►
From YouTube: IETF99-ALTO-20170720-1330
Description
ALTO meeting session at IETF99
2017/07/20 1330
https://datatracker.ietf.org/meeting/99/proceedings/
A
A
B
B
You
are
done
right,
yeah,
good
I
think
you
were
last
time.
I
met
you
sometime,
yeah,
I,
think
so
yeah,
okay,
so
I
think
we
should
start
and
welcome
everybody.
This
is
the
outer
working
group
and
my
name
is
hands
it
off
I'm
carrying
this
working
group
together
with
VJ
goo
Bonnie,
who
unfortunately
cannot
make
it
he
will
join
em.
B
I
hope
you
all
know
about
the
note
well
and
I
think
you
know
what
it
means,
but
just
to
recap
it
basically
means
that
if
you
are
aware
of
any
IPR
related
issues,
if
you
are
aware
of
any
IPR
that
relates
to
what
we
present
you
and
talk
about
here,
just
you
should
you
must
make
the
working
group
aware
of
it
and
if
you
don't
know
what
it
means.
Please
read
those
RFC's
they're
closely.
It's
basically
sets
the
ground
roots
for
this
meeting
works
and
so.
B
We
as
chairs,
let
me
just
work
this
there's
our
area
director
hai,
which
we
went
ahead
and
started.
Is
that
okay
yeah?
Sorry,
sorry,
that's
just
how
we
are.
B
Okay,
so
it's
a
good
idea
that
we
went
head
yeah,
good
so
and
anyway,
so
the
status
of
the
working
regen
I
had
a
Skype
call
just
the
other
day,
and
we
were
talking
about
how
we
see
the
working
groups,
oh
really,
and
we
really
want
to
focus
on
finishing
the
Charter
items.
I
think
this
is
what
we
mentioned
also
at
the
last
IDF
meeting,
because
a
lot
of
this
work
is
actually
quite
mature
and
we've
been.
B
We
have
not
been
dragging
it
along,
but
it
has
been
going
on
for
several
years
actually
now,
and
so
we
really
want
to
focus
on
finishing
the
Charter
items.
So
this
time
we
set
up
the
agenda
very
much
accordingly
to
that,
so
the
Charter
and
milestone
items
come
first
and
we
really
want
to
have
a
good
time
discussing
those
so
that
we
can
make
sure
which,
how
they
are.
B
What
are
the
next
steps,
so
we
can
really
move
towards
finishing
them
and
everything
else
on
the
agenda
is
kind
of,
and
we
are
of
course,
a
large
and
the
time
for
other
interesting
new
topics,
but
that's
best
effort.
Basically,
after
the
Charter,
Islands,
@bj
and
I
actually
am
really
have
the
goal
of
finishing
the
milestones
by
or
at
ITF
101,
so
I,
just
as
a
look
ahead,
it's
quite
likely
that
both
of
us
will
not
make
it
to
ITF
Singapore.
B
So
we
will
probably
have
a
virtual
interim
meeting
instead
of
the
next
ITF
meeting
just
for
people
interesting
and
for
people
doing
the
work
in
the
working
group
and
but
we
both
plan
to
be
at
ITF
London.
So
our
our
goal
is
by
that
time.
To
finish
all
the
milestones
comments,
Bertrand
asked
quick
question
so.
B
Working
group
or
let's
leave
it
at
finishing
right
now-
I
wouldn't
nail
it
down
so
much
I
mean
if
we
have
something
that
is
close
to
working
group
last
call
and
we're
waiting
on
one
final
review
or
something
that
that
that
should
work
I
mean
you
know.
You
know,
then.
If
let's
say,
if
one
item
is
not
finished,
we
can
always
finish
it
without
having
working
group
meetings,
but
we
really
want
to
concentrate
this
meeting
and
the
one
we
have
in
London
on
finalizing
the
milestones
we
have
and
and
obviously
yeah
I
mean
principle.
B
Finalizing
means
done
with
working
with
last
call.
But
let's
see
how
far
we
get
so
yeah
here's
the
agenda
and
as
I
was
mentioning
and
before
we
pretty
much
did
the
agenda
in
a
milestone,
oriented
fashion,
so
the
first
I
think
seven
or
eight
presentations
are
really
the
milestones
we
have
and
we
want
to
move
those
ahead
and
then
afterwards
we
have
I,
think
three
or
four
and
other
presentations
or
new,
interesting
stuff
or
stuff
that
has
been
around.
But
it's
not
chartered.
So
that's
like
I,
said
more
or
less
best
effort
for
now.
B
No,
that
doesn't
seem
to
be
the
case,
and
so
what's
the
progress
in
last
ITF
meeting,
we
actually
adopted
3m
working
group
up
working
group
items
for
actually
chartered
items
or
milestones
that
we
have
I,
think
the
CD
and
I
request
routing
and
that's
partially
my
fault
and
also
the
unit
I
fight
props.
The
authors
actually
made
updates
but
forgot
to
make
an
IETF
working
group.
You
know
make
the
new
name
with
ITF,
but
we
now
did
it
I,
think
Richard.
You
did
for
both
documents.
B
So
now
that
is
actually
done,
and
so
here
the
milestones
and
what
BJ
and
I
did
when
we
prepared
this
meeting,
we
we
we
try
to
give
you
an
overview
in
the
next
few
slides
on
how
we
start
with
respect
to
the
milestones,
so
the
submit
cause
property
extension
document
from
our
perspective
is
Stan
and
Sabina
actually
will
not
present
on
that
one
today,
because
multicast
is
now
in
RFC,
editor
queue
and
so
from
from
the
working
perspective,
we
are
done
with
it.
B
There
are
just
minor
knits
here
and
there
I
mean
I'm
following
the
emails,
but
it's
not
really
something
we
need
to
discuss
here.
If
you
look
at
the
other
M
charter,
my
item
set,
who
or
milestones
that
we
have
here
from
alternative
server
discovery
document
down
to
a
post
property
extension.
Here
you
see
the
documents
we
have
and
the
algebra
formance
metrics
and
we
think
needs
some
more
working
group
scrutiny,
but
the
other
ones
that
you
can
see
a
cost
calendar
and
especially
also
incremental
updates
and
Ekstrom.
B
We
would
like
to
move
to
working
group
last
call
during
this
meeting
and
BJ
also
send
in
mail
on
the
mailing
list
and
I've
seen
some
nice
discussion,
and
we
actually
had
a
lot
of
reviews
lately
also
for
these
documents-
I'm,
not
so
sure,
about
unified
props
and
maybe
when
you
present
Richard
M.
Maybe
that
probably
needs
also
a
little
more
time
until
they
can
go
to
worse
working
class
cause.
That's
not
entirely
correct.
Here.
Hat
factor
is
also
something
that
probably
needs
a
little
bit
more
work
and
the
CDN
request
routing.
B
It's
also
something
that
needs
a
little
bit
more
work.
So
basically,
we
have
six
core
milestones.
Three
of
them.
We
want
to
move
to
move
to
work
in
glass
column
we
actually
and
I've.
Seen
on
the
email,
there's
lots
of
agreement.
They
are
quite
mature.
So
let's
see
how
the
discussion
goes,
and
then
we
have
three
more
working
group
milestones.
Sorry,
where
we
actually
have
working
group
items
that
by
the
next
two
IDs,
we
really
want
to
get
done,
yeah
any
other
business.
Any
other
comments.
C
Doesn't
Richard
yeah
I
do
want
to
say
a
couple
of
things
that
would
be.
One
is
roughly
one's
good
time
to
talk
about
the
interim
meeting
for
IETF,
because
I
think
you
mentioned
you
and
we
cannot
attend
likely
not
can
IVF
100.
So
roughly
what
time
schedule
is-
and
secondly,
I
do
want
to
mention-
is
we're
running
a
time
mini
semi
with
people
from
Cal
Tech
from
a
lot
of
people
and
time
we
mostly
around
made
August
in
Los,
Angeles
relief
it
without
or
impractical,
Italy's,
auto
design
to
apply
the
ink
in
the
contact
HC.
C
B
I
mean
certainly
I
would
encourage
you
to
mention
the
stuff
you're
doing
in
regarding
the
large-scale
applications
and
it's
the
workshop
kind
of
thing
you're
talking
about
right
to
mention
that
on
the
mailing,
this
explicitly
so
that
people
are
aware
and
regarding
regular,
auto
entering
we
reach
and
I
were
thinking
about
something
in
well
instead
of
a
physical
meeting
in
Singapore.
So
we
were
thinking
about
something
in
October
or
a
million
knows
the
deadlines
more,
but
in
principle
you
can
have
a
virtual
meeting
instead
of
the
regular
one.
B
You
just
need
to
set
it
up
on
time
and
we
have
to
check
with
you,
but
basically
we
don't
want
to
meet.
If
we
don't
want
to
have
a
physical
meeting
in
Singapore
because
most
likely
we
will
not
be
there.
So
it
will
be
around
that
timeframe
and
that's
what
we
are
planning.
But
we
can
have
discussion
on
emailing
this.
If
you,
if
you
have
enough
people
that
go
for
a
different
meeting
already
earlier
for
sure
that's
we
can
have
that.
We
can
do
that,
want
to
comment,
but.
D
B
D
D
B
But
people
have
probably
closer
to
the
next
meeting,
so
probably
October
time
frame
is
then
what
we
have?
What
we're
thinking?
Yeah,
okay,
yeah
I
mean
if
there's
no
other
business,
anybody
that
anybody
wants
to
mention
I
think
we
should
go
quickly
to
the
first
presentation:
that's
cost
calendar,
so
Sabine,
I,
I,
hope
I.
Have
your
slides.
Are
those
the
correct
sites.
C
E
Yes,
it's.
This
is
about
the
latest
updates
on
coach
calendar
and
since
last
night,
it's
here
so
next
slide,
and
also
previously
it
was
version
1,
but
we
updated
on
February
we
collected
since
the
last
meeting
we
collected
whatever
feedback,
and
so
the
deeps
between
the
two
versions
are
mainly
InDesign
updates
for
backwards,
compatibility
and
compatibility
with
multi,
just
dropped
organization
and
rewarding
and
cleanup,
and
by
the
way,
great
thanks
to
the
whole
working
group.
