Minutes IETF100: ntp
minutes-100-ntp-01

================================
NTP/TICTOC Joint Meeting
November 13th, 2017, 13:30-15:30
================================

NTP WG Chairs: Karen O'Donoghue, Dieter Sibold
TICTOC WG Chairs: Karen O'Donoghue, Yaakov Stein (absent)
Note taker: Tal Mizrahi
Jabber: Kyle Rose

===========
NTP Session
===========

CHAIR SLIDES
------------
Presenter: Karen O'Donoghue
Slides:
https://datatracker.ietf.org/meeting/100/materials/slides-100-ntp-agenda-and-note-well/

Summary:
-   Karen presented the new note well.
-   Control Messages Protocol for Use with Network Time Protocol Version 4
draft:
    -   There are open questions how operators are currently using mode 6
    -   Would anyone be willing to help out Brian with the mode 6 draft?
    -   There are question about version numbering
    -   Question about the necessity to add additional commands to the mode 6
    messages -   Robert Nagy: I volunteer.
-   BCP
    -   is in the shepherding write-up stage
-   NTP extension field draft:
    -   We will be putting together a consensus call about this issue after
    further discussion with the authors. -   Harlan: the authors got together a
    couple of times. -   Karen: the authors still need to agree. There seem to
    be two alternatives, and we want to be able to present the options to the
    working group.

MESSAGE AUTHENTICATION CODE FOR THE NETWORK TIME PROTOCOL
---------------------------------------------------------
No slides were presented.

Discussion:
-   Aanchal: will add the security considerations section in the next version
of the draft. -   Tal: another comment was regarding whether this MAC could be
included in an extension field. -   Aanchal: that was already discussed on the
mailing list. -   Tal: another issue was there should be a subsection about
interoperability with previous implementations.

NETWORK TIME SECURITY (NTS)
---------------------------
Presenter: Dieter Sibold
https://datatracker.ietf.org/meeting/100/materials/slides-100-ntp-network-time-security-for-the-network-time-protocol/

Discussion:
- Kyle Rose: what is the plan for modes other than client/server?
- Dieter: In the last interim meeting we talked about moving the other modes to
another draft in the future. - Daniel Franke: I expect to write an experimental
draft about these other modes. - Karen: is there a time frame for moving
forward? - Dieter: not at this time. If we could have an interim meeting, we
can talk about a plan in the interim. - Karen: it could be great if we could
have a hackathon effort around NTS in IETF 101 in London.

A YANG DATA MODEL FOR NTP
-------------------------
Presenter: Anil Kumar (remote)
https://datatracker.ietf.org/meeting/100/materials/slides-100-ntp-a-yang-data-model-for-ntp/

Discussion:
- Harlan: I am assuming there should be compatibility between the YANG model
interface, the mode 6 interface, and the SNMP MIB interface. - Greg: why does
this need to be compatible? - Harlan: functionally equivalent, not necessarily
compatible. - Robert Nagy: I agree that it needs to be equivalent, but not
compatible. When can we finally get this released? We don't want to continue to
update this. - Greg: we want to standardize the base, and then we will be able
to add extensions in the future. - Karen: What is your view on the
compatibility with mode 6? - Anil: we did not try to have compatibility with
mode 6. We believe it is functionally equivalent to the MIB. - Karen: Harlan,
do you believe compatibility with mode 6 is really required? It seems that it
is not a design objective. - Harlan: we want to make it easier on the user by
having the three mechanisms as similar as possible. - Karen: it is not
reasonable to require these three to be compatible, as mode 6 and the MIB go
back a long time. - Harlan: let’s try to reach functional equivalence. - Anil:
I will look into it with the other authors.

GUIDELINES FOR DEFINING PACKET TIMESTAMP FORMATS
------------------------------------------------
Presenter: Tal Mizrahi
https://datatracker.ietf.org/meeting/100/materials/slides-100-ntp-packet-timestamp-formats/

Discussion:
- Greg: I believe this work is valuable. I would like to suggest clear
terminology that distinguishes resolution from accuracy. Also suggest to
discuss how to migrate to a higher accuracy. Another aspect that should be
discussed is how the control plane communicates the timestamp format. In future
we might need more resolution as provided by the truncated PTP timestamps. We
might add a foresight discussion how to migrate to higher resolution
timestamps. - Karen: what is the time frame for the next steps? - Tal: we hope
to complete the open issues by the next IETF meeting and be ready for WG last
call.

