►
From YouTube: IETF115-RAW-20221111-0930
Description
RAW meeting session at IETF115
2022/11/11 0930
https://datatracker.ietf.org/meeting/115/proceedings/
A
A
I
guess
it's
9
30.
correct
at
the
local
time:
okay,
it's
1
30
in
the
morning
in
California!
So
welcome
everyone,
no
matter
where
you
are
thank
you
for
coming
so
good
morning,
good
afternoon
and
good
evening
and
hope
you've
been
enjoying
the
ietf
week
in
London.
A
B
Well,
it's
like
tuna.
Oh
there
we
go
so
as
with
every
IDF
meeting.
This
meeting
is
covered
by
the
note.
Well,
if
this
is
your
first
session,
given
it's
Friday,
that's
that's
I
hope
unlikely
or
your
first
ietf
and
you
haven't
had
a
chance
to
read
this.
The
note
well
covers
the
IPR,
your
participation
in
meetings.
B
That's
a
bit
of
feedback
Please,
be
aware
that
anything
you
state
at
the
microphone
plays
onto
the
mailing
list
or
comment
in
general
or
is
covered
under
the
IPR
regulations
of
the
ietf,
and
do
you
have
agreed
by
that?
If
you
are
unaware
of
any
of
this,
please
check
the
relevant
pcps,
which
we
normally
call
out.
I.
Think
from
the
next
page.
B
The
next
page,
so
here
is
a
list
of
various
pcps
which
should
introduce
you
to
the
internet
standards
process
and
the
processes
of
working
groups.
Talk
about
your
about
copyrights,
patents
and
participation
rules,
but
also
I
will
draw
attention
to
the
Angie
Harrison
procedures,
harassment
procedures
and
code
of
conduct.
Although
we
promote
vigorous
debate,
bullying
and
unacceptable
behavior
is
obviously
unacceptable
and
there
is
a
process
to
address
this.
B
B
B
Because
we're
going
to
use
a
shared
queue
for
the
mics,
please
make
sure
you
use
the
little
app
on
your
phone
or
or
click
the
join
the
queue
button
on
your
laptop,
so
that
remote
participants
can
also
join
the
queue
fairly.
So
we
don't
have
two
cues
as
ever,
if
you're
speaking,
please
state
your
name.
First,
it's
just
useful
for
the
note
takers
mask
wearing
yeah
mask
in
the
room.
Please
it's
policy
unless
you
are
actively
speaking
at
the
microphone
myself
as
actively
speaking
at
the
moment,
also
to
avoid
feedback.
B
Can
you
mute
your
laptop
audio,
particularly
if
you're
remote
to
save
on
feedback
headphones
strongly
recommended?
If
you
are
at
home
yeah,
that's
it
so.
B
This
is
a
collaborative
tool.
I,
don't
believe
we
have
an
official
minute
taker.
So
please
either.
If
you
want
to
take
copious
minutes,
please
attack
right
now.
But,
more
importantly,
if
you
want
to
make
a
point-
and
you
don't
believe
it's
being
caught
or
your
name
is
not
being
captured,
spelled
correctly,
please
get
into
that
online
minutes
and
and
make
sure
that's
properly
recorded.
B
The
probably
collaboration
is,
everyone
assumes
everyone
else
is
going
to
collaborate
rather
than
you,
and
then
we
don't
capture
the
minutes
correctly.
So
it's
a
sort
of
group
responsibility
to
make
sure
we've
got
some
proper
minutes
taken
the
meeting
materials
I'm
assuming
by
the
fact
you
can
see
these
slides.
You
can
get
to
this
URL.
B
So
it's
a
bit
of
a
recursive
problem,
but
all
the
meeting
materials
are
available
through
the
data
tracker
and
through
the
low
speed,
Echo
integration
and
as
ever
we
are
an
ietf
working
group,
which
means
we
have
a
mailing
list
which
is
raw.etf.org.
I
hope
you
are
all
subscribed.
B
A
Sure
so,
for
those
of
you
who
may
not
be
tracking
on
the
mailing
list,
the
the
status
of
things
we
are
very
close
due
to
the
hard
work
of
the
authors
and
our
A.D
John
Scudder,
the
ldax
document
is
on
the
brink
of
ready
being
completely
ready.
It's
the
last
of
the
discuss
status
of
various
ads
and
reviewers
have
now
been
converted
to
abstains.
A
I
believe
John
is
wanting
to
give
it
one
last
very
detailed
read,
so
we
may
be
within
a
within
the
next
week
or
two
having
the
ldax
document
Advance
through
the
RFC
process.
So
we're
very
excited
about
that
and
again
thanks
to
the
hard
work
of
all
the
whoops,
all
the
folks
involved,
the
the
let
me
just
see
if
I
can
I'll
share
the
deck
it
went
away
again.
B
A
A
Here
you
go
so
and
the
use
cases
draft
has
similarly
begun
to
progress
through
the
iesg
and
has
received,
let's
see
a
routing
review
and
it's
been
through
the
tsv.
So
the
transport
and
gen
art
reviews
as
well
and
thanks
to
I
believe,
is
it
Karina.
Who
is
the
Shepherd
for
this?
A
We
are
now
at
a
REV
of
the
document.
I
think
it's
eight
having
Incorporated
all
of
the
feedback
from
the
three
reviews.
So
far
we
still
have
quite
a
few
that
we
await.
A
Then
the
Technologies
draft
has
begun
that
journey,
and
so
Carlos
bernardos
has,
you
know,
solicited
the
feedback
from
the
authors
on
IPR.
So
at
least
at
this
moment
in
time
we
haven't
gotten
any
there's.
Nothing
eventful
happening
in
on
the
IPR
front,
I
mean
no
one
has
any
anything
to
declare
there,
but
in
the
meantime,
we've
had
a
funny
exploration
of
the
draft,
but
I
think
once
things
open
up
again,
it
will
be
resubmitted.
C
Just
to
to
basically
apologize
for
not
taking
or
not
progressing
on
the
on
the
separating
for
that
document,
I
I
started
I
expect
to
have
the
the
right
up
in
the
next
one
or
two
weeks,
while
the
IPR
declarations
are
are
made.
So
I
just
wanted
to
update
a
bit
on
that
and
apologize
for
taking
a
bit
longer
than
it
should
have
been
clicking.
So.
A
Thanks
no
worries:
we
are
greatly
appreciative
of
you
taking
on
the
shepherding
and.
D
Let's
go
just
for
Carlos,
since
you
have
to
indicate
the
number
I
owe
janash
to
publish
a
slight
update
of
the
technology
draft
because
of
the
new
references
that
the
inertia
has
provided
and
probably
that's
what
Eve
is
alluding
to
as
well.
So
I
will
publish
it
like
next
week
arrive
so.
A
Perfect
and
for
those
of
you
who
may
not
know
that
there
was
a
detnet
study
group
in
the
3gppsa2
group,
and
and
so
they
have
a
new
publication
and
so
there's
some
slight
wording
updates
regarding
the
recommendations
there,
that,
once
things
open
up,
Pascal
will
resubmit
on
and
and
then
life
goes
on.
With
regards
to
the
other
documents,
we'll
hear
a
lot
today
about
the
architecture
document,
the
framework
document
is
a
little
bit
off
into
the
future.
Oam
is
progressing.
A
We
do
need
to
understand
what's
happening
with
our
industrial
requirements
document.
Our
last
correspondence
was
that
there
was
going
to
be
a
new
rev,
but
we're
we
need
to
close
on
that
and
then,
of
course,
there
are
three
documents.
I
think
they're,
all
Carlos's
at
least
he's
a
co-author
on
them
that
are
under
consideration
for
a
working
group.
Adoption
have
I
missed
anything.
B
C
Yes,
basically
I
I
was
waiting
a
bit
for
the
working
group
to
be
done
with
the
other
actual
Milestones.
To
then
ask
to
the
working
group
if
this
may
be
considered
as
something
interesting,
so
I
will
be
waiting
until
the
other
things
are
out
of
the
line.
A
Okay,
all
right
I'm
controlling
the
slides.
The
thing
that
the
sleepy
person
in
the
room
or
the
virtual
room
needs
to
remember-
is
to
advance
those
lives,
and
so
our
agenda
today
is
really
to
focus
and
give
time
to
the
architecture.
Draft
and
Pascal
has
done
a
tremendous
amount
of
work
on
it
with
his
co-authors
and
there's
been
lots
of
discussion
both
on
and
off
the
list,
and
he
will
be
giving
us
a
very
detailed
update
on
that.
F
B
This
point
so
I'm,
conscious
that
we
have
a
good
size
session
today
and
and
the
architecture
draft.
Is
it's
getting
a
lot
of
review?
It's
getting
a
lot
of
comment
and
it's
it's
slightly
blocking
our
work.
So
really
what
I
would
like
to
see
is
us
really
tackle
the
architecture
draft
almost
Workshop
it.
We
have
the
time
we
have
relevant
people
in
the
room
and
and
as
a
precursor
to
this
Pascal.
B
Myself
had
had
a
pre-meeting
to
talk
about.
How
can
we?
How
can
we
make
progress?
What
are
the
main
sticking
points
here?
Are
they
insurmountable?
Can
we
make
progress
or
have
we
have?
We
got
a
problem
here,
and
the
answer
was
actually.
We
didn't
have
a
problem.
We
could
make
progress,
but
the
recommendation
from,
in
fact,
the
decision
by
the
chairs
and
and
thank
you
Pascal
for
for
letting
us
make
that
decision
was
given.
B
The
architecture
document
actually
has
two
introduces
two
new
Concepts
one
very
much
about
how
does
raw
make
deterministic
forwarding,
plane
decisions
and
a
separate
novel
piece
of
work,
which
is
a
new
kind
of
service
function
or
a
forwarding
plane
protection
function,
which
is
the
the
pareo
function
by
introducing
both
of
these
new
pieces
of
work
within
one
document.
B
Firstly,
the
document
got
big
and
the
bigger
the
document,
the
harder
it
is
to
review
and
the
harder
it
is
to
maintain
readability
Etc,
as
we
all
are
well
aware,
and
secondly,
we
felt
as
chairs
that
the
the
introduction
of
the
new
forwarding
plane
control
ability
and
that
fast
feedback
loop
around
PSE
was
the
architectural
piece
and
the
perio
piece
could
be
introduced
in
parallel
at
a
slightly
slower
rate.
B
We
should
just
split
the
document
in
half,
and
that
means
we
would
have
a
lighter
architecture
document
that
would
talk
about
PSE
and
make
sure
it
was
a
good
quality
document
that
framed
a
bigger
pardon
that
framed
what
raw
was
hoping
to
achieve,
and
then
we
could
take
time
to
get
perio
nicely
described,
have
good
quality,
discussion
and
review,
and
so
on.
The
complication
when
splitting
a
very
large
document
is.
B
B
I
will
say
the
chairs
and
the
ad
are
very
happy
to
discuss
that
decision.
