►
From YouTube: IETF95-RADEXT-20160408-1220
Description
RADEXT meeting session at IETF95
2016/04/08 1220
A
A
Good
afternoon
and
thank
you
all
for
coming
all
four
of
you
to
the
red
X
meeting,
so
this
is
rad
except
ITF
95
and
if
you
want
to
be
asked
me
well,
hopefully
you
don't
want
to
be
as
we're
pleased
to
hear.
Otherwise
the
room
is
really
empty,
so
I'm
Stefan,
Lewis's
leonel-
and
this
is
the
agenda
for
today.
So
let's
take
a
look
at
the
agenda
is
published
on
servants
by
the
usual
time
on
preliminaries.
So
let
me
first
start
with
the
no
Tigers
in
javascript.
A
A
A
So
let
me
quickly
flip
back
to
the
note
well,
which
you
should
all
observe
I
have
seen
all
over
the
week.
So
here
we
go
I
hope
you
sing
it.
So
we
can
start
right
away.
The
first
thing
we
have
to
do
is
steal
the
agenda
bashing.
So
if
you
have
any
problems
with
your
agenda
or
something
change,
please
do
speak
up
now,
right,
okay,
so,
let's
start
with
the
document
status,
then,
and
actually
we've
come
a
long
way
here.
A
If
you
take
a
look
at
this
radix
draft
one
of
three
slide,
you
will
see
that
we
have
pretty
much
flushed
allowed
work.
You
all
the
documents
we
had
almost
done
in
the
last
idf's
on
how
pretty
much
often
published,
which
is
very
nice.
The
one
thing
we
have
a
tie
is
to
processing
right
now
is
the
larger
packets
for
radios
over
tcp.
That's
some
hard
times
work.
Sam
has
published
a
05
version
on
25
2015
December.
A
A
This
should
really
work
in
progress.
Lastly,
so
the
slide
radix
drafts
two
or
three-
this
is
now
we're.
Actually
work
is
happening.
We
have
the
radius
extensions
for
IP
port
configuration
and
reporting
from
Dean
Chang.
So
actually
leonel
wrote
the
Shepherd
right
up
already,
but
he
had
some
questions
about
it.
In
the
process
of
these
questions,
many
new
versions
of
the
draft
were
published.
We're
now
at
version
09
published
on
2016
March
18,
not
spending
much
more
time
on
this,
because
there
is
a
slide
deck
on
the
open
issues
on
this
draft.
A
A
A
We
also
have
two
documents
which
are
currently
in
progress
working
items,
one
of
which
is
the
dynamic
authorization
proxying
in
remote
radios
also
identicals
work
together
with
the
unique
conan.
The
last
draft
is
submitted,
it
16
in
January
11
again
we
have
a
slide
deck
on
that
to
discuss
the
remaining
open
issues
there,
and
then
we
have
my
draft
on
considerate
considerations
regarding
the
correct
use
of
response
identity,
which
triggered
a
fair
amount
of
discussion
of
the
mailing
list,
which
I
didn't
expect
at
this
stage,
but
never
mind
that
was
fun.
A
Lastly,
we
have
one
piece
of
new
work
which
is
not
currently
on
our
stack
but
might
become
on,
which
is
the
radius
extensions
for
network
assisted,
multipath
TCP,
which
is
from
image
buca
di
erent
and
colleagues.
Actually,
we
have
a
second
version
of
that.
The
01
was
submitted
january.
Nineteen,
twenty
sixteen
and
the
authors
have
a
slight
deck
on
it.
I
hope
somebody
is
going
to
present
doesn't
really
look
like
it
well,
anyway.
A
What
we
also
have
roughly
on
our
radar
is
some
people
who
submitted
drafts
with
the
radix
twerking
group
named
in
them,
some
of
which
we
talked
about
earlier,
but
we
haven't
heard
from
them
in
a
while.
So
there's
draft
Clemmie
reset
radix,
very
common
vs
a's
and
draft
erevan
try
text
message.
Bundling
these
two
have
seen
a
bit
of
discussion.
There
was
no
real
consensus
to
take
them
into
the
working
groups,
so
I'm
prepared
to
just
drop
them
off
the
radar
unless
somebody
really
thinks
this
needs
discussion
still.
A
I
also
noticed
that
the
data
tracker
has
one
other
new
draft
which
I
haven't
heard
anything
about,
but
it
sits
there,
which
is
called
draft
one
or
a
text,
multicast
radius
extensions.
So
it's
an
active
ID.
