►
From YouTube: IETF109-ALTO-20201119-0500
Description
ALTO meeting session at IETF109
2020/11/19 0500
https://datatracker.ietf.org/meeting/109/proceedings/
B
B
How
did
that
work
upper.
B
C
B
D
B
Me
the
opportunity.
Okay,
I
want
to
share
screen
two
okay,
great.
Oh,
that
looks
good.
There
is
google
there's.
You
should
see
my
my
browser
now
right
with
the
recharger
text.
C
B
Okay,
so
do
you
see
my
slides
now
yeah?
I
can
see
fantastic
great.
That
seems
to
work,
so
I
guess
we
can
get
started
everybody's
here,
so
our
air
director
is
here
quinn.
Our
new
chair
is
here
and
vijay
might
not
make
it.
I
say
something
to
that
in
a
minute,
so
I
guess
yeah.
It's
time.
We
should
start
welcome
to
the
ultra
working
group
at
itf
109.
My
name
is
jensirov.
B
I'm
co-chairing
this
working
group
together
with
our
new
chair
quinn,
wu,
who
is
here
and
also
with
vijay
gurbani,
who
might
not
be
here
vijay-
might
not
make
it
for
personal
reasons,
I'm
not
sure.
Maybe
he
can
drop
in
and
quinn,
and
I
discussed
with
him
that
we
are
going
to
take
care
of
business
today,
so
that
should
not
be
a
problem.
B
E
A
B
Okay
brought
that
up,
we've
been
mailing
a
lot,
but
we've
not
been
talking
in
person
so
far,
okay,
yeah
so
yeah.
I've
got
to
talk
and
get
to
that
also
in
a
minute
so
not
well.
I
think
everybody
is
aware
of
that,
and
this
is
the
ietf
general
rule
set,
which
basically
says
if
you
are
aware
of
any
ipr
or
if
you're
aware
of
other
important
things,
you
need
to
mention
that
we
are
here
under
full
transparency
in
the
itf.
B
So
if
you're
not
aware
of
this,
please
read
the
rfcs
that
I
mentioned
here,
and
this
is
a
general
rule
set.
That
applies
which
basically
means
yeah.
We
are
fully
transparent
and
you
need
to
mention
any
kind
of
related
ipr
information
or
other
things.
You
you
know
right
and
so
yeah
we
have
a
new
chair.
You
want
to
introduce
yourself
briefly.
E
Oh
okay,
yeah
thanks
yeah
and
for
kind
of
introduction
actually
for
people
who
don't
know
who
I
am
actually
my
name
is
qinhu
and
I
work
for
huawei
and
most
you
know
looking
into
the
enterprise
network
and
the
network
management
related
technology.
E
So
I
used
to
chair
l3,
sm
l2sm
in
ops
area,
I'm
currently
one
member
in
in
the
officer
directory
and
actually
author
is
not
new
to
me
and
and
six
of
seven
years
ago.
Actually
I
get
engaged
into
the
auto,
and
now
I
come
again
actually
very
excited
to
join
this
author
team
again.
Thank
you.
B
Thank
you
yeah
welcome,
as
I
said,
chin
welcome
here,
and
we
are
happy
to
have
you
on
board
and
I
think
you're
doing
great.
You
already
wrote
lots
of
emails
in
the
past
days
clarifying
many
things
preparing
the
session,
so
great
yeah,
and
so
before
we
go
on.
We
need
a
javascribe
and
a
notetaker.
Do
we
have
that.
B
C
So,
okay,
you
can
hear
me
I'll
volunteer
as
one
of
the
note
takers
we're
going
to
use
like
cody
md
is
that
right.
C
E
A
There's
actually
a
button
in
the
upper
right
corner
of
meat
echo
it's
for
the
note
taking
tool
and
that
can
get
you
directly
to
the
code
emd
as
well.
C
E
Yeah
yeah,
you
you're
right
so
richard
you,
you
actually.
B
B
Fantastic
great,
so
the
other
thing
we
officially
need
is
a
javascribe.
I
don't
know
if
it's
really
necessary,
because
we
also
have
now
meet
echo.
I
mean
java.
Javascript
is
for
me
the
two,
the
itf
used
to
use
a
lot
when
we
were,
you
know
at
actual
physical
meetings
and
we
didn't
use
meat
echo
and
now
we
have
meat,
echo
and
everything
is
being
recorded.
A
So
so
young,
the
the
chat
and
meet
echo
is
jabber,
and
so
the
purpose
of
the
purpose
of
having
a
jabra
scribe
is
if
the
chairs
do
not
feel
equipped
to
monitor
the
jabber.
At
the
same
time
is
sharing
the
meeting.
So
the
the
purpose
of
the
jabra
scribe
is
to
mention
important
stuff
in
jabber,
because
no
one
else
is
going
to.
B
Okay,
I
was
going
to
say
the
main
reason
is
to
so.
If
somebody
raises
a
hand
in
the
in
the
chat
that
we
we
don't
miss
it
so
so
I
was
gonna
ask
chin.
Can
you
have
a
look
at
the
at
the
at
the
chat.
E
B
Wonderful,
okay,
great,
I
think
we
are
we
are
we
have
the
logistics
set
up.
This
meeting
is
being
recorded.
I
notified
everybody
about
the
note.
Well,
we
I
introduced
our
new
chair.
We
have
note
takers,
we
clarified
the
javascript
thing
so
yeah,
let's,
let's
talk
about
our
working
group.
What
what
what's
going
on?
So
I
think
everybody
is
fully
aware
that
we
have
two
big
topics
and
that's
what
what
this
slide
basically
says.
On
one
side:
we
want
to
finalize
our
charter
items
and
we
are
making
good
progress.
B
B
Normally,
there
are
only
two
options:
either
the
working
group
is
finished
and
ends
or
there's
lots,
lots
of
interesting
stuff.
That
still
needs
to
be
done
and
we
need
to
recharter
them,
and
my
impression
is
clearly
that
we
have
a
lot
of
energy.
We
have
lots
of
people
with
good
ideas
and
we
had
a
recharger
discussion
already
last
time
and
in
the
meantime,
the
people
who
are
working
hard
on
this
have
really
putting
a
lot
of
effort
to
focus
this
discussion.
B
So
I
think
we're
gonna
have
a
good
recharge
discussion
today
and
that's
what
we
want
to
focus
most
of
our
time
and
if
you
look
at
the
next
slide,
oops
sorry,
so
the
agenda
has
been
set
up
accordingly.
So
I
will
talk
about
presented,
share,
slides
this
on
the
next
few
slides.
I
will
talk
a
bit
a
little
bit
about
the
current
milestone
and
their
current
status.
What's
going
on,
we've
had
lots
of
working-class
calls
and
we've
had
lots
of
revisions
lately,
and
so
we're
going
to
talk
about
that.
B
And
after
that
we
got
to
have
a
recharger
discussion
where
we
have
some
presentations
that
people
are
going
to
show
quickly
to
to
highlight
the
current
charter
items
that
are
on
this
recharger
discussion
and
also
richard
posted
a
link
on
the
mailing
list
that
I
want
to
highlight
with
the
current
recharter
proposed
text.
Let's
work
in
progress
text
so
and
but
people
can
have
a
look
at
it
and
we're
going
to
discuss
that.
B
So.
Regarding
our
general
progress,
the
last
meeting
was
itf.
One
way
there
was
a
virtual
itf
meeting
just
as
this
one
is,
and
during
the
due
to
the
corona
m
crisis.
I
think
everybody
is
fully
aware
of
that
and
we
have
published
two
rfcs
since
our
last
gathering
together,
which
is
great,
so
we
have
the
course
calendar
published
as
rfc889
and
we
have
incremented
updates
using
sse
published
as
rfc
8895.
B
So
congratulations
to
the
authors
and
I
think
that's
a
big
step
forward
for
us
two
important
work
items
and
milestones
that
we
had
so
yeah
we're
making
good
progress,
and
we
have
four
remaining
milestones
now
and
I'll
talk
about
each
of
them
on
the
next
slides
and
you
can
see
them
here.
It's
the
cdni
work,
the
path
vector
performance,
metrics
and
unified
properties
documents
and
regarding
alto,
cdna,
request
routing.
This
is
the
one
that
has
been
done
or
finished.
B
I
think
the
for
the
longest
time.
Actually
we
finished
the
working
glass
call
in
the
zero
nine
version.
There
have
been
no
objections.
We
had
quite
a
lot
of
comments.
We
have
addressed
them
all.
I'm
saying
we
because
I'm
a
co-author
of
this
document
and
yeah
speaking
as
chair
now
the
current
version,
zero
14
and,
I
think,
yeah
the
current
version.
Zero
14
is
ready
to
go.
B
The
only
reason
why
we're
not
moving
forward
with
it
right
now
is
that
we
agreed
that
we
want
to
move
it
as
a
bundle
with
the
unified
properties
document,
because
it
kind
of
relates
to
that,
and
we
want
to
make
sure
that
these
two
get
published
together.
So
basically
it's
ready
and
we're
just
holding
it,
and
we
chairs
already
started
with
a
shepherd
write-up.
So
vijay
is
the
shepherd
for
this
document,
because
I'm
a
co-author
and
but
I've
been
exchanging
emails
with
him
and
he's
already
making
progress
on
that.
B
So
the
shepherd
write
up
is
actually
almost
ready,
so
yeah.
Essentially,
this
is
done.
We're
just
waiting
for
unified
properties
to
move
ahead
and
chin.
You
made
a
good
comment
in
an
email.
I
think
I
it
was
not
animating
this,
but
sent
privately
to
me
or
the
authors,
and
that
we
need
to
make
sure
that
unified
properties
is
then
also
a
normative
reference
in
this
document.
B
It's
a
good
comment.
Maybe
you
can
resend
it
that
we
don't
forget
it.
I
am
yeah.
I
guess
that's
the
case
that
also
rfc
editor
will
notice.
So
in
the
shepard
writer,
we
usually
make
sure
that
there
are
no
non-normative
references
in
the
normative
references
section.
If
it's
the
other
way
around,
that
we
reference
a
current
draft
that
will
requires
a
bundle
and
we
also
be
published
as
an
rfc.
I
think
rfc
editor
will
will
take
care
of
that.
B
E
G
E
Yeah,
I
I
said
yeah,
as
you
mentioned,
I
already
you
know,
send
a
private
email
to
you
and
I
will
you
know,
posted
my
comments
to
the
list.
E
To
reminder
the
sitting
I
dropped
the
author
to
make
update.
Actually
so
main
commerce
actually
is
suggested
to
you
know:
move
unified
property
from
informative
reference
to
the
normative
reference.
B
Yeah
correct
yes
and
that's
a
good
comment
and
we
need
to
take
care
of
that,
but
the
problem
always
is
once
we
send
it
out
to
rfc
editor.
The
the
unified
properties
will
not
be
an
rfc
yet
either.
So
we
have
to
make
sure
that
rc
editor
notices
it,
but
I
think
it's
a
common
thing
that
two
rfcs
get
published
together
and
you
need
to
make
sure
there
are
normative
references
and
yeah
we'll
take
care
of
that.
B
B
So
vijay
has
been
shepherding
this
document
and
he's
not
here.
So
I
am.
I
want
to
make
sure
that
we
have
addressed
our
comments
and
what
the
status
is.
Can
the
authors
at
least
confirm?
I
mean
what
is
have
I
my
assumption
is
that
in
the
zero?
Sorry
in
the
version
12
you
have
addressed
all
comments.