To
make
sure
people
understand
that
the
authors
understand
that
clearly,
but
we'll
take
that
Offline
that
that
isn't
really
a
matter
for
the
working
group.
It's
a
it's
a
matter
for
the
chairs,
given
we
are
moving
to
the
there
is
an
editor
and
a
group
of
and
and
the
working
group
contributes
to
it
model.
B
You
have
to
name
that
editor
as
Pascal
has
been
carrying
the
editorial
pen,
at
least
over
the
last
couple
of
revisions,
and
he's
happy
to
do
so.
We
wish
to
make
Pascal
the
editor
of
these
documents
again,
if
you
feel
that
is
contentious
and
you
have
any
complaints
about
it.
Please
talk
to
the
chairs
and
the
ad
and
we
will
get
this
resolved
in
some
way.
B
As
far
as
the
working
group
is
concerned,
I
believe
this
decision
will
help
us
make
progress
on
Raw
architecture
and
let
us
do
proper
review
and
consideration
on
perio
and
unblock
some
of
the
workers
Carlos
alluded
to,
which
is
sat
in
the
backlog
waiting
for
adoption
or
for
those
of
us
to
find
review
time
to
look
at
when
we
have
a
nice
architecture
document.
We're
really
happy
with
that
can
frame
our
our
decisions
and
our
reviews
would
anyone
Carlos
is
at
the
mic,
go
for
it.
C
Thanks
Carlos
I
I,
don't
have
any
issue
with
the
decision.
I'm
perfectly
fine
I
think
this
is
the
right
one
to
keep
Pascal
as
a
leader
of
documents.
C
I
I
had
just
a
qualification
question,
so
is
this:
a
split
meant
to
be
like
one
document
is:
will
will
be
like
taking
care
of
how
the
raw
architecture
entangles
with
the
dead
net
and
ensure
that
everything
is
coordinated,
while
the
other
one
will
basically
take
care
of
the
very
much
Wireless
specifics
and
in
this
way,
kind
of
facilitate
the
integration
I
mean
because
you
mentioned
optimize
functions
will
be
in
the
new
work
document.
C
B
Yes,
I
believe
you
have
understood
correctly,
you,
you
said
one
thing
there
which
I
would
qualify,
which
is
the
architecture
document
is
the
you
said.
Less
one
is
focused,
I
would
disagree
there,
because
I
would
say
in
that
architecture
document.
We
still
talk
about
the
fast
feedback
loop
with
the
OEM
plane,
because
Wireless
moves
much
faster
than
than
in
general
wired
networks,
and
so
we
have
that
faster
feedback
loop
and
that
PSE
concept,
which
I
would
suggest,
is
very
much
Wireless
focused.
B
You
know
these
are
Wireless
adaptations,
but
then
the
the
details
of
the
the
protection
function-
perio,
specifically
Wireless
optimized
protection
function,
makes
more
sense.
Yeah.
C
B
Okay,
thank
you
I'm,
going
to
pause
here.
If
anyone
else
has
any
comment
on
on
this
decision,
this
is
a
good
moment
to
air
concerns.
If
you
have
any
otherwise,
Carlos
is
still
in
the
queue,
no
he's
gone.
Otherwise,
we'll
move
on.
What's
the
next
slide,
I
think
we're
going
to
get
straight
into
the
into
the
guts
a
bit.
Aren't
we
yeah
so
Pascal.
You
have
plenty
of
time.
B
E
D
D
Okay,
well
next
slide,
please,
okay,
so
I
will
start
with
the
draft
with
the
slides
I
prepared
before
the
meeting,
knowing
that
with
the
recent
discussion.
Some
of
that
is
partly
resolved
or
appears
to
be
on
good
way
of
resolution
to
be
agreed
with
the
group
and
so
we'll
see
where
we
are
for
which
those
slides
might
be
slightly
obsolete.
D
So
what
happened
recently
is
a
lot
of
great
reviews.
I
mean
a
lot
of
exchanges
with
lu
yeah.
Thank
you
all
for
for
the
time
you
you
spent
on
on
reading
the
document
and
and
considering
it
and
Etc
part
of
the
results
is
Lou
is
added
as
a
contributor
to
document,
and
but
now
she
strongly
acknowledged
I,
don't
remember
if
it
is
on
the
published
version,
but
you
are
certainly
in
the
acknowledgment
Village
and
then
there
were
lucaments
about
pario
that
we
are
still
need.
D
That
still
need
to
be
discussed
and,
with
you
know,
the
decision
that
the
chairs
made
those
commands
now
will
apply
to
a
separate
document,
and
so
they
will
be
made
separately
and
they
are
not
blocking
the
architecture
document
anymore.
D
We
did
a
lot
of
restructuring
to
help
the
new
readers,
because
we're
all
used
to
what
we
are
saying,
but
the
new
reader
may
want
something
more
serial.
So
we
we
reorganized
a
bit.
The
introduction
got
simpler
and
based
on
the
comment
by
Liu
of
what
is
raw
exactly
we,
we
extracted
a
new
sections
free
also.
The
term
UDA
appeared
many
times
that
it
was
a
fundamentally
new
thing,
so
that
has
been
there
for
like
50
plus
years
and
the
fact
that
it
respects
the
UDA
model
is
not
the
fundamental
fundamental
thing.
D
The
fundamental
thing
is
we
have
this
fast
control.
Loop
UDA
is
a
model
for
control.
Loop
is
the
model,
the
model
that
we
are
following,
but
it's
a
control
Loop.
So
we
we
renamed
the
section
row
control
Loop.
To
insist
it
is
a
control
loop,
as
opposed
to
insisting
that
it's
following
the
odometer
right.
D
There
was
also
these
points.
I
mean
it's
my
French
here
where
I
said
something
like
Row
versus
Net
in
one
section
title
and
for
me
it
meant
nothing
but
probably
in
English.
There
is
a
sense
to
that
and
we
I
really
made.
You
know
how
do
you
compare
this
and
that
that
was
the
meaning,
but
these
verses
that
appears
to
be
a
conflict?
D
Certainly
rowan.net
don't
conflict.
The
goal
is
that
you
know
it's.
We
are
providing
a
plug-in
functionality
into
that
net,
and
so
there
is,
it
cannot
conflict.
It
has
to
smoothly
integrate.
So
certainly
the
title
gave
the
wrong
impression.
Sorry
for
my
French,
so
yes,
we
clarified
that
you
know
row
is
a
specialized
announcement
that
mostly
applies
to
wireless,
but
since
we
are
layer
3,
it
could
apply
anywhere,
but
the
needs
really
appeared
in
the
context
of
ILS
through
RS
mesh
next
slide,
please
a
number
of
issues
that
we
discussed.
D
There
was
one
and
probably
it's
against
again
my
French
all
the
way
expressed
things
but
seemed
that
the
text
could
be
read
as
if
we
were
changing
something
on
wireless.
We
are
not
touching
Wireless.
That
would
be
a
layer
violation.
The
best
we
can
do
is
Prov
provide
hints
to
the
lower
layer
about
the
sort
of
thing
we
would
like
to
see
achieved,
and
we
currently
don't
have
even
the
apis
for
doing
this,
but
the
architecture
should
show
that
we
expect
that
someday.
D
We
might
have
something
and
we
need
to
position
it
and
that
something
is
basically
some
capability
to
hint
the
lower
layers
of
the
sort
of
result
that
we
would
like
to
see.
An
example
of
results
is,
if
you
send
something
on
Wi-Fi
today,
you'll
get
by
default,
64
retries,
which
could
take
forever
in
in
row.
That
doesn't
make
any
sense,
because
we
have
we
have
a
bounded
latency.
D
So
at
some
point
we
need
to
tell
hey
Mr,
Wi-Fi
I
know
you're
doing
your
best,
but
if
you
don't
send
that
within
that
bonded,
latency
drop
it
or
a.
If
you
can,
please
do
at
least
a
few
retries
right.
You
don't
know
you're
doing
that.
That's
Layer
Two,
but
try
to
make
the
reliability
of
this
packet
a
factor.
Five
I,
don't
know
how
to
express
that,
but
make
it
more
reliable
than
this
other
packet
by
doing
more
retries.
Doing
retries
faster,
I,
don't
know
the
way
you
do
it.
D
D
So
yes,
we
we
clarify
that
we,
we
are
the
architectural
layer,
so
what
we
are
doing
is
conceptual,
but
we
are
basically
adding
new
services
and
they
are
considered
and
meant
to
apply
at
least
to
the
technologies
that
we
we
support.
We
have
in
our
Charter
and
are
listed
in
the
Raw
technology.
D
D
We
already
have
a
lot
of
work
and
capabilities
for
getting
information
from
the
lower
layer,
and
then
we
use
that
in
writing.
We
compute
we
have
our
SSI,
we
lqi,
we,
we
can
do
it
TX.
There
are
tons
of
things
we
can
do.
We've
got
delete
and
we
want
to
build
on
all
this
and
at
the
same
time
we
need
to
to
do
it
over
Malta
help.
So
we
need
to
differentiate
this
concept
and
see
how
they
work
together.
That's
part
of
of
what
the
architecture
must
represent,
and
then
there
was
this.
D
This
question
from
it
was
coming
from
the
blue,
but
one
of
the
parallel
functions
is:
is
over
a
yearing,
meaning
that
one
sounds,
and
you
know
by
the
magic
of
of
the
radio
transmission,
there
will
be
multiple
receivers
and
again
it's
probably
my
French
I
must
have
written
it
such
a
way
that
it
looks.
Oh,
it's
completely
new
to
write,
jokes
radios
and
obviously
yellow
wires
were
like
that.
So
it's
not
like
it's
completely
new
to
radios.
D
If
you
configure
a
switch
with
promiscuous
mode
on
one
port,
that
bot
will
get
all
the
packets
mostly
we
do
it
for
tracing
not
for
forwarding,
but
but
that
exists
you
could
even
illumination
that
you
could
even
have
an
underlay,
which
is
a
multicast
to
send
your
unique
aspect.
I've,
never
seen
that
I,
don't
know
why
we
should
do
it,
but
it's
feasible.
D
So
yes,
over
a
yearing
could
happen
at
a
cost.
Just
happened
that
for
wireless,
yes,
one
transmission
from
A
to
B
is
less
C
compared
to
wires,
but
there
are
ways
which
are
other
properties
of
the
same
medium
by
which
this
disadvantage
can
be
compensated
and,
and
it
comes
from
the
same
magic
that
causes
the
loss,
is
that
others
can
listen
to
it.
D
Can
we
do
something
to
compensate
the
fact
that
one
point
to
point
radio
obstruction
is
low
C,
with
the
fact
that
in
fact,
physically
speaking,
it
is
a
pawn
to
multi-point
thing
and
that's
really
what
we
want
to
discuss
with
these
promiscuous
over
hearing
thing,
don't
say:
hey,
yes,
you
you
receive
ratio
drops.