NTP INTERLEAVED MODES
---------------------
Presenter: Aanchal Malhotra
https://datatracker.ietf.org/meeting/100/materials/slides-100-ntp-ntp-interleaved-modes/

Discussion:
- Greg: when you say “improve the accuracy”, what do you mean?
- Aanchal: the accuracy of the timestamp captured by the client or server.
- Greg: when you timestamp in software, there may be nondeterministic error.
The further you move from the demon, the less accurate it is. - Rich Salz: the
time stamping is already in the kernel space. The further you move from the
demon, the more accurate. - Tal: this is very similar to the PTP Follow_up
messages. PTP uses Sequence ID to match the timestamp and the corresponding
event message. Is there something similar in the proposed mechanism? - Aanchal:
yes. The receiver verifies the origin timestamp and the transmit timestamp. -
Tal: from a security perspective, it looks like if a MAC verification fails,
you should also ignore the previous and next packet, as they are bound by the
interleaved timestamps. Please consider this and add text about it to the
draft. - Harlan: there was a discussion about where you get the most accurate
timestamp: closest to the interface, or further from the demon. Both are
correct, depending on the specific system. There was quite a bit of work by
Dave Mills about Interleave mode. Regarding the MAC: either the MAC works or
not, and we drop if it does not work. Interleave mode is a wonderful thing. -
Kristof Teichel: this work is very important. Clarification question: does the
server have to keep per-client states? - Aanchal: yes. - Robert: what you did
is great. There are security issues that we need to think about. If we could
avoid adding an additional field for that, it would be great. - Greg: regarding
PTP Follow_up messages: what is the frequency of the messages from the client
to the server? By delaying the information to the next query, we increase the
interval. If we had used something similar to Follow_up messages, the
information would be used immediately. - Karen: please summarize this proposal
on the mailing list. - Aanchal: delay is not an issue here. - Harlan:
interleave mode is most useful in symmetric mode, and less useful in
client/server mode.

ON IMPLEMENTING TIME
--------------------
Presenter: Aanchal Malhotra
https://datatracker.ietf.org/meeting/100/materials/slides-100-ntp-on-implementing-time/

Discussion:
- Greg: a timestamp is not necessarily based on wall clock time. It can be
relative, or can be based on any reference. - Willem Toorop: a timestamp is a
point in time expressed in wall time. - Greg: it is important to agree on the
dictionary. - Harlan: it is unfair to describe some of these issues in the
context of NTP. - Kristof: this work is important. Where can the document be
found? - Aanchal: it is on the agenda. - Kristof: please use the term monotonic
more carefully – need to define it. - Greg: only NTP is used as a reference to
this draft. - Tal: it is important to clearly define the scope of the document,
and the target audience. Another issue is that the intended status should
probably be informational. - Karen: the scope is very important. - Ethan
Grossman: one of the things that was not clear was what the draft was focusing
on: implementation considerations, or security aspects. - Karen: can we have an
update to the draft before we start call for adoption? - Aanchal: we would be
happy if people could send their comments on the mailing list.

==============
TICTOC Session
==============

CHAIR SLIDES
------------
Presenter: Karen O'Donoghue
Slides:
https://datatracker.ietf.org/meeting/100/materials/slides-100-ntp-agenda-and-note-well/

A YANG DATA MODEL FOR IEEE 1588v2
---------------------------------
Presenter: Yuanlong Jiang
https://datatracker.ietf.org/meeting/100/materials/slides-100-tictoc-yang-data-model-for-ieee-1588v2/

Discussion:
- Karen: one remaining unresolved issue. Once we resolve that issue, there is
probably no need for another WG last call. - Yuanlong: we are looking into it,
and I will update the mailing list once we resolve it. - Suresh Krishnan: has
this been reviewed by a YANG expert? Will need to be. - Karen: there was no
official YANG doctor review, but it was reviewed by YANG experts. We can do a
formal YANG doctor review. - Suresh: please do that when the document is ready.
- Yuanlong: we sent version 05 to the NETMOD working group. We received
comments on the mailing list. We have not received any formal comments from a
YANG doctor.