[mpls] Re: [spring] Hop by Hop TE in SR networks (aka draft-beeram-mpls-rsvp-sr-01)

Vishnu Pavan Beeram <vishnupavan@gmail.com> Fri, 13 November 2026 22:07 UTC

Received: from mail-lj1-x22a.google.com (mail-lj1-x22a.google.com [IPv6:2a00:1450:4864:20::22a]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange x25519 server-signature ECDSA (prime256v1) server-digest SHA256) (No client certificate requested) by mx.ietf.org (Postfix) with ESMTPS id DA49850 for <mpls@ietf.org>; Fri, 13 Nov 2026 22:07:12 +0000 (UTC)
Authentication-Results: mx.ietf.org; dkim=pass header.d=gmail.com header.s=20260707 header.b=Be0H6g11; spf=pass (mx.ietf.org: domain of vishnupavan@gmail.com designates 2a00:1450:4864:20::22a as permitted sender) smtp.mailfrom=vishnupavan@gmail.com; dmarc=pass (policy=none) header.from=gmail.com; arc=pass ("google.com:s=arc-20260327:i=1")
Received: by mail-lj1-x22a.google.com with SMTP id 38308e7fff4ca-3a49a022221so18025901fa.3 for <mpls@ietf.org>; Fri, 13 Nov 2026 14:07:12 -0800 (PST)
ARC-Seal: i=1; a=rsa-sha256; t=1794607626; cv=none; d=google.com; s=arc-20260327; b=eEZvz4vPA/P+ajNr41xnPwou5f+aR+oAVgQWAWTyi9EdzH7MVdEGWPdaqFccxRAg9x 6K+QBoWg8GT5GogcLKou0J/ukGazJm7RYkrGTwPGCMDcy7O3VzrWwMgrCaxOZIFy/gjv FTFiC+u6ArLuqOyraUfnf+ILV33ZG5IMPN/c250Jg/3aQgbMwx/3ruLJSMsgba8CIAZ6 WNjZxjqlX679qKlceE0MLYLFDOv2oK4rBr8FpUoogdLMOX11JiKq424gMzAN3MI9I8iQ U4A+jBeXJyFAuakl9c0GYNzCJOUgwZPQ7qQZqwt/SzVCxl2cZNSnV9GsVTwP9+wmg1ZF 7V/Q==
ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20260327; h=cc:to:subject:message-id:date:from:in-reply-to:references :mime-version:dkim-signature; bh=oKCarfpp/KLWLCmWOwpmiWMeVuQw1Oai56gfYLkdnT0=; fh=/A66bmTstKUA5w0aHckum/SqSpJOqzYSuMfRTQhsSwI=; b=he5ql4Cj+6GPoOlYz9bd6SCXyTtZ8UTWt03i7azC7ZFBv8X89FeI1yNo/wn8t5rkoC /CVVx9aVfHtVtSwnB57MtTT1hsgFjo06k/1ZyzS5UsODiUDy9F/52t4GAClNaUTqPKSw hKNezY6yNRuq940OXIYWGNmzrB9cFu1EDkVOBjzQ1omw0dz+FTd7fj4w6OOfVkylU7OC pdITvPGDo0zfGHJPdELVytr0X4p3dWJixna7LLca5DunMES4JJNcZ4+f2dFJikNCDmTk //Cfwax2S0t/dgsqxFwwtkDJnjcNbFXJCIit9GOn16uf1ZI6wWV+Ttd9rDW3euDrt2lh qUpQ==; darn=ietf.org
ARC-Authentication-Results: i=1; mx.google.com; arc=none
