►
From YouTube: IETF Hackathon in London
Description
IETF Hackathons encourage developers to collaborate and develop utilities, ideas, sample code and solutions that show practical implementations of IETF standards. The IETF Hackathon in London on 17-18 March is poised to be the largest ever.
B
C
C
We
came
up
some
quite
some
interesting
issues
that
happen
when
people
run
their
own
mill
servers
with
weird
encoding
issues,
weird
dates
like
30
seconds
of
january
and
unix
time,
it's
beautiful
you
can
find
and
contribute
on
github
on
in
the
big
bang
repo
and
we've
been
working
on
this
with
a
lot
of
academics
and
civil
society
people.
So
we've
been
analyzing,
roughly
2
million
emails
from
almost
two
hundred
thousand
email
addresses
from
955
mailing
lists.
C
There
are
a
bit
over
a
thousand,
but
these
were
the
ones
we
could
crepin
that
were
not
too
weird
a
one
email
address
send
about
80,000
of
those
of
those
emails,
but
most
contributors
sends
one
or
two
messages
which
could
indicate
a
spam
issue.
What
is
interesting
is
what
you
see
here
is
these
are
most
of
them.
Our
mailing
lists,
but
also
some
individual
contributors,
is
that
you
see
that
the
community
is
quite
modularized,
but
also
quite
integrated,
which
is
probably
to
be
expected
and
some
good
news.
There
are
some
outliers.
C
We
won't
tell
you
who
they
are
so
another
question
we
had
of
a
presumption
is
that
mailing
lists
are
often
Domino
dominated
by
early
participants.
That's
not
true!
So
we've
been
researching
into
tenure.
It's
a
complicated
analysis.
Nik
will
tell
you
everything
about
it.
If
you
approach
him
after
a
session,
we've
also
been
analyzing
a
number
of
messages
sent
by
contributor
and
even
though
a
few
people
send
a
lot
of
messages,
there
is
a
pretty
good.
C
Tailor
there's
a
pretty
long
tail
and
a
few
people
are
not
dominating
all
the
lists
here
as
well.
So
what's
interesting
when
we
look
at
individual
email
list
such
as
IETF
at
IETF
dot-org,
if
you
look
at
the
longest
threats
be,
these
are
all
quite
recent,
so
this
indicates
there
is
still
quite
some
active
discussion
and
some
some
of
you
might
be
interested
in
the
discussions
that
actually
brought
up
so
much
stuff.
C
C
We
can.
We
can
also
do
word
trends
and
there
we
can
phrase
and
see
that
there
is
a
steady
increase
in
the
mention
of
privacy.
Security
got
much
more
uptake
and
access
got
popular
and
then
lost
popular
over
time.
You
can
do
with
it
whatever
you
want.
So
if
you
want
to
do
funky
analysis
and
come
to
cool,
tore
conclusions,
join
us
and
make
poll
requests.
B
D
D
Okay,
so
this
is
your
four
for
the
price
of
one
bumper
DNS
presentation:
I'm
gonna
cover
deep.
Can
you
hear
that?
Okay,
sorry
so
I'm
gonna
cover
DNS
dinner
service
discovery,
deprived
and
overall
in
three
minutes,
I'll
be
turned
up
yesterday
morning
about
half
of
the
people
here
we're
interested
in
working
on
DNS
over
HTTP,
which
is
doe,
so
they
looked
at
multiple
implementations
of
the
o3
version
of
the
draft
and
the
other
half
of
the
folks
here
wanted
to
work
on
smaller
projects
in
the
other
areas.
D
So
what
got
done
in
doe,
if
you're
interested
in
nitty
gritty
details,
follow
the
link
at
the
top,
which
will
take
you
to
a
repo
which
tells
you
everything
about
these
implementations,
the
in
overview.
We
had
seven
different
implementations
in
everything
from
C
to
JavaScript,
and
we
also
had
an
interrupt
test
framework
that
got
started
as
well.
D
What
was
learnt
there
will
be
a
full
report
at
the
DOE
working
group,
so
if
you're
interested
go
along
there,
but
the
main
thing
that
we
learned
is
Wow,
don't
you
learn
a
lot
when
you
put
DNS
folks
and
HTTP
people
at
the
same
tables
and
they
write
code
together?
The
headline
is
it
mostly
works?
Most
of
the
Interop
was
okay
and
there's
also
interrupts
with
the
public
servers
that
are
available.
D
For
this,
though,
two
very
specific
feedbacks
of
the
draft
one
was
that
the
draft
didn't
say
enough
about,
should
a
must
on
implementing
getting
posts.
That's
going
straight
back
to
the
authors
and
also
that
oops
error
handling
was
completely
unspecified
in
the
draft.
So
that's
going
to
go
in
there
as
you'd
expect.
D
We
worked
through
a
whole
bunch
of
minor
clarifications
and
bugs
we
had
one
fail
where
dough
was
not
a
good
fit
for
the
go
DNS
library
where
nice,
every
till
I
said
already
been
implemented,
and
also
who
knew
you
can
write
a
pearl
dough,
client
in
20
lines
of
code.
If
that's,
how
you
get
your
kicks
in
the
other
areas
for
DNS,
SD,
Ted
and
Stuart
spent
the
whole
weekend
working
on
the
mdns
relay
implementation.
D
D
There
was
an
epic
fail
in
trying
to
implement
edward
curves
validation,
but
we
thank
Nick
Harper
for
help
with
his
asm1.
In
that
respect,
some
issues
with
Rowan
I've
been
trying
to
implement
the
latest
KS
case
Sentinel
draft
and
there's
also
work
on
a
name
and
wind
up.
We
also
did
some
work
in
deprive
so
they
get
dns
library
and
stubby
can
now
connect
to
resolve
it's
just
by
name
and
this
works
for
URI
for
a
doe
server
or
name
for
a
Taylor
server.
D
We
also
did
a
bunch
of
implementation
work
around
TLS
in
system,
buddy,
resolve
D
and
bind
Dennis
perfect
benchmarking
tool,
those
more
compelling
and
also
an
experimental
chain
server
is
available
and
just
to
put
it
the
names
of
the
teams
that
I
haven't
mentioned
so
far,
we
had
two
remote
participants
and
five
first-timers.
Thank
you
very
much.
