►
From YouTube: IETF100-ANIMA-20171113-0930
Description
ANIMA meeting session at IETF100
2017/11/13 0930
https://datatracker.ietf.org/meeting/100/proceedings/
A
A
A
A
The
main
list,
as
euro
is
one
of
the
men
working
place
unite,
have
so
they
can
eat
and
the
disgusting
work.
We
continue
in
many
needs
and
all
those
working
group
t
saying
maybe
in
this
room
will
be
confirmed
in
many
lists
and
just
curious,
myself,
Shing
and
various
working.
So
we
also
have
the
website
for
antenna
tracker.
A
A
A
The
profits
management
document
also
passed
the
working
class
in
July
just
after
an
asteroid
meeting
and
it's
submits
for
publication
as
informational
document
in
August
its
highest,
which
is
Dean
now
as
well,
and
this
table
connectivity
as
another
one
of
to
use
key
validation,
use
cases
and
passed.
The
working
velasco
in
August
submit
to
ISD
for
publication
as
informational
document
in
October.
So
it's
working
for
ad
variation
and
we
still
have
three
document
with
the
working
group.
The
autonomic
control
plan
passed.
The
working
of
last
call
only
make
I
only
made
this
a
result
yesterday.
A
But
actually
that's
you
know:
two-week
working
class
copier
trust
two
week
extension
so
that
waiting
for
another
round
of
author
update
and
Shepherd
ready,
Hogg
I
was
submitted
to
I
asked
if
a
publication,
if
there's
no
significant
change
for
the
which
request
to
the
second
working
of
Alaska
and
risky
tariffs.
As
the
Shefford
already
down
one
round
off
review,
and
but
it
happened,
we
follow
amber
oil,
so
it
has
to
do
it
in
another
round
again
and
so
where
we
are
planned.
A
B
A
Six
new
work,
I
turn:
you
proposed
work,
some
of
them
already
discussed
in
the
working
group,
since
we
are
keeping
the
second
session
for
most
of
meetings,
so
some
of
them
are
the
amp
in
discussed
wheel
and
the
church
believes
at
least
was
three
of
them.
Actually
in
the
current
chatter,
which
means
you
know
we
we
may
be
able
to
add
the
milestone
without
modify
the
current
each
other
and
the
others.
A
We
may
need
to
discuss
that
with
our
ad
afterwards
to
say
how
much
you
know
they
are
relevant
to
the
current
chart
and
how
much
they
may
have
to
change
the
the
Charter
of
all.
Even
they
are
out
of
scope,
so
that's
welcome
to
doing
afterwards,
so
just
grasp
API
and
the
informational
distribution
over
grasp
and
the
guidance
for
at
Monaco
service
agent.
Also,
the
packet
data
over
grasp
and
towards
public
have
shop
and
storage,
integrating
in
an
American
concept
and
DNS
SD
over
a
grasp.
A
So
that's
six
plantation
we
had
today
for
the
non-charter
locator.
Each
of
them
have
50
minutes,
but
I
have
asked
the
printers
to
keep
their
presentation
shorter.
But
given
the
ideas
why
the
this
work
is
important
and
how
they
fit
into
the
working
animal
working
group,
chatter
and
hopefully,
leave
Li
enough-
is
cutting
for
the
working
group
to
see
whether
this
those
proposed
work.
Are
you
know,
interested
by
the
group
or
not
any
art
in
the
past?
A
B
A
A
A
D
C
This
is
this
was
based
upon
feedback
from
a
number
of
implementers
that
they
wanted
to
have
a
more
clear,
walkthrough
kind
of
point
of
view.
Since
then,
we've
been
working
on
essentially
clarifying
the
text,
making
it
a
little
bit
better.
We
fixed
the
proxy
discovery
registration
that
originally
had
been
conceived
of
as
a
typical
grasp
discovery.
It's
now
an
EM,
the
flood,
from
the
registrar
to
the
proxies.
C
And
then
that
leads
to
the
question
of
how
the
domain
itself
is
owner
is
pinned
in
a
voucher,
particularly
in
light
that
some
would
desire
to
use
a
web
PKI,
which
is
not
self
signed,
private
PKI,
but
rather
descends
to
them
something
like
GoDaddy
or
Verisign.
So
clearly,
the
domain
owner
is
not
Verisign
and
if
you
were
to
pin
the
top
cert,
which
is
what
we
said
to
do,
then
that
would
be
wrong.
So
we're
pretty
sure
that
we
want
to
pin
a
set
of
certs
and
not
a
say,
Verisign
Plus.
C
At
the
end,
we
want
to
pin
a
specific
thing,
but
current
tooling
either
says
send
the
entire
chain
or
send
just
the
last
piece
or
you
know
it
doesn't
really
let
you
do
what
you
want.
So
we
think
that's
really
a
tooling
issue,
not
a
text
issue,
but
we
really
want
to
just
pin
the
the
domain
owner.
It's
really
easy.
C
If
it's
a
private
PKI
and
it's
self
signed
the
domain
owners
who
they
say
are
it's
only
when
somebody
wants
to
use
a
something,
that's
anchored
into
some
public
PKI
that
there
becomes
an
issue
we
added
some.