You
said
we
it's
like.
E
The
Canada
attribute
object
in
a
resource.
Entry
now
has
a
mandatory
member
orko
spy
planes,
which
is
an
array
of
one
or
more
elements
indicating
the
custom
name
in
the
IRT
entry
to
which
these
attributes
are
tied.
This
is
typically
in
case
you
have
you
require,
or
you
want
to
require
more
of
multiple
metrics
and
also
to
be
self-contained.
So
next
slide.
E
Okay,
so
menu,
we
have
other
design
updates
and
the
request
format
so
that
we
set
now
our
rule
on
the
members
in
the
request.
So
given
that
n
is
the
number
of
requests
e
metrics,
so
we
introduce
a
an
array
of
n
boolean
value
that
just
say
whether
or
not
you
want
these
very
metric
or
not,
and
with
a
couple
of
rules
saying
that
we
already
must
coming
exactly
n
boolean
value.
Otherwise
the
server
will
return
an
error
and
in
this
field
is
not
present
at
all.
E
E
E
E
E
So
yes,
then
we
did
some
draft
organization
update.
So
we
dropped
the
very
lengthy
section
5
with
the
new
spaces,
because
all
these
use
cases
are
actually
for
the
other
sections
and
if
there
was
a
section
for
the
3,
that
was
a
recap
of
rules
related
to
outer
coat
calendar,
which
was
dropped
again
as
well,
and
the
text
was
moved
to
the
section
to
the
to
the
tool
in
the
design
part
and
on
liability
with
legacy
other
projects.
E
So
next
slide,
yes,
and
we
are
rewarding
well.
In
particular,
we've
changed
the
example
metric
name
which
is
available
bandwidth,
because
this
one
is
actually
used
in
the
performance
metrics
draft,
and
then
this
is
actually
a
metric
that
is
used
for
links.
So
we
prefer
to
pick
something
called
banded
score,
which
is
something
that
is
abstracted
again.
It's
an
example,
but
we
illustrated
it
is
as
a
completely
abstract
expulse,
a
between
0
and
20
of
0
and
500.
E
E
So
then
we
would
have
to
add
a
little
bit
of
text
to
specify
the
different
search
between
the
link,
labeled
metric
and
entry
and
path
metric.
Shall
we
also
consider
also
Canada
phone
for
endpoint
properties?
Do
we
consider
endpoint
properties
that
whose
values
are
changing
over
time,
such
as
I,
don't
know
color,
location
connection
time
with
respect
to
time
period
such
as
morning.
E
Evening-
and
there
was
a
working
group
ease
question:
is
it
possible
that
the
outer
circle
has
several
options
of
different
combinations
of
pain,
interval,
size
and
number
of
these
accounts
for
auto
clients
allowing
the
auto
client
to
choose
or
pick
from
all
fine-grained
codes?
Canada?
The
my
first
answer
was
yes,
the
server
can
afford
the
workload
it
may
propose
different
Canada
attributes
for
a
same
cosmic.
Do
you
have
a
menu
of
calendar?
E
You
pick
whatever
you
want,
but
then
the
issue
is
that,
first
of
all,
the
client
request
should
stay
simple,
and
this
would
actually
require
to
change
the
design
and
format
of
our
two
requests
and
responses,
and
so
the
basic
one
of
the
design
options
was
really
to
just
give
you
a
menu,
and
you
pick
what
you
have
on
the
menu
too
much
to
me
to
to
remodel
the
options
so
well
I'm
open
to
the
discussion.
But
my
point
right
now
is.
B
B
E
C
And
I
can't
go
back
to
one
slides,
so
here's
a
while
opinion
and
about
of
course
we
can
talk
about
applying
as
well,
is
several
options
of
a
time
interval
size
times.
Production.
A
central
number
of
interval
essentially
is
I.
Think
the
design
question
here
is:
do
you
think
we
should
allow
the
server
to
define
different
levels
of
granularity?
Essentially,
it's,
for
example,
so
say:
I
can
provide
a
one-minute
five
minutes,
20
minutes
one
hour
units
and
therefore
keep
it
still
tell
your
whole
thing
about
one
day.
C
So
one
so
I
agree
and
you
need
discussion
that
the
request
should
be
simple.
So
there
were
a
lot
of
options,
probably
complex.
So
it's
the
okay,
for
example,
for
the
working
group
to
define
the
following,
for
example,
design
decision
that
the
server
should
always
give
the
finest
green
result
and
therefore
the
client
can
do
the
aggregation.
C
Would
that
be
a
reasonable
compromise
space
I
mean
if
I'm
server
I
can
really
collect
data
at
a
one
minute
interval
our
provider
by
minute,
and
you
guys
can
essentially
to
aggregation
to
gather
five
minutes
or
one
hour
or
one
day,
interval
and
I.
Think
the
the
size
had
me
not
to
be
too
big
because
it
costs
Canada
by
itself
or
he's
another
China
data
probably
blow
up,
even
though
by
a
factor
of
five,
but
overall,
it's
small.
So
therefore
may
not
be
too
bad.
So
therefore,
that's
my
opinion.
B
So
this
is
the
a
members
chair
head
on,
so
in
order
to
move
this
forward,
I
think,
if
both
of
you
agree
and
there's
no
other
seems
to
be
no
other
opinion
in
the
room.
Why
don't
you
just
state
that
on
the
mailing
this
and
if
nobody
objection-
and
we
move
ahead
like
this-
for
me,
the
the
issue
is
rather
and
how
many
iterations
do
we
need
to
get
this
done
so,
do
you
guys
agree
that
you
are
close
to
getting
done
and
I
mean?
B
E
C
F
C
B
No
and
I
I
completely
understand
the
issue,
but
so
the
first
question
is
other
opinions
in
the
room
of
obviously,
and
if
they're,
if
they're,
not
I,
think
it
seems
to
be
that
you
and
sabine
richard
agree.
And
so
maybe
we
can
have
a
little
bit
of
discussion
on
this
on
the
mailing
list
and
you
just
state
your
way
forward.
B
And
since
it's
a
working
group
item,
then
everybody
has
to
call
a
chance
to
comment
and
once
we
are
done
with
this
issue
and
them
other
minor
issues
that
mean
that
you
mentioned,
then
we
can
move
to
working
class
call
and
it
doesn't
seem
to
be
completely
finalized.
But
I
would
like
still
like
to
have
a
ham
in
the
room.
F
B
Are
done
think
that
the
document
is
ready
for
working
class
call
I
can
from
I
can
take
my
chair
hat
off
and
I.
Think
there's
been
lots
of
discussion
on
the
mailing
list.
This
document
has
been
around
for
a
while
I
mean
we
definitely
agree
on
the
approach
and
it
seems
to
be
mature.
So
I
just
want
to
like
to
know
of
people
agree
that
once
these
issues
have
been
resolved,
we
can
go
to
work
with
less
call
yeah
yeah,
yeah.
B
Yeah,
thank
you.
Thank
you
for
reminding
me
of
my
powers.
I
feel
good
I
believe
in
democracy.
So
and
I
say
what
I
say
is
right,
but
since
we're
here
and
I
think
we
should
do
a
ham
again,
we're
anywhere
only
humming
off.
If
people
think
you
know
these
are
minor
issues
and
then
we
can
sue
move
to
work
in
class
call.
We.
B
Anyway,
think
mayor,
we
posted
this
already
on
the
list,
so
yeah,
so
M
can
I
am
who's
in
favor
of
moving
this
soon
to
working
class.
Call
please
now
who
is
not
in
favor,
please
hum
now,
okay,
so
we
have
the
backing
of
the
working
group,
which
is
always
nice.
You
know,
I
can
sleep
much
better
now
so
yeah.
If
there's
no
further
discussions,
I
mean
thank
you
very
much.
So
I
think
this
is
and
from
a
chair
perspective
or
from
a
moving
forward
perspective.
I
think
this
is
what
we
wanted.
B
It
seems
there's
some
minor
issues
so
I'm,
you
know
please
resolve
them
and
and
impost
how
you
resolve
them,
I
think
so
Richard.
What
you
were
saying
is
these
are
design
choices
that
the
working
group
needs
to
be
aware
of.
So
please
make
them
explicit
on
the
mailing
list
and
do
an
update
and
as
soon
as
you
make
the
next
update,
we
can
go
to
working
class
call
and
we
will
handle
it
on
the
mailing
list.
Then
yeah.
Okay,
are
there
any
other
comments
on
this
document?
B
If
not,
then
sabine
you
can
stay
you.
You
still
have
a
slot
for
em
context.
Right
calendar
context,
it
just
to
be
clear.
This
is
a
different
document,
but
since
it's
very
related,
we
put
it
on
the
agenda
right
afterwards.
It
doesn't
make
sense
to
present
it
later
on
because
it
very
much
relates.
But
for
now
the
document
you
just
presented
right
now
is
the
one
we
see
as
the
milestone
and
we
have
to
see
how
we
move
the
other
one
forward
right.
But
you
agree
with
that.
You
right
that's
mm-hmm,
so
that's
cost
context.
E
So
next
slide
please
so
this
actually.
So
this
is
about
the
use
cases.
I
skip
the
first
one,
which
was
the
first
motivation.
Well,
there
are
several
motivations
actually
to
use
them,
and
so,
as
you
mentioned,
there
was
the
the
the
need
to
have
something
more
fine-grained
than
a
cost
calendar,
even
with
very
small
time
slots
such
that,
such
as
like
10
seconds
or
15
seconds,
but
the
the
the
the
story
was
about
a
Yui
willing
to
have
a
calendar
of
I
know
how
nice
the
connection
quality
is.
E
For
example,
there
is
another
use
case
which
is
even
older,
which
is
related
to
CDN
applications,
and
so
some
people
raised
the
need
to
have
a
cost
map
where
values
were
depending
on
qualitative
criteria
such
as
policy.
So
the
feeling
is
that
it
is
cumbersome
to
define
a
metric
associated
to
context,
c1
or
c2
or
CN.
