►
From YouTube: IETF109-NTP-20201116-0900
Description
NTP meeting session at IETF109
2020/11/16 0900
https://datatracker.ietf.org/meeting/109/proceedings/
A
Okay,
it
is
the
top
of
the
hour,
so
we
will
go
ahead
and
get
started.
This
is
the
ntp
working
group
at
ietf109.
A
The
this
is
the
ietf
notewell
for
those
of
you
who
are
new
to
the
working
group.
You
will
have
acknowledged
and
accepted
this
when
you
registered
for
the
ietf.
A
A
A
A
I
put
this
down
as
ietf
109
hackathon
and
then
put
a
rough
time
report,
but
it's
actually
both
a
report
of
what
rough
time
did
at
the
hackathon
and
also
implications
and
next
steps
for
the
rough
time
draft
itself
and
then
we'll
talk
about
ntf
ntp,
v5
development,
rechartering
in
any
other
business.
A
A
Okay,
so
ditcher,
I
guess
you
get
the
minutes.
A
So
the
next,
so
the
first
part
of
the
first
agenda
item
is
our
ntp
working
group
status.
We
have
a
couple
documents
that
are
with
the
iesg.
Now
we
have
the
mode
6
commands
and
the
yang
data
module
model,
so
there's
no
updates
to
those.
Yet
we
also
have
two
documents
that
are
ready
for
the
iesg
that
have
not
been
submitted.
A
Yet
one
is
the
interleave
modes
document
and
the
second
one
is
the
ntp
port
randomization
and
then
I
originally
listed
the
next
two
documents
as
documents
that
we
had
no
updates
on,
but
we
will
have
an
update
on
rough
time
based
on
the
I
mean
on
the
hackathon.
I
just
didn't
move
it
off
this
list.
A
The
okay!
So
with
that,
the
next
part
of
the
agenda
is
the.
A
C
A
Yeah,
I
just
got
into
the
habit
of
doing
them
all
myself,
so
I'm
happy
to.
C
Sound
so
great,
it
looks
like
that's
working
all
right,
so
I'll
be
discussing
rough
time,
the
rough
time
hackathon
experience
and
the
next
steps
in
the
protocol.
I
noticed
that
a
number
of
people
are
new,
so
I'll
give
a
quick
recap
of
rough
time.
So
it's
essentially
certificate
transparency
for
time.
C
If
these
intervals
are
inconsistent
with
cut
with
the
evidence
of
causality,
then
we
know
that
at
least
one
server
was
lying
when
it
said
that
it
generated
the
response
at
that
time,
and
this
can
then
lead
to
a
distrust
of
the
server,
and
this
is
a
intended
to
provide
a
much
stronger
security
guarantees
for
an
approximate
time
with
applications
to
signature,
verification,
I've
heard
of
interest
in
iot
time,
synchronization
security-
obviously,
but
it's
not
going
to
be
as
accurate
as
you
might
want.
C
It
will,
however,
ensure
at
the
time
is
correct,
and
one
of
the
wrinkles
in
this
is:
you
need
a
monotonic
time
representation,
because
you
want
to
have
the
case
that
you
can
compare.
The
timestamps
and
you'll
always
have
a
later.
A
causally
later
timestamp
be
a
numerically
greater
value
and
to
adopt
this,
we
use
a
representation
based
on
the
modified
julian
date
and
microsecond
since
midnight.
So
there's
a
number
there's
some
bit
packing.
C
So
I
meant
the
hackathon,
so
worker,
cloudflare
and
cloudflare
has
a
client
and
server
implementation
and
there's
now
a
branch
that
implements
the
rough
time
version.
3
draft
there's
still
work
to
be
done
on
updating
the
interfaces
with
some
of
the
changes
that
have
happened
in
the
draft
and
fixing
some
of
the
usability
issues
around
the
library
that
we
provide,
but
that
you
know
general
cleanup,
but
it
does
implement
it
and
the
server
interoperated
within
it
with
a
client
from
johann.
C
There
were
some
issues
we
identified
so
modified
julian
date
in
order
to
compute
it.
