►
From YouTube: IETF106-HACKATHON-20191117-1400
Description
HACKATHON meeting session at IETF106
2019/11/17 1400
https://datatracker.ietf.org/meeting/106/proceedings/
A
A
That's
there
and
we'll
essentially
run
through
everything
that
that's
there
at
this
time
and
then
afterwards,
I'll
go
back
and
we
can
individually
pick
up
ones
that
we
missed
or
that
God
updated
or
something
like
that,
but
just
to
try
to
speed
it
up,
I'm,
going
to
pull
everything
now
and
we'll
just
run
through
it.
In
order
and
plus
that
way,
you
can
kind
of
see
where
you
are
in
the
list
and
when
your
your
slots
coming
up.
B
A
C
A
D
A
Okay,
so
so
the
way
this
is
going
to
work
is
you
know
we
have
that
three
minute
time
limit
as
I
mentioned
rod
here,
will
be
keeping
he'll,
be
the
timekeeper
and
he'll
give
you
kind
of
a
nice
like
sort
of
one-minute
warning
when
you
have
only
one
minute
and
then
he'll
give
you
a
less
nice.
You
know
notification,
I,
guess
when
your
time's
up,
if
you
don't
wrap
up
in
like
two
or
three
seconds,
there'll,
be
something
even
worse,
so
everyone
start
clapping.
Okay,
we
can
use
that
practice.
That
sounds
good,
so
yeah.
A
C
So
this
this
is
Sir
J
map,
as
it
can
be
used
right
now.
I'll
send
a
link
to
this
around
once
we're
happy
that
it's
a
little
bit
more
stable
and
everyone
can
have
a
play
with
it.
It
slopes
in
in
slightly
less
than
real
time
because
it's
pulling
from
our
map,
but
this
is
what
you
can
do
with
modern
IMAP,
a
modern
email
standards
I'm
at
replacement
feel
free
to
poke
around
in
ITF
topic
box
scratch
comm.
A
Okay,
so
different
than
usual,
all
just
seems
by
default.
It
put
things
in
reverse
alphabetical
order,
so
we're
gonna
go,
go
backwards,
which
maybe
is
a
good
way.
So
so
you
I'm
just
gonna,
go
in
order
here,
so
the
first
one's
going
to
be
Yang,
and
then
se
e
and
so
on.
So
you
kind
of
have
an
idea
when
your
slot
is
up
so
yang
mediator,
who's,
presenting
that
one.
A
F
F
So
here
is
our
hack
song
plan.
We
we
have
things
when
we,
when
we
developing
the
ietf
service,
modeling
and
also
the
device
model
we
we
have
facing
this
kind
of
problem
here.
On
the
left
hand,
side
there's
a
the
Sdn
controller
architecture
like
in
a
lot
of
deployment,
most
as
the
in
controllers
needs
to
do
like
different
vendor
adapters
to
use
different
vendor
native
young
to
manage
the
whole
network.
F
But
we
also
have
some
service
provider
have
other
requirements
from
them
like
the
they
they
have
their
requirements.
They
don't
want
you
like
facing
different
vendor
specific
yum
modeling,
so
they
thinking,
maybe
they
can
use
IETF
native
young
to
manage
all
the
devices
and
a
network
and
here's
our
proposal
to
this
kind
of
problem.
F
So
we
are
proposing
to
use
a
mediator
as
a
as
a
solution
to
to
migrate
the
vendors,
basically
young,
to
IETF
young,
like
in
the
right
that
right
now
like
there's
most
vendors,
they
don't
have
implemented
IETF
young
right
now,
but
they
can
using
the
mediator
as
the
proxy
to
the
ITF
young
modeling
when
managing
the
the
whole
network.
The
sd1
controller
in
the
network
could
could
only
use
like
idea
as
a
basics
or
PGP
SS
to
control
the
ho
s,
a
VPN
services.
F
Then,
when
they're,
when
most
when
right
now
at
moment,
they
can
only
use
vendor
specifically
on
to
develop
the
SI
basics
function.
Okay,
thanks!
So
here
is
the
topology,
and
this
is
our
hexam
topology
to
to
control
the
whole
network
and
and
here's
our
okay.
Okay,
thank
you.
Here's
our
demo,
rego
and.
A
G
H
H
We
just
like
to
address
some
of
the
existing
problems
by
revising
aqm,
so
the
traffic
scenarios.
We
wanted
to
look
at
we're
bursty
links
that
are
often
associated
with
Wi-Fi.
These
are
typically
problematic
for
high
fidelity
ecn
schemes.
We
first
check
verified
that
this
affects
SCE,
as
well
with
the
original
marking
scheme.
H
Now
we
introduced
nasty
queueing
to
have
minimal
extra
complexity
over
a
plain
FIFO,
and
that
helps
and
to
adjust
the
bursty
links
on
the
path
we
changed.
The
SE
marking
over
to
use
caudal
instead
of
a
read
play
about
algorithm,
and
this
turns
out
to
have
helped
wrath
of
a
lot
increasing
throughput
from
seven
to
forty
megabits
per
second.
I
So
what
cut
a
ton
and
is
a
custom
project,
so
we
fixed
some
compound
error,
so
the
the
Connor,
the
comparation
from
the
owner,
another
University.
We
tried
to
clear
out
the
make
file
error,
or
so
we
enable
the
PS
atiyah
up
and
also
we
enable
the
back
camera
driver.
So
we
can
successfully
a
computer
IP
version,
6
address
and
also
UDP
data
delivery
and
also
video.
There
are
delivery
okay.
So
this
is
a
testing
environment.
So
we
go
over
there,
we
have
a
two
or
laptop
computers.
I
We
can
insert
the
Ubuntu
Linux
and
then
we
can
enable
the
errors
that
nowadays
call
comm
Wi-Fi
module,
so
we
cannot
successfully
delivered.
So
this
is
the
video
on
the
snapshot.
We
unloaded
a
video
on
demonstration,
a
clip,
or
so
we
can
unload
it
the
menu
for
anybody
who
have
interesting.
You
can
install
intensity.
So
this
is
the
summary,
so
we
are
learned
so
how
to
configure
how
to
IP
version.
I
6
delivery
on
top
of
a
commercial
Wi-Fi
learning,
5.29
gigahertz,
ok,
also
during
this
hackathon
or
we
showed
current
the
neighbor
discovery,
I
P
wave
is
not
a
complete,
so
I
feel
great
working
group.
We
work
on
for
our
neighbor
discovery
to
enable
automatic,
illegal
Opera
Builder
s
also
neighbor
vicar
recognition
will
be
done
in
near
future.
So
we
have
a
champion
me
and
jung
yong,
kim
sun-shin
university,
also
post
student,
ok,
also
alliance,
ID
that
you
can
get
a
bit
of
clay
and
menu
and
also
original
contributor
is
check.
You
know,
conquer
University
impossible.
J
Okay,
hello,
everyone:
this
is
Satoru
presenting
the
IIT
Hakkasan
for
its
our
basics,
mobile
user,
praying
I
explained
it
a
bit
our
Hawks
on
plan
to
simplify
the
mobile
network.
Its
idea
is
describing
this
right.
It's
a
basic.
My
buddies
were
praying
while
DMM
working
group,
so
we
implement
mobile
user
plan
requirement
into
one
single
main
layer,
which
is
si
v6.
I
PB
square.
It's
our
basics.
To
be
able
to
integrate.
Mobile
is
a
friend
to
the
one
to
the
underlying
layers.
J
So
this
is
wrap
up
our
team
member.
It's
to
see
I'm
murakami.
If
he
who
is
your
past
time
to
join
the
Hakkasan
and
Kentaro,
I
Basava
have
become
a
leader
and
Phil
Berger.
We
also
have
the
skeptic
code
to
de
generate
the
SR
basic.
