►
From YouTube: IETF102-ALTO-20180716-0930
Description
ALTO meeting session at IETF102
2018/07/16 0930
https://datatracker.ietf.org/meeting/102/proceedings/
A
C
A
A
A
A
We
have
a
note-taker,
Richard
is
taking
notes
thanks,
Richard
I
do
need
one
more
note
taker.
Could
somebody
volunteer
Lyle
Lyle
is
the
second
note
taker,
so
thank
you
very
much
and
I
think
meet.
Echo
also
has
the
jabber
feed,
so
I
will
keep
an
eye
on
the
geography
in
meet
echo
and
if,
if
somebody
posts
anything
I'll
go
ahead
and
proxy
for
them.
So
this
is
an
agenda.
A
For
today,
as
usual,
we
are
trying
to
give
precedence
to
the
chartered
work
items
and
of
interest
to
us
should
be
the
remaining
issues
in
Aalto
and
the
auto
survey
and
next
steps,
as
you
met
as
you
can
see,
both
Yun
and
I
have
been
trying
to
close
some
of
these
working
group
agenda
items
and
move
them
forward.
I'll
talk
about
that
in
a
few
minutes,
but
the
intent
is
to
try
to
finish.
A
We
have
about
seven
drafts
that
are,
we
need
to
wrap
up,
so
the
intent
is
to
try
to
move
these
seven
drafts
ahead
and
then
figure
out
what
to
do
with
the
working
group
itself.
I've
been
sharing
the
working
group
since
its
inception,
I
think
in
2008.
So
that's
been
10
years
and
you
know
we
can
reach
harder
or
we
can
close.
The
working
group
after
the
current
deliverables
are
met.
That's
a
decision
that
we
will
have
as
a
working
group
and
some
of
that
aspects
we
can
discuss
later
on
today.
A
C
A
All
right,
so,
if
there's
no
changes
in
the
agenda,
let's
continue
so
we've
been
busy
since
ITF
101,
there
have
been
movement
in
a
few
drafts.
Performance
metrics
has
been
through
TSP
our
early
review
right
now
in
this
ITF.
There
is
no
discussion
for
it,
but
I
would
like
the
author's
to
inform
the
working
group
what
the
outcome
was
from
the
TSV.
A
Our
review,
probably
it'll,
be
great
if
you
could
put
it
on
the
mailing
list
on
what
you're
doing
to
address
the
TSP
art
review,
cost
calendar
is
in
publication,
requested
stage,
I
went
through
the
write-up
and
I'm
shepherding
it,
and
it
is
right
now
out
of
the
working
group
with
I
with
is
G
X
Thom
discovery
is
currently
being
prepared.
The
IPR
statements
have
are
in
from
all
of
the
author's
yon
has
prepared
the
proto
write-up,
and
there
is
a
new
revision
that
we
are
waiting
for.
A
Increment
update,
SSE
has
been
working
group
class
called
from
what
I
recall
and
people
can
correct
me
if
I'm
wrong.
There
was
nothing
substantive
with
the
contents
of
the
draft
itself
during
working
group
last
call
review.
There
was
some
editorial
comments
and
some
mid
comments
which,
as
suspect
authors,
are
going
to
address
and
put
a
new
track
forward
as
soon
as
a
new
draft
becomes
available
subject
to
vacation
schedules
between
me
and
yon
will
further
right
up
out
and
also
move
this
to
publication
requested.
A
So
that
leaves
us
with
the
following
four
remaining
tracks:
CD
and
I
request.
Routing
Shawn
is
going
to
be
talking
about
it
today,
so
we
will
see
what
the
status
of
that
is
and
how
to
move
that
ahead.
Half
vector,
Richard
and
Don
will
fill
us
in
on
what
the
status
of
that
is,
and
how
do
we
go
ahead
with
that
I'll
jump
to
unified
crops,
which
is
also
going
to
be
discussed
today
between
Sabine
and
I,
Billy
Richard,
so
we'll
find
out
the
status
of
that
as
well.
A
Performance
metrics,
as
I
said,
is
not
due
to
be
discussed
today,
but
it'll
be
nice
to
have
the
author's
post,
something
on
the
mailing
list.
Giving
us
an
idea
of
what
was
the
outcome
of
psvr
review
and
Jana
and
I
will
then
figure
out
what
to
do
with
respect
to
working
group
last
call
or
if
there
is
any
open
issues.
I
think
that
is
it
from
my
side.
Any
questions
before
we
go
into
the
first
presentation.
D
A
The
one
after
so
one
administrative
note
right
now
at
least
neither
Yan
or
I
are
planning
to
be
in
Bangkok.
So
there
is
not
going
to
be
a
face-to-face
meeting.
We
may
have
a
virtual
meeting
or
if
there
is
enough
mass
in
the
work
being
done
in
these
four
drafts,
and
people
feel
that
we
need
to
have
a
face-to-face
meetings
meeting
at
Bangkok,
then
we
could
always
have
a
temporary
chair
and
try
to
accommodate
a
face-to-face
meeting.
A
E
E
So
in
this
change,
let
go
of
us
is
to
make
our
document
more
clean
and
Sinko
based
on
the
working
group
members
Commons.
So
we
have
five
relative
big
changes
and
six
relatively
small
changes.
The
first
one
is
we
shorten
the
introduction
part
and
we
also
add
an
outline
of
this
document
and
then
the
second
one
is.
We
use
some
pointers
and
example
to
support
the
benefits
of
using
Auto
and
the
third
one
is
based
on
discussions
in
IETF
101.
E
We
update
a
little
CD
and
IFC
IMF
response
schema
and
the
first
one
is
we're
trying
to
add
more
description
based
on
the
auto
network
map
footprint
type,
and
we
described
the
defendants
ease
between
FCI,
Maps
and
network
maps
and
also
we
add
arrow
handling
for
filter,
cdnify
IMF's
and
for
some
consistency,
consideration.
We
also
update
to
media
type,
and
we
also
update
one
data
components
name
and
also
we
extend
the
security
consideration
and
make
some
description
of
the
example
more
clean
at
at
last.
We
also
have
some
small
text
edit.
F
E
The
first
part
is
the
relation
of
our
introduction.
First,
we
change.
We
change
no
less
important
list
into
the
plain
text
and
we
also
remove
unnecessary
and
necessary
descriptions
and
also,
we
add
the
outline
of
the
whole
document
English
section
and
follows
support
of
benefits
of
auto
for
benefit
1.
E
We
add
how
Auto
can
help
New
Zealand
to
select
a
proper
decision
and
for
the
second
benefit,
we
add
pointer
to
a
section
for
of
CDR
m
l-c
IMF
using
auto
network
map,
and
we
also-
and
we
also
emphasize
the
identification
between
the
you
CDN
and
decision
and
also
add
pointer
to
the
security
consideration
and
for
a
restful
design.
We
add
a
pro
pointer
reference
to
a
section
three
and
four
error
handling.
We
added
a
pointer
to
section
five
and
for
futures
map
service.
E
We
use
the
example
of
the
unified
property
map
to
show
that
we
can
directly
use
the
map
service
of
auto
and
for
the
benefits
X.
We
have
some
small
edit
on
the
SSE
part
and
for
the
benefit
aids.
We
also
address.
We
add
a
new
property
for
entities
in
section
6,
which
is
the
CDI
capabilities
and
follow
benefit
night.
We
also
add
a
reference
to
the
path
vector
extensions.
E
So
the
next
update
is,
we
update
a
CDN
IFC
I
met
with
response
schema,
based
on
a
discussion
based
on
discussion
in
IETF
101.
We
use
the
base
advertisement,
object
defined
in
RFC
830
or
a
to
encode,
the
city
RFC
IMF,
and
we
also
keep
the
original
encoding
of
Lopez
advertisement
object,
because
we
we
want
to
remember
them
for
self-contained
purpose
and
also
we
use
an
example.
E
This
example
is,
they
are
two
examples,
but
lay
describe
lesstm
see
capability
object
with
la
Stanford,
OS
la
same
footprint
constraint.
We
want
to
use
this
examples
to
show
that
the
optimization
of
base
advertisement
object
is
possible,
however,
follow
basic
intercity
and
interconnection.
The
specific
specific
of
such
mechanisms
are
outside
of
out
of
the
scope
of
this
document
and
the
next
one
is.
We
update
the
fruit
screen
types
description
of
Auto
network
map.
E
First,
we
indicate
that
the
corresponding
foot
cream
below
is
a
list
of
the
PID
names
and
least
PID
names
are
actually
the
reference
of
the
PIDs
in
a
network
map.
So
a
CD
on
IFC,
I
map
is
depends
on
network
man,
so
the
dependent
abandoned
vehicle
with
a
reference
to
a
related
network
map
must
be
included
in
the
city
on
IFC
imf's.
E
The
next
one
is
about
arrow
handling
of
the
future
CDN
RFC
I
maps.
We
first
emphasize
that
the
response
must
indicate
an
arrow
using
all
ho
protocol
arrow
handling
and
for
and
for
invalid
requests.
We
point
out
three
cases.
The
first
one
is
the
value
of
capability
type
is
now
and
the
second
one
is
the
value
of
capability
value
is
now
another.
Third,
one
is
lockable.
A
capability
value
is
an
inconsistency
with
capability
type.
So
when
this
situation
happens,
the
request
is
invalid
and
the
auto
server
must
return.
E
Next,
one
is
for
consistency,
consideration.
We
update
the
data
components,
theme
of
CDN,
FC
m,
f2,
c,
DF
c
MF,
and
also
we
updates
the
media
type
of
the
query,
unfiltered
city,
FC,
MF
to
application,
auto
CD,
f's,
MF,
chastens
and
and
also
lists
and
also
list
filter
service
and
follow
for
the
extension
of
the
security
consideration.
E
Despite
of
the
five
security
consideration
already
described
in
RFC,
seven
285
I
think
we
must
strengthen
let's
the
isolation
of
the
CDN
RFC
I
maps
for
different
usage
is
important
because
otherwise
it
may
lead
usage
intuitive
to
redirect
requests
to
around
this
area,
which
cannot
serve
this
request,
and
a
potential
solution
is
that
we
cannot
include
all
CD
f's
I
met
in
only
one
ard,
so
the
next
next
one
is.
We
have
some
relation
of
the
examples.
E
Basically,
we
add
more
explanation
for
incremental
update
in
sections
three
point:
seven
point
three,
and
also
we
have
more
description
on
how
to
fill
their
own
capabilities
in
section
five
point:
seven
point
two
and
next
part
is
about
the
discussion
and
we
are
thinking
about
what
other
auto
features
may
benefit
to
see.
Dna,
for
example,
first
calendar
or
pass
vector.
E
Can
it
help
in
CD
analyzed
situations
and
the
next
one
is
since
CDN
has
four
interface
if
FC
I
use
auto,
can
other
interface
also
use
auto
as
well
or
at
least
how
all
four
together
they
can
fit
together?
So
next
one
is
next
steps.
I
think
we
want
to
get
more
comments
from
working
group.
Members
and
I
want
to
know
what
was
other
opinions
or
suggestions
phone
from
you.
Oh
thanks.
D
D
Okay,
so
I
just
want
to
relay
one
one
comments
and
send
about
you,
ve
done
and
from
from
Tencent
and
his
comment
I'm
for
inspire.
There
was
the
late.
So
therefore
he
couldn't
really
land
here.
That's
why
I'm
really
in
his
comments
and
he's
a
mean
common,
was
next
genera
paradigm
called
capabilities
and
with
entities
for
example.
