►
From YouTube: IETF115-HACKATHON-20221106-1400
Description
HACKATHON meeting session at IETF115
2022/11/06 1400
https://datatracker.ietf.org/meeting/115/proceedings/
A
A
A
A
A
A
B
C
D
E
F
H
H
A
J
K
N
F
Right
ready
to
go
okay,
all
right!
Well,
thanks
everyone,
as
you
know,
the
the
hackathon.
The
whole
thing
is
a
bit
of
a
a
flurry
of
activity
and
getting
things
together
quickly
and
that
that
goes
for
our
the
way.
We
share
the
results
and
the
recording
of
all
that
it's
kind
of
always
a
a
test
and
a
a
bit
of
an
experiment,
so
I
think
we're
all
set
to
go
thanks
for
your
help
in
getting
the
presentations
uploaded
and
and
we'll
go
ahead
and
get
started.
F
First
of
all,
just
a
reminder:
I
know
some
of
you
weren't
here
yesterday
morning
when
we
started,
but
we
do
have
our
our
Mass
policy.
You
need
to
wear
a
mask
when
you're
presenting
up
here,
you're
free
to
take
it
off,
take
it
off
as
I
have
now,
but
when
you're
at
your
seat,
unless
you're
eating
your
drinking,
please
please
wear
your
mask
and
wear
it
properly.
Just
make
everyone
else
safe
and
comfortable.
We'd.
F
Appreciate
that
a
reminder
of
the
note
well,
the
the
presentations
you'll
be
doing
here
do
fall
under
the
note.
Well,
just
like
the
work
you
do
within
the
working
group
sessions
and
and
whatnot,
so
a
reminder
of
that
this
is
where
we
are
in
the
agenda
and
so
again
with
these
results.
Presentations
I
think
it's
great
that
so
many
of
you
have
put
them
together
and
you
are
doing
them.
F
Our
goal
is
to
be
able
to
to
get
through
them
all
relatively
quickly,
we'll
try
to
keep
the
things
flowing
smoothly
up
here
to
the
best
of
our
abilities,
and
we
really
ask
that
you,
you
keep
your
presentation
to
four
minutes
at
the
very
very
most
I'll
be
running
a
timer
there
and
I'll.
Let
you
know
when
you're
down
to
two
minutes
when
you're
down
at
one
minute
and
when
your
time's
up
we're
really
going
to
have
to
cut
you
off,
unfortunately,
just
just
to
leave
time
for
everyone
else.
F
So
sorry
about
that,
but
please
you
know
try
to
be
brief
and
know
that
there's
other
opportunities,
including
hack
demo,
happy
hour
tomorrow,
where
you
can
have
much
more
in-depth
conversations
with
everyone
and
with
that,
oh
one,
more
thing.
If
you
haven't
already
uploaded
your
presentation,
please
upload
it
into
GitHub.
A
A
A
A
P
E
E
E
E
What
we
try
to
understand
in
this
hackathon
if
the
quick
speed
is
supported
over
internet
by
the
mainly
by
the
Ott
or
any
kind
of
company
web
server,
and
we
made
that
implement.
We
have
already
made
an
implementation
of
in
our
open
source.
Browser
in
this
case
is
chromium
and
with
the
modification
for
support
of
the
quick
spin
beat
here,
we
made
some
tests
over
the
Internet
versus
the
over
the
top
web
servers
and
mainly
they
doesn't
support.
E
This
spin
beat
because
it's
an
optional
parameter
in
the
standard,
but
we
made
some
tests
versus
an
openly
speed
web
server.
That
is
an
open
source
project,
and
this
is
successful.
We
got
the
spin
bit
running,
as
you
can
say,
in
the
screenshot
in
the
right
bottom
of
the
right.
The
spin
beat
is
spinning
and
we
obtain
some
delay
in
solvents
with
a
left
LTT
that
is
from
the
client
side
and
on
the
right,
LTT
runt,
replace
it
is
related
to
the
mainly
on
the
internet
side
and
is
quite
presses
affordable.
E
This
is
the
other
thing
that
we
try
to
investigate
in
this
hackathon
and
we
try
to
implement
a
new
delay
measurement
technique
and
these
new
experimental
technique,
that
is
a
modification
of
the
spin
bit,
requires
only
the
the
the
modification
of
the
client
of
the
client
side.
It's
the
realized
on
internal
protocol,
LTT
evaluation.
We
always
use
okay.
We
always
use
in
this
case
the
quick
and
the
user,
the
same
square
wave
mechanism
that
is
the
with
the
alternate
marking
the
of
the
spin,
beat
with
some
preliminary
tests,
which
we've
done
here.
E
We
obtained
some
good
results.
That
is,
we
can
obtain
also
the
delay
measurements
and
the
mainly
goal
of
this
implementation
of
this
method,
and
we
can
use
this.
This
alternate
technique
for
the
delay
instrument
only
by
the
modification
of
the
client,
and
so
we
don't
need
any
way
any
anymore.
A
support
by
the
server
side,
that
is,
we
don't,
doesn't
need
to
have
the
the
support
from
the
reflection
bit
on
the
server
side.
E
E
A
A
Q
Okay,
how
do
you
do
the
next
one?
Oh
this
one:
okay,
thanks
oops
yep,
so
we're
working
on
three
different
extension
header
testing
things.
Two
of
them
will
present
him
at
V6,
Ops
and
one
of
them
at
ippm.
We
have
quite
a
quite
a
crew
working
on
this,
a
number
of
students
from
from
India
as
well
who
who
are
working
with
us,
and
so
our
plan
we're
trying
to
do
some
cloud
provider
testing
at
the
hackathon
decide
on
the
registration
protocol
that
we
need
for
our
encrypted
PDM.
Q
And
then
we
did
some
presentations
with
ebpf
and
free
router,
which
the
students
in
India
were
working
on.
Essentially,
the
the
the
point
of
our
project
is
to
figure
out
if
IPv6
extension
headers
can
be
used
on
the
internet
and
we
did
a
presentation
at
iepg
on
them
and
we
feel
that
it's
really
important.
Q
That
extension
headers
work,
because
it
is
a
very
good
way
that
we
will
have
to
measure
encrypted
Networks
and
what
we've
been
doing
is
to
figure
out
if
they're
being
blocked
exactly
where
are
they
being
blocked,
and
we
will
have
a
whole
side
meeting
on
Thursday
an
hour
to
talk
about
the
the
whole
situation.
It
seems
to
us
that
there's
three
different
topologies
involved
bottom
line.