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20260707; t=1794607626; x=1795212426; darn=ietf.org; h=content-type:cc:to:subject:message-id:date:from:in-reply-to :references:mime-version:from:to:cc:subject:date:message-id:reply-to :content-type; bh=oKCarfpp/KLWLCmWOwpmiWMeVuQw1Oai56gfYLkdnT0=; b=Be0H6g11QPV+S0/+vphaOdpx1lvAqcJHhCvX6/u403AJZYA74rkJQTZZU3diYlL9EM quNPKo+mFZ8aoJzbVNvVMUO5XgGzU4+awW/IfrOddwDfN4Axg0ChTCgoU8v4FRkYDq/r UxYhDp6g+4ZHFQkcsMD/iM92A+34/KlVnVxAebZFPG2dnAAblbpRfElyNd/QtKC2yMM/ ozVHCYTclRdH6m3wfnuOhe66ik8XY1QyTzj8btjJuXO5HLX3vh8P1AqzfHaQhKZRKpIr KGt5Etw2l2M4gC9Cwag+U5TUNxxWxfpTQ0GjXBSp0lvTATuIlH6hB8JJZ64zFv7XnDLO fveQ==
X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20260707; t=1794607626; x=1795212426; h=content-type:cc:to:subject:message-id:date:from:in-reply-to :references:mime-version:x-gm-gg:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to:content-type; bh=oKCarfpp/KLWLCmWOwpmiWMeVuQw1Oai56gfYLkdnT0=; b=DrIt+ggKjvhLvfu/DYvhFCOwMTCv1VKF8tGXTfrLMG7GF8oFZCNu2+GZdvVuuI/o0N WJPKHXdrBLAXP0BnWJb0Mhqb/EbnpDta6GO8i6RjbPWBxDqRuzvMNpUlg/ofQRypMCUh +aGWOOK2RY/ZCKSUqq2rW65Ev9m1VRI6qfb87LopNl6HCE29Y8AOz8Ti5kQU5D/m/55r 9X59yWGZjP/fdsrSnecwloHOnf7Hz7mlD/Ob54lQnfTDClQ5ij2Iyea+Nx2LOHEDbJ/+ zt0WgNh3ybiNSii/oVhBGRSONIOVCodbkFrQF/11UgFOYW5k7xBeXe2J+TLRWz3xRoX1 vYEg==
X-Forwarded-Encrypted: i=1; AL+hErGHJwgyr3ScFItLFEpIpsJOKfYv09ruNs8+nqZIs3tjlHtQECclFsqoI8aSXJKbCkOdfRi6@ietf.org
X-Gm-Message-State: APuU4Z9vQxvRuGwaxTQzlgyM1xQsvOw+DaHwmKdmW07Albdx08r1inzd rpF5qQ+xLumDq+prPDER8HbEkbTY4vBc1hHRixAdoxo3khqdvpf3VFcCmdDPS5aq0tALJlZGL9U fK1U8lyeg80FoW6rqzFwaf3wV06O7LUjnHRU8+aY=
X-Gm-Gg: AYBFou3u0kR8pAi2xUuGggkG7GytdhS/z41lUAuHbkFznlqUg6dau667SnSHMMPlOA5 wfE30nj40gDn+xbwpOAL0msnbVPDR5KJbqQV25M5//TyeDw2fuHFrDmuLpJnsFKJjMqXaNpUhIp Qpj0BEEtL2nIkuVdA253Xb0CwdCZQ+KXjnC5QxfjbvIH9hWV5o9z3gs0JwFqD5jtFQrEXtnodLC YegpJfcBLKhq1n9dXQEMUMe3x8CBlFCFNFqTx+42ltgvWVaGcsPxDLZPSFrXxSEVe8kyl86598T DDALBxvjuOdiBMYcgM9F3jC9DCQTOcTbwt4SjjslT+w/rdB1gnsrtjcpXjDrNOj1tzgTFgXi1RF RDuKbnOZcb95KXA5S8AjdgCgP4n39ak4qwpBbanGiV8bP/b0/CqRo9J01BqTyejGuzHY+uQ==
X-Received: by 2002:a05:651c:421b:b0:3ab:751:6ba8 with SMTP id 38308e7fff4ca-3ae2b5564f5mr11807041fa.28.1794607625864; Fri, 13 Nov 2026 14:07:05 -0800 (PST)
MIME-Version: 1.0
