►
From YouTube: CORE WG Interim Meeting, 2020-04-08
Description
CORE WG Interim Meeting, 2020-04-08
A
A
A
You
hello,
you
can
just
say
your
name
in
the
in
the
in
either
of
the
chats
and
say
that
you
want
to
do
the
note-taking
or
just
go
to
the
iPad
already
named
Thank
You
Francesca,
so
we
have
one.
Maybe
someone
else
can
help
Francesca
there
you
go.
They
tell
you
this
all
right,
all
right
right
now.
We
have
three
people
very
good.
A
Thank
you
so
go
to
go
to
the
ether
pattern
and
when
the
discussion
happens,
for
those
who
have
maybe
haven't
done
it
before
I,
don't
know
just
remember
that
the
important
part
of
the
discussion
is
not
so
much
the
presentations,
but
the
presentations
are
kind
of
self-explanatory
in
a
way
and
then
moving
on.
So
the
note.
Well
here
you
have
some
of
the
rules
that
the
IDF
works
on.
A
They
is
not
just
about
the
IPR
or
about
the
the
legal
items,
but
also
about
the
best
practices
and
the
best
way
we
can
interact
with
each
other.
So
being
nice
and
polite
and
tone
the
sessions
will
run
on
Wednesday
and
on
Thursday.
For
today
we
have
maybe
the
three
clusters:
the
cork
on
cluster,
the
group,
communications
and
CIN
ml,
and
then
on
Thursday.
We
will
have
the
spillover
of
Wednesday
plus
the
core
applications
and
some
discussions
that
will
happen
in
the
on
this
Friday
side,
applications
meeting
and
then
some
flex
time
for
other
material.
A
A
So
we
have
a
new
culture
Marco.
All
of
you
probably
know
him
already.
He
has
a
PhD
on
networks
and
communication
security
and,
and
he
got
a
Marie
Curie
a
scholarship
a
few
years
ago
for
his
postdoc,
which
is
a
very
prestigious
scholarship
in
Europe
and
currently
he's
a
senior
researcher
in
the
cyber
security
unit
in
rice
in
Stockholm,
an
IDF,
wise
I,
guess
you
have
known
him
for
a
long
time
in
involved
in
a
sand
in
core
and
nowadays,
especially
with
group
communication
in
core
as
well.
So
very
warm
welcome.
Marco.
B
A
And
it's
kind
of
a
strange
always
to
do
these
things
over
videoconferencing,
I
guess,
ideally,
we
should
be
face
to
face,
but
it
can
happen
and
then
also
we
have
I
mean
Carsten
raised
in
an
email
to
the
mainly
stone
on
his
ten
years
in
core,
which
is
quite
an
achievement
again
over
videoconferencing
is
it's
kind
of
hard
to
to
thank
him.
You
know
like
like
this,
so
hopefully
we
will
have
a
warm
celebration
when
we
meet
again
face
to
face
Avenue.
You
have
something
to
say
Carsten
by
the
way.
C
C
So
thank
you
for
everything
and
I
made
this
slide,
reminding
people
that
we
do
rotate
chairs
in
the
ITF
and
I
think
there
are
very
good
reasons.
People
refocus
on
their
jobs,
people
change
their
interests
in
other
ways
or
people
I
just
want
to
see.
New
people
coming
in.
My
marco
is
not
exactly
new
as
a
participant
of
the
group.
C
He
is
a
new
chair
and
I
think
that's
very
important
for
the
IETF
to
to
have
young
people
come
in
and
perform
leadership
positions,
and
so
yes,
thank
you.
Mark
II
Marco
for
taking
this
on
and
I,
have
already
seen
your
work
for
a
month
now
and
I'm
sure
the
two
of
you
will
manage
the
working
group
very
well.
A
Thanks
a
lot
speaking
of
rotation,
we
also
have
we
also
have
a
new
area
director
well
new
renew
in
a
way
so
Alexi
ends
his
turn
now
and
Bari
will
be
becoming
again.
The
core
will
course
re
the
area
director
for
core,
which
is
great
also
because
he
knows
the
material
and
can
help
on
on
that
too.
I,
don't
even
feel
like
say
you
have
some
words
as
well.
I
think
you
are
in
the
colors.
For
yes,
yes,
word.
D
A
That
ends
the
intro,
then
practicality,
so
we
will
have
interim
meetings
to
every
other
week,
a
certain
29th
of
April.
There
will
be
on
specific
topics,
so,
for
instance,
applications
is
a
very
hot
topic:
the
usage
of
resource
directory
with
a
sin
in
practice,
so
a
very
hot
topic.
So
we
are
kind
of
hammering
down
the
details
for
those
meetings
and
now
it's
a
good
time
to
have
proposals
we
haven't
send
invites
we
will
do
that
later
on
then
yeah
we
did
a
bit
of
cleaning
of
the
github
page,
the
landing
it
a
page.
A
So
there
you
should
have
now
a
bit
of
info
on
the
working
group
material.
So
you
have
all
of
this
light
source
on
github
and
all
the
man-eating
materials
as
well
as
in
the
data
tracker.
We
also
have
the
editors
page,
a
tracker
and
basically,
all
the
links
together
in
the
same
place,
which
is
kind
of
convenient.
I
already
said
this
as
well,
so
for
the
jabber.
Just
please
use
the
jabber
bot
when,
when
we
have
in
discussion,
so
it's
a
virtual
cube
use
the
Q
plus
for
yourself.
A
If
not,
you
can
do
Q
plus
on
the
WebEx,
and
if
you
don't
know
how
you
just
press
help
you
other
people
can
add
you
to
the
queue
you
don't
have
the
jabber
now
on
working
bite
them.
So
we
have
two
new
RFC's
coming
out
for
multi
port
city
and
hop
limit,
which
is
great
hope
limit
was,
if
you
remember,
it,
was
also
requested
by
another
working
group
called
dots,
so
they
are
using
coop
for
signalling.
That
is
great
that
they
can.
A
Then,
on
the
edit
oq
sin
MLH
and
more
units
which
are
pretty
much
done,
there
were
few
comments
or
more
units
that
sparked
a
discussion
on
cinema
versioning,
and
you
will
have
a
glimpse
at
that
later
today,
in
Iasi
processing,
we
have
resurrect
theory
which
by
now
everybody
I
assume,
has
read
multiple
times:
multiple
versions.