Q
If
you're
doing
just
client
internet
server,
things
seem
to
work
well,
if
you're,
if
you
have
a
CDN
provider
or
if
you're
in
a
cloud
provider,
environment,
things
work
less
well,
and
so
we
need
to
work
together
with
these
people
and
the
way
we're
doing
our
testing.
Is
we
used
ebpf
to
create
a
stack
which
will
append
an
extension
header
of
our
choice
to
each
packet
going
in
and
out
of
the
system?
Q
Q
R
We
discussed
that
the
care
query
type
should
be
text
instead
of
no,
so
you
could
have
different
ttls
for
different
errors
and
also
the
error
code
was
moved
towards
the
domain
name
to
facilitate
that
as
well.
We
thought
of
a
mnemonic
for
the
edns
option:
DRC
DNS,
error,
reporting,
Channel
and
overall,
the
the
draft
has
better
language
now,
so
there's
implementation
in
inbound
built
upon
Also
earlier
hackers
of
work.
You
can
find
it
in
this
GitHub
tree.
We
have
a
open
testing
resolver,
responding
to
the
edns
option.
There's
also.
R
We
also
have
a
authoritative
site
sending
option
in
ebpf
which
just
works
for
any
authoritative
survey.
It
augments
the
responses.
With
this
option
there
was
a
implementation
done
by
Stefan
bertzmeyer
in
the
authoritative
names
of
a
drink.
It's
a
dynamic
name
server.
It
can
do
stuff
when
it's
queries
for
stuff
and
it's
great
for
experimentation.
Here's
the
GitHub
link
for
it
and,
besides
sending
the
Eds
option,
Stefan
also
started
with
doing
DNS
errors,
reporting
processing
so
doing
something
with
the
reports.
R
R
A
A
B
B
Yeah
I
tried
that
okay,
so
what
strip
wireless
network
based
equivalent
of
a
auto
license
plate
for
unmanned
aircraft,
but
the
media
love
to
call
drones.
The
Baseline
standards
are
external
to
ietf,
but
the
problem
is
the
Baseline.
Standards
are
not
well
informed
from
the
perspective
of
cryptographic
network
protocols,
so
thus
the
drip
working
group
which
Builds
on
the
Baseline
external
standards
to
to
provide
those
properties
so.
B
So
the
goals
for
yesterday
and
today
we
we've
had
to
be
fairly
careful
with
our
language
in
our
graphs,
and
so
we
have
endorsements
that
are
largely
in
the
sense
of
the
rats.
Working
group
and
problem
is
a
lot
of
the
work
that's
been
done
by
Adam
and
myself
is
for
an
employer
who,
unfortunately,
will
not
release
it
as
open
source,
and
so
in
the
interests
of
moving
this
forward
within
ietf.
We
want
open
source,
so
our
volunteers
worked
on
an
independent
implementation
that
will
be
released
as
open
source.
B
It
already
been
uploaded
and
this
will
produce
endorsements
and
Import
and
Export
them
in
the
necessary
formats
for
over
the
air.
Did
the
work
in
Python
to
get
it
done
in
two
days
and
also
the
linkaping
university
implementation,
which
is
open,
source
of
drip
and
open
hip
needed
to
be
brought
up
to
the
latest
draft
level
and
yeah?
Our
volunteers
got
it
done
and
demonstrated
this
morning.
A
S
S
So
what
do
we
got?
Basically,
we
now
have
an
API
and
also
a
website
which
uses
this
this
API
and
you
can
enter
any
quick
domain,
for
example,
google.com,
and
then
you
see
the
results,
how
the
quick
handshakes
behave.
So,
for
example,
here
now
the
quick
handshake
with
Google
takes
multiple
round
trip
times
because
of
large
certificates.
S
We
have
cloutra,
which
has
a
slight
amplification
of
four
or
a
small
amplification.
Then
we
also
have
the
compression
which
I
talked
about.
So
basically,
we
show
how
efficient
the
the
compression
levels
are.
So
here
in
this
example,
we
save
around
30
of
data
if
we
activate
a
certificate
compression.
S
So
in
the
end
we
have
an
open
website
which
currently
is
under
accessible
under
my
private
domain,
but
you
can
also.
We
are
currently
moving
to
understand
the
quake.net.
So
please
check
out
our
tool,
our
API,
and
if
there
are
any
tests
which
you
think
are
missing
just
talk
with
us
and
in
the
end
we
this
is
not
about
finger,
pointing
we
just
want
to
Foster
the
discussions
about
quick
handshakes
and
improve
the
deployments.
P
P
There
are
three
modifications
about
my
our
project
and,
firstly,
yankalaga
Pro
Wife
online
young
customer
meditation
tool,
which
can
then
data
via
modules,
regardless
dependencies,
but
there's
no
offline
tool
can
do
it.
Second
only
provides
some
query
interfaces,
it's
difficult
to
fully
meet
the
requirements
of
a
young
model
of
Automation
and
thirdly,
no
two
is
a
variable
to
compare
two
revisions
of
your
model.
According
to
customize
the
compression
rules.
P
Our
goal
is
developing
a
young
automation
tool
to
compare
their
models
by
reserving
dependencies
automatically
and
provide
the
product
plugging
system
to
all
the
customized
functions,
for
example,
for
the
youngster
communication.
According
to
colonized
conversion
rule
the
related
drugs
and
active
and
leave
below
next.
P
This
figure
shows
the
architecture
of
a
young
compiler
the
model
to
be
compared
as
input,
and
then
compiler
will
search
that
the
dependencies
from
a
local
recall,
if
not
found
young
Cabela,
will
request
the
remote
Ripple
fetch
the
needle
difference
to
a
local
repo
and
then
after
build
building
the
young
female
contests.
Young
fellow
Will
will
call
the
plugins
to
perform
extended
functions
next.
P
P
U
T
T
Hello,
everyone
I'm
from
the
IRC
from
South
Korea,
so
our
hackathon
activities
is
to
verify
our
consideration
very
much
in
their
performance
container
infrastructure
drive,
so
we
he's
got
a
different
considerations
that
can
affect
cutting
a
level
performance,
including
network
access
and
model
and
different
department.
Configuration
settings
next
slide,
please.
T
So
in
the
previous
cartoon
we
have
already
start
initial
tests
of
the
pdpf
a
service
model
performance
with
the
OBS.
If
sdp
support,
V
switch
and
easy
hackathon,
we
continue
to
explore
other
variations
of
EPF
acceleration
model.