He
really
mentioning
in
the
context
of
IG-
and
you
have
all
the
capabilities
you
have
in
and
then
you
have
the
basically
you
have
devices
or
any
points
will
have
capabilities
and
his
common
would
be
a
CDI.
D
Somehow
you
look
at
another.
Bigger
picture
wooden
looks
like
essentially
is
deeply
encoding
of
capabilities
and
for
you
to
capability
which
footprints
would
have
that
capabilities.
So
his
comment,
of
course
he's
not.
I
don't
believe
he's
objecting
to
proceed.
He
asking
is,
can
clarify
a
kansas
city
and
I,
say
hello,
serve
as
template
or
general
design
for
general
network
capabilities
exposure,
which
is
reward.
Typically,
when
we
design
we
always
do
from
entities
to
properties,
and
here's
encoding
is
from
cue
abilities
to
footprints
to
end
to
end
it
this
so
I.
E
D
Sure
one
example
problem
you
can
easier.
Can
you
show
an
example
of
your
CBN
I?
Carry
yeah
then
might
be
good
yeah.
So
basically
is
encoding.
Is
capabilities
yeah
and
your
capability
type,
for
example,
now
FCI
delivery
protocol
and
value?
What
kind
of
this
capable
you
have
exam
here
it
to
be
1.1
and
they
talk
about
indicates
which
one
will
have
these
capabilities.
His
comment
would
be
okay,
and
what,
if
you
don't,
have
FC
adopt
your
protocol?
D
What
if
you
have,
for
example,
5g-
and
this
is
whatever
your
ability,
this
support
is,
can
our
values
maple
print
is
a
possible
to
really
do
that
design?
If
it
had
possible
design,
then
why
do
you?
Name
is
city
and
IFC
IMAP
current
should
be
generic
for
your
capability
map.
I.
Think
that's
the
question.
E
A
So
I
think
we
are
two
minutes
over.
So
I
will
ask
you
to
keep
your
comments
brief
and
probably
when
you
inve
is
here,
maybe
you
can
talk
with
him
and
try
to
see
how
to
solve
yeah.
It
looks
more
of
a
semantic
issue.
More
than
anything,
that's
substantive,
so
I
think
you
can
sort
it
out
with
them.
Okay,
thank
you.
Jensen.
G
G
E
H
Like
works
question
actually
for
the
chair,
so
if
we
take
this
submission
and
generalize
it
right
at
that
point,
understanding
that
we
would
have
very
strong
examples
supporting
CD
and
IFC
I
does
that
is
that
a
new
submission
or
what
would
happen
here
because,
clearly
going
to
the
is
GE,
you
know
with
naming
and
everything
else
very
targeted
right
and
then
kind
of
abstracting
it
up
right.
What's
the
best
way
to
go
for
if
that's
genericized,
so.
A
I
think
this
work
has
had
its
genesis
in
another
working
group
that
closed
down,
and
it
came
here
and
it's
always
been
associated
strongly
with
cDNA
footprints
capabilities
and
stuff.
If
you
want
to
make
it
more
general
I
think
so
long
as
the
primary
use
case
still
ends
up
being
cdnify,
I
I
think
we
should
be
okay,
but
we
could
always
have
that
talk
with
Maria
as
it
moves
forward.
But
I
think
we
need
to
move
this
trap
forward
because
it's
been
around
for
a
while.
So
Richard
quick
comment
and
yeah
you're.
D
Good
Connor
so
I
think
what
probably
address
it,
but
you
we
reading
a
check
with
him
with
which
will
be
here
and,
of
course,
I
think
also
one
of
common,
it's
probably
objecting
to
naming
cDNA
and
HFC
I'm
a
probably
can
make
a
generic
capability
map.
That's
probably
simpler
change
then,
which
make
ever
happy
right.
Maybe
that's
the
only
change
but
yeah
I'm,
just
okay.
I
So
simple
updates
is
already
in
the
plastic,
oh,
but
we
still
make
some
changes
to
this
job.
During
these
days
we
have
made
several
updates
and
the
first
to
us,
as
in
we
fully
sorry
for
these
some
managers
that
we
fix
some
many
like
typos
and
writing
in
short,
specs
in
this
drug
and
then
four
terms,
item
we
are
new
to
new
homes
like
the
are
based
in
Zurich
and
constant
control
server,
and
we
have
three
server
names
like
address
room.
Sir
words
during
country
server
and
the
channel
a
generic
term
named
Otto
server.
I
So
we
have
some
texts
describing
what's
differences
between
them
and
for
the
following
two
items.
I
think
we
see
as
the
major
items
that
we
may
change
for
for
the
updates
of
contracted
message
formats
for
the
decoupling
of
the
address
room
service
and
the
stream
control
service,
and
the
last
one
is
that
we
improved
response
of
this
to
services
listed
aboard.
So
we
were
started
with
the
updates
of
control
message
format
in
a
previous
version,
or
maybe
some
previous
versions.
I
We
are
happy
news
that
the
removed
the
removed
field
and
for
this
for
the
update
room
service
code
for
the
countries
controls
during
service
to
inform
the
client
what
resources
have
been
removed.
So
considering
that
we
have
a
blue
mood
field
and
we
consider
to
add
another
study
field
to
improve
to
inform
the
kind
what
resources
the
service
has
already
started
to
send
notifications
send
up
updates
to
the
clients
and
we
rename
the
remove
the
field
to
the
stop
field.
I
Another
is
that
we
added
a
string
name,
the
description
just
to
send
some
text
describing
why
this
update
event
is
happening.
The
last
major
items
that
we
made
is
that
we
made
a
major
modular
design
to
the
Kaposi
attesting
service
and
stream
control
service.
So
we
have
three
entities
in
the
SS
citizen,
Z,
auto
clients
at
the
stream
server
and
the
stream
country.
So
I
think
most
of
you
are
familiar
with
what
the
process
is.
I
So
the
Otto
Klein
first
sent
an
initial
credit
to
Z
a
base,
new
server
and
then
upstream
server
my
signed
and
adapted
messages
and
as
well
as
control/play
messages
back
to
the
old
clients.
Another
that
old
client
can
also
send
a
tour
request
later
to
the
stream
conscience
over
some
of
the
stream
conscious
or
might
communicate
with
the
stream
server
with
a
primary
combination,
communication
channel,
which
is
not
considered
in
our
draft
after
this.
I
After
dealing
with
this
or
processes,
the
other
stream
server
will
still
see
ends
that
adapt
the
messages
and
the
contract
images
to
the
auto
plant,
so
in
previous
and
draft
versions
will
have
a
constraint
on
this
process.
Is
that
whenever
a
stream
is
established
in
the
other
instance,
server
will
create
another
service.
That
is
assume
control
service.
That
means
they
are
combined
together.
So
in
this
version
we
remove
this
combination
into
that.
I
So
to
support
such
the
compost
become
the
coupling
we
add
a
new
field
in
the
cup
in
the
capability
that
is
a
sub.
That
is
a
support
stream
control
field.
That
means
if
the
server
as
the
server
ensures
such
set
status
says
the
field
to
true.
That
means,
yes,
always
supports
the
strong
control
service
to
be
together
with
the
optimal
service.
But
if
the
server
sets
this
field
to
force,
that
means,
on
your
may
say,
may
use
other
mechanisms
to
to
develop
the
encompasses
mechanisms
as
the
stream
control
service,
so
synapse.
I
The
last
one
is
that
we
update
the
response
format
of
the
updates
room
service
and
a
strong
conscious
of
it,
so
that
journal
processing
of
all
the
requests
is
that
the
seller
will
first
validate
whether
your
cause
is
valid.
So
if
not-
and
we
will
return
an
auto
semi,
auto
error
message
and
if
yes
for
the
update
stream
server,
we
will
return
the
adaptive
message
and
Contra
update
message.
I
So
for
the
extreme
control
service,
we
have
more
complex
that
that
is,
if
the,
but,
if
the
request
is
varied
with,
but
I
love
with
the
invalid
Associated
data
stream
from
the
silhouette
reaching
out
for
zero
for
not
for
no
response
to
the
clients.
So
if
all
this
goes
above
to
is
valid
and
the
request
is
successfully
processed
by
the
server,
the
server
will
return,
either
on
to
zero
to
accept
your
response
or
the
for
two
zero.
Four
knots.
I
No
content
response
to
this
differences
between
these
two
response
is
that
whether
the
other
stream
server
is
sure
about
that,
the
at
the
stream
server
has
also
processed
the
request
successfully.
I
think
that's
all
the
the
last
one.
It
is
the
what
a
standard
or
two
our
response
would
be
like
and
actually
thirst.
An
auto-response
format
is
already
defined
in
the
fc7
285.
So
we
just
to
make
to
underline
some
contents
to
be
to
be
used,
such
as
Co
response,
the
auto
L
response
field
response
would
be
starting
form.
I
Ght
p400
by
request,
and
the
field
will
include
meta
field.
Is
the
code?
What's
a
semi,
auto
error
code
for
this
message
and
what's
field,
and
what's
the
value
of
your
error
message,
the
last
two
are
optional
fields,
but
the
server
can
provide
them.
So
here
is
an
example
that,
if
see,
request
include
an
invalid
field
like
remove
the
current
requested
to
remove
the
client
IDs
as
the
resource
with
named
practiced
passes
as
a
resource
is
not
previously
added
to
the
Aptus
room,
so
sick
right.
I
So
the
server
will
regard
it
as
an
arrow
request
and
will
tell
so
return
concept
and
the
code
is
the
e
invalid
fuel
value
and
the
email
field
is
the
new
field
and
value
for
the
invalid
value
is
the
properties,
so
we
have
already
set
all
for
the
updates.
We
have
already
finished
the
world
new
version
of
this
artistry
of
these
incremental
updates
and
we'll
upload
it
to
Z
working
group
as
soon
as
possible.
This
is
amazing.
A
D
So
are
you
I
think
it?
Last
night
we
meet
with
Sadie,
we
talked
about
with
his
closure
and
there's
one
one
comment
which
actually
is
not
senator
home
unionist
is,
and
by
a
year
Nixon,
who
is
the
designer
of
SSE,
so
I
think
we
got
a.
We
got
a
review
from
him
and
he
sent
a
reply
on
July,
9th
and
so
his
comment
over
the
email.
I
took
a
quick
look.
It
seems
fine
in
terms
of
design
because
he's
really
looking
at
the
review
and
the
hidden
main
objection.
D
All
his
main
mean
comment
would
be
why
don't
you
use?
Why
don't?
Why
do
you
use
a
POS
and
one
possibility
is,
for
example,
said
the
relative
coding
here
s
then
decoding
HTTP,
then
decoded
utf-8,
then
decoding
SSE,
Lindy,
couldn't
consent
and
decoding
your
vocabulary.
Implementation
will
be
just
me
to
decode
TOS,
then
your
binary
protocol,
then
in
some
cases
utf-8.
So
therefore
he
is
saying
why
you
don't
switch
to
a
binary
encoding
yourself
and
I'll.
Respond
would
be
this
essential.
A
The
chance
to
go
to
a
straight
binary
protocol.
You
know
that
particular
train
left
the
station
many
years
ago,
right
so
I
think,
while
in
practice,
yes,
we
can
do
anything,
we
are
doing
on
top
of
HTTP
with
a
bespoke
binary
protocol.
The
problem
here
is
that,
as
you
noted,
we
are
building
on
top
of
existing
work,
but
so
yeah
I
think
we
can.
We
can
attend
to
that
comment,
but
I
doubt
there
is
anything
for
the
working
group
to
do
unless
I'm
mistaken.
Okay,.
D
A
H
Just
another
comment
on
that:
I
mean
at
at
the
very
least
right.
