Re: [Isis-wg] [Bier] WGLC: draft-ietf-bier-isis-extensions-04

Toerless Eckert <> Sat, 22 July 2017 13:34 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id 81536129AD1; Sat, 22 Jul 2017 06:34:27 -0700 (PDT)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -4.2
X-Spam-Status: No, score=-4.2 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, HEADER_FROM_DIFFERENT_DOMAINS=0.001, RCVD_IN_DNSWL_MED=-2.3, RP_MATCHES_RCVD=-0.001] autolearn=ham autolearn_force=no
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id uRJRE9Mq0B_G; Sat, 22 Jul 2017 06:34:24 -0700 (PDT)
Received: from ( [IPv6:2001:638:a000:4134::ffff:40]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by (Postfix) with ESMTPS id DC445131A78; Sat, 22 Jul 2017 06:34:23 -0700 (PDT)
Received: from ( []) by (Postfix) with ESMTP id 6A25758C4B0; Sat, 22 Jul 2017 15:34:19 +0200 (CEST)
Received: by (Postfix, from userid 10463) id 53108B0C66E; Sat, 22 Jul 2017 15:34:19 +0200 (CEST)
Date: Sat, 22 Jul 2017 15:34:19 +0200
From: Toerless Eckert <>
To: "Les Ginsberg (ginsberg)" <>
Cc: Tony Przygienda <>, "Hannes Gredler (" <>, Greg Shepherd <>, "" <>, " list" <>, Christian Hopps <>
Message-ID: <>
References: <> <> <>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <>
User-Agent: Mutt/1.5.21 (2010-09-15)
Archived-At: <>
Subject: Re: [Isis-wg] [Bier] WGLC: draft-ietf-bier-isis-extensions-04
X-Mailman-Version: 2.1.22
Precedence: list
List-Id: IETF IS-IS working group <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Sat, 22 Jul 2017 13:34:27 -0000

Thanks Les

When searching various terms in the doc to figure out what happens
i am not sure why i missed this one.

But: IMHO, RFCs can not only be the minimum number of words to get a
running implementation. It also needs to specify what this implementation
intends to achieve. Otherwise its not possible to do a useful review:
The reviewer can to verify whether the spec will achieve what it claims
to achieve is there no definitionn of what it claims to achieve.

If i understand ISIS correctly, my reverse engineering of the intent is:

- BIER TLVs stay within single ISIS areas. BFIR and BFER must therefore be
  in the same ISIS area: There is no inter-area BIER traffic possible
  with this specification. This is also true for ISIS area 0.

- The same BIER sub-domain identifiers can be re-used
  across different ISIS areas without any current impact. If these BFR-IDs
  are non-overlapping, then this would allow in the future to create a single
  cross ISIS area BIER sub-domain by leaking TLVs for such a BIER sub-domain
  across ISIS levels. Leakage is outside the scope of this specificication.

I actually even would like to do the following:

- If BIER sub-domains are made to span multiple ISIS areas and BFR-ids assignemtns
  are made such that all BFR-ids with the same SI are in the same ISIS ara,
  then it should be in the future reasonably easy to create inter-area BIER
  not by leaking of the BIER TLV but by having BFIR MPLS unicastBIER packets
  for different SIs to an appropriate L2L1 BFIR that is part of the destination area/SI.
  (if you would use SI number that are the same as ISIS area numbers then
   you could probably do this without any new signaling. Not quite sure if
   you can today easily find L1L2 border router for another area via existing

  Alas, this idea will probably be killed because of the BIER architecture
  intent not to engineer SI assignments in geographical fashions to
  minimize traffic duplication in the presence of multiple SIs.