Somebody
published
that
with
the
intent
of
coming
to
radix,
but
I
haven't
heard
anything
so
yeah.
If
somebody
wants
to
talk
about,
it
would
be
nice
to
raise
discussion
points
on
the.
A
A
A
F
Yes,
just
as
a
reminder
this,
this
dress
has
passed
working
with
classical
since
last
year,
so
with
the
foreign
key
for
the
review
that
has
been
come
there
recently
so
web.
We
integrate
the
comment
that
we
ship
from
the
union,
but
again
it
may
thanks
for
a
detailed
review.
Yes,
Jerry
I
very
appreciated,
so
we
haven't
agreed
all
the
Knesset
hub
that
has
been
raised
during
that
review
and
also
some
of
a
trail
coming
that
has
been
raised
also
from
from
Alan,
but
still
we
have
one
issue
there
that
is
displayed
in
the
next
slide.
F
It's
about
the
inner
section.
In
fact,
we
have
the
current
approach
in
the
draft.
We
all
the
attributes,
all
the
tvs,
have
the
same.
The
same
name
and
the
same
the
same.
The
same
ID
but
unit
has
a
point,
is
that
he
thinks
that
it
is
more
straightforward
to
to
to
define
dedicated
to
just
wait
for
the
nested
attribute,
so
that
you
that
this
attributes,
even
if
they
have
the
same
name
DM
we
have
identified,
will
depend
on
on
the
parent
attributes.
F
So,
as
author
of
this
graph
don't
have
in
the
am
you
know,
we
will
fill
out
the
recommendation
from
the
working
up,
but
despite
there
is
no
recommendation
in
existing
that
the
RX
is
to
how
to
proceed
with
this
issue.
So
it's
we
were
happy
adore
to
maintain
the
one
which
is
which
is
currently
in
the
draft
or
to
go
with
the
proposal
from
from
your
unit.
So
feedback
is
more
working
on
this,
so
that
we
can
fix
this
soon
and
ship.
The
draft,
hopefully
in
after
this
this
meeting.
B
Yet
yeah
it's
a
little
yes,
the
point
is
that,
as
it
was
defined
and
how
the
registry
is
maintained
are
meant
to
be
maintained,
I
think
it
just
about
codes,
it's
not
about
name
so
at
least
I.
Just
repeating
my
my
comment,
but
I
think
it
would
be,
it
would
be
easier
because
the
main
points
was
the
extensibility
and
to
be
able
to
create
any
number
of
new
attributes
using
a
speaks
name,
whereas
you
are
using
the
same
sub
attribute
in
different
parent
attribute,
will
restrict
the
number
of
possible
values.
B
So
it's
why
I
was
proposing
to
just
follow
the
recommendation
of
the
of
the
of
the
RFC,
defining
the
use
of
the
extended
AVP
attributes
and
to
and
to
just
have
a
unique
identifier
that
will
comply
with,
so
that
will
be
relevant
only
for
apparent
attributes.
So
it
would
mean
that
the
name
need
to
because
we
need
to
have
a
unique
name.
We,
you
will
have
one
name
pair
parents
actually,
so
it
is
a
proposal.
I
think
it's.
B
So
may
I
don't
know
if
Alan
is
saying
something,
but
my
recommendation
would
be
in
this
direction
now,
except
if
someone
has
a
strong
opinion
against
I
think
it
will
be
better
and
it
would
be
a
clear
also
because
it
is
the
first
time
that
we
need
to
use
cyber
tribute
to
multiple
parents
attributes,
but
it
may
come
with
the
next
draft.
We
need
to
have
a
clear
guidelines.
F
G
A
Well,
I,
wonder
and
how
far
we
are
shooting
ourselves
in
the
foot
with
any
other
two
suggestions:
it's
just
allocating
a
different
set
of
integers
for
the
same
things.
I,
don't
see
any
interrupt
problems,
no
matter
which
approach
we
choose.
It's
mostly
beauty
contest.
So
anyway,
I
don't
have
many
steaks.
F
In
yourself
and,
let's
all
say
or
take-
and
in
fact
the
current
approach
in
the
draft
is
more
simple
from
the
cell
phone
there's
no
dedicated
to
register
for
each
of
the
air
tubes,
as
we
already
defined
in
the
draft,
so
I
think
it's
more.
It's
more
simple,
but
I
understand
also
the
one
the
proposed
that
has
been
made
by
value
net,
but
please
give
us
one
day,
I
feel
so
so
that
we
can
speed
up
so
I
think
that's!