If
you
consider
the
link
as
point
to
point
like
we
would
on
wires
typically,
but
on
this,
on
the
other
hand,
there
is
a
zero
four
ratio
on
the
other
guy
here
that
we
could
exploit
because
he
got
it.
D
If
we
consider
that
in
the
overall
routing
a
maybe
the
the
loss
here
is
compensated
by
the
received
is
there
and
rho
is
a
lot
about
doing
that,
and
one
of
the
methods
effectively
that
we
can
Leverage
is
this
of
our
urine
sec.
So
we
we
insist
that
it's
not
it's
something
that,
yes,
it
was
always
there.
It's
not
the
question.
D
The
question
is
here:
we
intend
to
really
use
it
and
build
on
it
and
consider
that
it
exists
and
include
it
now
rotting
thinking
so
that
if
the
the
packet
is
received
by
the
wrong
guy
and
the
wrong
guy
is
on
the
right
path,
yeah,
the
packet
can
continue
see
that
that's
that's
part
of
raw
next
slide,
please,
okay!
So
when
I
said
that
some
of
these
slides
and
that's
basically
a
third
of
them
is
probably
obsolete
by
now.
This
is
this:
is
it
so
we
need
to
discuss
Barrio.
D
We
need
to
go
into
the
details
of
how
it
works.
What
it
does,
whether
it's
new
or
not,
is
not
so
relevant,
more
relevant
is
how
we're
going
to
use
it,
how
we're
going
to
abstract
it,
how
we're
going
to
talk
to
the
lower
layers
to
get
them
to
do
things
for
us,
because
we
don't
do
it
for
the
most
Peace
by
ourselves
at
day
or
three,
but
it's
another
document,
so
this
document
is
supposed
to
give
the
big
picture.
So
we
have.
D
We
have
this
tension,
it's
not
on
the
slide,
but
I
realize
we
had
this
tension
with
Lou
about
a
new
I
mean
you
know,
you're
free
to
come
to
the
mic
anytime,
if
I'm,
not
misspeaking,
but
we
have
this
tension
where
I
had
an
intention
that
this
architecture
is
the
big
picture,
I
did
not
focus
so
much
on
what's
new.
What's
not
new
I
focused
on
here
is
how
it
works
right
here
is:
there
is
the
map
of
London
and
if
the
the
London
Bridge
has
been
there
for
200
years,
it's
still
on
the
map.
D
It's
still
a
fundamental
component
of
this
architecture,
whereas
another
reader
could
say
hey
all
these
things.
I
know
them
well.
I
just
want
this
document
to
focus
on
exactly
what's
new.
Well,
if
I
just
give
you
what's
new,
you
don't
have
the
map
of
the
city.
You
just
have
the
shot
right.
So
there
is
this
tension
about.
You
know
wasting
time
of
the
reader
by
saying
something
he
already
knows,
and
yes
in
everywhere,
and
the
fact
that
you
still
need
to
provide
the
map
of
the
city.
D
So
that's
something
we
have
to
arbitrate,
but
personally,
I'm
happy
with
the
current
content
of
the
document.
I
would
not
like
to
strip
it
too
much.
So
I
focus
really
only
on
the
very
new
things.
I
want
to
to
show
how
the
components
are
integrated
together,
the
new
and
the
old,
and
then
there
was
this
track
segment.
D
D
We
have
agreed
that
the
term
path
was
so
overused,
overloaded
that
it
lost
a
lot
of
its
significance
and
we
kind
of
when
we
talk
about
path.
We
end
up
not
really
figuring
out
what
the
guy
is
talking
about.
So
for
the
thing
that
we
built
for
radio
multipath
Etc,
we
crafted
this
term
of
truck
and
so
I
as
I'm
coming
from
that
world
I've
been
using
it
and
using
it
and
using
it
it's
using
the
60s
architecture.
D
It
is
in
the
role
documents,
it
is
everywhere
and
if
you
look
at
it,
it
does
not
represent
a
path,
as
most
of
us
are
used
to
call
a
path.
A
path
in
the
past
research,
working
groups
and
stuff
is
an
observable
thing.
It
is
the
result
of
a
packet
going
through
a
network.
The
path
is
where
this
packet
has
been
to
go
from
A
to
B.
It
is
a
consequence.
It
is
an
observable
I.
D
I
have
listed
in
the
document,
the
quote,
because
if
I
don't
speak,
it
correctly,
at
least
the
code
is
in
the
architecture.
Now,
if
you
go
to
other
documents,
you
will
see
other
definitions
of
path
and
they're.
They
got
more
and
more
wide
and
more
and
more
elastic,
and
then
Lou
understands
the
definitions
that
are
done
in
teas
Etc.
So
but
but
my
concern
was
path
is
supposed
to
be
an
actual
thing.
It's
supposed
to
be
where
the
packet
is
going
to
be,
and
it's
observable,
it's
measurable.
You
know
what
it's
going
to
be.
D
D
For
those
who
have
seen
that
it's
still
predictable,
we
know,
will
control
we'll
know
we're
going
to
do
that,
we'll
throw
the
truck
out
at
the
Moon,
it
will
change
its
orbit
and
we'll
know
the
new
orbit
now
track
is
not
like
that.
The
truck
is
the
aggregate
of
all
the
possible
place
the
packet
can
be,
but
at
any
point
of
time
you
don't
know
where
it
is.
The
packet
can
be
Network
coded,
meaning
that
it's
turned
into
smaller
chunks,
which
are
sent
on
different
paths.
D
There
is
not
one
place
where
the
packet
is
okay,
and
because
of
the
PSC
activity
at
for
for
this
packet,
and
maybe
a
few
packets,
it's
going
to
be
a
subset
of
all
those
hubs
which
are
going
to
be
used
for
this
packet.
It's
not
the
whole
track.
The
whole
story
of
what
we
are
doing
here
is
to
not
use
the
full
track,
but
always
use
a
subset
of
it.
D
That
is
enough
to
get
2.5
copies
of
your
packet
to
the
end,
but
it's
never
the
same
until
you
stop
the
system
and
observe
one
package,
you
never
know
what
the
packets
will
be.
So
that's.
Why
I
like
this
Quantum
thing,
this
Quantum
analogy,
because
the
track
is
to
an
orbit.
What
an
orbital
is.
You
know
for
electrons,
the
electronic
orbital
is
a
statistical
representation
of
where
quote-unquote
the
electron
can
be,
but
unless
you
stop
the
system
to
observe
it,
you
will
never
know
where
the
electron
is
and
in
fact
we
take
some
work.
D
Even
if
you
stop
the
system,
if
you
don't
know,
but
the
analogy
is
there,
the
analogy
is
the
track
represents
a
statistical
aggregate
of
possibilities
where
our
path
is
the
observable
result
of
a
packet
I've
I've
been
that
has
been
flowing
through
the
network
and
if
we
use
the
same
term
I
was
concerned.
That's
why
I
had
all
that
discussion
that
it
be
the
term
path
now
is,
is
even
more
abstract
and
people
less
and
less
understand
what
they're
talking
about
so
I
was
all
for
a
special
term.
D
Now
Lou
told
me
yesterday
that
maybe
there
is
the
term
that
net
pass
or
something
that
is
in
the
detnet
documents
that
we
could
use.
The
bottom
line
is
we
kind
of
agreed
that
the
terminology
must
provide
all
the
terms
and
provide
the
synonyms
and
kind
of
explain
that
this
actually
comes
from
different
place,
but
means
the
same
thing
as
that
kind
of
so
I'll.
D
Try
to
craft
that
and
I
hope
Lou
will
help
me
do
that
terminology
then
we'll
have
to
the
thing
which
will
be
around
on
me,
but
I
have
probably
have
to
do.
It
is
going
through
the
text
and
see
every
time.
I
use
the
term
track
and
change
it
to
this
new
thing
which
comes
from
the
cheese
world,
it's
less
familiar
for
me
all
of
a
sudden
I
feel
that
I
don't
Master
what
I'm
talking
about
anymore,
but.
B
So
can
I
jump
in
at
this
point.
I
am
in
the
queue
as
well
as
kind
of
with
my
chair
hat
on
as
I
see
it.
Rory
has
a
very
close
relationship
with
detnet.
We
are
very
much
a
sibling
group
of
of
debt
net,
so
I
think
there
is
relevance
in
making
sure
the
terminology
we
use
in
our
documents,
Maps
very
closely
to
the
terminology
used
in
detnet.
So
someone
in
in
the
world
of
deterministic
networking
can
understand
what
that
net
does
and
what
raw
does
using
the
same
terminology.
B
I
absolutely
understand
the
point
of
within
the
Wireless
World,
completely
different
terms
are
used
for
things
which
may
or
may
be
close
to
terms
used
in
the
debtnet
world.
So
I
absolutely
agree
with
the
suggestion
that
some
kind
of
glossary
to
make
sure
that
there
is
a
mapping
between
the
different
terminology
universes
out
there,
but
I
would
suggest
strongly
that
the
main
terms
are
you
that
are
used.
B
G
D
G
D
You
cannot
say
do
this
from
it's,
not
from
a
local
Loop
or
anything
that
that
is
kind
of
hidden
inside
the
system,
but
it's
something
which
is
recontrolled
by
the
outside,
then
for
a
long
time
quote
unquote
in
those
orders
of
magnitudes,
it's
just
a
matter
of
orders
of
magnitude
and
time,
but
for
for
a
long
period
of
time,
which
is
observable
by
the
PC
Etc.
The
the
network.
Behavior
will
be
exactly
that
and
then
you
can
decide
for
whatever
reasons
most
reliability
to
change
that,
and
then
there
will
be
an
action.
D
G
They're
often
called
primary,
secondary
or
protection
or
alternate,
and
a
the
the
actual
path
that
is
used
and
that
may
be
one
or
more
is
based
on
a
the
protection.
Switching
function,
that's
being
used
yeah,
so
we've
had
one
plus
one
protection
of
which
pre-off
and
net
is
a
new
type
of
packet.
G
One
plus
one
we've
had
one
for
end
protection
and
in
for
various
Technologies
to
me,
Paro
sounds
like
a
new
one
for
n
protection
scheme
that
is
being
used
being
defined
for
wireless,
and
maybe
it's
not
one
friend,
maybe
it's
one
plus
or
fourn,
because
sometimes
you
might
use
multiple.
Sometimes
you
might
use
one.
G
I
prefer
to
stick
to
terms
that
we've
used
in
the
ietf
and,
really
you
know,
we're
operating
within
the
context
of
ITF
traffic
engineering
from
teas.
That's
further
downscoped
to
detnet,
which
is
further
downscope
to
Raw.
So
it
would
be
good
if,
as
we
sort
of
proper,
you
know
populate
up
in
terminology
that
we
inherit
terminology
from
the
groups
that
we
are
inheriting
technology.
I.