Is
that
the
case.
C
I
Okay,
I
just
try
to
speak
louder,
yeah.
I
The
latest
revision,
and
but
after
the
submission
richard
also
gave
some
other
comments,
so
we
might
just.
Let's
see
he
also
proposed
some
changes
to
address
some
comments,
so
we
might
accommodate
the
changes
after
basically
after
the
new
submission
window
is
open,
and
then
we
will
submit.
C
Right
so,
by
the
way,
kai
you,
you
probably
want
to
try
to
figure
out
how
you're
handling
a
microphone.
I
think
a
chi
props
amount
of
very
few
people
who
use
this
linux
and
open
time
you
have
all
kinds
of
issues
and
so
on,
so
because
otherwise
you
you'll
present
later,
I
don't
know.
C
Cool
yeah,
so,
basically
let
me
clarify-
and
I
think
the
small
issue
is
the
following:
there,
there,
there
there's
various
very,
very
good
comments
from
luis
from
the
last
review,
and
I
think
kai
addresses
the
reviews,
and
I
also
address
the
reviews-
there's
some
like
overlapping
part.
So
therefore
we
address
slightly
differently
and
so
on.
So
I
think
we
need
to
kai
and
I
we
need
to
go
over
the
our
basically
the
revision
and
murder
of
revision
in
the
final
version.
I
think
that's
that's
all
right.
C
I
think
that's
the
one
so,
basically
just
make
sure
we
both
of
our
revisions,
will
be
merged
into
a
single
coherent
one,
so
we
slightly
took
a
slightly
different
way
to
address
lewis,
but
I
think
I
checked
my
my
answers
with
luis.
I
think
I
believe,
replied.
I
think
louise
said
if
he
was
okay.
With
my
comments
I
think
kayaking
there
also
some
come
very
nice
revision
by
kai,
very
nice.
We
just
need
to
very
quickly
probably
write
up.
Ipf,
maybe
take
a
couple
days,
we'll
finalize
it,
but
they're
not
hacking.
C
Mostly
they
are
awarding
you
not.
B
Like
potential
okay,
thanks
richard
that
would
that
would
have
been
my
question.
So
I'm
also
under
the
assumption
that
there
are
no
technical
issues
that
we
need
to
discuss
right
here
now
and
there
are
also
no.
This
also
implies
that
there
does
not
need
to
be
another
working
class
call.
So
basically
it's
editorial
things
and
that
you
as
authors,
need
to
sort
out
okay,
so
great.
So
then
please
do
that
as
soon
as
you
can
so
we
are
taking
notes.
Oh.
B
Oh,
I'm
sorry,
then,
maybe
that's
the
cut
and
paste
error
on
the
slides
when
I
did
them
yeah.
Probably
that
is
possible,
I'm
sorry
that
can
very
well
be,
and
so
that
actually
the
reason
why
that,
where
this
error
on
my
side
happened,
is
that
that
there
has
been
lots
of
traffic
on
the
mailing
list
lately
and
that's
that's
a
good
thing
so
and
we've
had
many
working-class
calls.
So
maybe
that's
a
cut
and
paste
error.
I
apologize
yeah
so,
but
but
to
to
wrap
this
up.
B
So
we
are
also
vijay
and
I
have
already
started
the
shepherd
write-up,
and
so
we
are.
We
are
good
to
go
here.
So
basically,
my
assumption
is
the
working
last
call
is
over
no
objections.
E
So
so
yeah
for
past
available.
Actually
I
did
have
a
quicker
review
to
this
draft.
I
I
think
this
path,
vector
is
a
you
know
very
important
worker
in
the
old
working
group,
because
the
original
base
order
protocol
is
designed
for
the
peer
selection.
Now
we
actually
move
toward
to
the
path
selection,
so
pass
vector,
is
a
funimation
worker
for
this.
So
it's
very
important
to
deliver
work,
but
I
have
additional
comments
because
I
I
didn't
rate
the
abstract-
I
I
think
it
would
be
desirable
to
split
it
into
several
paragraph.
E
I
I
can
you
know,
send
my
proposal
to
the
list
to
have
the
author
to
maybe
also
to
con,
take
a
look
at
that
we
can
to
try
to
make
abstract
more
clearer
and
more
shorter.
That
would
be
great,
of
course,.
J
E
In
addition,
I'm
just
wondering
what,
for
the
next
step,
do
you
think
it
will
be
good
to
cross
cross
post
to
some
other
working
group
like
a
pc,
or
we
need
to
ask
some
directory
like
a
routine
director
to
to
give
a
quick
review
all
this
is
a
all
the
next
step
since,
after
working
with
law,
school.
B
B
So
that's
the
normal
process.
We
have,
and
I
think
it
works
as
well
here.
Martin,
do
you
want
to
add
something.
A
Yeah,
so
you
can,
you
can
always
request
cross
area
review,
it's
actually
just
the
button
in
data
tracker.
If
you
want
ops
or
whoever
to
look
at
it,
you
just
ask
and
it
goes
into
their
queue.
So
you
don't
have
to
wait
for
it.
I
mean
you
could
do
it
like
on
a
zero
zero
draft.
If
you
want
it's,
it's
totally,
you
don't
have
to
wait
till
after
working
class
call
or.
B
Yeah,
I
understand,
but
I
guess
but
the
point
here
is,
since
we
are
really
really
really
close
to
being
done
with
this
one.
Do
we
want
to
have
this
cross
directory
review
triggered
or
manually
now
by
us,
or
do
we
just
push
it
out
of
the
working
group?
Because
it's
going
to
happen
then,
anyway
doing
the
isg
review
right
so
well,.
A
So
when
first,
we
have
itf
last
call,
which
is
probably
which
is
actually
the
best
opportunity
to
get
good
cross-area
view.
It
is
true
that
iesg
is
also
across
area
view,
but
it's
a
different
person.
So
if
you
want
like
the
right
ops
person
to
look
at
it,
I
would
suggest
that
in
in
last
call
you
request
an
officer
interview.
They
might
just
do
it
on
their
own
anyway,
but
you
can.
B
Request:
okay:
well
then,
that's
verified,
then
please
for
the
note
takers.
Also,
let's
take
a
note
that
we
do
this
and
we
were
so.
I
think
chin.
This
also
asks
you
a
question
answers
your
your
comment
or
your
proposal.
Then
we
will
request
it
during
itf
last
call
great:
okay,
yeah,
okay,
yeah,
so
again,
authors
of
path,
vector-
and
please
finalize
this
as
soon
as
you
can,
so
that
we
can
really
get
this
done
before
we
recharge
her.
That
would
be
great.
B
Performance
metrics,
so
this
one
I
would
like
to
get
some
feedback
from
the
authors,
so
I
had
a
close
look.
We
had
a
working-class
call
on
the
12
version,
therefore,
no
objections
which
is
great,
but
we
have
the
three
individual
reviewers
thanks
to
kai,
danny
and
jensen.
I
think
these
this
time.
The
reverse
are
correct,
that
I
wrote
here
and
yeah.
My
understanding
is
that
the
authors
are
still
working
on
revising
the
documents.
B
So
as
far
as
I
saw
on
the
on
the
mailing
list,
but
also
in
the
data
track
and
a
new
version
has
not
been
posted
yet
I
also
had
a
look
at
preparing
the
shepherd
writer
when
I
found
some
normative
references
issues
that
you
should
fix,
I
posted
it
on
the
mailing
list
as
well.
So
please
have
a
look
richard.
Yes
right.
C
That's
right
so,
basically
we're
addressing
some
of
the
comments
and
I
think
we
missed
because
wrestling
reviews-
and
basically
I
I
took
some
responsibility.
You
know,
since
I
think
I
missed
the
review
from
I
found
out.
I
found
out
that
there
there
were
reviews
by
jameson
like
two
or
three
weeks
ago
during
the
weekly
meeting,
and
I
realized
I
was
missing-
some
emails
from
jensen
to
the
auto
meetings.
C
So
therefore,
therefore
I
I
I
started
to
really
address
it,
then
I
missed
the
the
we
missed
the
deadline
of
the
update,
so
we
do
not
want
to
read
up,
for
example,
right
away
during
ietf
week,
but
we
will
finish
all
the
reviews.
We
don't
see
any
techno
technical
comments.
I
believe
they
are
mostly
clarification
issues
and,
of
course,
the
nominal
mutual
address
and
so
give
us
basically
one
week
of
atf
and
we
will
send
upload
the
final
version
and
then
you
can
do
the
red
tap.
B
Great,
so
thanks
richard,
for
that
summary,
this
is
exactly
what
what
I
was
hoping
for,
and
it's
good
that
you
also
pointed
out.
There
are
no
technical
issues,
that's
also
my
understanding.
B
So
basically
there's
a
many
editorial
comments
that
you
just
need
to
work
out
and
great
yeah,
please,
if
you
can
do
it
as
soon
as
you
can,
that
would
be
great.
I've
already
prepared
a
shepherd
write-up,
and
so
it's
almost
done
and
then
we
just
need
to.
I
need
to
just
ask
the
ipr
question
to
all
authors,
and
then
we
are
pretty
much
done
with
this
one,
great
okay,
so
moving
to
unified
properties,
we
had
a
working
class
call
in
the
12
version.
H
Do
you
hear
me?
Yes,
oh
okay,
so
while
I
was
activating
my
mic,
I
lost
your
questions.
You
were
asking
whether
I
want
to
comment.
B
H
No
just
want
to
to
to
answer
to
your
questions.
So
all
the
comments
have
been
addressed.
We
posted
the
version
14
a
couple
of
days
ago
because
there
were
a
couple
of
pending
comments
from
luis.
So
thank
you
to.
G
H
And
danny
for
the
review,
we,
what
misses
is
just
the
one
table
that
summarizes
that
is
listing
all
the
new
features
that
are
introduced,
and
we
want
to
put
this
in
the
beginning
of
the
document,
and
so
we
with
the
jensen.
We
have
prepared
the
list
and
the
just
the
references
to
the
text
sections
and
we
will
inject
it
like
tomorrow
or
let's
say
monday.
B
Okay,
thank
you
also
here
for
the
for
the
good
summary
and
again
I
my
understanding
is
that
these
are
non-technical
issues
that
we
need
to
discuss
here.
It's
editorial
issues,
and
so
yeah,
please
as
soon
as
you
can
and
do
this,
but
my
understanding
is
also
this
one
is
done
you
just
you
publish
a
new
version
with
the
last
outstanding
editorial
issues.
I've
saw
some
emails
also
on
the
mailing
list.
B
Regarding
this
table,
I
think
bj
gave
some
comments
and
vj
already
posted
a
shepherd
write
up,
so
he
has
it
ready
so
yeah.
We
are
also
very
good
to
go
here,
yeah.
So,
in
summary,
I
think
all
the
remaining
milestones
we
are
really
really
close.
Now
to
to
being
done.
We
have
finished
working
glass
corner
call
on
all
of
these
and
for
none
of
you.
A
First
of
all,
thank
you
for
this
report
on
on
how
it's
going,
and
I
just
could
I
ask
the
chairs
to
update
the
state
and
data
tracker
correctly
for
all
these
documents.
It's
kind
of
like
where
I.
A
Myself,
where
things
are
and
and
I'm
going
to
reach
for
that
much
sooner
than
I'll
reach
my
notes
from
this
meeting.
Thank
you.
B
E
Yeah,
so
so
yeah.
Actually
I
have
a
additional
comments
for
this
job.