Oh
no
gtp,
you
a
5g,
our
pocket,
which
is
very
new,
and
this
is
the
lack
of
the
the
point
URL
to
show
the
people
code,
which
is
also
available
on
the
idea
of
Hakkasan
github
library.
So,
please
check
out.
Thank
you.
D
K
My
god,
sister
Michael
from
hallway
and
in
today,
wave
concern
is
about
to
implement
one
even
condition
action
young
day
tomorrow,
and
this
is
our
hack
plan.
We
will
we
well
find
relatives
and
the
possibility
of
the
Union
mother
for
our
policy,
basically
even
amendment
to
implement
and
the
test
an
event
in
easily
folder
location
as
of
holding
application
in
the
network
device,
and
we
also
implement
in
test
and
the
chain
reaction
of
the
coordinate
the
humans,
and
here
is
a
reference.
K
If
you,
there
are
two
rates,
our
document
appraiser
a
date,
and
if
you
have
some
stressor
in
the
comments
they
are
welcomes-
and
here
is
my
use
case-
and
scenarios
is
a
automatic
detect.
If
you
don't
have
the
possibility,
hunted
module,
we
need
to
do
this
work
step
by
step
and
it
rely
on
operators
experience.
But
now
we
have
a
Jungian
module.
We
can
generalize
a
young
instance
and
submitted
to
the
at
was
ninety
was
implemented,
come
rated
to
in
person,
scraped
and
subscribe
to
the
relative
teacher
and
our
sizes
either.
K
If
it
meets
the
condition,
it
will
ask
you
some
specified
action
to
use
as
Detective
works
out,
some
self-healing
works
and
what
we
do
you
system
away.
The
wrapper
on
is
a
photo
localization
and
the
self
held
implication
did
not
work
device
and
to
provide
the
out
who
are
localization.
Self-Healing
such
as
architect
and
here
are
way
also
to
our
one
network,
with
application
to
display
that
he
was
changing
over
time.
The
last
one
is,
the
young
instance
is
a
is
a
script
general
read
by
you,
say,
models
and
write.
One
is
a
scenario.
K
What
will
end?
Okay,
we
collected
many
useful
comments
and
we
know
that,
yes,
you
say
young
ml,
the
yonder,
the
mother
can
develop
not
an
image
to
spot
automatic,
photo
localization
and
self
Harry,
and
also
across
many
other
use
available
use
case
such
as
mass
filter,
and
it
is
computing
and
we
have
a
great
member
sentence,
their
comments
and
contribution,
and
also
we
will
have
shown
our
demo
in
tomorrow's
happy
hours
and
if
you
are
interested
in
this
topic,
it's
just
thank
you.
L
Hey
folks,
Santosh
I'm
part
of
crabs
information
security
team.
So
the
topic
is
on
okay,
so
I'm
Santosh,
I'm
part
of
grabs
information
security
team,
which
is
a
right
healing
company
based
in
Singapore
I'm
sure
quite
a
few
of
you
would
might
might
have
used.
There
were
companies
services
to
reach
the
venue
from
our
hotel.
So
the
topic
is
on
a
personal
information,
identification
and
logging.
So,
basically,
you
know
like
in
in
today's
world
they
might
have
several
of
the
log
sources
could
be
that
servers
or
any
of
the
other
applications.
That
might
be
inadvertently.
L
Logging,
PII
data-
and
this
is
one
of
the
challenges
that
we
face,
and
you
know
like
regular
I
know
like
my
day-to-day
job
stuff.
So
how
do
we
effectively
identify
and
take
right
actions
on
PII
data?
It
could
be
redaction
masking
encryption
or
trimming
it
out
from
privacy
reasons,
as
well
as
data
and
compliance
reasons.
Before
we
send
this
data
out
to
any
third-party
processing
could
be
external
sock
knock
or
you
know
like
a
vendor
who
is
trying
to
debug
your
appliances
right?
L
So
you
might
want
to
sort
of
trim
that
out
so
the
hackathon
plan
was
to
you
know
like
identify
the
conventional
methods
of
you
know
like
how
do
they
work
when
you
want
to
sort
of
parse
the
logs
conventional
methods,
like
largely
regular
expression
based,
but
not
a
lot
of
contextual
data?
You
know
like,
as
opposed
to
the
proposal
that
we
are
putting
on
the
table
with
respect
to
tagging
the
data
at
the
source
before
it
is
sent
out
for
any
sort
of
processing
or
parsing,
and
then
probably
put
this
logs
out
and
lovely.
L
L
A
data
detection
under
action
is
one
of
the
use
cases
that
we
are
tried
and
we
can
clearly
see
that
you
know
like
you,
probably
would
have
a
lot
lower
accuracy,
but
the
conventional
methods,
especially
the
red
one,
indicating
at
the
bottom,
is
showing
the
hash
index,
which
is
actually
matching
with.
Probably
a
national
ID,
see
Singapore's
NR
IC
or
something
is
what
we
were
using.
That
also
gets
redacted
as
part
of
this
so
yeah.
L
I
Okay,
thank
you
Taurus.
Well,
this
is
a
poor
jung-gu,
so
this
is
the
tenth
consecutive
I
to
an
SSR
hackathon.
So
over
all
three
years
we
work
for
ohm,
I
turn
SF.
So
previously
we
are
demonstrated
that
the
major
interfaces
such
as
registration
to
page
consumer-facing
and
natasha
key
function
MSF
facing
also
we
can
demonstrate
T
the
script
avala
generator.
So
this
time
we
want
to
demonstrate
NSF
the
monitoring.
So
this
is
not
a
complete.
I
Yet,
as
we
are
keep
going
for
next
year
and
comedy
yeah,
so
we
use
the
open
source,
especially
comedy
will
be
I
used,
especially
the
notification
picture
will
be
used
for
monitoring
okay.
So
this
is
the
cuss
on
project,
so
we
have
so
collaborator
and
also
student,
and
also
the
newcomer,
the
two
persons
joined
so
the
right
side.
We
have
some
information
for
this
hackathon,
so
this
is
the
our
eye.
I
Tuna
type
of
framework,
the
left
hand
side
our
going
to
provide
the
intent
of
base
descriptive,
surpise,
a
right
side
we
implemented
using
on
table
and
actively
and
also
Sdn
technology
based
upon
their
light
and
also
of
the
stack.
So
this
time
we
try
to
implement
I
turn
SF
a
nap
and
I
set
up
the
monitoring
so
on
waiting
work.
So
this
is
our
basic
are
tracked.
Okay,
so
I
am
an
editor
for
destruct.
We
based
on
our
castle
experience.
We
will
revise
this
trap
so
we're
under
how
to
implement.
I
M
M
We
were
planning
to
do
a
little
bit
of
a
improvement
on
the
mud
visualizer
we
had,
we
actually
were
going
to
do
do
a
visualization
of
one
particular
device
called
yes,
SC
comm
recorder,
Brian
showed
up
to
implement
mud
and
grasp
and
I
was
doing
some
debugging
on
an
internal
implementation,
and
then
our
fault
of
the
folks
at
c-calm
have
been
looking
to
develop
a
model
to
use
mud
for
to
actually
make
devices
accessible
and
in
a
scalable
fashion.
Such
the
device
doesn't
have
to
implement
a
lot
of
mechanism.
M
So
a
lot
of
this
got
done
turns
out.
Brian
didn't
need
to
work
too
hard
at
this.
He
got
the
initial
work
going
in
about
30
minutes.
We
still
have
some
signature
work
to
do.
I'll
talk
about
that
savio
from
the
Federal
Institute
Brazil
built
out
some
capabilities
in
the
visualizer.
If
you
haven't
seen
the
visualizer,
it's
pretty
cool,
you
can
actually
visualize
mud
files
and
and
see
what
sort
of
access
of
device
we'll
get
I
did
my
debugging
I
have
more
debugging
to
do.