E
Hello,
everyone:
this
is
a
tempura
junk
from
seung-hyun
University
in
South
Korea
this
time
my
student
PhD
student,
Jin,
Yong,
Kim,
Kim
and
I
championed
our
eye
to
another
project,
so
I
20s
f
is
for
how
to
apply
neck,
complete
yang
elastin,
complete
yang
to
import
a
high-level
security
policy
into
network
security,
function
level,
security
policy.
So
this
time
7th
grade
student
at
sanguine,
University
six
students
are
participating.
E
This
hackathon,
my
PhD
student
and
I
participated,
and
this
hackathon
and
our
son
code
located
at
this
kid,
have
a
link.
You
can
access
it.
So
this
is
the
poster
and
shows
the
brief,
so
our
city
prefers
professors
and
other
Research
Institute,
dr.
Park,
and
also
Korea
Telecom,
mr.
an
participate
and
also
student
list.
E
So
in
this
custom
project
we
show
the
two
of
things.
The
first
one
is
a
dynamic
configuration
which
means
map
a
security
function
to
actual
metal
key
function
located
at
a
security
controller.
The
second
one
is
a
consumer
facing
interface
based
on
Lester
comp
and
the
lady
lady
stood
our
young
tailor
model.
So
let
me
explain
one
by
one
at
a
time,
so
you
can
see
I
to
NSF.
User
is
other
Minter
writer
of
program,
so
it
has
some
policy
such
as
some.
E
During
our
work
time,
employee
cannot
access
some
other
web
sites
such
as
Google,
YouTube
and
Facebook,
so
that
kind
of
security
policy
delivered
to
security,
con
controller
using
this
kind
of
the
Yankee
remodel.
So
we
can
improve
import
that
policy,
so
a
security
controller
using
dynamic
computation
which
caste
the
security
our
service
is
a
map
into
actual
at
expiry
function
such
as
fire
and
the
Web
Filter
translate
leader,
and
then
that
policy
translate
key
divide,
is
spirit,
controller
and
actual
if
I
were
web
filter.
E
Naturally,
function
are
mapped
so
that
kind
of
skill
function
automatically
a
map,
so
we
used
this
kind
of
open
source
and
also
we
demonstrated
our
concern
using
this
network
configuration
page
so
on
a
mini
net.
Okay.
So
if
access
are
blocking
according
to
user
group,
so
this
is
the
project
github.
This
time
we
showed
our
data
based
on
their
come
from
adult
works.
Thank
you
for
your
attention.
Okay,.
E
However,
let
me
introduce
the
achievement
of
thoughts
interrupts,
so
our
Hakkasan
plan
is
here.
We
tested
the
ability
between
independent
implementations
in
order
to
see
the
maturity
of
these
core
aspects,
specification
of
thoughts
protocol.
Now
we
have
four
implementations.
One
is
a
phase
by
entity,
so
it
is
open
to
the
public
and
rest
three
are
from
in
the
different
vendors
in
the
city
grid
era
and
from
way
so
this
spreadsheet
shows
our
achievement.
We
check
the.
G
E
So
here
is
our
status
so
thoughts
again,
adidas
open,
threading
protocol
protects
the
internet
from
DDoS
attacks.
It
makes
a
distribute.
It
makes
DDoS
protection
more
effective
with
its
programmatic
capability.
So
now
we
confirm
that
we
can
do
cooperative,
DDoS
protection
operation
between
at
least
two
individual
implementations.
E
E
So
here
is
what
we
learned
we
can
idea.
I
can
say
we
can
meet
the
expectations
for
adult
protocol
for
the
markets
in
because
the
core
most
important
specifications
is
almost
stable,
stable.
Then,
in
this
hackathon
we
tested
based
on
the
latest
specification,
so
it
is
proven
to
work
then
also
a
we
discussed
a
lot
about
current
draft
also.
E
B
F
F
Authentication
protocol,
it's
used
typically
for
network
access.
Authentication
here
is
one
example
deployment
where
you
would
use
heap
or
you
might
use.
If
is,
if
you
are
trying
to
connect
to
access
point,
then
you
first
authenticate
yourself
a
deep
and
then
then
you
can
actually
connect
to
it
and
get
internet
access.
So
EEP
runs
end
to
end
between
your
laptop
and
the
server
there's
a
bunch
of
EEP
method.
It's
it's
not
new.
So
over
the
past
10
years
we
have
a
TLS.
We
have
EEP
a
K
which
uses
sim
credentials.
F
So
some
of
you
might
know
that
EEP
method
update
group
was
reach
out
and
restarted.
It
was
actually
just
reach
hearted
last
month
and
during
the
recharter
discussions
there
was
people
reporting
on
the
mailing
list
that
there
are
some
deployment
issues
with
EEP
TLS
and
they
know
they
notice
that
certificates
have
been
growing
larger,
especially
are
with
RSA
keys
of
4
0
9
6
bits
and
the
chains
are
also
getting
long,
and
this
is
the
case
in
many
enterprise
deployments,
and
what
this?
F
What
happens
is
that,
even
even
though
both
the
client
and
server
have
valid
valid
certificates
to
authenticate
each
other,
they
fail
because
access
point
implementations
drop
EEP
sessions
that
don't
complete
within
40
packets.
So
our
goal
was:
can
we
reproduce
this
error?
So
in
that
terms
we
were
probably
the
only
team
that
was
trying
to
break
something
that
already
works.
F
So
we
started
in
the
morning
yesterday
by
trying
different
types
of
configuration.
So
we
had
like
many
intermediate
CS
and
took
to
reproduce
error.
We
try
to
have
very
long
chains
and
large
certificates,
so
we
were
only
using
RSA
keys,
but
we
were
still
unable
to
produce
the
error,
so
it
was
still
authentication
successful
and
we
were
very
disappointed.
So
what
we
did
next
is
there
is
a
configuration
parameter
that
allows
you
to
reduce
the
maximum
a
packet
size.
The
default
was
1396
bytes
and
we
reduced
it
to
200,
bytes
and
voila.
F
We
have
a
failure,
so
the
AP
bars,
as
was
reported
on
the
mailing
list,
but
we
also
noticed
that
for
this
to
actually
happen
with
the
default
1396
bytes
fragment
size,
you
would
need
to
have
chain
that
is
roughly
55,000
bytes.
So
that's
yeah.
We
aren't.
We
send
this
input
on
the
mailing
list
and
we'll
probably
discuss
this
tomorrow
when
he
meets.
We
are
also
continuing
our
tests
with
with
ECC
based
certificates,
to
see
how
how
things
are
we
have.