I
also
think
this
dropped.
You
know
more
major
enough.
Actually
you
know
he
is
a
reusing
component
actually
can
be
used
by
the
cdni
and
also
pass
vector.
So
but
I
my
suggestion,
maybe
we
should
align
with
the
passive
vector
job
because
for
the
title
you
know
for
unified
property.
What
is
the
unified?
So
I
I
know
this,
you
know
unified
property,
really,
you
know
extended
and
endpoint
property
try
to
generalize
this
endpoint
property
better
than
the
naming.
E
Maybe
you
know
not
very
clear.
Maybe
you
know
just
you
know,
use
a
similar
title
as
as
a
pass
vector.
Actually
that
will
be
more
more
easily
understand,
understood.
B
I
have
personally
no
problem
with
it,
but
I
guess
it's
a
question
also
to
the
authors
or
to
the
working
group.
If
we
want
to
change
the
title,
maybe
changing
title.
E
To
the
property
mapper,
I
don't
know.
B
Or
any
of
the
authors,
do
you
have
any
any
yeah
any
feedback.
H
Yeah
hello,
so
we
had
a.
We
did
have
a
discussions
about
the
title
for
this
document
and
so
well.
We
can
resume
the
discussion
if,
if
if
this
is
something
really
that
brings
like
that
like,
if
the
title
lacks
clarity,
we
can
reconsider
it
and
we'll
have
a
discussion
with
a
richard
and
yes.
B
Yes,
maybe
include
also
jin
in
this
discussion,
because
he
he
thought.
J
B
Perfect
yeah.
That
sounds
good
to
me
exactly
clarified
on
the
list.
I
I
have
no
objections
and
I
guess
it's.
It's
also
very
good
comment.
If
the
title
is
clearer,
that's
good,
but
of
course
the
authors
need
to
be
involved
and
that's,
let's
have
the
discussion
on
the
mailing
list.
Okay,
but
also
that
is
for
me.
It's
not
it's.
Of
course
it's
a
non-technical
issue,
so
it
doesn't
well
it
it's
still
in
line
with
what
I
said
before,
that
we
are
done
with
all
the
remaining
milestones
we
have.
B
We
are
done
with
working
class
call
and
we
are
really
really
close
to
publishing
these
and
and
we
will
update
the
status
on
the
on
the
data
tracker.
I
took
a
note
yeah
so
that
that
that's
good
and-
and
we
are
actually
perfectly
in
time
we
we
set
this
goal,
it's
going
to
take
half
an
hour
and
it
exactly
took
half
an
hour.
So
the
question
is
we.
B
I
would
now
move
to
the
recharger
discussion,
but
before
we
do
that
the
question
is:
is
there
any
other
business
anything
and
that
we
forgot
to
mention.
C
B
C
B
Thank
you
very
much
yeah.
We
should.
I
know
that
other
working
groups
do
this
kind
of
thing
where
if
we
would
have
physical
itf
meetings
for
sure
we
should,
we
would
go
for
a
drink.
Yes
now
we
will
have
virtual
drinks
yeah.
I
guess
it's
a
thank
you
very
much.
I
think
it's
a
good
idea.
If
we
will
have
a
physical
meeting
soon
again,
we
should
maybe
have
a
gap
together.
That's
a
good
idea.
I
think.
C
Right
right,
because
the
reason
I
want
at
this
point
is
given
that
and
young
and
vegeta,
I
don't
know
who
might
need
to
you
know
and,
for
example,
switch
a
little
bit
and
because
I
really
want
to
say
that
I
think
the
both
vg
and
young
are
just
wonderful
cheers
to
organize
this.
For
example,
just
look
at
how
we
organize
all
these
working
group
documents
and
then
greater
leadership
in
particular,
for
example,
as
a
good
example.
C
Lately
we
were
writing
this
recharger
text
and
we
we
read
the
current
version
and
just
incredible
clear
direction
about
what
charters
would
be
and
then,
of
course,
always
very
steady
and
very
well
organized
just
wonderful,
I
think,
really.
As
a
working
group,
we
probably
should
set
this
one
up
as
some
kind
of
probably
informal
business
to
appreciate.
B
B
Much
it's
yeah,
I
appreciate
it
and
I
will
also
I
think,
vijay
does
too.
I
will
talk
with
him
about
it
thanks
a
lot
and
I
will
take
a
reminder
if,
if
we
have
a
physical
meeting
soon,
who
knows
with
the
vaccinations
coming
out,
maybe
you
know
already,
and
I
we're
going
to
have
an
itf
meeting
sooner
than
we
think
again
in
an
actual
at
an
actual
location,
we
we
can
celebrate
for
sure
yeah.
E
E
E
B
B
Why
not?
Oh
it's
good!
I
mean
it
was
actually
quite
some
work.
I
mean
of
most
of
most,
of
course,
by
the
authors,
but
also
sharing.
Is
this
work
and
yeah?
It's
a
good
mice.
It's
a
big
milestone.
If
we
have
finished
this
work
and
it
looks
like
we're,
gonna
get
there
really
really
soon
so
yeah.
Take
a
note
we
should.
We
can
set
up
a
special
card
to
celebrate.
That's
a
good
point.
B
Yeah
I
mean
martin
knows
more
than
than
I
might,
but
regarding
upcoming
itf
meetings,
I
think
the
official
schedule
is
itf
march
will
still
be
virtual,
but
itf
july
next
summer
is
officially
scheduled
to
be
taking
place
in
san
francisco.
C
B
Yeah
yeah
and
probably
that's
realistic.
I
guess
I
guess
that's
realistic
summer
is
maybe
I'm
too
hopeful
for
summer
already.
Yes,
okay,
but
let's,
let's
get
to
the
recharger
discussion.
Thank
you
richard
for
this
comment.
I
appreciate
it.
Let's
get
to
the
recharger
discussion
so
yeah.
How
are
we
going
to
do
this?
You
guys
uploaded
slides.
Do
you
want
to
go
through
the
items
and
then
briefly
present
five
minutes
your
slides
and
then
have
a
discussion
on
each
item.
I
guess
that's
how
we're
going
to
do
this
right.
E
Yeah,
I
I
think
yeah
as
we
discussed
overnight,
maybe
yeah
we
maybe
we
need
to
just
focus
on
the
child
charter.
You
know
text
proposal
text
and
if
we
have
any
you
know
detail,
we
need
to
get
a
detailed
discussion.
We
can,
you
know,
run
the
slides.
I
I.
G
E
C
Right,
I
think
you
you
set
up
this
order
like
overview.
I
think
you'll
give
a
very
quick
for
other
people
who
who
did
not
attend
the
meeting
like
every
tuesday
nowadays-
and
I
think,
probably
basically
last
tuesday,
which
is
like
around
two
days
ago,
and
I
think
we
did
agree
on
that
order
will
be
qing
would
give
the
overview,
and
then
each
one
of
the
champion
will
give
a.
I
think
we
talk
about.
I
think,
and
young
you'll
give
us
like
10
minutes
to
go
over
our
some
details
and
then
post.
C
B
All
right,
that's
that's
how
we,
how
we
will
try
this
so
it
just
for
everybody
to
to
be
clear,
and
I
just
shared
now
this
proposed
charter
text
or
the
the
current
text
on
the
table.
Richard
has
also
posted
this
link
on
the
mailing
list.
So
everybody
you
should
be
able
to
access
this
google
docs
text
that
I'm
sharing
here
yourself.
So
it's
it's!
The
link
has
been
posted
on
the
mailing
list.
B
E
E
J
E
Okay,
so
as
we
discussed,
actually,
we
want
to
set
an
order
for
these
several
topics.
Yeah,
I
haven't
followed
the
previous
discussion.
We
have
15
proposals
now
we,
you
know
converge,
you
actually
have
a
vibrable.
I
I
think
we
make
a
lot
of
progress
and
so
the
reason
to
set
this
order.
We
think
you
know,
based
on
the
maturity
of
each
proposal
and
and
also
based
on
the
dependency
of
this
proposal,
so
we
set
this
order.
E
The
first
proposal
is
a
cellular
network
exposure
proposed
by
chen
chung
and
from
tencent
and
gandhi
from
china
mobile,
and
I
I
think
you
know
foundation
for
this
cellular
network
exposure
is
3gbp
worker.
They,
you
know
they
focus
on
to
define
network
exposure,
function,
architecture
and
requirements
and
related
api,
so
this
will
provide
the
foundation
for
this
cellular
network
exposure,
and
the
second
proposal
is
about
the
general
extension
which
is
proposed
by
the
saving
from
nokia,
and
I
I
think
this
is
you
know.
E
As
luis
mentioned
this
is
you
know,
optimization
to
the
existing
protocol.
We
now
we
already
have
you
know
past
vector
chapter
which
help
you
to
to
to
do
the
past
selection,
and
we
may
need
additional
pieces
to
to
help
you
to
to
optimize
this
path.
Selection,
and
so
so
so
we'll
ask
her
sabine
to
introduce
his
proposal
and
another
one
is
about
a
new
transporter.
You
know
for
base
auto
protocol,
you
know
it
originally.
E
You
know
designed
for
the
p2p
application,
setting
application
and
build
on
top
of
the
http
one.
Now
we
have
hdb2
and
hd3,
and
also
we
have
quicker
so
how
we
can
migrate
from
http
1
to
http,
2
and
3.
So
this
is
something
we
can
discuss
and
the
fourth
proposal
is
market
domain
support.
You
know
we
already,
you
know
for
author.
We
can
do
the
peer
selection
path,
selection,
mostly
focus
on
single
domain,
and
but
how
do
you
extend
to
the
market
domain?
So
this
is
something
we
should
discuss.
E
It
may
require
the
server
to
server
communication.
The
last
one
is
of
operation
automation
and
we
already,
you
know,
published
author
deployment
jobs
and
as
auto
deployment
jobs
said.
Actually,
we
can,
you
know,
gather
different
data
source
and,
and
you
know,
fed
into
the
auto
server
and
do
the
aggregation
abstraction.
E
How
to
you
know
you
know
automate
these
auto
server
operation
management
automatically.
You
know
so
so
this
is
still
very
challenging.
So
so
we'll
ask
you,
know
lewis
and
jason
to
discuss
this,
and
so
we
set
this
order
and
so
that
that's
all
and
any
objection
to
this
order.
A
So
I
want
to
be
clear,
so
there
you
have
five
proposals
and
you're
not
proposing
that
all
five
of
them
go
in
the
charter.
We're
supposed
to
select
some
number
of.
E
Yeah
yeah
you're
right
actually
there's
a.
I
think
we
hope.
Actually
we
can.
You
know
pick
a
two
or
three
proposal
from
all
these
five
and
we
can
have
a
good
start
and
we
will
not.
You
know,
promise
we'll
accept
all
these
proposals
so
based
on
our
discussion,
our
consensus
will
decide
which,
which
two
or
three
can
can
be.
You
know
put
into
the
charter.
You
know.
A
Okay,
I
I
did
have
one
question.
You
said
something
in
cellular
about
3gpp
and
is
that
work
is
3gpp
asking
us
to
do
any
work
in
this
area,
or
is
that
does
it
just
yeah
going
to
coincide
with
go
ahead?
Yeah.
E
This
is
just
my
question,
so
I
will
ask
transcension
and
company
to
clarify
this,
because
I
also
raised
the
same
question
on
the
list.
I
haven't
get
a
response
and
I
my
feeling
my
expression
is:
they,
you
know
defined
the
nef
network,
exposure,
function,
architecture
and
requirements.