M
We
actually
did
test
the
device
and
we
were
able
to
visualize
it,
and
we
did
some.
We
did
some
initial
modifications
for
the
mud
manager
for
parsing
the
file
extension.
This
is
a
little
view
of
Brian's
work
that
he's
done
with
grasp
top
UI
he's
not
yet
comfortable
pushing
this
into
a
pip
repository.
Given
that
yeah
we
have
a
has
a
couple
more
things
he
wants
to
do
and
we
really
want
to
bind
it
to
one
of
the
open
source
mud
manager,
implementations.
M
Here's
like
an
approval
flow
that
I've
been
working
on,
such
that
one
of
the
key
things
that
we
have
for
administer
network
administrator's
we've
got
to
make
it
simple
for
them
to
understand
what's
on
their
network
and
have
a
simple
approval
flow.
This
is
what
we
were
working
on.
We
learned
a
fair
amount,
actually,
the
first
of
which
is
that
autonomic
computing.
One
of
the
things
we
said
was
that
on
and
I'm
an
autonomic
competing
needs
to
give
some
thought
to
ad-hoc
authorization
models
in
terms
of
what
devices
can
talk.
M
We
reinforce
that
mud
is
not
meant
for
device
configuration,
but
network
configuration
did
individual
individual
eyes
device
that
it's
not
meant
for
device,
identity,
open
source
version
and
management
is
still
a
pain.
Everybody
still
knows
this,
and
the
Python
API
is
really
need
a
little
bit
of
help
regarding
verification
and
CMS
and
here's
the
people
who
are
involved
there's
an
updated
version
of
this
presentation
that
has
a
nice
picture
and
correct
spelling
in
some
places.
So
thanks
very
much
and
see
you
in
the
next
hackathon.
A
D
N
N
We
identified
some
gaps,
namely
regarding
local
originated
routes
in
the
BMP
local
it.
It
seems
that
RFC
42
71
doesn't
really
specify
what
the
next
stop
activate
attribute
should
look
like.
We
take
that
to
the
NGO
working
group.
Another
thing
which
we
identified
was
regarding
the
pier
up
messages
when
in
adjacency
rip
out
pre
and
post
policy,
configuration
was
applied
depending
on
implementation,
either
one
or
multiple
messages
were
sent
so,
and
this
ambiguity
me
meant
some
some
hiccups
on
the
on
the
data
collection
side
and
on
the
BMP
d
sector.
N
We
take
that
as
well
to
the
working
groups
and
see
how
other
vendors
implemented
that
and
yeah
the
good
thing
was
lab
environment
was
working
and
the
bad
thing
in
the
test
took
longer
than
expected.
So
next
time
we're
gonna
do
in
107.
We
want
to
improve
our
testing
and
also
at
now,
these
additional
three
drafts
which
are
building
up
on
the
current
one,
which
we
test
it
this
time
thanks
a
lot,
and
these
are
all
hackathon
new
timers
thanks.
D
O
I'm
max
I,
mixed
from
taps
and
MB,
which
are
closely
related,
the
sake
of
fun.
So
some
of
us
worked
on
taps,
which
is
like
a
replacement
for
sockets.
Basically,
we
have
been
here
before
yeah.
Those
are
the
drafts
we
had
interactions
with
and
then
we
also
try
to
work
on
Emme,
which
is
the
new
draft
by
Jake.
And
so
what
did
we
get
done?
In
taps?
O
We
managed
to
implement
an
endpoint
racing,
which
is
neat
we've
had
between
agates
once
you
implement
the
protocol
Racing's
in
a
way
of
endpoint
racing
and
protocol
racing,
and
also
working
on
local
endpoint
racing.
There
have
been
a
lot
of
fixes
for
taps
during
a
heck
of
fun
and
we
found
some
new
interesting
interactions
and
issues
for
Framus
which
are
going
to
take
to
the
working
group.
Then
for
mb
we
implemented
the
first
draft
implementation.
O
D
P
Yeah,
what's
goal
is
to
design
an
open
source
correct
for
recovering
packet
losses
and
reduce
latency
compared
to
traditional
real
ribbon,
for
instance,
block
codes.
There
are
strong
relationships
with
the
Korean
efficient
paper
communication
research
group,
and
we
have
been
working
on
also
on
specification
of
the
ILC
correct
for
that
source
code
for
that
which
is
almost
published
as
RFC
86-81.
It's
a
matter
of
days,
so
the
goal
yes
I,
kept
on,
as
I
said,
is
to
design
this
reference
open
source,
free,
correct
and
also
to
challenge
generic
API
internet
draft.
P
That
explains
how
to
use
such
collects.
What
we've
been
working
on
this
time
is
to
centrally
the
decoder
side,
which
is
the
complex
part
of
a
correct.
This
is
where
you
have
to
manage
and
do
linear
system
decoding
using
some
kind
of
Gaussian
elimination.
So
we
took
advantage
of
an
existing
library
following
that,
but
this
library
was
done
was
designed
for
different
context
with
different
logic,
and
so
this
this
is
complex
to
line
this
this
library
with
what
we
are
doing
here.
So
this
is
basically
what
we've
been
doing.
P
Thanks
to
Siri
can
do
Mema
and
we
also
I
don't
find
an
issue
in
the
generic
API
internet
right.
That's
right!
We
continue
working
on
the
Python
wrapper,
as
well
as
the
testing
C
application,
so
small
team
free
people,
including
myself
that
I
can
do
Mima
if
you
are
interested
in,
is
the
github
repository
and
you
can
also
attend
our
research
group
meeting
on
Thursday
afternoon.
Thank
you.
N
Q
So
we're
looking
to
set
up
a
create
a
development
environment
for
us
to
look
at
some
of
the
usability
issues
and
the
problems
we
face,
and
so
we
set
up
this
marionette,
which
is
a
goal:
language
implementation
of
something
called
parts
of
a
finite
state
machine
that
does
format
transforming
encryption.
So
we
can
make
one
kind
of
traffic
look
like
another
kind
of
traffic,
and
this
gives
us
many
ways
to
do
that.
Q
So
it's
a
good
environment
that
has
many
options
for
the
user
to
select,
all
of
which
are
completely
opaque
if
you're,
just
a
regular
user,
and
you
need
to
get
something
done.
So
we
want
to
understand
how
to
you
know
what
the
limitations
of
the
current
model,
this
marionette
project
is,
are
along
with
what
the
global
issues
are,
that
users
face,
and
can
we
put
these
things
together
to
give
us
give
ourselves
a
good
development
model
for
looking
at
user
experience
and
making
some
claims
about
future
user
experience?
Q
This
is
how
the
idea
of
pluggable
transports
works.
Overall
here
it
is
in
in
in
marionette,
where
you
have
a
client
that
runs
low,
is
locally
hosted
on
your
device
that
helps
you
select
a
format,
transforming
encryption
mode
that
that
is
collected
by
a
server
which
itself
links
to
the
application
server.
So
there's
a
number
of
things
in
here
and
our
current
model
they're,
fixed
and
Static,
and
they
need
to
get
a
little
bit
more
dynamic.
So
we
have
to
look
at
how
do
we
bootstrap
the
client
properly?
Q
Do
you
select
that'll
implement
that
technology
in
the
way
you
need
so
I'd
like
to
thank
the
team
at
marionette
who
did
this
environment
based
on
some
some
university
research
that
was
done
first
in
Python,
and
then
another
team
implemented
this
and
go
our
sponsor
is
inter
news,
and
we
thank
idea
for
giving
us
a
chance
to
talk.
There's
a
lot
more
on
pluggable
transports
at
pluggable
transports,
dot
info
when
we
have
an
internet
draft
on
it,
also
in
the
usual
place.
Thank
you
very,
very
much.