It
is
cumbersome
to
get
any
cost
map,
especially
because
of
time,
because
of
volume,
and
actually
the
client
would
be
better
off
just
having
one
metric
value
with
n
options.
So
next
slide.
E
Ok,
so
I
introduced
here
the
new
use
cases
that
are
motivated
by
the
discussion
with
the
path
aware
networking,
research
group-
and
there
was
a
meeting
yesterday
with
a
couple
of
nice
presentation
and
one
of
them
was
actually
addressing
the
problem
of
an
end
system
and
host
trying
to
figure
out
the
capability
or
the
quality
of
the
several
party
possible
path.
It
could
actually
take
well,
especially
here
with
respect
to
connection
so
so
this
is
actually
related
to
the
use
case
to
in
the
previous
slide.
E
So
you
would
like
to
figure
out,
shall
I
use
I,
don't
know,
shall
I
use,
LAN
shall
I
use
my
phone,
shall
I
use
Wi-Fi
and
on
which
equipment
and
actually
do
I
have
an
option
at
all,
so
that
was
actually
yeah.
It
relates
to
multi-party,
but
the
purpose
is,
is
not
here
to
establish
multiple
paths
it's
to
have
to
allow
the
end
host
to
have
a
clue
of
what
how
the
properties
on
some
of
the
paths
are
so
next
slide.
Please.
E
Okay,
so
the
approach
to
address
these
use
cases
is
to
extend
the
cost
information
specified
in
the
base
protocol
by
providing
so
far
a
same
cost.
Metric
several
possible
cost
value
where
each
value
does
depend
on
qualitative
criteria
as
opposed
to
time
based
criteria.
So
it
does
not
depend
on
time.
It
depends
on
like
access
or
SLA
or
UGA
or
what-have-you.
E
So,
to
this
end,
we
propose
to
add
a
member
called
cost
context
associated
to
a
cost
type
name
in
the
IRT.
So
this
cost
context
has
members
cost
ID
names
and
and
a
set
of
context.
Parameters
which
are
is
a
an
array
of
attributes
array
who
just
indicate
with
respect
to
what
you
can
have
the
value
and
where
these
attributes
can
be
combined
among
the
attributes
arrays.
So
we
also
add
a
member
context.
E
E
E
Some
Wi-Fi
line
land
with
SLA,
one
SLA
to
exit
a3
and,
for
example,
the
client
can
request
to
combination,
so
I
want
value
for
sale
and
SLA
3,
and
also
for
Wi-Fi
and
SLA
3,
and
then
in
the
response.
You
have
something
like
that.
So
this
is
the
context,
parameter
combination
that
has
been
selected
by
the
client,
and
so
it's
the
same
as
in
the
request
input-
and
here
you
have
the
two
values
corresponding
each
to
the
combination.
So
next
slide.
E
So
since
the
first
version
presented
at
the
in
Chicago,
we
have
addressed
the
working
group
concerns
in
particular
on
the
near
real-time
changes
in
wireless
networks.
So
in
section
2.1
we
have
clarified
the
meaning
of
the
example.
Cellular
connection
cost
and
we
have
named
it
abstracted
air
f
cost,
because
air
f
cost
has
a
connotation.
That
may
be
worrying
too,
especially
as
auto
is
not
supposed
to
be
a
real-time
information
provider.
E
So
we
also
added
definition
on
drools
on
context,
attributes
and
backwards
capability
compatibility.
Sorry.
So
in
section
4
we
mentioned
that
an
auto
service
must
provide,
of
course,
values
as
in
in
the
base
but
protocol.
If
no
context
attributes
have
been
requested
or
if
it
is
not.
If
the
client
puts
a
combination
that
is
not
supported,
so
we
updated
the
example.
Combination
in
the
ILD
use
case,
1
and
updates
it
also
the
filtered
cost
map
requests
and
clarified
some
rewarding.
So
next
slide.
E
E
Related
to
a
list
so
next
slide
so
again
in
the
in
the
draft.
So
you
there
is
also
the
possibility
that
the
there
is.
The
client
just
extensively
indicates
all
the
combination
it
wants,
meaning
you
da
you
da
and
uplink,
and
so
and
so
the
response
of
course
conveys
all
this
combination,
but
we
feel
that
this
may
be
like
too
long,
so
we
are
trying
to
come
up
with
a
smart
design.
So
every
suggestion
is
more
than
welcome.
So
next
slide.
E
Remember
if
I
remember
so
about
the
flexibility,
the
lack
of
flexibility
to
have
like
a
list
of
hard-coded
parameters
so
because
the
first
idea
was
to
have
only
a
little
number
of
values.
So
the
question
would
be:
can
a
context
parameters
take
numerous
values,
and
then,
in
that
case,
is
it
isn't
it
actually
rather
a
qualitative
property
or
a
qualitative
metric?
So
that's
the
discussion.
Actually,
we
I
think
we
need
to
have
and
yeah
if
you
couldn't
go
back
one
slide.
E
F
F
Yes,
this
one,
this
wrong
is
that
the
combination
in
each
array
is
an
O,
and
the
combination
between
the
array
is
an
end.
One
suggestion
is
that
maybe
we
should
make
a
specification
for
the
order
of
this
combination,
for
example,
and
the
return
value
of
the
order
of
the
returned
value
can
be
the
UDN
and
uplink,
then
your
the
end
and
downlink,
but
it
can
also
be
the
UDN
and
uplink
and
then
the
yurina
beauty
in
a
uplink
right.
F
E
Started
to
play
with
the
three
arrays
and,
of
course
we
need
to
specify
rules
allowing
the
client
and
the
server
to
understand
how
the
combination,
what
the
combination
is
and
what,
how
we
map
it
to
the
array
of
return
values,
that's
so
sure,
and
so
definitely
well.
This
is
actually
more
at
least,
and
of
course,
we
should
also
set
rules
saying
that,
okay
within,
for
example,
you
should
not
have
an
array
with
UDI
and
uplink
I
mean
if
you
refused.
E
B
F
B
E
Some
of
this
stuff
is
private,
but
yeah.
There
is
the
concern
that
somebody
can
capture.
So
if
my
application
within
my
phone,
for
example,
sends
a
request,
somebody
could
capture
and
find
out
while
which
SLA
I
have,
but
on
the
other
hand
it's
well,
then
I
think
the
client
should
be
aware
of
that,
and
if
the
client
does
not
want
to
get
the
cost
with
respect
to
the
SLA,
then
it
does
not
request
it.
So
yeah.
B
On
so
I
remember
in
the
beginning
of
the
other
working
group
ridgemount,
he
always
would
say
that
he
doesn't
want
other
people
to
know
what
ISP
contract
he
has,
because
he
would
think
if
he
has
the
premium
contact
and
burglary
is
more
probable
and
things
like
that.
Whatever
that
means
I
mean
just
trying.
F
F
B
E
C
So
type
student,
of
course,
is
a
very
general,
very
big
prom
in
computer
science,
I
get
a
big
part
of
a
computer
science
program.
Language
is
about
type
systems
right
because
the
philosopher
can
process.
All
word
lately
is
used
syntax
to
impose
semantics
and
catch
all
bugs,
and
so
that's
why
you
move
to
other
strong
types,
design
and
so
on.
The
very
early
part
of
ours.
Ours
is
seventy
to
eighty
five
uses.
C
What
I
will
probably
call
a
hybrid
type
system
hybrid.
You
know
for
nonsense.
Example
for
every
cost
type.
We
have
two
members
and
the
Casa
metric
and
Cosmo
Cosmo
indicate
the
datatype
exam,
a
numerical
or
ordinal
and
cosmically
denoted
semantics.
We
do
a
cross
product
to
indicate
like
a
type
system.
So
that's
way
we
import.
This
is
called
a
hybrid
and
because
we
don't
want
to
go
over.
For
example,
like
a
Java
which
in
haven't
you
fine,
you
cannot
even
define,
for
example,
of
this
type
of
system.
C
You
must
say,
for
example,
you
cannot
see
end
or
wanting
for
Alice
Amanda.
Give
you
cannot
see
is
the
temperature
temperature
would
give
you
most
man
takes
like
type
system
and
for
this
one,
if
we
add
I,
bad
or
I,
think
a
form
cementing
polynomial
contacts.
A
parameter
is
wonderful,
is
very
helpful
or
from
the
grammar
point
of
view,
and
then
we
probably
don't,
will
start
to
read
a
break.
C
The
types
in
a
sense
that
the
return
value
now
would
be
a
array,
but
a
Cosmo
will
still
be
in
America
ordinal,
so
which
means
now
one
part
of
type
students
say:
okay,
it's
a
number
and
the
other
halfs.
Oh,
my
grades
are
ready.
We
start
to
really
see
such
a
break.
I
don't
have
the
subs
from
this
very
complex
problem
and
but
overall
I
think
that's
probably
your
merging
is
that's.
Why
you
like
clever
encoding
but
very
likely,
my
god,
the
feeling
is.
C
If
we
could
move
beyond
to
substantial,
we
probably
come
up
with
another
type
system
and
I.
Try
to
look
at
around
even
per
exam
types
in
some
Yamato
and
so
on.
They
also
cannot
solve
your
problem.
It's
mostly
elaboration.
You
want
it.
You
want
to
avoid
generate
a
here
a
number
of
times.
For
example,
some
strong
types
is.
B
C
C
Fewer
discussion
very
likely,
for
example,
what
I'm
thinking
of
but
there's
a
big
reason
why
Java,
eventually
all
C
plus
a
go
to
a
fog,
lamp,
directly
attacks,
complete
stuff
and
that
proudly
conserves
from?
But
where
you
probably
should
talk
about
this
one.
And
of
course,
if
we
are,
we
want
we're
willing
to
abandon
or
because
I'm
fully
stretch
and
they
type
system
we're
okay,
but
otherwise.
I'll
probably
need
to
go
to
more
complex
type
system,
but
I
think
from
a
simplicity,
is
cement
upon
where
this
is
a
very
useful.
There's.
B
F
B
E
There
would
be
a
need
to
be
registered
because,
as
father
the
the
other
working
draft
may
end
up
that
we
were
required
to
instead
of
using
Wi-Fi
and
LAN
to
use
something
like
blue,
oh
sweet
and
salt.