There
were
pending
comments
by
Alexey
I,
think
a
Christian
has
addressed
them
or
will
address
them
very
soon,
either
way
they
will
be
discussed
on
the
Thursday
session.
A
A
Other
days
also
many
remarks
to
be
said
on
this,
that
they
are
moving
forward
and
then
on
post
working
group
last
call,
so
we
have
a
core
request.
Talk
echo
request,
talk
which
basically,
we
haven't
formally
closed,
but
suppose
we
need
to
actually
do
that
and
I
believe.
This
version
also
address
all
the
comments
that
were
raised.
So
we
just
need
to
close
that
and
start
with
the
Shepherd
right
up
that
send
it
to
Bari.
A
So
I,
don't
know
different
is
pretty
much
done
as
well
and
with
that
we
are
in
there
with
some
comments
on
the
ABN
F
formatting
of
the
urine
and
I.
Think
the
well
I
know
that
the
offers,
from
the
point
of
view
is
done.
So
we
just
need
to
formally
start
the
working
class
call
and
we
can
continue.
So
that's
so
much
for
the
very
mature
document,
and
now
we
can
go
into
some
other
items
that
we
have
in
core.
A
My
plan
is
by
the
way
for
the
Alpha
for
the
presenters
that
I
can
pass
the
slides
so
that
we're
gonna
need
to
switch
slides.
So
you
just
tell
me
next
slide
and
now
we'll
do
that
for
you.
Okay,
so
I'll
meet
myself
and
unless
there
is
some
comments
about
the
previous
drafts,
yes,
you
can
go
ahead,
a
very
low
percent.
A
A
E
C
E
F
A
E
G
C
G
G
G
G
So
one
of
the
questions
is
whether
we
want
to
begin
linked
in
any
particular
way
and
the
possible
change
of
semantics
in
young
files.
So
what
is
being
discussed
in
net
notice
that
maybe
in
the
future,
some
young
files
might
change
the
semantics
of
previous
revisions
of
the
Semyon
pile?
So
whatever
we
want
in
this
document
to
have
any
particular
hand
link
of
such
situations
or
we
can
assume
that
we
can
just
allocate
anos
it
range
for
those
seed
files
for
those
young
files
and
have
that
as
the
solution
is
there.
G
Some
reviewers
noted
that
is
a
little
bit
Chico
to
understand,
and
others
say
that,
maybe
it's
not
exactly
in
the
spirit
of
early
application,
as
in
the
era
see
where
the
indication
is
described.
It
is
mentioned
that
it's
for
documents
that
are
going
to
be
RFC's,
and
here
we
are
not
mandating
this.
We
are
requiring
only
expert
to
review
state
of
the
document.
So
what
I
have
discussed?
G
G
C
C
C
So
we
probably
need
a
little
bit
of
text,
because
the
earlier
location
RFC
is
the
number
of
which
I
forget.
Of
course,
it's
very
much
oriented
towards
an
is
G
based
process
and,
of
course
the
designated
expert
doesn't
operate
exactly
in
the
same
way.
The
is
Vedas
so
I
think
we
we
may
have
to
write
white
some
text
there
to
explain
all
the
system.
G
H
I'm
sorry,
two
circles
also
the
feedback
I
think
from
that
an
expert
review
tip
please
very
fast,
so
you
can
expect
like
in
two
weeks
or
three
weeks,
location
decision.
You
don't
need
in
fact
an
early
allocation
anymore
because
she
gets
location.
Anyhow,
that's
valid
point
here,
but
there's
also
something
else
to
consider
is.
C
That
the
future,
so
early
locations
have
the
interesting
property
that
they
are
reviewed
after
a
year
and
and
they
go
away
if
nothing
happened
so
by
by
mirroring
the
earlier
location
process
here,
we
would
also
be
able
to
issue
provisional
locations
very
quickly
and
then
get
them
back.
If
the
thing
actually
didn't
happen,
which
is
different
from
just
quickly
allocating
things
where
essentially
the
code
point
has
come.
C
Yeah
I'm
not
sure
that
that
the
the
effort
is
worth
it
because
of
SIDS
aren't
quite
as
scarce
as
other
core
points
that
we
are
doing.
An
education
for
so
I
think
we
have
to
think
whether
we
actually
want
this.
But
I
was
just
trying
to
explain
what
we
gained
by
having
an
early
education
kind
of
process.
G
G
A
G
G
Most
of
them
are
already
resolved.
I
haven't
received
formal
confirmation
from
Mexico,
but
he
will
for
sure.
Let
me
know
if
there
are
still
some
things
that
need
to
be
resolved
there.
The
question
that
arose
from
their
answer
to
the
comments
was
how
exactly
we
want
to
connect
this
specification
with
young
to
savour
a
specification,
and
the
question
is
that.
G
We
could
mandate
that
the
seats
that
we
use
in
the
young
people
are
exactly
the
seats
that
come
from
the
Sidra.
Currently,
this
is
not
the
case
for
now.
We
are
just
talking
about
integers
that
represent
elements
items
in
the
young.
We
don't
say
that
they
need
to
be
proficient
in
the
same
way.
They
are
provisioned
in
the
situa
and
the
idea
has
been
that
people
might
decide
that
they
want
something
else
in
the
future
and
it's
the
common
track
that
actually
links
the
things
together.
G
So
that's
the
question
whether
we
want
to
whether
we
want
to
have
that
link
that
normative
reference
from
the
young
people,
or
maybe
the
other
way
around
from
the
city
draft
to
the
app
Siebel.
In
order
to
have
this
facility,
this
piece
of
use
there
in
other
circumstances,
do
you
want
to
discuss
this
now
or
maybe
we
can
continue
on
the
mailing
list.
I,
don't
know
well.
G
G
A
C
I
have
a
very
strong
opinion
on
normative
references,
because
I
I
was
chair
of
this.
Like
group,
when
we
had,
we
had
to
wait
for
a
year,
RFC
72
5:20
we
published,
because
we
had
a
normative
reference
that
we
shouldn't
really
have
had
and
then
right
now,
I
have
a
document
in
the
RFC
interview
that
sitting
there
for
140
weeks,
which
is
almost
three
years
now,
because
we
have
a
what
normative
reference
that
we
probably
shouldn't
have
had.