They
could
provide
the
you
know,
foundation
to
provide
a
good
data
source
to
the
other
server,
and
so
so
shenzhen
can
present
later
in
his
called
the
first
half
proposal,
and
we
can
discuss
this.
B
And
thank
you
yeah
good,
good
clarification,
martin,
but
also
from
my
perspective,
regarding
selecting
a
few
of
these.
We
could
also
think
about
prioritizing
them.
Regarding
on
the
feedback
we
are
having
during
the
discussions
today
right,
so
I
don't
think
we
need
to
date
trim
down
to
a
subset
of
this
five.
But
if
we
can
also
say
this,
one
got
better
feedback
or
seems
to
be
more
important.
That's
also
a
good
progress
or
a
good
outcome,
but
yeah
thanks
for
the
clarification
guys.
C
E
Okay,
so
yeah,
so
let's
move
to
the
presentation.
Actually
we
already
assigned
each
topic
to
the
champion
owner
and
so
transcension
are
you
there?
Do
you
want
to
present
your
topic.
J
Yeah,
can
you
make
the
test
more
lighter?
It's
almost
smaller,
okay.
J
I
can
read
it
yeah
the
main
target
of
the
song.
C
Starting
interrupt
so
I
I
think
team
probably
won't
share
his
slide
right.
I
believe
he
uploaded
his
lights.
J
J
E
Okay,
let
me
check
his
latest
portfolio,
yellow,
slides.
B
Yeah,
this
is
something
you
have
to
get
used
to
in
this
auto
working
group.
People
send
slides
really
close
to
the
meeting
all
the
time.
Yeah.
J
This
one,
this
is
also
the
higher
this
one
hang
on
next:
okay,
okay,.
J
Yeah
this
one,
the
the
proposed
that
we
want
to
attention
of
ato,
to
provide
secular
network
information
to
the
applications,
because
more
and
more
application
move
from
the
fixed
ip
network
to
the
mobile
network
and
a
lot
of
mobile
devices
are
used
by
the
people
and
attention
that
have
used
this
terra
saturday
level.
J
Information
can
help
demonstrate
that
we
can
leverage
the
information
from
the
student
network,
such
that
from
the
presentations
through
the
surrounding
status
between
the
ue
with
the
base
station
can
improve
the
application
performance
cycles
by
dynamically,
changing
the
bid
rates
to
according
to
the
status
of
the
radio
state
radio.
J
So
the
work
you
will
investigation
extension
or
the
basic
auto
information
service
to
equal
to
the
mobile
clients
or
mobile
application
or
mobilization
in
the
cloud
about
cellular
information
or
an
english
states
between
the
unit
and
the
station.
For
example,
the
information
also
only
a
throughput
delay
or
the
latency
of
the
data.
J
J
The
attention
should
be
created
to
address
the
dynamic
issue,
because
the
cellular
cellular
lingus
status,
like
the
throughput
delay
in
a
data,
will
be
changed
very
quickly
and
we
needed
to
provide
a
real-time
information
and
also
we
should
consideration
the
issue
whether
this
information
is
private
or
how
we
include
a
private
data.
J
So
the
the
work
of
were
firstly
defined
clearly
in
school,
empirically,
regarding
its
correlation
with
the
cpp
how
to
leverage
the
cv
network
employee
function,
that's
the
propose.
J
B
A
J
From
my
understanding,
normally
cellular,
because
it's
cellular
have
more
and
more
cpu
power
and
they
have
a
smart
enough
to
dynamity
to
detect
the
new
status,
maybe
it's
less
important
for
the
cellular
device,
but
for
the
cloud
gaming
server
because
he's
located
in
a
cloud,
he
has
no
information
for
the
radio
link
status,
so
he
needs
some
some,
the
dynamic
information
of
the
new
status
inside
the
case.
J
Normally
we,
I
said
that,
for
the
face
and
network
the
benefits
and
the
detail
on
delay
is
some
static,
not
not
changed
very
quickly,
but
a
cellular
network
is
changing
very
quickly
inside
the
case.
If
we
can
provide
this
information
to
the
cloud
gaming
server
or
a
srvr
server,
they
can
deliberately
change
the
bid
rate
to
the
user,
in
some
cases
no
lagging
or
and
nagging
and
interruption
of
the
service
and
have
a
good
ue
quality,
because
the
ue,
the
user
are
well
alloyed
for
the
the
lagging.
A
Okay,
thanks
that
clarifies
that
so
I'm
glad
you
put
in
language
about
the
dynamicity
issues,
because
I
think
that
is
really
my
chief
concern
about
this
work.
You
guys
understand
ulta
better
than
me,
but
you
know
these
things
are
not
usually
scaled
to
to
be
delivering
real-time.
You
know
information
at
this.
This
granularity
like.
Are
we
confident
this?
This
is
actually
workable,
or
are
we
just
going
to
end
up
finding
out
that
the
that
the
frequency
issues
are
are
are
insurmountable.
B
B
B
So
that's
okay,
but
yeah
in
this
case,
things
may
so
jitter
in
a
mobile
network
might
change
quite
often,
but
so
I'm
fine,
because
the
the
the
word
dynamicity
is
explicitly
in
here,
but
I
guess
the
question
for
martin
is
a
valid
one.
Are
there
some
metrics
that
don't
change
that
often
that
can
be
conveyed
from
base
station
to
mobile
application
right?
That's
the
question.
I
guess.
A
Well,
so
I
mean
so
so
the
word
is
in
there
and
that's
great
and
like
I'd,
be,
I
would
you
know,
stomp
my
foot
if
it
weren't
in
there,
but
at
the
same
time
I
I
don't
want
to
I
I
would
rather
not
bring
in
a
research
project
if
we're
not
sure
this
is
viable,
and
I
don't
know
if
it's
viable,
I'm
just
that
I'm
asking
like
have
we
have
we
have
someone
worked
through
the
problem
of
like
does.
C
Right
so
martial,
I
I
don't
know
if
I
can
jump
and
answer
a
little
bit,
because
I
was
involved
a
little
bit
and
of
course
probably
the
other
good
person
to
get
involved
will
be
gangly
from
panamobo,
and
there
was
this
like
a
sitcom
ni
paper,
I
called
them
away
paper
actually
of
a
reported
results,
and
so
the
evaluation
of
the
I
think
was
a
large-scale
quote-unquote
large-scale
deployment
of
this
information
in
the
china
mobile
network.
C
I
think
in
georgian
province
is
that
right
and
I
believe
they
can
approach
to
about
like
if
the
information
changes
about
the
collecting
information
about
one
minute
is
that
right
to
five
minutes
per
update
and
it
can,
it
could
get
good
performance
and
also
the
overhead
here
also
is
overhead.
Actually,
information
is
not
already
sent
through
the
cellular
network.
Actually,
the
information
is
collected
at
a
base
station,
but
using
backhop
to
directly
send
it
to
the
servers.
So
therefore,
there's
no
you're
not
consuming
the
the
information
in
the
cellular
channel,
basically,
information
collected.
C
Of
course,
information
collected
will
be
about
the
cellular
network,
but
the
channel
the
giving
information
pushed
to
the
server
would
not
really
go
through
the
cellular.
Echo
would
go
through
the
back
call
and
therefore
you
don't
have
the
bandwidth
constant
over
over
there.
So
that's
why
essentially
called
split
the
channel
split
architecture
and
there
was
this
demo,
I
believe,
very
good
performance,
but
I
guess
governor
was
a
guy
really
physically
running
in
the
the
the
demo.
So
probably
I
should
yield
to
him.
A
E
Okay,
that's.
I
also
want
to
make
some
clarification.
I
I
did
do
some
homework,
for
you
know,
3gbp
worker
called
nef.
Actually
they
actually
tried
to
use.
You
know
publishable
model
they
can
get
gather.
You
know,
celebrate
information
or
buy
network
information
or
again
very
related
information.
You
know
they
can
get
the
real-time
information
more
efficiently.
So
so
we
can
separate.
You
know
this
problem.
You
do
two
issues.
How
one
is
how
we
can
get
the
information
put
put
into
the
auto
server?
E
The
second
is
how
frequently
we
can
expose
this
information
to
the
client.
So
so
I
I
think
in
3gbp
they
focus
on
the
first
part
of
the
issue.
They
they
can.
You
know
using
the
public
mechanism,
they
can
connect
real-time
information
or
some
real-time
static
information
and
fade
it
into
the
auto
server.
So
if
we
can,
you
know
have
this,
you
know
assumption
we
can.
You
know,
provide
this
information
to
the
client.
E
You
know
focus
on
the
interface
between
all
client
and
auto
server.
They
may
complementary
to
what
3gb
is
doing
and
so
probably
so
shenzhen.
You
should
answer
martin's
question
how
this
works
relate
to
the
3gb
is
any.
You
know
there
is
an
exchange
or
is
any
feedback
from
3gbp.
J
Yeah,
I
think
that
we
should,
as
I
thinking
we
should
as
we
should
start
navigate
a
nef
without
any
reason
to
the
cpp,
because
for
my
understanding,
nef
have
defined
a
well
open
api
to
to
net
the
ef
client
to
subscribe
the
service.
They
can
use
the
subscribe
notification.
This
requested
response
for
the
subscriber
notification.
You
can
defend
one
time
notification
and
a
long
time
and
how
and
as
a
task
attack
spam
you
can
notify
the
status.
J
If
the
anything
is
changed,
then
you
can
look
up
you.
You
can
notify
back,
not
need
a
poor,
again
game,
so
they
can
notify.
Oh,
what
can
you
can
define
how
long
how
their
frequency
you
are
notified?
For
example,
you
can
notify
10
milliseconds.
Even
there's
no
changes,
you
can
overlap
back.
J
Does
not
provide
very
frequency
change
the
parameters
because
they
also
know
it's
a
parameter
or
valuable
consumes
a
lot
of
resource
because
they
tended
to
frequently
very
consume
a
lot
of
resource.
Also,
they
cannot
provide
very
sensitive
information,
because
you
cannot
ask
please
open.
J
A
Okay
yeah,
so
my
other
question
was
so
I
mean
like
when
you,
if,
if
you're
talking
about
you,
know
feed
gaming
fee
or
any
sort
of
feedback
for
an
active
base,
adaptive
bit
rate,
like
I
mean
you
know
the
obvious
solutions
to
this-
that
don't
involve
lots
of
intermediaries,
like
you
know,
looking
at
connection
control
signals
looking
at
direct
user
feedback
from
from
the
from
the
from
the
gaming
client
et
cetera,
like
could
someone
sort
of
explain
intuitively
why
this
is
like
why
this
would
have
better
dynamics
than
something
more.
D
I
just
I
want
to
say
something
about
the
time
granularity
right,
so
I
I
think
it's
really
depends
on
the
the
demands
from
from
the
application
right
if
they,
for
example,
if
for
the
cloud
I
mean
server,
even
the
cloud
game
server
want
to
change
it.
To
read
for
for
seconds
right
might
be.
Second
level.
Information
might
be
very,
very
useful,
I
think,
from
our
experiment
in
our
live
network.
We
think.
G
D
So
I
think,
as
a
100
minutes
level
worth
second
level
kind
of
fluctuation,
I
think
it's
it's
not
quite
to
to
refine
compare
with
a
is
a
they
they.
This
singular
interaction
between
the
three
digit
network.
I
think
it
is
facebook
to
us
for
the
auto
server
to
obtain
those
kind
of
information
and
distribute
those
kind
of
images
they
auto
client.