D
R
Everybody
I'm
Seth
can
I
just
got
a
quick
show
of
hands
of
who
uses
Gmail
regularly.
That
is
a
lot
of
people
hi
I'm
Seth.
This
is
Neil
from
pass
fail.
I,
so
Neil's
worked
for
the
last
good
while
to
make
a
beautiful
new
application
protocol
for
email
clients
to
be
able
to
work
effectively
because
IMAP
is
terrible.
R
Honestly,
unfortunately,
all
of
the
people
in
the
room
here,
which
is
like
about
two-thirds
of
people
who
use
Gmail
today,
can't
take
advantage
of
beautiful
new
Jay
map
API
and,
as
a
result,
you
have
awful
loading
bars
when
Gmail
loads
and
you
don't
have
a
lot
of
options
because
third-party
email
clients
need
to
implement
email
functionality
specific
for
Gmail
or
they
use
IMAP,
both
of
which
are
vaguely
terrible.
So
so
we
thought
well,
wouldn't
it
be
great
if
we
had
some
sort
of
map
from
Gmail,
API,
Gmail
husband,
a
snake,
you
guys
too
well.
R
S
R
Browser
email,
so
that's
what
we've
spent
that
weekend
doing
and
J
map
if
anyone's
curious
to
look
at
the
slides
is
actually
quite
beautiful.
You
can
do
these
kind
of
recursive
RPC
queries.
Unfortunately,
Gmail's
api
is
oh
yeah.
Gmail's
api
is
kind
of
awkward
and
slow
and
we
kept
running
into
rate-limiting
issues.
We
quietly
suspect
that
Gmail
doesn't
actually
use
their
own
API
to
build
their
own
email
clients
and
it
sort
of
hasn't
been
exercised.
R
R
Gmail
included
jam
up
and
that
way
we
could.
We
could
have
all
of
the
stuff
working
well
so
yeah
if
you're
interested,
there's
all
this
stuff
stuff
on
github,
although,
unfortunately,
as
we
said
it's
all
very
proof
of
concept
and
I'm
really
be
great.
If
we
could
throw
up
a
website
which
was
you
know,
gem
up
with
Gmail
already
working,
but
it's
really
slow
and
after
you
refresh
it
a
couple
times,
it
starts
taking
or
for
a
minute
to
load.
Just
because
it's
the
right
limit.
S
U
We
were
not
as
many
people
this
time
as
usual
in
Asia
that
we
were
in
Europe
and
so
on,
but
we
still
had
some
interesting
project
program,
so
we
had
apple
Intel
and
a
few
other
coming
today
to
try
to
bring
a
new
codec
in
the
RTP
part
of
the
stack
everyone,
even
though
it's
done
at
I/o
media,
it
touches
RTP
which
is
done
at
IETF.
So
we
wanted
also
to
you
to
you,
know,
share
the
copies
and
look
at
all.
We
can
make
him
better
and
so
on
and
so
forth.
U
Alright,
so
the
usual
suspects
plus
a
few
first-timer
Intel
join
us
for
the
first
time
thanks
and
welcome,
I
hope
it's
going
to
be
good
enough
to
make
them
come
back
and
participate.
We
were
certainly
happy
to
have
yet
another
an
implementer
providing
feedback
to
for
the
specification
and
another
company
based
focusing
on
security
and
encryption
that
we
wish
is
going
to
join
us
again.
Oh
thanks.
V
Thank
you.
This
is
teep
three
minutes
talk
doesn't
mean,
I
have
to
talk
three
minutes
just
going
through
slice.
Great
team
will
sorry
misspelling
and
what
we
plan
is
evaluating.
The
draft
with
origin
is
story
from
authority
without
a
version
and
then
in
what
highly
modified
to
have
a
tea
proposal,
and
we
brought
few
implementation
to
Tam,
which
is
as
works
as
a
server
and
the
device
to
the
different
kind
of
device
and
got
done,
is
all
pictures,
so
I
don't
have
to
type
on
the
slides
so
type
on
the
table.
V
This
is
Eli
on
the
tip
Pam.
It's
able
to
upload
the
TAS
over
here
is
Oba
sons.
Time,
and
this
is
the
device
running
on
arm
opti
and
hacking
in
the
bucking
and
many
things
didn't
work
and
finally,
yeah
t
transit
application
was
able
to
install
on
the
T
and
what
we
learned
is
many
things:
it's
it's
written
in
the
slides,
but
I'm
honestly
for
the
implementation.
V
It's
long
long
away,
Dave
stamps
and
device
is
pretty
much
implemented,
but
we
need
a
strategy
how
to
make
it
as
a
properly
in
the
future
of
the
reference
implementation
and
that's
the
wrap
up.
But
before
finishing
it,
this
implementation
will
cost
a
lot
of
a
different
kind
of
knowledge
te.
How
does
it
work
in
different
kind
of
CPU
and
json
seaboard
and
HTTP
stack,
or
have
to
be
working
perfectly
to
implement
on
something
on
top
of
it
and
how
needs
the
knowledge
of
how
the
device
boots
sick?
So,
yes,
it's
a
lot
of
work.
W
W
Actually
they
exist.
Existing
Janusz
telemetry
provides
a
mechanism
to
select
and
subscribe.
Your
operational
data
object
it's
based
on
a
selection
filter,
but
there
is
no
document
to
discuss
how
the
selection
filter
can
be
specified
to
address
this
issue,
we
introduced
the
self-cleaning
conception,
which
can
be
interpreted
as
instruction
that
data
no
the
selection
principles
so
with
the
self
explanation
information.
W
We
develop
a
running
code
and
deploy
during
the
telemetry
collector
and
provides
module
data
node
or
to
automatic
tagging
automatic
alluring
and
automatic
subscription
mechanisms.
The
left
figure
is
out
is
our
tag
model.
We
call
the
OPM
model
the
over
object
type,
for
example,
for
example.
The
interface
is
an
object,
type
and
P
for
property,
the
interface
name.
We
call
it
is
the
property
and
M
for
Mature,
the
CPU
usage
or
the
receiver,
or
the
stand.
Bytes
is
a
matrix.
W
The
right
finger
is
how
we
realize
our
OPM
model
in
our
demo.
You
can
see
we,
we
collect
the
CPU
usage
from
the
two
devices
generally
say.
Without
any
pre-configuration,
the
monitoring
data
can
be
collected
and
displayed,
can
chemical
acted,
matically
and
displayed
on
the
GUI
error
results.
We
imported
our
collected
data
to
the
two
in
flux
to
the
influx
TV
as
the
data
source
and
we
process
and
analyze
the
telemetry
characteristic
data
using
Grovner.
Both
the
two
software's
are
open
source.
W
D
X
X
Big
plan
was
to
get
some
work
done
on
SSH,
mostly
X
MSS,
which
is
a
first
quantum
signature,
and
it
was
also
on
our
plan
to
finish
HTTP
for
5-1
Drupal
module,
which
was
taking
forever
so
so
what
got
done
HTTP
for
5-1
were
final
fixes
for
Drupal,
but
upstream
maintenance
requested
TLS
1.3
support
for
an
open
source
project
known
as
check
firewall
loads
of
XM
SS
and
SS.
F,
p.
Ss
h,
FP
support
added
to
a
bunch
of
tools,
so
quite
a
few
of
them
were
relying
on
Association
just
wanted
updated.
X
X
X
So
we're
still
it's
very
new
and
I
think
we're
going
to
be
busy
next
year
with
more
obvious
kind
of
work,
so
the
people
who
do
it
did
it
in
from
Mauritius
or
actually
quite
young
between
15
years
old
to
19
years
old,
so
I'm,
just
a
picture
of
the
group
and
thanks
to
Rogers
Capital
for
hosting
very
wet
emerges
and
was
overly
soft
team
members.