We
have
it's
one
year
that
we
are.
B
Yes
video,
but
if
you
want
to
say
that
this
value
will
be
unique,
whatever
the
apparent
attribute,
you
will
need
to
declare
for
all
the
existing
or
next
parent
attributes
that
this
value
can
only
be
used.
This
name
will
be
used
whatever
the
parent
attributes
and
there
is
no
way
to
the
closets
in
the
registry.
Yes,.
F
A
F
F
You
will
have
the
main
issue
so
that
it
will
be
a
guideline
so
to
say
that,
so
we
don't
have
this
kind
of
discussion
again
in
the
future,
so
the
proposed,
unless,
if
you
you
know
as
an
objection,
is
agree
that
we
go
for
the
current
option
that
we
have
on
the
draft
and
during
the
discussion
with
Diana,
we
can
you
know
once
the
draft
aspect
we
passed
the
East
your
review.
Look
at
we
can
we
can.
We
ask
ask
again
this
this
question
if
this
with
Corey,
oh.
B
A
B
But
it
is
something
anyways
that
will
need
to
clarify
because
it
cannot
be
left
for
finance.
It
will
be
in
the
report
of
this
meeting,
but
we
should
have
something
here
because
how
it
is
defined
in
the
RFC.
It
is
said
that
it
is
only
relevant
pair
parent
attributes.
So
we
need
to
make
clear
that
if
it
is
not
the
case
at
least
to
put
an
errata
or
something
as
it
could
be
used
by
anyone
after
that,
because
we
don't
have
to
figure
out
each
time
how
to
do
that.
For
for
the
next.
G
So
is
the
the
real
organization
here:
is
the
ability
to
spoof
the
replies
coming
back
so
when
you
send
the
package
from
a
visited
network?
Oh
my
god,
that's
great
straightforward!
That
works
andreas
its
hardest,
because
there's
the
event
occasion
credentials
in
that
packet.
You
can't
spoof
of
how
users
password
the
packets
coming
back.
The
other
way
the
authentication
data
like
stated,
is
visible
to
all
of
the
proxies.
G
That
means
that
the
whole
proxies
can
spoof
it
minimize
it
by
not
prevent
it.
In
addition
to
that
3576
and
51
others,
it's
written
assume
that
Russell
as
possible,
but
not
really
talking
about
it
and
that
all
the
mass
treated
all
the
attributes
is
intended
on.
The
problem
is,
as
you
start,
adding
/
leaves,
which
are
specific
to
proxying.
You
have
to
make
sure
that
the
local
radius
server
romanum
is
all
those
attributes
earth
where
it
sends
them
fact.
This
is
only
what's
happened
to
the
real.
Whatever
many
people
do
this.
G
If
the
mass
received
proxy
state,
then
it
generates
an
error,
saying
unsupported
attribute:
even
they
fit
a
1
706
says
you
are
required
to
have
a
proxy
silent
and
bad
knees.
We
added
with
a
bit
of
a
text
of
pain
and
Andy,
actually
tickets
ahead,
Emma,
knowing
every
band
has
to
be
able
to
move
coming
room
at
the
workman's
back
to
operate
other
than
Peter
T.
Keep
us
in
comments.
I
will
try
you
like
further
clarification
on
and
I
think.
A
So
one
question
for
my
site
is
who
suggested
two
alternatives
for
the
text
here.
I
can
get
back
here
on
the
second
one
says
all
server
attributes-
and
you
put
it
in
quotes
here-
I-
think
that
from
the
wording
this
should
need
some
clarification,
because
what
is
itself
right
route,
it's
kind
of
undefined
technology
at
this
point
are
so
I'm,
not
totally
happy
with
that
formulation,
but
the
first
one.
Actually
are
it's
nice
to
me.
G
G
After
betraying
you
is
when
you
do
proper
to
a
radius
prophecy
was
and
I
hadn't,
certainly
not
remember
the
related
seal
April,
a
piss
movies
after
a
winter
ass.
If
you
assume
that
a
radius
proxy
and
see
me
so
click
here
in
college,
they
each
know
what
the
other
is.
You
have
to
learn
that
there's.
A
lot
is
that
welcome
to
the
strange
when
your
action
in
ought
to
be
on
the
program
that
makes
sense.
So
it's
a
little
hard
to
talk
to
do.
There's
a
delay
in
the
okanagan
are
very
difficult.