References: <CAOj+MMGNFOvZvsyoMU1oDi78PUs_xfe2aSNJWpE6uPRvjCGzsA@mail.gmail.com> <F55A428D-47B4-4EC4-A184-0CA2733F051C@arrcus.com> <SJ2PR84MB3370BA05E77F332076730FBD92702@SJ2PR84MB3370.NAMPRD84.PROD.OUTLOOK.COM> <CAOj+MMGKSdDF1Q+wf9Z7ctUe5hGwKVCiXjBf++Uy1ZJBGboeSw@mail.gmail.com> <SJ2PR84MB3370249C98105C5D535F1605927F2@SJ2PR84MB3370.NAMPRD84.PROD.OUTLOOK.COM> <CAOj+MMEFjfi7ZYN=L-it_DW7wyqwT=BzpZmdzt2zquAqJOpn_A@mail.gmail.com> <SJ2PR84MB3370CC9613660295114981EC927F2@SJ2PR84MB3370.NAMPRD84.PROD.OUTLOOK.COM> <CAOj+MMF4uWQpNgv4Z2AJxYP8XhrX9Vi=5N9fW0LKdZtOz5-xUQ@mail.gmail.com> <02669E2E-2A9A-4544-9762-0FDC6A9578E6@arrcus.com> <CAOj+MMHfz9TV2m-BWMExuUdeJjT8gGujDpaJx7TjBQfJMp5kSg@mail.gmail.com>
In-Reply-To: <CAOj+MMHfz9TV2m-BWMExuUdeJjT8gGujDpaJx7TjBQfJMp5kSg@mail.gmail.com>
From: Vishnu Pavan Beeram <vishnupavan@gmail.com>
Date: Fri, 13 Nov 2026 14:06:53 -0800
X-Gm-Features: AclHuK9i_BNauHjjJiZK1Uxw3ehbOVcAyKVg0ZvV9DPw-_xarknOHADI-akqhR8
Message-ID: <CA+YzgTve36KxvpHLOYc-nmxRLwmYhTqtH=v8f10-7cZb6aiqJQ@mail.gmail.com>
To: Robert Raszuk <robert@raszuk.net>
Content-Type: multipart/alternative; boundary="0000000000000d07f506602fa6c5"
X-Spamd-Bar: ---------------------
Message-ID-Hash: GMAYYTI2PCEZVK5E5YZW7I7FGQRIWWCJ
X-Message-ID-Hash: GMAYYTI2PCEZVK5E5YZW7I7FGQRIWWCJ
X-MailFrom: vishnupavan@gmail.com
X-Mailman-Rule-Misses: dmarc-mitigation; no-senders; approved; loop; banned-address; header-match-mpls.ietf.org-0; emergency; member-moderation; nonmember-moderation; administrivia; implicit-dest; max-recipients; max-size; news-moderation; no-subject; digests; suspicious-header
CC: Andy Smith <andy@arrcus.com>, "Barth, Colby" <jonathan.barth@hpe.com>, mpls <mpls@ietf.org>, TEAS WG <teas@ietf.org>, SPRING WG <spring@ietf.org>, "Andrew Stone (Nokia)" <andrew.stone@nokia.com>
X-Mailman-Version: 3.3.10
Precedence: list
Subject: [mpls] Re: [spring] Hop by Hop TE in SR networks (aka draft-beeram-mpls-rsvp-sr-01)
List-Id: Multi-Protocol Label Switching WG <mpls.ietf.org>
Archived-At: <https://mailarchive.ietf.org/arch/msg/mpls/SQTb0em_90bebOFH0rUopUwh970>
List-Archive: <https://mailarchive.ietf.org/arch/browse/mpls>
List-Help: <mailto:mpls-request@ietf.org?subject=help>
List-Owner: <mailto:mpls-owner@ietf.org>
List-Post: <mailto:mpls@ietf.org>
List-Subscribe: <mailto:mpls-join@ietf.org>
List-Unsubscribe: <mailto:mpls-leave@ietf.org>

Hi Robert,

Sticking to RFC 8577 here, whose delegation procedure the SR-MPLS draft
preserves (Section 5).