They
are
VDP,
V
switch
with
a
memory
interface,
the
Intel
quality
data
plane
with
a
coded
I,
have
HD
plugin
and
a
CD
xip.
T
V
R
T
For
the
cndp
and
the
Champions
limited
indicate
users,
pay
user,
hdb
socket
original
one
and
the
Chipmunk
is
displaying
container
using
the
their
own
communist
plugin.
So
these
are
plugin
move
the
network
device
from
the
horse,
necklace
ready
to
support
network
display
directly
attacked
the
network
device
support,
so
the
quad
can
socket
next
slide.
Please
for
serums
and
Serum
is
the
only
one
that
utilize
the
EBP
app
for
both
north
south
and
Eastward
traffic
acceleration,
the
EPF
at
the
network,
big
drivers,
before
the
muscle
and
at
the
soccer
layer
for
Israel
next.
W
T
T
This
is
why
you
learned
so
we
compare
OBS,
we
host
VP
in
a
bit
and
cndp,
so
VP
outperform
obvious
because
of
the
better
performance
of
memory
against
the
V
host
and
the
cndb
auto
will
have
different
implementation
of
ETF.
You
still
cannot
catch
up
the
performance
of
PvP
with
Hyper
package
sign
this.
W
T
For
the
psyllium
serums,
I
have
already
published
their
own
very
much
gain
for
their
cni,
and
this
year
you
can
simply
refer
to
their
result,
so
the
figure
here
so
the
most
cell
and
Israel
traffic
acceleration
using
eppm
on
silu,
let's
type
in
so
for
the
future
work
we
in
finalize
the
Trap
and
I
saw
reviews.
We
would
like
to
welcome
any
questions,
comments
and
contribution
to
our
draft
to
start
the
publishing
adoption
process,
the
service.
A
J
J
J
This
we
have
two
parts:
next,
okay,
and
we
have
two
parts
of
this
work.
One
is
the
capital
platform
of
the
simulation
platform
and
the
next
is
the
our
platform
from
Nanjing
University
and
about
the
quality
platform.
We
have
three
passions
and
one
is
the
generation
of
satellites,
the
topology
of
the
network,
and
the
second
is:
we
use
the
topology
to
calculate
the
routine
in
the
N3
and
the
third
part
is.
We
will
give
a
resolution
of
the
satellites
and
about
the
actual
Star
platform.
J
Okay,
our
development
is
based
on
the
Ubuntu
links
and
use
the
pattern
and
the
G
and
G
CC.
Okay,
and
this
is
we
have
three
yeah:
let's
just
introduce
them
before
and
and
in
the
future.
Maybe
we
have
more
work
about
the
like
the
orbit,
Dynamic
and
the
evaluation
of
the
constellation
performance
and
the
simulation
of
ICT
and
other
traffic
performance
in
the
future.
Maybe
we
can
do
and
there's
a
lot
of
balance
and
routine
awesome.
Look
at
this
okay.
This
is
our
demo
result.
We
have
no.
We
have
three
types.
J
Okay:
okay
from
the
duties
of
the
action,
we
I
think
the
simulation
platform
is
very
helpful
to
the
solid
solution,
Network
and
other.
Maybe
in
other
other
area
we
can
use
the
platform
simulation
another
thing
and
in
the
future,
I
think
we
will.
Maybe
we
will
have
the
adoption
more
routine
hotels
to
the
simulation
platform
and
to
the
satellite
networker,
and
we
have
another
TBR
go
phone
yeah.
This
is
both,
and
maybe
we
have
some
salad
use
case
and
to
discuss
in
the
off
and
maybe
the
topology
change.
J
K
K
Yeah,
so
the
goal
is
to
optimize
XR
flows,
the
steering
of
XR
Flows,
In,
The
Edge
cloud,
and
the
the
idea
of
this
hackathon
is
to
use
Alto
to
get
the
state
of
the
network
and
complement
that
with
bottleneck,
structure,
graphs
which
are
currently
not
in
the
standard
it's
been
discussed,
but
it's
not
yet
in
the
standard.
So
it's
exploratory
in
terms
of
being
able
to
explore
and
leverage
some
of
these
capabilities
from
the
DSG
graphs.
K
These
are
the
the
RCs
and
the
including
a
draft
and
bond
like
a
structure,
graphs
and
so
the
the
challenge
here
is
about
integrating
end-to-end
the
whole
workflow,
which
I'm
going
to
present
next
yeah.
So
that's
the
workflow
so
starting
from
a
multi-domain
network,
mostly
The
Edge
Cloud,
which
could
be
Wi-Fi,
5G,
4G
and
eventually
6G.
It
is
to
close
the
loop.
So
on
the
right
side,
we
have
the
host
and
the
user
equipment.
K
If
you
will
running
an
XR
application,
then
you
would
have
you
know
some
kind
of
source,
routing,
algorithm
segment,
routing
and
then
to
close
the
loop.
You
know
we
have
netflow
or
you
could
use
also
as
flow.
Then
we
have
the
computational
bottleneck,
structure
graphs
here,
which
is
the
what
that's
sort
of
the
math,
and
then
we
fit
that
into
the
alto
server,
which
is
the
the
standard.
K
So
yeah,
that's
the
the
topology
that
we've
been
testing
against.
It's
a
5G
topology.
It's
actually
extracted
from
an
actual
deployment
in
Philadelphia,
but
it's
everything
is
running
in
an
emulation
mode.
Actually,
this-
and-
and
here
we
have-
you
know
there-
is
an
edge
Computing
resource
here.
This
could
be.
K
So
what
we
show
here
is
you
know,
you
start
the
Excel
application,
it's
getting
getting
10
units
of
bandwidth,
then
congestion
kicks
in
on
that
path
and
then,
while
next
structural
graphs
recompute
in
real
time,
the
the
Dynamics
of
the
system
find
that
there's
a
better
path
that
still
maintains
the
latency
requirements
so,
and
so,
as
congestion
builds
up,
then
eventually
finds
the
better
path
and
then
switches
back
to
to
find
to
to
higher
throughput
pass.
So
the
blue,
the
blue
flow
here,
is
the
XR
flow.
K
You
know,
as
congesting
kicks
in
then
eventually
increases
back
to
finding
to
reroute
reroutes
and
finds
a
higher
throughput
path
to
maintain.
You
know
at
any
point
in
time
giving
you
the
highest
throughput
available,
so
what
we
learned,
yeah
so
bound
next
structures.