So
I
think
we
well
with
respect
to
everything
relating
to
seminar
and
wireless
with
a
the
next
step
will
also
include
looking
at
the
other
IETF
work,
to
see
whether
they
have
registered
such
property
somewhere
and
also,
and
if
not
or
if
needed,
consider,
also
registering
some
of
these
properties
to
IANA.
E
B
I
think
we
close
this
discussion
now
because
we're
running
a
little
bit
out
of
time-
and
this
is
not
currently
not
a
my
son
item,
but
thank
you
very
much
Sabine
from
from
a
chair
perspective,
I
think
this
is
interesting
work
and
but
again
it's
not
so.
Currently
the
milestone
is
addressed
by
your
other
documents.
B
So
we
have
to
see
how
this
moves
along
I
guess
for
now
we
just
you,
you
should
move
it
ahead,
because
certainly
there
seems
to
be
interested
in
it,
but
we
have
to
see
how
that
fits
in
the
current
charter
or
if
we,
how
to
proceed
with
it
right,
okay,
okay,
so
next
up
is.
Let
me
look
at
the
agenda.
The
next
one
is
incremental
updates,
Richard.
B
C
I'll
talk
about
incremental
updates
and
I
believe
this
is
a
very
important
extension
to
also,
and
we
have
done
a
lot
work
already
and
I
will
always
see
that
is
essentially
the
main
driving
force
is
always
windy
and
share
your
very
first
version
and
mostly
I'm,
trying
to
take
over
as
to
do
a
lot
of
revisions
about,
of
course,
that
you
give
should
come
back
soon.
Okay,
so
next
slide,
please
so
here's
overview
of
what
we
have
changed,
because
it's
a
very
important
capability.
Therefore,
it's
making
a
lot
of
cheese.
C
C
C
A
drive
to
the
change
is
handle
new
use
case,
for
example,
the
city
I
use
case
turns
out
to
really
be
very,
very
beneficial
to
occur
between
author
working
group
and
see
a
working
group,
because
a
change
allowed
made
a
lot
of
decisions
eventually
to
be
coming
some
other
way.
We
can
anime
clothes
design
decisions,
and,
to
summarize,
we
made
four
types
or
changes
for
this
document
and
while
make
more
generic
statements
and
could
be
more
generic
number
two
is
we
want
to
make
a
benefit
to
be
more
expedient.
C
Some
people
still
not
fully
clear
about
the
benefits
and
I
restart
a
document
and
make
the
text
more
friendly,
but
during
launch
the
day
and
we
stop
being
and
maybe
I
get
bad
job
go
back
some
lachan
this,
because
I
will
try
and
make
it
either
great
and
last
major
change.
Is
we
finally
decide
to
allocate
some
patch,
so
that
is
a
major
thing
to
grecian
effort.
So
what
else?
If
you
give
more
words,
so
next
slide,
please
so
now.
C
Let
me
first
say
a
few
words
about
the
generic
changes
and
I
would
like
to
get
approval
from
the
working
group.
So
here's
compares
I'm
using
our
c
diff
to
really
show
that
type
of
changes,
I'm,
making
I
hope.
Other
people
are
also
easier
to
see,
because
there's
some
comments
last
time
was
is
incremental.
Updates
it's
actually
really
limited
to,
although
all
can
be
more
broadly
and
therefore
the
major
change
is
we
want
to
emphasize
that
the
designs
will
come
jr..
C
So,
therefore,
if
you
look
at
this
document,
for
example,
here's
I'm
only
giving
you
a
very
simple
example
of
changes
which
we
made
and,
for
example,
the
right
hand.
Side
is
an
older
version.
Oh
five,
the
life
and
inside
is
a
newer
version,
which
is
oh
seven
and
even
look
at
the
first
paragraph.
For
example,
look
at
first
paragraph
I,
love,
abstract
and
the
old
abortion
would
say
alcohol
making
decisions
until
I
end
Otto
server
provides
Network
and
custom
apps
and
user
maps.
C
Another
client
can
determine
the
cost
and
make
a
decision
and
so
on,
and
if
you
look
at
the
newer
version,
we
suggested
also
it's
providing
information
resources
and
networking
apps
and
the
customer
is
essential,
just
simple
examples.
So
there
are
a
lot
of
such
changes
across
the
document
and
we're
trying
to
make
sure
that
we're
always
talking
about
the
generic
resources
and
not
specific
in
the
custom
apps
or
so
that's
a
one
type
of
meter
change
next,
one
please.
C
C
Therefore
we
want
to
get
you
two
component
across
now
for
solid
story
structure
document
to
make
sure
that
those
pounds
are
get
across
very
clearly,
for
example,
even
look
at
document
over
here,
we
said
to
achieve
two
benefits:
wine,
and
you
know
Chris
document
we're
trying
to
use
a
little
slightly,
less
structured,
working
and
so
on.
I
know
we're
talking
that
this
one
it
really
can't
get
across.
So
that's
the
second
immediate
chain
is
again
around
across
the
whole
document,
which
is
make
it
clear
next
slide,
please.
C
So
next
one
is
a
major
restructure
and
I
would
like
to
get
a
feedback
and
get
a
comments
from
people.
So
here
is
the
right-hand
side.
Eighth
is
old,
structure
and
live
hand.
Side
is
a
newer
structure.
So
if
you
compare
this
to
structure
changes,
you
can
see
the
later
product.
We
don't
make
a
lot
of
changes.
C
For
example,
one
big
part
is
we
move
a
lot
of
stuff
into
a
single
background
session,
so
we
add
a
term
section.
We
move
a
lot
of
atom.
Ss
e
JSON,
merge
patch
and
do
some
hatch
into
the
single
background
section
in
Section.
Four
and
you
know,
careers
design,
I
could
they're
spread
out
all
over
the
other
sections
and
nobody
said
that's.
C
It
really
move
everything
into
the
single
background
and
then
we
have
overview
of
the
background,
and
you
know
the
old
version
is
we
have
the
ghini
of
assistant
exam
section
to
not
become
section
five
and
this
is
may
or
may
not
design,
because
maybe
people
now,
for
example,
when
people
reading
this
and
your
version
is
Oh
what
exactly
bigger
picture.
So,
therefore,
if
people
can
really
give
us
a
little
bit
feedback
about
which
won't
work
out
better
I
think
we
are
more
than
happy
to
really
take
a
feedback
and
revise.
So
that's
a
major
change.
C
Okay
next
slide,
please
so
next
one
and
which
I
really
want
emphasized
is
finally,
we
have
decided
to
add
into
JSON
into
okay.
The
motivation
is
very
simple:
all
working
group
all
well
isla
falling
multipack,
IP
personal
I,
believe
it's
quite
elegant.
In
many
many
ways
you
have
a
object
model
and
you
say
you
send
a
multipack.
We
don't
even
see
much
at
all
and
you
March
the
car
model
and
you
pack
incremental
you
get
a
dart
encoding,
very
beautiful,
very
elegant
and
non
nonsense.
C
Words
at
all,
and
but
unfortunately,
this
one
cannot
handle
or
race
very
well
for
every
single
array
you
between
a
single
element
you
might
catch.
You
have
to
replace
the
whole
array
and
array
can
be
a
lot,
for
example,
1000
or
2000
numbers,
and
you
have
to
replace
the
whole
thing,
the
benefit
of
JSON
packed
with
RFC
69
OH.
C
But
you
cannot
do
like
a
Delian
element,
but
by
introducing
jeezum
patch-
and
you
can
see,
operation
is
replaced
and
path
is
essentially
it's
essential
in
the
XPath
model
and
go
to
all
wheel
all
the
way.
For
example,
this
element-
and
you
can
see
remove
so
there
was
suddenly
now
we
can
handle
all
or
risk,
which
is
incredible.
It
new
capability
and,
of
course,
if
it
as
a
comparison
for
people
who,
like
elegance
personally
less
elegant,
become
now
for
every
single
case
being
specified.
C
C
So
because
of
this,
a
change
and
there's
a
basic
cascading
changes
across
the
document.
For
example,
if
you
look
at
the
old
document
in
the
capability
in
terms
of
information
resources,
you
know
the
one
you're
always
able
go:
hey
I'm
a
server
I
can
specification
more
patch,
based
on
as
my
immediate
I
could
really
change
it
in
the
dual
incrementally
encoding,
but
not.
C
More
so
now,
suddenly
we
have
to
really
the
server
say:
I
can
encode
all
kind
of
types
now
and
all
authority
station
will
become
more
complex.
Okay
next
slide,
please,
okay,
so
those
are
the
major
changes
alibis.
The
document
has
been
pretty
pretty
stable
for
a
while
now
and
those
are
the
major
next
steps
which
were
planning
and
number
one
is
we
want
to
take
a
full
pass
and
I
out
terms
in
the
term
section
and
there's
a
review
today.
C
I
think
all
yesterday
asking
a
section,
and
now
we
have
placeholder,
but
we
have
now
that
you
find
all
the
terms
yet
because
we
don't
have
your
terms,
some
utilize
them
somewhat
inconsistent.
So
therefore,
we're
trying
to
take
a
pass
and
number
two
is
feedback
from
couple.
People
is
the
following:
that
they
want
us
to
really
have
a
consistent
usage
of
all
the
require
eases
mass
should
may
and
so
on,
and
somehow
completely
more
consistent
way.
So
therefore,
I
think
we're
going
to
take
past
will
address
all
those
comments.
C
Number
two
and
number
three
is
that's.
Actually
a
comment
which
we
decide
to
really
accept.
Is
that
finish
section
1105?
This
a
new
section
in
in
document
we
should
say
for
all
of
you
if
you
wanna
use
incremental
push
what
kind
of
requirements
you
must
have,
even
though
we
see
what
I
want
to
be
future
proof,
I
don't
believe
we
can
be
future
proof.
C
Okay,
if
you
want
to
this
one,
we
recommend
that
you
do
the
service
things
certain
ways,
and
so,
for
example,
it
should
be
tax
base
recent
base
and
so
on
alibis
problems
will
not
do
very
well,
and
our
plan
is
a
final
model,
final
version
by
August
of
force
and
that's
our
next
step
and,
of
course,
one
issue
which
we
don't
plan
to
will
address
in
this
document,
but
the
more
and
more
we
look
at
it
stock
we
realize
is
probably
working
group
should
think
about
this.