So
again,
thank
you
very
much.
A
Y
Hi
I'm
Murray
I'm,
representing
at
the
best
table
for
computing
in
a
network,
we're
a
proposed
research
group,
but
we're
trying
to
do
things
we're
about
data
playing
programming
and
adding
computation
and
the
network.
So
our
original
problem
that
we
had
started
dressing
at
our
first
hackathon
in
Montreal
was
to
do
packet
filtering,
and
we
started
with
that.
Y
As
an
idea,
we
had
some
tons
of
inputs,
actually
there's
a
big
p4
org
that
has
a
ton
of
templates
and
all
kinds
of
things
for
us
to
get
going
and
virtual
machine,
and
we
had
a
great
participant
Michelle
who,
because
of
his
own
work,
had
also
a
lot
of
tools
to
get
people
getting
going
we're
still
at
a
set
of
individual
projects,
because
our
team,
we
don't
have,
we
have
a
team.
We
have
a
bunch
of
people
getting
together
and
trying
to
have
fun
and
but
in
the
future,
work
for
the
next
hackathon.
Y
We
intend
to
have
a
project
that
will
keep
everybody
together
and
we
intend
to
have
that
defined
by
by
the
group.
So
what
got
done?
Obviously?
Who
editorials
for
people
like
me
and
others
who
don't
really
know
what
p4
is
about
or
not
enough?
We
had
Michelle
who
continued
his
work
on
service
function,
chaining
and
he
had
great
improvement
on
his
code.
So
we
were
very
proud.
There
was
a
just
for
fun
in
coin.
Y
We
address
a
lot
of
industrial
networking
issues,
and
so
we
had
coy
who
actually
designed
something
to
do
packet
filtering
for
sensor
networks,
where
you
actually
can
define
what
a
packet
and
how
a
packet
is
treated
based
on
the
measurements.
From
a
sensor,
we
also
had
Aris
who
started
looking
at
some
BGP
origin
validation.
They
had
a
snafu,
they
figured
out
that
p4
wasn't
good
enough
for
that,
and
so
they
took
an
alternate
route.
Y
Y
We
could
do
that
at
interim
meetings,
decide
what
type
of
setup
we
want
to
have,
bringing
a
VM
on
a
USB
would
be
great
and
by
the
way,
yes
should
be
more
poster
boards
with
pens
and
paper,
because
it's
fantastic
to
communicate
with
the
people
around
your
table
and
not
just
the
people
online.
On
slack-
and
this
is
the
fantastic
team
we
had
actually
all
the
start-
people
or
new
stars-
the
great
people
I
see
I.
Y
Z
I
Me
summarize
the
support
project,
so
this
project
is
related
to
our
general
research
group,
and
the
support
is
open-source
software
platform
enabling
shishi
and
communication
fit
is
comparing
two
ways.
These
two
are
F
sheets.
The
main
objective
of
this
project
is
to
enhance
support
functionalities
in
the
context
of
the
two
proposed
draft
shishi,
an
info
and
the
network
holding
for
shion
and
a
DM.
I
So
more
precisely
in
this
hackathon
we
implemented
to
support
functions.
One
is
in
network
controller
using
incision
info,
which
come
some
operate,
which
can
enable
some
operator
to
check
out
in
network
condition
and
the
configure
which
data
packets
should
be
cached,
and,
second,
is
a
Content
producer.
Application
called
Campari,
which
can
you
which
can
utilize
donor,
according
in
the
context
of
shishun
communications
as
a
shipment,
we
built
and
tested
the
basic
problem
modules
related
to
the
two
functions
and
regarding
lessons
we
got.
I
V
Hello:
everyone,
my
name
is
KJ
from
University
in
Korea
and
secretin,
so
we
try
to
define
the
words
of
our
draft
in
the
BMW
G
containerized
infrastructure,
benchmarking.
So
for
that
draft
to
reroute
that
the
superior
allocation
are
the
technology
for
this,
these
days
is
and
not
guaranteed
to
the
deterministic
results
so
or
we
want
to
verifying
this
or
to
use
some
tools.
So
the
first
is
a
native
kubernetes,
CPU
scheduler
and
then
second
is
comparing
with
CPU
pining
technology
using
the
cmk
by
Intel.
V
So
we
thought
that,
firstly,
we
thought
that
just
for
informs
switches
to
the
container,
but
there
have
so
many
the
configuration
is
needed
and
then,
on
the
other
end,
we
learned
more
than
understanding
for
technology
and
then
configurations
and
all
the
options
we
tried
and
or
and
or
one
the
one
thing.
One
more
thing
is
the
checks
power
consumption,
so
our
site
server
was
shut
down
during
the
hackathon
for
some
for
why
or
run
when
running
checks
with
100%
CPU
usage.
A
Z
Thank
You
Charles,
yeah,
ok,
hi,
I'm,
mal,
Morton
and
I
was
working
with
Lynch
avataan
remotely
on
our
measurement
project,
and
here
it
is
we.
We
talked
a
bit
about
this
in
Montreal
we
were,
we've
been
working
on
an
internet
draft
now
we're
into
the
third
version
of
maximum
IP
glare
capacity,
metric
and
methods
of
measurement,
and
it's
a
UDP
based
method
at
the
ITF
we're
trying
to
gain
experience
with
really
challenging
access,
one
gigabit
per
second
with
lots
of
users
that
won't
go
away.
Z
That's
you
guys,
so
we
compared
that
with
TCP
Kubik
this
time.
Last
time
we
compared
it
with
you
cos
system,
which
is
also
TCP
based,
and
we
run
the
test.
We
iterate
the
measurement
parameters
and
do
it
again,
so
we
were
also
going
to
examine
the
crossover
with
the
benchmarking
project
that
you
just
heard
about,
but
we'll
have
to
wait
for
later
in
the
week.
So
here's
what
got
done
this
time.
Looking
over
here
at
the
right,
we
had
our
measurement
system
implemented
in
this
nice
little
PC
minute
mini
PC
system.
Z
So
we're
able
to
run
that
bare-metal
and
we
also
had
a
Google
Cloud
VM,
located
right
here
in
Singapore,
less
than
4
milliseconds
away,
and
it
was
still
a
15
to
18
hops.
The
measurement
system
sends
traffic
in
one
direction
and
it
gets
feedback
in
the
other
direction.
So
it
actually
makes
adjustments
very
quickly
and
we
use
the
UDP
and
IP
3.6
clients
and
servers
to
do
that.
You
can
see
on
the
on.
The
right
is
the
I
guess.
Z
On
the
right
hand,
side
of
the
graph
is
the
percentage
delivery
for
packets
that
we
were
able
to
measure
using
the
UDP
system.
You
can't
get
that
kind
of
information
readily
from
a
TCP
measurement
or
the
delay
variation
very
easily
as
well.
What
we?
What
we've
got
here,
we're
measuring
a
one
gigabit
per
second
bottleneck
somewhere
between
us
and
our
server
and
our
UDP
measurements
in
blue.
Z
There
are
pretty
close
to
theoretical
the
lowest
one
I've
plotted
here
is
998
megabits
per
second,
when
corrected
for
Ethernet
header
and
one
VLAN
tag,
the
other
ones
are
actually
higher
and
the
last
one
there
had
loss
ratio
of
practically
nothing
so
and
then
the
three
TCPS
are
not
doing
so
good.
So,
let's
see
here
23
seconds.
Z
Okay,
so
we
learned
a
lot
today
was
testing
was
really
messy,
so
don't
test
on
Sunday
many
PC
was
good,
a
good
bare
metal
platform.
We
need
to
do
better
load
adjustment,
the
search
algorithm
uses,
loss
and
delay.
We
need
to
have
a
better
threshold
for
loss
and
we
got
some
actions
for
the
existing
draft
and
so
forth.
