Folks,
Below are the draft minutes for the recent tictoc meeting. Thanks to
Dave Marlow for the timely production of the minutes. Please review and
submit any comments or changes as soon as possible.
Regards,
Karen
Minutes for TICTOC meeting @ IETF-80
28 July 2011, 15:20 EDT (19:20 UTC)
The meeting started at 15:20 EDT. Karen O'Donoghue and Yaakov Stein chaired
the meeting
Dave Marlow took minutes. Karen was jabber scribe.
Karen bashed the agenda and the blue sheets were distributed. Karen provided
the status
for the Working group since the last meeting: 3 working group drafts, 3
individual
submissions, and no interim meetings.
Stefano Ruffino provided slides for an ITU-T SG15/Q13 update which Yaakov
presented.
SG15/Q13 had an interim meeting in May 2011. There were two primary topics of
interest to TICTOC: (1) Packet timing performance aspects for frequency (G.826x
series);
and (2)Time Sync in packet networks (G.827x series). The frequency work is
maturing
while the time/phase work has its requirements document (G.8271) updated and
other
documents are identified and being started. SG15/Q13 is also working on a
definition
and terminology document for both frequency and time of day (G.8260).
Yang Cui provided a Security Requirements discussion based on IPsec security
for packet
based synchronization, draft-xu-tictoc-ipsec-security-for-synchronization-01.
As described
by its Abstract, this document analyses the need for security methods for
synchronization
messages distributed over the Internet and gives a solution on how to mark the
synchronization message when IPSec is implemented in end to end frequency
synchronization.
It was pointed out that the Introduction has requirements language (i.e.
SHALLs) and
these will need to be taken out of this section. There was considerable
discussion on
the IEEE 1588 PTP use case across the Internet where Yaakov and Greg Dowd
pointed out you
cannot have transparent clocks you must tunnel. On a discussion of whether to
encrypt
packets across 3GPP, someone from the jabber room pointed out that with 3GPP
you must
encrypt. Yaakov asked whether the authors have talked with anyone in the
closing IPsec
WG about this draft and Yang indicated that he has and they do not have any
questions.
Karen asked whether the Working Group thought this work should be pursued, but
got little
feedback. Peter Lothberg said that in some use cases this could be useful but
this
provides no value for the Internet use case. Yang said that the femtocell is
getting
greater use and there is no protection in the femtocell. Greg Dowd said that
femtocells
are meant to provide telephone calls and that they can be stopped by dropping
all data.
He said we needed to clarify the threat models we are building the security for.
Karen discussed time synchronization protocol security requirements. The
expired
TICTOC Requirements draft had recently been resubmitted
(draft-ietf-tictoc-requirements-01). This draft identifies three security
mechanisms to
consider. Karen went over a survey that was done at the beginning of the
TICTOC work
where four questions were identified that cover the different aspects for
possible
security services. Karen said that a volunteer editor has been identified to
work on
general time synchronization security requirements. The goal is to have a
draft by the
next meeting.
Greg Dowd provided a PTPv2 MIB discussion based on
draft-ietf-tictoc-ptp-mib-00. This
MIB goes beyond the earlier drafts that only covered the PTPv2 telecom profile.
This
MIB covers all the PTPv2 devices. The current draft is in its third version and
Greg felt it was now semantically correct. Asked whether this MIB has been
deployed,
Greg said that it has. Measuring the performance of a PTPv2 node is outside
the
present scope, but this could be added later. Yaakov asked whether there was
commonality between this MIB and the NTPv4 MIB (RFC 5907), Greg said that this
MIB was PTPv2 specific with little commonality to the NTP MIB. Dave Marlow
said
that he had read the draft and found little commonality with the NTP MIB but
was
very supportive of this draft which appears very thorough. The Chairs said
that
there needs to be MIB Doctor review (which they would pursue) and review by
TICTOC participants for this draft to progress.
Yaakov lead a discussion on Transporting PTP messages (IEEE 1588) over MPLS
Networks,
draft-ietf-tictoc-1588overmpls-01. He said that the draft very mature with
three
individual drafts and now one TICTOC version. There has been considerable
discussion
on the list. Yaakov asked what was needed before this goes to WGLC? Luca
Martini
said that there are some text formatting issues and that some sections have
given a
a wrong impression. Of particular concern were the two modes (IP over MPLS and
Pseudowire), Pseudowire is not needed if timing is a service of the network and
there
is only one clock over the network. Pseudowire is needed if PTP is used
directly over
Ethernet, and Yaakov pointed out that in many parts of the world there are many
clocks
that a carrier transports separately. There were some concerns from the MPLS
community, George Swallow pointed out an issue with fast reroute to address a
fault.