*How delegation points are signaled. *Not via loose hops. For explicit
delegation, the ingress sets the LSI-D flag in a HOP_ATTRIBUTES sub-object
(RFC 7570) attached to the ERO entry for the chosen hop (RFC 8577 Sections
5.2 and 9.4). For automatic delegation, the ingress sets LSI-D in
LSP_ATTRIBUTES and signals an ETLD in the Path message. Hops select
themselves as delegation hops as the ETLD runs out, so all delegation hops
are known by the time the Path message reaches the egress. During Resv
processing, each delegation hop selects or allocates its delegation label
and records it in the RRO with the Delegation flag (Sections 5.3.1 and
9.5). In my experience with an implementation that has been shipping RFC
8577 support for more than 7 years, automatic delegation is the default,
and it has proven adequate.

*Loose hops.* In intra-domain TE deployments, any loose hops a user
provides are typically expanded by the ingress's path computation before
signaling, so the ERO that is signaled is usually all strict hops. Either
way, the Path message follows the ERO hop by hop and each hop admits
bandwidth on its own link. A delegation hop only imposes labels; it doesn't
compute or change the path. Bandwidth reservations are based on the
traversed TE link, so whether a hop is a delegation hop is irrelevant to
admission control.

*Delegation label sharing.* This depends on the stacking approach the
ingress chooses (Sections 5.1.1, 5.1.2 and 9.6). With "stack to reach
delegation hop", a delegation label includes the next delegation label, so
it is shared only among LSPs to the same egress over the same downstream
path. With "stack to reach egress", it is shared among all LSPs crossing
the segment between two delegation hops. In both cases the state is per
segment or downstream path, not per LSP. That same implementation defaults
to "stack to reach delegation hop", and in my experience that default has
proven adequate.

Regards,
-Pavan
P.S. For background on automatic delegation of label stack imposition to
transit hops ("Pop&Go"), see the NANOG 72 presentation and video (February
2018): https://nanog.org/events/nanog-72/content/1510/

On Fri, Nov 13, 2026 at 12:43 PM Robert Raszuk <robert@raszuk.net> wrote:

> Hi Andy & Colby,
>
> Ok this is useful to know. Frankly this should be copied and posted into
> the Operational section of the draft IMO.
>
> Now I see that it is actually this draft which also defines RSVP extension
> ... I missed this upon first reading !
>
> 5.4.1.  SSI-D
>
>    Bit Number TBD3: SRv6 Segment-list Imposition - Delegation (SSI-D)
>
>    Automatic Delegation:  The presence of this flag in the
>       LSP_REQUIRED_ATTRIBUTES object of a Path message indicates that
>       the ingress has requested automatic delegation of segment-list
>       imposition.  If a transit hop does not support this flag, it MUST
>       reject the Path message per the LSP_REQUIRED_ATTRIBUTES semantics
>       of [RFC5420].
>
>    Explicit Delegation:  The presence of this flag in the HOP_ATTRIBUTES
>       sub-object [RFC7570] of an ERO in the Path message indicates that
>       the hop identified by the preceding ERO sub-object has been picked
>       as an explicit delegation hop.  Processing follows the same
>       procedures as Section 9.4 of [RFC8577], adapted for SRv6.
>
> And in fact you have two flavors of it !
>
> As I skipped this I was under the impression that your idea of
> "delegation" is the same signalling wise with loose ERO objects. But it is
> obvious now that it is not.
>
> Now how do you signal in the RSVP-PATH msg the delegation points if not by
> loose ERO object ?
>
> You say that loose hops in ERO are locally at the (loose hop) expanded and
> then you say that this is independent of delegation.
>
> > The delegation hop builds its sub-segment-list locally from the route
> RSVP recorded,
>
> Well I guess not really. In the case where there is more then one
> delegation it only MUST to fill the SIDs/uSIDs to the next delegation point
> - otherwise there is still issue with SID depth.