It
would
require
us
to
probably
go
back
and
do
some
structure
in
the
vase
bag,
so
I
would
just
advocate
if
we're
going
to
do
something
that
we'd
have
a
comment
that
you
know
the
encoding
may
seem
ownerÃs
right
and
if
someone
wants
to
pursue
a
binary
protocol,
they're
more
than
welcome
with
full
understanding
do
not
fall
into
the
spec
or
I.
A
D
H
J
I
All
right,
yes,
so
instead
with
an
extra
C
working
document,
path
back
to
have
building
blocks
to
express
past
fact,
information,
yokas
type,
which
is
already
I,
think
it's.
It
doesn't
need
to
be
changed
more
and
new
NT
domain
on
which
will
still
remain
stable,
and
we
have
been
struggling
with
the
how
to
how
to
express
a
combined
information
of
Kozma
and
N
and
the
proper
map
along
with
it,
and
we
have
adopt
the
multiple
response
to
Z
have
adopted
T.
There's
a
multi-party
response.
I
Format
in
I
think
the
last
version,
and
maybe
the
last
last
version,
but
we
have
already
removed
that
part
and
move
that
path
to
a
new
service.
That
is
the
modified
service,
draft
and
I.
Think
and
it's
it'll,
also
more
dependent
and
drop
set
will
express
how
to
enquiry
the
related
information
in
a
single
request
and
how
to
return
the
relating
information
in
a
single
response.
I
This
will
be
represented
by
Johnson
and
so
long
as
as
long
as
we
have
moves
each
part,
and
so
we
have
to
we
adopt
a
new
I
think
it's
previously
discussed
part
so
with
up
to
extended
cosmic
and
the
endpoint
cost
map
service
to
support
such
capability,
and
some
other
updates
is
that
to
turn
to
step
along
with
the
extending
the
cosmic
and
implant
cost
maps
and
coasters
of
a
service.
So
we
also
make
some
updates
to
the
examples
and
also
to
the
compel
comfort
comfortably
holiday.
I
If
we
can
support
such
extensions
and
also
see
updates
on
the
security
considerations,
if
we
choose
to
express
some
information
like
past
vector
to
the
clients,
so
the
first
one
will
and
start
from
the
extended
across
map
service,
we
earn
your
Manzi
media
type,
HTTP
method
and
your
sees
as
the
same
as
the
customer
service.
Defining
the
base
of
the
protocol
by
the
way
extensively
days
filled
with
a
new
with
two
new
fields,
and
one
is
the
resource.
Id
dependent
problem
app
and
another.
I
Is
that
whether
we
allow
a
compound
in
response
in
a
single
response,
format
and
so
and
another
one?
Is
that
to
allow
the
query
of
properties?
We
add
another
import
input,
field
name
the
component
properties
and
your
clients
can
add
some
protein
names
in
this
view
to
to
ask
server
for
the
specific
properties.
So
for
our
component
response,
we
will
add
a
new
thing
where
we
will
require
the
auto
server
to
add
a
dependent
of
a
text
field
and
a
property
map
field
providing
a
proper
map.
I
So
an
example
of
separate
response
and
request
and
response
will
be
the
following,
and
the
client
will
see
on.
The
request
is
exactly
the
same
as
the
base
on
Kozma
to
say
what
cost-effectiveness
and
what
PID
PID
sets
that
it
wants,
and
the
formats
and
response
format
of
the
Kozma
is
is
the
same
as
the
base
based
course
map,
except
for
that
the
cosa
type
is
extended
to
support
a
path
vector
so
only
after
the
objects.
I
After
the
clients
received
receiving
such
information,
it
wasn't
requested
to
zip
up
the
map
to
ask
for
the
specific
properties
of
the
property
map
and
and
the
server
of
the
property
map
service
will
and
still
routine
returns.
The
same
format
as
the
unified
protein
labs
defined,
so
another
extension
is
that
we
updates
the
extended
we
update.
We
extended
as
a
compound
cost
service
actually
is
extension
to
the
endpoint
code.
I
Service
is
much
similar
to
the
cost
map
service
and
similarly,
we
add
this
dependent
protein
map
and
whether
we
support
allowed
compounds
and
response
in
these
capabilities
and
also
the
same
field
name.
The
component
properties
in
the
accept
input
parameters
and
the
response
is
also
exchange
as
the
same,
but
we
we
provide
a
different
example.
That
is
since
we
say
that
we
allow
a
compound
response
by
the
way
did
not
elaborate
it
is
it
in
the
previous
example,
so
I
will
show
it
in
this
example.
I
I
The
next
update
is
there
about
compatibility,
since
we
have
made
some
extensions
to
his
base
protocol,
so
it
it
will
introduce
some
updating
shows
so,
but
we
don't
think
there
are,
because
we
can
just
exchange
some
fields
so
for
based
on
clients
who
did
not
work
would
not
recognize
these
new
fields
and
it
will
not
introduce
extra
compatibility,
play
dangers
and
the
force
based
auto
protocol
who
do
not
and
add
such
fields
to
the
cup
to
the
to
the
its
capability.
There
is
no
other
compatibility
issues
with
the
client
another.
I
Is
that
when
the
syrup
and
will
consider
the
compatibility
issues
with
a
multiple
multi
cost,
so
we
added
an
extra
concern
here.
That
is,
we
do
not
things.
We
provide
an
array
element.
So
we
do
not
allow
constraint
test
array
elements
in
this
part
in
this
document,
so
for
uncomfortability,
with
incremental
updates
and
things,
we
have
a
dependent
issue.
That
is
the
see
map
see.
One
map
will
depend
on
another
map.
So
for
this
dependency
problem
we
have
to
follow
the
dependency
rules
and
specified
in
the
incremental
updates.
I
The
last
one
is
that
the
security
comes
durations
about
this
new
document
and,
as
we
all
know,
that
past
perspective,
information
actually
exposes
more
Network
information
to
the
from
the
server
side
to
the
client
side.
So
the
court
will
consider
on
whether
and
the
network
will
receive
more
attacks
because
of
the
much
exposure
of
the
information
and
another
is
available.
I
Availability
of
auto
service
using
the
competition
of
perspective
information
might
consume
a
lot
resources
on
the
server
side
so
on
whether
suicide
called
insistency
such
unlike
DDoS
attacks
and
due
to
the
Freak
frequent
frequent
requests
from
the
upon
shrubs
and
clients.
This
will
be
considered
as
a
security
conservation,
part
I.
Think
that's
all.
A
Ok,
any
I
want
to
move
up
ahead
since
we
are
a
little
bit
behind,
but
any
questions
on
on
path,
vector
path
vector,
is
one
of
the
most
important
deliverables
from
this
working
group.
So
if
you
have
not
read
it
I
strongly
advise
that
everybody
at
least
go
through
it
once
it
has
a
lot
of
interesting
ideas
and
and
and
and
paradigms
built
into
it,
especially
as
we
move
from
a
single
node
view
to
you
know
multiple
node
views.
So
if
you
haven't
read
it,
we
do
need
eyeballs
on
the
strapped.
A
I
Actually,
our
consideration
is
that
for
the
delivery
of
PV
information,
actually
this
dropped
and
the
extended
services
can
provide
such
such
information.
So
for
the
multi-part
service,
if
the
multiple
service
is
developed
much
well,
it
can
support
the
path
vector
information
too.
So
we
don't
have
to
stress
in
this
document
that
we
have
to
support
such
a
service.
Okay,.
A
D
Okay,
I
think
that's
a
great
question
so
so
far,
I
think
in
case
that
we
only
have
there's
no,
not
a
concept
called
set
right.
So
therefore
everything
become
array
and
so
on,
and
so
maybe
we
still
want
to
maintain
a
semantics
of
a
race
but
I
think
he's
raising
a
very
good
question
and
I.
Don't
think
I
have
a
strong
opinion,
I
don't
about
da,
oh,
but
that's
a
good
question.
Actually,
okay,.
A
So
I
think
we
probably
best
take
it
to
the
list
because
it's
appears
to
be
a
non-trivial
question
probably
require
some
reflection,
so
Jiang.
If
you
can,
please
put
it
on
the
list,
then
we
can
have
a
discussion
Richard.
You
can
comment
on
it
on
the
list
and
then
we
can
move
from
there.
Okay,
thank
you
much
and
Sabine
you're
next.
K
K
So
and
as
I
said,
there
are
two
technical
issues
that
are
remaining
so
we
have.
A
version
of
iPad
is
planned
for
the
next
well,
probably
not
weeks.
So
the
main
updates
here
about
the
consistency
of
the
registries
and
we
have
chosen
manual
consistency
method,
and
this
part
is
detailed.
Now
in
the
section
on
our
considerations-
and
we
got
that
we
have
designed
this
thing
upon
discussion
with
Elana
people
that
we
had
in
number.
So
we
also
did
some
error,
updates,
updates
on
error
handling
and
we
updated
terms
to
clarify
and
avoid
ambiguity.
K
So
now
on
the
manual
consistency
between
the
two
registries,
so
we
have
added
a
new
section
that
describes
the
procedure.
So
this
section
defines
the
two
conditions
at
which
both
registries
can
be
consistent
as
consistent,
and
there
is
a
rule
that
we
also
added
our
that
if
an
auto
domain
has
to
say
my
identifier
as
an
alto
address
type,
their
addresses
encoding
must
be
compatible.
So
in
a
nutshell,
so
I,
of
course
invite
you
to
read
the
other
section.
But
in
a
nutshell,
does
this
consistency
procedure?
K
And
if
yes,
so,
your
new
domain
domain
identifier
must
be
the
same
as
the
one
already
registered
in
the
art
web
websites
and
if
no,
you
can
define
a
new
domain
entity
domain
identifier
and
use
it
to
register
to
to
register
it
with
the
Ayana
so
and
okay.
Now,
if,
if
such
an
address
type
is
not
present
in
the
out
to
address
type
registry,
same
thing,
you
go
ahead
and
define
a
new
identified
by
the
way.
We
notice
that,
okay,
this
in
the
algorithm
of
the
last
version,
we
can
simplify
a
tralala.
K
The
algorithm
so
the
domain,
Reggie
domain
name
registration
process
is
specified
in
another
section
in
the
next
section
of
this
document.
So
to
make
it
simple
three
examples:
for
example,
you
want
to
register
domain
entity
domain
ipv4
I
can
resist.
So
you
check.
Does
this
match
with
some
known
address
that
yeah
we're?
This
is
known.
Second,
first
there
does
this
address.
Is
this
address
type
already
registered
in
the
ultra
integrity?
K
Incise?
Yes,
so
the
new
domain
name
here
Oh
must
be
the
same
as
the
one
already
registered
in
the
auto
address
type
registry.
Second
example:
you
want
to
register
a
domain
name
hopeful,
okay.
You
know
that
this
does
much
have
no
other
side
unknown
address
format
such
as
race
or
ethnicity,
okay,
but
it
is
not
specified
yes
in
the
auto
address
type
registry,
so
you
create
a
new
domain.
K
So
there
you
go
ahead.
You
define
a
new
entity
domain
and
identifier
and
you
register
this
entity
domain
with
IANA
without
pain.
So
now
our
update,
so
we
there
is
in
section
5.6
five
minutes:
okay,
okay,
so
well
we
did
some.
We
updated
the
error
handling
process
on
the
responses
to
a
property
map
request
we
also,
and
so
we
updated
the
examples
for
more
clarity
as
well,
and
so
these
is
the
remaining
issue,
but
I
will
give
the
floor
to
reach
other
audiences.
G
G
There
is
a
issue
about
when
the
client
requested
a
send
the
request
for
a
grass
block
to
the
filter.