Graphs,
as
I
mentioned,
are
currently
been
discussing
the
alto
working
group.
They
are
not
part
of
standard,
but
this
is
a
demo
that
allows
us
to
sort
of
bring
some
some
of
the
the
foundations
or
the
understanding
of
how
this
would
work
in
integrated
end-to-end.
K
With
with
to
close
the
loop,
you
know,
integration
with
with
that
flow
with
as
flow
and
and
Alto
and
the
source
routing
algorithm.
The
demo
show
that
it's,
you
know,
that's
feasible
to
run
this.
Actually,
you
know
in
real
time
finding
the
the
you
know
the
congestion
dynamics
of
the
problem.
We
have
actually
a
deployment
of
these
introduction
Network
at
the
national
research
platform
in
the
US,
and
so
the
demo
provides
some
practical
feedback
into
the
auto
working
group.
K
The
the
idea
is
that
bottom
infrastructure
graphs
provide
a
compact
and
a
scalable
approach
to
incorporate
traffic
engineering
information
into
the
out
of
standard,
which
is
something
that
the
ultra
standard
currently
doesn't
do,
but
it's
something
that
we're
consuming
and
that's
the
theme.
So
this
is
most
people
from
Qualcomm
Yale,
University,
Caltech,
UCSD
and
Sichuan
University
and
Huawei.
B
A
I
I
I
Then
the
third
problem
was
the
automatic
selection
of
a
crypto
backend,
the
pending
of
the
hardware
capabilities,
and
that
is
now
integrated
into
the
build
system
of
Riot
with
kconfig
and
make
files
and
the
problem
of
integrating
the
PSA
architecture.
Test
suit
was
solved
by
integrating
the
test
as
a
so-called
Riot
package.
I
I
Then
the
third
project
we
were
working
on
is
was
a
Schick
plug
test
between
the
riot
implementation,
which
uses
the
external,
lipstick
and
openshik
at
the
openshack
table
on
Saturday.
We
basically
just
did
preparations
and
tried
to
find
out
what
was
possible,
and
then
we
found
out
that
on
both
sides
with
compression
rules,
there's
some
handling
that
still
needs
work
on
the
lipstick
side.
I
H
H
We
got
just
in
time
one
implementation
for
Huawei,
Huawei,
vrp
and
another
implementation
done
by
inza
of
the
VPP,
so
the
first
topology
we
built
for
the
Huawei
vrp
topic
was
composed
of
eighth
notes.
We
see
the
egress
nodes
on
the
left
upper
corner
there,
where
we
configured
three
three
different
srv6
traffic
engineering
policies.
Where
we
push
these
City
lists,
which
accomplished
then
the
three
different
colored
paths
which
we
want
to
see
then
individually,
visualization
at
the
end
of
the
pipeline.
H
L
The
second
topology
we
tested
was
the
vpp1,
so
we
had,
the
topology
is
a
simpler
one.
We
have
only
three
nodes
with
a
cement
routing
policy
between
the
first
note
and
the
third
one
third
one,
and
we
are
using
traffic
engineering
to
send
the
packets
so
through
the
topology.
The
goal
here
was
to
implementing
the
ipfix
export
of
this
samaritin
header.
So
on
each
note
at
ingers
interface,
we
are
aggregating
the
data
for
for
each
SM
routing
flow
and
exposing
it
through
ipfix.
H
Exactly
so
at
the
end
of
our
pipeline,
which
actually
had
also
to
be
enhanced
to
decode
these
new
elements,
we
have
this
nice
visualization.
Where
we
see
you
cannot
see
it,
but
you
can
see
it
if
you
download
the
PDF
later
on.
You
see
in
the
first
number
the
segment
list
so
which
corresponds
to
the
politics
of
the
three
different
policies
we
had
before
and
then
for
this.
L
Yeah,
so,
to
conclude,
what
we
achieved
through
this
hackathon
was
testing
this
semi
routing
policy
within
different
topologies.
We
validated
the
this
ipfix
draft
and
also
testing
the
whole
pipeline
So
reading
from
the
database,
the
collected
data
and
a
funny
thing
we
we
also
tested
the
actual
use
case.
We
had
two
two
members
of
the
team
working
on
the
database
and
two
members
working
on
the
topology
and
the
database
team
actually
saw
these
changes
and
oh,
there
was
a
change
in
the
network.
L
L
A
X
So
I'm
Maxim
this
is
Francois,
so
we're
a
PhD
students
from
luva
working
with
our
advisor
officer,
Olivia,
Bonaventure
and
so
we've
been
working
on
on
Nesquik
and
Nesquik
is
not
just
a
cocoa
powder
that
you've
been
seeing
on
the
table.
It's
also
a
tool
that
we
are
shaping
up
to
do:
Network,
testing
or
internet
performance
testing
with
quick.
So
the
goal
for
the
akaton
was
to
try
to
do
quick
measurements
with
mlab
NDT,
so
mlab
NDT
is
a
tool
from
measurement
lab.
X
So
we
did
actually
quite
well.
So
we
have
a
working
prototype
which
integrates
quick,
go
the
the
quick
implementations
into
the
entity
server
and
on
the
client
side
we
have
switched
from
websockets
to
web
transport
in
the
NDT
JavaScript
client-side
code,
and
so
thanks
to
Google,
Chrome
and
and
and
web
transfer
go.
We
were
able
to
make
the
two
discuss
together
and
so
now
we
have
some
proof
of
concept
for
download
and
upload
tests
using
HTTP,
3
and
quick
right
from
the
from
the
browser.
X
So
that's
just
one
of
the
first
points
so
why?
Why
are
we
doing
this?
Are
we
interested
into
doing
a
quick
internet
performance
measurements
because
we
speak?
We
can
have
much
more
precise,
metrics
about
delays,
loss,
loss
patterns
and
stuff
like
that,
and
then
we
will
need
to
in
that
way.
We
will
need
to
update
also
like
how
we
get
metrics
from
the
quick
Stacks
up
to
the
to
this
measurement
performance.
X
So
look
at
what
we
can
have
similar
to
TCP
info
and
benefit
from
all
that
and
to
do
in
end-to-end
measurements
over
the
Internet,
because
quick
goes
through
any
path
that
is
in
the
way.
So
that's
it
I
guess.
If
you're
interested
in
that
go,
go
come
chat
with
us,
we
will
be
staying
at
the
next
quick
table
for
for
a
while.
Thank
you,
foreign.
