Folks,
Below are the draft minutes for IETF 80. The minute taker sent them to
me in a timely manner, and I dropped the ball in getting them posted. My
apologies. Comments welcome.
Karen
Minutes for TICTOC meeting @ IETF-80
The meeting started at 17:30 CEST. Karen O'Donoghue and Yaakov Stein chaired
the meeting. Dave Marlow took minutes, and Karen was jabber scribe. The
agenda was bashed, and the blue sheets distributed.
Eliot Lear provided an information brief on Maintaining the Time Zone Database
which covered information in draft-lear-iana-timezone-database-01. This was
provided to the TICTOC wg because time zones were of interest to many of the
TICTOC participants. The Time Zone Database goes back to 1980's, and is based
on contributions of hundreds of people. This contains perhaps the most
comprehensive and complete Time Zone history. There are about 500 distinct
entries in the Time Zone Database where local time is not a continuous
function; for example, Daylight Savings time transitions are a step function.
Eliot's primary issue is related to the future of the Time Zone Database. The
person who is maintaining it is retiring in 2012 with no successor identified.
Currently there are many problems with the future for the Time Zone Database
including no license structure and no place to store the database, including
the historic information. TICTOC participants were invited to comment on
Eliot's ID which proposes to transfer hosting of this database to IANA.
Stefano Ruffino provided an ITU-T SG15/Q13 update. He went over a number of
updates including an updated structure of their documents. ITU-T efforts as
they relate to TICTOC: (1) The TICTOC IEEE 1588 over MPLS effort does not
impact the ITU-T frequency synchronization initiatives; (2) There is no
specific efforts on-going in the Q13 on security, though the outcome of TICTOC
in this area could be used and referenced by Q13
Recommendations; (3) Management work is in the scope of ITU-T activities (e.g.
Q14).
Packet timing management will also need to be addressed. A TICTOC 1588 MIB
might be referenced by future ITU-T documents, coordination might be
appropriate on this point. Yaakov asked whether discovery is of interest in the
ITU or is every clock already manually configured? Discovery is not an ITU-T
subject at this time.
Yixian Xu provided a Security Requirements discussion based on
draft-xu-tictoc-ipsec-security-for-synchronization-00. Of specific interest is
whether timing packets need to be identified and understood by Intermediate
Systems transferring across an IPsec tunnel. The assumption is that the IPsec
tunnel is already in place. Two proposals were discussed, one allowed
Intermediate Systems to identify time packets and the other did not. Discussion
on this was moved to the mailing list. One needs to be able to recognize
timing packets in order to use hardware assists. In discussing whether the
time information was encrypted or not, someone said that RFC 3401 has a policy
on tunnels where policy decides what goes through the tunnel and what does not.
Karen had a short discussion on general security requirements. When the agenda
was set for this meeting it was thought that an editor was available to start
work in this area; however, an editor is still needed. This discussion was
deferred until an editor is found. Karen mentioned that the interim
requirements worked out at the Ireland IETF are not available, and that she was
looking for notes from Paris interim meeting that covered this area. Someone
said that security should go fast since IPsec is closing down.
Tim Plunkett provided a brief on Network Time Mechanisms for Improving Computer
Clock Accuracy based on draft-marlow-tictoc-computer-clock-accuracy-00. This
effort is looking for new mechanisms beyond those identified in NTPv4 (RFC5905)
that provide increased time synchronization accuracy for operating system
clocks' time and frequency. The ID identifies three candidate mechanisms for
experimentation. The use case targeted is a dense concentration of computing
elements connected by a gigabit network where a satellite-based time source
(i.e., GPS) is used for synchronizing primary time servers and secondary time
servers and leaf computing elements are synchronized via the network. In this
use case, some of the computing elements need to synchronize to each other to
within a microsecond while other computing elements only need to be
synchronized to each other to within a millisecond. Initial experiments have
been started in the use of NTP Interleave. This is an option similar to the
mechanism used in IEEE 1588 and is available in recent NTP distributions. Two
other mechanisms were discussed as possibilities including an approach that
merged the better capabilities of NTP and IEEE 1588. Stefano Ruffino asked why
NTP broadcast mode was used in the initial experiments, and Tim mentioned that
this was the option provided in the recent NTP distribution. Tim suggested the
start of a standards track effort to develop time synchronization performance
metrics that allows different experimental efforts be performed in a way that
the results are comparable. Yaakov thought that performance metrics is really
an ITU-T effort. Tim Frost said that existing ITU metrics are
frequency based. His company has made contributions for time based metrics to
ITU. Karen said that metrics related to those worked in the Benchmarking WG
could be worked in the IETF. Tim Frost was asked to see whether his company
could contribute the metrics to the TICTOC work. Vladimir Smotlacha said he
performed experiments with NTP servers using
OCXO oscillators showing much higher synchronization levels than those measured
by
Tim Plunkett in his initial experiments. Vladimir was asked to provide
references to his papers, which he later passed through the list. Tim
solicited comments and contributions on the mechanisms described and on
additional mechanisms that should be considered.