C
So,
even
even
if
it
looks
all
very
obviously
and
great
at
the
time
you
you
submit
things,
it's
really
worth
thinking
about
which
normative
references
you
want
to
have
in
a
document
and
since
basically,
Yangtze
bar
is
a
free-standing
document.
Of
course
it
sits
on
the
shoulders
of
gang,
but
it
doesn't
really
need
all
the
other
machinery
that
we
are
creating
here.
C
I
think
it
would
be
best
to
processes
with
the
smallest
of
amount
of
normative
references
we
can
have,
and
we
have
to
balance
that
is
I
with
something
that
yearning
said,
which
is
that
it's
probably
a
good
idea
to
make.
The
document
also
freestanding
in
the
sense
that
somebody
who
just
wants
to
use
yangtze
bow,
can
reference
it
and
doesn't
have
to
reference
all
the
other
documents
just
to
get
little
things
like
media
type,
definitions
and
so
on
and
so
on.
G
G
A
G
G
G
While
we
had
a
few
questions,
one
was
what
is
exactly
korkin,
why
it's
the
complete
clothespin,
whereas
is
where
it
is
just
the
protocol
like
rest
confidence,
just
the
protocol,
the
same
way
that
would
have
commonly
here
and
not
the
complete
cluster.
So
there
was
some
confusion
around
this
and
it
came
pretty
much
from
anyone
that
commented
on
this
draft,
and
even
some
of
the
other
were
just
so.
One
may
be
simple
solution
would
be
at
least
to
include
core
coughing.
G
If
you
have
some
other
opinion
do
letters,
we
are
going
to
progress
in
this
direction
and
the
other
part
two
is
how
we
want
to
actually
handle
this.
Take
considerations
for
sure
one
point
that
could
be
possible
to
do
is
add
some
references
to
Oscar
I.
Think
it's
missing
right
now
and
possibly
to
some
other
things
like
case,
but
I'm,
not
exactly
sure
whether
we
want
to
add
references
in
the
security
considerations
in
for
his
documents.
G
So
the
last
draft
from
the
cluster
is
a
library
we
still
held.
It
was
doing
working
delasco.
We
had
some
positive
reviews
and
some
reviews
that
a
score
some
more
clarifications
I
believe
I
answered
yes
before
the
meeting
and
akamai
from
Tom
pitch
so
sure
he
would.
He
might
have
others
because
I
his
command.
G
He
said
that
he
didn't
get
in
that
I
into
the
chart,
so
we
might
have
some
more
discussions
there,
but
the
only
other
part
that
I'm
our
office
again
considering
the
security
considerations
and
I
believe
before
we
just
link
it
to
the
comedy
draft
that
will
be
good
way
forward.
It
is
I
believe
how
it
was
handled
in
the
young
module
library,
RC.
G
A
G
I
believe
I
should
have
a
new
I
mean
I
will
have
or
the
editorial
changes
in,
maybe
by
tomorrow,
at
work
by
the
end
of
the
week,
and
then
it
depends.
If
some
discussions
need
to
be
ending
more
details,
it
might
take
a
little
bit
more
time,
for
example,
for
seed
draft
writing
down
the
part
about
earlier
location
might
take
a
little
bit
of
time,
but
in
any
case
my
intention
is
to
try
to
have
a
new
version
by
the
end
of
next
week.
A
So
then,
in
that
case,
I'll
be
good.
If
you
try
to
reach
out
to
other
potential
reviewers
that
you
think
would
be
should
be
aware
of
this,
that
you
have
reach
already
and
send
them
that
latest
version
and
other
than
that
I
guess.
We
can,
because
we
haven't
formally
closed
they're
working,
Robles
Cal.
Let's
call
on
this
one,
so
we
can
just
do
another
one
once
the
new
documents
are
submitted.
G
A
A
H
H
Don't
we
that
one
include
all
the
contents,
also
updated
latest
insights
and
have
standards
track
document
and
it's
okay,
what
can
be
used
in
systemic
reference
implementations,
and
so
what
we
have
in
scope
is
really
a
good
communication
offer
and
be
secured
using
no
score
or
unsecured
and
also
including
Nicole
protocol,
and
any
recent
developments
that
might
impact
example
serve
twice.
Their
security
is
a
clear
one,
maybe
also
the
new
methods
that
we
have
and
there's
the
use
cases
on
a
move
to
the
appendix
they're
suited
as
a
reference.
H
H
Your
authorship
and
next
I
will
talk
about
topics,
basically
that
we
made
changes
on
it's
on
the
next
slide.
So
all
that
fuel
and
what
has
seen
quite
a
bit
of
change,
is
the
definition
section.
There's
a
new
section
there
that
defines
Asian
groups,
including
two
new
figures,
there's
also
the
concept
of
discovery.
It
is
now
added
in
section:
3
refers
to
the
resource
directory.
Can
we
use
for
that?
H
It's
also
a
security
section
that
got
some
comments
is
now
rewritten.
It's
the
lone
countering
attacks,
also
with
more
details
on
what
to
actually
do
to
counter
the
attacks.
There
are
many
other
fixes
and
clarifications,
for
example,
description
of
what
what
our
seas
are
obsolete.
It
and
updated
is
no
go
to
the
next.
H
This
figure
here
shows
example
of
concepts
in
the
draft
there
will
be
also
the
more
generic
model
is,
so
the
touch
of
groups
are
showing
area
the
scope
group,
which
is
the
consisting
of
a
move,
cause
address
and
civic
port
used
by
this
group.
It
can
be
linked
to
a
scroll
core
group
to
provide
security
and
within
that
combination,
you
can
also
have
more
education
group,
so
you
think
of
example,
indicating
an
application
group
on
the
right
path
shown
here.
H
H
Yeah,
we
still
have
open
issues
in
the
draft,
so
we
have
a
new
issue
page
and
I
get
up
and
I
call
working
group.
It
was
three
there's
also
still
the
previous
page,
it
lab
yeah.
So
this
number
one
show
here
is
from
the
github
page
email
threats.
There
was
some
discussion
about
yeah.
What
is
it
exactly
the
most
costs
and
point
concept
and
can
the
server
actually
changed
UDP
ports
in
support?
