Re: [Lsr] I-D Action: draft-ietf-lsr-isis-srv6-extensions-10.txt

["Ketan Talaulikar (ketant)" <[e-mail]>]

Hi Joel,

Please check inline below.

-----Original Message-----
From: Lsr <[e-mail]> On Behalf Of Joel M. Halpern
Sent: 25 September 2020 03:18
To: Acee Lindem (acee) <[e-mail]>; [e-mail]
Subject: Re: [Lsr] I-D Action: draft-ietf-lsr-isis-srv6-extensions-10.txt

First, there is a slight confusion in the way I formed the quesiton, but I think it still applies.

The piece of this draft is section 9, which advertises the length of the arg portion of the SID.  But does not provide specific meanings for specific values.
[KT] This is quite appropriate for this draft since it is only specifying a generic SID structure and not associated with any specific behavior.

The example of an ARG in the network programming draft does provide part of the explicit interpretation of the ARG.  It says that it is a list of k items, each of x bits, where each x bit blob identifies an OIF.
[KT] The net-pgm draft in sec 4.12 introduces a specific End.DT2M behavior which includes support for ARG. That said, I am not quite sure about that text in that section which talks about how the ARG bits are formed and what they signify. I believe the ARG in this case is a locally assigned identifier that maps to an ESI so that it can be used for ESI filtering - much the same as an ESI label for split-horizon filtering. I see a comment from one of the ADs on this and I expect that the authors will clarify.

This leaves two gaps, and a more general question.
1) How does the receiver know the meanings of the OIF indices so that he can correctly fill them in?
2) The NP draft says that k and x are defined on a per SID basis.  But I do not see anywhere in the isis draft to advertise the values of k and x, only arg (which is k*x).
[KT] I hope the previous comment explains.

The more general question is, is there a requirement we can write down about how receivers will be able to understand ARG fields in general? 
One can argue that it would belong in the network programming draft; I would prefer not to delay that with a significant technical addition.
[KT] I don't believe the handling of ARG is something that can be generalized. It has to be something specific to the behavior that it is associated with. Therefore, each behavior that supports an ARG needs to specify its handling. The net-pgm draft is doing it for End.DT2M and future documents that introduce other behaviors requiring ARG would be expected to the same.

Thanks,
Ketan

There is a related question that I came across while trying to explain this question.

END.T must be associated with a forwarding table.  I presume this is done by where one puts the END.T (however-many-subs) TLV.  But I can not find anything in this draft that says this.  There is precisely one reference to End.T in the draft.

Thank you,
Joel

On 9/24/2020 5:25 PM, Acee Lindem (acee) wrote:
> H Joel,
> 
> Can you reference the specific section in the IS-IS SRv6 draft you are commenting on? I seem to remember this discussion but it was at least a month back, if not more.
> 
> Thanks,
> Acee
> 
> On 9/23/20, 6:31 PM, "Lsr on behalf of Joel Halpern" <[e-mail] on behalf of [e-mail]> wrote:
> 
>      The announcement prompted me to look again and think about an
>      interaction between this and the network programming draft.  To be
>      clear, I am NOT objecting to either this or the network programming
>      draft.  I am just wondering what I am missing.
> 
>      The NP draft, and the advertisement mechanism allows a router to
>      advertise the number of bits for the ARG portion of a SID.
> 
>      Q1: The point presumably is to avoid needing to advertise each of the
>      individual values?
> 
>      An example of this is, I think, and ARG for the table selection where
>      the ARG is the table number for the packet to be looked up in?
> 
>      Q2: If so, how does the head end know what table number corresponds to
>      what meaning?    If this requires a separate advertisement there seems
>      to be no savings.  if this requires out-of-band knowledge then we seem
>      to have lost the benefit of advertising all of this in the routing protocol.
> 
>      I suspect I am simply missing a piece.  can someone explain please?
> 
>      Thank you,
>      Joel
> 
>      On 9/23/2020 4:40 PM, [e-mail] wrote:
>      >
>      > A New Internet-Draft is available from the on-line Internet-Drafts directories.
>      > This draft is a work item of the Link State Routing WG of the IETF.
>      >
>      >          Title           : IS-IS Extension to Support Segment Routing over IPv6 Dataplane
>      >          Authors         : Peter Psenak
>      >                            Clarence Filsfils
>      >                            Ahmed Bashandy
>      >                            Bruno Decraene
>      >                            Zhibo Hu
>      > 	Filename        : draft-ietf-lsr-isis-srv6-extensions-10.txt
>      > 	Pages           : 25
>      > 	Date            : 2020-09-23
>      >
>      > Abstract:
>      >     Segment Routing (SR) allows for a flexible definition of end-to-end
>      >     paths by encoding paths as sequences of topological sub-paths, called
>      >     "segments".  Segment routing architecture can be implemented over an
>      >     MPLS data plane as well as an IPv6 data plane.  This draft describes
>      >     the IS-IS extensions required to support Segment Routing over an IPv6
>      >     data plane.
>      >
>      >
> 
>      _______________________________________________
>      Lsr mailing list
>      [e-mail]
>      https://www.ietf.org/mailman/listinfo/lsr
> 

_______________________________________________
Lsr mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/lsr