Yep
you
and
me
are
synchronized,
so
we'll
try
to
do
that
later,
with
the
keep
going
with
the
benchmarking
group
this
week.
D
A
AA
Everybody,
since
we
are
getting
to
the
end
of
this
session,
I
decided
to
post
it
the
last
slide.
First,
so
here's
the
project
rep,
oh
and
the
team
picture.
We
had
a
dozen
contributors
this
time,
quite
a
big
team
compared
to
the
other
times
the
usual
suspects
draft
authors
and
some
good
software
developers.
We
had
so
eg
from
Japan
who
usually
attends
remotely
you
flew
in
at
his
own
expenses,
just
to
be
with
us
this
time.
AA
AA
AA
AA
We
even
tried
running
the
the
test
code
on
memory
device
itself
and
yes,
it
fails.
It's
a
big
big
with
the
test
environment,
so
we
need
to
split
the
code,
but
that
gives
us
some
way
forward
and,
of
course,
we
need
more
tests
more
test
automation.
We
also
spotted
a
missing
element
in
the
data
model,
which
is
another
draft
that
we
work
on
at
the
working
group
and
also,
we
think
we
are
ready
for
interrupt.
AA
Z
A
AB
All
right
welcome,
so
these
are
the
results
of
the
did
the
DNS
table
good.
So,
as
usual
in
the
DNS
with
a
DD
Venus
table,
we
don't
have
one
single
plan.
We
have
a
group
of
individuals
with
their
own
plans
and
we
work
together
and
we
talk
a
lot,
but
we
have
a
number
of
themes:
the
privacy,
DNS
privacy.
AB
Of
course,
it's
important
in
the
past
years,
DNS
protocol
improvements,
container
DNS
and
service
provisioning
and
submitting
some
measurements
good,
genus
privacy,
continuing
work
on
dinners
of
atilla's,
but
previous
in
the
previous
years
we
worked
from
from
from
the
stuff
from
the
clients
to
the
resolver
new
work
in
the
DNS
privacy.
Deprived
working
group
is
from
the
resolver
to
the
tortoise,
a
dot,
also
gold.
AB
So
there
wasn't
hack
kind
of
proof
of
concept,
implementation
inbound
how
to
implement
that
using
a
subnet
specification
and
a
host
name
fela
date
it
to
have
an
authenticated,
tillis
connection
to
an
authoritative
nameserver.
How
to
bootstrap.
This
list
is
something
the
deprived
working
group
has
to
work
on.
So
if
you
have
ideas
in
this
room,
you're
welcome
to
join
the
discussion
on
deprived
working
group.
AB
Friday
morning
Dinah's
over
HTTP,
the
pi9
has
and
made
a
design
for
implementing
the
OS
in
by
9,
and
with
that
we
have
in
2000
2008
in
by
9
and
bounds
in
not
resolver
and
in
power
Dinah's
this.
So
we
have
for
open-source
implementation
of
do
H,
which
is
good.
So
if
there's
a
choice,
there's
an
options,
different
options
for
deployment,
also
a
proof
of
concept,
implementation
of
DNS
over
hp3,
quick
working
progress
based
on
one
minutes,
things
in
case
some
protocol
improvements,
cookies
server,
Dena,
cervical
keys.
Why
should
we
care
about
DNS
server?
AB
It's
a
way
to
mitigate
DDoS
spoofed
addresses
you
can
filter
them
out,
so
DNS
server
cookies.
We
did
had
an
in
specification.
It
wasn't
very
Interpol.
You
could
make
different
implementation
decisions
with
the
base
specification.
We
solve
that.
We
have
three
four
implementations
that
are
interoperable
and
extended
error.
Well,
skip
that
not
a
draft
working
on
service
provisioning
I
want
to
mention
that
it's
important
something
we
are
thinking
L
in
the
Dinah's
community.
Thinking
a
lot
about
how
to
solve
this
problem.
In
the
last
ITF
there
was
a
new
draft.
AB
A
AC
Hi
everyone
so
I'm,
Remy
I,
was
working
here
also
with
the
Salva
and
hang
were
sitting
over
there
on
the
captive
portal.
Some
captive
portal
proof
of
concept
for
the
Capitol
working
group
for
people
who
are
not
familiar
with
the
work
they're.
Basically,
the
first
objective
of
the
the
what
they
have
kept
kept
pulled
API,
let's
discuss
in
the
working
group,
is
first
to
make
captive
portals
this
bad
because
they're
like
very
often
really
bad
and
because
the,
for
example,
detecting
that
there's
a
portal
is
really
tricky.
AC
But
apart
from
that,
it
also
enables
number
of
use
cases,
for
example,
knowing
how
much
time
you
have
left
on
a
portal
if
it's
limited
or
like
knowing
how
much
data
you
have
left
also
having
a
way
for
you
to
know
if
you're
still
connected
or
not.
If,
like
the
the
network,
just
is
connected
you
and
you
have
no
idea
about
it-
general,
it's
not
great
and
also
it
enables
use
cases
such
as
you
know,
having
in
fight
on
containment
in
the
plane.
We
don't
have
internet
anything
like
the
schedules:
tight,
iam
timetables.
AC
If
you
are
at
the
HTF,
knowing
like
where
you're
making
is
oh,
like
the
train
at
the
train
station
stuff,
like
that,
so
general
flow
of
the
demo
we
built
was
first,
you
will
discover
the
API
endpoint
for
that
API,
so
it
from
the
draft,
it
could
be
a
ready,
HTTP,
every
six
hours,
all
the
link,
relational
header
as
a
backup
we
use
we
implemented
as
DHCP,
then.
Basically,
the
demo
probes,
the
endpoint
to
Detective
or
portals
of
is
open
or
closed
I,
get
the
session
information
and
show
info
in
a
notification.
AC
So
this
is
a
demo
that
was
implemented
on
an
Android
phone
like
with
the
open
source,
iOS
P
and
on
the
access
point
sight.
So
we
have
so
a
team
from
Google
session
which
who
is
here
today
who
provided
so
one
implementation
that
they
have
you
know
for
for
for
this
captive
portal,
API,
so
yeah,
and
they
provide
whether
the
client
is
captive
or
not.
They
provide
the
captive
portal
page.
How
much
time
is
left
and
also
a
URL
for
information
which
happens
to
be
the
IETF
information
page.
AC
So
here's
what
the
demo
is
I
can't
show
you,
because
on
a
screen
would
be
way
too
small.
But
here's
what
it
is.
Basically
you
connect
to
a
Wi-Fi.
It
tells
you
tap,
you
have
to
sign
in
the
usual
thing
when
you
open
the
portal,
which
is
you
know
the
in
this
case
the
Google
Station
provided
portal,
you
can
log
in
you
get
free
Wi-Fi,
and
then
you
get
this
notification
that
tells
tap
to
manage
and
when
you
tap
on
this,
you
can
get.
A
AD
Everyone
I'm
an
angel
from
away,
and
now
we
are
working
on.
We
worked
on
the
Institute
flow
information
telemetry
the
I
feet
provide
a
reference
framework
to
acquire
data
about
a
packet
on
its
a
forwarding
path,
but
when
we
deploy
it
in
the
network,
there
are
several
challenges
for
deployment
in
Carroll
Network.
AD
And
this
is
what
we
implemented
last
time
in
Montreal
and
we
implements
the
postcard
based
telemetry
and
achieve
several
use
cases,
including
the
delay
monitoring
for
no
link
an
end-to-end,
and
we
also
implement
the
packet
loss,
monitoring
and
the
past
tracking.
This
time
we
add
two
functions:
why
is
the
scratch?
Guided
elephant
flows
direction,
we
can
dynamically
insert
the
sketch
in
in
the
device
and
select
the
elephant
flow
for
high-precision
monitoring
and
second,
we,
the
the
exporting
data
packing.
