destination SID,
     >      >      >     or something else is deciding to use the
    destination SID.
     >      >      >
     >      >      >
     >      >      > KT> The ingress PE is deciding. Something else (e.g., a
     >      >     controller) may
     >      >      > decide the path (e.g., SID list for SR Policy) but the
     >     Service/VPN
     >      >      > context is signaled via BGP from egress to ingress PE.
     >      >      >
     >      >      >
     >      >      >     Ignoring the issue of argument manipulation, if the
     >     Ingress PE is
     >      >      >     deciding on its own, doesn't it have to
    understand the
     >      >     meaning of the
     >      >      >     behavior in order to decide that it wants to
    invoke it?
     >      >      >
     >      >      >
     >      >      > KT> The ingress PE is not invoking anything and
    hence it
     >     doesn't
     >      >     need to
     >      >      > understand the meaning of the behavior (with some
     >     exceptions like
     >      >     when
     >      >      > it needs to supply the argument). Ingress PE is simply
     >     setting the
     >      >      > received SRv6 Service SID as the IPv6 DA in the outer
     >     encapsulation
     >      >      > (let's keep aside SR Policy for now). When this packet
     >     reaches the
     >      >      > egress PE, it ends up invoking the behavior
    corresponding
     >     to the
     >      >     locally
     >      >      > instantiated SRv6 SID on the egress PE. As an analogy -
     >     whether the
     >      >      > label signaled by the egres PE is per-VRF or per-CE
    does not
     >      >     affect the
     >      >      > processing at ingress PE.
     >      >      >
     >      >      >     If something else provides the SID list and the
    rules for
     >      >     which traffic
     >      >      >     should use it (e.g. the SR policy or similar)
    then the
     >      >     Ingress PE would
     >      >      >     not seem to need such understanding.
     >      >      >
     >      >      >
     >      >      > KT> The situation is the same even in this case.
     >      >      >
     >      >      > Thanks,
     >      >      > Ketan
     >      >      >
     >      >      >
     >      >      >     Yours,
     >      >      >     Joel
     >      >      >
     >      >      >     On 3/20/2022 1:37 AM, Ketan Talaulikar wrote:
     >      >      >      > Hi Joel,
     >      >      >      >
     >      >      >      > There is no implicit assumption such as the
    one you
     >     refer
     >      >     to. The
     >      >      >      > ingress PE does not need to do anything specific
     >     with the
     >      >     choice
     >      >      >     of the
     >      >      >      > behavior picked by the egress PE except
    where the
     >     behavior
     >      >      >     involves the
     >      >      >      > use of argument. Ingress PE does need to
    know & support
     >      >     the specific
     >      >      >      > behavior when it needs to supply the
    argument based
     >     on the
     >      >     behavior
     >      >      >      > definition.
     >      >      >      >
     >      >      >      > Thanks,
     >      >      >      > Ketan
     >      >      >      >
     >      >      >      > On Sun, Mar 20, 2022 at 10:56 AM Joel M. Halpern
     >      >      >     <[email protected]
    <mailto:[email protected]> <mailto:[email protected]
    <mailto:[email protected]>>
     >     <mailto:[email protected] <mailto:[email protected]>
    <mailto:[email protected] <mailto:[email protected]>>>
     >      >     <mailto:[email protected]
    <mailto:[email protected]> <mailto:[email protected]
    <mailto:[email protected]>>
     >     <mailto:[email protected] <mailto:[email protected]>
    <mailto:[email protected] <mailto:[email protected]>>>>
     >      >      >      > <mailto:[email protected]
    <mailto:[email protected]>
     >     <mailto:[email protected] <mailto:[email protected]>>
    <mailto:[email protected] <mailto:[email protected]>
     >     <mailto:[email protected] <mailto:[email protected]>>>
     >      >     <mailto:[email protected]
    <mailto:[email protected]> <mailto:[email protected]
    <mailto:[email protected]>>
     >     <mailto:[email protected] <mailto:[email protected]>
    <mailto:[email protected] <mailto:[email protected]>>>>>> wrote:
     >      >      >      >
     >      >      >      >     I keep reading the description of the
    handling
     >     of unknown
     >      >      >     endpoint
     >      >      >      >     behaviors.