It
should
be
a
new
patching
method
for
Jason.
C
Multipath
is
very
elegant,
but
unfortunately
it
has
limitations
here.
Some
pan
is
very,
very
like
essentially
powerful,
complete
a
sample,
encodings,
not
an
elegant,
and
maybe
I
could
give
others
good
use
cases.
Maybe
people
should
you
find
a
new
incremental,
dart
and
coding
algorithm
or
specification
for
JSON
objects,
but
record.
This
is
now
the
problem.
Alpha
working
group
right,
therefore
be
something
for
other
people
think
about
and
for
ours
right.
B
Mic
so
I
can
make
a
personal
comment.
I
think
the
addition
of
Jason
patch
was
very
important
for
the
CDN
I
use
case,
which
personally
I'm
very
interested
in,
and
so
this
is
not
a
chair
comment
by
the
first
no
comment,
and
the
second
comment
which
I
had
now
on
I
would
like
to
make
is
also
this
one
Vijay
and
I
discuss.
We
really
want
to
move
this
to
working
last
call,
but
as
with
Sabine's
document,
it
seems
you
need
to
have
another
round
of
minor
comments.
B
At
least
you
have
to
address,
but
also
I
should
have
mentioned
this
before
I.
Think
lady
we've
seen
good
discussions
on
the
mailing.
This
so
I
think
the
working
group
is
becoming
more
active
and
we
having
like
cross
comedy
some
people,
read
other
documents.
I
think
it's
very
good.
So,
in
fact,
what
we
have
here
is
all
the
result
of
the
good
discussion
we
have
so
as
chess.
We
want
to
encourage
that
still.
We
would
like
to
you
know,
move
this
sooner
or
later
to
be
ready.
B
So
it
seems
to
me
again
here:
I
would
like
I
know,
Mia
I
don't
have
to,
but
I
would
like
to
hear.
A
hum
of
the
working
group
agrees
with
Vijay
is
in
my
statement
that
this
should
be
ready
for
working
class
call
after
the
minor
issues
that
you
still
have
here
need
to,
you
know
are
have
been
addressed
so
I'm.
B
If
you
think-
and
this
is
ready
for
working
last
card-
pretty
soon
M-
please
Sam
now
lo
nobody's
having,
if
you
don't
think
this
is
ready
for
working
with
last
call,
please
now
I
think
people
were
not
aware
that
I'm
taking
a
hum
and
so
for
the
record.
There
was
no
reaction
either
way,
but
I
think
the
discussion
on
the
mailing.
That's
actually
a
lot
of
people
have
commented
that
they
think
this
is
ready
and
the
chairs
think
so
too.
E
Would
say,
though,
we
had
a
discussion
on
that
draft
at
lunch
and
most
of
the
content
is
here,
but
I
think
the
main
one
of
an
important
point
would
be
really
to
structure
and
clarify
the
role
and
between
the
the
rick
who
sends
a
request
for
what
this
this
needs.
Some
additional
rewarding
or
restructuring
wonderful.
C
So
let
me
respond
to,
I
think,
starting
trying
to
summarize
the
the
wonderful
wonderful
comment
during
a
lunch
and
I
believe
what
she
ever
come
after
the
most
substantial.
Then
there's
a
very
short
comment
over
here
and
if
you
look
at
this
increment
updates,
this
one
can
not
only
be
between
a
alcohol
clan
and
auto
server.
We
always
talk
about.
We
want
by
the
way,
I
think
probably
it's
about
you
will
talk
about.
It
is
like
your
main,
synchronization
everyone,
multiple
server
to
be
synchronised.
C
This
one
can
be
the
protocol,
so
this
one
can
be
not
only,
for
example,
between
auto
clan
and
auto
server
can
be
a
protocol
for
in
out
of
one
auto
server
to
push
updates
to
an
auto
server
to
allow
synchronization
in
a
very
compact,
very
nice
way.
The
encoding
to
Compaq
and
approach
is
immediate,
so
there
will
be
very
powerful,
I
think
the
sobbing
Sir
John
one
position
will
be
sold
a
moment.
We
somehow
imply
assume
is
client-server.
Therefore,
a
lot
of
like
any
faction
would
be
assumed
to
be
implicit,
I.
C
Think
I
suggest
we
make
this
one
more
X
place,
therefore
make
people
go
that
one
can
be
bumpy
eventually,
if
you
applied
for
setting
of
even
synchronization,
of
course,
for
this
setting,
which
one
the
client,
which
is
server,
and
it
can
be
a
client,
they
can
be
a
server
I,
think
that's
a
wonderful
comment
and
I,
don't
believe
I
can
send
a
common
online
or
maybe
any
wording
most
ideal
like
a
bourbeau
but
I,
think
that's
a
wonderful
comment
which
we
plan
to
address,
probably
immediately
in
the
next
few
days.
I.
F
B
It's
not
a
design
change.
One
agree:
okay,
if
there,
if
there
are
no
other
and
comments,
I
would
like
to
because
the
agenda
we
already
a
little
bit
behind
so
I'm,
going
to
push.
Thank
you
Richard,
and
the
next
document
is
cDNA
footprints
and
capability
advertisement
and
Richard.
You
want
me
to
present
this
one
right,
yeah,
so.
B
Yeah,
this
is
a
document
that
actually
am
Richard.
Thank
you
very
much.
You
resubmitted
today,
under
the
working
group
item
name
so
actually,
Richard
and
I
had
a
Skype
call
and
a
couple
of
weeks
ago,
and
we
were
so
busy
with
the
content
of
the
document
that
we're
issued
a
new
version,
but
we
totally
forgot
that
it
got
accepted
us
working
who
by
them.
So
we
did
this
now
and
the
system.
B
But
the
CDN
I
working
group
does
not
specify
how
the
news
JSON
objects
get
from
client
to
server,
or
vice
versa,
actually
server
to
client,
and
so
there
has
been
a
session
to
define
a
new
address
service
and
the
city
and
I
working
group.
Estes
I
decided
this
and
we
we
have
moved
this
as
a
new
milestone
for
the
alto
working
group,
and
so
the
specification
we
need
is
a
document
that
specifies
how
you
transport
these.
We
call
them.
B
Fci
JSON
objects,
foot,
building,
capabilities,
interface
of
the
CDN
I
interfaces,
how
you
put
transport
these
JSON
objects
via
Al
job,
so
yeah
I
mean
last
polity.
Of
course
you
want
to
do
as
much
as
you
can
with
the
regular
algebra
Tikal
we
have,
but
in
some
cases
we
need
some
extensions
and
that's
exactly
what
we're
exploring
here.
B
So
I,
don't
think
I.
Think
I
explained
this
last
time.
Also
and
I.
Don't
think
we
have
to
go
into
much
detail
here,
but
I
think
Richard.
You
made
this
guys.
Thank
you
very
much,
I
think
they're,
very
good.
They
give
people.
What
we
want
to
do
here
is
give
people
an
example
how
this
looks
like
so
the
semantics
are
pretty
much
specified
in
RFC
8,
zero,
zero,
eight
and
we
have
a
another
CDN.
B
I
am
document
that
says
the
media
type
is
applications,
CDN,
I
and
so-
and
here
you
have
an
example
also
on
the
right
side.
How
these
you
know,
Jason
objects
look
like,
and
we
have
another
RFC
that
you
fans
for
the
footprint
is
I
think
this
is
RFC
8
0,
0
6
in
CDN,
I,
metadata
and
document
CDN
I
was
a
bit
complicated
where
we
put
it
also
Richard
we
had
a
discussion.
The
Syrian.
B
Our
working
group
was
not
very
good
at
putting
all
the
registries
into
a
into
a
single
document,
so
they're
scattered
around
a
bit.
So
this
is
why
it
might
be
a
little
bit
confusing
to
this.
Just
to
summarize
what
has
been
defined
before
well,
what
we
need
to
do
now
is
come
up
with
I'll
just
service
that
can
you
know,
transport
these
documents,
and
this
is
rather
generic,
and
so
this
is
actually
a
summer
summary
of
what
the
city
and
I
working
group
has
decided.
Why
use
Alto
because
their
service
discovery?
B
B
B
This
I
think
is
a
better
and
for
people
who
are
not
familiar.
This
is
what
we're
actually
talking
about
so
you're
looking
at
an
FC
I
object
and
will
be
of
type
application,
slash,
CDN,
I
and
and
well.
It
would
use
the
regular
altro
error,
handling
and
I
think
Richard.
The
point
we
should
discuss
this
maybe
later
on.
The
second
point
you
make
here:
we
need
to
agree
on
a
name
of
this
envelope.
I
think
Kevin,
you
seven
all
draft
where
it's
called
Sita
and
I
FC,
I
and
I.
B
You
have
actually
done
a
good
job
at
looking
at
error
handling
with
your
team,
because
you
are
the
expert
on
the
alto
error
handling
and
it
seems
to
be
that
the
actual
alpha
error
codes
we
have
are
sufficient,
so
we
might
have
something
that
is
very
seedy,
nice
specific,
but
also
from
my
perspective,
if
the
regular,
auto
error
codes,
work
more
or
less
I
think
we
shouldn't
do
much
more
here,
because
we
don't
want
to.
You
know,
specify
too
many
little
things
if
this
works-
and
this
seems
to
be
the
case
in
regular
error.
B
Handing
is
pretty
okay,
the
thing
that
we
did
need
to
update
and
Richard.
You
just
presented
that
this
is
the
incremental
updates,
so
it
turned
out
that
Jason
merge
patch
was
not
sufficient
to
do
incremental
updates
with
the
CNI
objects.
So
now
the
latest
SSC
document,
the
ultra
incremental
updates
document,
also
has
Jason
patch
every
Jason
patch.
This
is
very.
We
can
also
do
incremental
updates
for
CNI
objects,
which
was
actually
a
key
requirement.