For
my
how
the
plant
can
gather
cracked
a
property
Wyler
from
the
server,
and
the
second
issue
is
about
how
how
the
crack
interpret
the
dependency
result
from
the
other
server
information
directory
and
when
they
win,
when
they
property
might
have
the
multiple
dependency
between
the
resources.
G
One
issue:
one
yeah,
thank
you
so
the
first
issue
about
we
will
take
an
example.
For
example,
if
this
server
had
the
food
for
him
I'd,
like
the
our
left
side,
they
had
the
three
entities
and
now
the
clan
who
want
to
request
the
property
for
the
entity,
for
example,
2.0
slice
27
and
from
the
Carter
protocol
specification
the
several
return.
G
Only
this
entity
and
its
properties,
the
property
will
being
herded
from
the
PID
zero
and
when
the
clan
kaddish
with
past
trans
well
seen,
every
single
addressed
in
this
address
block
will
have
the
PID
provider
with
the
PID
zero.
But
actually
it's
the
result
is
different
from
the
foot
problem,
life
so
yeah.
So
we
propose
we
have
group,
we
have
three
routes
to
make.
This
server
can
currently
the
clan
powered
can
always
get
the
correct
properties
from
the
request.
G
So,
for
example,
if
the
clan
to
requested
stressed
at
27
and
first
first
rule,
the
server
will
return,
they
will
include
these
slides
and
is
an
entity
and
derive
its
property,
should
be
PID
0
from
the
curtain
rule,
and
then
they
found
the
syrup
out
the
sorry
their
table.
The
2.0
stresses
28
and
2.16
/
28
will
be
covered
by
this
request,
r/a
sprouts
so
and
pass
they
have
different
purple
iris.
A
G
Yeah
yeah
just
to
my
medicine
yeah,
so
so
the
process
like
for
every
scene,
very
single
entities
in
the
end
it
domain
Indian.
The
clan
will
take
the
full
uses,
lists
and
scan
the
list
for
re.
Maybe
the
resource
specific
party
and
gathered
pendency
results
in
order,
so
they
take
the
example,
a
scene
as
a
clarinet.
D
Do
want
to
mention
actually
make
a
man
made
a
comment
on
the
first
one
can't
go
back
at
first,
quick,
first
yeah,
you
sure
number
one
and
I
think
issue
number
one
actors
is
more
important
part
effects
in
this
document
and
the
issue.
Actually,
this
issue
was
somehow
not
caught
by
the
last
by
the
working
group
reviews
and
it
was
character.
I
was
talking
to
a
couple
people
from
here
and
I.
D
This
is
a
common
issue
in
program,
language
II
that
you're
going
to
talk
about
inheritance,
so
your
bigger
set
and
overall
objects,
and
you
look
at
their
attributes
and
then
somehow
would
be
a
new
also
can
partition
you're
doing
heritage,
the
refinement
and
my
inherent
use
under
patent
economics,
some
of
the
subsets
and
have
all
kind
of
properties.
If
you
query
the
property,
like
given
property
P
for
a
larger
set,
how
do
return?
D
You
return
the
base
property
or
you
return
essential
that
you
buy
properties,
and
so
that
turns
out
to
be
a
major
issue
here.
I
think
eventually,
for
our
to
be
correct,
is
we
make
a
decisions?
I
really
essentially
return
all
the
refined
properties
and
with
understanding
that
it's
really
that
people
want
to
cash
the
result
and
utilize
them
locally,
but
I
do
wanted
the
working
people
to
really
consider
and
give
a
share
comments,
because
s
very
important,
and
why,
because
to
fully
solve
it,
the
specification?
It's
not
like
a
cup
of
sentence.
D
It
need
to
be
a
lot
more.
For
example,
you
have
a
bigger
set
and
durations
would
have
all
kind
of
smaller
set
and
smaller
cell
can
over
can
overlap
if
the
overlap,
and
now
the
order
is
important,
who
appears
person
essentially
essentially
some
kind
of
like
a
like
open
flow
specification.
You
have
all
the
world
cards
and
a
different
bar
card
we're
doing
priorities.
We
essentially
need
to
talk
about
how
they
match
and
so
on.
D
A
A
Base
class
and
derived
classes
and
you
can
use
certain
casting,
etc,
to
cast
a
derived
class
to
a
base,
class,
object,
etc.
But
the
basic
idea
is
that
you
can
always
go
from
top
to
bottom,
but
you
know
you
may
not
be
able
to
go
from
bottom
to
top,
so
I
think
we
need
to
just
work
it
out
on
the
mailing
list
and
and
see
what
makes
sense.
I
So
I
have
a
question
that
is,
and
you
have
already
considered,
a
relationship
between
the
auto
entity,
domain
registration,
auto
entity
domain
registry
and
the
auto
address
type
registry
and
another
relationship
is
the
auto
endpoint
prototype
property
type
registry
and
the.
G
I
And
state
domain
and
state
property
type
registry,
so
on
in
the
auto
endpoint
property
type
registry,
coz,
property
values
is
to
be
is
assumed
to
be
a
JSON
stood
yet
string.
So
it's
registrar,
a
States
as
following.
It
has
identify
and
address
file
of
the
proper
name
and
the
intended
semantics
of
this
property.
It
does
not
have
to
define
what
a
value
type
the
value
type
of
the
property
it
is,
but
in
these
documents-
and
you
say
that
it
will
palsy
and
in
the
best
case
it
will
like.
We
change
the
name
of
the.
A
G
Actually,
I
think
we
will
well
specify
how
to
register
the
entity.
Property
I
think
we
cannot
just
a
borrow
the
end
point
property.
That's
right,
we're
different,
for
example,
in
the
issue
to
we
need
to
focus,
we
need
to
specify
which
purpose
is
resource
past
week
on
out
and
actually
it's
very
important,
nothing.
A
Ok,
I
see
Don,
not
quite
convinced
so
I'll
ask
you
both
to
take
it
to
the
mailing
list,
so
I
do
want
to
focus
now
on
the
next
three
slots.
The
first
coming
up
is
remaining
issues
in
also
from
Richard,
and
then
we
have
an
auto
survey
that
we
are
will
put
up
before
we
go
into
the
next
step
for
working
groups.
So
our
thinking
is
that
the
Altos
Erick
and
maybe
inform
us
what
the
next
steps
for
the
working
group
would
be.
D
So
I'll
talk
about
Romania
issues
and
actually
will
not
be
only
the
remaining
issues
and
I
think
a
good
kind
to
also
figure
out
what
we
have
accomplished
so
far
in
terms
of
Apple
structure,
and
so
therefore,
is
not
only
so,
therefore
that
first
part
of
the
free
market.
What
exactly
we
have
accomplished?
What
can
I
never
talk
about.
D
From
read,
beginning
of
the,
for
example,
from
a
problem
statement
and
requirements
and
then
utilizing
Sparkle
from
this
protocol
we
have
covering
and
in
the
deployment
and
that's
exactly
what
ever
comfortable
ready.
They
already
are
seats,
and
so
on
and
the
right
now
you
look
cute.
We
have
about
seven
drops
working
documents
and.
D
There
are
seven
documents
and
the
with
our
credit,
really
Richard.
Can
you
speak
in
the
mic?
Also
Mike,
yes,
okay,
great
good
idea,
okay
and
then
we
have
is
a
quite
a
few
number
of
like
a
draft
in
the
video
drafts
and
they
are
not
working
with
documents
and
people
still
maintaining
them,
updating
them
and
so
on.
So
that's
it
really
see
what's
going
on
and
how
to
read.
D
Fine,
we
really
talked
about
that
and,
firstly,
I
want
to
do
is
somehow
to
understand
exactly
somehow
the
structure
of
current
documents
so
already
I
mentioned
little
bit
your
problem
statement
and
then
your
from
requirements
from
requirement
you
give
our
base
protocol
unless
we
started
to
a
large
number
of
essential
extensions.
Of
course
here
is
we
really
have
the
discovery,
this
current
public
employment-
and
here
we
have
a
large
number
of
extensions,
but
that's
still
relatively
clear.
I
was
very
to
look
like
thing,
but
now,
let's
take
a
look
at
add,
leave
framework.
D
So
therefore,
but
now,
let's
take
a
look
at
it.
Essentially,
what
exactly
is
the
conceptual
framework
we
have
done
and
I
think
one
thing
we
learned
from
beginning
is
why
you
guys
don't
develop
a
young
programming
model
and
so
on
and
some
obviously
ok
may
be
treated
as
something
generic.
So,
let's
see
what
exactly
generic
mechanism
we
have
already
developed.
I
never
can
talk
about
between
users.
So
now,
let's
imagine
we
are
someone
teaching
this
outer
unit
class.
Let's
talk
about
itself.
What
can
a
product
from
work?
We
have?
D
We
already
introduced
as
a
generic
powerful
framework,
so
here
are
essentially
the
motor
unit
frameworks
we
have
introduced.
We
essentially
into
the
concept
of
networking
information
pick
a
basic.
The
protocol
from
network
to
applications
is
that
networking
information
is
divided
into
information
resources.
That's
a
fundamental
design
decision
for
this
working
group
for
this
essentially
the
framework
and
therefore
we
do
the
division.
We
can
account
our
media
types
become
all
arity,
we're
all
different
information
sources
at
different
media
types.
D
That's
not
the
way
to
organize
it
and
then,
if
you
divide
into
different
information
resources,
we
want
modularity,
you
reusability.
When
you
do,
you
introduce
little
issues
of
dependent
dependency,
a
consistency
so
therefore
include
is
already
the
mechanism
kind
of
dependencies
and
handle,
for
example,
consistency
of
different
resources,
and
then
we
continue
said:
oh
ok,
most
resources
are
dependent.
How
that
is
covered.
D
What
can
resources
I
can
really
access
from
this
network
and
then
what
you
could
just
concept
card:
the
information
resource
directory
you're
on
the
server
you
announce
all
the
resources
you
have
and
therefore
there's
a
resource
concept.
And
then
what
happened?
Is
you
have
the
resource
information
resource
virtually
and
how
people
funding
for
with
directory
and
to
have
is
we
have
solved
the
issue
observer
discovery
so
think
about
that
they
are
all
actually
not
really
dependent
with
auto
at
all.
It's
you
wait
at
the
generator
framework.
Let's
talk
about
applications.
What
can
I
can
provide?
D
That's
what
how
we
divide
and
conquer
modularity
and
so
on,
and
then
we
continue
in
terms
of
in-stream
work
and
what
we
said.
Okay,
now
for
using
home
a
free
resource,
let's
provide
address
for
device
service,
that's
another
decision
we
made
and
then
we
spell
about
a
basis
in
programmer
to
specify
essential
JSON
grammar
delivery,
specify
what
can
information
you
really
provide.
So
that's
a
second
progress
we
have
made
and
then
what
we
talk
about
is
okay.
D
We
have
resources
and
directories
and
Nairobi
rest
or
interface,
and
now
can
we
spare
a
grammar,
and
now
we're
talking
could
use
a
general
concept
called
filtered.
It's
okay.
Information
resources
can
be
future.
So
therefore,
that's
that's
very
nice
concept
and
then
what
we
did
was
introduced
to
map
a
general
concepts
in
this
framework,
and
one
is
a
generic
SSE
based
framework
to
stream
control,
push
increment,
update
information
resources,
it's
okay
for
not
for
every
single
individual
source,
again
grandmary
compute
alone-
and
you
can
say
hey.
D
I
want
to
push
server
push
and
I
want
to
want
to
only
push
I
won't
have
incremented
art
encoding,
not
for
encoding
and
I
can
even
control.
What
can
I
want
to
do
that?
