[pim] Ketan Talaulikar's Discuss on draft-ietf-pim-p2mp-poli cy-ping-20: (with DISCUSS and COMMENT)

Ketan Talaulikar has entered the following ballot position for
draft-ietf-pim-p2mp-policy-ping-20: Discuss

When responding, please keep the subject line intact and reply to all
email addresses included in the To and CC lines. (Feel free to cut this
introductory paragraph, however.)


Please refer to https://www.ietf.org/about/groups/iesg/statements/handling-ballot-positions/ 
for more information about how to handle DISCUSS and COMMENT positions.


The document, along with other ballot positions, can be found here:
https://datatracker.ietf.org/doc/draft-ietf-pim-p2mp-policy-ping/



----------------------------------------------------------------------
DISCUSS:
----------------------------------------------------------------------

Thanks to the authors and the WG for their work on this document.

I have a few points that I would like to discuss with the authors and the WG.

This is an updated ballot for the v20 posted by the authors. Note that the
numbering of open points in the DISCUSS position is being retained as in the
original ballot for ease of tracking.

discuss#2 This is related to a point of discussion that I've also raised on
the p2mp policy document. It arises from the lack of clarity on whether the
SR P2MP Policy construct is instantiated on the root node or not. There is
text in section 3.1.1 which seems to leave this critical aspect to
implementations and will result in interoperability problems. The base spec
needs to be very clear on this point and then this document updated to reflect
it. I believe having the construct instantiated in the root will greatly
benefit and simplify OAM operations. And things would then become very
similar to RSVP-TE P2MP trees?

< update for v20 > Note that the base document now clarifies that the P2MP
Policy construct is instantiated on the root. This document is not in sync with
the base - this point remains open for discussion.

Quoting some text from section 3.1.1 that is problematic:

"Only the Root node, where the P2MP SR Policy is programmed, may have visibility
 into the leaf nodes."

"In the case of P2MP SR Policies, the Root of the tree may have full visibility
into the egress nodes if the P2MP SR Policy is PCC-initiated. If the P2MP SR
Policy is PCE-initiated, the Root may or may not have visibility into the egress
 nodes, as this depends on the specific implementation and configuration of the
PCE. "

Further dependencies like the following seem unnecessary:

"Based on this, a P2MP SR Policy SHOULD follow the recommendations in Section
4.3.1 of [RFC6425], depending on the level of visibility the Root has into the
 egress nodes. For example, in a PCC-initiated P2MP SR Policy, the Root can
learn egress node identities through Next-Generation MVPN procedures and BGP,
as described in [RFC6514]. In contrast, for a PCE-initiated P2MP SR Policy,
the PCE may not provide the egress node information to the Root, making this
process optional and implementation-specific."

The lack of clarity hurts interoperability and would affect operations in a
multi-vendor network. I would like to discuss why all of this cannot be
simplified by ensuring that the SR P2MP Policy construct is instantiated on
the root node.

discuss#3 My understanding is that the P2MP MPLS trees that are setup by MLDP
or RSVP-TE are hop by hop in nature. While in this case, the packet can travel
multiple hops from one node to the next intermediate node using that next
intermediate node's Prefix SID. In this case, how would operation like
traceroute (or even errors in the case of ping) work when the packet is
exposed at a node that is doing unicast forwarding and has no replication
segment context for that specific P2MP Tree? Now, section 3.1.3 is covering
this, but talking about it as "unicast SR domains" is very misleading since
there is only an SR domain and it is just that the specific P2MP tree context
is not required to be instantiated on a transit node. Does this mean that this
mechanism works only when the P2MP Tree is setup up hop-by-hop? If so, this
should be clearly called out as a caveat upfront and the text in 3.1.3 updated
appropriately.

< update for v20 > This point remains open for discussion.


----------------------------------------------------------------------
COMMENT:
----------------------------------------------------------------------

Please also find below some comments provided inline in the idnits format of
the v18 of this document. On all editorial and minor comments, I will leave it
to the authors discretion. On the major ones, I would appreciate responses and
clarifications.

Note: the comments are updated by removing those that were addressed in v20

Please look for <EoRv18> at the end of this review and if it is not there, then
likely the email has gotten truncated by your client (please refer to the
mailing list in that case).

142     3.  Motivation

144        A P2MP Policy and its corresponding Replication Segments are
145        typically provisioned via a centralized controller or configured
146        using NETCONF/YANG or CLI.  The root and the leaves are discovered in

<minor> Perhaps you mean that the network topology that includes the root and
leaves is discovered?

319        *  Address Length: (1 octet) specifying the length of the Root
320           Address in octets (4 octets for IPv4, 16 octets for IPv6).

<major> reserved is missing; also I believe it MBZ

< update for v20 > Still needs a change ... s/and it should be/and - this is to
make the "MUST" apply to both sender and receiver.

<EoRv18>



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