D
So
I
think
it's
it's
donable
right
for
us.
A
The
question
is,
why
would
you
not
just
use
something
like
direct
client
feedback,
or
you
know
existing
regression
signals
in
the
network
to
to
do
your
apr
tuning.
C
F
Yeah
I
I
can-
I
can
answer
that,
so
I
I
think,
maybe
you
would
you
know
we
when
we
think
about
3gpp.
We
also
probably
want
to
keep
in
mind
a
gsma
effort
on
operator
platform
called
opg,
and
so
there
in
that
work.
We've
been
looking
at
two
things
that
are
related
really,
and
I
think
can
help
answer
the
question
here,
and
that
is
regardless
of
how
the
information
is
acquired.
F
We
don't
need
in-band
specific
user
feedback,
but
the
gaming
servers
tend
to
require,
for
example,
kind
of
a
general
information
on
the
average
cqi,
really
the
average
shirt,
but
the
average
latency
that
a
device
happens
to
be
connected
to,
because
you
know
the
connection
has
been
established
yet
between
the
client,
the
server
in
this
case,
and
so
we're
really
looking
for
kind
of
that
coarse
grain
type
of
information
that
we
would
expect
from
an
alternate
server
and
then
based
upon
that
before
the
game
started
right.
F
It's
really
that
matchmaking
design,
where
it
may
look
at
multiple
mobile
clients
and
then
make
kind
of
an
intelligent
decision
about
the
best
location
where
there's
multiple
options
or,
in
many
cases,
multiple
bad
options
on
where
to
actually
host
the
game
in
terms
of
the
network
placement.
So
so
that's
one.
So
we
have
to
remember
that
you
know
this
data
really
isn't
known
between
the
client
and
server
right
at
the
beginning
and
and
so
we're
really
needing
kind
of
just
some
generalization
based
upon
the
location.
F
The
other
thing
I
I
want
to
keep
in
mind
is
in
5g,
there's
a
couple
of
huge
changes,
so
one
is
that
fixed
wireless
is
now
you
know
a
primary
product.
So
when
we
say
mobile
nowadays
that
we
probably
should
really
not
use
that
term,
it's
it's
kind
of
a
kind
of
an
old
view.
The
other
thing
is
in
release
16
and
it's
some
new
documents.
F
Wireline
convergence
is
now
a
part
of
the
the
integration
of
the
network,
and
so
you
can
directly
add
what
they
call
alternative
or
other
access
networks,
such
as
g-pawn
or
or
docsis
networks
directly
into
the
5g
cores.
F
So
you
know
we'll
have
to
all
kind
of
start
shifting
our
thinking
in
terms
of
you
know
when
we
say
a
mobile
network
when
we
say
a
mobile
core,
what
that
is
from
from
the
point
of
view
of
the
application
servers,
it
makes
no
difference
right
as
to
what
it's
looking.
So
I
I
think
the
granularity
is
not
as
big
of
a
concern.
In
fact,
I
think
100
milliseconds.
F
This
are
things
maybe
overkill,
at
least
from
what
we've
seen
in
the
ops,
the
gsma
obg
group
in
terms
of
their
needs,
and-
and
you
know
I
I
think
really-
you
know
what
they're
looking
at
the
beginning
really
isn't
any
of
that
in-band
work.
So
I
hope
that
that
helps
maybe
answer
some
of
the
questions.
C
All
right
so,
like
I
really
want
to
add
some
one
thing
very
quickly
and
martin
asked
your
question:
why
not
directly
client
feedback?
I
think
just
one
thing,
I
think
that's
what
I
heard
from
you
from
tencent
people
and
the
kind
of
mobile
we
oftentimes.
The
business
relationship
is
between
the
network
and
the
server.
So
therefore,
information
basically
really
is
flowing
between,
for
example,
attention
it's
paying
for
the
services
and
they
want
sales
channel
standards
themselves.
They
decide
to
pay
or
not.
C
You
really
don't
have
to
go
through
a
client
because
they
don't
even
know
that's
basically
somehow
there's
negotiation
between
the
network
and
the
application
provider
directly.
So,
for
example,
that's
one
more,
I
don't
know
that's
a
slightly
just
they're
compiled
a
reason
to
talk
about.
Why
not
directed
client
feedback.
I
think
that's
one
question
you're
asking
as
well
right.
B
Okay,
guys,
I
think
we
should
move
on
to
the
next
yeah
and
we
are
already
way
into
the
session
and,
as
I
mentioned
before,
20
minutes
before
the
end
I
will
I
would
like
to
start
the
wrap
up,
so
I
think
we
should
move
on
to
the
next
yeah
charter.
Okay,.
C
So
sorry,
can
I
ask
a
quick
question:
how
much
notes
should
I
take
about
all
these
disco
they're,
not
really
working
group
action
items
and
should
I
catch
try
to
catch
all
these
questions?
Yeah?
That's
that's.
B
B
B
H
H
H
So
basically,
the
idea
would
be
to
to
to
to
go
along
the
the
the
path
costs
and
also
with
the
qualitative
attributes
and
add
time,
extend
properties
to
time,
dependent
properties,
more
or
less
mimicking
alto
cost
calendar.
So
we
would
have
property
calendars.
H
And
then
the
working
group,
of
course,
will
specify
these
are
like
gener,
general,
so-called
general
extensions,
because
we
feel
that
this
is
useful
to
not
for
the
sake
of
having
an
extensions,
but
this
will
be
useful
for
anybody
in
the
working
group
that
can
be
that
will
have
use
cases
involving
such
features,
and
so
this
is
why
we
will
specify
such
extensions
in
coordination
with
both
autoworking
groups.
H
Group
items,
sorry
and
also
other
ietf
working
groups
that
have
a
focus
on
the
related
use
cases,
and
the
scope
of
this
extension
is
not
limited
to
those
identified
by
the
working
group
items
and
then
working
groups,
but
also
limited
by
the
criteria
set
out
below
which
are
the
ones
the
the
usual
questions
that
are
to
be
addressed
within
the
working
group.
In
terms
of
is
it
doable?
H
So
in
the
next
in
the
following,
slides,
if
you
like,
the
following
slides
proposal,
use
cases
that
we
are
motivating
these
proposals
and
but
over,
which
I
can
go
up
on
your
questions.
E
B
H
Yes,
indeed,
I
this
is
done
on
purpose,
because
the
the
idea
of
of
resu
reusing
as
much
as
possible
the
text
of
the
current
charter
was
to
to
say
that
there
are
further
optimizations
that
are
needed
on
these
protocols.
So
it's
it's
a
con.
It's
a
continuation
of
the
previous
optimizations
that
were
that
brought
us,
for
example,
the
calendar
so
decision
on
when
to
connect
that
extended.
The
end
points
to
properties
that
were
allowing
to
define
entity.
H
Sorry
that
we're
extending
endpoints
to
entities
in
particular
ines
that
allow
too
detailed
particular
parts
of
a
path,
but
not
only
and
so
and
and
that
also
brought
to
path
vector.
So
so
there
is,
we.
H
There
are
use
cases
that
where
it
was
clear
that
some
additional
some
additional
some
additional
details
on
the
cost
value
could
be
necessary
or
useful,
and
by
the
way
I
would
like
to
comment
on
the
feedback
received
by
luis
and
so
just
to
say
that
this
proposal
does
not
mean
to
be
restrictive,
and
I
agree
it
should
be
a
placeholder
for
other
protocol
optimizations
that
we
might
find
necessary
in
particular
upon
making
the
examples
in
this
slide
set.
I
run
again
into
that
issue
of
the
cross
product
product.
H
When
you
do
so,
it
seems
simple
technically,
but
it
is
necessary
because,
for
example,
if
you
request
a
require
sorry
a
cost
map,
so
you
have
first
a
set
of
sources
and
then
a
set
of
destination.
I
mean
that
has
been
discussed
and
brought
by
by
richard
years
ago,
and
so
we
and
then
we
we
only
ca
the
current
protocol
with
the
current
protocol.
H
So
we
just
get
the
the
data
on
the
cross
products
of
source
and
destination
and
given
the
recent
publications
by
ingmar,
for
example,
on
the
deployment,
the
network
map
and-
and
we
saw
it
when
we
when,
when
we
measure
the
what
is
circulating
on
the
network
and
when
we
measure
the
volume
of
a
network,
it
gets
huge,
very,
very,
very
quickly.
So
whatever
we
can
do
to
to
diminish
the
the
volume
of
exchange
data
is
a
good
thing.
E
Yeah,
I
I
think
a
key
use
case
for,
for
this
proposed
academy
is
for
path
vector.
You
know
you
we
actually,
you
know
you
may
have
a
multiple
pass.
You
you
want
to
select
the
best
pass
and
the
limitation
for
existing
past
vectors
it
may
it
may
doesn't
describe
some
past,
concentrate
you
know,
as
you
mentioned
about
you,
know,
access
type
you
know
from
which
kind
of
x
type
you
can.
You
can
select
the
the
connection
paths
you
know,
or
you
may
select
a
end-to-end
path
with
some
network
performance
requirements.
E
So
this
proposal
seems
to
me
actually
is
the
you
know
additional
pieces
that
we
can
add
for
the
past
vector.
H
Yeah,
thanks
and
and
also
if
we
go
back
if
we
think
about
the
first
proposal
on
sale.
Actually
that
was
a
a
major
use
case
that
really
motivated
this
proposal,
because
the
the
the
use
case
was
a
a
cellular
network
I
presented.
I
think
this
use
case
a
couple
of
years
ago
when
it
when
we,
when
it
really
appeared
when
we
ran
into
it.
H
So
there
was
the
use
case,
was
an
application
on
a
on
a
mobile
phone
that
was
using
an
alto
client
and
of
at
that
time.
So
we
we
felt
that.
H
We
could,
for
example,
for
each
video
chunk.
H
We
could
get
auto
information
for
this
video
chunk
every
two
seconds,
but
I
mean
we
did
not
need
that
frequency,
but
we
could
achieve
up
to
this
frequency,
but
anyway,
the
alto
information
is
by
essence
meant
to
be
like
maybe
a
30-second
20
seconds,
just
for
the
sake
of
avoiding
network
instability,
but
still
it's
quite
long
compared
to
the
pace
of
network
information
that
a
cellular
network
can
generate
and
a
ceiling
network
can
generate
real-time
indicator,
and
for
this
case
it
was
a
an
indicator
saying
from
3gbp
release
13
that
was
telling
you
whether
unattended
data
is
allowed
or
not,
and
it
was
like
a
boolean.
H
So
well,
if
it's
not
a
load,
it
means
the
connection
is
poor.
It's
a
if
it's
a
load.
It's
me
it
means
the
connection
is,
is
fine,
so,
of
course
the
alto
server
cannot
extend
this
dynamicity,
this
high,
real-time
dynamicity,
but
still
if
we
have
the
the
path
cost
given
by
alto,
which
is
like
smooth
like
aggregated
over
time
and
space
and
this
path
cost
can
be
affected
by
this
indicator
that
it's
provided
real
time.
H
So
this
can
provide
you
in
a
sense,
a
detail
on
the
path
cost
because,
okay,
you
have
the
the
long
term
alto
information
like
30
seconds,
which
is
long
term
for
cellular
network,
and
then
you
can
always
if
the
ultra
server
says:
okay,
that's
the
cost,
but
if
whatever,
depending
on
the
value
of
this
indicator,
well,
this
path
cost,
which
would
be
normally
four
well,
if
it's
not
a
load.