B
Also
from
the
city
and
I
working
group
so
for
CDN
is
actually
very
important
that
you
can
do
incremental
updates,
yeah
I,
don't
want
to
discuss
if
it's
more
important
than
for
the
peer-to-peer
use
case.
That's
a
lengthy
discussion
you
can
have,
but
it's
certainly
very
important
for
CD
and
I
for
CDN
sat
interconnect
to
do
incremental
updates
and
server
initiated
updates,
and
so
there
we
have
Jason
patch
now.
B
So
that
issue
is
basically
soft
yeah
here
and
now
Richard
I'm
presenting
this
I
might
have
a
slightly
different
opinion
than
you
I
think
it
I
appreciate
that
you
look
into
advanced
design.
So
maybe
you
can
also
comment
here
a
bit.
There
are,
of
course
things
you
can
do
better.
You
can
do
advanced
filtering
of
FCI,
because
the
current
filtering
is
not
designed
to
do
that
in
the
ultra
protocol,
and
you
can
do
other
things
that
are
you
know
integrating
ultra
with
FCI.
B
There
are
certainly
certain
things
who
can
do,
but
maybe
to
the
next
slide
that
the
current
drive
focuses
on
the
basic
design.
So
I
would
like
to
emphasize
this
because
again,
this
is
the
Meissen
we
have
on
the
Charter
and
I
was
saying
in
the
beginning.
Bj
and
I
really
would
like
to
finish
this.
So
now,
I
have
my
chair
head
on.
B
So
we
think
the
basic
design
is
actually
feasible
in
the
next
couple
of
weeks
months
to
get
done
and
and
then
we
have
to
go
from
there,
but
I
I
personally
think
that
the
the
current
document
we
have
would
just
do
the
basic
design
and
leave
further
optimizations
for
further
documents,
and
we
really
intend
to
work
on
this,
and
obviously
you
have
a
time
frame
that
really
try
to
move
this
ahead.
So
so.
D
B
So
what
about
what
we
try
to
do
in
this
presentation
give
the
working
group
an
overview
of
what
we're
doing
since
it's
a
working
group
item
now,
and
so
we
actually
think
that
we
can
what
we
call
the
basic
design
move
ahead.
Pretty
quickly,
so
error
handling
seems
to
be
done.
Incremental
update
seems
to
be
done.
We
just
need
to.
If
you
look
at
the
current
document,
you'll
see
that
we
need
a
little
bow
it's,
but
we
don't
envision
major.
You
know
design
choices
anymore,
so
the
the
major
ones
we
just
present
it
so
yeah.
B
We
want
to
move
this
ahead
and
then
yeah
get
it
done
as
fast
as
possible,
but
you
never
know
if
we're
doing
the
work
you
find
out
something
that
needs
a
little
more
attention
than
you
should
raise
in
the
millions
and
since
we
already
running
a
little
little
bit
behind
I
would
stop
talking
now.
Are
there
any
comments
on
this
document
and.
C
Richard
make
one,
can
you
go
to
slide
number
seven,
so
I
do
want
Caterpie
working
group
opinion.
Yes,
oh
one
more,
maybe
yeah
I'm,
sorry,
another
lie
can
make
a
number
smaller,
like
maybe
five,
oh
yeah
yeah
this,
like
so
I,
don't
want
get
the
people
feeling
about,
for
example,
the
the
envelope,
for
example,
for
the
object.
Forget
them
look
at
all.
C
For
example,
we
have
network
map,
we
have
a
custom
map
where
all
stuff
the
convention
we're
using
in
the
past
always
is
we
take
a
media
type
and
extract
a
media
type
as
a
model
name,
for
example,
you
know
possibly
have
application,
slash,
Auto,
/
network
and
Mac
plus
Jason
and
application
slash
Auto
cost
map,
slash
just
annual.
That's
why
we
got
an
envelope,
but
unfortunately,
a
seeing-eye
naming
is
that's
all
I
have
class
Jason
sort
of
all.
C
B
B
F
B
F
Pass
back
to
her
job
is
to
address
network
graph
milestone
and
we
have
achieved
an
agreement
to
adopt
assist
dropped
as
a
working
group
document,
because
that
three
main
issues
are
raised
and
some
english
resolution
was
up
proposed
at
ninety
eight
meeting
and
during
the
ninety
airmedia.
An
internal
meeting
efforts
are
made
to
finalize
this
design
issues.
F
Let's
review
those
three
designs:
it
shows
the
three
teachers
are
divided
into
two
categories,
so
response
format
and
the
request
format.
The
response
format
includes
the
information
structure
of
the
of
providing
perspectives
and
how
to
prove
how
to
encode
each
part
of
this
information
structure.
The
request
format
in
sokurah
format
for
clients
to
get
the
perspective
information
next
so
start
with
an
information
structure
to
provide
perfectly
informations.
F
So
the
problem
is
how
to
provide
these
two
maps
to
clans.
A
Sturman
approach
is
that
we
extend
the
current
cost
map
structure
to
enable
a
cosmic
data
and
a
problem
map
data
into
a
single
response.
This
solution
is
actually
breaks.
The
current
structure
of
the
Kozma
and
the
media
type.
So
another
solution
that
is
adopted
at
the
last
meeting
is
that
the
perfect
information
that
is
the
cosmic
and
the
property
map
and
send
with
two
separate
responses
that
is
encoded
into
two
separate
messages.
F
F
So
our
decision
made
this
time
is
that
we
will
keep
on
solution
to
that
is
we
keep
the
auto
cost
map
media-type,
keep
its
structure
and
we
still
use
the
property
map
to
provider
properties
in
the
2-step
query
mentioned
last
time
is
the
applied
to
support
that
client
may
independently
query
the
property
maps
for
several
times
as
it
likes
and
major
change
set
is
that
we
introduce
a
new
media
type
to
deal
with
the
problem
of
snapshot.
Consistency.
F
It
is
that
the
media
type
is
named,
multipad,
slash
related
and
the
details
of
this
media
type
is
defined
in
the
IFC
27
823
87.
This
media
type
enables
multiple
resources
in
a
single
response,
so
let's
use
an
example
to
see
houses.
Media
type
works.
If
the
client
needs
multiple
resources
in
a
single
response,
it
has
to
declare
in
its
accept
field
herbs,
HTTP
requests
to
include
three
media
types
together,
the
multi
path
related
and
together
with
the
autocross
map
and
automotive
net.
F
F
After
dividing
after
deciding
the
information
structure
of
the
coast,
pass
factors
with
need
to
decide
how
to
encode
each
part
of
past
actors
so
to
provide
perfectly
information
in
a
cosmic.
We
have
to
know
what
the
exact
cost
type
is
a
concert.
I
p--
includes
a
cost
mood
and
a
cosmetic
in
the
existing
design
idea.
A
custom
mode
indicated
data
types
of
relating
value,
for
example,
Scott
mode
being
numerical
or
ordinal-
implies
that
the
cost
value
is
a
float
or
int
number
respectively,
and
a
cosmetic
indicates
the
semantic
meaning
of
the
return
value.
E
F
Pass
after
the
after
dealing
with
the
response,
format
of
pass
vector,
information
become
to
discuss
about
query
format
current
design.
The
ideas
is
that
we
only
support
the
cross-product
query
format
of
the
auto
service.
This
pure
format,
intuitive
system
problem,
for
example,
when
a
client
is
only
interested
in
two
dependent,
independent
floors,
but
when
using
course,
product
queries,
it
has
to
query
for
related
flows.
F
F
We
found
that
without
extending
the
call
flow,
clear
format,
the
server
is
still
able
to
provide
the
correct
information
of
path
factors
and
once
the
client
wants
to
use
this
perspective
information
efficiently,
it
can
just
settle
unrelated
flows
to
zero.
To
remove
the
influence
of
those
unrelated
follows.
B
F
With
all
the
decisions
made,
we
have
come
to
the
specific
specification
part
we
extended
extension.
Current
version
tag,
object
to
include
an
optional
field,
curly
ID
and
an
extension
to
the
ID
resource
entry
to
include
an
optional
field,
name
the
property
maps
for
clients
to
query
the
property
maps
and
look
unless
the
response,
part
of
the
cost
map
or
the
endpoint
course
map
is
extended.
That
is
the
V
tag.
Field
must
be
included
in
the
cosmic
or
an
endpoint
cost
map
being
able
to
provide
the
query
ID
information.
B
Yeah,
thank
you
Don,
so
and
I
want
to
actually
say
that
I
very
happy
that
we've
seen
lots
of
discussion
on
also
this
document
on
the
marinus
in
comments,
and
so
my
question
is:
are
there
any
other
questions?
We
are
running
quite
a
bit
behind
schedule,
so
if
there
are
no
important
issues
that
somebody
wants
to
raise,
I
would
rather
like
to
move
on
because,
as
we
said
in
the
beginning,
we
don't
want
to
ham
on
this
one
today
anyway,
it's
it's
good
work
in
progress
and
I.
B
C
I,
don't
want
to
actually
read.
One
thing
would
be
probably
single
baixa
feedback,
which
one's
thinking
from
working
group
is
the
following:
is
the
adoption
or
multi-part
encoding?
So
if
we
accept
that
one,
then
we're
opening
and
I
don't
believe
it
as
a
lot
of
complexity
is
central
to
following
for
exam
RFC
or
the
email.
There
are
all
your
the
multi-part
encoding
and
if
we
allow
this
one
suddenly
we
have
tremendous
a
new
capability
to
transfer
all
the
network
information
resources.
C
G
B
G
B
Are
there
any
other
opinions,
maybe
John
Richard?
You
want
to
comment
on
that.
G
It
would
be
helpful,
you
know,
for
someone
who
hires
people
to
have
a
consistent
skill
set.
That's
very
consistent
with
with
engineers
who
actually
my
multi-part
and
and
a
lot
of
the
mine
related
stuff
is
understood.
It's
a
little
esoteric,
but
I
can
point
them
to
specs.
The
SSE
spec
looks
a
little
odd
from
from
that
perspective,
as
a
hiring
manager
for
devs.
So.
B
C
G
B
So
so
yeah,
let's
discuss
it,
and
on
the
mailing,
this
maybe
or
further
and
I
mean
SSE,
is
still
not
done.