D
I
understand
that
desire,
but
at
the
same
time
understand
it
I'm
all
for
it
conceptually
incompetual
for
it
I'm
discussing
because
each
time
groups
like
traffic
engineering
have
taken
the
same
time
as
before
and
extending
its
meaning.
The
the
sense
of
that
world
has
become
to
started
to
become
more
and
more
fluid
or
something,
and
so
two
people,
unless
they
really
speak
within
the
same
context,
start
not
talking
about
the
same
thing
and
they
don't
even
realize
it.
D
That's
the
danger
of
overloading
terms,
and
here
I
have
the
cells
that
we
make
a
small
step
overloading
path,
yet
a
little
more
because
in
my
view,
I
mean
I.
I
want
to
use
these
I
I
want
to
do
it
anywhere.
There
is
one
desire,
I
have
to
change,
and-
and
you
would
not
even
need
to
ask-
is
for
everything
I
called
a
sub
truck,
which
is
a
definitely
well
organized
subset
of
the
truck
I
wanted
to
call
it
a
t-path
or
use
your
words.
I
mean
a
path,
because
that
piece
is
deterministic.
D
So
I
have
this
term
subtract
that
I'm
very,
very
happy
to
remove,
because
it's
effectively
in
my
mind,
matching
what
you
define
and
and
for
exactly
the
reasons,
the
exact
reasons
you
say,
I
want
to
do
that
change
now
for
the
track
itself,
which
is
the
set
of
all
those
things,
the
the
coalescence
of
all
those
t-path
I.
Think
it's
it's
it's
it's
a
little
more.
C
G
Was
created
well,
no,
it
is
when
it
was
rechartered
and
when
T's
broke
off
it
was
its
scope
was
set
to
help
new
technologies,
reuse,
iatf,
traffic
engineering
techniques,
terminology
Solutions,
and
that
old
presentation
talked
about
how
it's
very
common
for
someone
who
has
a
new
technology
to
show
up
spend
a
whole
bunch
of
time
adapting
to
what
they
think
and
saying.
Okay,
we're
we're
ready
to
go.
This
is
our
solve
salute.
G
This
is
our
solution,
it's
time
for
an
RFC
and
then
to
enter
the
c-camp
group,
and
that's
the
start
of
what
ccamp
does
is
help
take.
What
is
this
unique
technology?
Well
from
the
perspective
of
those
people
who
are
coming
in
with
the
new
technology,
here's
a
unique
solution
and
saying
how
much
of
this
solution
is
actually
unique,
and
how
much
is
generic
and
part
of
the
the
process
is,
is
teasing
out
the
terminology
to
get
the
terminology
aligned
with?
G
What's
in
it
already
there
with
ietf
traffic
engineering
and
then
by
getting
the
terminology
right,
you
help
identify
which
mechanisms
exist
and
clear
used
which
mechanisms
are
brand
new.
We
are
in
them
in
that
process.
Now
we're
not
quite
at
the
very
beginning,
we've
been
the
working
group's
been
running
a
little
while,
but
I
don't
think.
We've
had
that
good
synchronization
and
we
would
might
benefit
from
talking
to
the
Sea
camp
chairs
and
that
C
Camp
working
group
and
maybe
even
doing
like
a
joint
interim
and
say
help
us.
G
You
know,
merge
our
now
well-formed
architecture
document
into
ietf
traffic
engineering,
so
we're
repeating
a
process
that
we've
seen
many
times
for
many
different
Technologies
by
the
way
microwave
was
one
of
them.
It's
it's
a
it's
a
wireless
technology.
That
is
point
to
point.
So
you
don't
have
overhearing,
you
know,
so
it
has
different
characteristics
than
what
you
want,
but
it's
still
Wireless,
so
they've
been
down
this
path
before
I
think
we
could
benefit
by
trying
to
work
more
closely
with
them.
D
G
So
From
a.net
perspective.
We
have
something
called
a
service.
The
service
may
be
supported
by
a
set,
a
forwarding
layer
paths,
I.
G
D
D
C
F
For
my
French,
when
we
started
in
the
death,
Network
group
TSM
worker
already
done
in
93
plea-
and
we
had
exactly
a
very
similar
discussion-
those
days
because
TSN
already
had
the
terminology
in
that
net.
This
was
okay
in
in
that
we
have
not
yet
had
the
ietf
terminology
for
all
this
stuff.
So
there
were
lengthy
discussion
and
and
similar
methods
like
like
Loop
reports.
We
definitely
had
the
luxury
that
having
the
TSN
guys
also
in
the
room
during
the
discussion,
and
we
have
made
that
happen
in
the
architecture
document.
F
So
there
you
can
find
also
some
IEEE
technology
and
terminology,
and
it
is
pointed
exactly
what
is
that
net
relation
that
the
terminology
relation
to
those
terminologies,
so
something
similar
here
I
think
would
would
be
excellent
to
that.
The
the
second
comment
is
is
about
distract
track
story,
so
I
think
in
the
document.
You
have
also
complex
path
and
then
I
have
read
the
document.
Yes,.
D
F
Understanding
was
that
it
is,
it
is
describing
your
track
terminology,
and
then
we
have
a
terminology
with
path.
This
is
describing
what
you
wanted
to
have
with
with
that,
and
and
also
also
regarding
the
flows
we
have
defined
in
that
net.
The
member
flow
and
the
compound
flow,
which
is,
of
course
on
flow
level,
but
something
very
much
related
to
that
root,
just
not
to
use
track
or
path
but
root
to
take
over
the
network
for
the
packets
belonging
to
a
flow.
D
I
can
answer
the
second
one.
First,
if
you
don't
mind,
I
think
there
is
a
bug
in
that
net.
I
try
to
raised
it
a
number
of
times,
and
the
bug
in
that
net
is
that
there
is
a
huge
confusion
between
the
water
and
the
pipe.
The
flow
is
the
water.
That's
what
goes
from
A
to
B.
The
pipe
is
quote
unquote
of
this
by
path
discussion.
It's
where
it's
gonna
go.
What's
gonna
happen
to
the
packets
on
on
the
flight
Etc,
the
the
flow
is
in
that
net.
Typically,
not
the
compound
flow.
D
The
base
flow
is
is
identified
by
the
five
tuples
meaning
it
is
a
application
layer
construct.
The
application
decides
what
goes
in
the
flow.
At
some
point
there
is-
and
it's
written
like
this
in
the
architecture
I
know
because
I
wrote
it
that
says
at
some
point
the
flow
is
mapped
to
a
path,
and
that's
intentional.
D
The
intention
is
to
say
there
is
a
function
somewhere
that
we
decide
that
this
water
goes
in,
that
pipe
and
the
cool
thing
about
separating
the
water
and
the
pipe
is
that
you
can,
for
instance,
decide
that
this
oam
goes
in
the
same.
Pipe
will
have
the
same
treatment
as
this
water.
So
all
of
a
sudden,
the
OEM
can
absorb
the
water.
D
If
we
do
things
based
on
application,
signaling,
like
the
five
tuples,
we
are
screwed
because
of
a
sudden
we
have
all
those
plants
hey.
How
shall
we
put
my
OEM
on
the
same
falling
operation?
You
can't
because
you
made
that
mistake
of
signaling
stuff
with
the
application.
I
mean
it's,
not
your
stuff
that
belongs
to
the
application.
D
That's
a
bug,
so
I
I
went
to
the
mic.
Several
times
tried
to
to
say:
hey,
hey,
we
confused
the
water
and
the
pipe
I
wrote
this
IPv6
that
Network
IPv6
document
which
I'm
not
maintaining,
but
when,
when
the
group
is
finding
interested
in
doing
that,
I
will
be
happy
to
revive
it.
Where
I
write
this
down
in
words
and
I,
show
how
IPv6
can
signal
not
the
flow,
because
that's
already
what
the
application
did
in
the
packet,
but
how
you
process
it
in
the
network,
which
is
the
pipe?
D
E
F
The
age
node
at
the
age
of
the
that
net
domain,
you
are
just
doing
the
death
net
flow
identification,
so
it
is
not
an
application
related
signaling.
This
is
something
but,
but
you
are
deciding
at
the
age
of
the.net
node
and
within
the.net
domain.
You
are
using
those
credentials
to
identify
the
flow.
D
You
fix
my
back
as
long
as
we
relate
what
that
net
does
I'm,
not
talking
about
the
Ingress
function.
That
puts
the
data
in
I
mean
that
exists,
but
once
we
have
decided
that
this
thing
goes
in
that
pipe,
we
should
not
be
looking
at
the
application
information
anymore.
We
should
signal
our
home
pipes
and
if
we
put
multiple
pipes
in
a
bigger
pipe,
we
just
seen
all
that
it's
our
one
domain.
We
can
decide
that
we
signal
it
and
now
the
compound
flow
is
just
a
pipe
inside
another
pipe
right.
D
D
Yes,
the
flow
identification
will
record
will
have
some
mapping
table
of
sort
which
will
do
several
things
recognize
the
application
flow.
If
it's
not,
if
it's
stacked
by
five
double
stacked
by
five
double
it
will
assign
a
path,
it
will
do
some
shaping
if
the
flow
doesn't
respect.
You
know
the
shape
of
the
pipe,
because
we
insert
that
flow
in
that
pipe.
The
shaping
must
be
correct.
D
So
there
are
all
those
things
which
happen
at
the
Ingress
Edge
and
once
that
is
done,
we
should
not
see
look,
observe
think
about
any
application
layer,
signaling
we
are
in
our
world
and
I.
Don't
think
that
that
net
specifications
are
clear
and
enough
on
that,
but
at
least
my
IPv6
hot
by
Hub
document
is
very
clear
on
that.
F
Okay,
let's
have
further
discussion
on
that.
Maybe
just
one
final
comment
is
that
in
that
net
we
have
two
data
plane,
not
only
IPv6.
We
have
also
mpls
data
plan
so.
D
But
in
mpls
I'm
sure
that,
naturally,
you
fix
this
bug
because
you
label
it's
a
label
inside
label
so
very
soon,
you're
in
your
own
labels,
exactly
and
so
by
Nature,
even
without
thinking
about
it
or
wording
it,
you
kind
of
solve
that
problem,
but
still
a
lot
of
the
discussion.
I'm
hearing
and
the
all
these
things
about.
Five
top
also
know
lead
to
the
impression
that
we
have
this
back
and
I
want
to
be
very,
very
clear
in
everything.
I
write
that
I'm
I'm,
not
writing.
For
that.
H
I
John
Scutter
Juniper
Networks,
so
I
somewhat,
regretfully
want
to
return
to
the
the
previous
conversation
you
are
having
with
Lou,
where
which
you
know
made
me
think
a
little
bit
of
Alice
in
Wonderland,
and
you
know
a
word
is
exactly
what
I
mean
it.
You
know,
use
it
to
mean
neither
more
nor
less
and
I
I.