H
H
For
example,
legato
is
interacting
with
presto,
and
presto
is
communicating
with
that
underlying
infrastructure
to
create
a
flow
or
create
a
link
between
those
switches.
So
here
we
use
mininet.
Hence
we
don't
have
get
time
to
use
that
actual
network
box
devices,
so
we
used
min
in
it
and
in
the
mignonette
we
created
the
flows
which
have
their
two
switches
and
between
the
switch
one
and
switch
two
we're
just
creating
that
EVC.
H
This
is
that-
module.
Ii
is
a
young
based
module
and
is
a
flexible.
We
can
install
it
and
uninstall
it
at
any
time,
and
if
we
install
it,
we
can
see
the
list
of
api's,
for
that.
Legato
is
visible
in
that
API
doc
exploded
and
you
can
see
that
all
those
card
operation
for
that
legato
we
can
get
update
post
all
kind
of
operations.
We
can
do
from
that
either
from
that
Explorer
screen
or
we
can
do
from
that
postman
script,
and
this
is
that
implementation
over
there.
H
So
there
is
a
two
offices
located
one
is
San
Diego.
Another
is
that
Los
Angeles
and
those
two
have
that
two
units
are
defined
and
inside
the
service
provider
network
we're
creating
point-to-point
EVC
over
within
uni
and
uni
two,
and
it's
also
possible
to
create
multi
point
or
EVC
tree.
But
right
now
we
are
just
creating
that
point-to-point
EVC.
So
we
are
establishing
relation
between
link
between
that
even
a
one
and
even
a
and
I'm
sitting
over
here.
If
you
want
to
see
any
demo,
just
let
me
know
come
to
my
spot.
H
I
will
show
that
what
the
steps
we
did
and
I
can
run
that
small
demo
for
that
in
the
decorating
of
EVC
bit
into
a
nice
after
that
uni
we
can
verify
the
flow
that
data
is
going
from.
One
end
point
to
another
end
point
and
after
that,
a
deletion
of
the
DVC
that
flow
traffic
will
drop
down
and
then
again
we
can
verify
that.
Is
it
working
or
not?
Okay,
that's
it
from
my
side.
B
I
So
the
the
plan
for
me
was
to
bring
along
my
java
implementation
that
was
almost
finished
and
used
as
inspiration
for
other
people.
We
with
implementations
of
their
own
in
their
own
languages,
what
we
actually
achieved.
Well,
the
plan
changed
slightly
and
we
developed
an
email
reflecting
service,
this
it's
an
MTA
and
it
transforms
a
regular
incoming
email
and
returns
a
an
email
to
the
the
sender,
which
is
exactly
the
right
kind
of
format
for
a
multilingual
email.
I
According
to
that,
the
new
spec
we
also
developed
some
client
code
that
sits
inside
near-match
or
will
do
when
they
merge
it
in
and
that
correctly
select
the
email
part.
Given
a
new
preferred
language
configuration
that
you
can
set
in
the
client
and
that
will
select
the
right
type,
the
right
part
of
the
email
given
that
configuration
and
also
we
had
a
little
bit
of
feedback
into
my
Java
implementation.
That
was
very
useful.
So
this
was
the
original
plan
that
we
circulated
between
us
before
the
event,
there's
obviously
quite
a
lot
on
there.
I
So
we
selected
two
things,
so
the
top
one
is
the
the
code
in
the
client
side
for
near
mark
so
that
when
you
receive
an
email,
it
shows
the
right
part.
Given
those
preferences,
the
one
in
the
bottom
is
the
reflector
code
so
that
you
can
and
create
those
correctly
formed
emails.
I
was
gonna
skip
over
this
one.
I
So
to
use
a
reflector,
it's
used
just
send
an
email,
plain
text,
email
with
a
specially
formatted
a
recipient
address,
and
then
the
that
code
looks
at
that
address
and
then
sends
you
back
a
an
email,
multilingual
email
with
all
the
parts
and
the
translations
obtained
from
Google
Translate.
So,
for
example,
this
has
got
English,
French
and
Spanish,
and
it
gets
those
from
Google.
Translate
creates
the
multilingual,
email
and
returns
that
to
you,
as
for
the
near
Matt
client
side,
part
selection,
stuff
so
on
the
left
is
what
it
would
look
like
before.
I
B
Just
a
reminder
to
everyone:
I
really
want
to
encourage
you
to
I
know.
Three
minutes
is
not
nearly
a
long
enough
time
to
do
justice.
Please
do
consider
joining
us
tomorrow.
Yeah
Monday,
6
to
8,
we'll
have
the
happy
hour
thing
where
you
can
come
and
talk.
You
know
in
depth
about
your
your
presentation
so
network
time,
security.
J
J
J
Both
are
still
in
progress
and
have
proof
of
concept
character
and
we
performed
yesterday
is
the
first
interoperability
test
to
find
out
if
it
works
or
not.
Our
test
scenario
is
pretty
simple:
we
have
a
small
NTS
client
and
entire
server
two
devices
with
two
different
implementations
and
the
test
results
very
interesting.
We
have
two
part
in
the
communication.
The
first
pass
works
over
TLS
1.2.
J
No,
it
works,
but
the
second
way,
the
time
exchange
over
NTP
failed
at
the
first
test,
because
we
have
misinterpretation
of
CNTs
to
aft
on
the
server
side.
So
this
is
supposed
to
be
a
little
bit
tricky
yesterday,
but
we
fixed
it
and
we
improved
the
NTS
to
aft,
so
in
the
second
one
it
works
correctly
and
we
can
avoid
such
problems
in
the
NTS
draft
in
the
future.
J
Ts
draft
works
is
correct,
other
implementations
works
correctly
and
we
have
some
small
issues.
We
have
to
discuss
some
small
problems,
but
it's
not
a
big
thing
and
yes,
we
have
to
finish
a
tests.
We
have
tests
only
client
and
so
of
this,
our
implementations,
but
we
have
to
test
the
other
way.
Okay,
so
it
was
for
more
informations.
You
can
shown
us
and
github
or
yes,
okay,.
K
So
the
the
area
that
we
actually
worked
on
is
having
to
do
with
delegation
of
HTTP
video
session
and,
in
particular
the
at
the
hackathon.
We
worked
on
a
CDN
I
use
case
of
delegation,
basically
of
content
from
an
upstream
CDN
to
a
downstream
CDN
and
specifically
use
case
of
upstream
CDN,
delegating
HTTP
content
to
the
downstream
CDN.