A
So
yeah,
so
what
we
have
in
the
first
one
is
I
are
well-defined
whiteness,
because
you
can
actually
observe
what
the
client
was
in
before
for
what
the
natural
setting
before
and
the
second
one
is
a
not
quite
so
well
defined
blacklists
covering
all
sorts
of
things
and
enumeration.
Those
is
probably
hard.
Oh
yeah.
A
A
So
this
is
about
the
populating
of
identity,
and
it
is
the
thing
that
clogged
up
the
mailing
list
in
the
last
couple
of
days,
so
this
was
finally
accepted
as
a
working
group
item
and
the
changes
in
the
meantime
were
that
this
is
now
going
on
the
bcp
drag
arm.
There
is
a
terminology
section
in
which
I
make
up
some
identifiers,
but
my
choice
maybe
was
not
the
best,
so
we
actually
discussed
this
on
the
manliest
quite
a
bit.
A
The
important
thing
about
this
is
that
we
very
much
converge
that
we
only
care
about
the
one
identifier
that
is
transported.
It
sides
eep
response
identity
and
we
don't
care
about
anything
that
is
happening
inside
the
method
and
that
should
be
about
the
guidance
for
all
following
discussions
are
because
we
have
in
no
way
trying
to
interfere
with
eat
methods
as
they
are
deployed.
A
Well,
the
basic
advice
of
the
document
is
still
unchanged
from
the
times
of
my
individual
draft,
so
you
can
do
on
the
supplicant,
whatever
you
like,
you
can
use
and
holdings
wherever
you
find
them.
The
thing
is,
as
soon
as
you
actually
choose
to
put
a
string
on
the
wire
inside
the
response.
Identity
make
sure
it
is
utf-8
at
that
point
for
the
supplicant.
This
means
yes
maintain
stage.
A
The
second
point,
which
is
probably
bit
more
worthy
of
discussion
as
we
have
learned
last
days,
if
there
is
a
pool
of
more
than
one
method
and
more
than
one
realm
identifier
as
a
consequence,
you
can
choose
from
those.
It
may
be
that
one
obsession
is
not
enough
to
get
a
sexual
authentication,
so
the
one
who
chose
maybe
just
lot
of
work,
and
then
there
is
a
second,
maybe
third,
fourth
alternative,
and
you
should
try
this
out
if
the
first
one
failed-
and
this
means
you
have
to
run
an
entirely
new
EAP
conversation.
A
C
A
So
we've
had
a
lot
of
discussion
on
the
mailing
list,
on
much
of
which
actually
led
to
pending
changes
in
the
draft.
I
was
creating
these
likes
a
few
days
ago
with
many
more
rent
to
be
discussed
points,
but
actually
I.
Think
in
terms
of
what
goes
into
the
draft.
Most
of
the
text
is
now
excellent,
converged
at
and
solved.
A
So
one
point
was
that
I
was
using
these
inner
and
outer
identity
notions
which
many
eat
methods
actually
use,
but
maybe
it
is
better
in
indy
to
to
invent
some
new
terminology
to
steer
clear
of
all
kinds
of
misunderstandings,
in
terms
of
which
method
meant
what
has
an
identity.
So
the
thing
is
that
we
now
have
user
identifiers
and
realm
identifiers,
and
these
are
the
things
I
will
use
in
the
next
revision
of
the
document.
A
Then
the
other
point
was:
can
we
actually
tell
our
z
3748
that
everything
has
to
be
utf-8
at
this
point?
It's
arguably
slightly
beyond
our
mandate,
because
we
are
only
supposed
to
clear
about
the
AP
vividus
transported
over
a
triple
rennsport,
but
most
of
the
time
it
is
and
actually
been
at
a
boba
said.
It
would
be
better
if
somebody
actually
made
that
global
statement
inside
okay,
eep
response
identity
is
always
under
all
circumstances,
utf-8
or
you
doing
something
wrong.
A
So
we
can
have
an
updates
relationship
to
3748
and
we
can
just
write
it
in
there
and
that
should
be
it
local
and
coatings
of
identifiers.
You
know
what
I
said
before.
Oh
it's
true,
so
there
are
no
restrictions
on
what
happens
in
this
applicants.
So
talking
us
Vegas.
What
happens
in
the
supplicant
stays
in
the
supplicant,
but
as
soon
as
you
communicate
with
the
outside
world,
you
use
utf-8
and
that's
it.
A
Then
we
had
an
interesting
discussion
about
what
is
the
relationship
between
the
realm
identifying
the
user?
Identifier
and
one
point-
is
there?
