[mpls] MPLS-RT Review of draft-gandhi-mpls-ioam-sr-04

["Andrew G. Malis" <[e-mail]>]

I've been asked to provide a pre-adoption MPLS-RT review of
draft-gandhi-mpls-ioam-sr-04.

I have a major concern that I believe needs to be addressed, either before
or after WG adoption (I defer to the WG chairs to make this decision). My
personal preference is that it be addressed by the authors prior to
adoption, but if it occurs following adoption, I would like to see it
addressed before it gets much further in the WG process.

My concern is as follows:

In Section 6 and Figure 1, 0x0010b (2 decimal) is used for the first nibble
following the MPLS label stack in order to avoid ECMP. This intent is fine,
but there is an issue with choosing this particular value. The first nibble
following the label stack is often (as we know) interpreted as an IP
Version Number. According to
https://www.iana.org/assignments/version-numbers/version-numbers.xhtml ,
0x0010b (2 decimal) is currently unassigned, so it COULD be assigned by
IANA, creating a future conflict.

We could request IANA to assign IP Version number 2 for this purpose, but I
believe that would be a very difficult sell to both IANA and the IESG, as
there are only a small number of IP Version numbers available.

Instead, I would suggest either of the two following alternatives:

1. Use the MPLS ACH (RFC 5586), starting with 0x0001b, and alter the packet
format in Figure 1 of this draft accordingly so that it follows the ACH's
general format but also includes the necessary fields for the draft's
purpose.

2. Use one of the IANA reserved IP version numbers instead of 0x0010b. I
would recommend 15 (0x1111b). There is reasonable certainty that this would
never actually ever be assigned by IANA.

The first alternative is my personal preference, but I would be OK with the
second as well.

Other comments:

Other than this issue, I found the draft to be well-written and easy to
follow, and generally ready for WG adoption.

Cheers,
Andy