K
The
key
goal,
of
course,
for
for
us
is
interoperability.
Testing
was
to
ensure
trust.
Boundaries
are
maintained
to
the
right,
as
you
see
here
in
the
picture,
and
also
the
end
user
privacy
is
not
breached.
A
couple
of
other
things
to
call
out
here
is
that
in
the
delegation
model,
the
the
trust
boundaries
are
maintained
and
then
the
redirection,
as
it
happens,
whether
it's
DNS
or
it's
just
simple
HTTP,
redirect
that
can
occur
also
that
the
ISP
cache
is
only
holds
a
certificate
without
the
private
key
of
the
content
provider.
K
In
this
case,
video
at
example.com
and
and
all
the
decryption
material
and
everything
in
the
keys
are
hold
are
held
by
the
large
server
which
basically
returns
the
decrypted
master
secret,
but
not
the
private
keys.
So
the
use
case
really
is
that
we
have
an
end
user
requesting
to
watch
a
content.
From
example,
dot
video
comm
and
the
content
is
hosted
on
a
CDN
of
choice
by
the
content
provider
and
then
delegated
by
it
for
the
downstream
to
the
downstream
CDN,
which
in
this
case
is
the
I
speak
cache
on
the
edge
of
the
network.
K
So
the
the
flow
really
is
that
the
user
would
request
content
and
then
the
through
the
DNS
see
named
allegation
in
this
case
CDN.
Would
the
content
provider
would
redirect
to
CDN
and
the
CDN
would
redirect
it
to
the
to
the
ISP
for
for
the
TLS
handshake,
overlark,
here's
what
we
have
done?
Basically
they're
the
two.
K
The
two
key
aspect
is
that
the
TLS
handshake
occurs
with
the
turn
with
the
lure
client,
which
is
which
we
did
on
the
nginx
server
and
the
between
the
nginx,
which
is
the
lower
client
and
the
log
server
that
happens
over
the
lark
protocol
and
so
just
to
quickly
wrap
up
as
to
what
we.
What
we
did
here
was
for
the
interoperability
standpoint.
We
used
a
client
browser
and
for
the
trial
over
the
last
two
days.
We
really
just
used
chrome
and
ie
the
the
edge
server
software
was
basically
nginx.
B
L
Hello
I'm,
a
teen,
leanness
and
yeah.
Basically,
I
wrote
DHCP
client
for
riot
on
my
own
this
weekend,
so
the
goal
was
to
provide
a
minimal
implementation
for
prefix
delegation
on
the
basis
of
the
current
draft
for
an
update
for
the
dhcpv6
RC,
and
the
motivation
were
that
we
currently
have
a
very
quick
hack
version
of
something
that
resembles
DHCP,
but
isn't
really
the
HTTP,
which
is
called
you
HTTP
to
configure
the
prefix
of
a
6lo
border
router
so
that
it
can
disseminate
its
information
throughout
paper
and
so
for.
L
Our
implementation
will
require
the
message
behavior
for
client
for
solicits,
advertisers,
requests,
replies
and
renews.
We
had
a
very
simple
set
up
to
so
we
just
had
a
write
native
instance,
which
means
that
we
can
basically
run
riot
as
a
process
on
linux
and
kia
dhcp
server
to
run
our
tests
against
and
yeah.
L
So
it's
a
little
bit
more
camera
already
finished.
The
implementation
of
prefix
delegation,
then
maybe
later
on
down
the
road,
provide
more
support
for
other
DHCP
features
and
I
already
PR
to
pull
requests
to
the
riot
repository.
So
if
you
want
to
have
a
look
at
the
code,
it's
about
thousand
lines
of
code,
you
can
do
it
under
this
URL.
Thank
you.
B
B
M
Hi
everybody
I'm
Lucas
party,
I
work
for
BBC,
R&D
and
I
was
running
a
little
umbrella
project,
so
to
speak,
called
making
multicast
IP
more
successful.
So
if
we
think
about
what
our
problem
was
for
the
table
to
solve
is
that
mass
audiences
for
IP
be
IP
based.
Tv
distribution
causes
scalability
issues
for
the
internet,
so
do
some
mathematics,
20
megabits
per
second
per
stream
times.
M
20
million
streams
is
unhappy
for
various
people
around
the
world,
so
one
means
to
a
solution
for
that
is
multicast,
IP
or
IP
multicast,
and
we
think
that's
a
scalable
way
to
address
audiences,
but
no,
it's
been
used
to
date,
but
success
has
been
limited
across
unmanaged
and
heterogeneous
networks.
So
we'd
like
to
reach
audiences
across
disparate
kinds
of
networks.
You
know
we
realized
that
in
two
main
multicast
is
difficult,
are
not
universally
deployed
and
we
have
this
idea
of
multicast
islands.
So
we
want
to
improve
discoverability
of
multicast
beyond,
say.
M
Maybe
traditional
means
like
SDP
and
I'll
come
on
to
that
in
a
little
bit.
One
solution
to
some
of
these
problems
is
automatic,
multicast
tunneling
and
that
helps
you
reach
clients
that
maybe
don't
support
native
multicast
reception,
but
moving
over
to
the
the
streaming
formats
say.
We
think
that
HTTP
based
is
is
the
future.
It's
pretty
good
for
unicast
and
fall
delivery,
enabled
new
kinds
of
content
experiences.
So
what
we
wanted
was
a
solution
to
deliver
this
over
IP
multicast.
So
we
looked
at
something
like
quick
and
thought
that
this
was
you
DB
paste
HTTP.
M
Could
we
make
it
work
over
IP
multicast?
We
wrote
up
a
draft
that
kind
of
does
this
in
a
server
push
context
and
a
BBC
R&D.
We
built
an
end-to-end
prototype
using
Google,
quick
and
we've
had
a
demo
on
our
table
this
week,
showing
that
on
a
Raspberry
Pi
our
ambition
was
to
move
to
IETF
HTTP
over
quick.
So
to
enable
us
to
do
that,
we
took
our
current
demo.
We
want
to
deliver
it
into
this
building.
M
We
had
some
of
my
colleagues
working
on
the
implementation,
a
new
client
library
or
client
and
server
library
to
do
HT
to
be
over
quick
built
on
top
of
ng
TCP
and
we're
also
looking
at
different
kinds
of
clients.
