Re: [mpls] [Technical Errata Reported] RFC4928 (5396)

Copying George on his correct address.

As I remember the point being made was that IP ECMP would in that case 
be performed on the Ethernet payload which might be constant. It might 
also of course be effectively random.

We spent a lot of time on that text so it is worth some careful thought 
before taking any action on the errata.

- Stewart

On 29/06/2018 15:32, BRUNGARD, DEBORAH A wrote:
> MPLS WG,
>
> We have this errata on RFC4928 - any thoughts?
>
> Deborah
>
>
> -----Original Message-----
> From: RFC Errata System <rfc-editor@rfc-editor.org>
> Sent: Monday, June 18, 2018 5:37 AM
> To: stbryant@cisco.com; swallow@cisco.com; loa@pi.se; BRUNGARD, DEBORAH A <db3546@att.com>; aretana.ietf@gmail.com; martin.vigoureux@nokia.com; loa@pi.nu; n.leymann@telekom.de
> Cc: jsharma@ciena.com; mpls@ietf.org; rfc-editor@rfc-editor.org
> Subject: [Technical Errata Reported] RFC4928 (5396)
>
> The following errata report has been submitted for RFC4928,
> "Avoiding Equal Cost Multipath Treatment in MPLS Networks".
>
> --------------------------------------
> You may review the report below and at:
> https://urldefense.proofpoint.com/v2/url?u=http-3A__www.rfc-2Deditor.org_errata_eid5396&d=DwIBaQ&c=LFYZ-o9_HUMeMTSQicvjIg&r=6UhGpW9lwi9dM7jYlxXD8w&m=zz6ovHLVAtV_M2JDhQbYVwIhYVehe5aI6IVO8gDY6fU&s=ZeHGqLSL8ZArKF83wT85HuuT78PHErIz0Lfn0bk8280&e=
>
> --------------------------------------
> Type: Technical
> Reported by: Jitendra Kumar Sharma <jsharma@ciena.com>
>
> Section: Section 2
>
> Original Text
> -------------
>     A less obvious case is when the packets of a given flow happen to
>     have constant values in the fields upon which IP ECMP would be
>     performed.  For example, if an Ethernet frame immediately follows the
>     label and the LSR does ECMP on IPv4, but does not do ECMP on IPv6,
>     then either the first nibble will be 0x4, or it will be something
>     else.  If the nibble is not 0x4 then no IP ECMP is performed, but
>     Label ECMP may be performed.  If it is 0x4, then the constant values
>     of the MAC addresses overlay the fields that would have been occupied
>     by the source and destination addresses of an IP header.  In this
>     case, the input to the ECMP algorithm would be a constant value and
>     thus the algorithm would always return the same result.
>
> Corrected Text
> --------------
> <This paragraph should be removed>
>
> Notes
> -----
> The example stated here seems incorrect. It talks about an L2VPN case where Ethernet frame starts immediately after the last label in the stack. But had it been an IP packet instead, the same initial 12 bytes, which is the place for MAC addresses in an Ethernet Frame, would not be the place of IP addresses, as IP addresses are placed at the end of 20-byte IP header (not start). Hence it would still be subjected to ECMP if precautions (as recommended in this RFC) are not been followed.
>
> Instructions:
> -------------
> This erratum is currently posted as "Reported". If necessary, please
> use "Reply All" to discuss whether it should be verified or
> rejected. When a decision is reached, the verifying party
> can log in to change the status and edit the report, if necessary.
>
> --------------------------------------
> RFC4928 (draft-ietf-mpls-ecmp-bcp-03)
> --------------------------------------
> Title               : Avoiding Equal Cost Multipath Treatment in MPLS Networks
> Publication Date    : June 2007
> Author(s)           : G. Swallow, S. Bryant, L. Andersson
> Category            : BEST CURRENT PRACTICE
> Source              : Multiprotocol Label Switching
> Area                : Routing
> Stream              : IETF
> Verifying Party     : IESG
> _______________________________________________
> mpls mailing list
> mpls@ietf.org
> https://www.ietf.org/mailman/listinfo/mpls