So
that's
generic
protocol.
They
don't
help
you
apply
for
also
at
all
they're.
Just
essentially
is
a
framework
and
then
I'm
copied
what
we
did
was
essentially
a
generic
framework
called
entity
properties
supporting
inheritance
and
decomposition.
So
everything
is
entities
properties
with
your
values,
Roger
general
framework.
They
do
not
have
it
to
be
auto
at
all.
D
They
are
generic
product
framework
to
provide
information
in
a
systematic
way.
So
that's
the
one
and
now,
if
you
look
at
early
slides-
and
here
are
essentially
all
of
related
protocols
and
really
the
concept
and
if
imagine,
for
example,
I
think
there
are
several
designs
already
for
Alsop
examine
each
other
class
for
a
framework.
Those
are
the
major
frameworks.
So
now,
let's
take
a
look
at
what
we
accomplished
continues.
Let's
continue
next.
One
we
talked
about
essentially,
is
what
exact
structure
we
provided
right.
D
We
have
do
not
protocol
framework,
and
now,
let's
do
look
at
what
exactly
is
specific
for
networking
and
into
application
for
auto
purpose.
Here,
actually
is
set
of
very,
very
simple
abstractions
we
introduced
in
Auto.
So
what
what
he
is,
we're
certain
abstractions,
very
simple.
A
network
consists
of
set
of
nodes,
and
the
path
is
that's
always
that
essentially
we
don't
have
anything
fancier
and
this
moment
so
far
our
abstraction
is
a
network
consists
of
set
of
nodes
instead
of
Hattie's.
D
That's
always
that
and
what
kind
of
nodes
can
we
have
and
the
other
protocol
said?
Ok,
the
nodes
can
be
endpoints,
I
could
wait
for
address,
ipv6,
address
and
so
on,
and
it
can
be
aggregation
about
endpoints
partition,
ID
and
and
provide
E.
And
then
you
can
abstract
network
elements,
but
there's
always
that
you
can
have
those
a
concept.
No,
that's
all
we
having
an
abstraction,
all
the
rest
complex
it
they're
gone
okay
and
then
what
do
we
do?
D
We
also
said
something
very
simple,
very,
very
simple:
full
obstruction
is
a
path
is
from
one
node
to
another
node,
and
for
this
for
this
essentially
for
this
path,
you
have
passed
properties,
we
need
to
the
concept
cause
of
matrix
and
macro
can
have
calendars
and
in
pretty
okay,
so
right
now
we're
doing
is
a
pass,
can
be
a
vector
and
program,
vector
and
vector
is
set
of
abstract
network
elements.
That's
always
so
far
with
all
the
protocols.
All
we
said
is
it's
a
very
simple
abstraction?
D
Okay,
just
report
this
overall
abstraction
of
network
and
so
on,
and
now,
let's
take
a
look
at
what
other
between
issues.
If
we
could
redesign
a
system
again,
remember
we're
with
ourselves
second
system
issue:
is
you
design
system,
your
commbuys
out,
then
I
did
bad
job.
Let
me
refine,
let
me
redo
it.
Oh,
actually,
I
think
you
probably
cannot
redo
in
terms
of
what
we
accomplished.
What
about
constraint?
I
think
we
did
a
pretty
good
job,
but
of
course,
still
we
have
a
quite
a
few
places
where
I
have
remaining
issues.
D
Now,
let's
take
a
look
at
them
and
in
particular,
let's
look
at
my
from
a
concept
of
what
we
have
already
accomplished.
So
here
is
a
generic
free,
not
networking
specific,
it's
a
framework,
and
now,
let's
take
a
look
at
what
exactly
we
are
really
somehow
missing
and
one
thing
we're
missing
this
moment.
Is
we
talk
about
resource
dependency,
but
somehow
dependence
is
so
far
specified.
It's
relatively
rigid.
D
The
idea
is
centrally
bottom-up,
us-based,
driven
in
start
from
networking,
networking
maps
and
cosmetic,
our
dependency,
we
do
so
dependency
just
we
write,
for
example,
dependent
using
that
concept.
So
it's
not
very.
Is
it
possible
to
make
these
dependencies
dependence
to
be
much
inert
and
also
make
like
transport
to
be
more
generic,
its
central
into
some
kind
of
concept
of
transactions
and
so
on
as
a
generic
mechanism
and,
for
example,
llamado
and
so
on
will
have
such
mechanism,
but
my
press
open,
it
is
slightly
way
too
complex.
Then
we
really
want
it.
D
So
that's
one
remaining
issue
and
second
issue,
of
course,
is
we
have
a
grammar
so
far,
working
out
very
well
and
I've
got
me
there
about
a
young
model,
and
you
said
all
the
grammars
and
so
on
of
course
were
based
on
Jason,
and
so
the
question
is:
how
long
do
we
really
want
to
formalize
all
Grandma
or
not
do
run
have
write
down
so
oh
here's,
a
grammar
I
mean
gonna,
be
very
useful
for
ours
and
also
useful
for
other
people.
So
therefore,
that's
our
ii
mean
you.
D
Two
people
can
think
about
the
second
one.
Essentially,
if
we
talk
about
essentially
is
we
have
this
increment
to
updates
and
pushes
which
is
very
powerful
mechanism
now
the
question
is:
can
we
do
more
junior
flex
or
update,
and
so,
for
example,
were
based
on
SSC
how
about,
for
example,
GTP
to
come
on
elements?
So,
therefore,
if
it
has
a
remaining
issue
it
when
you
think
about
it,
we
may
or
may
not
design
so
far
working
out
very
well
and
for
example,
why
does
the
trigger
trigger
about
comments
from
a
guy
from
from
SSC
said?
D
Okay,
if
you
use
another
transport,
maybe
it
can
be
a
lot
more
efficient.
Maybe
we
even
develop
I
want
in
the
framework
of
these
overall
framework.
Remember
this
jedoug
framework,
and
then
we
talked
about
is
the
following.
We
have
user
concept,
we
can
filtering
in
multiple
places,
but
somehow
they
are
all
specific
there's,
not
a
jinuk
filtering
concept.
We
introducing
auto
framework.
D
We
are
with
specify
the
future
and
and
other
stuff,
and
for
example,
there
is
a
generic
fusion
concept
example,
but
it's
incredible
in
your
vision,
I
believe
that's
the
way
for
getting
a
typical
future,
a
data
tree,
for
example
a
low
grammar
but
less
efficient.
So
therefore,
question
is
I
mean
issue
is
mechanism?
D
Do
you
specify
enough
a
same
time
actual
to
be
be
generally
nutball
context,
I,
think
that
be
very,
very
useful
and
very
helpful
can
simply
I
make
a
lotta
a
lot
more
beautiful,
so
that
those
are
the
remaining
issues.
The
second
set
of
remaining
issues
we
related
with.
Essentially
they
are
obstructions,
rectory,
not
a
protocol
framework
and
in
terms
of
the
abstractions.
Let's
we
can
one
by
one
well,
of
course,
is
if
we
talk
about
all
kinds
of
nodes.
D
Okay,
so
we
talked
about,
for
example,
all
kind
of
entities
and
so
on,
and
now
our
intent
is
what
cause
be.
Very
simple
would
be
like
a
typical
domains,
ipv4
ipv6
paid
and
also
for
examine
every
abstract
elements.
We
probably
need
to
be
to
start
looking
to
allocate
domains,
for
example
like
for
telephone
or
5g
and
for
all
kind
of
domains,
and
they
are
all
very
low
hanging
fruit,
because
why
we
have
a
whole
framework,
Oh
some
movie
very,
very
simple
cling
and
which
would
probably
shouldn't
start
to
really
think
about
that.
D
And
now,
let's
look
at
all
the
paths
with
support,
and
one
thing
we
are
doing
this
moment
is:
we
have
all
you
can
like
a
Casa
map
between
two
aggregations
and
informacast
map
is
between
two
endpoints
and
then
has
factor.
Essentially
the
properties
is
pass
vector.
The
question
is
for
the
I'm.
Sorry
is.
D
We
can
think
about
the
generic
pass
properties
from
one
node
to
another
node
or
set
up
note.
The
second,
for
example,
can
be
flow,
multicast,
multi
path.
They
are
now
somehow
starting
to
be
a
little
bit
like
low-hanging
fruit,
and
it
can
be
all
very
beautiful.
The
design
I
think
if
you
are
interesting
and
solve
those
issues,
I
think
there
are
key
remaining
issues
and
now,
let's
take
a
look
at
the
other,
its
outliers
and
one
other
I,
of
course,
is
the
one.
We
talked
about
ready
calendar
concept.
D
I
think
this
is
very
bad
call
coincide,
weather
cannon
concert,
but
unfortunately
kind
of
concept
is
introduced.
Is
it
also
piecemeal
here
talk
across
America
calendar?
Can
we
make
a
calendar
concept
to
be
generic,
essentially
jinuk
annotation
in
the
system,
design
I
need
to
do
I.
Think
that'd
be
very
beautiful.
We're
not
doing
that
and
also
another
concept,
if
you
think
about
it,
and
it's
where
a
future
in
this
moment
is
very,
very
specific.
D
Can
we
make
it
generic,
of
course,
Janaki
you
come
over
language,
express
or
maybe
all
too
not
into
more
obstructions
design,
very
abstract
and
very,
very
clear
way,
I
think
there's
some
like
a
unified
representation.
We're
trying
to
do
it
yesterday
at
six
and
so
on,
trying
to
look
into
the
path,
but
probably
can
be
very
interesting.
D
One
more
thing
is:
if
you
look
at
this
look
at
this
bigger
picture
you
realize
FCI
map
is
somehow
it's
an
outlier
in
some
sense
nodes
and
the
path
is
linking
nodes,
and
here
at
CI,
essential
knows
which
properties
but
reverse
it.
That's
called
capabilities,
I
think
what
we
really
need
to
start
with.
We
think
about
it,
because
it's
really
motivated
by
the
comments
by
you
feedback
these
people
from
5g
from
Chennai
mobile
and
so
on,
is
maybe
we
really
show
the
design
is
to
want
to
be
to
not
network
capability
capability.
D
So,
therefore,
now,
suddenly
our
system
becomes
a
very
powerful
design.
So
those
are
the
two
ones
and
extension
remain
issues
in
the
junetta
protocol
framework
and
remain
issues
in
terms
of
abstractions
and,
of
course,
propolis
is
the
most
important
and
is
also
look
at
also
issued
in
a
bigger
picture.
So
what
are
the
bigger
picture?
Issues
number
one
of
course
really
fundamentally
is.
D
It
should
be
use
cases
and
requirements
and
no
matter
how
beautiful
your
framework
is,
how
elegantly
abstraction
is
and
you're
all
happy
and
so
on,
but
you've
actually
is
in
the
determination
is
how
good
job
you
do
to
make
it
useful.
You
could
be
very
ugly
protocols
to
be
useful
with
other
around
may
not
work.
So
therefore,
I
think
one
major
use
case
and
major
essential
remain
issue
is
look
at
use.
Cases
and
requirement,
for
example,
I
think
allows
really
arguing
very
strongly
that
we
should
really
lose
this
man.
D
What
we
look
at
yours
cases,
org,
attrition,
I,
think
now
real
time
we
have
all
the
base
protocols
based
abstractions,
not
less
and
I'll,
really
try
all
sound
like
a
real
use
cases
and
also
deployment
I'll,
be
very
useful
and,
of
course,
other
settings,
for
example,
Murphy
gomez
and
computerization
every
cloud,
IOT,
and
so
people
proposal
can
be
discussions
about
it.