Are
situations
where
you
can't
do
with
one
real
man
in
the
file
could
have
multiple
methods
and
they
have
multiple
different
identifiers
in
the
methods,
and
there
has
to
be
some
kind
of
link
between
what
is
inside
the
method
at
what
is
sent
as
if
response
identity,
it
does
not
mean
they
have
to
be
identical,
they
can
be
totally
different
strings.
A
So
for
that
we
just
say
there
is
a
dependency
of
some
sorts.
It
might
be
that,
let's
say
you
have
five
methods
and
two
of
those
actually
can
be
catered
for
with
the
same
same
realm.
Identifier
lucky
you
so
there
is
not
necessarily
a
one-to-one
relationship
are
could
be
less
relative
fires.
Then
there
are
methods
on
the
system
are,
but
you
don't
know
that.
A
Okay,
yeah
the
logic
in
terms
of
selecting
the
realm
identifiers.
First,
you
select
the
method
with
which
you
want
to
authenticate
with.
Then
you
look
up
which
of
the
realm
identifiers
is
suitable
for
using
that
you
responds
identity
there,
and
then
you
send
bad
realm
identifier
in
response,
identity.
A
In
terms
of
the
user
identifier,
that
was
probably
the
most
controversial
pointed
at
this
point.
So
if
somebody
were
to
create
a
new
eat
method,
probably
it
would
be
a
good
idea
if
that
person
used
nai,
identify
us
for
its
user
identifiers
but
like
if
that
is
for
some
reason,
not
possible
and
also
especially
for
existing
methods,
we're
not
suggesting
any
changes
to
the
deployed
reality
out
there.
So
this
is
only
advice
for
the
future.
A
If
you
happen
to
design
a
new
eat
method,
it
would
be
better
if
you
use
a
magnet
nai
and
finally,
the
last
issue
are
the
realm
identifiers.
What
about
the
local
parts
and
what
about
the
structure
we
want
to?
It
goes
on
those
one
point
which
came
out
of
this
is
that
aah?
What
the
nai
document
suggests
is
that
the
local
part
should
be
empty,
and
there
is
this
backward
compatibility
for
I'm
universe,
that
this
is
just
this
one
spring,
which
is
suggested.
A
If
you
think
you
need
it,
if
you
cannot
do
with
the
low
with
the
empty
logo
part
and
then
the
second
thing
is
that
we
really
want
to
see
any
eyes
are
on
the
Internet
in
the
response
identity
are
so
you
can
have
good
routing
rules.
It's
maybe
too
much
to
ask
it
for
everybody
on
the
planet,
even
inside
enterprises,
because
inside
an
enterprise
you
don't
have
to
hear
about
what
the
rest
of
the
world
does,
but
the
suggestion
was.
It
must
be
in
LA
I.
A
Yeah,
the
last
thing
that
we
didn't
touch
on
the
mailing
list,
yet
I
guess
we
have
to
at
some
point
I
remembered
honolulu,
and
it
was
this
point
discussed
in
the
meeting
that
when
we
talk
utf8
were
you
may
also
want
to
talk
about
normalization,
actually
I,
don't,
but
the
idea
there
was
that
we
should
follow
the
ni
document.
Then
say:
okay,
every
peer
things.
It
would
also
want
to
normalize
the
string
before
putting
in
each
response
identity.
Then
it
can
do
that.
A
A
A
Is
it
in
any
way
normal,
or
can
it
be
expected
that
you
have
more
than
one
method
to
use
for
a
single
network,
and
that
was
pretty
exotic
out
until
recently,
because
if
you
configured
your
supplicant,
let's
say
for
Wi-Fi
enterprise
network,
the
configure
in
your
sample
can
typically
ask
you
so
which
method
do
you
want
to
use
for
this
SSID
and
that's
it.
So
there
is
a
one-to-one
relationship
between
SD
and
the
eat
method.
A
Unfortunately,
the
word
is
getting
getting
more
complex
and
not
true.
Everybody
actually
are
has
realized
that
or
actually
cares
enough
about
this
world.
So
there
is
the
arguably
802
11
interworking
chapter,
which
basically
has
as
a
as
an
underlying
theme
the
SSID
listening
a
string
and
it's
burnt,
and
it's
not
good
enough
to
select
the
network.
We
devise
completely
new
ways
of
arm
for
network
to
show
which
kind
of
identifies
it
supports.