Also
I
still
feel
like
a
new
guy
in
this
group
and
like
I
kind
of
walked
into
the
living
room
when
the
conversation
is
well
underway.
I
So
you
know
forgive
me
as
I
continue
to
come
up
to
speed,
but
it
it's
sounded
to
me
listening
to
you
guys,
Converse,
like
you're,
saying
I
really
want
to
not
reuse
your
words
when
they
have
a
different
shade
of
meaning
from
what
I
mean
I
already
have
a
terminology
set.
It
means
just
what
I
want
it
to
mean,
and
it's
much
easier
for
me
to
write
a
precise
description
of
my
architecture
using
the
terminology
that
is
precisely
tailored
for
that
purpose.
Fair.
E
I
I
think
what
I
heard
Lou
saying,
among
other
things,
was
we
have
a
a
terminology
set.
That's
really
pretty
close
to
what
you
mean
and
you,
if
you
use
that
it
will
make
the
document
a
lot
more
accessible
to
a
wider
audience.
I
I
It's
obviously
good
for
an
architecture
to
be
precise
and
unambiguous.
It's
obviously
good
for
an
architecture
to
be
accessible
to
a
new
reader,
and
it
furthermore
seems
to
me
that,
with
anything
complicated
the
first
time,
a
new
reader
reads
through
it.
No
matter
how
precise
you
have
been
in
your
choice
of
terms
they're,
going
to
misunderstand
something,
because
that's
human
nature,
so
it
seems
to
me
like
one
way
of
squaring
the
circle
is
and
I'm
sure
that
this
has
been
part
of
the
discussion.
I
Maybe
even
right
now
is
to
have
a
careful,
exhaustive
and
precise
set
of
definitions
and
to
say
look,
you
can't
really
understand
the
document
unless
you
have
understood
and
internalized
all
the
definitions
and
applied
them
as
you're
reading
the
document
right,
so
you
know
you,
you
got
to
expand
all
the
pound
defines
sorry
to
date,
myself.
D
D
I
So
so,
actually,
I
I
don't
mean
this
as
a
as
a
criticism
of
the
document
as
it
stands,
I'm
just
sort
of
trying
to
set
up
one
more
thought,
which
is
once
you've
accepted,
that
the
only
way
that
your
document
is
really
going
to
be
properly
understood
is
by
somebody
who
has
carefully
read
your
definitions
and
is
carefully
applying
them.
Then
you
can
use
either
terminology
set
right.
You
can
use
your
native
terminology
set
that
you
have
been
thinking
in
you
can
use.
I
So
it
seems
to
me
that
you
know
if
you're,
going
to
bias
towards
one
thing
or
another,
it's
maybe
more
valuable
to
bias
towards
making
the
document
accessible
and
towards
really
focusing
on
as
you've
done
towards
focusing
on
using
definitions
to
make
sure
that
your
your
meeting
is
completely
clear
and
I
should
have
said
up
front
that
I'm,
you
know,
don't
intend
to
be
in
imposing
any
ads
opinion
here
or
anything.
This
is
just
me
sitting
in
the
room
listening.
D
Yeah,
so
why
I'm
I
am
so
sensitive
to
not
giving
that
wrong
impression
to
the
Casual
reader?
The
big
big
thing
with
the
way
I
see
it
works
right
because
I,
it's
not
just
the
architecture.
I
have
my
mind
image
of
what's
exactly
going
to
happen.
What's
going
to
happen?
Is
you've
got
this?
Probably
this
PC
I
mean
the
distributed,
is
in
the
architecture
slow
pointed,
but
I,
don't
I,
don't
have
any
clue
of
how
to
make
that
work.
Distributedly!
D
So
so
say
you
have
this
controller
and
the
controller
will
observe
the
links
every
hub
through
a
periodic
statistical
knowledge.
So
every
hour,
what's
the
ratio
of
delivery
on
that
Hub?
So
so
it's
going
to
get
statistical
values
because
it
cannot
get
so
many
full
pictures
of
exactly
what's
going
on.
So
it
will
have
statistical
views
of
all
those
links
right
and
it
will
be
asked
to
build
something
which
allows
every
packet
from
a
to
Rich
B
like
2.5
times
every
packet
should
reach
should
arrive
2.5
times
average.
D
Something
like
that,
and
you
will
say:
Okay
2.5
if
I
just
build
three
parallel
paths
with
all
the
laws
and
the
statistics
I
get
1.1,
which
means
that
some
packets
would
be
lost,
not
good
enough.
So
let
me
build
more
of
those
things
now
if
I
build
too
much
yeah
Spectrum
energy
blood.
That's
why
we're
here?
D
Okay,
but
the
PSC
cannot
be
that
smart
as
to
decide
if
exactly
on
each
packet,
which
hops
are
going
to
be
followed,
so
I
need
to
hint
and
that's
that's
the
Orient
thing
in
UDA
I
need
to
hint
the
PSC.
If
you
see
this,
you
do
that
and
each
of
those
that
is
a
sub
topology
of
the
overall
thing.
So
the
overall
thing
is
all
those
extra
links
that
the
PC
has
decided
may
be
used
from
A
to
B.
D
The
PSC
doesn't
just
give
that
it
gives
a
set
of
T
path,
each
one
mapping
a
certain
condition
of
the
network
and
I'm
willing
to
call
this
thing:
t-path
and
I'm.
Sorry
I
called
that
sub
truck,
and
so
now
I
have
a
clear
usage
of
the
term
t-path.
It
is
this
subset
now,
oam
doesn't
care
only
on
that
subset
because
we
are
doing
out
of
Bando
am
as
well
to
see
if
there
is
another
path
in
that
bigger
thing
that
we
could
use
better
than
what
we
are
using
currently
Okay.
D
D
Actually,
when
I'm
going
to
receive
from
the
pce,
all
those
different
cases,
I
will
instantiate
a
service
for
each
of
them,
because
in
runtime,
I
will
just
use
one
of
those
services,
but
in
my
code
I
will
have
one
instance.
I
will
just
say
flip
that
one
on
flip
that
one
off
as
the
immediate
reaction,
so
each
of
the
service,
the
blue,
the
red,
the
green,
the
whatever
they
will,
they
will
be
activated
and
deactivated
dynamically
based
on
OEM.
D
That's
exactly
what's
going
to
happen
so,
for
each
of
those
Services
I
will
have
a
t
path.
So
what
I'm
gonna
do
is
activate
the
t-path
deactivate
that
t-path
that's
exactly
what's
going
to
happen
and
then
again,
I
need
to
represent
the
set
of
all
those
T
path
and
what
I
call
coalitions
is
When
You
observe
all
of
them
from
above,
and
if
you
have
twice
the
same
link,
it
looks
like
one,
but
at
the
end
you
observe
the
set
of
all
that
that's
the
track.
D
H
H
Er,
since
today,
the
first
time
I
came
over
because
and
then
the
first
time
I
saw
your
terminology
track.
It
immediately
comes
to
my
mind
in
the
tracking
area
that
being
used,
since
this
is
the
raw
the
radio.
The
world
is
hard,
so
that
has
a
different
meaning
to
tracking
area.
H
For
the
channel
you're
talking
about
the
drv
and
then
I
continue
reading,
and
you
say:
okay,
you
statistically
and
combinational
comment
combinatory
aggregation
of
the
blah
blah
everything,
and
then
at
that
moment,
okay!
Well,
you
are
talking
about
the
PSE
and
going
to
determine
a
path
so
with
for
me.
I
have
both
background
when
I
read
it,
it's
the
first
impression
I
got
is
a
little
bit
confused.
H
H
I
fully
understand,
but
you
know
for
I
have
background
both
and
then
the
first
time
I
got
coffee.
D
That's
wonderful
because
we
are
lacking
this
feedback
because
everybody
here
kind
of
understands
what
we've
been
working
on
for
a
long
time
and
when
I
said
that
they
made
a
rework
to
to
try
to
make
the
new
reader's
life
easier.
Seems
that
I'm
not
through
yet.
So.
If
could
you
write
down
what
you
just
said
and
maybe
suggest
you
know
how
things
could
be
reordered
or
where,
where
you
you
add
those
times
of
confusion,
yeah.
H
Well,
I:
try
to
ask
you
one
particular
question:
we're
gonna
talk
about
the
end
to
end.
Are
you
only
talk
about
up
to
the
base
station
and
Beyond
on
the
layer?
3
part,
oh
you're,
talking
things
you're
talking
this
at
the
raw,
the
radio,
the
well
the
wireless
part.
Are
you
also
mentioning
from
the
UE
all
the
way
to
the
server
the
whole
path.
D
Okay,
so
so
there
are
two
things
here:
yeah
one
of
them
is.
There
are
two
use
cases
in
a
row,
because
that's
the
use
cases
that
were
presented
to
us
in
life
and
there
might
be
a
lot
more
to
wireless
than
those
two
use
cases
and-
and
you
talk
about
UE
and
then
base
station.
So
that's
use
case
one
when
row
protects
the
access
only
the
access,
but
it
observes
the
end-to-end
path.
So
basically,
you
get
feedback.
D
From
the
other
hand
whether
the
packet
was
received,
but
the
only
thing
you
really
act
on
is
the
first
top.
So
if,
if
the
server
doesn't
get
your
data
over
Wireless
Wi-Fi
you'll
go
to
5G
regardless
if
the
problem
is
really
with
the
Wi-Fi
or
if
it
is
somewhere
else
right.
So
that's
when
the
oam
is
layer,
three
and
and
the
selection.
The
action
that
you
can
take
is
just
on
the
first
half,
because
the
rest,
you
don't
control
anyway.
D
So
so
that's
that's
half
of
of
the
discussion,
and
so
the
other
Alpha
is
what
I
lost
myself.
What
was
the
core
of
your
question?
Please
I
need.
H
To
cover
the
thing
here
got
the
the
PSE
okay,
so
from
the
IP
and
side
I
understand
it's
like
like
some
place
to
calculate
on
the
layer,
three,
the
tunnel,
but
the
thing
here.
If
you
look
at
the
radio
architecture,
especially
5G,
the
layer
3
is
you
can
consider
you
can
consider
under
the
genob
side
and
then
so
for
your
internal
pass.
Yes,.
D
Well,
even
on
the
the
first
use
case,
the
PC
will
tell
you
about
the
statistics
of
the
links,
but
it
does
not
compute
the
end-to-end
path
that
we
are
talking
about.
It
just
says
you
switch
this
interface
or
that
interface
or
that
interface
based
on
observable
qualities
or
whatever.
So
if
you,
if
you
see,
for
instance,
that
your
Wi-Fi
reports
that
the
speed
is
this
or
whatever
you
switch
to
5G
just
an
example,
that's
the
kind
of
reaction
that
the
PC
could
tell
to
the
end
user,
the
user
equipment.
H
Let
me
just
to
make
it
very,
very
clear
here
for
your
statistic:
terminology
used
to
the
track
suppose
this
is
the
radio,
the
5G
scenario
on
the
down
link
from
The
genome
B.