AD
D
AE
Weird,
a
quick
measurements
team
working
on
the
spin
dump
tool
so
for
this
hackathon
we
plan
for
adding
some
more
functionalities
to
the
spin
dump
tool,
which
is
basically
a
network
latency
monitoring
tool.
So
we
wanted
to
implement
support
for
what
is
called
a
delay
bit.
It's
an
experimental
proposal
of
adding
another
bit
to
improve
latency
measurements.
We
also
wanted
to
improve
our
RTT
statistics
and
we
wanted
to
add
more
more
spin
bit
supports
to
quick
stacks.
AE
So
what
we
got
done
was
that
we
added
support
for
this
delay
bit,
so
we
can
get
improved
RT
measurements
for
four
endpoints
that
support
that
one.
We
also
improve
the
statistics
by
adding
some
some
more
statics
like
standard
deviation
of
the
RTD
measurements,
and
we
also
implemented
some
RTT
filtering
based
on
that,
and
we
also
managed
to
get
spin
bit
support
into
the
move
fast,
quick
stack
from
Facebook.
AF
Okay,
hello,
everyone,
here's
a
report
from
we
see
our
work
on
IOT
semantic,
hyper
media
interoperability.
Our
plan
for
this
this
time
was
work
on
data
model,
convergence
and
automate
translation,
in
particular
with
the
1bm
language
and
tools.
Also,
we
plan
to
work
together
with
w3c
web
of
things.
We
had
a
joint
meeting,
this
Friday
together
with
them
and
they
joined
here
on
their
web
of
things,
block
fest
activities.
And
finally,
we
plan
to
work
on
a
new
P,
which
is
this
translation
between
binary
and
and
JSON
based
data
models.
AF
What
cut
down
the
over
this
weekend,
we
had
a
lot
of
work
on
the
1dm
tools
and
concepts
in
particle
are
doing
conversions
between
1
DM
SDF
and
live
with
them
to
emit
somewhere
else.
So
this
is
the
reverse
from
the
what
we
did.
One
last
idea
where
we
did
it's
at
1
a.m.
but
will
also
enhance
that
kind
of
translations.
We
also
have
a
linter
for
1
DM
model,
so
you
can
check
the
validity
of
your
models
and
we
worked
on
a
CI
in
github
to
actually
check
real
time.
AF
When
you
commit
new
models,
they
are
actually
valid.
Most
of
code
for
above
is
already
available
on
on
github
on
the
link
above,
we
also
had
very
good
discussions
on
architecture
use
of
semantics
how
we
can
handle
dynamic
descriptions
and
in
particle.
How
should
we
work
on
convergence
across
SDS
on
various
specifications
in
this
area
on
U
P?
Now
we
have
also
basic
encoding
from
chasing
the
binary,
so
we
can
go
both
ways
on
between
JSON
based
and
binary
based
models
and
web
of
things.
We
have
no
first
version
of
TD
discovery
using
coop
Rd.
AF
One
thing
that
we
agreed
to
work
on
going
forward
is
this
Symantec
proxy?
So
the
idea
is
to
go
from
one
DM
descriptions,
potentially
with
use
of
some
RDF
to
thinking
descriptions,
and
then
we
could
be
using
something
like
Symantec,
proxy
or
a
gateway
to
work
between
applications
and
devices
from
different
ecosystem,
see
if
you're
interested
to
work
on
this
topic,
you're
very
welcome
to
join
our
ongoing
Vichy
activities
at
the
tingling
Archie,
something
we
learned
again.
Of
course
we
knew
about
it.
You
shouldn't
do
models
and
and
examples
by
hand.
AF
AF
A
AG
It's
it's
essentially
a
transition
mechanism
from
today
to
a
position
where
everything
can
just
be
low
latency
without
having
to
particularly
do
anything
about
it
and
we
mean
really
low
latency
links
there
to
existing
code
and
the
specs.
We
had
a
number
of
remote
people
as
well
today,
and
also
three
new
people
to
appear
on
site
and
three
remote
people
who
were
remote
last
time.
So
it's
it's
not
all
bad.
When
your
remote,
some
people
do
it
twice.
AG
What
we
got
achieved
really
three
projects,
the
first
two
of
those
bullets
about
testbeds
and
the
first
one
was
one
of
the
new
people
getting
them
to
work
through
the
read
million
check.
You
know
it.
A
fresh
set
of
eyes
can,
through
everything,
without
anyone
else,
telling
them
site
information
and
also
the
the
other
one
going
on
on
the
test.
AG
They're
also
that
a
complete
remote
team
working
on
NS
3
added
first
start
up,
got
that
running,
but
still
more
to
do
on
that
need
a
testing
framework
around
it
updated
the
g
oq
to
the
spec,
although
that's
still
in
progress,
and
he
added
configuration
code.
So
if
the
auto
configures,
when
you
all
the
ECM
DSP
stuff
for
the
NS
three
implementation,
a
piece
of
TCP,
so
we
learn,
we
found
a
few
missing
steps
in
the
readme
accuracy.
N
found
it's
more
flexible
in
first
appears.
This
was
also
with
the
SE
e
team.
G
So
hi
everyone,
my
name,
is
accused
of
an
eagle
I
worked
for
a
net
mode
in
Sweden,
so
I'm
gonna
talk
about
about
NTS,
it's
security
firm
for
MTP,
basically,
okay,
sorry
so
time
is
important.
Is
it's
basically
the
message
and
I
think
everybody
knows
that
so
I
didn't
have
to
come
up
with
a
lot
of
examples
and
when
it
comes
to
time
of
the
internet,
MTP
is
the
protocol
that
everybody
uses
it's
been
around
since
1985
I
think
it
even
predates
BSD
sockets,
so
NDP
is
pretty
old.
G
Ntp
does
have
some
security,
but
it's
based
on
symmetric
keys,
pre-shared
keys,
so
it
doesn't
really
scale
to
a
lot
of
users.
Nps
adds
the
kind
of
scaling
we
need
to
be
able
to
do.
Secure
NTP
to
everybody
and
I
wants
to
meet,
haven't
done
a
lot
of
progress
today,
which
is
in
a
way
good.
It's
because
we
got
a
lot
of
things
done
at
early
hackathons
in
I
had
my
first
hackathon
was
in
Prague
and
I
felt
that
we
made
a
lot
of
progress
at
that
time.
G
So
today,
what
I've
done
is
really
just
been
working
on
scripts
to
verify
the
functionality
of
the
entire
servers.
That's
because
we
had
netted
her
MTS
servers
up
and
running
now,
and
we
intend
to
keep
them
up
and
running
so
I'm
trying
to
build
tools
so
that
our
IT
operations
department
can
see
if
anything,
fails
and
quickly
corrected.
G
It
might
be
interesting
to
other
people,
so
I'll
try
to
open
sources.
I'm
I've
been
also
working
on
how
to
on
how
to
actually
start
using
NTS
in
a
client
and
I've
been
using
entity
SEC
as
an
example,
and
this
how-to
is
suppose
to
or
I'm
going
to
show
how
to
do
it
with
the
NTS
services
for
a
university,
not
real
anger
has
done
those.
We
have
cloud
fire
officially.
G
We
said
that
they
have
entire
service,
that
they
intend
to
keep
running
I've
done
a
an
implementation
for
net
mode,
which
is
a
Frankenstein's
monster
of
my
Python
code
and
the
time
stamping
from
Crowley
and
NTP
SEC.
They
was
around
a
couple
of
servers
and
sort
of
good
news.
The
NTP
NTS
draft
was
submitted
for
publication
on
them
the
7th.
G
So
hopefully
this
will
actually
become
a
standard
pretty
soon
and
not
much
to
say
today
it's
been
me,
Krista
fighting
on
deepest
today,
people
who
have
been
working
on
NTS
and
I'll
be
at
the
academia
tomorrow.