We're
not
really
looking
at
we're
focused
out
based
protocols,
but
now
probably
time
to
look
at
it,
then
at
the
top.
D
If
you
also
look
at
infrastructures,
I
think
it
early
days
we
have
all
kind
of
people
presenting
and
so
on
later
we
all
focus
on
based
protocols.
Most
people
start
to
really
disappear,
but
they
are
also
very
important
is
an
abstraction
it
does
not.
It
doesn't
see
at
all
where
abstraction
come
from
and
it
just
hasn't
been
there
a
secret,
and
so,
but
there
are
a
lot
of
a
commenting,
of
course,
a
lot
of
it
really
been
there
specific.
D
That's
the
way,
you're
smart,
that's
where
you
divide
a
product,
that's
where
you
really
become
money
and
so
on.
But
there
are
a
lot
also
like
a
comment
in
which,
if
you
can
do
it,
for
example,
and
one
example
would
be
proposing
I
think
there's
also
draft
now,
because
we're
not
pushing
it
because
on
base
product
is
smart.
D
We
have
all
the
obstructions,
but
how
other
really
provided
in
an
efficient
scalable
way,
for
example
a
smart
team
and
even
though
it
structure
rubber
start,
we
will
think
about
that.
If
we
will
continue
the
work
on
here
and,
of
course,
is
to
really
integrate
information
from
other
places
provide
a
most
capable
packing.
D
A
Yes,
all
right,
so
lots
of
information,
nice
taxonomy
of
where
the
drafts
are
and
what
the
work
is
in
also
with
respect
to
the
large
abstractions
that
we
have
in
also
I
what
I'm
going
to
do
is
I'm
going
to
go
with
the
survey,
and
then
we
have
an
open
discussion
slot.
So
if
you
have
any
questions
on
on
what
which
are
presented,
we
can
roll
that
into
the
discussion
on
what
to
do
with
the
next
work.
You
know
what
to
do.
Next
with
the
working
group.
Wonderful.
D
D
E
E
I
want
to
describe
our
goals
and
scope.
Our
goal
is
we
want
to
find
out
the
auto
increment
patience,
the
possible
issues,
and
we
also
want
to
find
out
the
current
use
case,
sorry
and
future
use
case
on
Paris
research
opportunities
and
full
of
scope.
We
have
two
parts,
one
part
is
about
auto
implementation
and
the
second
one
is
about
application
scenario
of
Auto
English
literature
work.
E
So,
for
the
first
part,
it's
about
the
implementation.
The
methodology
of
us
is
the
questionnaire.
Actually,
we
design
questionnaire.
It
includes
three
parts.
The
first
one
is
the
basic
information
and
it
include
project,
name
and
motivation
system,
architects,
implement
map
surveys
and
the
second
one
is
the
four
entities
in
Auto
implementations,
which
is
data
source,
auto
server
or
the
client
and
information
consumer.
The
third
one
is
about
the
benefits
or
issues.
E
The
next
slide
is
about
methodology
of
the
auto
application
scenario
in
leadership
world,
including
papers,
projects,
patents
and
RFC's.
Actually
at
least
parts.
We
do
a
lot
of
heavy
paper
reading
and
we
trying
to
class
classify
them
and
we
use
an
auto
survey
form
which
we
classified
the
papers
into
three
category.
The
first
one
is
the
paper
who
is
describing
the
solution
of
implementations
or
deployments
of
auto,
and
the
second
one
is
the
papers
mentioned.
E
Also
as
recommended
protocol
and
the
least
part
could
be
a
potential
use
case
for
auto
and
the
third
one
is
some
papers
compete
with
Auto
and
the
least
paper
is
also
could
be
the
research
opportunities
and
for
each
papers.
We
were
trying
to
dig
out.
Is
this
paper
extending
the
autos
and
how
they
use
auto
or
mention
auto,
and
is
there
any
challenge
or
research
opportunities
or
lease
flavor
is
in
a
single
domain
situation
or
a
multi
domain,
and
also
the
application
scenarios?
E
So
next
next
part
I
will
summarize
the
current
stakes
of
our
survey
for
implementations.
Currently,
we
found
about
6
implementation
layer
from
opendaylight,
auto
and
auto
base.
Progress
is
mojito,
Mac
situation
and
big
nose,
and
also
auto
in
Sdn,
a
mobile
network,
and
also
from
Cisco
Network
positioning
system
and
also
a
unique
own,
and
we
also
know
two
more
potentially
potential
company
which
already
implement
autos.
They
are
Tata
and
Canada
barrels,
and
here
we
may
need
some.
E
We
may
need
help
from
were
Pico
member
to
to
let
us
get
more
connection
with
small
companies
or
institutions
so
that
we
can
English
our
surveys
in
the
next
part.
I
will
summarize
the
current
state
of
literature
work.
First,
we
trying
to
classified
by
the
applications
narrow
the
numbers
in
the
branch.
It
is
the
count
of
lead
of
that
kind
of
candles
of
papers,
actually
layer
for
forced
application
scenarios
in
black
font.
E
They
are
relatively
common,
which
is
p2p,
CDN
Sdn
and
a
cloud
and
the
rest
of
it
in
a
red
phones
is
relatively
new
application
scenario,
which
include
NF
we
and
wireless
blockchain
and
IO,
and
also
up
multi-domain
the
next
part.
We
also
trying
to
classify
the
auto
body's
functionalities.
They
are
four
functionalities.
E
Currently,
we
summarized
the
first
one
is
about
a
peer
selection
include
peer
selection
in
p2p,
surrogates
direction
in
CDN
and
DC
and
selection,
and
also
location
selection.
A
sec
one
is
about
the
past
election.
It's
mentioned
by
the
past
election
in
multipath
transport,
transport,
layer,
application
relay
analyst.
This
one
is
about
a
resource
placement,
it
include
virtualized
service
function,
chaining
placement
and
intelligent
virtual
machine
placements,
and
also
the
service
placements
in
IOT,
and
the
last
one
is
about
a
measurement
result.
E
Interface
also
can
be
used
as
an
interface
to
query
on
a
large
scale
measurement
results
and
for
existing
use
case.
We
can
refer
to
this
link,
so
the
next
one
is
about
some
future
use
case.
The
first
one
is
mentioned
by
Bell
Canada
in
IETF
101
pa
pa
energy
is
about
it.
Mention
Auto
as
a
network
behaviour
exchanger
between
providers
and
the
second
one
is
mentioned
by
the
es
net
6e
x
net
is
energy
science
network
of
American,
and
it
provides
a
nation,
a
nationwide
high
speed
network,
to
help
the
researcher
to
do
some.
E
Research
and
Leigh
mentioned
Auto
as
being
used
to
solve
unsquare
problems
in
the
last
third,
and
the
third
use
case
is
also
as
an
information
source
of
Lazarus
placement
in
IOT.
Actually,
authors
can
help
to
to
select
the
location
of
News
Service
instance
in
IOT
and
the
first
one
is.
The
auto
can
be
used
as
a
surveys
to
guide
the
peer
selection
in
the
blockchain
and
I
think
which
had
already
mentioned
some
of
this
challenge,
and
they
are
too
large.
E
So
if,
if
there,
if,
if,
although
can
also
I
mean
if,
if
also
the
network
provider,
can
get
some
information
from
the
application
so
that
they
could
do
more
optimization
of
their
resource
allocations,
analyst
second
one
is
the
mistrust
between
the
network
provider
and
application
provider
list.
Seo
is
also
important
because,
if
like
an,
if
this
can
be
soft,
I
think
it
will
make
the
make
the
auto
can
be
applied
in
a
more
generic
scenario.
E
A
Okay,
I
noticed
the
blockchain
stuff.
There
was
actually
a
draft
here
in
ITF
in
an
altar
working
group
that
looked
at
block
chains
couple
iterations
ago.
I,
don't
think
they've
updated
it,
but
we
could
start
with
that,
but
any
any
questions
from
the
working
group
or
any
feedback
before
we
go
into
our
discussion
session.
A
We
are
trying
to
finish
the
current
deliverables
and
I
want
to
open
up
the
you
know
the
time
to
the
rest
of
the
working
group
to
decide
on
next
steps
mariya
not
to
put
you
on
the
spot.
If
you
want
to
kick
us
up
with
kick
us
off
with
what
you
think
you
know,
are
the
next
steps
beyond
finishing
the
current
work
and
then
seeing
what
to
do?
Maybe
that
will
open
up
the
and
and
and
cause
other
people
to
come
and
give
their
comments
as
well.
Yeah.
J
So,
for
me,
this
is,
of
course,
a
processing
issue.
So
for
me,
it's
really
important
to
see
that
there's
enough
energy
and
interest
in
the
group
that
the
group
can
take
on
work
and
then
also
sufficiently
in
a
timely
manner,
finish
the
work
having
enough
people
to
review
it
and
guarantee
certain
quality
of
work
and
I.
Think
what's
important
here
is
really
to
have
people
in
the
room
who
kind
want
to
breed,
or
at
least
on
the
mailing
list,
but
like
more.
J
Ideally,
it
would
also
be
really
people
from
industry
who
come
here
and
have
a
specific
use
case
where
they
want
to
deploy
Auto
and
have
problems
that
need
to
be
addressed.
I
think
this
is.
This
is
really
interesting
work
to
see
what
the
different
opportunities
are
for
Auto,
but
it's
it's
research
work
right.
What
we
really
need
is
people
who
want
to
deploy
Alto
and
have
like
the
manpower
and
energy
to
work
on
the
protocol
and
have
an
industry
need
for
deployment
and
related
to
that
point.
J
So
there's
a
trade-off,
so
you
have
to
figure
out
what
is
actually
the
use
cases
that
people
want
to
address,
because
if
you
have
to
implement
a
generic
framework
for
one
specific
use
case-
and
you
don't
have
an
and
benefit
from
like
all
this
general
functions
in
there,
it
might
actually
be
burden
for
deployment
because
it's
more
complex,
then
you
need
it
right.
So
that's
something
to
consider
as
well.
H
Lyle
Bert's
speaking
in
part
from
industry,
so
we've
been
in
several
of
these
meetings,
but
we've
been
told
to
kind
of
hold
back
right,
given
the
backlog
right
that
we
have
so
it's
it's
not
for
lack
of
pushing
and
then
you've
got
like
NPS
from
Cisco
and
others
that
have
it
in
production
in
the
system.
So
and
we've
had
a
system
that
we
actually
retired
after
a
few
years
where
it's
actually
been
in
production,
we
retired
the
system
not
because
of
the
use
of
Aalto,
but
just
it
was
time
to
retire.
H
The
system,
in
fact
I
think
we're.
We
finally
figured
out
that
2011
technology
and
stuff
we
want
to
move
to
and
as
operators
right,
we're
very
slow.
So
interest
is
still
there,
but
I
would
also
say
because
of
kind
of
the
pushback
or
the
appearance
that
you
know,
the
framework
wasn't
done.
Those
sort
of
things
right.
The
appearances
is
that
we're
not
as
present
and
just
realistically
speaking,
we've
heard
you
and
the
chairs
loud
and
clear,
and
we've
just
held
back
so
now.
H
J
H
All
along
I
think
the
industry
is
now
catching
up,
so
I
think
the
crossroads
we
have
is.
We
can
still
provide
specific
application
context,
which
is
great
but
I,
think
the
industry,
or
at
least
the
operators
now
are
awake
enough
to
kind
of
say
that
would
be
good,
but
the
more
interesting
one
is.
Can
we
go
after
the
generic
application
deployment
so
basically
tie
it
more
into
orchestration,
so
I
think
you'll
get
individual
apps,
but
what
the
industry
would
be
looking
for,
unfortunately,
is
it's
kind
of
weird
right.