H
So
that's
shown
in
figure,
schematically
clients
sending
requests
from
a
specific
duty
port
to
the
destination,
and
that's
not
just
what
the
server
does
here.
It
receives
it
at
one,
end
point
and
then
internally
it
delegates
the
request
to
another
endpoint,
that's
the
one,
but
also
serious
at
five.
Six.
Eight
three
responds
with
coop
response
in
this
case,
so
yeah.
H
The
client
doesn't
really
look
at
that
for
number,
because
by
the
earlier
rules
in
RFC
75
to
it
already
knows,
I
should
only
match
all
the
children
not
endpoints
case,
because
it's
still,
there
was
some
discussion
on
the
waiting
list,
because
some
people
thought
that
it's
counterintuitive,
that
the
server
would
actually
change
the
endpoint
role.
Delegating
turn
on
into
another
hand,
that's
fully
concluded
yet
issues,
and
what
we
want
to
do
is
actually
clarify.
H
These
two
issues
are
from
the
previous
subtractor
from
Wavelab.
There
was
a
comments
here
coming
from
Thomas
about
using
your
right
host.
I
mean
patient
groups,
yeah,
that's
something
we
have
to
clarify,
maybe
a
bit
more
in
the
text,
so,
if
I
suppose
to
use
the
right
host
for
that
fine
application,
loop
or
not,
there's
one
issue
with
the
right
host
option
is
that
you
cannot
extract
from
your
I,
because
if
you
give
regular
co-op
your
eye
and
you
parse
it
into
options,
then
by
the
default
parsing
media
right
host
gets
taken
out.
H
So
you
get
shortest
possible
our
message
and
to
use
that
option
you
have
to
after
your
right
person,
you
have
to
somehow
put
it
in
again
and
so
so
reliable
way
to
put
in
to
negate
the
application.
Look,
that's
issue.
We.
We
also
are
happy
to
take
more
comments
on
that
course,
and
the
second
issue
is
about
response
suppression.
H
H
H
A
H
A
H
H
A
H
H
A
B
B
B
So
just
the
quick
highlight
of
the
main
updates
already
incorporated
in
version
8,
we
improved
pretty
much
rules
about
processing
of
messages
was
requesting
responses,
and
especially
across
group
tracking.
So
in
case
the
response
is
protected
with
a
new
context
different
from
the
one
used
for
the
request.
The
server
has
to
use
its
own
sequence
number
also
included
as
partial
ID.
B
We
also
detail
the
support
for
observation
in
line
for
the
different
message
processing
as
the
Oscar
RFC
does,
and
we
have
also
more
security
considerations,
especially
not
recommending
using
the
group
key
material
to
send
a
unicast
message,
which
can
be
the
case
actually
as
a
need
in
case
the
option
is
used
or
block
wise
transfer
is
used
and
we
have
kind
of
solved
the
problem
introducing
some
more
content
in
this
document.
I'm.
Coming
to
that
next
slide,
please
in
particular,
we
have
defined
additional
modes
of
operation
of
score.
B
B
Well,
you
can
do
that
and
can
be
convenient
in
use
cases
where
you
don't
have
intermediaries,
for
instance,
for
verifying
signatures
like
gateways
next
slide,
please
and
as
to
the
actual,
a
pairwise
keys.
They
are
derive
using
the
same
construction,
adopting
an
all
score,
starting
from
the
sender
key
of
the
sender
device
and
a
static
static.
B
The
field
manager
secret,
which
is
in
turn,
computed
out
of
the
a
symmetric
keys
of
the
TX
of
devices,
and
this
is
compatible
for
easy
to
say
and
abbs
a
we
are
signaling
this,
so
that
the
message
is
in
particular,
protected
with
pairwise
keying
material
setting,
one
of
the
yet
unused
bit
in
the
in
the
flag
bite
of
the
oscar
option
that
we
are
now
defining
and
every
serving
for
this
next
slide.
Please
so
I
have
now
a
list
of
open
points
with
point
to
address
for
the
next
update
I'll
give
an
overview.
B
Please
comment
especially
if
you
have
concerns
about
the
way
we
propose
to
address
them
so
out
of
the
review
summers
that
we
should
be
more
clear
in
what
to
do
with
the
sender
sequence.
Number
of
the
note
after
a
group
routine
occurs,
there
are
pros
and
cons.
We
both
approach
is
we
prefer
to
not
reset
the
sender.
Sequence
number
just
keep
it
growing,
at
least
as
default
behavior,
but,
of
course,
application
policies
on
a
particular
node
can
just
override
this
and
instant
reset
it
to
zero.
B
B
So
all
considering
the
instant,
the
pairwise
mode
seemed
to
be
very
well
accepted.
We
have
this
in
mind.
We
can
forget
about
the
optimize
mode
as
as
a
mode
as
such
keeping
static
content
about
the
optimized
request
for
for
more
thoughts
on
the
raised
issues
and
moved
optimize
requests
as
an
appendix
in
the
document.
B
B
B
Before
previous
one
just
slide,
thank
you
this
one.
We
want
to
clarify
a
bit
better
about
the
nature
of
the
pairwise
mode,
because
yeah,
it's
not
a
group
of
score.
Of
course,
if
you
are
willing
to
pay
the
price
of
having
essentially
a
group
manager
deployed-
and
this
made
me
convene
our
setups,
where
you're
not
interested
in
having
messages
just
only
one
to
one,
but
this
can
be
helpful
for
the
sake
of
key
management.
In
fact,
the
note
can
just
run
on
the
wire
a
single
establishment
with
a
group
manager
and
then
following
local
derivation.
B
H
B
Christian
wondered
about
the
flag
bit
in
particular
if
it
has
any
other
meaning
than
signaling
this
usage,
and
not
that
we
can
think
it
was
really
thoughtful
of
having
group
of
score
in
mind,
but
Christian
proposed
also
to
flip
the
way
we
use
the
value
of
that
bit
right
now.
So
it's
proposing
to
have
it
set
to
0
to
indicate,
in
fact
that
the
message
is
protected
with
pairwise
key
material
that
that's
in
a
sense,
what
happens
in
all
score
where
the
bit
has
value
0
1
instead
setting
the
bit
to
one
to
signal.