A
One
of
those
is
that
the
Wi-Fi
beacon
can
say
hey
if
you
have
a
cellular
sim
card
and
it
is
from
the
operator
mcc,
123
and
MNC
e45.
Then
you
can
log
into
this
network,
never
mind
the
SSID.
We
don't
care
about
the
SSID
and
there
are
other
things
like
roaming
consortium
identifiers
with
the
same
idea.
A
If
I
come
to
networks
which
support
my
sim
card,
I
will
use
my
sim
card
and
then
the
unpleasant
part
is
when
the
Wi-Fi
network
actually
has
an
its
beacon.
Why
I
society
is
at
your
own
and
you
can
also
connect
with
your
sim
card
and
all
of
a
sudden,
with
these
multiple
independent
ways
of
finding
out
how
I
can
authenticate
to
the
network.
A
The
supplicant
is
stranded
with
a
choice
of
heat
methods
and
they're
totally
independent,
and
there
is
really
no
no
easy
way
to
choose
and
it
might
be
that
one
of
those
doesn't
work
and
the
advice
in
the
document
here
is
then
simply
that
if
you
have
a
choice
and
your
first
choice
didn't
work,
it
didn't
authenticate
you
to
the
network.
Maybe
you
just
want
to
run
a
second
obsession
and
try
this
other
way
of
connecting
to
the
network.
A
That's
all
I,
don't
think
I
can
can
do
much
in
terms
of
text
in
the
draft
dextrous
says
if
you
have
multiple
opportunities
to
use
be
sure
to
iterate
through
all
of
them
on,
but
I
think
adding
this.
This
use
case
of
a
Wi-Fi
network
with
these
multiple
networks,
election
identifiers,
I,
hope
it
gets
things
before
clear
hands
are
well
helps
to
converge
in
the
discussion
on
that
point,
and
that's
it.
D
B
A
I
think
I
think
I'm
the
dot
knew
enough
about
3gpp,
but
at
least
you
have
the
lucky
case
that
the
realm
identifier
for
both
eat
methods
would
be
the
same
course
to
be
at
VG
PP
network
org,
so
both
eat
methods
probably
terminate
at
the
same
server
and
then
eat
method.
Negotiation
can
can
kick
him
and
so
yeah.
B
A
So
I'm
pretty
sure
we
should
keep
the
text
in
the
draft
for
this
extra
piece
of
advice.
Actually,
I
have
a
war
story
about
why
I
actually
wrote
the
text
that
was
years
ago
before
the
work
was
there
that
wasn't
a
potential
problem
of
once
applicant
vendor,
but
I
don't
think
it
matters
anymore.
The
interworking
thing
is
the
much
better
reason
so.
B
Now
have
a
specific
question
because
I
have
not
been
through
all
the
gmail
exchange
recently
on
the
definition
of
the
identity
or
identifiers.
So
first
of
all
is
a
bia,
not
okay
with
the
last.
So
what
is
the
last
status
on
that
when
you
said
so,
it's
means
that
Bernard
is
okay
with
what
was
said
or
when
I
said,
Bernard
is
valid.
Ababa
yeah.
A
I
think
he
is,
but
you
would
have
to
confirm
himself
now.
The
real
problem
is
that
I
used
to
turn
method,
specific
identity
for
the
realm
identifier,
which
gives
in
the
deep
response,
because
it
is
kind
of
specific
to
the
method.
But
it
is
not
the
reserved
method,
specific
identity,
and
that
was
just
the
source
of
much
confusion
and.
B
A
F
Yeah
so
yeah
this
would
be.
It
should
be
quick,
so
yeah
and
yeah
yeah.
This
is
a
proposal
yet
to
the
specific
us
reviews
for
in
order
to
provision
than
in
participate
concentrator
to
m2
to
SCP
next
slide,
please
yeah.
So
this
slide
is
just
providing
you
the
context
of
this
of
this
of
this
work.
Why?
We
are
why
we
are
trying
to
define
why
we
are
trying
to
provision
this
information
to
add
to
DCP.
H
F
Upgrade
to
exit
technology
to
provide
more
bandwidth
and
more
capacity
to
customers.
So
the
strategy
for
for
some
of
this
of
our
network
segments
is
to
aggregate
multiple
access
networks
so
that
we
provide
an
enhanced
quality
of
service
to
the
customers.
So
one
of
the
solution
that
that
is
currently
integrated
in
our
project
is
called
MP
TCP,
and
this
protocol
is
meant
to
aggregate
various
connection
that
are
sent
over
various
access
networks
and
present
them
as
one
single
connection
to
attitude
to
the
customer.