You
need
to
know
whether
or
not
you're
in
a
leap
second,
and
that
information
is
known
to
the
kernel
on
unix-like
systems,
but
there
is
no
portable
interface
to
obtain
it.
C
C
So
what's
remaining
with
the
draft,
so
there
there's
an
area,
that's
been
a
sort
of
to-do
list
for
a
while,
which
is
the
transfer
of
proofs
of
server
malfeasance.
C
We
need
to
wait,
we
need
to
serialize
them
and
there
needs
to
be
some
idea
of
where
you
send
them
if
you're
interested
in
deploying
roughly,
I
want
to
hear
from
you.
Even
something
like
this
is
fine
would
be
very
useful.
C
I
I
think
at
that
point,
we'll
we'll
sort
of
be
ready
for
a
working
glass
call
didn't
find
any
issues
with
interoperability.
I
don't
think
from
from
the
draft
tax:
that's
pretty
pretty
nice,
some
particular
areas
of
concern.
So
I
mentioned
earlier
that
there
was
this
problem
with
minimum
with
getting
the
information
out
of
the
kernel.
C
Well,
if
your
minimum,
you
can
only
that
only
changes.
The
time
by
one
second
so
is,
is
having
a
server
say.
Okay,
I
only
know
the
time
within
a
second.
I
am
just
using
get
time
a
day
enough
or
maybe
it
needs
to
be
a
little.
It
needs
to
be
a
little
bigger
with
a
modified
julian
date.
Representation.
C
C
You
need
to
know
whether
what
the
last
seconds
number
is
there's
also
invalid
representations.
The
number
of
seconds
in
a
day
is
not
a
power
of
two
and
because
it
is
not
a
power
of
two,
there
are
a
number
of
values
that
should
never
appear
and
the
invitation
when
you
program
this
is
to
do
your
arithmetic
on
a
degenerate
representation.
Just
take
the
set
microseconds
part
and
add
in
the
timestamp
normalize
afterwards.
C
What
happens
if
a
server
produces
signatures
that
some
parties
verify
and
others
do
not?
Can
this
be
used
to
lie
and
get
away
with
it?.
C
C
There's
a
small
number
of
earlier
versions
deployed
mostly
of
the
google
version
that
was
kicking
around
for
madame
langley,
not
enough
financial,
three
servers,
I
I
think,
there's
exactly
one
of
them
and
it's
not
working
right
now
and
got
restarted
every
day
and
there
need
to
be
places
to
report
malfeasance
there
needs
and
we
also
need
clients
and
just
as
a
this
is
a
certificate
verifying
client
needs
to
have
policies,
certificates,
verifies
or
a
certificate.
Transparency,
implementation,
a
client
that
uses
certificate.
Transparency
needs
to
trust
her
in
logs.
C
The
client
in
in
rough
time
needs
to
have
a
policy
of
acceptable
servers,
and
that
is
a
bit
of
an
obligation,
and
it's
so
that
that
is
something
it
needs
to
at
some
point
happen:
are
there
so
at
this
point
I
think
I'll
open
the
floor
to
questions
if
there
are
any,
because
that's
sort
of
us
done
with
the.
C
D
I
have
a
question,
but
you
you
mentioned
some
areas
of
concern,
but
you
all
also
mentioned
that
you
think
this
draft
is
waiting
for
working
group
plus
calls
so.
C
D
A
So
watson,
the
during
the
hackathon,
did
you
get
any
participants
in
your
project.
C
Yes,
johan
lindquist
showed
up
and
his
client
interoperates
with
the
with
the
server.
So
that
is
good.
A
A
Do
you
have
a
rough
estimate
on
the
the
time
that
you
think
like?
What's
your
next
plan
for
your
next
update.
C
I'd
like
to
say
as
soon
as
the
submission
window
reopens,
but
I
that's
that
might
be
me
being
too
optimistic.
Probably
within
the
month.
A
A
A
To
so,
there's
no
other
questions
on
the
rough
time
work.
Okay,
so,
on
ntp,
v5
development
work,
we
do
not
have
a
published
draft.