G
N
Yeah,
so
a
lot
of
interrupt
testing
has
been
done
too
much
to
explain,
I
think
so,
with
the
mainly
a
lot
of
congestion.
Controls
have
been
tested
on
different
platforms
and
also
a
lot
of
congestion
controls
I've
been
testing
against
each
other,
whether
they
converge
to
the
same
rates,
which
is
of
course
important
if
multiple
applications
share
the
same
bottleneck
that
they're
evenly
dividing
the
the
flows-
and
we
still
are
here
for
a
few
days
more
after
the
hackathon
to
do
further
interrupt
tests
going
on.
N
Maybe
one
of
the
results
and
to
show
you
why
it's
important
to
have
alpharettes
on
a
typical
5G
Network,
you
see
at
the
top
the
latencies
which
are
with
classic
TCP
up
to
400
in
reality,
even
above
a
second.
If
you
really
do
downloads
with
detail
for
us
here,
we
have
a
99
that
below
a
few
milliseconds,
two
three
99.99,
four
milliseconds,
you
can
see
it
in
the
slides
for
the
details
and
I.
Think
then.
G
Sure
so
we
have
quite
a
number
of
organizations
participating
this
time.
This
is
on
the
slide.
I
want
to
read
through
them
all,
but
and
in
terms
of
actual
individuals
listed
here,
21
folks
that
are
joining
us
most
in
or
many
in
person.
There
are
a
few
that
are
remote,
participating
as
well
and
then
finally
to
wrap
up
this
isn't
the
end
of
interoperability,
testing
for
l4s
we,
this
is
actually
this
I
mentioned
the
second
one
at
ietf.
G
There
actually
was
one
in
the
interim
between
last
IHF,
and
here
that
was
hosted
at
cable
labs
in
in
Denver.
We
do
plan
a
second
one
in
Denver
at
in
January
and
then
we'll
look
at
scheduling
at
ihf116
at
Yokohama.
So
if
you're
interested
in
participating
reach
out
to
me
and
I'll
help
you
get
involved,
thank
you.
A
H
A
Y
Y
Y
Okay,
I
pressed
the
button
and
nothing
happened.
Yes,
okay!
Suddenly!
So
what
have
we
achieved?
Over
this
weekend?
We
cleaned
up
the
the
GitHub
repo
for
project,
merge
a
few
development
branches
cleaned
up.
You
know
a
dead
code,
Etc
improved
the
documentation.
We
have
a
tutorial
to
go
with
the
the
code,
which
is
a
step-by-step
instructions
on
how
to
run
examples
and
get
acquainted
with
the
code.
Y
We
also
had
one
participant
designing
a
connector
between
openshack
and
a
real
Network
server
for
sigfox
Network,
so
we
can
exchange
packets
over
a
real
radio
network
and
we
did
interrupt
testing
with
the
riot
team.
As
was
mentioned
before
we
exchanged
a
few
compressed
packets,
both
ways
very
simple.
It's
a
start,
but
good
news,
and
also
working
on
the
interrupt
between
the
micropy
and
clean
slate
implementation
and
the
Legacy
open
check
implementation
and
what
we
learned.
Having
a
newcomer
with
the
rfc's,
you
know,
the
fresh
look
is
always
refreshing.
Y
Lots
of
good
questions
came
up
about.
You
know
how
the
rfcs
worded
and
also
design
choices.
Where
did
you
do
this
this
way?
Or
that
way?
And
you
know
we
could
do
things
in
a
slightly
different
way,
using
the
same
toolbox,
because
it's
a
generic
protocol,
and
so
this
led
to
some
ideas
we
might
extend.
Y
The
young
model
do
little
protocol
extensions
to
do
things
differently,
which
might
be
better
on
a
very
constraining
device,
for
example,
and,
of
course,
we'll
bring
that
feedback
to
the
working
group
during
the
week
or
at
an
interim
later
on,
and
that's
it.
That's
a
team
with
two
newcomers
and
happy
to
talk
to
you
about
open
check
if
you're
interested.
U
Thank
you,
okay,
yeah,
so
T
cozy
is
a
an
implementation
of
cozy.
Cozy
is
a
seaboor
object,
signing
encryption,
a
a
new
message,
a
new
format
for
signing
in
encrypting
messages.
So
tikosi
stands
for
trusted
cozy.
It's
a
c
implementation
of
1952
and
1953
suited
for
small
devices
and
small
memory
use
it's
aiming
for
commercial
quality
and
it
uses
it
can
use
either
the
open,
SSL
or
embed
TLS
crypto
libraries.
U
So
that
was
our
progress
for
the
for
the
hackathon
and
then
briefly,
there's
other
progress.
It
wasn't
kind
of
really
hackathon
related
since
114,
and
that
was
RSA
in
eddsa
was
integrated
into
ticos,
a
1.0
that
will
eventually
move
to
ticos
a
2.0
and
just
this
GitHub
project,
T
cozy,
so
happy
for
to
use
it
or
or
contribute.
Thank
you.
A
Z
C
C
So
the
open
Alto
project
is
an
open
source
implementation
of
iitf's
Alto,
and
then
we
also
worked
on
openoutso.org,
which
is
a
running
deployment
of
this
open
Alto
project.
Next
slide,
specifically
open
alto.org
is
used
with
lhc1
a
layer,
3
VPN
for
high
energy
physics
data
out
of
CERN
next
slide,
please.
C
C
Next
slide.
Please
please.
Our
second
achievement
was
on
obtaining
routing
paths.
So
there
are
many
ways
to
accomplish
this,
but
we
focused
on
the
data
plane
using
looking
glasses,
G2
snapshots
and
equivalent
classes.
Next
slide,
please,
on
the
Looking
Glass
implementation,
we
were
retrieving
forwarding
information
bases
from
looking
glasses
in
CERN
and
giant
and
were
able
to
query
the
path
vector
through
that
next
slide.
C
For
the
data
path,
sampling,
driven
implementation,
we
were
able
to
obtain
similar
information
using
equivalent
classes,
and
this
was
in
reference
to
National
research
platform.
Next
slide,
please,
our
third
achievement
on
which
we
made
good
progress
but
was
not
completed
by
the
end
of
the
hackathon,
is
using
open
alto.org
as
a
global
query
orchestration
platform.
So
the
idea
is
that
this
allows
multi-domain
query
processes
in
lhc1.
C
The
process
is
generally
looking
up
the
source
IP
in
irr
to
obtain
the
source
as
clearing
the
appropriate
Alto
server
associated
with
the
source
as
to
obtain
the
as
path
and
then
refining
that
as
path
into
a
general
path,
representation
which
I
won't
go
into
detail
here,
but
a
good
description
can
be
found
in
recent
working
group.