So
could
we
have
a
browser-based
version
of
this
and
try
and
address
some
the
multicast
into
a
browser
considerations
for
these
things
and
anticipating
and
I
mentioned
AMT
earlier,
and
we
had
Jake
Holland's
earning
us
continuing
his
work
at
well
on
AMT
at
the
hackathon
I've
got
an
architecture.
M
Diagram
I
won't
go
into
this
in
too
much
detail,
but
you
know:
multicast
delivery
is
hard
so
to
get
multicast
from
our
home
base
about
15
minutes
away
from
here
we
had
to
go
over
a
GRE
tunnel
and
have
support
from
the
ITF
network
team.
That
was
great.
We
got
that
delivered
to
our
table
and
we
had
various
kinds
of
clients
consuming
that
stream
successfully
and
what
we
got
done
was
we
continued
the
work
we
did
that
end
to
end
test
successfully.
We
made
some
progress.
M
N
Hello,
everyone,
so,
to
give
you
a
little
bit
of
context,
the
problem
we
are
trying
to
solve
here
is
that
on
the
internet
today,
when
you
want
to
retrieve
some
unicorn
picture,
you
just
have
too
many
ways
of
doing
it.
You
can
do
it
through
your
phones
via
Wi-Fi,
so
you're
wired
link
and
sometimes
our
wire
link
is
multi
ohms.
You
can
even
go
through
a
VPN
it
satellites.
There
are
of
many
ways
of
doing
it.
N
You
know
which
wanted
to
use
an
expensive,
the
captive
portal
upstream,
the
fast
upstream
or
the
secure
through
the
VPN,
and
in
order
to
present
that
information
to
to
the
application,
we
use
the
neat
project
which
gives
middleware
such
that
application
can
as
well
retrieve
this
information
from
the
upstream
the
uplink
characteristics.
Now.
One
aspect
is
that
if
one
of
your
uplink
happens
to
be
a
captive
portal,
well
we
have
something.
N
Finally,
one
aspect
is
that
in
ipv6,
when
you
have,
when
you
have
a
multi
home
network,
you
don't
want
to
do
Nats.
So
there
is
a
there
is
some
work
going
on
at
ITF
and
it's
using
a
source
test
routing,
which
basically
means
that
you
are
not
only
going
to
look
at
the
source
the
destination
address,
but
also
the
the
source
address
to
route
the
packets.
So
during
this
hackathon
we
made
a
complete
test
apology
with
all
those
components
plug
together
the
PVD
object.
We
had
three
different
PV
DS.
One
of
them
was
captive.
N
We
had
some
Linux
routers,
some
Cisco
routers
doing
Seder
array
DVD
some
Linux
kernel
was
patched
PDD
GDP
was
patched.
We
also
have
an
implementation
in
open
WT.
So
now
not
all
that
was
done
during
the
hackathon,
but
we
plugged
them
all
together
and
had
some
fixed
system
to
make
it
all
work,
as
well
with
the
neat
project
and
know
we
have
a
wire
Shakti
sector
and
you
can
find
all
that
on
the
github.
We
also
had
some
work
with
the
captive
portal
done
by
Kyle.
B
O
Okay
good
afternoon,
so
we've
been
working
on
scalable
privacy-preserving
in
network
bandwidth,
independent
technology
or
spin
bit
measurements
in
quic
protocol.
So
first
question
is
why
well
what
the
spin
bit
gives
you?
It's
a
explicit
signal
that
you
can
use
for
passive
per
flow
RTP
measurements.
So
we
need
this
because
quic
is
not
exposed
transport
internals
on
the
wire,
which
generally
is
good,
but
it
also
means
that
if
you
want
to
do
passive
performance
measurements,
the
quality
of
your
data
is
much
worse.
O
So
the
spin
bit
can
be
used
to
replace
this
loss
of
visibility,
and
it
effectively
gives
you
something
similar
back
that
you
would
get
anticipate
from
looking
at
sequence,
number
and
acknowledgment
acknowledgement,
sequences
or
times
ten
sequences.
It's
also
Sapir
dand.
If
you
would
just
do
a
handshake
measurement,
because
you
get
samples
all
through
the
flow
and
you
don't
measure
any
overhead,
that's
related
to
setting
up
the
connection.
O
So
how
does
this
work?
Well,
when
a
client
initiates
the
connection
it
starts
sending
packets
to
the
server
which
spin
zero,
then
the
server
will
always
reply
to
the
client
with
packets
that
have
the
same
spin
as
they
lost
saw
from
the
server
and
then
the
client
will
always
do
the
opposite,
so
it
will
always
flip
the
bit
and
then
send
it
back
to
the
server.
So
that
way,
you
get
this
bit
that
spins
through
the
network
and
if
you
then
have
an
observer
on
the
network
and
you
just
look
at
what
these
bits
are.
O
You
extract
the
waveform
and
then
you
just
measure
what
the
half
RT
or
the
half
period
is
of
this
waveform.
Then
you
get
the
RTT
of
your
network.
So
what
we've
did
is
we've
implemented
this
into
a
couple
of
quick
end
points,
so
that's
Ming,
Pico,
quick
quickly
and
quick,
oh,
and
what
was
really
nice
is
that
we
solve
for
each
of
these
ones.
O
Basically,
what
you
have
to
do
is
you
have
to
spend
some
time
familiar
familiarizing
yourself
with
the
code
base,
and
then
you
need
like
10
minutes
and
10
lines
of
code
to
add
to
spin
bits.
So
it's
really
super
simple
to
implement,
and
then
we
also
looked
at
a
bunch
of
implementations
for
observers
for
this,
so
we
have
one,
that's
really
minimal.
Just
built
around
go
packets.
O
We
have
one:
that's
used
around
VPP,
which
is
a
sort
of
or
or
around
an
experimental
measurement
platform
that
we
have,
that
is
based
on
VPP,
and
then
we
also
have
a
p4,
parser
yeah
and
then
a
whole
bunch
of
other
projects,
but
the
results
of
those
we
will
hopefully
be
able
to
present
that
hackathon
one
or
two.
Thank
you.
P
Okay,
we
that's
just.
We
worked
on
firmware
updates
for
IOT
devices,
and
this
is
a
recently
formed
working
group
and
deals
with
the
basic
problem
that
you
start
off
with
an
ID
device
comes
out
of
the
factory
and
later
you
figure
out
that
you
have
to
update
that
device,
because
there's
some
new
bugs
that
have
been
found.
There's
some
new
features
that
you
want
to
add,
and
so
that's
what
the
this
group
does
in
describing
the
manifest
metadata
to
the
firmware
and
providing
that
firmware
update
to
the
device.