>
> > A Delegation SID can be shared by all LSPs crossing the same segment
> between two delegation
> >  hops, whatever their egress, so the state is per segment rather than
> per LSP / TE-Tunnel.
>
> True but only between delegations. On the last delegation node where LSPs
> go to different egress nodes that no longer applies.
>
> So to summarize ... I think we reached the clarity of this proposal - well
> at least in my view I see how this is planned to work.
>
> I think even if each separate piece looks pretty straight forward putting
> this entire jigsaw puzzle together does look complex to me. And we only
> talked about primary paths. When you publish a draft on how fast
> connectivity restoration is supposed to work in this network at each hop
> with BW awareness it is going to be fun !
>
> Many thx,
> Robert
>
>
> On Fri, Nov 13, 2026 at 9:17 PM Andy Smith <andy@arrcus.com> wrote:
>
>> Hi Robert,
>>
>> It is your Option 2 in substance, with one difference: the "chunks" are
>> not sent from the source. RSVP collects them itself as the Resv travels
>> back upstream.
>>
>> The path is computed once, by the ingress or a PCE, and signaled end to
>> end. The Path message visits every hop in the ERO, including the hops
>> between delegation points, and every hop admits the reservation on its own
>> link. Delegation changes only which node imposes which SIDs in the data
>> plane; it changes neither what is signaled nor where bandwidth is admitted.
>> A delegation hop runs no CSPF and does not choose its own sub-path, so the
>> reverse-direction case in your Option 1 does not arise. (Loose hops in the
>> ERO are expanded as RFC 3209 describes, but that is a choice of the ingress
>> and independent of delegation.)
>>
>> The chunks come from the RRO. Each transit records its End.X SID in the
>> RRO of the Resv. A delegation hop takes the SIDs recorded downstream of it,
>> up to the next delegation hop, binds them to a Delegation SID it allocates
>> (End.B6.Encaps or End.B6.Encaps.Red, or their NEXT-CSID variants), and
>> records that Delegation SID in the RRO it sends upstream, with the
>> Delegation SID flag set. The ingress then imposes its own portion and the
>> Delegation SIDs, and omits the SIDs they cover. Section 7.2 of
>> draft-beeram-teas-rsvp-srv6-01 walks through a twelve-node example, and
>> Section 6.1 (step 4) gives the ingress rule.
>>
>> Delegation hops are picked either explicitly by the ingress, with the
>> SSI-D flag in the ERO's HOP_ATTRIBUTES, or automatically: each hop signals
>> its Effective SRv6 Segment-List Depth, the SRv6 counterpart of the RFC 8577
>> ETLD, and the decision is made during Resv processing from the recorded
>> SIDs and depths (Section 7.3 and Appendix A).
>>
>> On the SR architecture point: no SRH is carried in the control plane and
>> nothing is inserted. The delegation hop builds its sub-segment-list locally
>> from the route RSVP recorded, and End.B6.Encaps pushes a new outer IPv6
>> header exactly as RFC 8986 defines it. It is a binding SID in the RFC 8402
>> sense; only the source of its segment list is new.
>>
>> You are right that a delegation hop holds forwarding state for its
>> Delegation SID. Two things bound it. A Delegation SID can be shared by all
>> LSPs crossing the same segment between two delegation hops, whatever their
>> egress, so the state is per segment rather than per LSP / TE-Tunnel. And
>> with compressed SIDs (RFC 9800) delegation is needed far less often: a path
>> that fits in one container needs no SRH at all, and each further container
>> adds a single 16-byte entry. The other transit nodes hold no per-LSP /
>> TE-Tunnel forwarding state in any case.
>>
>> cheers
>>
>> Andy
>>
>>
>>
>> -
>> Andy Smith
>> Arrcus Distinguished Engineer
>> andy@arrcus.com
>> mobile +1 267 253 5697 <(267)%20253-5697>
>>
>>
>>
>>
>> On Nov 13, 2026, at 15:05, Robert Raszuk <robert@raszuk.net> wrote:
>>
>> Hi Colby,
>>
>> Yes I read that section but this only explains the mechanics of
>> delegation.