Here,
when
the
Downing
traffic
come
onto
the
genome,
B,
it's
going
to
select
based
on
the
sum
of
5qi,
it's
going
to
select
even
the
drbs,
the
data
radio
barrier;
okay,
so
that
one
is
going
to
be
well
depending
on
the
qls
perform
without
being
start
that'll,
be
the
selection,
okay.
Well,
it's
not
random.
H
D
The
statistics
in
the
access
case,
the
statistics
are
obtained
through
the
end-to-end
oam
and
whatever
did
those
two
layer,
2
does
is
not
observed.
Normally,
it's
not
observable.
We
don't
really
have
access
to
what
layer
2
does
so.
The
the
that
is
the
use
case
where
the
oam
is
layer,
three
from
the
client
to
the
server,
and
if
that
goes
bad,
we
just
switch
the
radio
okay,
but
we
don't
know
what
better
we
don't
even
care
if
it's
Wi-Fi
or
5G,
okay.
H
B
So
I'm
in
the
Q
next
Rick
Taylor,
with
chair
hat
off
really
and
Lou,
it's
kind
of
a
follow-up
question
to
a
statement
you
made
earlier
and
I'm
glad
you're
in
the
queue
behind
me
going
back
to
the
track
versus
path,
question.
B
This
is
we've
got
to
get
to
the
bottom
of
this.
So
actually
quick
comment
to
concerning
John's
statement,
I'm
I'm
in
complete
agreement
that-
and
this
goes
back
to
my
idea-
that
that
we
are
a
sibling
group
of
debt
Nets,
so
making
sure
that
if
you
come
from
a
debt
net
background,
you
can
immediately
understand
because
there
is
commonality
of
terminology.
But
if
specifics
need
to
be
pulled
out
in
order
to
make
a
coherent
argument,
that's
absolutely
fine
and
should
be
front
and
center
so
that
it
those
differences,
can
be
noted.
B
So
my
question
to
Lou
actually
was
given
you
have
layers
of
service
protection
functions
and
effectively
a
service
protection
function
does
something
to
paths
or
hops.
You
know
you've
earned
it's
going
to
be
Parts
I,
think
where
a
service
protection
function,
which
does
some
sort
of
duplication
and
and
eliminate
replication
and
elimination
pairing
so
that
service
function
is
creating.
What
do
you
call
a
link
of
service
functions
from
end
to
end
if
a
service
function
does
something
to
makes
choices
between
paths?
B
G
G
It's
still
the
end-to-end
service,
because
when
you're
operating
at
a
layer
you
see
what
layer
you're
operating
at
so
switching
you
know
the
across
the
horizontal
providing
service
up
to
the
vertical,
where
you
know
when
you're
going
up
vertically
that's
to
the
application
and
when
you're
going
down
you're
going
to
get
receiving
service
from
the
network
usually
represented
as
interface,
but
you
know
there's
lots
of
ways
to
implement
that.
It's
not
standardized.
G
E
B
B
B
H
B
B
G
D
B
Yeah
yeah
sorry,
when
you're
done
I'll
make
my
point:
okay,
so
oh
God,
you
just
interrupted
me
so,
and
this
is
kind
of
a
question
to
Pascal.
Now,
if
you
call
it
a
end-to-end
service
at
a
particular
layer,
is
that
a
pipe
or
is
that
still
a
flow
because
I
agree
with
you
on
pipes?
I
actually
do
agree
with
you
on
pipes
and
it's
I
know.
That's
a
TechNet
consideration,
not
particularly
a
raw
one.
I
think
there's
a
big
difference.
D
We
we
spent
some
time
yesterday
on
that
picture.
I'm,
sorry
for
the
rest
of
the
group,
but
I
I
presented
it
at
the
last
iatf
I'm,
it's
very
hard
to
build
it's
a
very
simple
picture,
but
it's
very
hard
to
build.
What
it's
trying
to
say
is
that
the
PSC
operates
outside
of
the
flow
of
the
packets.
Remember
it's
something
which
will
switch
from
Service
Blue
to
service
Red,
because
the
radio
conditions
somewhere
have
changed
and
you
could
see
it
as
a
fork
of
packet,
but
no
because
it's
not
in
the
data
plane.
D
That's
my
answer
to
you
and
Lou!
That's
why
I
say?
No!
It's
because
this
thing
happens
asynchronously
to
the
traffic.
This
happens
in
this
oam
plane.
If
we
really
want
to
to
have
an
oem
plane
where
we
will
switch
the
service,
but
the
next
packet
will
only
see
one
service
on
the
path
of
the
packet.
There
won't
be
a
service
selection
that
will
and
then
a
sub
service
selection.
So.
G
So
it
you
really
I,
think
you're,
saying
you're,
not
switching
between
Services
you're,
switching
between
forwarding
paths
in
the
recursive
architecture,
I
I
think
it
would
be
super
helpful
to
draw
the
same
picture
for
a
wired
Network.
We
have
a
protection
switching
function
in
defined
for
detnet.
We
have
one.
The
architecture
allows
any
so
we
could
have
a
one
for
n
service
defined
in
bednet.
We
don't.
We
only
have
a
one
one
plus
one,
but
that
is
a
type
of
switching
function.
That's
a
type
of
PSE!
G
B
G
In
line
and
I
hear
you
yeah
so
in
detnet,
it's
a
very
rigidly
controlled
PSE
controlled
by
a
Yang
model
and
it's
expected
to
be
driven
by
a
provisioning
system,
but
there's
no
reason
that
couldn't
be
driven
by
you
know.
Take
your
AIML
system
that
automatically
changes
it
based
on
what
it
sees
is.
You
know
the
weather
pattern
of
the
network.
D
G
Think
model
fully
supports
what
we've
talked
about.
I
think
that
the
architecture
defines
different
ill
allows
for
different
protection
mechanisms,
we've
only
defied
one,
we've
always
talked
about.
We
may
be
doing
others,
we
identify
Network
coding,
as
maybe
the
second,
but
it's
a
maybe
second
we're
contribution
driven.
So
if
someone
came
with
something
new,
you
know
we
could
talk
about
that,
but
I
think
from
an
architectural
standpoint,
you're,
not
adding
something
fundamentally
new
you're,
adding
a
new
protection,
switching
which
already
fits
into
the
demo.
D
In
the
same
plane,
because
we
are
doing
it,
that's
that's
why
that
picture
is
important.
What
you're
saying
is
all
true,
but
at
the
end
of
the
day,
how
can
we
Express
in
this
document
that
we
are
not
folding
plane?
We
are
not
doing
steps
in
the
lay
going
through
layers
for
the
packet
operation.
There
is
only
one
layer
or
this
is
a
service
well
and
then
and
then
going
down
to
Fourth,
but
we
never
go
through
the
control
layer
for
the
packet
processing
right
and
we
need
to
expose
that
to
the
reader.
G
Yeah
so
I
don't
think
we're
arguing
about
the
net
effect
and
the
behavior
on
The
Wire
right
we're
talking
about
we're,
arguing
about
how
to
represent
it
on
paper,
so
it's
more
accessible
to
the
reader
as
well
as
allows
us
to
reuse
existing
mechanisms
documentation.
So
we
can
reference
rather
than
have
to
define
a
whole
new
set
of
things.
That's
what
we're
really
talking
about
right,
I'm
suggesting
take
this
picture,
draw
it
for
Wired.
D
B
G
To
because
I
I
think
they're
separable,
but
you
know
I
I
he's
the
one
driving
slides
he's
the
one
showing
a
slide
so
I'm
responding
to
the
slide.
The.
D
D
But
the
service
access
point
exists
in
that
net.
It's
in
the
model
Etc
and
that's
neat,
that's
perfect!
Now
this
we
have
this
new
user,
which
is
effectively
not
the
one
we
expected
when
we
wrote
initially
the
architecture,
but
the
architecture
is
open
to
having
it
and
now
row
is
introducing
that
new
consumer
to
this
providers
app.
G
So
you're,
actually
taking
me
to
the
point
that
I
wanted
to
come
up
and
actually
make
my
point,
which
is
we
have
an
art.
It's
true.
We
have
an
architecture
document.
You
wrote
A
lot
of
it,
but
what
came
through
the
working
group
is
what
the
ietf
agreed
to
and
I've
had
this
set
to
me
by
the
way
for
documents
that
I've
written.
Is
it
doesn't
matter
what
I
wanted
it
to
say?
It
says
what
it
says
yeah.
So
we
have
to
accept
that
that's
what
the
ietf
standardized
yeah.
G
D
G
G
G
Is
I
I
agree
with
you
completely,
and
that
goes
to
the
next
point,
which
is
different
protection
functions.
From
my
my
perspective
as
a
reader
I,
don't
think
I
was
an
author
on
that.
I
did
comment
a
lot,
but
I
don't
think
I
was
an
author.
I
saw
the
ability
to
do
any
number
of
protection
functions.
You
should
take
it.
G
D
D
D
When
you
could
it's
a
control
domain
service,
if
you
like,
but
it's
not
the
services
that
we
have
today,
because
the
only
service
that
we
Define
and
where
we
placed
it,
we
placed
it
in
the
forwarding
plate
and
because
this
thing
is
in
the
control,
oam
or
whatever
plane
you
want,
then
it
doesn't
fit
unless
we
take
our
definition
of
service
and
extend
it
to
control
plane
activity.
So
it's
that's.
That's
that's
the
thing,
but
it
is.
It
is
what
you
said.
It
is
a
Protection
Service.
It
just
doesn't
live
in
this.
J
I
I
think
that,
like
you've
spent
a
lot
of
time
on
this
and
and
but
you've
you've
you've
taken
a
Step
Beyond.
What
what
some
of
us
can
keep
up
with
with
a
new
terminology,
I
would
suggest
like
lose
the
term
track
and
describe
it
in
like
you,
like
you,
have
the
definition
in
terms
of
a
path.
You
know
whether
it's
an
aggregated
path
or
a
collection
of
paths.
J
And
let
the
working
group
decide
if
there's
a
term
that
we
need
to
describe
that
in
the
end
and
and
for
a
couple
of
the
new
terminologies.
Do
the
same
thing
and
that
would
add
Clarity
to
the
document
for
me,
because
a
conceptually
I'm
not
driving
with
it
when
I
see
a
track
and
I'm
sure
a
few
others
have
the
same
problem.
But
if
we
use
all
the
same
terminology
that
we've
used
before-
and
maybe
it's
a
big
bunch
of
words
that
we
have
to
use
for
this
version
of
the
draft.
J
And
we
would
decide
that
we
want
a
term
that
just
describes
that
then
I
think
we
would.
We
would
reach
a
consensus
on
this
faster.
B
That's
very
wise
advice.
Thank
you,
I'm!
Next,
in
the
queue
so
I'm
I'm,
gonna
sort
of
go
back
to
the
argument,
not
the
the
polite
discussion
between
Lou
and
Pascal.