If
anybody
has
some
questions
or
wants
to
see
anything,
what
has
any
one
is
about
NTS
and
there's
a
link
to
the
draft
all
all.
So
if
anyone
wants
to
see
it,
thank
you.
D
AH
So
we're
working
on
a
problem
that
comes
from
the
real
physical
world
you're
at
a
emergency
scene.
You
see
this
unmanned
aircraft
hovering
above
the
building,
that's
on
fire
and
you
want
to
know
whose
it
is.
So.
You
can
figure
out
whether
it's
like
another
fire
department
and
has
some
business
being
there,
or
maybe
it's
the
guy
that
started
the
fire.
AH
So
there's
another
standards,
development
organization,
ASTM,
International,
that's
been
working
on
this
problem
for
the
US
Federal,
Aviation,
Administration
and
various
other
civil
aviation
authorities
in
the
rest
of
the
world
problem
is
their
standard,
does
not
provide
for
information
that
is
immediately
actionable
and
privacy-preserving.
So
we're
looking
at
leveraging
some
ITTF
work
mostly
related
to
hip
we're
going
to
need
to
update
hip
with
some
new
crypto
and
make
it
hierarchical.
So
fundamentally,
this
is
to
make
the
claims
that
come
across
one
of
these
remote
ID
systems
trustworthy.
AH
AH
We
started
discussing
observer
to
pilot
emergency
comes
via
hip
proxies
back
in
New
York.
We
flew
a
Bluetooth
broadcast
UAS
from
what
ID
of
currently
standardized
flat
hit.
What
we're
gonna
demo
here
right
now
and
you
here
in
the
room,
aren't
gonna,
be
able
to
see
this,
but
those
who
are
playing
along
at
home
since
the
cameras
right
here
will
be
able
to
get
a
better
experience.
AH
I'm
gonna
start
a
scan
here
and
then
Adam
is
gonna,
start
broadcasting
using
his
cell
phone
as
a
surrogate
for
an
aircraft
and
I
hope
that
the
Bluetooth
range
reached
it
did
my
screen
paint.
Yes,
it
did.
Those
of
you
who
are
playing
along
at
home
can
see
that
the
UAS
ID
type
is
a
host
identity
tag
which
is
not
in
the
ASTM
standard,
but
we're
lobbying
for
it
to
get
added
to
that
standard.
We
need
some
more
reviewers.
We
need
some
inter
opera.
AH
The
experimenters
some
of
the
work
that
was
done
in
hip
a
few
years
ago
on
an
experimental
basis,
really
needs
maturation.
Some
of
the
drafts
need
maturation
when
I
say
UAS
from
what
I
D
roles
of
who
is
I'm
really
talking
about
the
whole
domain
registration
process,
not
so
much
the
who
is
protocol.
Probably
the
new
are
DAP
protocol.
How
are
we
gonna
use
DNS,
because
the
privacy-preserving
aspect
is
really
important.
AH
So
the
new
IETF
work
will
be
mostly
in
the
hip
working
group,
but
we're
also
going
to
need
to
do
some
coordination
with
other
standards,
development
organizations,
including
ASTM,
the
International
Civil
Aviation
Organization,
and
the
radio
technical
Commission
for
aviation,
and
so
all
the
team
members
are
noobs
except
for
Bob
Moskowitz
and
please
join
us
set
up
off
Tuesday
morning.
Thank
you.
AI
Okay,
so
our
goal
is
to
implement
RFC
2544
tests
in
simple
command
line,
two
that
connects
the
devices
part
of
the
test
setup
over
net
conf
and
those
devices
had
the
young
model.
So
we
have
a
draft
that
describes
this
solution,
which
is
described
by
this
diagram
here,
and
the
execution
of
the
test
should
be
as
simple
as
running
a
command
line,
providing
a
configuration
based
on
the
UNK
models,
mainly
the
topology
ITF
model,
augmented
with
data
about
connecting
and
specifying
what
kind
of
parameters
the
test
should
actually
take
into
account
when
executing
so
this.
AI
AI
To
do
a
RFC
2544,
you
need
the
hardware
to
actually
execute
it
correctly,
and
then
we
had
to
make
hardware
for
that.
So
we
we
have
two
testers.
One
is
a
proprietary
tested
on
the
bottom
and
the
other
one
is
the
open
source
open
hardware
tester,
which
is
running
Debian.
Both
of
them
are
running
Debian
and
you
may
want
to
try
net
conf.
AI
AC
AJ
Usual
suspects
were
in
the
back
corner.
There
interrupting
their
quick
implementations,
I
think
we
had
seven
clients
and
eleven
servers
actually
participating.
That's
a
little
worse
than
usual,
a
couple
reasons
for
that
number
one.
We
I
think
with
lower
attendance
here
in
Singapore,
muted
in
Montreal
and
some
others,
and
secondly,
draft
24
drop
only
a
couple
weeks
ago,
when
I
think
we'd
several
implementations
that
weren't
quite
caught
up
yet.
Nevertheless,
the
story
is
fairly
similar,
a
lot
of
a
based
on
functionality
we're
pretty
strong
Interop
on,
even
though
some
of
the
details
are
changing.
AJ
We
have
a
few
implementations
that
are
doing
some
of
the
advanced
IP
address
migration
stuff
and
some
of
the
more
advanced
HTTP
3
features
like
push
and
and
so
on,
but
probably
not
as
many
as
we
would
like.
Overall
I
think
things
are
moving
forward.
A
lot
of
production
implementations
are
are,
maybe
don't
quite
have
all
the
all
the
interoperability
checkboxes
that
some
of
the
some
of
the
you
know
the
for
fun.
Implementations
are
having
but
they're
getting
there
and
there
they're
getting
closer
to
do
something
can
actually
deploy
questions.
A
Ok,
any
presentations,
I
miss,
ok,
one
one
thing
I
forgot
to
mention
earlier:
if
some
of
the
teams,
if
you
want
you're
like
a
team,
photo
taken
or
something
we
have
our
friendly
photographer,
who
will
hang
around
for
a
bit
afterwards
and
you're
welcome
to
stick
around?
Let
them
know
you
want
to
get
a
team
photo
taken.
A
A
The
code
lounge
will
be
here
kind
of
this
half
of
the
room,
so
I
know
some
of
you
said
we
didn't
get
a
chance
to
finish
everything
or
you
know
we
have
other
things.
We
want
to
continue
working
on.
You're,
welcome
to
use
that
space.
If
it's
helpful,
you
can
also
schedule
time
there's
a
page
for
the
code.
Lounge
you'll
find
it
in
the
hackathon
wiki,
and
so
you
can
sign
up
there
and
welcome
you
to
do
that.
A
How
to
thought
you
know
next
time.
I
think
what
we're
going
to
do
is
we'll
give
a
time
penalty
to
anyone
who
doesn't
get
their
slides
in
like
well
one
less
minute
or
maybe
one
more
minute.
If
you
do
get
your
slides
in
in
time
and
a
bit
of
an
incentive
there
other
than
that,
I
think
we're
we're
done.
Thank
you
all
for
coming
here.
Early
spending
your
weekend
with
us
I
hope
you
got
a
lot
of
value
out
of
the
hackathon
I.
A
Think
I've
certainly
seen
a
great
sort
of
impact
of
the
hackathons
had
on
on
what's
happening
in
the
working
groups.
Please
do
bring
back
what
you
learned
into
your
working
group
feel
free
to
upload
more
like
your
presentations.
If
you
have
things
you
want
to
add
to
them,
you're
free
to
do
that
even
after
the
hackathon
ends,
that's
all
good,
because
that's
a
great
resource
and
we
have
meet
echo,
recording
all
this
so
we'll
have
that
too.
In
case,
you
want
a
recording
of
your
presentation.