P
We
wanted
to
get
the,
as
the
figure
shows
we
wanted
to
get
the
firmware
image
with
the
manifest
to
the
device,
in
our
case
using
UART,
but
we've
also
had
Alex
to
work
on
a
Laura
based
delivery,
and
we
use
these
this
one
specific
board
called
K,
64,
f,
NX,
Peapod
and
there's
a
bootloader
on
that
device
that
reads
that
stuff.
There
is
verifies
to
digital
signature
and
installs
the
the
new
firmware.
We
had
a
couple
of
people
participating.
P
It
was
a
pretty
ambitious
up,
so
we
initially
started
off
with
the
setups
of
the
development
environment.
Online
IDE,
desktop
IDE
command
line
environments.
We
got
the
code
working
for
manifest
creation,
which
is
in
peyten.
We
max
worked
on
absoluto
bootloader
on
the
IOT
device
and
also
on
the
on
the
update
mechanism
using
UART,
the
Laura
one
is
still
working
progress.
We
got
all
the
setup
with
the
keys
and
the
examples.
The
code,
the
C
code
for
the
verification
is
still
working
in
progress.
P
We're
still
dealing
with
some
issues
which
I
I
will
describe
on
the
next
slide.
We
produced
a
write-up
which
you
can
find
on.
This
is
a
bad
link.
If
you,
if
you're
interested
in
the
details
and
want
to
reproduce
this,
we
ran
into
some
challenges.
First
of
all,
for
those
for
the
for
the
guys
who
haven't
been
involved
in
this
work,
setting
up
the
environment
and
the
bison
and
other
embedded
development
environment.
Wasn't
that
easy?
P
B
Q
Q
Q
So
what
was
our
plan?
Our
plan
was
actually
to
check
whether
referential
implementation
Micronesian
meets
following
drafts.
There
is
bundle
protocol
version
7,
which
is
going
to
turn
into
standard
pretty
soon.
We
hope
it's
gonna
be
proposed.
Actually,
this
Friday
and
the
render
its
DTN
TCP
called
convergence
layer
protocol,
which
is
RFC
7
to
42.
Q
We
focused
on
two
problems
to
be
solved.
The
first
one
was
whether
there
is
enough
information
in
the
draft
to
do
the
bundle
encoding-
a
decoding-
that's
quite
complicated
in
this
protocol,
because
we
are
trying
to
save
as
much
space
as
we
can.
So
there
is
Tibor
and
all
kinds
of
optimizations
regarding
sides
of
the
of
the
bundle.
Q
Q
We
also
had
difficulties
to
reproduce
test
cases,
so
we
virtualized
integer
containers
all
the
work
we
did
so
how
we
did
it.
We
created
another
implementation
based
on
specs
and
we
let
it
talk
to
the
referential
one
to
find
out
whether
or
not
there
are
some
difficulties
or
other
it
works,
so
key
results.
We
had
some
interesting
discussions
with
our
co-operators.
Q
As
said
bono
protocol
was
originally
designed
for
space
communication,
but
they
pointed
out
there
might
be
useful
use
case
is
also
on
earth,
which
was
kind
of
surprising.
For
us.
We
were
discussing
providing
delight
or
and
peer-to-peer
data
network
in
developing
countries.
So
basically
the
idea
would
be
that
you
have
ad
hoc,
Wi-Fi
or
Bluetooth
or
something
and
he
would
distribute
information
hop-by-hop.
Q
Q
We
have
a
new
code
which
you
are
welcome
to
access
called
by
DTN
and
during
the
inter
operational
testing,
we
found
an
issue
that
micro
PC
and
actually
uses
special
header
which
does
not
meet
the
draft,
and
surprisingly,
it
was
for
historical
reasons:
yeah
okay,
we
are
running
out
of
time,
also
feel
free
to
actually
see
us
tomorrow.
To
answer
your
questions,
this
is
our
team.
Thank
you.
B
B
R
So
we've
been
working
to
validate
HTTP
four
five
one,
which
is
illegally
withheld
HTTP
status
code.
This
is
RFC,
seven,
seven,
two
five,
it's
it's
our
work
as
part
of
the
Human
Rights
HR
PC
group
and
Public
Interest
Technology
Group.
This
time
we've
had
new
contributors
from
around
the
world.
Some
from
Mauritius
we've
also
got
a
bigger
group
working
on
this
technology.
We've
got
three
new
code
modules,
which
makes
now
seven
we're
trying
to
answer
a
question
camp
protocols,
increased
transparency
and
serve
to
protect
human
rights.
S
Thank
you,
and
so
we
had
three
modules.
We
have
implemented
that
this
haxon
so
to
evaluate
to
measure
the
how
much
website
extra
using
the
451
status
code,
we
made
a
caller
to
called
webpage.
Actually
we
are
calling
the
reddit
web
website
right
now
and
also
if
he
found
the
4-5-1,
we
need
to
have
some
character,
so
we
implemented
the
corrector
and
also
to
visualize
the
results
we
made
a
dashboard
of
that.
Thank
you.
So
this
is
a
dashboard.
S
So
actually
we
haven't
find
any
4-5-1
say:
that's
called
on
the
production
website,
but
if
we
find
the
son
a
week
and
crawler
will
send
to
the
results
the
collector,
so
we
can
see
that
what
what's
happening
on
the
those
website.
Thank
you.
So
we
have
to
worked
on
from
the
ITF
ninety-ninth
product,
but
and
we
have
that
Singapore
water
cooler
and
now
we
have
the
character
to
corrects
the
results
results
from
the
tools,
browser,
extensions,
improving
and
Oh
Sakura.
Thank
you
send.
S
T
T
So
what
we
achieved.
Well,
we
achieved
dissemination
of
the
knowledge
of
the
sports
team
members
got
to
play
with
us
and
understand
and
master
them.
We
discussed
the
shake
principles
with
some
of
the
people,
newcomers
and
other
team
members
around
the
room.
We
run
the
existing
pieces
of
code,
everybody
got
to
experiment
with
the
existing
pieces
and
then
we
started
the
real
stuff
first
implementing
on
the
server
side,
replacing
JavaScript
code
with
Python
fixing,
bugs
that
we
found
in
the
compression
implementation
also
sending
fragments
over
laura1
using
new
board
brought
by
the
right
team.