Here,
the
splitting
of
the
document
I
think
was
made
to
separate
the
pario
service
function
from
the
pse's
role
in
the
architecture,
because
I
agree
with
Pascal.
Here
they
are
two
separate
things
and
I.
Think
where
I
disagree
with
Lou
is
I.
Believe
he's
telling
me
that
they're
just
one
thing:
I,
that's
what
I'm
picking
up
Lou
I
I!
B
Welcome
to
my
world
yeah
we'll
consult
the
minutes
fundamentally,
there's
there's
the
beautifully
young
modeled
detnet
service,
a
Control
Function
or
service
control
model
and
as
I
understand
it,
the
PSE
can
be
a
user
of
that
in
order
to
change
service
function
based
on
some
management
and
monitoring
information
that
flows
across
the
control
plane
policy,
whatever
that
is
a
unique
decision
or
a
unique
part
of
the
raw
architecture,
because
Pascal
and
the
working
group
by
extension.
B
Across
whether
we
call
it
a
path
to
track
a
stream
of
flow
I,
don't
care
within
the
debt
net
service,
sub
layers
and
splitting
those
two
documents
may
help
clarify
that.
Does
anyone
disagree
with
me
on
that
Pascal
are.
G
Burger
sorry
I'm
not
in
queue
I
just
want
fast
enough
with
the
app
your
words.
Just
maybe
remember
that
we
spent
a
huge
amount
of
time
going
over
these
drawings
of
protection
models.
G
As
we
were
developing
the
Yang
model
to
make
sure
we
were
appropriately
modeling
what
the
data
plane
could
do
and
Don
was
involved
in
that
I
was
seeing
if
Don
had
a
pointer
to
any
of
those,
but
maybe
we
can
grab
those
and
give
it
to
the
working
group
here
and
then
we
can
start
extending
them
to
draw
the
pictures
of
what
we
mean,
and
then
we
can
say
we're
pointing
right
here,
and
this
is
the
new
thing
or
we're
pointing
here,
and
this
is
the
existing
thing.
I.
B
D
Since
Lou,
you
brought
up
the
young
model,
first
things
down
for
the
ug
front,
and
you
know
that
that's
huge
I
mean
if
that
contributed
on
the
meaning
list.
But
that's
that's
a
huge
thing.
Now
a
model
gives
you
what
you
can
use
and
not
how
you
can
use
it.
That
is
the
architecture
which
will
tell
you
what
you
can
do
and
what
you
cannot
do.
D
For
instance,
the
model
of
a
car
will
tell
you
that
if
you
press
the
pedal
it's
going
to
go
faster,
it
doesn't
tell
you
whether
or
not
you
go
when
the
light
is
red
or
green.
That's
the
architecture,
and
so
because
the
young
model
could
be
is
wide
enough.
To
do
stuff
doesn't
mean
that
we
we
do
all
those
stuff
that
the
young
model
allows,
but
we
wanted
the
young
model
to
allow
everything
you
want
to
do,
but
it
cannot
necessarily
be
limitating
to
the
to
to
to
what
we
allow
so
so
did.
D
B
Pascal
can
I
reply
to
that.
I
I
think
you're,
slightly
meshing,
two
concepts
together
there.
What
I
meant
to
say
was
in
an
architecture
document
you
can
say
there
is
this
well-defined
control
path
between
the
service
control
functions
down
to
the
services
themselves?
We
are
very
lucky
that
there
is
a
good
quality
Yang
model
that
can
describe
that
and
in
the
architecture
we
can
say
we
are
going
to
use
this.
It
exists
and
people
familiar
with
the
prior
work
will
go.
Aha,
that
is
perfect.
D
So
so
for
me,
it's
fine
that
if
the
model
allows
that
it
doesn't
mean
that
we
use
that
what
what
we
use
is
what
the
architecture
says
we
use
right.
Even
if
the
model
is
I
mean
the
model
has
to
represent
everything
we
could
do
and,
and
the
architecture
says
everything
we
do.
Hopefully
the
young
model
is
a
superset
of
of
that
yeah
and.
B
E
D
B
So
Don,
are
you
still
in
the
queue
or
have
you?
Okay,
I'll
bring
you
out
of
the
queue
Stu
you're
in
the
queue.
K
Yeah
Stu
card
and
I'm,
not
sure
which
of
my
affiliations,
I,
want
to
tag
on
this
so
I'd
just
like
to
give
a
perspective
of
a
noob
I
have
attempted
to
parse
the
net
documents
and
I
have
attempted
to
parse
the
raw
documents
and,
despite
what
is
obviously
an
enormous
amount
of
work
by
both
groups
and
an
enormous
amount
of
intelligence
underneath
that
work
I
find
neither
of
them
particularly
accessible
to
a
newbie
such
as
myself.
K
Okay,
they
appear
to
be
things
that
probably
can
be
read
easily
by
someone
who
is
involved
in
the
work
from
the
first
and
can
trace
its
Evolution
and
not
so
easily
read
by
someone
who
was
not
involved
in
the
work
from
the
first
and
is
coming
in
late
and
trying
to
wrap
his
head
around
this
okay.
So
from
that
perspective,
I
have
some
sympathy
for
what
Lou
is
saying
about,
trying
to
make
them
more
accessible.
K
On
the
other
hand,
I
I
do
like
precision
and
by
the
way
we
already
in
ietf
and
even
in
irtf
abuse
the
word
path.
Okay,
if
you
go
to
the
mathematical
graph
Theory,
we
have
walks
Trails
traces
circuits,
Etc,
et
cetera,
Etc
right.
E
K
So
so,
to
draw
an
example
from
a
different
area
with
an
ietf
locator
identifier
split,
you
know,
there's
hip,
there's
ilnp,
there's
lisp
and
there's
probably
others.
K
Okay
and
each
introduces
its
own
set
of
nomenclature,
which,
on
the
one
hand,
is
annoying,
but
on
the
other
hand,
not
only
are
they
potentially
attempting
to
address
a
different
set
of
requirements
even
to
the
extent
that
they
address
the
same
set
of
requirements
they
may
partition
and
allocate
those
requirements
to
different
subsystems
within
the
overall
system
architecture
and
the
use
of
the
same
term
to
refer
to
a
box
that,
over
here
on
the
left,
includes
a
b
and
c
and
over
here
on
the
right
includes,
b,
c
and
d
I
think
ultimately
can
be
more
confusing
than
attempting
to
reuse
terminology.
K
K
D
You
you
Nest
service
layers.
When
you
do
compound
flows,
you
do
that,
but
the
operation
the
packet
sees
that
operation.
The
packet
fails.
Every
of
those
operations
the
packet
doesn't
fill
the
PSE.
D
The
PSC
has
been
done
under
the
carpet
while
the
packet
was
not
looking,
okay
and
and
that's
the
reason
why
it's
not
falling
plane
and
that's
the
reason
why
I
don't
call
it
service.
Now,
if
we
decide
that
we
have
control
planes,
reliability,
Services
as
well
as
data
plane,
reliability,
Services
all
of
us,
then
we
agree
right.
So
so
my
question
is
really
what
we
Define
service
layers.
Oh
it's
all
right,
but
but
I
want
to
be
honest
to
this.
Drawing
and
then
again
it
looks
simple.
I
mean
loose.
D
L
Hi
this
is
joseon
Pascal.
Thank
you
for
your
explanation.
Just
because,
as
the
quarter
of
the
then
the
yellow
mode
or
I
just
want
to
ask
some
clarification,
questions
to
make
me
more
clear.
What
do
you
try
to
describe
about
the
PSE
I
just
read
through
the
document
about
the
concept
and
the
description?
L
Does
it
mean
that
is
a
control
Loop
it
does
it
mean
that
the
PSE
itself
includes
the
oam
process
and
the
controller
plan
process,
and
also
the
maybe
something
like
forward
implant
or
service
layer,
often
that
no
it's.
D
The
combination,
no,
the
PSC
is
this
little
box
on
the
top
left
of
that
screen
and
the
PSC
talks
to
somebody
who
represents
what
who
I
am
finds,
which
I
call
the
oam
supervisor
or
something
just
that
thing
which
sits
in
the
control
plane
that
observes
the
oim
activity
and
get
reports
from
OEM
activities.
So
it
knows
something
happened
on
the
other
side
of
the
network.
Now
PSC
needs
to
take
action,
so
PSE
is
just
this
and
it's
written.
You
know.
If
you
look
at
the
again,
that's
why
I
like
this
UDA
reference
model?
D
L
L
Uh-Huh
and
it
does
the
decision
kind
of
like
pce,
yes,
okay,
I
I.
D
Try
to
understand
the
PC
is
the
Orient
and
Orient
is
the
only
letter
that
is
intelligent
in
the
UDA
Loop
Orient
is
the
one
that
has
some
historical
understanding,
for
instance
the
statistics
that
we
bring
to
the
PC.
So
it
knows
this
link
is
feeble.
This
link
is
solid,
so
when
I
construct,
those
alternate
paths,
I
will
rely
more
on
this
one
than
on
that
one,
and
expect
that
this
one
fails,
in
which
case
I've
prepared
discipline
B
for
you,
which
is
service,
Red,.
L
F
L
D
B
No,
no,
no!
No!
No,
not
now,
but
over
over
time.
If
you
find
areas
that
you
find
difficult
to
to
make
that
thinking,
coherent
and
understandable
for
yourself
in
the
document,
please
record
the
pieces
and
get
them
on
the
mailing
list,
because
a
document
that
isn't
as
easy
as
possible,
given
the
technology
to
help
people,
do
that
thinking
and
get
correct
needs
to
be
better.
So
you
are
a
perfect
reviewer,
because
you
are
trying
to
understand
it
for
a
reason
rather
than
just
for
interest.
So.
D
F
H
H
No,
no
and
now
try
to
downplay,
but
they
try
to
just
because
I
I
do
both
work
on
the
awareness
and
the
Wireline,
especially
here
you're
talking
about
like
the
past
the
track
and
then
PSE
here
so
for
and
also
you
mentioned,
for
your
awareness
or
I,
you
I
think
you
mentioned
about
something
like
some
Wi-Fi
switch
over
to
5G.
You
and
one
moment
you
mentioned,
but
I
tried
to
map
the
5G
architecture
through
the
three
GPB
access
and
the
non-3
GPU
access,
including
the
Wi-Fi,
and
so
here
your
path.
H
Let
me
make
the
clear,
is
your
path,
the
the
DNA,
the
data
network
has
beyond
the
n6
it
all.
It
will
pass,
including
the
the
back
hall
path
between
the
genome,
B
and
the
UPF,
or,
if
your
path
end
to
end
from
UE
up
through
the
through
the
back
hole
through
the
5gc
and
then
all
the
way
beyond
the
n6,
including
DNN,
so
just
try
to
get
it
clear.
Thank
you.