     >      >      >      >
     >      >      >      >     It seems there is an implicit assumption
    that I
     >     would
     >      >     think
     >      >      >     it would be
     >      >      >      >     helpful to make explicit.  As far as I
    can tell, a
     >      >     head end
     >      >      >     would never
     >      >      >      >     choose based purely based on local policy to
     >     make use
     >      >     of an
     >      >      >     advertised
     >      >      >      >     SID with an unknown behavior?  However,
    a head end
     >      >     might use
     >      >      >     such a
     >      >      >      >     ISD,
     >      >      >      >     without knowing what it was really
    asking, if so
     >      >     instructed
     >      >      >     by a policy
     >      >      >      >     engine (e.g. SR Policy)?
     >      >      >      >
     >      >      >      >     Yours,
     >      >      >      >     Joel
     >      >      >      >
     >      >      >      >     On 3/19/2022 11:32 PM,
    [email protected] <mailto:[email protected]>
     >     <mailto:[email protected]
    <mailto:[email protected]>>
     >      >     <mailto:[email protected]
    <mailto:[email protected]>
     >     <mailto:[email protected]
    <mailto:[email protected]>>>
     >      >      >     <mailto:[email protected]
    <mailto:[email protected]>
     >     <mailto:[email protected]
    <mailto:[email protected]>>
     >      >     <mailto:[email protected]
    <mailto:[email protected]>
     >     <mailto:[email protected]
    <mailto:[email protected]>>>>
     >      >      >      >     <mailto:[email protected]
    <mailto:[email protected]>
     >     <mailto:[email protected]
    <mailto:[email protected]>>
     >      >     <mailto:[email protected]
    <mailto:[email protected]>
     >     <mailto:[email protected]
    <mailto:[email protected]>>>
     >      >      >     <mailto:[email protected]
    <mailto:[email protected]>
     >     <mailto:[email protected]
    <mailto:[email protected]>>
     >      >     <mailto:[email protected]
    <mailto:[email protected]>
     >     <mailto:[email protected]
    <mailto:[email protected]>>>>> wrote:
     >      >      >      >      >
     >      >      >      >      > A New Internet-Draft is available
    from the
     >     on-line
     >      >      >      >     Internet-Drafts directories.
     >      >      >      >      > This draft is a work item of the BGP
    Enabled
     >      >     ServiceS WG
     >      >      >     of the IETF.
     >      >      >      >      >
     >      >      >      >      >          Title           : SRv6 BGP based
     >     Overlay
     >      >     Services
     >      >      >      >      >          Authors         : Gaurav Dawra
     >      >      >      >      >                            Clarence
    Filsfils
     >      >      >      >      >                            Ketan
    Talaulikar
     >      >      >      >      >                            Robert Raszuk
     >      >      >      >      >                            Bruno Decraene
     >      >      >      >      >                            Shunwan Zhuang
     >      >      >      >      >                            Jorge Rabadan
     >      >      >      >      >       Filename        :
     >      >     draft-ietf-bess-srv6-services-13.txt
     >      >      >      >      >       Pages           : 34
     >      >      >      >      >       Date            : 2022-03-19
     >      >      >      >      >
     >      >      >      >      > Abstract:
     >      >      >      >      >     This document defines procedures and
     >     messages for
     >      >      >     SRv6-based BGP
     >      >      >      >      >     services including L3VPN, EVPN,
    and Internet
     >      >     services.  It
     >      >      >      >     builds on
     >      >      >      >      >     RFC4364 "BGP/MPLS IP Virtual Private
     >     Networks
     >      >     (VPNs)"
     >      >      >     and RFC7432
     >      >      >      >      >     "BGP MPLS-Based Ethernet VPN".
     >      >      >      >      >
     >      >      >      >      >
     >      >      >      >      > The IETF datatracker status page for this
     >     draft is:
     >      >      >      >      >
     >      >      >
     > https://datatracker.ietf.org/doc/draft-ietf-bess-srv6-services/
    <https://datatracker.ietf.org/doc/draft-ietf-bess-srv6-services/>