B
B
B
The
server
to
group
identify
I
think
we
need
to
think
a
bit
more
about
this,
and
if
this
is
any
shooters,
the
different
of
submissions
are
are
bound
to
two
different
talking
values
and
they
play
a
role
in
retrieving
the
security
context
also
on
the
client
when
getting
a
response,
so
it
may,
it
may
not
be
an
issue
actually,
but
we
need
to
think
more
of
this
next
slide.
Please.
B
Final
slide
on
open
points,
we
have
different
opinions
from
a
Christian
and
Jim
about
an
appendix
section,
describing
one
of
the
approaches
for
the
server
to
synchronize,
back
with
a
sequence
number
of
clients
based
on
a
first
trusted
request,
whose
sequence
number
is
taken
as
baseline
by
the
server
essentially
so
christian
purpose
is
to
also
can
process
the
requests,
such
while
jim
believes
it's
actually
vital
to
just
take
its
sequence.
Numbers
baseline,
yes,
but
then
discard
that
request.
B
So
we
have
different
opinions,
and
this
definitely
deserves
more
discussion.
Starting
from
the
data
Bishu,
we
create
easier
things.
We
got
some
text
from
christian
things
for
the
appendix
related
to
the
pairwise
mode,
where
we
describe
very
a
general
simple
discovery
mechanisms
to
the
benefit
of
the
client
to
find
out
servers
in
the
group
and
information
related
to
them
to
talk
to
them
using
the
pairwise
mode,
and
this
is
in
particular
for
their
IP
address.
B
We
think
that
text
cover
pretty
well.
These
need,
but
feedback
are,
of
course,
welcome.
If
anything
is
missing,
we
need
to
add
more
text
now
to
make
it
possible
also
for
group
members
that
are
only
silent
servers
to
support
the
pairwise
mode
and
use
it
because
of
course,
they
will
also
need
now
to
provide
the
public
key
to
the
group
manager,
possibly
have
a
k
ID
as
identifiers
of
the
public
key.
This
will
have
side
effect
that
we
can
totally
handle.
Also
on
on
the
ice
Draft
taking
care
of
the
key
management
in
ace.
B
We
will
also
remove
the
Ayana
registers
we
created
here
Oh
long
ago
cause
now.
It
seems
to
be
time
to
point
to
the
new,
updated
cozy
registries
and
we
think
of
having
new
appendix
describing
a
bit
more
other
than
sequence
number
synchronization.
What
the
device
should
do
in
case
of
reboot,
essentially
next
slide.
Please
yeah.
We
plan
to
work
on
these
open
points
from
the
previous
reviews
and
the
latest
revision.
We
got
already
own
version.
B
I
I
I
I
Regarding
the
pairwise
I'm
having
problems
saying
that
EDD
si
should
be
the
mandatory
to
implement
and
pairwise
is
mandatory
to
implement,
because
that
basically
says
I
need
to
completely
separate
implementations
of
EDD
si
on
my
hardware
when
in
front
of
Montgomery
in
one
for
Edwards
I,
don't
think
that's
very
kind.
B
J
A
I
Enough
to
ask
it
in
terms
of
doing
Edward
signatures,
there's
a
draft
coming
out
of
a
wig
which
talks
about
the
EDD
essay
and,
if,
with
the
mapped
signet
with
the
mapped,
coordinates
where
you
can
do
both
encryption
and
signing
both
the
key
agreement
and
the
signing.
So
that
could
become
the
mandatory
to
implement
rather
than
generic
EDD.
Si.
A
A
F
F
Because
the
score
right
now,
if
you
have
been
following
that
discussion,
only
allows
H
Mac
based
HKD
F
to
be
used
for
key
derivation,
and
there
was
some
discussion
about
is
this.
Why
was
that?
And
there
was
no
real
reason
why
only
H
Mac
based
H
KDF
are
used.
So
we
wanted
to
fix
that
and
say
that
all
KDF
registered
in
Cosi
can
be
used
and
John
suggested
that
we
would
do
that
in
this
draft.
So
in
this
document-
and
that
will
mean
that
we.
F
Specify
or
registering
codes
it
can
be
used
for
all
score,
and
that
will
mean
that
this
document
updates,
Oscar
and
I
agree
that
we
need
to
do
that.
We
need
to
make
these
fix
for
a
KDF,
but
I'm
not
sure
doing
it
in
this
document
is
the
best
option,
because
I
would
like
to
do
it
the
fastest
way
possible.
A
A
F
A
I'm,
assuming
that
the
office
of
group,
common
no
score
will
be
the
ones
having
well-versed
opinions
to
me
is
a
bit
like
Mike.
What
Michael
Richardson
was
saying
that
I,
don't
know
enough
about
cryptography
to
be
able
to
have
an
opinion
and
I
guess
for
the
rest
of
the
group
is
similar
regarding
the
reference.
Maybe
we
can
move
it.
We
can
check
it
a
bit
later.
We
are
not
it's
not
urgent.
I
would
say
at
this
point,
so
my
other
thoughts
were
that
for
this
draft
I
guess
we
were
looking
for
reviewers
right.
A
J
Think
that
would
be
yes
yeah,
so
hello.
This
is
just
announcers
and
I'd
like
to
give
you
a
short
rundown
on
what
has
changed
on
a
score
discovery
now.
To
recap:
this
is
the
next
slide,
please.
The
goal
of
what
we're
doing
here
is
to
allow
devices
to
join
a
group
communication
network
with
a
very
limited
set
of
information
about
the
group,
because
some
of
that
information
might
not
even
be
there
by
the
time
their
Commission,
so
they
might
not.
There
might
not.
J
J
The
point
where
we'd
like
to
store
the
information
and
make
it
available
to
the
device
is
the
resource
directory,
because
that
has
the
nice
advantage
of
allowing
devices
to
observe
where
they
can
join
even
before
the
group
manager
has
arrived
on
the
network
next
slide,
please
so,
we've
received
really
from
from
Jim
on
the
last
document.
Thanks
for
that,
most
of
that
has
been
addressed.
I'll
come
back
later
to
what
has
not
oh
well.
We
still
not
want
to
talk
about
with
the
group.
J
We've
reshuffled
a
bit
of
of
text
about
where
how
will
the
registration
and
happens
and
what
we
described
by
the
group
manager
we've
added
a
way
of
discovering
the
authorization
server
which
it
means
that
the
device
when
it
actually
comes
to
join
in
chrome
kit.