H
We
came
in
with
a
lot
of
very
specific
apps
a
few
years
ago,
and
now
we
be
more
along
the
lines
of.
Could
you
just
support
us
with
generic
orchestration,
which
is
a
little
bit
more
generic,
so
I?
It's
it's
going
to
be
kind
of
this
push-pull
between
a
very
specific
narrow
use
cases
that
are
still
out
there,
which
would
be
more
happy
to
provide
versus
right,
as
so
long
as
it's
still
coming
in
kind
of
the
generic
framework
of
orchestration
or
other
items
right.
L
D
I
just
wanna
point
I,
think
those
are
actually
exercising
comments
and
one
thing
I'm
proposing
and
with
of
course,
with
several
people
and
which
I
collaborate
with
people
from
and
several
operators,
and
is
we
I
think
we
thought
our
ad
and
yesterday
very
briefly,
and
it
was
a
very
good
guidance,
so
I
think
one
thing
really
is
the
industry
and
so
on.
We're
organizing
or
trying
to
organize
a
workshop
and
I
think
if
anyone's
interested
Malinois
could
tell
us
very
like
unfortunate,
because
I'm
founded
in
the
u.s.
so
I'm,
basically
not
in
place
a
year.
D
So
therefore
we're
likely
I'll
be
founded
about
the
US
government
and
we
can
you
might
people
operators
and
we
talk
about
use
cases
and
so
on,
and
we
identify
issues
I
think
that's
probably
a
good
way
to
do.
This
is
really
a
face-to-face
meeting.
I
understand
what
kind
of
like
a
ribbon
usually
were
not
addressing
so
so
I,
don't
claim
anything.
I
know
I'm
just
trying
to
talk
about
an
opportunity,
sweet
eyebrows,
10
million,
an
email
to
the
meaning
is
to
really
discuss
about
very
likely.
D
K
They
cross
all
sorts
of
networks
and
in
real
life
you
have
like
I,
don't
know
if,
in
a
simple
case
you
you
have
at
least
dozens
of
Sdn
controllers,
you
have
dozens
of
Orchestrator
and
these
elements
need
to
talk
to
each
other
to
access
network,
topology,
informations
and
capabilities.
So
I
mean
we
may
design
a
homemade
interface
to
to
to
look
at
it,
but
I
think
well
also
seems
so
well.
It
is
actually
easy
to
use
it
I
mean
it
works.
So,
but
definitely
we
need
to
extend
the
auto
topology
to
this
diversity
of
endpoints.
K
So
this
is
why
we
need
to
look
at,
for
example,
those
cellular
address
multi
path,
multi
path
capability,
so
this
was
like
touched
in
I
believe
Chicago.
So
we
well
with
the
possibility
to
have
several
cost
in
one
path
between
one
source
and
one
destination,
but
definitely
we
need
to
look
at
it
like
more
in
detail
and
see
what
what
we
still
need
to
do.
The
other
points
we
really
need
to
address
I
think
is:
is
a
country
flex
a
big
concern
from
many
operators?
Many
of
them
they
say.
K
Well,
we
are
not
so
great
about
advertising
our
net
ma
our
capabilities
in
terms
of
performance
related
metrics
like
bandwidth,
throughput
latency.
So
definitely
we
don't
really
want
to,
but
we
are
interested
in
okay
trading,
our
costs
and
trading
those
capabilities
and
also
interest
from
applications
to
look
at
our
outcomes
to
take
our
cost
in
to
extent.
So
there
is
we
a
next
big
topic
will
be.
What
is
the
level
we
want
of
aggregation
of
abstraction?
So
we
want
to
announce
our
advertise
all
what
an
application
needs,
but
only
what
it
needs
application.
K
They
don't
care
about
complexity.
They
don't
want
complexity.
So
I
think
this
is
a
real
big
topic,
so
it's
not
about
how
to
aggregate
this
is
research.
This
is
I
would
have
scope
of
the
outer
protocol,
but
a
way
to
give
the
applications
or
the
user
an
understanding
on
how
how
to
interpret
this
information
a
hint
on
how
it
was
aggregated
and
also
we
need
to
care
about
confidentiality,
and
we
need
to
care
about
complexity.
I
I
think
I
have
a
1-point
from
previous
discussion
in
several
members
of
our
working
group.
That
is,
and
also
is
now
some
much
like
South
to
nose
and
protocol
that
is
providing
the
network
information
from
bottom
and
opportunity.
Applications
so
in
the
recently
mentioned
draft
stay
mentioning
some
implications
like
which
will
contains
the
multi
domain.
A
K
J
Yeah
and
I
also
have
the
feeling
that
this
maybe
goes
a
little
bit
beyond
the
scope
of
how
the
Alta
working
group
is
currently
chartered
and
that's
another
question.
I
have
because
it
seems
like
there
might
be
a
point
where
we
can
go
away
from
like
developing
the
protocol
to
a
point
where
it's
more
about
maintaining
or
operating
the
protocol.
J
A
All
right,
thank
you
folks,
so
this
is
unfortunately
the
meeting
is
not
over
yet,
but
this
is
the
the
chartered
part
of
of
the
meeting.
So
now
we
are
going
to
go
into
these
individual
draft
that
certain
people
have
been
driving
for
a
while,
and
these
are
interesting
drafts
because
they
give
you
an
idea
of
where
the
work
is
headed,
so
I
believe
the
first
one
is.
G
Okay,
thank
you.
So
this
document
is
derived
from
the
our
previous
path,
vector
document,
because
with
racing
they
see
the
general
problem
service.
We
want
to
provide
an
extent
in
the
auto
and
yeah.
We
take
how
much
yeah.
Thank
you.
So
the
motivation
why
we
want
to
provide
this
new
service
is
that,
because
we
found
many
use
cases
using
auto,
we
involved
multiple,
auto
resources
to
complete
their
own
traffic
out
mini
optimization
tasks.
G
G
Yes,
so
the
in
this
document,
which
is
list
three
use
cases
to
using
the
stories
like
the
we
can
curate
a
multiple
resource
in
a
single
cure
in
impact,
and
we
can
consider
one
client
request
a
proper
team
at
first
and
then
felt
her
the
entity
with
the
some
specific
property
riders
and
then
they
curie.
The
cosmic
are
of
the
network
map
for
these
entities
and.
G
We
also,
we
also
consider
how
to
use
this
general
service,
to
suppose
a
pass
factor
extension
and
to
provide
this
service.
We
consider
the
auto
shield
they
also
service.
You'd
met
these
three
additional
requirements
execute
so
the
first
one.
We
need
to
consider
how
to
how
the
new
response
came
up
to
provide
a
multiple
resource
and,
second,
why
is
how
to
provide
the
general
fury
schema
to
filter
that
arbitrary
resources
and
the
major
part
is
how
to
how
to
provide
a
relation
Curie
and
our
given
example-
was
the
current
solution,
our
new
service
yeah?
G
They
called
adjacent
a
queue
and,
of
course,
you
can
use
some
other
language
and
it's
a
general
design
missing
and
all
the
multi-part
messages
very
simple
say
this
server
can
bundle
the
that
will
result
in
a
single
response.
I
can
provide
the
content
tab
and
the
we
saw
a
response
data
entry
for
each
resource.
G
We
consider
the
impulse
by
2phi
a
curry
language
program
and
it
is
people
writing
in
the
language
specified
by
the
client
and
in
this
example,
like
the
client
want
to
cure
the
first
accurately
problematic
of
the
prop
map
location
and
it's
a
cat
mode
request,
and
then
the
second
resource
is
for
the
clan
want
to
occur.
The
mighty
forecast
map
and
the
input
will
be
generated
by
the
curry
program
written
in
the
key
snack.
G
So
just
like
this
reference
workflow
for
this
new
service.
If
we
consider
how
to
implement
it,
so
the
plant
just
specify
send
request
waste,
multiple
requests
for
the
multiple
resulted
and
first
in
the
server
side,
Sarah
will,
firstly,
restore
the
dependency
between
days
to
request.
So
at
this
example,
they
request
to
and
the
requests
three
will
depend
on
the
response
of
the
request
to
one.
So
the
server
will
first
labor
say:
first,
if
they
request
one
I
gather
it's
at
1
and
thence
in
the
request
and
then
ask
you.
G
G
G
G
D
Do
want
to
give
a
common
I
think
this
is
very
interesting
by
the
same
time
and
UK
justify
this
particular
Creole.
Angry
design,
for
example,
for
example.
Any
is
throwing
in
something
very
simple:
why
sicko
is
nautical
language,
and
oh,
it's
more
general,
or
why,
for
example,
I
guess
your
syntax
looks
very
similar
to
like
a
spark
all
these
kind
of
like
it's
in
a
functional
language
where
the
D
Drive
occur
languages.
D
Why
is
that
a
really
good
language,
design,
lighting
and
audio
stuff
and
in
what
sense,
for
example,
because
we
do
have
their
simple
abstractions
in
what
sense
this
language
derived
from
those
very
simple
nodes
path,
is
clear
language.
Essentially,
the
people
also
have
like
a
graph
based
query
languages
right,
like
all
these
blue-screen
or
whatever
they
have
more
clear
language,
I'm,
just
kind
of
and
trying
to
I
guess,
I'm.
A
question
is:
how
do
you
justify
this?
The
red
language?
G
A
All
right,
I
think
this
is
an
interesting
work,
but
I
think
from
the
point
of
your
question
on
whether
we
can
adopt
it
as
a
working
group
item,
it's
a
little
bit
premature,
at
least
right
now,
given
the
deliverables
we
have
I
think
this
is
fair
to
ask
that
question
as
we
go
in
to
reach
ordering
and
and
other
discussions,
etc.
Yeah.
G
G
Thank
you.
The
major
update
is
consider
how
to
support
the
multicast
flows
in
the
tropical
coast,
curity
LAN,
and
also
we
want
to
meet
this
dungeon
general
for
some
specific
information.
So
we
add
a
flow
specific
information
on
sment
in
the
current
flow
based
bathroom
and
we
currently,
we
only
define
the
new
flow
specific
information
announcement
like
the
sense
mention
tab
and
it
can
be
the
unicast
and
multicast.
We
redefine
how
to
provide
a
multicast
fury.
G
And
the
Cremonese
like
it
for
the
current
out
who
resource
curie
it
come
out
even
in
the
cosmic,
the
entrant
code
met
all
the
way
proposed.
The
flow
based
call
security.
It
can,
it
can
only
suppose
a
unicast
unicast
flow
Curie,
so
it
cannot
spoil
the
multicast
and
if
we
consider
the
podcast
may
be
the
result
of
it
will
be
very
different.
G
The
capacity
region
like
the
bandwidth
of
the
for
a
while
and
flow
to
their
dear
son,
should
be
less
than
or
equal
to
D
100,
but
if
the
flow
line
and
flow
to
at
the
Americas
flow,
so
they
as
well
and
the
link
as
one
key
weapon
out
the
ball
tonight
link
because
the
flow
will
be
abrogated
in
this
link.
So
actually
the
capacity
region
should
be
every
the
penalized
for
every
individual
flows
should
be
less
than
the
greater
than
D.
G
So
we
attend
the
current
flow
based
car
to
redesign
to
introduce
a
new
felt
like
the
record,
is
a
flows
by
pronounced
and
if
the
extensible
didn't
and
the
plan
can
specify
some
non
point,
attributes
for
the
flows.
So
the
currently
we
just
defined
the
transmission
tab
and
the
tab
can
be
the
unicast
and
multicast,
and
if
this
condition
is
the
multicast
for
the
flows,
the
other
server
will
inherit
the
destination
address
as
a
multicast
flow
multicast
groups
and
the
disturber
will
each
post
all
the
masks
as
close
to
the
multicast
the
targets.