Let's
say
it
will
be
three
so
far.
H
If
sorry,
the
cost
will
be
higher
five
and
if
it's
a
load,
okay,
it
can
be
say
three.
So
this
is
actually
a
way
to
combine
the
long-term
information
like
20
or
30
seconds
by
the
auto
server,
with
some
real
time
indicator.
That
is,
that
you
get
outside
the
alto
channel.
We
via
a
normal
network,
channel.
H
E
C
Okay,
so
I
I
do
want
to
make
one
one
comment
about
this
particular
extension.
So
here's
one
use
case,
which
I'm
always
thinking-
and
I
think
that's
also
my
comment
we
related
with.
I
believe,
martin
asked
two
questions
right.
Two
questions
all
right.
Two
following
one
is
leprom
and
of
course,
typically:
what's
the
problem
and
what's
the
usual
kind
of
result,
typically
use
case
and
number
two
is
feasibility
right,
so
we
will
have
solution,
architecture
and
coming
a
richer
extensible
set
of
attributes
and
so
on.
C
So
I
think
my
comment
would
be
related
with
both.
So
what's
the
issue
here,
I
think
the
problem.
The
way
I
look
at
what
subbing
is
proposing
here
is
actually
quite
similar
to
an
architecture,
because
we
want
to
talk
about
architecture.
Visibility
is
architecture
of
like
say,
for
example,
premier
prometheus,
right
p-r-o-m-e-t-h-u-s,
which
is
very
very
popular
data
management
premium.
I
think
a
lot
of
networks,
a
lot
of
enterprise
network
using
this
framework
to
collect
data
process,
data,
query
data
and
so
on,
and
so
basically
for
that
data
model.
C
What
you
do
is
you
have
a
metric
and
then
you
have
an
unlimited
key
value,
store
it's
important
called
attributes
and
then,
of
course
you
you
apply
the
data,
I
think
that's
a
very
nice
architecture
can
really
address
a
lot
of
use
cases
and,
for
example,
one
use
case
or
issue,
maybe
even
kill
it
from
one
right
out
over.
Basically,
for
example,
very
simple
one:
is
you
ask?
What's
the
perform?
C
Okay,
you
ask
what's
the
latency
for
this
particular
flow
and
eventually,
if
anyone
follows
let's
say,
ippm
for
all
the
product
eventually
gets
very
ugly
and
and
you
you,
you
become
endless
of
specification
of
standards
and
so
therefore
that's
very,
very
difficult,
because
the
latency
could
be
which
one
which
follow
http,
http,
2
and
repeat,
three
and
a
quick
and
and
what
kind
of
traffic
and
all
these
can
attributes
will
go
in
eventually
to
really
understand
what
performance
metric
potential
can
be
server,
become
very
messy
system
and
promise
to
solve
the
problem.
C
That
basically
said:
okay,
here's
a
metric
here.
Let
me
indicate
a
central
and
open
flat,
flexible,
a
key
value
store
and
you
can
query
and
so
on.
So,
therefore,
if
we
do
this
richer
and
extensible
set
of
attributes-
and
we
immediately
would
be
able
to
introduce
promiscuous
capabilities
into
auto
framework
and
therefore
we
have
a
very
powerful,
very
reachable
extensible
framework
to
solve
a
lot
of
issues
and
also
framework
is
already
over
there,
which
you
can
just
borrow.
I
think
that's
my
comment.
Why
I
buy
this?
G
E
Maybe
saving
you
can
take
a
look
at
that
and
see
whether
you
can
get
a
useful
input
from
there.
So,
okay,
so
let's
move
to
the
next
and
new
transport
and
enemy
stops.
I
E
I
Yep,
I
think
that's
the
one.
So
I
think
for
this
which
item
the
motivation
is
relatively
simple
and
I
think
it
based
on
the
fact
that
we
might
be
providing
similar
information
to
clients.
So
we
might
have
more
requirements
on
the
performance
of
the
auto
service
itself,
so
and
and
as
hp
and
also
so.
I
Http
2
is
already
an
rfc
and
http
3
is
being
developed
as
an
rc,
but
but
it's
still
in
the
progress
and
I
think
they
provide
some
features
that
can
improve
the
performance
of,
but
first
they
have
some
benefits
over
hp
1.x,
which
is
used
by
auto
and
second,
they
can
help
eliminate
the
some
workarounds
that
we
use
in
the
current
extension,
such
as
pass
vector
and
not
equipment
updates
and
then
might
help
simplify
the
protocol
itself.
I
And
then
the
purpose
of
this
item
is
to
provide
an
alternative
way
to
transport,
the
auto
information
on
top
of
http
2
and
above
and
yeah.
Basically
that
that's
it
and
we
see
we
need
to
specify
how
the
servers
and
clients
should
exchange
information
and
especially
for
past
factory
and
equipment
updates.
And
then
we
might
also
introduce
an
extension
to
the
ird
to
specif,
allow
clients
and
servers
to
kind
of
negotiate
how
they
want
to
transport.
The
auto
information.
E
So
thanks
guys,
so
I
I
have
a
quick
question.
You
know
for
passive
vector,
you
know
it.
You
know
actually
use
multi-part
to
return
multiple
responses,
so
my
multi-part
message
also
is
obviously
actually
widely
used
and
it
can
be
used.
E
You
know
you
know
build
on
top
of
this
http
one,
but
can
we
use
these?
You
know
multipart
messages
in
hdb
too.
Also
it
applied
to
to
you
know
for
cdni.
You
know
draft,
actually,
the
actually
the
reference
sse
you
know.
Sse
is
also
widely
used.
Why
so,
since
your
proposal,
you
want
to
obsolete
sse
or
multiply
the
message,
what
do
you
think
about
that.
I
I
First,
yes,
if
possible,
so
first
answer
the
question:
we
can
still
use
the
current,
for
example,
multiplier
and
sse
in
even
on
top
of
http
2,
but
that
would
basically
not
fully
leverage
the
benefits
of
http
2,
especially
for
the
ssc
part,
so
the
sse
kind
of
multiplexes,
multiple
resources
into
a
single
data
stream,
and
so
from
the
service
point
of
view.
From
the
from
the
communication
point
of
view,
it's
a
single
resource
and
for
the
single.
I
If
you
deliver
a
single
resource
over
h32,
it
will
basically
will
be
considered
considered
as
a
single
stream,
and
if
you,
if
you
want,
if
you
split
the
resources
and
treat
every
increment
up
for
through
every
increment
update
for
a
single
resource
as
a
single
stream,
then
there
can
be
multiple
streams
delivering
us
data,
and
I
think
that
could
be
more.
I
E
Okay,
so
so
follow
up
a
question
you
know
for
encoding,
you
know
so
http
they're
using
binary
encoding,
for
you
know,
header
format
and
data
format
for
me.
Actually,
so
how?
How
do
you
see?
You
know
this
is
actually
you
know
for
ultra
baseball
protocol.
You
know
they
use
a
json.
I
Actually,
actually,
I
think
we're
now
going
to
introduce
a
new
binary
format
of
auto
messages.
We
just
basically
like
put
the
clear
json
text
or
the
encrypted
json
text
as
a
binary,
and
then
the
server
or
client
could,
for
example,
read
the
data
directly
from
the
binary
object.
Actually
I,
if
you
go
to
the
next
slide
and
yep
it's
like
so
on
the
left
hand,
side
is
actually
the
json
json.
I
Basically,
it's
a
message
for
the
multiple
message
for
m
pass
vector
example
and,
on
the
right
hand,
side.
Actually,
it's
not
the
same
co.
I
just
basically,
I
used
a
python
library
to
put
these
two
messages
into
a
single
http
request
and
here's
what
the
data
transfer
from
the
server
to
the
client
will
look
like.
I
First,
there
will
exchange
some
settings
frames
and
then
the
server
the
server
will
send
a
push
promise
frame
to
the
client,
notifying
that
one
of
the
resources
delivered
is
going
to
be
delivered
to
the
client
through
stream
id
2,
and
then
it
returns
the
pass
the
course
map
part,
of
course,
the
data.
Here
I,
I
mainly
extract
the
json
information
from
the
binary
format,
but
actually,
when
I
was
developing
the
library,
the
the
banner,
the
the
binary
basically
contains
the
the
same
json
text.
So
I
think
this
wouldn't
be
a
problem.
I
B
E
B
E
E
E
C
Need
to
wrap
yeah,
so
I
already
I
have
four
minutes
left.
Okay,
so,
basically,
and
basically
I'm
trying
to
say
is
a
multi-domain
actually
was
designed
to
be
very
beginning
and
also
should
support
multi-domain,
that's
specified
in
rc,
79,
7971
and
chin.
I
think
quarterly
center,
which
is
great.
But
what
I
want
to
argue
is
that
the
current
auto
services
actually
don't
really
handle
multi-domain
very
well,
and
there
are
quite
a
lot
of
document
talk
about
the
issues
next
slides.
C
Please
next
place
yeah.
So
this
is
a
slice
of
actually
illustrator
issues.
Multi-Domain,
and
this
is
a
slide
icon
made
by
sebastian.
I
don't
believe
he
joined
today,
that's
early,
and
so
all
this
kind
of
conflict,
timing
and
so
on.
C
So
basically
here
is
the
current
auto
would
work
we
have
here
x
here
on
left
hand,
side
we
have
bunch
of
resources
at
the
pier
1a,
1b,
2a2b,
3a3b
and
current
essential,
auto
particle
is
only
fully
specified
when
the
prx
can
query
the
auto
server
or
first
hop,
essentially
one
hops
and
then
wouldn't
be
able
to
get
information
about.
C
I
speed,
for
example,
here
suppose,
look
at
a
simple
case
where
the
the
auto
server
isp1
would
be
neutral
about
pier
2a
and
pier
2b,
and
there's
no
way
to
really
get
losing
information.
It's
not
specified
at
all.
So
therefore,
I
could
perform
here.
Maybe
sp2
would
prefer
2b
or
2a,
but
there's
no
way
to
know
about
that.
So
therefore,
basically,
the
protocol
is
relatively
neutral
and
specific
for
only
a
single
domain.
Second
exercise:
please.
C
Okay-
and
so
I
got
this
transformed
to
be
a
major
problem
because
to
answer
the
relevant
question,
would
people
be
interested
yes,
of
course,
and
left
hand
side
here
is
a
paper
the
slides
by
steering
the
hyper
giants
traffic
at
scale?
I
think
I
believe
one
called
next
best
paper.
Also
the
irtf
apply
networking
price
and
one
of
the
major
issues
we
need
to
figure
out
and
we
have
a
lot
of
discussions
with
engema
and
sparcia
and
so
on.
C
Is
you
want
solid
issue
of
federated
fellow
director
which
is
actually
based
on
auto
and
so
on,
and
essentially
is
multi-speed
collaboration?
How
do
you
really
have
multiple
outdoor
servers?
The
right-hand
side
is
lc
opn,
which
call
network.
You
can
see
that
it
has
a
lot
of
networks
and
therefore
you
also
have
a
lot
of
auto
servers.
How
to
handle
the
issue.
The
auto
really
does
not
really
have
given
a
full
specification.
C
Oh
okay,
next
slides,
please,
okay,
and
what
I
want
to
say
would
be
their
discussion
during
the
meetings
was:
would
release
really
become
a
research
problem
or
not?
