Thank you for what looks like a thorough effort to address my concerns.
I will look at the draft this week, and let you know what I conclude.
From your summary, I expect you have addressed the points I raised.
Yours,
Joel
On 11/2/2026 5:13 AM, Satoru Matsushima wrote:
Hi Joel,
Thank you for the early Rtgdir review. We took your main points as
clarity and scoping issues: what this document standardizes was not
stated clearly enough, and the relationship to the BGP work was left
ambiguous.
We have now posted the -03 revision, which tries to address them:
https://datatracker.ietf.org/doc/draft-ietf-dmm-mup-architecture-03/
On the Standards Track question: this document standardizes the
semantics of MUP segments and routes and the behavior of MUP PEs and
the MUP controller; it does not define wire encodings. That follows
the model of architecture specifications such as the Segment Routing
architecture [RFC 8402]. Keeping this document on the Standards Track
also avoids a downref problem for draft-ietf-bess-mup-safi, which
normatively references this architecture.
We agree that the -02 text made this harder to see: -02 did not cite
draft-ietf-bess-mup-safi at all, and described the BGP extensions "as
an example", which understandably read as if this document were
defining them. The -03 revision states the scope and the division of
labour explicitly. The BGP-MUP SAFI, NLRI, and MUP Extended Community
encodings and IANA assignments are specified in
draft-ietf-bess-mup-safi; this document defines only the semantics
(what the community identifies and what metadata it carries).
The main changes in -03 are:
* Scope and deployment model are stated up front. The Abstract and
the Introduction now say explicitly what the document defines (the
architecture principles, the semantics of segments and routes, how
the routes are distributed and resolved, and the behaviour
required of a MUP PE and of a MUP Controller), that those
definitions are normative, and that the document defines no wire
encoding and makes no IANA request. The Introduction and Section 3
also give the deployment model before the procedures: a MUP
network is typically deployed within a single administrative
domain, with the discovery and session-transformed routes
distributed over iBGP between the MUP PEs and the MUP-C, and
Section 3 explains why BGP is recommended to carry the routing
information.
* "MUP segment" is defined as a logical abstraction, not a portion
of the network. Section 4 now opens with that definition and
states that a segment is a symmetric abstraction: a segment is
described, discovered and used in the same way at either end of a
session, and the document does not characterise a segment by
whether it sits on an access side or on a non-access side; such
roles are decided by the deployment. The two segment types are
distinguished only by whether reaching the segment requires
interworking with a user plane protocol of a mobile service
architecture. Section 4.1 clarifies that, for SRv6, a SID is not
the segment itself: it identifies a forwarding behaviour executed
at the MUP PE that hosts the segment.
* The session information that the third principle of Section 3
talks about is now defined in the Terminology (with the F-TEID as
the 3GPP example of a tunnel endpoint identifier), and the
principle itself is written out as a full statement starting from
that concrete picture. A short statement of the BGP division of
labour is repeated in Section 3: the BGP extensions that realise
the routes, including the BGP-MUP SAFI and the MUP Extended
Community, are specified in draft-ietf-bess-mup-safi.
* Section 5 defines the semantics of the MUP Extended Community
only; its encoding and IANA allocations are in
draft-ietf-bess-mup-safi. For the ISD route, Section 5.2 states
that the community may be attached, and that an ISD route without
one is treated as belonging to the default Interwork Segment.
* Sections 6.1 and 6.2 describe the ST1 and ST2 routes in parallel
terms. An ST1 route carries the prefix of a UE or MN; an ST2 route
is keyed by the tunnel endpoint identifier of the session (the
matching endpoint), and the destination endpoint is resolved
through the segment discovery routes, as illustrated by the
home-routed roaming example in Section 7.5.
We would welcome your view on whether this addresses the scoping
questions you raised, and on any remaining clarity issues.
Thanks again for the review.