Email
next
slide.
C
And
then
our
final
achievement
was
integration
of
this
visibility
that
we'd
established
into
russio
Source
selection
next
slide,
please
so.
This
first
involved
the
configuration
of
Alta
resources
to
fetch
information
about
resume
replicas
and
then
expressing
how
we
needed
these
Russia
replicas
to
be
sorted
and
final
slide.
Please,
and
we're
able
to
deploy
this
versio
implementation
integration
into
mininet
that
support
that
partially
simulated
lhc1
networks
and
saw
success
in
disintegration.
A
AA
So
this
is
the
poster
for
our
project.
So
basically,
this
failure
shows
the
Premo
of
iTunes
framework.
So
this
time
we
want
to
implement
ipsec
Pro
protection
based
on
sdn,
so
basically
RC
90
61
is
for
one
domain,
IP
stack
upload
protection,
but
this
time
we
proposed
West
and
Easter
Bond
security
facing
interface
for
multi-domain
environment.
AA
So
this
figure
you
can
see.
So
we
have
two
of
domain
domain
and
B,
so
security
control,
a
carbon
domain
a
and,
on
the
other
hand,
security
control,
B
carbon
domain
B,
so
basically
nsf2
and
another
three
are
belonging
to
two
different
domains.
So
we
want
to
set
up
ipsec
Pro
using
ituna
set.
So
basically
you
can
see
security,
controller,
B,
providing
ipsec
Ike
so
parameter
through
security
controller.
A
and
finally,
an
asset
2
also
NH3
can
get
a
security.
AA
Related
parameter
can
get
from
B,
so
this
one
can
be
done
using
our
proposal,
so
this
figure
is
p80
pure
authorization,
database,
Ike
and
security
policy
database.
We
can
snapshot
so
this
one
or
also
snapshot.
So
once
Ike
wasn't
to
protocol
exchange
is
done
and
then
you
know
ESP
packet
traveled
between
two
network
security
functions,
so
in
the
middle,
nobody
can
catch
the
information
because
the
ESP
packet
delivered
okay,
so
we
got
done
so.
During
this
hackathon
we
showed
the
IP
stack.
AA
Security
Association
can
be
done
using
our
itunesf
extension
and
next
step.
We
want
to
also
implement
the
icon
list
case,
which
means
security
control
can
generate
sa
parameters
instead
of
IQ
button
2..
So
this
is
our
open
source
guitar.
This
video
clip-
and
this
is
our
team.
Thank
you
for
your
attention.
Thank
you.
AA
A
E
AA
Okay,
hello,
oh
this
is
the
second
presentation.
The
ipmont
is
a
new
Bob
I
tried
to
introduce
this
buff,
so
the
basically
iPod
6,
moving
object
and
networking
is
after
ipoable
working
group
work.
So
we
want
to
provide
the
v2x
b2b2i
among
many
kinds
of
moving
beakers
such
as
a
weaker
thrusterial
vehicle
or
area
vehicle
drones,
also
the
Marine
Vehicles.
So
this
one
based
on
our
IP
wave
working
good
problem
statement.
This
is
already
almost
ready,
approved
by
IC,
almost
okay,
so
you
can
see
many
kind
of
objects
here.
AA
So
we
want
to
build
the
particular
architecture
using
many
kinds
of
vehicles.
You
can
see
pedestrian
terrestrial
vehicle
entrance
area
vehicle
and
also
Marine
Vehicles.
Something
like
that.
So
we
also
take
advantage
of
storage,
PP
5G
v2x
protocol,
in
addition
to
OCB
mode.
So
this
is
a
poster.
So
basically
we
want
to
prove
the
concept
context
aware:
navigation
protocol
for
IP
based
the
particular
Network.
AA
Also
we
implemented
our
work
using
omnip,
plus
plus
simulation
and
5z
simu
5G,
for
basically
you
can
see
so
the
drones
can
communicate
with
each
other
also
deliver
its
position.
Information
to
server
through
the
in
the
middle,
we
have
a
genoto
b
and
connect
to
the
server
so
cover
collect,
and
then
they
provide
some
efficient
navigation
for
three-dimensional
drone
paths.
AA
So
what
got
done
so?
We
checked
whether
our
5D
stock
can
be
implemented
for
context
of
your
navigation
protocols,
so
we
proved
the
5G
similarity
can
be
used
also,
so
in
the
next
step
we
try
to
implement
a
multi-air
address
as
configuration
and
also
loading
efficiently.
This
is
open
source
and
this
is
a
demonstration.
AA
AB
Okay,
okay,
hi
I'm,
John
Gray
from
entrust-
and
this
is
the
first
hackathon
for
myself
and
I-
think
pretty
much
everyone
on
our
table.
So
we
had
a
lot
of
fun.
So
our
goals
for
this
event
was
essentially
interrupt.
Testing
of
PQ
Keys
certificates,
pkcs
10,
essentially
x509
based
artifacts,
and
also
using
the
new
nist
crypto
Primitives,
dilithium,
Falcon
and
sphinx,
and
also
alone,
but
also
in
combination
with
and
composite
combinations
with
traditional
and
with
traditional
crypto
as
well.
AB
Also,
we
wanted
to
solve
asn1
encoding
issues
to
help
clarify
specifications
and
obtain
experience
using
these
new
algorithms
and
also
provide
an
artifact
repository
so
that
other
people
can
also
use
these
artifacts
for
their
own
testing
purposes.
I
just
referenced
some
RFC
drafts
that
that
we
used
there
was
a
lot
of
them
and
also
there's
been
standards
around
x509
for
like
25
years,
so
we're
also
using
those
okay.
So
what
got
done
so?
We
did
create
a
GitHub
artifact
repository,
that's
actually
in
the
hackathon,
so
you
can
see
There's
the
link
right
there.
AB
AB
We
also
agreed
on
public
and
private
key
encoding,
so
a
little
bit
more
on
that
later
and
the
other
thing
we
have
seven
different
implementations
that
we're
testing
so
I
think
this
is
a
great
success
for
you
know.
First
time
hackathon
we
have
four
vendor
Implement
implementations
and
also
three
open
source
implementations
involved,
bouncy
castle,
openssl
and
python.
AB
So
what
we
learned.