The
group
can
spare
another
round
trip
on
the
network
for
discovering
that
information,
and
there
are
now
examples
in
Coral
next
slide.
Please.
J
So
here
is
an
example
of
of
what
that
link
looks
like
this.
Is
the
information
in
there
is
very
similar
to
what
the
device
would
the
joining
device
would
get
when
it
eventually
tries
to
register
at
the
group
read
at
the
group
managers,
membership
resource
yeah.
It
would
fail
in
its
first
attempt
because
it
might
not
be
authorized
and
that
information
can
be
put
right
into
the
resource
directory.
It's
covering.
J
If
you
remember
from
the
last
presentation
on
this,
there
was
a
bit
of
confusion
about
maybe
the
last
before
about
what
we
mean
by
the
group
name,
because
the
concept
of
a
see
rowed
out
group
identifier
was
still
in
there
that
is,
can
now
and
replaced
with
the
term
that
also
lands
with
what
is
used
for
the
A's
process
of
joining
edger
and
yeah.
Some
some
of
the
other
definitions
were
improved
next
slide.
Please.
J
If
you
look
at
this
example
here,
that
shows
basically
that's
very
similar
registration,
as
we
had
before.
So
on
the
right
side
and
the
in
the
RT
item.
We
see
all
that
metadata.
That
is
there
to
inform
the
client
ahead
of
time
about
which
algorithms
and
which,
which
key
parameters
are
used
in
the
group,
so
that
it
doesn't
need
to
fall
back
to
to
trial
and
error,
and
it
also
has
a
link
further
on
that
describes.
If
that
a
resource
is
accessed.
J
Where
can
you
get
a
token
for
this
from
an
authorization
server,
and
this
is
expressed
as
the
a
su
RI
in
here,
which
is
not
something
that
is
really
particular
to
to
this
document,
but
something
that
we
would
probably
want
to
advertise
in
more
general?
This
is
not
directly
a
link.
You
can
navigate
as
you
click
on
there
and
then
you,
you
are
in
a
different
state
in
your
state
machine,
but
this
is
a
method
item
that
happens
to
be
expressed
in
links.
J
So,
of
course,
the
your
eyes
in
there
all
have
their
usability,
but
not
in
terms
of
hey
I'm.
On
this
on
this
thing
and
I
do
on
to
that
thing,
but
I
am
on
this
thing
and
I
see
that
it
has
a
particular
AAS
and
then
I
can
use
it
in
that
way.
So
I
think
this
is
aligned
with
all
the
all
the
general
ideas
of
of
how
coral
can
be
used,
but
I'm
happy
to
be
correct
in
there.
J
We
do
have
one
example
in
there
that
came
across
as
a
bit
confusing.
It
is
basically
expressing
something
a
set
up,
that's
similar
to
what
is
done
in
pack
net.
This
example
has
a
very
interesting
aspect
to
it.
That
is
that
the
device
doesn't
know
anything
about
the
group
at
all,
but
it
only
knows
the
name
under
which
it
becomes
a
member
of
the
network.
That
is
the
name
under
which
it
registers
in
the
resource
directory.
J
Now
the
information
that
this
device
is
a
member
of
that
is
a
member
of
that
group
can
potentially
be
expressed
in
a
resource
directory.
It
is
not
with
the
terminology
that
is
right
now
and
we
just
write
free,
but
it
could
be
extended
that
way
in
the
last
iteration
of
that
document
and
I
think
even
in
the
in
the
current
one,
we
use.
J
Sorry
we
used
such
a
mechanism
without
being
very
explicit
that
this
is
just
an
example
mechanism
that
could
also
be
used
along
here.
So
there's
the
open
question
of
whether
that's
a
piece
of
information
that
devices
may
want
to
extract
for
menarche
in
a
more
general
context
than
in
the
context
of
of
dancing
Oscar
Oscar
group
memberships,
but
that's
basically
up
to
you.
So
if
you
have
more
use
cases
about
that,
maybe
we
should
pull
this
out.
J
J
So
yeah
we've
we've
asked
for
reviews
already
last
time
we
only
got
one
big
one
so
far,
so
I
think
it
would
help
a
lot
if
we
could
have
a
few
more
reviews
here
and
in
parallel
process.
The
reopen
issue
that
I've
mentioned
I
think
that's
it
from
my
side.
A
A
The
talk
so,
but
you
need
the
the
ast.
Stick
you
talking
before
and
you
register
with
these
extra
parameters.
Like
the
token
a
second
place
on
the
payload,
then
you
specify
the
group
that
you
want
to
reduce
that
the
group
that
the
GM
wants
to
register
the
end
point
to
and
the
token
that
is
being
used.
A
A
J
So
the
slide
of
the
bottom
right
is
what
Francesca
proposed
in
ITF
104
as
one
of
the
many
ways
in
which
perhaps
of
could
develop,
and
this
is
part
of
what
inspired
this
document
and
also
alliance
as
I
think
quite
well
with
what
the
pub
sub
does
right
now.
So
we'll
come
later
to
how
this
could
all
could
all
be
orchestrated
to
pubs
to
into
a
pub
subsystem
to
the
extent
that
the
actual
broker
doesn't
have
to
deal
with
individual
clients
anymore.
At
all,.
J
J
J
J
Now
that
request
cannot
be
utilized
or
IP,
but
conceptually
that
request
was
near
at
some
point
in
time,
and
then
we
distribute
that
phantom
request
or
the
information
containing
that
phantom
request
out
to
the
clients,
and
that
can
happen,
and
here
I'm,
coming
back
to
the
first
point
on
the
slide.
This
can
happen
either
from
the
start.
So
the
server
knows
when
it
creates
a
resource
that
this
might
potentially
be
observed.
J
The
the
reducers
so
when,
when
a
client
tries
to
subscribe
to
return,
tries
to
observe
a
resource
it
receives
and
and
that
resource
is
P
is
available
for
multicast
notifications
and
the
server
doesn't
really
want
to
send
out
individual
responses
than
the
server
in
the
first
place
sends
an
arrow
response,
and
that
error
response
tells
the
client
that
sorry,
you
can't
direct,
observe
that
resource.
But
here
is
some
information
that
would
help
you
achieve
the
same
thing,
and
that
includes
the
serialization
of
the
phantom
request,
which
includes
all
the
the
request.