So
there's
still
a
good
opportunity.
If
it
were
already,
you
know,
M
Lyle
if
it
were
a
little
bit
further
down
the
road
okay.
But
at
this
point
it's
probably
still
a
good
chance
that
we
yeah
it's
good
I
could
put.
The
time
is.
C
H
F
F
It
is
it
a
power
to
the
security
problem.
Yeah
I
think
there
is
some
trade-offs
yeah
and,
for
example,
and
cross
product
in
from
cross
products
curious
assent.
The
server
is
going
to
send
a
more
detailed
information,
maybe
about
network,
and
if
we
use
this
cross
product
information,
clear
format,
what
can
benefit
I
think
it
may
be
that
the
the
client
is
able
to
hide
it's
possible
datasets,
for
example,
if
the
client
query
exactly
with
the
call
flow
query
format
yeah.
It's
me
maybe.
B
B
So
basically,
it's
saying
cross
product
query
might
have
a
privacy
implication
and
PID
flows,
don't
and
but
I
think
you
should
just
consider
discuss
it
with
your
with
maybe
on
the
mailing
this,
but
also
with
the
authors
and
either
way,
I
mean
either
you
you
consider
it
and
you
go
for
one
or
the
other
or
if
you
go
for
something
that
has
privacy
vacations
as
I
was
mentioning
whose
have
been
before
just
mentioned
them
explicitly
right,
added
privacy
considerations,
discussion,
section
and
discuss
it.
Yeah.
F
B
D
B
F
And
another
thing
is
that,
with
declares
that
the
learners
to
met
prefix
match
rule
for
the
ipv4
and
ipv6
domain
will
introduces,
but
not
introduces
multiple
inheritance
and
a
small
mistake
corrected
is
that
value
for
the
property
value
for
the
this
ipv4
address
is
the
v1
rather
than
v2,
and
if
there
is
in
network
maps,
so
server
is
able
to
provide
maybe
and
pass
one
Maps
or
Internet
address,
and
today's
to
prop
the
values.
H
F
The
uses
field
in
the
resource
entry
is
that
a
so
if
a
papdi
map
videos
case
is
that
if
a
problem,
a
property
map
depends
our
network
map
resources.
It
has
to
declare
that
it
only
depends
on
a
specific
network
element
resource.
But
no,
not
you
know,
maybe
two
or
more
and
to
general
case-
is
that
the
uses
field
must
not
contain
two
resources
of
the
same
type
and
then
which
clarifies
that
the
value
for
the
type
property
value
to
be
defined
to
have
no
value
must
be
a
Johnson
non
value.
F
The
section
five
point:
five
is
the
same
as
a
section.
Four
point
five
is
to
declare
the
u.s.
is
field
in
the
use
of
entry
and
the
ending
domains
that
we
propose.
A
new
domain
named
a
ne
and
the
domain
specific
and
the
trashiest
is
the
same
as
PID
name
specified
in
RFC.
Seventy
to
eighty
five,
there
is
no
hierarchy
or
inheritance
rule
associated
with
this
domain
and
the
relationship
to
cost
Maps
and
an
pankot
map
needs
to
be
discarded.
More
ok.
F
B
F
The
next
steps
for
for
this
draft
is
that
we
we
are
going
to
specify
more
clearly
the
relationship
between
the
N
domain
and
the
cost
map
or
on
and
particles
map,
and
we
have
interesting
example
of
Hadoop's
system.
You
make
use
of
the
property
map,
this
may
introduce
a
new
domain
and
we
have
not
stated
the
system.
F
The
distance
is
computed
by
the
client
self,
but,
however,
on
the
client,
because
the
network
location
property
of
each
domain,
this
is
provided
by
the
service,
but
not
the
client
feel
so.
The
unified
public
map
is
going
to
provide
the
network
location
property
of
each
node
and
a
possible
request,
and
the
response
format
will
be
like
this.
F
We
are
use
the
property
mapped
query,
a
with
entities
in
one
info
and
the
properties
being
network
location
and
the
property
map
will
return
to
the
client
where's
that
there's
no
locates.
It
may
be
locating
data
center,
Wang
and
rec
1
and
node
1.
Oh,
it
may
be
located
yet
at
the
data
center
and
the
rack,
and
no
the
for
don't.
B
F
B
F
B
Is
the
last
one?
Okay
so
and
since
we're
running
behind
schedule,
I
don't
think
we
should
have
a
discussion
on
or
I
would
refer
discussions
to
the
mailing
list
and
to
the
author's
as
my
chair
Don
now
I'm
I'm
I'm,
not
sure
I'm
happy
that
you're
introducing
new
use
case.
It's
a
new
stuff
because
we
want
to
move
this
ahead.
B
F
B
Okay,
but
anyway,
I
think
they
were
also
here.
We
had
some
discussion
and
I
think
it's
moving
ahead.
So
it's
good.
It's
good
work
and
I
just
fear
that
it
gets
out
and
it
will
not
get
done
by
the
time
we
want.
Okay,
Thank,
You,
Don,
and
so
the
next
presentation
is
actually
I'm
presenting
on
behalf
of
Sebastian
and
Martine
on
the
cross
domain.
Server
discovery-
and
this
can
actually
be
very
short
presentation,
I,
think
or
I
hope
so
and
I
would
go
to
the
mic
and.
B
So
just
to
remind
everybody
what
what
this
is
all
about
is
basically
to
find
a
new
server.
We
are
DNS
records
or
sorry,
an
agile
server
and
basically
I'm
reading
here,
the
intended
semantics
you
can
say
if
you
are
a
nice,
PR
IT
department
who
controls
something
you
can
say.
Okay,
if
you
want
to
optimize
your
traffic
to
or
from
this
prefix
square,
the
outer
server,
which
you
can
find
here
right
so
and
yeah.
It's
a
classic
and
used
to
be
called
third-party
server
discovery,
but
for
various
reasons
and
change.
B
The
only
recent
changes
are
really
small
ones,
so
the
algorithm
has
been
changed
that
it
can
also
handle
IP
prefixes
as
a
parameter
and
not
only
IP
addresses
is
really
small
change
and
there
have
been
some
clarifications:
aguar
regarding
the
usage
of
alto
maps
and
map
filtering
services.
So,
as
I
was
saying
in
the
beginning
of
the
this
session,
and
we,
the
chairs,
think
this
is
ready
for
working
last
call
and
because
again
it
has
been
around
we've
discussed
it.
B
C
Richard
go
ahead
yeah,
so
I
did
send
out
that
my
part
where
a
review
I
think
this
morning
Oh
yesterday,
I
forgot
and
I,
like
it
I
like
I,
think
this
is
very
useful.
My
only
I
guess
roughly
common,
mostly
the
editor
would
be
somehow.
If
we
can
come
out
of
the
principle,
for
example,
I
believe
is
using
somehow
expand
a
the
ring
search,
something
like
that
or
gradually
like
how
to
search
make
it
a
principle
more
clear,
probably
be
helpful,
because
mostly
this
editorial
and
I.
B
Just
I
saw
the
discussion,
I
haven't
read
all
the
emails,
but
I
think
Sebastian
actually
replied
and
said
that
they're
going
to
look
into
it
right.
That's
at
least
what
my
understanding
so
and
it's
a
good
comment.
Obviously
we
don't
want
to
push
things
if
there's
two
comments
or
concerns,
and
but
I
understood
your
comment
is
something
that
is
also
something
that
can
be
handled
and
rather
quickly
so
yeah
I
think
this
book
is
really
so
beachy
and
I
think
this
is
ready
for
working
class
call.
B
So
probably
it's
a
prescient
saying
something
now
he
joined
the
queue
he
wants
to
say
something
I'm
on
the
mic
and
somebody
Sebastian.
If
we
want
to
say
something
typing
in
jabber
and
maybe
I
can
read
it
or
something,
I
am
I.
The
only
one
in
jabber
I
can
check
and
hold
on
and
can
I
present
and
read
my
jabber.
At
the
same
time,
guys.
B
So
I
think
the
authors
and
the
chairs
think
this
is
almost
ready
for
work
in
Glasgow
or
pretty
much
so
I
think
they're.
They
will
review
Richards
comments
and
issue
a
new
version
rather
soon,
and
then
we
can
go
to
working
class.
Colin
I,
don't
hear
any
other
opinions
in
the
room.
So
I
think
this
is
pretty
much
it.
B
B
We
have
15
more
minutes
and
we
have
three
scheduled
presentations
that
are
not
chartered
items
so
I
would
say:
I
give
each
of
them
five
minutes
to
be
fair,
so
the
first
one
is
Sabine
with
cellular
address
extension
and
then
we
have
J
with
flow-based
cost
query,
and
then
we
have
Q
with
resource
orchestration
and
and
I
think
the
last
ones
are
remote,
so
we
see
how
we
can
handle,
but
Sabine.
Please
go
ahead
with
cell
your
address
extension
and
again
everybody.
Please
stick
to
your
five
minutes
to
be
fair
right,
so.
E
This
is
actually
resuming
a
couple
of
discussions
we
had
since
a
couple
of
months
ago.
So
next
slide
please
so
here
I,
the
draft
is
very
short,
actually
I
wanted
to
start
a
mailing
list.
Discussion
and
I
found
it
more
convenient
to
just
drop
a
draft.
So
this
is
a
list
here
of
related
out
of
work.
That
involves
some
of
our
networks,
and
this
list
is
pretty
much
the
same
as
the
one
in
the
cost
context
a
draft,
so
we
need.
E
E
If
I
want
to
query
some
properties,
understand
I
want
one
I'm
connected
to
and
given
the
auto
based
protocol
in
space
specified
with
endpoint
addresses
endpoint
having
cellular
IP
our
RV
for
our
ipv6
addresses,
but
nothing
precludes
in
the
protocol
to
have
something
else
than
IP
IP
addresses
so,
and
we
also
had
discuss
it
in
the
last
IETF.
So
next
slide,
please
so.
E
E
There
was
some
previous
out
of
work
and
design
choice,
so
there
was
the
work
from
line
and
the
work
presented
in
IETF,
91,
I
guess
for
also
for
wireless
networks,
so
typically
cell
is
associated
to
a
PID
and
you
want
would
like
to
have
cellular
address
on
this
PID.