So
there
was
a
lot
of
discussion
in
the
mailing
list
about
octet
string
encodings,
but
there's
actually
it
turns
out
there's
a
sec
one
draft
from
a
while
ago
that
basically
says
to
to
treat
a
bit
string,
an
octet
string
identically.
So
once
you
do
that
it
solves
the
problem,
you
say
four
bytes
as
well,
so
you
know
over
time.
Four
bytes
might
save.
You
know
gigabytes
of
data
so
anyway,
that
was
a
good
way
to
solve
that
and
we
agreed
with
that
for
the
private
key.
AB
We
also
were
having
some
issues
with
octet
strings
wrapping
octet
strings.
We
decided
there's
no
point
in
doing
that
and
we'll
use
the
representation
from
59.58.
The
other
thing
we
talked
about
was
object.
Ids
at
this
point,
there's
no
standard
object.
Ids.
We
know
this
is
going
to
be
standardizing
them
soon,
but
until
then
we
need
to
be
flexible
at
this
point.
So
we've
also
need
to
make
our
implementations
flexible
to
read
different
types
of
voids.
AB
We
also
suggest
using
an
arc,
a
version
and
a
security
level,
so
you
can
actually
encode
some
data
into
the
oids
and
the
most
issues
found
that
we
found
anyway
are
not
related
to
the
PQ
algorithms,
so
the
last
slide
I
just
wanted
to
wrap
up
again.
This
was
pretty
much
a
team
of
first-timers.
There
were
16
of
us
and
yeah.
We
had
a
lot
of
fun.
We
are
planning
to
continue
to
meet
on
a
monthly
basis,
so
the
next
meetings-
Monday
December
5th-
that
12
UTC
time.
AB
A
M
We
then
started
to
plot
the
broad
trends
that
we
could
see
in
the
data
set
both
over
time
and
for
individuals
and
and
answering
other
questions
about
the
data
that
we
gathered.
We
developed
some
tooling
previously
and
during
hackathon
we
identified
places
where
we
could
perhaps
improve
the
documentation
or
the
packaging
for
that
tooling.
M
M
M
The
the
the
sort
of
other
takeaway
that
we
had
was
that
sentiment
analysis
over
technical
language
is
pretty
difficult.
You
know,
phrases
like
Drop
packets
and
kill
process
and
abort
transmission
are
all
on
the
face
of
it
negative
using
the
tilling.
Of
course,
there
are
fairly
neutral
phrases
given
the
given
the
context,
and
we
need
to
work
on
that.
We
need
to
build
a
lexicon
of
these
phrases
that
we
can
identify
as
being
neutral
now,
just
to
wrap
up.
M
We
had
team
members
from
the
University
of
Glasgow
and
from
Queen
Mary
University
of
London.
We've
got
lots
more
information,
we've
published
papers
on
various
other
data
sets
around
the
ITF
at
that
link.
If
you're
interested
in
any
of
this
there's
going
to
be
a
side
meeting
on
researching
internet
standards
processes
on
Thursday
afternoon,
there's
more
information
at
that
link
just
want
to
conclude,
with
a
request
for
your
help,
we're
trying
to
build
a
tool
that
will
help
working
groups
to
identify
suitable
reviewers
for
their
drafts.
M
That's
based
on
context
from
emails
and
from
other
data
sources
that
lets
us
know
the
interests
of
different
participants.
We
then
scan
the
drafts
to
identify
the
topics
and
then
try
and
match
the
two
up.
We've
got
a
tool
that
will
let
you
look
at
different
drafts
and
to
see
if
our
suggestions
are
correct.
So
if
you
scan
that
QR
code
or
visit
our
website,
you
can
use
that
tool
and
give
us
some
feedback
on
it.
Thank
you
very
much.
W
Thank
you,
hello.
My
name
is
Thomas
Dan
and
I
worked
on
vcons
for
this
hackathon
I'd
like
to
I'm
sure
vcons
are
new.
To
most
of
you,
a
v-con
is
a
Converse
is
a
container
for
a
conversation,
and
it
has
four
parts
and
I
really
love
the
sentiment.
Analysis
of
the
last
group
I
want
to
remind
them,
though,
that
if
you
have
a
model
or
a
conclusion
that
you've
used
for
with
customers
data
with
people's
data,
if
they
take
their
data
away,
you
have
to
retrain
your
model.
W
W
So
what
we
did
was
we
used
vcons
to
implement
transcriptions
and
what
we
did
was
we
used
open,
ai's
whisper
program
to
transcribe.
It
then
we
also
use
Capital
one's
data
profile
to
identify
which
personal
information
existed
in
the
the
the
transcript.
Can
we
use
those
two,
those
two
together
to
redact
the
transcript?
W
We
could
have
used
text-to-speech
to
really
depersonalize
it
or
you
wanted
to
save
something
for
ietf
116..
So
what
did
we
learn?
It
basically
Works,
but
we
also
learned
that
maybe
specifying
and
standardizing
a
transcription
object
would
allow
us
to
use
different
transcriptions
engines
and
be
able
to
test
them
more
reliably.
W
I
want
to
show
you
what
we
came
out
with
so
here's
an
example
of
our
of
our
work.
We
use
the
at
sign
for
the
redaction,
and
this
is
actually
from
my
my
particular
job.
Thank
you.
My
my
company
sells
new
cars
for
a
living
and
we
about
a
million
times
a
year.
We
take
people's
personal
information
and
we're
looking
for
better
ways
to
handle
them.
Here's
a
transcript
of
a
customer
conversation,
a
couple
wins
a
couple
losses.
W
Apparently
the
agent's
name
is
not
private
and,
of
course,
that
it
is
but
the,
but
the
car's
name
apparently
is
considered
to
be
private,
and
it's
not,
but
I
will
tell
you
one
thing,
which
is
really
the
point
behind
the
vcon.
If
you
notice
part
of
this
conversation,
he
says
I'm
putting
the
volume
up
because
you're
so
loud,
we
have
100
agents
that
work
all
around
the
the
world
and
we
have
no
idea
what
the
audio
quality
is
between
any
two
participants.
W
W
Okay,
so
for
more
information,
we
have
a
mailing
list.
We've
just
submitted
our
internet
draft,
we're
going
to
be
at
the
heart
RFC
later
on
today,
but
the
best
one
is
the
vikon
bar
buff.
It's
going
to
be
on
Thursday
if
you're,
thirsty
or
interested
in
vcons
or
just
thirsty
come
visit
us
at
the
at
the
buff.
Thank
you,
foreign.