E
D
Okay,
so
what
you
see
here
is
probably
the
drawing
that
corresponds
the
best
in
the
document
to
to
the
situation
we
we
are
talking
about.
So
the
situation
we're
talking
about
is
we
don't
have
an
end-to-end
service.
We
don't
have
end-to-end
visibility
on
the
path.
The
only
thing
we
see
is
what
we
see
at
layer
2
and
what
we
see
at
layer,
3.
D
and
in
the
case
of
5G,
because
now
you're
modeling,
the
5G
Network
as
a
deterministic
TSN
type
of
network.
We
could
say
that
the
debt
net
forwarding
extend
to
the
G
note
B,
basically
Unit
B,
whatever
it
is
so
so
this
this
relay
node
would
be
the
end
of
TSN
quote-unquote,
the
the
the
we
the
5G
Network.
B
Pascal,
can
you
come
back
to
the
mic
and
see
what
I
think
was
amazing,
yeah
the
little
red
one.
E
D
D
We
don't
know
how
you
build
the
network
behind
it
and
even
TSN
doesn't
see
it
because
TSN
at
least
talks
to
to
to
your
whole
5G
Network
as
one
switch,
but
we
don't
see
what's
what's
inside,
we
don't
have
visibility,
so
the
bottom
line
is
we
have
layer,
three
oam,
so
we
can
observe
when
we
send
something
whether
it's
received,
but
we
don't
know
what's
going
on
here,
because
it's
completely
opaque
is
the
internet.
It
can
be
anything,
and
so
we
we
cannot
take
actions
that
happen
here
like
inside
your
core.
D
We
don't
see
it.
The
only
thing
we
see
is
when
we
send
a
packet.
We
send
away
packets
over
this
interface,
whether
they
are
received
or
not
so
cdoam
activity
in
my
phone
as
being
oh
I've
got
those
two
or
three
interface
and
my
phone
one
can
be
3gpp
that
can
be
one
or
two
Wi-Fi
that
can
be
Bluetooth,
whatever
I'll
get
those
interfaces
and
I'm
sending
away
in
packets
on
all
of
those
out
of
band.
D
Oh
I
am
packet
and
all
of
those
interfaces
and
I'm
trying
to
measure
if
some
of
them
reach
my
server
reach
the
other
hand
and
based
on
that
I
can
make
decision
of
whether
I'm
using
5G,
Wi-Fi
or
whatever
else
from
the
abstraction
that
we
live
on.
We
don't
even
care
what
type
of
radio
that
is.
It's
just
access
a
accessb
access,
C
and
if
I
send
a
packet
on
an
xsa
xsb
access,
C
I'm.
Getting
this
sort
of
reliability
should
I
use,
two
of
them,
which
two
should
I
use.
D
H
D
D
B
Chair
hat
on
later,
once
we
get
the
architecture
and
these
core
documents
done
having.
There
are
linked
layer
protocols
for
for
some
types
of
radio
having
an
ability
to
understand
better
to
get
better
oam
from
those
lower
black
boxes,
so
to
make
them
gray
boxes
in
some
way.
That
would
be
very
interesting
and
can
make
can
provide
more
information
on
the
OEM
flow
so
that
we
can
make
smarter
decisions.
B
B
So
yeah
thank
you.
Lou
you're
up.
J
J
D
J
D
Over
the
networks
are
nested,
so
you
don't,
they
will
all
depend
on
what's
going
on
below
yeah,
so
so
the
the
the
intelligence
behind
all
this
is
the
pce.
D
But
if
you
do
this,
recursively
I
mean
each
each
of
the
nested
levels
will
have
to
offer
an
abstract
reliability
level
between
Ingress
and
egress,
and
if
that's
used
as
one
hub
for
the
inner
flow
well
the
goal.
The
goal
is
basically
this
right:
you've
got
in
and
out
and
for
one
end
you
want
2.5
out
something
like
that,
just
to
ensure
that
your
loss
is
10,
minus
five
or
ten
minus
whatever.
So
you
you
should
average
to
the
drive.
D
It's
just
the
number
I'm
giving
now
you
should
say
a
I'm
going
to
do
compound
flow
and
between
here
and
here,
I
have
yet
another
track.
Well,
this
other
track
will
be
abstracted
as
something
which
gives
me
2.5
and
one.
So
that's
a
PC
doing
this
right.
It
would
build
this
new
inner
track,
get
to
2.5
out
of
it
and
compute
that
this
is
a
virtual
link.
Now
that
has
2.5
and
make
the
bigger
end-to-end
service
for
the
for
the
inner
flow
using
that
2.5
link.
J
D
D
I
I
did
not
visualize
it
with
multiple
services
like
this.
It's
just
blue
is
a
set
of
things,
but
then
you
select
blue
as
a
whole.
I
did
not
sell
I,
don't
think
you
would
select
some
some
little
bit
of
blue
and
some
little
bit
of
yellow
I
thought
that
blue
would
be
a
consistent
set
of
activities
and
and
that
you
know
it's
not
like
you
have
multiple
of
them
which
operate
independently.
It's
just
a
set.
It's
the
parameters.
D
B
E
D
I
agree
with
you
get
crazy,
I,
never
figured.
You
could
run
two
things
in
parallel
on
the
same
packet
without
knowing
what
you're
doing
on
each
side.
That
looks
crazy
to
me,
even
if
we
don't
switch
services
but
okay,
because
if
they
all
have,
if
you
can't
do
two
control
loop
on
the
same
physical
system,
right
I
mean
even
in
the
real
world.
If
you
have
something
that
decides,
if
your
rating
system
goes
up
and
down
another
thing,
that
makes
another
decision
for
the
same
meeting
system.
You
have
a
problem,
no.
J
E
J
D
B
So
I've
locked
the
queue
because
we
have
six
minutes
left
Lou
you
have
just
squeezed
in.
Can
you
keep
it
short?
Please
just.
G
Yeah,
this
is
a
question
to
the
chairs.
Can
we
ask
to
get
a
sense
of
the
room
of
whether
or
not
we
want
to
want
going
back
to
what
Stu
was
saying?
Do
we
want
to
focus
on
new
terminology
that
may
be
more
precise
or
to
try
to
leverage
existing
terminology
and
then
refine
that.
D
I
So
I
would
like
to
point
out
that
this
week,
I've
seen
various
things
that
went
poorly
because
the
show
of
hands
tool
is
a
blunt
instrument.
Unless.
C
B
And
really
we
want
to
censor
the
room.
Does
anybody
object
to
the
to
what
I
think
my
understanding
of
the
room
is
I?
Think
I
heard
strong
consensus
that
people
wanted
a
terminology
section
which
unified
familiar
topics
for
people
from
the
two
re,
the
two
main
inputs
from
the
wireless
world
and
from
the
debt
net
world
and
added
some
richness
to
say
in
this
document?
B
B
So
my
following
question
is
given:
Pascal
is
not
a
native
English
speaker,
and
this
is
very
much
about
and
I'm
going
to
say
us
English,
because
British
English
doesn't
help.
Sometimes
is
there
a
is
there?
Is
there
a
U.S,
English
speaker
who
would
like
to
contribute
text
to
this
and
I'm
not
saying
that
people
where
there
are
great
people
who
speak
English
with
a
second
language
and
so
I'm
not
going
to
discount
them?
But
given
we
are
spending
a
lot
of
effort
on
the
news
and
I'm
using
a
French
word
for
of
of
this
terminology.
B
B
B
That
was
two
cards
saying
he'd,
be
thrilled
to
review
and
help
tweak,
but
he's
he's
the
wrong
man
to
draft
right.
Thank
you.
Pascal
you
can
oh
and
balash
as
well
has
has
raised
his
hands,
which
is
a
fantastic
great.
Thank
you.
Pascal.
Do
you
want
to
add
any
last
points?
Otherwise,.
D
I
I
just
have
one
last
point
is
because
you
said
something
at
some
point,
though.
I
did
not
want
to
interrupt
the
discussion
on
that,
but
he
traced
the
image
that
people
might
misunderstand.
Something
and
that's
something
is
some
Protocols
are
based
on
sessions.
There
is
one
endpoint
which
does
something
the
other
end
points
that
has
to
be
aware
of
it
and
that's
you
know
the
matching
thing.
D
The
pre-off
doesn't
work
like
that,
for
instance,
you
might
have
one
replication
point
that
makes
two
copy
and
later
on
the
right.
That
may
be
another
replication
point
that
makes
two
copies
somewhere.
In
the
end,
there
will
be
a
single
elimination
point
which
will
eliminate
all
the
copies
that
single
elimination
point
doesn't
talk
to
either
of
the
other
two
they
are
independent
and
it
doesn't
care
whether
those
replication
nodes
succeeded.
There
is
no
session.
D
So
at
some
point
there
was
this
image
like.
Oh,
if
I
make
a
decision
on
my
hand,
I
have
to
tell
somebody,
on
the
other
hand
of
the
network,
because
that
impacts
him.
No,
they
must
be
independent
because
there
is
no
time
to
talk
to
him
anyway.
Whatever
we
do,
he
needs
to
absorb,
we
can
never
negotiate
and
and
to
design
different.
The
design
is
is
distributed,
the
operations
can
happen
anywhere
and
the
nodes
which
absorb
the
copies
don't
have
to
know
who
made
those
copies.
B
Perfect,
thank
you
Pascal,
so
we
are
very,
very
close
on
time.
First
I
want
to
apologize
for
the
fact
that
we
missed
the
two-minute
silence
because
we
were
too
excitedly
talking
about
things.
I
can
only
apologize
to
those
who
wanted
to
hold
that
I'm.
So
sorry,
any
other
business.
Does
anyone
have
anything
that
they
wished?
They
had
talked
about
not
concerning
the
raw
architecture,
draft.
B
B
Thank
you
all
for
your
contributions.
Can
you
please
maintain
the
momentum
on
the
list,
so
we
can
just
get
on
and
get
this
raw
draft
of
this
raw
architecture
done,
because.
A
A
I'd
also
like
to
thank
those
of
us
who've
been
heads
down
on
the
the
note
taking
because
it
was
a
really
Dynamic
and
nuanced
discussion.
We
will
no
doubt
have
to
consult
the
recording
multiple
times,
but
thank
you
for
the
group
of
people.
Who've
been
note
taking,
and
please
put
your
name
into
the
notes
that
you
that
you
were
part
of
this.
B
And
one
last
thing:
thank
you
very
much
to
everyone
involved
with
getting
ldax
out
of
the
door
that
was
surprisingly
painful
and
behind
the
scenes
a
lot
of
work
happened.
So
congratulations
for
getting
it
kind
of
out
of
our
hands
and
thank
you
for
everyone
who
helped
make
that
happen,
and
that's
it
thanks
guys
end
of
the
last
session
go:
have
your
exercise,
meetings
and
check
out
of
your
hotel
room,
which
is
on
my
to-do
list.
Otherwise,
thanks
guys.