T
So
not
quite
integrating
the
fragmentation
on
riot,
but
bridging
between
the
two
and
actually,
this
was
used
at
the
suit
the
table
to
do
the
firmware
update
over
Laura.
When
so,
we
would
send
the
fragments
of
the
firmware
and
yeah
I
think
we
also
got
focused
time
to
work
on
our
implementations
to
discuss
and
have
fun
really
reuniting
as
a
team.
T
What
we
learned
yes
once
again-
and
we
said
before
and
in
previous
hackathon
as
well-
it
takes
time
to
set
up
the
Laura
when
Gateway
is
the
lp1
gateways
in
general.
The
modes
debugging
takes
longer,
but
we
got
that
working
yesterday
and
then
we
fix
typos,
discussed
interpretation
of
the
draft
discuss
the
implementation
of
this
draft.
So
I
think
we
are
in
better
shape
now
than
yesterday
and
more
coding
to
be
done.
The
other
references.
U
Thanks,
so
my
name
is
yo-yo:
I
worked
with
our
term
Tarot
and
Brian
on
various
kinds
of
measurements
tools
or
little
tests
that
we
thought
were
interesting
for
some
reason
or
the
other,
and
they
were
actually
couple
of
different
things.
So
the
first
thing
that
that
we
worked
on
and
I'd
be
working
on
a
little
bit
for
the
last
couple
of
weeks
is
basically
experiment
with
different
algorithms.
That
would
do
a
similar
thing
as
traceroute.
U
So
here
in
the
picture,
for
instance,
you
see
the
traceroute
algorithm,
basically
just
increased
the
TTL
decrease
it
and
until
you
hit
the
right
destination
and
that
that's
sort
of
grows
linearly
as
you,
your
destinations
further
away,
the
more
packets
you'll
need
to
send,
and
then
we
came
up
with
this
idea
of
binary
search.
You
can
sort
of
try
and
guess
where,
where
the
destination
might
be,
and
then,
depending
on
the
answer,
you
will
go
either
on
the
lower
the
higher
half
and
that
actually
obviously
scales
better.
U
Now
whether
that
actually
matters
depends
on
whether
you
you
care
about
sending
many
packets
on
the
network
or
not.
If
it's
just
debugging
your
own
network,
it
may
not
matter
in
some
other
cases
might
actually
matter,
and
so
the
random
looking
graph
here
is
a
comparison
algorithm,
just
doing
a
random
test
on
a
different
kinds
of
TTL
values
and
then
then
deciding
based
on
that
and
that
obviously
keeps
random
results.
U
The
other
thing
that
we
were
testing
on
was
or
making
making
this
little
tool
that
would
figure
out
if
the
destinations
reports
which
TLS
version
and
it
can
be
used
for
many
ways,
for
instance,
testing
across
the
internet
for
popular
sites
what
what
they
support.
So,
for
instance,
this
this
afternoon,
I
did
this
test,
where
included,
I,
think
60
most
popular
websites
on
the
on
the
internet
and
came
up
with
the
answer
that
the
average
TLS
version
across
the
internet
for
those
sites
is
one
point.
U
B
V
V
B
B
V
L
V
Jones
I'll
connect
it
so
the
goal
here
is
mercy
around
application.
Integration,
open
SSL,
pass
without
the
release
of
open
SSL
with
TLS
1.3,
and
the
goal
here
is
to
work
the
track,
23
and
work
with
upstream
open
source
project
so
that
you
can
principle
take
something
active
you
get
and
just
specify.
I
want
to
get
dearest,
1.3
only
connection.
So
the
idea.
V
Okay,
so
this
is
what
we've
got
done,
w
get.
It
was
done
just
a
few
days
before
I
tear
friend
her
than
one
acaba
money
was
done
since
ITF
founder
at
hackathon.
This
time
we
just
did
test
with
religious
OpenSSL
release,
Nagios
plugins
check
HTTP
the
pull
request
via
was
reviewed
positively,
and
it's
going
to
be
in
the
next
release.
The
HTTP
per
will
request
already
merge.
Nice
thing
is,
but
now
we've
get.
V
You
can
specify
what
you
want
here
as
one
for
free
and
something
but
I
didn't
mention
also
is
a
period
Eston
l2-cache
up
in
central
meeting
this
event,
based
on
over
nagas
legging
check,
said,
which
is
which
has
been
murdered
the
first
day
of
a
hackathon
and
then
there's
an
event.
Woman
asked
for
to
see
which
is
a
HTTP
downloader
and
when
for
interesting
one
is
a
flips.
V
Powell
is
basically
a
MQTT
library
and
it
knocked
TLS
1.3
support,
so
it
will
go
only
up
to
jealous
1.0,
so
it's
useful
machine
to
actually
from
acacia
and
then
there's
hitch,
which
is
a
basically
a
TLS
proxy,
the
fascists
they're
working
progress.
So
what
we
learn.
Basically,
we
had
to
work
with
upstream
projects,
because
if
you
don't
follow
up
with
your
batch,
your
bash
becomes
there
and
that's
an
issue
also
when
it
comes
to
integration
application.
V
We
also
have
to
take
into
account
compatibility
with
all
the
OpenSSL
releases
which
X
integration
order
and
interoperability
testing
with
vertical
TLS
1.3
serve
optimization.
We
build
a
number
of
bugs
actually
so,
for
example,
in
what
particular
case
we
were
able
to
development
is
one
whole
freaking
town,
client,
hello,
pocket
which
was
actually
not
well
constructed.
So
with
Joseph
workshop.
V
This
makes
it
really
really
help
you
to
develop
there
at
one
point
and
what
were
the
things,
but
we
learn
is,
but
we
should
have
sent
the
announcement
earlier
so
that
we
fill
up
the
movie,
so
fatigue
is
completely
removed
from
Mauritius,
so
we
met
in
a
place
called
Agatha
English's.
So
there's
five,
you
people
were
involved.
V
B
Okay,
thank
you.
Logan
enter
the
whole
team
there
all
right,
no
one
else
right
or
are
we
we're
done
with
presentations?
Anyone
else
last
chance.
You
can
run
up
here
and
remember,
even
if
you
weren't
doing
a
presentation
here,
if
you
have
some
results,
are
you
waiting
to
do
a
presentation?
No,
okay,
still
upload
your.
If
you
have
a
some
slides
that
are
useful
and
you
don't
feel
like
presenting
them
here
there.