>>
>> I was looking more for explanation where the chunks of stack are carried
>> in the RSVP. I can envision two options (and reading the RFC and yr draft I
>> am not sure which one you are after):
>>
>> *Option 1) *
>>
>> So far and including the text from the subject draft (where suggestion to
>> use BSIDs is made) it looks to me like ingress RSVP is setting a path
>> between ingress and BSID1 (delegation 1) and BSID2 (delegation2) without
>> actually traversing (in the control plane) any hops between BSID1 and BSID2
>> or BSID2 and the Egress
>>
>> If you are suggesting that we now need to run three CSPFs to get from Src
>> to Dst then this is a very topology dependent TE. In fact if Delegation
>> node 1 runs his CSPF it may choose a path in the reverse direction (towards
>> the src) depending on the network topology and IGP configuration - say in
>> flat ISIS/OSPF domains end to end with links following physical
>> availability.
>>
>> *Option 2) *
>>
>> If however you have a way to still compute full path Src to Dst on
>> ingress, but to cut it in chunks due to hardware limitations and send those
>> chunks from Src to both delegation points that may not be a bad idea (just
>> from TE perspective). Still operational complexity grows, but with modern
>> tools maybe this would be something to accept. Note also that if you push
>> BSID and in the control plane carry SRH to insert when given BSID is
>> noticed that IMO would be an extension to SR architecture.
>>
>> KInd regards,
>> Robert
>>
>>
>> On Fri, Nov 13, 2026 at 7:20 PM Barth, Colby <jonathan.barth@hpe.com>
>> wrote:
>>
>>> https://www.rfc-editor.org/rfc/rfc8577.html#page-8 for a description of
>>> delegation.
>>>
>>> BW reservations continue to operate the same regardless if delegation is
>>> used or not because reservations are in the control-plane.
>>>
>>> -- Colby
>>>
>>>
>>> ------------------------------
>>> *From: * Robert Raszuk <robert@raszuk.net>
>>> *Date: *Friday, November 13, 2026 at 8:18 AM
>>> *To: *Barth, Colby <jonathan.barth@hpe.com>
>>> *Cc: *Andy Smith <andy@arrcus.com>; mpls <mpls@ietf.org>; TEAS WG <
>>> teas@ietf.org>; SPRING WG <spring@ietf.org>; Andrew Stone (Nokia) <
>>> andrew.stone@nokia.com>
>>> *Subject: *Re: [mpls] Re: [spring] Hop by Hop TE in SR networks (aka
>>> draft-beeram-mpls-rsvp-sr-01)
>>>
>>> Hi Colby,
>>>
>>> <cb> The delegation procedures described in the draft and in RFC8577 can
>>> be leverage to ensure the number of SIDs/uSIDs/Labels is kept to a
>>> minimum.  In the SRv6 case, you could avoid an SRH altogether even for very
>>> long paths.  This is one of the values of having a signaling plane.
>>>
>>>
>>> I do not see a clear description in the subject draft of how you perform
>>> end to end strict hop by hop BW admission control in the case of delegated
>>> SID impositions. Needless to say that delegated SID impositions in the end
>>> to end TE increases significantly operational complexity when it comes to
>>> troubleshooting this type of architecture.
>>>
>>> Yes, using binding SID is ok .. it has been part of SR for a long time.
>>> Of course this actually is now a data plane state per TE LSP/Path.
>>>
>>> But the question related to the draft if using such imposition
>>> delegations is about bw allocations. Also OAM tools now need to be aware of
>>> new signalling protocol too.
>>>
>>> Regards,
>>> Robert
>>>
>>>
>> [EXTERNAL]
>>
>>
>> _______________________________________________
> mpls mailing list -- mpls@ietf.org
> To unsubscribe send an email to mpls-leave@ietf.org
>