J
Id
data
and
the
client
can
even
decrypt
that
and
make
sure
that
it's
that
it
that
its
signature
is
correct
and
so
on,
and
it
contains
the
multicast
address
and
can
just
as
well
container
current
representation
of
the
resource
which
is
kind
of
making
the
response
successful,
but
at
least
not
in
the
way
the
client
intended
it
to
be
originally.
So
it's
still
an
error
response.
This
also
makes
things
like
I'm
selecting
content
for
much
much
easier.
J
Next
slide,
please,
since
the
the
last
version
of
the
document
that
has
been
presented,
we've
added
a
section
on
congestion
control.
This
is
largely
picking
picking
topics
that
are
already
about
what
stating
that
they
apply
now
here
as
well,
we've
been,
we
are
now
a
lot
more
explicit
about
how
those
and
from
how
those
responses
are
encoded.
That
is,
there
is
a
similar
format
that
tells
where
which
details
go
where
on
the
message
and
there's
registry
that
allows
extension
of
that
and
gives
the
emptyness
out
the
short
numbers
and
next
slide.
Please.
J
J
We
would
try
to
start
an
observation
and
the
server
would
fail
that
with
the
information
on
how
to
rather
proceed
in
a
multicast
setup.
That's
not
always
the
case,
so
I'm,
just
so
one
one
way
of
the
for
the
client
to
discover
that
information
that
is
in
the
error
response
enough
in
other
cases,
is
when
it
does
topic
discovery
in
a
pops
up
program.
J
Another
example
where
this
could,
where
this
information
could
come
from,
is
explicitly
asking
the
server
about
a
token,
but
that's
more
for
illustrator
illustrator
purposes
to
show
that
it
could.
It
could
happen
in
a
variety
of
ways
and-
and
there
is
not
necessarily
a
single
single
way
of
containing
that
information.
J
J
A
I
mean
yeah,
it's
just
I'm
saying
that
this
this
job
is
really
very
related
to
Thomas's
problem
draft.
The
the
error
description
at
least
and
also
read
the
pops
up
draft,
and
basically
I
was
just
thinking
that,
if
you
are
aware,
was
the
pops
up.
I
know
you
are
even
if
you
are
aware
of
Thomas
s,
draft
for
the
response
and.
J
I
know
we
have
Thomas
is
draft,
it's
not
exactly
the
same
topic,
so
the
the
the
error
response
draft
is
for
things
that
I'd
call
actual
errors
in
the
sense
that
there
is
something
there.
There
is
something
really
going
wrong
here.
This
is
more
like
this
is
not
really
good
and
it's
wrong.
It's
Maura
and
I
kind
of,
but
not
the
way.
You
think
it
would
align
quite
well,
probably
with
with
the
coral
version
of
the
problem
of
the
coop
problem.
J
A
J
K
A
We
haven't
that's
a
good
point.
This
is
a
long
discussion,
but
they
say
that
there
is
a
lot
of
chances
or
is
looking
that
we
could
go
in
that
direction.
Anyways,
yes,
so,
but
my
main
point
is
that
this
sounds
like
an
interesting
topic
for
an
interim
II.
We
already
don't
have
a
schedule,
one
just
like
this
common
areas
between
this
draft,
the
pops-up
draft
and
the
problem
description
draft
I
think
they
are
touching
upon
each
other
a
bit
so
that
that
was
my
main
coming.
That
I
wanted
to
say
in
the
queue.
J
Just
to
summarize,
my
point
here
is
that
I
don't
see
that
there
is
any
particular
application
area
that
would
be
expressed
in
parallel
with
that
you
could
probably
but
I,
don't
think
that
it
will
be
the
common
case.
So
in
those
can
those
can
coexist
and
and
yeah.
It
would
be
good
to
have
a
look.
J
So
what
else
do
we
have
you?
One
thing
that
we
had
as
well,
since
that
or
four
is:
how
do
you
cancel
all
those
observations
which
is
inherently
tricky,
because
you
don't
have
the
good
minutes?
It's
a
group
you're
sending
you
don't
want
to
spam,
the
back-channel,
so
you
don't
have
good
grip
on
the
on
the
on
the
on
who
is
in
that
group,
and
the
proposal
now
is
to
have
a
an
option
in
there
that
gives
that
basically
communicates
this.
J
I
think
that
the
most
tricky
point
around
here
is
that
the
mechanism
we
are
using
to
get
the
responses
back
is
that
we
ask
the
clients
to
register
for
the
observation
in
a
unicast
way
again,
there's
the
recommendation
that
they
do
this
with
no
response
options,
but
in
a
way
we
are
relying
on
side
effects
of
something,
that's
otherwise
a
safe
operation.
Now
we
are
implementing
observation,
so
the
the
fact
of
registering
and
observation
is
something
that
inherently
has
some
side
effects,
but
so
in
in
essence,
I
think
that
this
is
viable.
J
J
C
So
the
idea
is
to
have
something
like
like
cash
value
or
something
that
the
sender
sends
that
essentially
selects
a
small
subgroup
to
send
the
response
like
ree-ree,
observe
or
something
like
that.
So
you
essentially
need
to
maintain
something
like
a
group
size
estimate
at
the
group
head
at
the
side
where,
where
you
are
sending
to
the
group
and
based
on
that,
you
can
choose
the
size
of
that
Center.
So.
J
That's
pretty
much
what
we
are
doing,
the
only
thing
we
do
differently
and
we
might
take
just
take
your
input
and
and
use
that
to
make
it
better.
Is
that
we're
describing
it
in
terms
of
generating
a
random
event?
So
we
say
that
if
I
have
one,
if
I
think
I
have
1,000
observers
and
I
want
to
get
10
responses,
then
I
send
100
out,
and
everyone
causes
a
coin
between
0
and
100,
and
if
that
coin
toss
comes
up
with
a
zero,
then
they
respond
and
statistically
I
get
back.
J
J
J
J
So
but
like
the
question,
the
the
question
is:
do
we
want
to
have
this
in
a
more
dissected
way
in
Quran
versus?
Do
we
just
take
the
coop
message
and
put
it
into
a
binary
field?
So
this
is.
This
is
one
of
the
open
questions
and
the
other
is
the
topic
of
the
the
estimations
that
we
just
talked
about,
where
we'll
just
have
to
go
through
everything
once
again,
to
make
very
sure
that
we
can't
make
make
demises
programming
Smurf
attacks.