Actually,
it's
not
and
just
like,
like
maybe
like
three
weeks
ago
and
during
a
weekly
meeting,
we
did
introduce
a
concept
called
automotive
domain
abstractions
we
did
introduce
quite
a
few,
very
interesting,
abstractions
and
so
on.
So
therefore,
feasibility
and
architecture
seems
to
be
over
there.
C
Next
slides,
please
just
skip
it,
give
me
time
and
basically,
oh
and
I
can't
go
back
to
one
slice,
so
basically,
what
we
proposed
would
be
for
this
particular
design.
Only
we're
not
saying
this
must
be
the
one
and
the
solved
issue
multi-domain,
and
we
need
two
extensions
to
auto.
One
is
segment.
Discovery
basically
means,
let's
really
discover
the
whole
path
from
the
given
source
to
the
destination.
C
If
you
you
spam,
multiple
domains,
and
we
feel
that
this
is
a
very
interesting
good
extension
and
this
extension
even
would
be
useful
by
itself
essentially
somehow
open
the
bgp
interface
would
be
very
useful
by
itself
and
then,
of
course,
after
you
found
the
segments
with
auto
servers
involved
using
some
kind
of
chaining,
and
then
you
also
want
to
be
able
to
discover,
extend
auto
your
ecs
and
custom
app
services.
C
Therefore,
you
can
really
discover
all
the
whole
whole
vector
and
then,
of
course,
you
can
then
also
would
extend
the
auto
with
multiple
operating
modes.
For
example,
iterative
recursive
hybrid
for
people
who
know
dns
is
highly
highly
based
on
ds,
recursive
required
multiple
servers
and
integrated
multiple
servers,
and
now
can
you
go
to
the
next
next
slide.
Basically,
I
can
go
to
paragraph
yeah,
you
can
skip,
you
can
skip
milestones
and
we
probably
need
to
see-
and
here
is
essential
paragraph.
C
So
basically,
what
we
want
to
do
is
extension
of
auto
services
to
super
multi-domain
settings
and
clearly
the
current
auto
framework
has
made
it
clear
how
to
provide
information
for
single
server
for
single
network,
but
devices
might
traverse
multiple
networks
and
not
in
the
same
domain.
So
therefore,
we
want
to
essentially
extend
two
things.
C
One
is
specify
multi-auto
server
protocol
flow
and
user
guidelines.
When
auto
service
involves
network
passes
spanning
multiple
domains,
a
number
two
extend
or
introduce
auto
services
allowing
east
west
interfaces
for
multiple,
auto
servers
to
integrate
and
collaborate.
That's
a
basic
proposal,
I
think
important
issue,
and
so
on
and
specification,
of
course,
should
reuse
it.
Whenever
possible
and
and
the
of
course,
you
would
always
read
the
issue
of
incremental
deployment,
because
multiple
domains
would
have
women
that
will
have
issue
with
increment
deployment,
and
then
we
must
address
the
issue
of
dynamicity.
C
Of
course,
we
also
must
address
israel's
security
and
privacy,
because
right
now
always
a
major
issue
even
from
very
beginning.
I
believe
the
first
issue
always
reads
about
atp
and
comcast
email
very
early
on
was
okay,
and
what,
if
there
are
all
the
other
people
want,
wanted
to
claim
to
be
my
to
be
the
auto
server
for
my
domain?
So
therefore
we
want
to
address
this
issue.
E
Flow
so
richard
I.
I
think
this
is
a
I
I
think
it's
a
very
important
word,
but
also
very
challenging,
because
you
discuss
you
know.
Cross
domain,
you
know,
for
each
domain
may
have
one
auto
server.
How
may
for
each
domain
they
may
own
by
different
operator.
How
you,
you
know,
ask
a
different
operator
to
work
together
to
share
the
information.
C
Correct
correct
so
as
a
first
step,
and
I
think
our
design
can
you
go
back
to
the
like
a
dummy
design
just
very
quickly
excellent
question:
a
go
back,
slides,
more
more,
more
okay,
yeah!
I'm
sorry!
I
think
you
have
to
go
forward
yeah
next
next,
next
yeah!
C
One
yeah:
no,
I'm
gonna
talk
about
the
simple
extension
next
one
I
think,
there's
oh
there's,
okay,
next
one
next
one
yeah,
so
absolutely
very,
very
good
comment,
and
I
think
number
one
is
we
by
the
way
clear
that
this
is
only
a
a
draft
proposal,
and
so
we
clearly,
this
karma
audio
audit
service
will
be
very
difficult.
So
therefore,
we
thought
one
way
to
solve.
This
issue
would
be
using
the
trust
of
all
like
a
bgp
or
peer
relationship
already
established
among
the
domains.
C
Therefore,
if
you
look
at
this
like
the
dummy
design-
and
they
can
give
us
some
exponential-
they
can
even
tell
us
if
there's
their
black
hole
and
how
do
we
even
have
an
extra
solar
for
really
discover
audio
like
a
server.
So
I
think
that's
one
way
to
address
it
so,
therefore,
and
how
they
can
work
together
and
to
really
answer
that
question
is
we
want
to
have
multiple
operation
modes?
C
Iterative,
for
example,
always
collect
all
the
information
and
therefore
the
client
will
be
essentially
making
decision
itself.
We
don't
solve
the
issue.
We
want
to
specify
the
mechanism,
not
the
power
say
so,
therefore
collect.
We
only
provide
a
mechanism
to
collect
all
the
information,
but
we
don't
make
decisions
and
then,
of
course,
we
can
run
like
a
recursive
or
hybrid.
Therefore,
we
can
introduce
essentially
this
like
like
a
proxy
and
and
frameworks
through
automatic
build.
Therefore
we
can
specific.
C
So
I
think
that's
what
we
want
to
solve
number
one
is
give
the
complexity
to
other
clients,
but
introduce
the
mechanism
and
second
one
at
the
same
time,
make
it
possible
that
mechanism
can
also
provide
a
solution,
at
least
in
a
initial
setting.
I
don't
know
if
I
answer
a
question
or
not,
but
that's
essentially,
our
current
design.
E
J
G
G
So
it's
true
that
there
are
already
some
some
best
practice
document,
and
so,
but
from
that
point
till
now
there
were
newer,
auto
protocol
stations
that
the
ones
that
we
have
discussed
today
and
some
others,
because
calendar
unified
properties
and
so
on
so
far
and
also
there
are
newer
use
cases
that
we
we
are
foreseeing
like
the
closed
loop
automation
that
could
come
from
the
ideas
from
etsy
csm
or
the
some
other
drafts
that
are
available.
G
G
What
could
be
those
considerations
to
to
take
into
account
well
decisions
on
the
information
resource
exposed
and
notification
of
changes,
either
in
practice
or
reactive
mode,
and
I
mean
the
interaction
with
the
with
the
application
and
so
to
to
deploy
to
develop
new
capabilities
for
an
alto
for
working
in
those
different
modes,
aggregation
processing
of
the
corrected
information.
So
to
what
extent
we
need
to
to
aggregate
the
information
collected
and
exposed
in
a
certain
manner,
et
cetera,
et
cetera.
So
this
will
be
one
of
the
of
the
the
sides
of
the
proposal.
G
D
E
Yeah,
so
thanks
luis
and
good
representation,
actually
so
my
very
curious,
so
how
fast
we
can
get
data
from
southbound
in
the
interface.
Actually,
I
think
this
may
relate
to
this
proposal
and
also
for
for
the
second.
Actually,
I
think
you
want
to
propose
a
new
protocol
for
some
extension.
I
think.
Maybe
you
know,
maybe
you
you
only
focus
on
like
a
service
edge,
computing
service
discovery,
or
you
know
what
kind
of
product
extension
you
expect
there
for
this.
This
will.
G
This
will
be,
this
will
be
one
of
them.
At
least
this
is
on
ongoing
work.
There
are
some
some
droughts
on
that
and
regarding
the
refreshment
of
the
information.
Clearly,
we
have
dependencies
on
the
on
the
protocols
and
and
the
mechanisms
that
we
do
for
c,
for
that,
for
instance,
going
directly
to
the
point
of
of
the
compute
case
computed
case.
If
we
leverage
on
bgpa
mechanisms
for
that
for
sure
we
have
the
dependency
on
the
timing
of
these
of
the
dates
of
of
lpp.
G
G
We
can
refresh
and
collect
the
information
this.
This
is
a
topic
to
be
analyzed,
but
we
already
have
a
work
in
the
direction
of
having
clear
that
there
will
be
some
dependencies
and
this
we
these
will
questions
so
not
question,
but
with
influence
the
the
the
speed
in
which
we
can
provide
the
information,
but
as
mentioned
before,
by
by
having
probably,
we
would
not
consider,
although
for
something
in
real
time,
at
least
yet
in
this
space,
so
we
could,
I
think
we
could
live
with
with
that
timings
by
now.
G
E
So,
for
for
second,
you
know
you,
you
propose
a
new
political
extension,
I'm
just
the
one
and
whether
you
attack
it
to
like
a
service
agent
service,
discovery
cases
you
want
to.
G
G
Yeah,
let's
say
that
this
is
is
dual
in
the
sense
that
once
you
have
newer
capabilities,
for
instance,
the
identification
of
compute
environments
into
the
network,
you
need
also
to
enable
ways
of
for
the
application
to
ask
for
it.
So
so
how
do
you
you
have
to
consider
as
an
extension?
How
do
you
collect
the
information
and,
and
also
how
do
you
expose
that
information,
how
the
applications
could
ask
for
that?
So
it
would
be,
let's
say,
dual.
G
Yeah
in
this
slide,
for
instance,
you
can
see
so
there
would
be
some
way
of
collecting
the
the
computing
information
that
could
be
the
next
information.
This
could
be.
For
instance,
there
is
some
proposition
in
in
this
working
group
for
doing
that
as
a
station
of
the
te
topology
model.
G
So
we
could
retrieve
the
data
center
information,
maybe
in
the
form
of
these
bundles,
as
in
the
case
of
cntt
or
either
the
the
raw
information
about
the
cpu
and
and
all
of
that,
and
also
consider
alto
server
as
the
the
way
of
let's
say,
processing
this
information,
together
with
the
topological
information
from
the
network
collecting
from
from
bgp
such
a
way
that
we
could
offer
both
the
view
of
the
data
center
plus
the
view
of
the
transport
network
together
in
such
a
way
that
we
could
allow
applications.
G
B
B
A
We
could
do
a
show
of
hands,
I
suppose,
for
for
each
of
the
I
mean
I
don't
know
I
mean
I
would
first
maybe
ask
do
people
feel
like
they
understand
these
enough
to
let's
go.
I
guess.
A
B
Good
question
so
yeah
comments.
So
for
me
personally,
I
think
speaking
now,
not
as
chair,
but
personally
I
think
most
of
these
extensions
were
pretty
clear
or
at
least
charter
items.
I
should
say-
and
I
think
the
text
already
is
already
in
in
somewhat
good
shape,
but
that's
my
personal
opinion.
Others
may
want
to
speak
up.
E
Yeah
my
opinion,
actually
you
know
I,
I
did
a
full
follow
the
you
know
some
discussion
in
the
design
team,
and
I
this
makes
me
a
better
understanding
of
what
they
want
to
propose,
and
I,
I
think
for
cellular
network
proposal
and
general
extension.
They
seem
to
you
know,
have
some
overlapping
because,
as
serbian
mentioned,
they
may
cover
some
widest
network
of
mobile
network
cases.
So
this
need
to
be
consolidated
and
also
for
operation
automation,
actually,
as
a
you
know,
louis
represent.