On Sat, Jul 22, 2017 at 06:03:53AM +0000, Les Ginsberg (ginsberg) wrote:
> Tony/Toerless ???
> There is an explicit statement as to scope:
> <snip>
> Section 4.2
> ???
>    o  BIER sub-TLVs MUST NOT be included when a prefix reachability
>       advertisement is leaked between levels.
> <end snip>
> Tony seems to have forgotten that we had a discussion about how BIER might be supported across areas and the conclusion was we did not know how to do that yet.
> (Sorry Tony)
> Note this is ???consistent??? with Section 2.5<> - which limits the flooding scope of BIER information to a single area unless it can be validated that the best path to the prefix with BIER info can be validated to be to a router which itself advertised the BIER info. This is not something IS-IS can do since a single IS-IS instance only supports one area and therefore does not have the Level-1 advertisements of the originating router when that router is in another area.
> A few more responses inline.
> From: BIER [] On Behalf Of Tony Przygienda
> Sent: Friday, July 21, 2017 5:17 AM
> To: Toerless Eckert
> Cc: Hannes Gredler (; Greg Shepherd;; list; Christian Hopps
> Subject: Re: [Bier] WGLC: draft-ietf-bier-isis-extensions-04
> Terminology is a bit nits  IMO since the doc is reading clear enough for someone who read BIER & ISIS. I can reread it or Les can comment whether we should tighten glossary ...
> With the scope I agree, that got lost and the doc should be possibly rev'ed before closing LC. Yes, we flood AD wide was the agreement but something mentioning that this could change in the future is good so we are forced to give it some thought how that would transition ...
> Thinking further though, in ISIS we have a clean document really. The  BIER sub-TLVs go into well defined TLVs in terms of flooding scope. Normal L1-L2 redistribution can be used to get the info to all needed places AFAIS. So maybe nothing needs to be written. I wait for Les to chime in.
> OSPF I would have to look @ scopes again & think whether we need to write something or maybe Peter can comment ...
> --- tony
> On Fri, Jul 21, 2017 at 8:27 AM, Toerless Eckert <<>> wrote:
> Sorry, past the two weeks, but hopefully  benign textual comments:
> We tried to find an explicit statement about the scope of BIER TLVs - eg:
> are they meant to stay within an area, have some redistribution across
> areas/levels or not.
> Tony said WG agreement was to have these TLV be flooded across the whole
> ISIS domain for now (this draft). So an explicit statement to that effect would
> be great (All BIER sub-domains TLVs are flooded across all ISIS areas/levels,                     so they span the whole ISIS domain).
> Also, if future work may/should could improve on that maybe some sentence
> about that (i guess one could just have ISIS intra-area BIER sub-domains ?).
> Also: Do a check about possible ambiguity of any generic terms like                               sub-domain, level, area, topology so that reader that don't know the terminology ofall protocols (ISIS, BIER) by heart can easily know which protocol is referred to.
> [Les:] There is no mention of ???level??? in the document.
> The use of ???sub-domain??? is clearly always associated with ???BIER???.
> ???topology??? is always used as an RFC 5120 topology ??? therefore clearly an IS-IS topology.
> There is only one use of the term ???area??? (in Section 5.1). That text might deserve a bit of clarification given this might be either a Level 1 area or the Level2 sub-domain. I???ll take a pass at it.
> (BTW ??? I am talking about IS-IS area/L2sub-domain Toerless. ???)
> I don???t see that any other clarification is needed ??? but Toerless ??? if you can point to any specific sentences/paragraphs which you find confusing - I???ll take a second look.
>    Les
> I guess there are no BIER level, area or topologies, but still makes reading easier if the
> doc would say "ISIS level", "ISIS area", or at least have them in the
> Terminology section. And probably in terminology say "domain -> in the context
> of this document the BIER domain which is also the same as the ISIS domain"
> (which i hope is the correct statement, see above).
> Cheers
>     Toerless
> _______________________________________________
> BIER mailing list
> --
> We???ve heard that a million monkeys at a million keyboards could produce the complete works of Shakespeare; now, thanks to the Internet, we know that is not true.
> ???Robert Wilensky