There were concerns whether the draft was ready for last call or not, and what
procedure should be taken to get comments from the MPLS community. George
suggested
to at least announce to the MPLS list about a WGLC in TICTOC.
Tal Mizrahi briefed UDP Checksum Trailer in Timing Protocols,
draft-mizrahi-tictoc-checksum-trailer-00. This draft proposes a Checksum
Trailer
extension to NTP, OWAMP and TWAMP that allows intermediate nodes to reflect the
checksum modification in the last 2 octets of the packet rather than in the UDP
checksum field. This is to bring a capability already in IEEE 1588 into these
other
protocols. Yaakov asked why this is being brought into OWAMP and TWAMP. Tal
said
these are performance protocols that transfer time and could be benefited as
well.
An issue related to the NTP extension field had been raised and discussed on
the
mailing list. This proposed technique requires an extension field without
authentication (i.e. without a MAC) to be practical; however, the NTPv4 spec
(i.e. RFC 5905) REQUIRES all extension fields to include authentication.
Discussion on the list included opinions that the RFC was wrong and this is not
a requirement for NTPv4. Additional effort is needed to determine if NTPv4
spec
should be changed or not. Yaakov pointed out that both OWAMP and TWAMP have
authenticated modes. This draft is intended for non authenticated packets
only.
Yaakov said that in the case of OWAMP this is very limiting because its default
is
to be authenticated. In addition Yaakov said that OWAMP and TWAMP would
require
an extension to their control protocols. OWAMP and TWAMP are IPPM protocols,
the
Chairs took an action to send this draft to the IPPM chairs. Tal was asked
whether the mechanisms discussed in the draft had been implemented and he
said the mechanisms had not yet been implemented.
Dave Marlow discussed Network Time Mechanisms for Improving Computer Clock
Accuracy,
draft-marlow-tictoc-computer-clock-accuracy-00. This draft had not been
updated
since the last meeting but Dave briefed the discussion that had been on the
list.
Vladimir Smotlacha provided to the list, references to his papers describing
NTP
servers which use OCXO oscillators to achieve much higher synchronization
accuracy
than the experimental results described in this draft. The use case in the
draft
covers client accuracy and thus is outside the Vladimir's use case which
addresses
server accuracy. In June, Karen had forwarded a note from Dave Mills to the
list
which mentioned that additional experimental results with NTP Interleave are in
his
book. Dave Mill's book provides experimental results for both an unloaded
scenario
(which is a very similar to the experiment described in the draft) and a loaded
scenario. Dave Mills, in his email, pointed out that Interleave showed greater
improvement in the loaded scenarios and with digest computations (e.g.
Autokey).
Dave Marlow and Tim Plunkett are looking at the experimental results in the
book and
comparing this to their results. Tim Frost provided a paper on Minimum
Time-Dispersion
Metrics to the list. Dave indicated that this provides direction towards
identifying
a common set of metrics for network time synchronization experiments. A common
set of
metrics or perhaps a benchmarking draft could be a candidate for future TICTOC
work.
Greg Dowd said that there are products on the market that can provide a
hardware time
stamp based on a programmable sequence off of a packet data network, this
provides a
direction for the second mechanism outlined in the draft. Yaakov mentioned that
his company tried NTP interleave on their highly optimized products and did not
see a significant difference in performance. Dave said he and Tim Plunkett
would
like to update their draft with the new information. He solicited comments and
contributions on mechanisms to achieve greater client accuracy.
Karen led an additional NTP discussion. The NTP control protocol which is an
appendix of RFC 1305 (NTPv3) is not in a current standard since RFC 1305 was
deprecated.
Volunteers are needed to get this important work documented and standardized.
Brian
Haberman (along with Karen, co-chair of the NTP working group) suggested in
line with
Dave Hart to get rid of mode 7 and just document mode 6. Harlan Stenn, via
jabber
requested that mode 7 be documented as well. Dave Hart, via jabber, said that
mode 7
is fragile and vendor specific, so that there is no use in standardizing it.
Harlan
volunteered to document the NTP control protocol. It was pointed out that RFC
5905 needs
to be redone since mode 6 is identified for future use; however, if this is an
IANA
controlled field then just a change by IANA is needed. Karen asked for
volunteers to
document the NTP interleave extensions. There has been email on list
discussing Autokey
bugs/vulnerabilities being found by PTB (Germany's Standards Organization).
There were
no details discussed at this meeting but this is a concern that must be
followed. There
was a short discussion on what work would be carried out in the NTP WG and what
should
be done in the TICTOC WG. Brian said that all NTP control protocol work should
be done
in the NTP WG but he had no opinion at this time as to where Interleave work
should be
done. All co-chairs agreed that email on any NTP topic should be sent to both
lists.
The meeting adjourned at 17:20 EDT (21:20 UTC).
_______________________________________________
TICTOC mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/tictoc