A
O
O
O
We
did
some
work
on
draft
identification
for
ntpv5
so
that
once
we
start
to
make
a
lot
of
revisions
to
these
drafts,
we
can
keep
track
which
implementations
we
use,
which
drafts
and
which
servers
you
switch
drafts.
So
that's
nice
We've
identified
a
few
minor
bugs
in
particularly
the
ntpd
RS
implementation
of
nwv5
and
we've
had
a
lot
of
discussions
on
time,
scales
and
leap
seconds,
because
those
are
always
difficult.
O
O
D
Okay,
so
give
a
brief
overview
of
some
stuff.
I
looked
at
this
weekend
next
slide,
please
so
looking
at
RFC
8032
and
one
of
the
issues
raised
from
the
last
hackathon
is,
that
has
been
suggested,
a
new
series
of
test
vectors
which
might
give
different
results
with
different
implementations
of
adsa
next
slide.
Please.
D
A
F
V
T
V
V
So
what
we've
learned
mostly
is
that
we
had
some
gaps
in
some
of
our
formats.
For
example,
for
in
the
in
the
evidence
we
needed
to
include
some
inbound
reference
to
the
verification,
key,
the
testing
key
and
also
the
interface
between
Verizon
and
Matt
TLS.
We
needed
to
embed
some
sort
of
that's
going
to
be
used
in
our
in
the
TLs
handshake,
so
yeah.
That's
that's
something
to
to
do
and
change
the
documentation
as
well
and
yeah.
V
A
F
Okay,
great
well,
well,
thank
you.
Everyone
for
those
great
presentations
been
a
fantastic
job
running
through
them.
All
may
have
been
a
record
number
I'll
have
to
go
in
and
check
our
previous
totals,
but
really
appreciate
all
of
you
joining
the
hackathon,
all
the
great
work
you
did
and
and
sharing
your
results.
That
was
really
fantastic.
Another
thing
I'd
like
to
invite
you
to
do
is
tomorrow
we
have
hack
demo
happy
hour,
it's
from
six
to
seven
it'll,
be
in
this
room.
F
I
think
this
room
won't
be
quite
as
large
it'll
be
divided
up,
but
we'll
have
some
space
here
with
tables
we'll
have
a
cash
bar,
and
so
what
you
do
is
you
sign
up
and
then
you'll
get
some
space
where
you're
welcome
to
set
up
laptops
whatever
signboards,
and
you
can
talk
to
people
about
your
projects
in
more
detail.
Right
you'll
have
a
whole
hour
where
you
can
go
a
more
in-depth
conversation
about
what
you
did
and
and
discuss
with
them.
F
Some
of
the
finer
points
around
it
so
I
encourage
any
of
you
who
are
able
and
wanting
to
to
do
that
to
take
advantage
of
that
opportunity.
We
usually
get
at
least
10,
or
so
teams
sign
up
and
have
a
good
crowd
of
people
coming
through
some
of
who
participated
in
the
hackathon,
but
a
lot
of
them
who
who
aren't
here,
who
are
just
arriving
now
and
and
maybe
heard
about
what
you
were
doing
so
it
usually
ends
up
being
a
really
good
opportunity,
a
really
good
experience.
F
The
other
thing
I
think
you
heard
the
l4s
team
is
going
to
remain
with
their
stuff
set
up
and
they'll
be
continuing
to
do
some
work
during
the
week.
All
of
you
are
welcome
to
do
that.
We'll
have
space
in
this
room
set
up
kind
of
as
the
the
ITF
Lounge,
but
also
as
the
code
Lounge,
a
part
of
it
will
be
kind
of
separated
off,
we'll
make
sure
we
have
ample
power
and
whatnot
for
you
to
come
down.
F
If
you
want
to
you,
can
you
don't
need
to
reserve
space
or
anything
like
that?
But
there
is
a
a
sign
up
board
where
you
can
just
advertise
to
everyone
else,
but
when
you
plan
to
be
here
just
to
help
coordinate
schedules
a
bit,
so
you
can
go
to
the
the
webpage.
That's
there.
You
can
sign
up
and
you're,
also
free
just
to
pop
down
here
anytime.
So
that'll
continue
throughout
the
week
Monday
through
Friday.
F
F
So
you
know,
thanks
to
the
our
our
sponsors,
for
that
we
were
able
to
to
make
that
happen.
I
think
we
all
appreciated
it
and
for
each
hackathon
we
always
need
new
sponsors
or
are
looking
for
for
new
and
additional
sponsors,
as
you
can
see
here,
we're
able
to
have
more
than
one
that
we
have
sponsorship
at
different
levels.
F
So
that's
all
very
appreciated
and
enables
us
to
continue
to
to
do
this
and
and
to
have
a
nice
setup
and
all
that.
So
please
encourage
if
you
work
for
an
employer
or
anyone
else
who
who
might
be
able
to
sponsor
us
in
the
future.
That'd
be
great
and
I
want
to
give
an
early
reminder
for
the
next
opportunity.
The
next
hackathon
it'll
be
in
Yokohama.
So
what
I
have
up
here
are
the
dates
of
the
entire
iitf
meeting,
I
guess
the
Saturday
and
the
Sunday.
F
It
will
be
the
hackathon,
so
the
25th
26th
that
weekend
of
March
maybe
mark
your
calendars
and
start
planning
your
your
time,
so
that
you
can.
You
can
be
there
we'd
love
to
have
you
there
and
with
that.
You
know
we're
we're
wrapped
up
here,
but
I
do
want
to
say
a
big
thanks
to
I
think
Barry
had
to
run
for
another
commitment.
F
He
was
back
there
until
just
about
a
minute
ago,
but
he
was
helping
behind
the
scenes
with
a
lot
of
your
requests
to
get
added
to
to
mailing
lists,
to
get
added
to
the
GitHub,
to
upload
your
when
you
uploaded
your
presentations
to
the
GitHub,
he
was
getting
him
into
mitecho
so
that
that
no
could
actually
bring
them
up.
The
miteco
folks
were
working
really
hard
because
you
know
it's
it's
a
lot
of
presentations
all
coming
together
at
one
time.
F
That's
not
a
typical
thing
for
a
meet
Echo
session
and
and
I
think
it
worked
relatively
well.
So
so,
thanks
to
everyone
who
helped
make
that
happen,
also
the
Secretariat
and
the
knock
team.
That's
been
helping
us
tremendously
throughout
the
whole
day.
So,
thanks
to
everyone,
congratulations
on
a
great
event
and
I
hope
you
have
a
fantastic
ietf
week,
foreign.