G
G
But
for
this,
for
this
prospective
information
announcement
is
done.
There's
some
issues
like
if
we
consider
the
clan
want
to
request
them
some
different
flow,
but
they
flow
have
the
same
between
the
same
endpoint,
but
they
have
the
different
flow.
Specific
information,
how
they,
how
the
server
can
respond,
can
return
to
the
response,
because
in
the
current
autorickshaw
schema,
it
can
only
response
the
cost
for
the
endpoint
to
endpoint,
but
cannot
come
out,
provide
the
information,
the
cost
for
the
noun
in
contact
review.
G
So
can
we
extend
the
response
format
without
changing
the
current
request
format,
because
we
we
want
to
make
the
flow
based
category
decouple
with
the
response
format,
and
also
we
want
that
invite
whether
it's
difficult
to
implement
in
the
real
system
and
next
step.
We
want
to
check
my
case
for
the
flow
based
Kure
to
see
how
to
proceed
it
from
yeah
question.
A
A
A
A
A
M
Yeah,
thank
you.
So,
in
the
last
meeting
we
submit
a
word
in
Tier
one,
and
this
make
we
in
the
draft
will
make
lots
of
updates
and
including
the
details
of
our
three
phase
discovery
process,
to
discover
the
resource
from
different
domains
or
different
member
networks
and
to
discover
the
network
resource
sharing
between
flows
across
multiple
networks
and
but
and
in
the
meantime,
to
maintain
the
privacy
of
each
network.
We
propose
a
privacy
preserving
aggregation
service
at
the
Auto
extension
next,
please.
M
So
in
in
this
ITF
with
updated
our
craft
with
0-2
version
and
which
includes
two
major
technical
updates,
the
first
one
is:
we
update
the
design
of
our
privacy,
preserving
resource
aggregation
service
into
a
new
design,
which
does
not
require
a
changing
aggregation
process
so
that
we
can
substantially
reduce
the
latency
and
the
second
one
is
a
projection
technique
which
we
introduced
to
improve
the
scalability
of
our
system.
Next,
please.
A
A
M
Yeah,
so
maybe
can
you
go
to
slide
six
yeah,
so
in
slide
six
I
want
to.
In
the
example,
I
showed
you
in
slide,
six
I
show
in
our
new
design.
We
have.
We
have,
instead
of
using
a
changing
process
for
auto
servers
to
commune
with
each
other.
We
directly
like
with
directly
light
a
each
a
is
to
send
a
disguise
set
of
feasible
regions,
but
directly
back
to
the
auto
plan
and
after
the
auto
clan
collect
all
sets
of
linear
equations.
It
sums
up
all
sets
together
and
eventually
got
the
same.
M
These
guys
set
of
linear
equations.
So
in
this
way,
in
this
way
that
latency
the
aggregation
latency,
is
reduced
to
the
maximum
of
all
auto
server
client
latencies.
So
in
so
this
substantially
reduce
the
latency
from
owen.
201
I
can
go
to
slide
number
eight!
Please
thank
you.
So
you
know
the
in
addition
to
reduce
the
aggregation
latency.
M
We
also
did
develop
another
feature
for
scalability,
Purser's
and,
and
the
basic
idea
is
to
let
the
auto
server
in
each
network
can
pre-compute
the
resource
sharing
information
for
a
set
of
flows
whose
source
and
destination
is
a
combination
of
all
about
ingress
and
egress
of
the
network
and
use
this
pre-compute
results
to
do
projection
to
get
the
actual
results
for
Curie
from
the
client
Slenderman
I.
Please
all.
A
Right
I
think
we
should
probably
start
wrapping
this
up
and
maybe
when
you
start
your
next
show
next
slide
show,
then
you
can
mute
your
video,
if
possible,
so
go
finish
this
in
the
next
30
seconds.
Okay,.
M
M
Two
equals
two
for
ingress
egress
pairs
resulting
a
set
of
linear
inequalities
and
when
a
real
actually
comes
from
that
client
for
two
flows
as
when
to
D
1
and
s
2
D
2,
the
auto
server
can
protect
that
the
pre-computer
result
and
to
directly
get
the
actual
accurate
on
for
is
to
flows
only
and
return
to
the
return
to
the
client.
So
to
summarize,
I
know
the
time
so
we
have.
There
are
several
data,
we're
going
move.
K
M
M
We
need
to
move
forward
with
our
scalability
design
to
make
sure
it
can
integrate
into
the
other
phase
of
our
resource
discovery,
part
and
additionally,
just
a
very
quick
update.
Our
overall
design
is
under
the
final
review
of
stage
4
the
supercomputing
conference
in
2018,
and
that
is
all
for
this
draft.
A
M
M
A
M
We
can
go
to
our.
He
go
to
slide
number
two,
so
this
is
a
similar,
a
somatic
part
of
our
generic
scurry
service.
This
part
is
this:
unified
resource
recognition
is
also
very
early
stage,
but
so
the
back
one
we
have
is.
Is
we
start
from
the
base
out
of
protocol?
It
provides
use
this
one
big
switch
abstraction
to
provide
your
networking
rich
with
network
information
to
applications,
but
this
abstraction
cannot
review
the
resource
sharing
among
a
set
of
flows
or
uneven
pairs,
so
in
DC.
So
that's
one
of
them.
M
M
M
A
A
A
A
A
A
B
B
B
First
slide
with
your
name
on
it:
okay,
nice,
so
so
it
is
a
quick
update
on
the
broker
system.
With
you,
my
orchestration
draft,
the
rationale
of
jidoka
means
to
pressing
on
into
the
my
brokering
approach
on
the
top
of
the
manage,
aiming
an
orchestration
play
angle,
design,
resort
of
the
alto
services
with
the
proper
set
protocol
extension
tour
farang
enter
to
my
mekinese
for
resource
service
and
connectivity
information
discovery
explicit,
which
number
are
you
in
the
number
three?
B
Now
I
am
number
three
yes
right,
so
you
just
show
what
we
have
been
doing
on
the
single
is
life
basically
to
change
is
seeing
see
Roberto
Phil's.
Do
you
require
is
tension
section
for
the
other
property
map
was
removed
a
second
it's.
What
added
a
section
on
benefits
and
open
question
game
or
composite
activity?
B
B
Yes,
I'm
on
slide
five
right:
okay
in
the
description
section
when
Eliza
benefits
an
opening,
Mesa
single
professor
architecture,
for
example.
Here
we
describe
something
a
physical
block
addresses
in
your
detector,
such
as
about
the
distribution
of
the
topology
and
resource
information
in
a
peer-to-peer
fashion
and
the
step
taken
for
measurement
of
a
net
resultant
service
I'm,
maintaining
in
each
local
net
Orchestrator
on
a
privacy,
preserving
information
model
based
on
the
Alta
to
provide
apologies
and
service
information.
B
B
For
example,
what
kind
of
organization
will
manage
also
for
this
cooperation
of
a
broker
entity
for
this
question?
That
is
that
the
broke
identity
must
be
trusted
by
each
operator
sings.
It's
a
store,
a
hungry,
sensitive
information.
The
idea
is
to
hope
for
futures.
The
plumbing
of
Software
Defined
Networking
on
internet
checkpoint
can
be
use
it
as
a
trusted
party
platform
to
support
rich
business
model
between
difference
operator
in
case
of
the
peer
to
peer
information
reaching
model.
B
If
any
order,
of
course,
in
the
broken
entity,
the
information
reaching
among
all
involve
a
demand
will
be,
but
the
how
about
the
single
point
of
failure?
If
you're,
currently
team
and
they
df2
demanding
a
centralized
database,
local
restoration
will
application
option
may
be
applied.
Another
point
is
that
the
operator
have
a
preference
to
share
different
view
about
the
resource
service
formation
towards
differing
operator,
for
example,
a
detailed
view
for
the
operator
bellowing
to
the
same
operator
group,
a
high-level
information
towards
the
other
operator.
How
is
the
high
level
low
level
information
exchange
Anglin?
B
Lights:
seven,
that's
like
yes,
like
seven
red
seed,
basically
at
red
steps,
because
these
are
potential
introduced,
a
new
actor
services.
The
context
of
multi
domain
orchestration
is
enough
that
you
have
to
provide
stronger
scale
examples
to
support
the
creation
of
such
services.
Another
points
to
end
divide
witnesses
need
to
further
discussions
so
I,
just
such
as
the
problem
statement.
Challenges
is
openness
whose
benefits
immunology,
etc,
and
the
question
is
of
the
request
for
the
working
groups
to
provide
a
free
per.
Even
we
got
a
very
good
to
be
from
richer.
B
A
M
You
hear
me,
yes,
can
you
go
to
so.
M
A
N
M
You
have
a
minute
okay,
so
number.
Six
is
either,
as
example,
that
we
we
can
propose
a
design
extension
to
support,
multi-party
and
pass
vector
by
introduce
a
new
entity
domain
root
segment.
A
new
cost
man
mode
set
probably
is
not
supported
in
JSON,
but
it
conceptually
is
a
set
and
the
new
matrix
called
RSS
and
total
and
number
seven.
K
M
Either
it
is
an
extension
to
another
extension
of
auto
and
if
we
keep
doing
this
for
every
important
use
case,
we
may
end
up
with
many
extensions
with
the
chaos
of
dependency
and
compatibility
issues.
So
the
driving
question
we
are
trying
to
answer
in
this
draft
is
converted
and
a
unified
resource
representation
framework
in
Auto
to
provide
accurate
compact
resource
information
to
applications
who
may
have
a
very
wide
range
of
requires
requirements
and
objectives.
Next
slide.
Please.
M
M
What
kind
of
what's
the
least
of
the
certain
variables
related
to
this
pair
related
at
Curie,
and
then
we
introduced
a
different
entity
domain
called
constraint
and
use
a
constraint
problem
a
to
send
a
set
of
mathematical
constraints.
So
this
is
a
very
first
step
that
very
first
very
storm
and
solution.
We
propose
next,
please.
A
M
M
I
is
the
last
like
so
so
again.
This
is
a
very,
very
early
and
first
step,
and
there
are
lots
of
issues
need
to
be
addressed,
for
example,
if
this
design
of
maybe
using
generic,
but
it
may
be
two
giant
generic
and
we
saw
that
it
caused
privacy
and
security
concern
and
we
start,
we
started
started
with
the
returned
refunded
patients
and
by
the
auto
server,
but
I've
wanted
in
Johnson's
early
draft
apology,
judge,
judge
generic
multi-part
service.
M
Another
important
missing
piece
is
how
odd
hook,
client
Enterprise
the
requirements
of
what
kind
of
the
immersion
of
what
kind
of
information
is
needed
and
different
from
Johnson's
proposal.
We
have.
We
have
a
different
kind
of
CQ
like
a
language
grammar
specified,
but
we
haven't
fully
read
it
down,
because
we
are
also
debating
with
ourselves
why
this
is
a
good
language.
Why?
This
is
a
good
language,
the
expressor
so
are.
M
M
A
No
problem,
thanks
for
presenting
and
I,
think
we'll
take
the
discussion
on
this
draft
if
any
to
the
working
group
thanks
for
staying
a
few
minutes
over,
so
we
will
be
having
a
virtual
and
definitely
a
face-to-face
in
Prague,
and
if
there's
a
need,
we
can
see
what
to
do
for
a
face-to-face
in
Bangkok,
but
definitely
a
virtual
before
then.
Thank
you.
Much
have
a
nice
summer,
guys.