What
we
do
have
is
a
draft
in.
A
So
this
is
the
contribution
from
marislav
that
will
get
submitted
at
some
point.
The
other
thing
that
I
wanted
to
point
people
to
was
the
track.
The
wiki
which
we
haven't
updated
in
a
while.
But
if
you
look
at
the
wiki,
there's
a
potential
ntp
work
items
and
there's
basically
a
list
of
version,
4
issues
and
a
version
5
feature
requests.
A
A
C
I
don't,
I
think
at
this
point,
it's
pretty
early
to
say
much
about
it,
it's
in
part
because
it
intro
it
talks
a
lot
about
the
data
format,
but
doesn't
talk
about
gathering
for
some
of
the
more
interesting
extensions,
but
I
I
think
it's
in
reason.
You
know,
I
think
it's
in
reasonably
good
shape
for
your
starting
point.
C
A
I
think
one
of
the
concerns
is
it
narrows
the
scope
from
ntp
v4
and
we're
gonna.
We're
gonna
need
to
have
a
serious
conversation
about
backward
compatibility
at
some
point.
D
C
Okay,
watson
yeah,
I'm
not
sure
what
was
meant
by
narrowing
the
scope.
If
I
I
think
there
was
some
degree
of
agreement
that
we
should
specify
the
network
lay
in
the
network,
wire
format
and
the
algorithms
for
steering
somewhat
independently,
basically
split
split,
the
current
structure,
where
it's
algorithms
for
treating
the
clock
and
the
network
protocol
in
one
big
document
and
split
it
up
in
part,
because
we
of
operational
experience
where
you
can
use
other
algorithms
that
works
fine,.
A
Right,
that's
not
actually
what
I
meant.
I
do
believe
there
was
agreement
to
do
that.
I
was
in
particular
talking
about
like
reducing
it
to
server
client
mode
only
as
opposed
to
supporting
other
types
of
operations.
A
I
guess
scope
is
not
really
the
right
word,
but
deprecating
some
features
that
are
not
being
used
as
much
now
as,
but
that
are
part
of
the
original
specification.
So
that's
more
what
I
meant
by
narrowing
the
scope.
A
E
Here
we
go
okay
great,
so
I
I
just
wanted
to
follow
up
on
that
last
comment
about
other
modes
of
operation
and
the
possibility
of
marking
marking
them.
As
you
know,
no
longer
recommended
or
something
you
know.
I
think
I
think
this
is
a
discussion
worth
having
you
know.
We
want
backwards
compatibility,
but
we
could
say
we're
phasing
these
loads
out.
E
A
Have
all
right,
I
guess
I'm.
A
And
but
we
don't,
we
aren't
having
a
good.
We
haven't
had
enough
discussion
to
make
any
consensus
calls
on
like
things
like
whether
things
should
be
deprecated
or
not.
So
it
would
be
helpful
to
have
some
of
that
discussion
all
right
anything
else
on
ntp
v5,
I
still
don't
see
marislav.
A
We
have
been
promising
some
rechartering
texts
for
a
while,
but
we
have
not
done
that
yet
I
just
had
that
open.
I
have
created
an
ether
pad
to
start
collaborating.
We
got
some
volunteers
at
our
last
virtual
interim
to
work
on
the
to
work
on
some
chartering
text
and
we
did
not
make
any
progress
on
that.
That's
mostly
my
fault
for
not
pulling
it
together.
A
A
That's
the
wrong
button
too
many
buttons
with
that
we
are
at
the
any
other
business.
This
is
going
to
be
the
shortest
ntp
meeting
ever
eric
go
ahead.
B
The
link
was
available.
I
think.
A
Yeah-
and
I
had
it
like
30
minutes
ago,
and
now
I
can't
find
it
it's
somewhere
in
the
millions,
I
will
send
it
to
the
mailing
list,
because
I
cannot
find
where
I
put
it.
A
In
all
likelihood,
in
mid-january,
I
don't
know
that
it
makes
much
sense
to
try
and
get
it
in
in
december.
A
A
A
A
For
this
it'll
be
on
youtube.