Tim Frost discussed the Precision Time Protocol Version 2 (PTPv2) Management
Information Base, draft-vinay-tictoc-ptp-mib-00. This is a new draft covering
all PTPv2 clocks (ordinary clocks, boundary clocks, and transparent clocks) and
goes beyond the telecom profile allowing for multiple ports and multiple
instances. Remaining issues to resolve include: (1) How should we handle
different profiles, such as the G.8265.1 telecom profile; (2) How should we
handle associated data; (3)Do we want to configure PTP clocks via MIBs, or
simply collect management information? It was pointed out that SNMPv3 security
capabilities can be used. Yaakov asked Tim if he would refresh his Requiremens
doc, Tim answered that that he would if there is help and interest. Stefano
Ruffino asked how to handle profiles, which Tim answered it would be via a
family of MIBs. Ron Cohen asked the diference between the clock and domain,
but this was taken to the list. Yaakov said that he wants a 1588 MIB which
covers all cases. The Chairs requested review of this draft. There is
interest in moving this to a Working Group draft.
Shahram Davari discussed the Transporting PTP messages (1588) over MPLS
Networks, draft-ietf-tictoc-1588overmpls-00. Shahram reviewed the contents of
the draft, including IS-IS and OSPF signaling extensions for 1588 aware LSRs.
Open issues with this draft are: (1) Support 1588 over P2MP LSP, which requires
co-routed return PTP LSP. There are drafts proposed for LDP and RSVP-TE. (2)
Encapsulation Types, Ethernet and UDP/IP Encapsulations are defined. The draft
describes mechanisms which are a layer violation. Does the group need PTP PDU
directly over MPLS? For the Ethernet encapsulation, Yaakov said to ask the
IEEE. Stefano Ruffino raised a concern about modifying the payload.
(3) FRR may not be usable since it is not bidirectional, some thought that this
was not a concern, just the way it is. Recommend 1:1 or 1+1 protection or
redundant master? (4) Is PHP support required? Some thought it would be easier
to just say that PHP is not supported. If yes then BGP and Targeted LDP will
need extensions.
Yaakov Stein discussed Generic MPLS Time Correction Field,
draft-stein-tictoc-mpls-00.
Shahram pointed out that this is a competing draft to
draft-ietf-tictoc-1588overmpls made by the Chair. Yaakov said that he did not
want affect the current Working Group draft but to begin working with the MPLS
WG to develop a more general solution. The idea is instead of a
1588-specific CF mechanism, to define a generic MPLS mechanism that can be
applied to any
MPLS packet. This would enable applying CF to any 1588 over MPLS encapsulation
or applying CF to NTP or Time Protocol. There are two encapsulation
possibilities: (1) Follow-up CF using GAL and (2) In-packet CF. This will
require coordination with PWE who control the first nibble registry. The
signaling for OSPF, IS-IS, or RSVP-TE can be carried out similarly to in
draft-ietf-tictoc-1588overmpls. Ron Cohen, via jabber, said that a similar
idea had been suggested in IEEE for an Ethernet time correction field. The
planned MPLS Discussions was moved to the next TICTOC telecon since time had
run out.
Karen is planning to continue monthly discussions via telecon on Tuesdays,
these will
be announced on the list. WG milestones and charter are badly out of date and
these
need to be updated.
_______________________________________________
TICTOC mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/tictoc