Best regards,
--satoru
On Sun, Aug 16, 2026 at 10:25 AM Joel Halpern via Datatracker
<[email protected]> wrote:
Document: draft-ietf-dmm-mup-architecture
Title: Mobile User Plane Architecture for Distributed Mobility
Management
Reviewer: Joel Halpern
Review result: Not Ready
Hello
I have been selected to do a routing directorate “early” review of
this draft.
https://datatracker.ietf.org/doc/draft-ietf-dmm-mup-architecture/
The routing directorate will, on request from the working group
chair, perform
an “early” review of a draft before it is submitted for
publication to the
IESG. The early review can be performed at any time during the
draft’s lifetime
as a working group document. The purpose of the early review
depends on the
stage that the document has reached.
This review is provided with the understanding that the draft is
approaching
completion.
For more information about the Routing Directorate, please see
https://wiki.ietf.org/en/group/rtg/RtgDir
Document: draft-ietf-dmm-mup-architecture-02.txt
Reviewer: Joel Halpern
Review Date: 15-Aug-2026
Intended Status: Proposed Standard
Summary:
Choose from this list...
I have significant concerns about this document. It needs more
work before
being submitted to the IESG. I believe most of these issues are
matters of
explanatory text and grammar, not issues of technical flaw. But it
is hard to
be certain. There are also issues of appropriateness of scoping
as it seems to
define BGGP procedures.
Comments:
The document describes an alternative architecture for a mobility
infrastructure. It describes it as being related to the 3GPP
architecture, and
attempts to use 3GPP terms appropriately. However, the document
is quite hard
to read and understand.
Major Issues:
It is unclear to this reviewer if this draft is actu
Give as much context information as possible with your comments
(e.g., section
numbers, paragraph counts).ally informational or standards track.
It describes
an architecture. It describes that there need to be protocol
extensions to
support that architecture. Presumably for reasons of scoping and
appropriate
responsibility, this document does not standardize protocol
extensions. So
what is being standardized? The requirements are stated to be
taken from
another document.
It is also odd, in a document that is not defining protocol
extensions, to
provide example extensions. It seems to invite readers to
treat those as
extensions to be used, even though they are not formally
adopted by this
document. This becomes even more confusing when the document
proceeds to
state that it defines a MUP extended community. There are
very specific
BGP rules limiting how extended communities can be defined,
and as far as I
know it needs to be done by the IDR working group.
Apparently, the architecture intends to require the use of
BGP. Such a
requirement is understandable. But it was not stated before
the text
suddenly introduced reliance on extended communities. And
then in section
6 introduces further BGP dependence. I suspect that the
assumption is that
the PE devices correspond to the AS boundary devices, and this
is all
assumed to be used in iBGP. But I can't find the text that
explains that
before it is relied upon.
As a reader, I am quite confused as to what constitutes a MUP
segment.
From some of the diagrams, it seems to be a portion of the
network,
delimited somehow. On the other hand, there is text that an
SRv6 SID can
correspond to a MUP segment. An SRv6 SID identifies a node,
not a portion
of the network. (Yes, this is a general issue with segment
routing where
we often treat the node as if it represents the path portion that
terminates there. Within an SR SID list of any type, this can
usually be
understood. The usage here is not in the context of a SID
list, so I am
unsure what is meant.) This makes review of the subsequent
discussion of
discovery somewhat more difficult.
Minor Issues:
In section 3 on architectural principles, there appear to be
grammatical
problems in the core statements of the principles. This is
most easily
seen in the third principle where the text starts with the
apparent
sentence "The third principle is that transforming session
information to
routing information." That is neither a sentence nor a principle.
I found the explanation in section 4 of the difference between
a direct
segment and an interworking segment very hard to parse. I am
not confident
enough of my understanding to suggest specific new text. I
strongly urge
that it be revised for English clarity.
In section 6, section 6.1 states that type 1 information is
for carrying
addresses of nodes to be reached over the access network. I
would then
expect that type 2 information is for carrying information abut
destiantions / prefixes reach over the non-access network. As
far as I can
tell the text does not say that, but leaves the reader
guessing as to what
type 2 information is for.
Nits:
_______________________________________________
dmm mailing list -- [email protected]
To unsubscribe send an email to [email protected]