Hi Reshma,

Pls find inline replies

Brgds,

 

From: Das, Reshma <[email protected]> 
Sent: Tuesday, March 3, 2026 12:21 AM
To: Jeffrey Haas <[email protected]>; [email protected]; Stephane 
Litkowski (slitkows) <[email protected]>; [email protected]; [email protected]
Cc: BESS <[email protected]>; idr-chairs <[email protected]>
Subject: [bess] Re: draft-ietf-bess-ebgp-dmz - preparation for working group 
last call

 

Hi Jeff,

 

Regarding the IPR call, I'm not aware of any IPR related to this draft.

 

 

 

Reviewing the document I have couple of comments:

 

1.       In Section 4.2:

“for consistency reason, all BGP paths of a prefix MUST use a contributing 
bandwidth based on the same source (e.g.: all paths are using contributing 
bandwidth from the same function)."

Should this be a MUST? In the context of EIBGP peering, the source of the LBW 
value may differ depending on the type of session. For single‑hop EBGP peers, 
the contributing bandwidth may correspond to the directly connected interface, 
whereas for multi‑hop IBGP peers, the contributing bandwidth would be derived 
from the LBW received from the upstream node. In both cases, the attached LBW 
values can be used for WECMP calculations.
However, enforcing that all BGP paths of a prefix MUST derive contributing 
bandwidth from the same source appears unnecessarily restrictive.

[SLI] I agree, that we should not be so restrictive, we can probably relax 
that. Would a SHOULD be ok ? I could put a note for the EIBGP case.

2.       In section 4.3:

a.       Example:

“    P/m

      Path1:

        NH=R1, Link-BW extended-community: 1000Mbps,

        local-link-bandwidth: 2000Mbps, contributing-bandwidth: 2000

        best, multipath

 

      Path2:

        NH=R2, Link-BW extended-community: 2000Mbps,

        local-link-bandwidth: 2000Mbps, contributing-bandwidth: 4000

        multipath

 

      Path3:

        NH=R3, Link-BW extended-community: 3000Mbps,

        multipath"

This example is a bit confusing. Could you elaborate on how the contributing 
bandwidth is calculated? Although Path3 has received a Link Bandwidth (LBW) 
extended community, its contributing bandwidth is shown as 0. It appears that 
the contributing bandwidth for all three paths is calculated differently, which 
contradicts the second MUST statement in Section 4.2.

[SLI] Yes, I think contributing BW have been set randomly here I think which 
doesn’t help and doesn’t match with 4.2 reqs. Good catch. I’ll fix this to make 
it more readable.

b. Further, according to Section 4.2, since Path3 does not have any 
contributing bandwidth, WECMP cannot be done on this prefix. 

Although the DUT does not perform WECMP, it is still aggregates and advertises 
the LBW value to the peer, which can be misleading

In my opinion the cumulative or aggregated LBW should be the sum of the LBW 
value that is used locally for WECMP. This provides the peer with an accurate 
representation of the bandwidth available downstream.

[SLI] In fact it’s may not always be the case. Cumulating BW across layers is 
causing problems of churn. If a remote link (far away) flaps, it would be 
reflected everywhere in the fabric. But it’s not always suitable. Some people 
prefer when cumulating hiding what’s happening upstream and cumulate only based 
on local link BW available even if they use something different for WECMP.

 

Thanks & Regards,

Reshma Das 

 

Get Outlook for Mac  <https://aka.ms/GetOutlookForMac> 

From: Jeffrey Haas <[email protected] <mailto:[email protected]> >
Date: Monday, February 23, 2026 at 11:19 AM
To: [email protected] 
<mailto:[email protected]>  <[email protected] 
<mailto:[email protected]> >, Das, Reshma <[email protected] 
<mailto:[email protected]> >, Stephane Litkowski (slitkows) 
<[email protected] <mailto:[email protected]> >, [email protected] 
<mailto:[email protected]>  <[email protected] <mailto:[email protected]> >, 
[email protected] <mailto:[email protected]>  <[email protected] 
<mailto:[email protected]> >
Cc: BESS <[email protected] <mailto:[email protected]> >, idr-chairs 
<[email protected] <mailto:[email protected]> >
Subject: Re: draft-ietf-bess-ebgp-dmz - preparation for working group last call

Note that IPR attestations are still pending for Reshma, Stephane, and Arie.  
Ajay, Akshay has made one note related to Arista, but please confirm your own 
understanding.

 

Authors, the following still needs to be addressed:

"[NOTE: IS THAT TRUE IN OUR CASE ?]"

 

:-)

 

-- Jeff

 

On Feb 13, 2026, at 18:38, Jeffrey Haas <[email protected] <mailto:[email protected]> 
> wrote:

 

https://datatracker.ietf.org/doc/draft-ietf-bess-ebgp-dmz/shepherdwriteup/ 
<https://urldefense.com/v3/__https:/datatracker.ietf.org/doc/draft-ietf-bess-ebgp-dmz/shepherdwriteup/__;!!NEt6yMaO-gk!F0JKYQ9w22XTRB5wHBiBLajKvCfqbO7QAryzC1Vqk8dy5tdx4HMJ4vFyZAfWj4b83A0KEg0idUoE$>
 

In November 2025, the IDR chairs corresponded with the BESS chairs about the 
disposition of the DMZ draft.  As you know from the long history of this draft, 
the link-bandwidth one, and the work that came to be part of RFC 4360-bis, 
we've discussed leaving the DMZ draft in BESS and running concurrent WGLC's on 
this draft along with RFC 4360-bis.

As part of this concurrent WGLC, I've completed an initial shepherd report for 
the DMZ draft.  Please review its contents for correctness.  A few items have 
been identified that will be needing to be addressed prior to starting the last 
call.  They are:

- Author count > 5. I've not recommended changes since I've not been a 
participant in those discussions.  The chairs have been urged to discuss what 
to do about it.

- Proposed status.  Ketan, as part of the link-bandwidth draft AD review, 
suggested that the DMZ draft should likely be informational.

- IPR declarations were partially gathered during document adoption.  While I 
have not been deeply thorough in my review of the IETF mailarchive for search 
terms covering this draft, my cursory review suggests we're missing IPR 
declarations for some of the authors.  If you're one of the authors identified 
in the shepherd writeup, please either make an appropriate declaration or 
perhaps provide a pointer in the archive for a prior one.

- The authors have left a NOTE in the draft suggesting a bit of doubt about the 
disposition of one of the points regarding zero link-bandwidth.  Discussion 
about this was part of our last minute link-bandwidth document changes in IDR - 
so this isn't surprising.  The authors are requested to please address this 
lingering point in the draft.

- While there are no normative procedures specified in the draft, the 
procedures do raise questions about supporting implementations.  I won't speak 
for the chairs in terms of expectations of formal implementation reports for 
their working group.  If such an implementation report exists for one of your 
implementations, please disclose it so it can be added to the report.

- The document may benefit from a routing-directorate review. I was unable to 
identify a prior review for this document from the mail archives.

In general, the document is in good shape.

I'll be working on a similar shepherd report for RFC 4360-bis in IDR.  IDR does 
have a requirement for at least two implementations in order to progress the 
document.  Since part of the motivation for RFC 4360-bis was addressing the 
transitivity considerations needed for this document, I find it highly likely 
that there are implementations from authors participating in the DMZ draft for 
the RFC 4360-bis work.  Please contribute to IDR's implementation report so 
that we can send both documents to the IESG at the same time when they've 
completed their reviews.

-- Jeff (as document shepherd)

 

_______________________________________________
BESS mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to