So
the
proposal
comes
from
the
draft
outer
in
wireless
access
networks
and
the
idea
is
to
use
to
mimic
the
format
of
this
GI,
which
stands
for
EU
transfer,
global
identifier.
So
next
slide.
E
So
you
have
the
mobile
country
code
and
then
you
have
the
mobile
network
code
and
then
you
have
the
EU
transyl
identifier,
so
the
discussion
for
future
version
will
be
about
those
leading
zeros
I'm,
not
so
sure
about
how
we
do
given
the
fact
that
there
you
have
MCCS
coded
0,
20
MC,
cisco
d
20,
so
I
leave
it
to
the
to
the
specialist
so
and
so
next
slide.
Please.
E
Much
to
say
about
that,
and
next
slide,
please
yeah!
So
these
cellular
this,
the
cell
ID
format
for
auto,
could
have
three
formats
either
just
about
mobile
country
code
or
the
mobile
country
and
Nick
Moore
code
or
the
total
with
a
mask,
and
so
this
is,
and
with
some
rules
saying
that
okay,
so
some
matching
rules
obliging
that
you
have
at
least
a
match
between
the
MCC
and
MNC.
E
E
It's
just
summarizing,
so
this
is
again
an
example,
so
you
can
skip
it
and
the
next
slide.
Yes,
next
step
will
be
to
look
at
other
IETF
proposal
for
the
cell
address
format
validate
these
formats
or
another
one
with
respect
to
working
with
discussions
introduced
cell
addresses
in
working
group
documents,
especially
performance
metrics
and
cell
entity
domain
identifier
for
the
unified
property
draft
use,
also
endpoint
properties,
filter,
cost
map
and
endpoint
Kozma,
and
also
use
this
format.
G
H
I
I
Yeah
I
can
see
it
so
I
will
quickly
go
through
our
update.
Our
flow
based
calculus
in
our
lab
meeting
next
week,
and
here
is
our
overview
since
ITF
98,
we
have
updated
to
variants
of
this
document,
introduce
a
bigger
architecture
to
claim
why
we
need
the
auto
extension
such
elefseus,
and
we
make
some
reason
to
target
more
extensible,
sensible
design
and
the
consistent
semantics.
Next,
please.
I
I
I
I
From
this
inside,
we
consider
the
general
requirement
at
two
items:
the
more
flexible
input
and
more
practical
output.
A
many,
including
Auto
extensions
like
test
vector,
unified
property,
mat
cost
and
the
Condor
are
targeting
the
most
flexible
output,
but
FCS
is
focusing
on
more
flexible
input.
Specifically
FCS
has
two
requirements.
The
first
way
is
to
provide
more
flexible
shape
of
the
pure
space,
and
the
second
way
is
to
provide
more
pricey
encoding
of
the
cure
entry.
Next,
please.
I
And
if
it
goes
and
details
about
the
curie
schema,
we
make
some
small
change
on
the
Christ
based
schema
since
I
have
98.
Currently,
the
wire
of
the
endpoint
flow
spelled
is
our
REO
and
one
filter
instead
of
the
array
of
just
destination
pairs
which
can
reduce
some
overhead
and
the
next,
please
I
think
it's
a
pity.
I
I
To
turbit
mark
pricey
entry,
including
we
have
created
some
new
auto,
tries
tab.
Rigid
3,
for
example
the
eth
HTTP,
ftp,
tcp
and
UDP
are
the
new
accepted
address
types
but
hidden,
but
it
contacts
some
further
issues.
For
example,
if
an
wanna
filter
controls,
ftp,
tapped,
source
address
and
HTTP
tablet
destination
christ,
this
curie
should
be
invalid,
because
ftp
complete
with
HTTP
so
next
place,
I'm.
B
Sorry
to
interrupt,
you
can
wrap
up
a
bit
because
we.
I
Like
it
ok,
so
there
is
a
remaining
issue.
If
we
consider
this
example,
then
this
curio
have
two
unwanted
felt
her
first
client
valid,
but
the
second.
Why
didn't
violate
legacy
auto
/
co'?
Will
it
turn
our
message
directly?
So
the
auto
clan
has
to
resend
filter
one
and
fell
to,
but
in
our
question
is
it
is
possible
to
return
every
passed,
the
valid
part
with
error
message
for
the
invalid
part?
We
have
a
simple
solution
at
option
one,
but
we
are
considering
whether
can
provide
an
automatic
conflict
avoidance
mechanism,
for
example.
I
In
this
case,
UDP
is
a
specific
type
of
activity
for
our
ipv6,
so
the
auto
server
can
reduce
the
certain
point
address
to
an
aquifer,
Christ
and
returns
cost
between
it
and
the
ATP
or
crest
okay.
So
the
last
page
is
the
next
type.
We
also
analyzed
the
design
decisions
and
we
want
to
make
the
new
skete
services
clearer.
So
an
actual
thing
already
I
have
done.
B
I
C
B
J
H
J
Yeah,
this
is
a
quick
takeaway
from
last
ITF
meetings
and
we
propose
that
not
only
Otto
can
be
used
to
provide
lateral,
New
York
emissions
so
that
to
speed
up
dataset
transfers.
It
can
also
provide
information
on
different
type
resources
too,
so
that
we
can
improve
help
improve
the
performance
of
data
and
any
applications,
and
based
on
this
proposed
already
done.
The
a
multi
resource
or
treater
for
the
compact
muon
solenoid
at
kitchens
are
called
EXO
and
next
piece.
J
So
this
is
a
quick
update.
The
latest
update
for
the
draft
and
the
week's
first.
We
expand
application
scenarios
from
the
resource
alteration
for
science,
application
to
a
more
generic
unified
resource,
Austrians
for
geo,
dispersion,
modeling
data
and
linking
data
analytics
and
we'll
name
the
framework
to
be
called
unicorn.
We
describe
the
key
component
of
unicorn.
It
consists
of
a
set
of
tools
to
manage
the
dynamics
of
the
resource,
demand
and
supply
and,
of
course,
it
will
maintain.
J
It
has
has
a
Orchestrator
to
match
the
demand
with
with
the
resource
supply
and
we
restructure
the
document
by
adding
some
by
updating
the
abstract
and
destruction
sections
and
we
add
some
several
examples
and
describe
the
people,
the
workflow
of
the
whole
framework,
the
work
group
to
review.
We
can
jump
to
the
slide
7.
J
So
yeah
30.
So
this
is
why
we
want
to
use
our
tool
in
our
framework.
This
is
mainly
because
the
existing
resource
management
system
such
as
H,
Condor,
mesos
and
so
on.
They
only
provide
course
growing
the
informations
on
resources
and
may
lead
to
Union
efficient
resource
allocation
decisions,
but
on
the
on
the
country,
auto,
provide
on-demand,
foundering
information
on
different,
have
resources
to
support
the
optimal
resource,
appreciation
for
data
analytics
jobs,
and
we
can
jump
to
a
slide
number
10.
J
So
yeah,
you
know
not
mean
we
already
gave
a
quick
example
about
how
Auto
can
provide
certain
information.
So
here
is
the
architecture
of
the
unicorn
so
to
manage
the
resource
diamond
dynamic.
We
provide
a
work,
workflow
converter
and
the
amount
esta
meter
to
automate
to
automatically
find
the
optimal
configurations
for
different
data,
analytic
jobs
and
under
resource
supply
that
are
on
the
resource
supply
side.
J
We
use
Auto
client
to
I,
will
use
auto
plant
to
cure
a
to
issue,
a
series
of
issue,
a
series
of
surgeries
to
the
auto
servers
to
get
there
with
the
resource
informations,
corresponding
to
certain
to
correspond
to
certain
data,
analytical
jobs
and
we
use
a
resource.
Will
each
structure
to
assemble
and
the
comprise
is
such
information
into
an
abstract
wheel
of
resources.
We
can
go
to
the
next
slide.
J
So
so
yeah,
so
you
know
pitching
to
the
base
to
the
base
services
de
proposing
RFC
70
to
85.
We
also
a
unicorn
also
benefit
substantially
from
several
auto
its
extension
included,
including
the
unified
RP.
The
past
record
cost
calendar
are
all
this
year
has
been
adopted
and
also
it
the
resource
will
extractors
I
used
red
I,
wasn't
proposing
the
auto
route,
instead
attraction
top
and
also
we
include
the
flow
cost
services
to
probably
show
more
fun
going.
The
information
about
different
type
of
flows
go
to
slide.
14.
J
Absolutely
so
so,
but
so
yeah,
just
just
quick
summary,
so
unicorn
basically
provides
a
template
architecture
for
data
center
resource
orientations,
and
this
is
a
major
use
case
for
auto
again.
The
working
group
chatter
and,
like
I
just
said,
in
addition
to
the
base
auto
protocol.
It
applies
several
auto
extensions
and
to
click
resource
information
from
different
sites,
and
this
is
a
informational
job
and
it
provides
key
insight
and
experience
in
the
deployment
of
different
type
of
auto
services,
and
you
know
very
large
and
public
that
an
elective
project.
J
So
the
next
step
for
us
is
to
continue
to
document
on
the
design
and
the
experience
of
unicorn
and
also
we
are.
We
are
still
on
the
right
track
with
the
previous
prepared
with
the
mouse
joint
for
pre
production
deployment
on
by
I'd
have
one
hundred,
and
hopefully
two
for
production
front
by
I
tracked
one
100
or
three.
That's
it.
A
B
Over
time
so
I'm
also
here,
if
people
have
comments,
please
do
them
on
the
mailing
list.
But
thank
you
for
this
update
and
I'm
happy
that
we
managed
actually
all
the
scheduled
presentations
and
I
think
we
made
good
progress
also
with
the
chartered
items
so
yeah
with
that
I
close
and
again,
most
likely
we're
gonna
have
next
virtual
interim
meeting,
but
we
have
to
see-
and
we
will
post
it
on
the
mailing
list.
So
thanks
everybody
and
yeah.
Let's
keep
moving
the
drafts
long,
okay,
bye-bye
thank.