F
So
this
is
for
better
quality
of
experience
and
bitters,
including
service
continuity,
and
also
they
increase
them.
Increased
bandwidth
for
forest
customer
next
type
is
yeah.
So
NP
tcp
in
in
principle
is
an
internal
protocol
which
allows
two
to
signal
in
TCP,
syn
TCP
session
themselves,
the
information
so
that
you
can
tell
the
other
endpoint
that
here
I'm
all
the
set
of
my
IP
addresses
and
my
set
of
port
number
so
that
you
can
establish
multiple
connections
and
then
you
can
glue
all
these
connections
so
that
it
can
be
perceived
as
watching
a
connection.
F
But
the
problem
is
that
for
the
service
providers,
that
we
don't
control
the
neck,
the
terminals-
and
we
don't
want
all
the
servers-
and
today
the
current
rate
of
the
penetration
of
of
impedance
observer
is
that
there
are
only
few
servers.
So
if
you
want
to
talk
to
benefit
from
the
functionality
than
the
filter
or
provided
by
IP
tcp,
we
we
are
currently
investigating
the
need
to
introduce
any
new
features
into
the
CP
itself
and
to
introduce
some
dedicated
service
functions
at
the
network,
which
we
call
the
concentrator
so
the
overall
field.
F
Overall,
there
will
be
three
tcp
legs,
the
first
one
between
the
hot,
the
internal
host
and
the
CP,
and
disappear
from
slate.
In
fact,
the
disappearin
action
into
a
multipath
TCP
connection,
so
the
data
will
be
split
into
available
network
attachment
and
then
the
concentrator
will
terminate
the
sessions
and
then
from
Staten
BAM
into
an
to
a
TCP
connection.
Next
slide,
please
yeah!
So
then
this
document
is
a
focusing
on
the
provisioning
part.
At
the
network
network
side,
there
will
be
multiple
in
participate,
contrast,
concentrator
instances
that
will
be
deployed
in
the
network.
F
Yet
just
one
important
aspect
here
is
that
we
are
not
dealing
with
in
participate
in
general.
For
all
the
confusion
is
work
of
aggression.
We
are
dealing
with
one
specific
configuration
which
is.
We
are
controlling
those
network
accesses,
so
there
is
no
conflict
between
the
Khufu
graph
that
would
be
coming
from
various
networks
because
we
are
controlling
both
the
mobile
one
and
also
the
fix
defects
connection.
The
in
participation
criteria
is
sort
in
a
triple
a
server.
F
This
is
very
important
for
us
because
we
don't
want
to
introduce
another
authorization
phase
between
the
CP
SF
and
the
constant
writer,
and
we
want
to
leverage
on
existing
on
existing
techniques
during
the
network
attachment
as
both
for
the
fixed
and
for
for
the
mobile
itself.
So
this
document
is
about
radius
extensions
and
not
about
MP
disappear
in
general.
Next
slide,
please-
and
yet
this
let's
just
summarize
the
rationality
and
the
design
that
we
adopted
in
this
in
this
draft
is
that
to
say
that
we
are
following
closely
the
recommendation
from
this
working
group.
F
F
Next
slide,
please,
this
is
just
as
there
is
nothing
new
there.
Nothing
specific
to
attitude
to
the
case
is
just
to
tell
you
how
we
can
we
are.
We
are
cancelling,
including
this
attribute
in
ind
architecture,
sweet.
It
will
be
an
so
the
dhcp
server
or
the
last
faction.
Will
we
contact
the
server?
Then
we
see
if
the
customer
is
is
subscribe
to
the
server
we
will
contact.
F
We
will
convey
to
him
the
information
about
the
HTC,
the
concentrator,
ipv4
or
ipv6
addresses
that
will
be
translated
into
a
dedicated,
dhcp
or
dhcpv6
option
so
that
the
CP
can
use
that
the
information
that
we
are
returned
to
active
next
slide,
please
so
I
think
yeah.
We
have
already
received
an
interesting
review
from
Veronica
and
we
integrated
all
the
commands
that
here
he
raised,
we
are
still
can
will
come
in
all
the
commands
that
we
can
receive
from
the
working
group,
and
we
would
like
to
water
progress
this
this
item
in
radix
command
question.
B
B
B
B
A
I
also
have
a
question
about
that
same
diagram,
so
it
I'm
not
sure
where
this
information
exchange
that
happens
in
the
excess
acceptance
here
is
related
to
triple-a.