E
B
C
C
Okay,
so
number
one.
What
I
wanted
to
say
was
all
these
five
use
cases
and
the
what
five
extensions
that
were
all
involved,
a
lot
of
weekly
discussions,
and
so
basically,
I
think
the
this
fiberbody
among
the
many
many
choices
like
a
relative
well
understood.
So,
therefore,
I
think
that's
why,
personally,
I
won't
support
all
five
us
five
extensions
and,
of
course
I
I
do
agree
with
a
team
comment
that
there
looks
like
some
over
life.
C
You
need
to
be
clarified,
some
like
a
better
delineation
of
the
boundary,
so
among
some,
for
example,
I
I
do
buy
that
and
if
I
really
have
to
pick
and
for
example,
personally,
for
example,
I
personally
think,
for
example,
the
cellular
is
incredibly
important
extension,
which
actually
would
give
us
a
very
good
opportunity
from
ietf,
typically
is
tended
to
slightly
stay
above
layer,
two
earlier
three
layer,
four
and
now
really
abstract
all
the
information
and
work
together
with
3gpp
on
layer,
2
and
so
on.
C
I
think
the
wonderful
operator
would
definitely
should,
for
example,
support
the
cellular
network
and
so
on.
So
if
we
really
have
to
only
pick
3
and
so
on,
I've,
probably
at
least
I
would
personally
vote,
for
example,
cellular
because
that's
such
an
important
use
case
and
such
a
good
opportunity.
So
those
are
just
my
quick
comments.
E
Yeah
yeah,
I
have
an
additional
sword.
Actually
I
I
think,
based
on
auto
protocol.
Actually
you
know
they
design.
You
know
originally
designed
for
the
pwp
application
city
application.
Now,
actually
we
see
you
know
many
people
talk
about
5g
and
edge
computing
and
they
could
be
the
bigger
driver
for
for
us.
You
know
to
extend
auto
program,
but
we
see
a
lot
of
you
know
many
proposals
distributed
in
other
working
groups.
E
So
what
I
like
to
you
know
really
want
to
see
is
is
all
the
working
group
can
have
good
collaboration
with
other
working
or
some
some
problems
in
other
directions.
So
we
can,
you
know,
build
synergy
for,
for.
C
G
Yeah
I
I
would
like
that
simply
to
comment,
maybe
a
way
of
avoiding
overlappings
or
or
mixing
things.
Maybe
maybe
it
could
be
to
to
look
at
the
putting
alto
server
as
a
central
point,
maybe
considering
one
could
be,
let's
say
what
could
be
the
mechanisms
for
feeding
alto
server
with
information,
what
would
be
the
mechanisms
for
exposing
that
information
to
applications
and,
and
so
basically
the
the
let's
say,
the
normal
interface,
the
software
interface
and
maybe
the
east-west
interface
could
be
also
an
angle
to
look
at.
G
We
have
the
discussed
also
in
the
past
in
that
way,
so
maybe
in
that
way
also,
we
could
see
could
let's
say,
make
the
the
avoid
to
overlaps.
Basically,
it's
just
my
fault.
E
Yeah
yeah
louise,
you
have
good
comments.
You
know
for
feeding
in
the
face
right
now
we
seem
to
leave
this
out
of
scope,
but
you
know
we
think
we
make
an
assumption.
You
know
also
it
will
connect
all
the
information
intelligently.
You
know
smartly,
you
know,
but
we,
but
so
for
so
that's
why
you
know
operation
automation.
You
know
they
may
target
to
address
this
issue
to
automate
to
get
all
the
information
so.
H
Okay,
yeah,
I
am
also
hesitant
if
I
had
to
choose,
I
think
it
is
for
me.
The
fact
that
there
is
overlaps
is
is
is
a
a
good
thing,
because
this
means
there
is
also
synergy,
and
yes,
indeed,
the
the
the
purpose
of
general
extension
was
really
to
to
to
to
to
reach
out
to
other
working
group
items
and,
for
example,
an
extension
so
cost
attributes.
H
They
are
definitely
useful
for
cellular
use
cases,
but
they
may
also
be
useful
for
other
use
cases
that
are
mentioned,
for
example,
that
were
mentioned
by
luis
in
one
of
his
drafts.
So
it
is
kind
of
I
think
it
it's
it.
It's
it's
a
good
thing
that
to
have
like
a
placeholder
on
extensions
and
maybe
motivated
by
other
people
working
on
other
working
group
items
in
the
in
the
in
alto.
B
Okay,
so
we
have
just,
we
have
just
a
couple
more
minutes
left
and
the
question
is:
do
we
want
to
show
do
a
quick
show
of
hands
for
the
for
the
items
or.
H
Sorry,
it's
me
again,
sorry
again,
yes,
I
actually.
I
want
also
wanted
to
add
that
I
think
to
better
understand
possible
synergies
and
then
that
could
be
consolidated
in
the
proposals.
H
I
think
maybe
it
would
be
interesting
for
each
person
to
carefully
take
a
look
at
the
slides
at
the
related
material.
H
B
H
To
take
the
time
to
leave
some
time
to
to
to
yes
for
everybody
to
to
take
a
closer
look
at
the
proposals.
B
Sabine,
so
that
is
not.
That
is
not
what
we
are
debating
at
all
here.
So
for
sure,
we
gonna
have
further
discussions
on
the
mailing
list
and
we
are
not
making
any
decisions
here
right
now.
What
we
can
make,
however,
in
the
remaining
two
or
three
minutes
that
we
have
is
have
a
you,
know,
consensus
or
try
to
find
a
like
martin
said
a
show
of
hands
on
what
is
what
is
the
current
status
quo?
B
What
are
people
thinking
of
the
charter
items
on
the
table
for
sure
we're
not
making
any
decisions
and
further
confinement
is
necessary.
That's
that's
totally
clear.
C
Yeah
so
yeah
yeah,
I
agree
yeah.
I
totally
agree,
I
think
sorry
slightly
different
problems.
You
know
I
I
think,
like
a
more
discussion.
Often
time
means
delay
one
or
three
more
months.
That's
probably
a
horrible
thing
and,
if
possible,
let's
make
a
decision
soon,
because
they're
already
designing
getting
started
and
so
on.
All
these,
like
momentum,
let's
don't
say,
let's
discuss
even
more
for
basically
means
we
mislead
meeting.
Today
we
probably
wait
for
three
more
months
or
four
more
months
right
so.
B
I
guess
we're
getting
close
to
the
to
the
end
of
the
session
and
martin.
You
want
to
say
something.
A
Yeah
I
mean
like
I,
I
I
I'm
not
sure
how
effective
you
know,
having
a
show
of
hands
on
five
different
things
is
going
to
be
in
in
30
seconds
or
whatever
we
have
all
right.
A
I
mean
you
could
my
advice
would
be
to
have
somebody
whether
it
be
a
chair
or
somebody
like
richard
or
or
really
anyone
else,
maybe
have
some
informal
discussions
and
and
see
if
there
is
a
a
proposal
for,
like
you
know
about
one
to
two
years
of
work
built
out
of
these
five
five
proposals,
and
then
you
know,
do
a
consensus
call
on
that,
rather
than
just
kind
of
have
a
completely
open-ended.
E
Martin,
actually,
we
can,
you
know,
begin
to
prepare
the
charter,
how
to
consolidate
all
these
proposals
together
to
to
document
each
other
have
a
strawberry
proposal,
and
actually
so
we
can
have
a
meeting
to
to
to
discuss
this,
and
but
I.
F
A
Yeah
yeah
I
mean
I
I
I
don't
think
it
needs
to
take
a
long
time,
but
but
we're
not
going
to
have
time
for
like
a
lot
of
people
to
come
up
on
the
mic
and
and
sort
of
say
what
they
want.
So
so,
maybe
an
offline
way
like
a
not
in
a
meeting
way
is
a
good
way
to
kind
of
gauge
what
people
think.
So.
A
The
other
thing
I
want
to
say
is
I'm
not
an
ultra
protection
practitioner,
so
I
don't
have
a
ton
of
insider
opinion
on
where,
where
would
you
go
with
any
of
these?
A
Some
of
them
sound
harder
to
me
than
others,
but
if,
if
everyone
sort
of
wants
to
do
everything
which
is
kind
of
what
I'm
hearing
and
it's
sort
of
being
forced
to
choose,
some
of
these
are
probably
much
shorter
than
than
than
others
and
picking
up
the
short
you
know
having
it
rechartering
and
doing
a
couple
things
that
can
wrap
up
in
six
months
or
a
year
and
then
moving
to
the
harder
things
is
a
way
to
keep
everyone
engaged.
A
Whereas
if
you
like
pick
the
hard
things-
and
you
know
that
drags
on
for
four
years
or
something
because
there's
conflicts
and
people
can't
reach
consensus
and
they're
hard
problems,
then
the
people
with
easy
problems
are
gonna
go
away,
so
all
else
being
equal.
I
might
recommend
that
as
a
strategy
on
how
to
organize
this,
but
obviously
you
know,
the
consensus
of
the
group
is
more
important
than
that
advice.
B
No,
but
for
sure,
that's
good
advice
and
also
to
take
into
account
that
these
items
are
different
and
and
the
time
they
need
the
time.
I
think
we
should
close.
I
think
I
I
I
agree
with
what
basically
chin
and
martin
just
agreed
to.
We
should
take
this
offline
and
have
some
people
lead
this.
I
think
shin
and
and
and
richard
are
probably.
E
E
C
Yeah
yeah,
I
would
like
to
volunteer
and
or
of
course
with
king
as
the
chair
and,
of
course,
work
with
everyone,
and
we
can
give
a
proposal.
Probably
within
you
know,
a
couple
weeks
max
and
then
probably
we
can
pre-authorize
according
to
the
time
frame,
and
you
know
which
one
shorter,
which
one
easier
we
arrange
a
little
bit
and
definitely
we
want.
Personally,
I
really
want
to
get
you
know.
The
proposal
to
to
you
know
to
martin's
table,
and
you
know
like
yeah,
definitely
king
and
and
getting
his
guidance
I'll
work
with
him.
B
Good
to
me,
I
think
this
is
a
good
agreement.
Yeah,
I
think
we
can
close.
Martin
martin
is
in
the
queue
any
last
words
martin,
you
wanted
to
add.
A
But
but
yeah
sure
so
so,
hopefully
like
in
order
a
month
or
two
like
we're
able
to
reach
some
sort
of
consensus
and
then
bring
that
consensus
to
me.
We
can
wordsmith
the
charter
and
then
get
it
through
the
process.
Thanks.
Everybody.
B
Cool
yeah
thanks
everybody,
so
we
close
yeah.
I
wish
wish
you
all
a
good
good
time
until
the
next
itf
meeting,
whether
I'm
sure
it's
going
to
be
a
virtual
one
again
and
yeah
we're
going
to
have
lots
of
discussions
on
the
mailing
list
and
vijay,
and
I
will
take
mostly
care
of
finishing
the
current
milestones
and
chin
can
help
a
lot
richard
and
others
to
with
the
new
charter.
I
think
this
is
kind
of
the
way
we
as
chairs
support
the
current
and
future
work
and
chen.
You
want
to
add
something.
E
Yeah,
I
think
that's
my
top
priority
work
to
you
know,
have
a
concrete
proposal
and
and
bring
this
is
a
proposal
to
to
the
modern
table,
and
we
can
see
whether
we
can
you
know
recharge
her.
This.