B
B
B
So
you
know
that
they
really
did
a
great
job
of
meeting
as
many
of
our
requirements
for
the
hackathon
special
Network
set
ups
and
and
whatnot
as
they
could.
But
the
one
thing
they
did
ask
me
was
next
time.
If
we
could
get
those
in
even
earlier,
they
ship
all
their
gear
about
a
month
before
so
for
IETF
102,
they're,
gonna
ship,
it
all
the
gear
out
a
month
before,
if
they,
if
they
have
requests
for
special
things,
they
do
the
best
they
can
with
what
they
have
here.
B
But
if
they
learn
about
those
requests
ahead
of
time,
then
you
know
they
can
do
even
more.
For
us,
for
example,
some
people
asked
for
four
VMs
if
we
had
the
ability
to
spin
those
up.
If
you
let
us
know
ahead
of
time,
then
yeah
we'll
send
a
server
and
be
able
to
spend
those
up
for
you
and
get
them
on
the
network,
just
as
an
example.
B
B
You
don't
need
to
read
my
email,
sorry
about
that,
okay,
so
what
I
wanted
to
show
was
just
a
couple
reminders
for
what
we
still
have
to
come
so
after
we
leave
this
room,
it's
going
to
get
converted
into
the
IETF
lounge
and
part
of
the
IETF
lounge
is
going
to
be
the
code.
Lounge
I'm,
not
sure
exactly
where
it'll
be
but
it'll
be
in
this
room.
It'll
be
pretty
easy
to
find.
It's
just
going
to
be
a
portion
of
this
room.
B
That's
kind
of
set
up
for
coding
or
maybe
it'll,
be
the
whole
room,
but
anyways.
If
you
come
to
this
area,
you're
sure
to
find
it,
and
you
can
schedule
a
time
if
you
click
on
this
link
and
there's
a
link
in
the
in
the
wiki
for
the
hackathon.
You
can
set
a
time
when
it's
mainly
just
for
scheduling
to
help
your
team
know.
Hey
we're
gonna,
be
here
at
a
specific
time
and
to
let
other
people
who
are
interested
in
coming
and
working
with.
B
B
The
other
thing
is
what
we
haven't
done
before.
This
is
an
attempt
to
find
a
nice
thing
to
replace
what
we're
missing
now
by
not
having
bits
and
bytes
anymore,
so
we're
calling
it
hack
demo,
happy
hour,
it's
gonna,
be
in
here
as
well.
The
thought
is,
you
can
set
up
your
whatever
your
demo
is,
or
you
know,
just
set
up
some
space
here
during
the
day
tomorrow
and
then
from
6,
not
6,
to
8
we're
just
going
to
have
some
light
refreshments
here
and
there'll
be
non-alcoholic
drinks.
B
There
will
be
a
cash
bar
if
you
feel
like
having
a
beer
or
a
glass
of
wine,
but
really
the
idea
is
to
have
a
couple
hours
where
people
can
come
and
talk
to
you
in
depth
about
your
project.
If
you
have
a
demo,
you
can
show
the
demo,
you
can
run
them
through
your
slides.
They
might
want
to
see
your
code
and
learn.
You
know
what
you
did,
how
to
get
access
to
it,
how
to
build
your
project.
They
just
might
want
to
ask
you
questions.
B
You
know
whatever
so
it's
kind
of
a
social
atmosphere
where
you
can
have
those
more
in-depth
conversations
that
you
know
we
couldn't
do
here
with
a
3-minute
presentation.
So
for
that
we
would
really
appreciate
that
you,
you
sign
up
again.
It's
not
so
much
that
we
have
a
shortage
of
space,
but
that's
the
only
ways
that
people
will
know
what
projects
are
going
to
be
here.
So
the
sooner
you
sign
up
the
sooner
anyone
else
interested
in
that's
thinking,
hey
should
I
go
to
that
thing
or
not.
They'll
see!
B
Oh,
hey
the
you
know
the
quick
projects
going
to
be
there.
Cool
I
really
want
to
see
that
or
you
know
whatever
your
project
is
so
I
recommend
you
register
as
soon
as
possible
if
you're
planning
on
doing
it.
If
you
haven't
already,
you
know,
please
go
ahead
and
do
that.
The
links
for
that
are
also
in
the
hackathon
wiki.
B
B
So,
as
we
always
do,
I
would
love
any
feedback,
good
feedbacks,
always
nice
to
hear
but
any
constructive
feedback.
Any
criticism.
That's
that's
good
to
hear
too
so,
if
you
feel
like
sharing
something
or
have
an
idea-
and
you
want
to
share
it
publicly
now-
that's
that's
fine,
you
can
you
can
shout
it
out
or
run
up
here.
Grab
me
anytime
during
the
week.
That
would
be
fine
as
well
yeah
we
ran
out
of
shirts,
so
I
you
know,
did
a
great
job.
B
B
I
want
to
thank
I,
do
have
a
slide
thanking
my
team
because
they
put
up
some
money
so
that
I
could
get
shirts,
but
only
enough
money
to
get
200
which,
which
seemed
like
more
than
enough.
But
the
response
here
and
the
turnout
was
was
fantastic
and
I
really
want
to
well
apologize
for
the
lack
of
shirts,
but
thank
all
of
you
because
you
know,
as
you
know,
that
this
hackathon
doesn't
happen
unless
we
have
champions
coming
here.
B
Leading
projects,
participants
participating
in
those
projects,
and
it's
really
amazing
the
way
all
of
you
turn
out
a
sacrifice
your
weekend
and
fortunately
we
didn't
have
to
compete
with
the
weather
so
much
because
I
don't
think
anyone
wanted
to
be
out
in
the
snow.
Maybe
that
helped,
but
but
no
really
seriously.
I
appreciate
everyone.
You
know
taking
advantage
of
this
opportunity
and
we're
seeing
great
work.
Take
what
you
learned
back
into
your
working
group.
I.
B
Think
that's
fantastic
when
that
happens,
and
you
spread
the
word
seeing
so
many
first-time
ietf
people
here
is
fantastic
and
for
those
of
you
who
came
and
you're
not
staying
for
the
meeting.
I
hope
you
do
stay
involved
with
the
IETF
and
other
ways
through
the
mailing
list.
You
know
whatever
and
maybe
consider
a
future
IETF
meeting
we'll
definitely
be
having
the
hackathon
in
Montreal
almost
positive.
The
dates
will
be
exactly
the
same:
the
Saturday
in
the
Sunday
same
format,
hopefully
we'll
get
a
really
big
room.