It
looks
like
the
being
ass
realizes.
Oh
I
would
like
to
send
the
HTTP
option,
but
I
have
no
idea
where
the
concentrators
are
so
it
it.
Does
this
request
just
to
figure
out
something
about
the
network
topology
not.
F
F
A
F
A
B
F
F
B
F
Yeah
this
is
the
activation
of
the
MP
disappeared.
Said
that
mean
that
this
information
about
the
rich
ability
of
your
concentrator,
you
will
receive
that
through
network
has
emerged
by
the
same
entity,
which
means
that
there
won't
be
any
conflict
between
provision
information
if
you
receive
from
various
Network
attachment.
B
F
Cp
dissipate
is
not
aware
about
the
relative
change,
he's
only
what
it
needs
to
receive
information
through
DGP
or
dhcpv6,
or
TR
69,
or
whatever
means
you
are
using
to
configure
your
recipe.
So
the
interoperability
issues
is
something
which
is
again.
This
is
not
the,
for
instance.
If
you
take
the
radish
extension
to
to
configure
to
this
light,
a
siegean,
for
instance,
it's
exactly
the
same
thing
at
you,
you
we
are
doing
here.
Mike.
F
A
G
A
G
G
G
So
this
is
the
affiliate
and
it's
the
only
real
man
am
I.
Gonna.
Find
wireless
network
of
the
essence.
Id,
which
can
be
saving
prepared,
is
an
authentication
the
only
method
of
identifying
the
eep
server.
That
is
beyond
the
present
certificate
when
everyone
knows
to
click
harm
those,
we
can't
trust
the
user
collecting
so
eat
administrator
to
your
revenge.
G
G
G
G
So,
as
it
turns
out,
you
can
generate
a
pretty
convincing
fake
certificate
apply
by
using
local
/
leb
site,
so
something
consented
you've
got
identity
notes
are
just
a
hat
example:
you
can
you
hear
it
to
connect
to
date,
LOL
your
own
clothes,
the
field
or
connect
to
the
Corporal
I
by
if
you're
close
enough,
creates
a
new
radio
server
certificate
for
tensed
user.
All
the
fields
are
corrected
and
leave
you
sign.
Excellent.
Bear
sign
of
people
on
frontier
of
the
difference
between
any
damn
seems
to
be
a
valid
certificate.
User
click.
G
So
what
this
means
pretty
much
anyone
is
capable
of
creating
interacts,
isn't
even
a
story
about
any
other
devices.
Only
the
first
highlighted
someone
move
do
ssi
geek.
This
is
something
that
had
built
very
experienced
with
its
a
whole
lot
of
fun.
Ssit
angels
cannot
stewart
terrific
administration
site
again
so
q0
world,
for
example,
there's
new
cfa
ssit
need
some
international
heights.
Holograph
attacks
on
windows,
at
least
is
something
good
here
and
shows
you
a
little
grape
jelly
on
the
head
of
the
holographic
action.
G
G
F
A
A
A
G
G
A
G
G
A
Well,
okay,
for
at
least
for
our
group
in
engine
room,
I
will
discuss
this
in
our
development
team
and
then
see
if
we
will
actually
put
up
that
advice.
There
are
some
some
ways:
I
can
prevent
that
people
use
the
attack
setup
method,
even
if
you
are
about
it
to
happen.
But,
as
I
said
it's
about
drastic,
oh
well,
let's
see
how
it
works.
B
B
So,
as
you
said,
it
should
not
be
part
of
a
ETF,
but
the
implication
of
relying
on
EP
meter
and
so
on,
and
how
to
use
this
API
method
should
be
I,
think
documented
somewhere.
So
maybe
we
can
see
also
with
with
em,
if
it
could
be
interesting
to
have
some
kind
of
a
information
all
at
least
or
bcp
even
document
describing
the
issue
and
and
proposed
way
to
fix
this
issue.
At
least
we
can
we
can
at
least
we
can
ask
for,
and
we
will
see
what
will
be
the
answer.
B
We
will
discus
at
so.
We
need
to
close
the
meeting
now,
because
we
need
to
rush
to
the
airport
and
for
possible
them
at
least
was
the
chairs.
So
thank
you
allows
thank
you
all
of
you.
People
still
in
the
room
and
people
that
were
in
the
room
and
see
you
in
Berlin
and
please
keep
react,
keep
being
active
on
the
mailing
list,
especially
we
are
regarding
the
existing
documents
that
we
need
to
to
push
forward.
Thank
you.