Well,
everyone
would
start
sending
their
responses
to
a
place.
A
A
A
H
You
so
the
motivation
of
the
draft
was
actually
the
analysis
of
communication
that
we
did
for
the
draft
of
this
describe.
The
early
use
of
proxies
also
introduces
a
number
of
issues
for
group
communication,
previous
rc7
390.
We
all
I
said
well,
just
don't
do
it
or
it
has
some
issues
now
we
want
to
make
it
work
these
graphs,
so
the
issues
are
that
the
clients
need
to
be
whitelist
and
authenticated
on
the
proxy,
preferably
because
the
proxy
will
send
out
the
podcast
request
on
behalf
of
the
client.
So
it's
better
to
do
that.
Tonight.
H
H
The
guy
may
not
be
able
to
distinguish
responses,
so
if
it
comes
through
the
proxy,
you
cannot
distinguish
the
response
any
more
based
on
the
IP
address
that
sent
back
to
this
was
all
the
responses,
look
as
if
they
are
coming
from
single
device,
which
is
the
proxy
and
also
so
far
with
current
definition,
support.
She
also
doesn't
know
when
to
stop
handling
responses.
H
More
detail
next
time,
please.
So
what
we
described
here
is
a
contribution
that
addresses
the
issues
discussed
and
the
approach
we
consider
is
the
individual
or
infected
clients
assumption
that
obviously
can
do
this,
but
only
if
it's
explicitly
or
figurative
supports
this
method
of
action
is
to
support
also
D
the
options
that
we
are
going
to
define
so
clients.
Our
right
list
of
approxi
can
be
authenticated
by
the
proxy
and
group
or
score
is
also
used
here
or
Qi
as
a
secure
to
secure
the
good
communication
go
to
the
next
right
now,.
H
This
defines
is
a
new
trick,
the
link
protocol
with
too
many
options,
and
so
basically
clients
indicates
to
the
proxy
D.
That
is
that
it
is
interested
and
on
multiple
responses,
because
it
knows
that
in
zoonotic
last
requests
going
to
be
unicast
request.
Those
are
the
time
how
long
the
proxy
should
be
active
and
collect
these
responses,
and
we
also
need
the
server
still
do
something
so
basically
server
and
that
achieves
the
most
awesome
quests
as
to
indicate
its
IP
address
in
the
response.
H
This
option
looks
like
basically,
he
went
encoded
in
there,
that's
taste
time,
so
she
to
be
active.
Collecting
responses
and
the
presence
of
the
option
alone
indicates
that
client
is
interested
to
receive
the
inner
core
which
bounces
signals.
Basically
I
use
multicast
responses
that
come
out
of
this.
H
H
H
You
will
take
the
time
word
value
and
start
the
timer
and
then
note
costs
in
the
request
service.
I
have
to
say
the
time
is
basically
normal.
Sponsors
there's
also
possible,
observe
notifications.
These
are
an
exception.
You
also
need
to
be
handled
beyond
the
time,
so
this
key
observe
client
state
is
cleared.
H
Let's
go
to
the
next,
so
we
see
what
is
serving
us
is
also
signaled.
Hey.
There
is
clients
behind
a
proxy,
that's
created.
They
suppress
that
you're
handling
just
that
by
detecting
the
presence
of
the
not
receiving
name
option
response.
The
server
can
also
include
that
response
forwarding
option
in
there
its
own
IP
address.
H
You
works
again
forward
responses
affecting
client
individually
as
from
ignore
latex
for
late
responses
again
for
the
share
of
notifications,
which
things
need
to
be
passed
through
and
now
the
clients
can
retrieve
the
identity
of
server
from
this
response.
Forwarding
option
and
if
asked
or
security
is
used,
this
identity
or
this
identifier
of
the
clients
is
also
protected
here,
but
it's
for
security
and
after
certain
number
of
time,
when
the
token
can
be
freed
again,
the
client,
please
result
its
token
go
to
the
next.
H
Open
points
because,
since
we
published
a
really
good
review
from
thanks
for
that
points,
to
consider
this
is
led
to
propose
alternative
design.
So
that's
you
could
say
it's
simpler
design
where
the
proxy
only
has
to
be
aware
of
this
model,
our
signal
it
can
remove
it
from
the
rest
and
also
the
proxy,
can
add
the
response,
forwarding
options
to
the
responses
so
that
your
responses
can
be
identified.
H
H
Also,
the
end
server
doesn't
know
that
that
it's
a
proxy
to
request
would
also
be
more
difficult
if
you
consider
a
chain
of
proxies,
maybe
that's
not
not
really
relevant.
Perhaps
it's
not
cause
case.
You
have
to
see
that
another
open
point
was
that
we
thought
about
identification
of
the
client
to
the
proxy
using
all
score,
and
in
this
case
you
have
a
kind
of
double
usage
of
also
score.
It's
a
good
question
to
people
and
should
be
define
at
this
be
appropriate
in
this
document
or
maybe
in
another
one.
H
H
A
Thank
you.
Thank
you,
so
Jimmy's
in
the
queue
I
had
Jim.
I
H
H
B
J
J
If
you,
if
your
client
is
running
in
a
web
browser
and
is
keeping
the
web
browser
through
our
web
sockets
kummel,
then
it
might
ask
the
proxy
to
just
look
around
what
is
there
and
then
it
would
send
a
basically
via
or
coop
all
these
local
co-op
devices
a
broadcast
address
to
that
proxy
and
that
avarice
would
not
even
make
sense
to
itself
to
the
client
itself.
But
the
proxy
could
send
out
the
request
and
the
responses
that
come
back.
Their
addresses
make
sense
in
the
context
of
that
proxy
could.
A
Okay,
I
see
the
queue
is
empty
and
actually
there
won't
be
any
time
for
the
further
presentation.
So
we
will
move
into
the
next
Thursday
and
let's
put
a
pin
on
things,
because
the
Sivir
session
isn't
starting
right
in
two
minutes.
So
thank
you
very
much
for
your
time
and
hope
to
see
you
next
Thursday
at
the
same
time,
and
we
will
talk
about
core
applications
and
about
a
bit
about
cinema
and
a
bit
about
link
relations.