There
will
be
some
sample
example
vouchers,
including
private
keys
and
voucher
requests
in
the
next
revision,
which
I
think
that
Max
and
I
will
be
posting
today
next
slide.
Please.
C
Can't
read
your
side:
oh,
he
ripped
out
all
the
appendix
D,
so
we
think
that
between
revision,
six
and
revision,
nine
that
we
managed
to
essentially
decide
find
all
the
text
that
we
thought
might
be
still
valuable.
We
merged
it
back
in
and
finally,
we
ripped
out
that
that
collection
of
text
that
we
have
been
accumulating
and
kept
around
for
a
while.
We
changed
the
definitions
of
proximity,
registrar
search
and
prior
signed
voucher
requests.
C
At
one
point,
these
were
kind
of
overloaded
between
whether
or
not
it
was
a
voter
request
versus
a
voucher,
so
proximity
registrar,
cert
is
always
the
pinned
name
of
the
domain.
Id
and
prior
signed
voucher
requests
can
contain
a
well
prior
signed,
voucher
or
request,
and
that's
actually
how
the
pledges
signed
or
unsigned
voucher
request
is
passed
by
the
Registrar
to
the
masa.
So
there's
some
good
text
in
there
to
explain
how
that
works
and
why
it
happens
and
I
think
we're
pretty
happy
with
that.
C
At
this
point,
we've
put
it
a
lot
of
text
into
a
number
of
sections
to
even
them
out
a
little
bit
to
remove
some
things
out
of
one
sections.
The
proxy
got
its
own
section
again,
which
is
a
little
bit
of
a
regress
to
how
the
previous
organization
was,
but
the
proxy
is
kind
of
a
separate
thing.
It
doesn't
really
participate
in
the
in
a
in
a
semantically
meaningful
way
the
rest
of
the
transaction,
but
yet
it
does
exist
next
slide,
please
we
went
through
a
bunch
of
it
cycles
and
iterations
with
Diana
registration
details.
C
C
As
point
10
says,
we're
actually
going
to
be
an
application,
pkcs
mine,
and
then
we
would
have
a
new
s,
mine,
type,
voucher
or
voucher
request,
but
just
to
be
clear,
it's
it
may
be
that
that
it
might
be
that
the
that
it's
really
going
to
be
application
CMS.
So
the
voucher
document
has
really
moved
in
that
direction
and
we're
going
to
fix
that
and
we're
absolutely
sure
that
the
voter
text
has
passed
all
the
reveals
and
everyone's
happy
with
that.
D
E
C
You
have
eight
and
then
me
and
then
see
and
so
it'd
be
bad
if
somehow
I
started
doing
Stevie
or
a
was
finished.
So
that's
really
building
your
client
correctly
and
it
shouldn't
be
a
problem.
But
but
I
can
imagine
somebody
that
got
their
client
to
work
correctly
with
HTTP
one
one
sequentially
and
then
breaks
with
HTTP
two.
C
So
in
the
end
we
decided
there
wasn't
really
a
lot
to
say
about
this
implementation
issue,
but
we
did
tighten
their
carbons
to
make
it
pretty
clear
that
we
were
expecting
HTTP,
1.1,
persistent
connection
and
that
if
you
have
to
renegotiate
the
connection
of
the
TLS
connection,
then
got
some
problems
that
they're
going
to
happen.
That's
really
yet.
A
B
C
B
And
in
warning,
with
my
experience
with
the
author's
on
the
first
round,
so
they
have
a
great
system
of
driving
changes
to
the
document
through
get,
and
so,
instead
of
only
you
know
sending
to
the
mailing
list,
which
obviously
is
an
IETF
requirement.
It's
always
good
for
for
this
draft
to
bring
up
considerations
as
issues
on
the
kit
itself.
C
G
F
B
All
right
so
just
wanted
to
give
a
quick
update
on
the
changes
since
Brock
as
Chang
mentioned.
The
document
is
a
past
working
group
last
call,
so
this
is
just
for
information
purposes,
so
there
was
a
thorough
reviews
by
Brian,
Ching
Michael
and
some
smaller
points
from
others.
So
thanks
a
lot
for
that.
We
had
the
discussion
of
these
points
and
feedback
to
the
mailing
list,
and
so
I
think
Shang
is
going
to
do
the
the
final
chat
right
up.
So
this
is
primarily
you
know
for
for
the
history
books,
just
the
summary
of
things.
B
B
You
know
in
consistent
terminology
device
host,
physical
and
so
on,
and
so
I
try
to
fix
up
the
whole
terminology,
that
everything
is
a
note,
so
that
it's
more
logical
that
it
could
be
a
virtual
router
vrf
function
as
well
and
then,
instead
of
physical
interface,
I
use
the
term
native
interface,
because
I
really
don't
know
what
a
physical
interface
in
the
virtual
machine
is.
But
that
is
novel
in
the
idea.
So
I
don't
think
we
have
define
terminology
for
that,
we're
keeping
the
the
term
loopback
interface.
There
has
been
discussions.
B
B
So
right,
so
basically,
the
ACP
now
demands
that
all
nodes
must
support
certificate
renewal
through
the
ACP
via
est,
which
is
distinct
from
brewski.
So
when
we
started
this
out
couple
years
back
most
folks
working
on
the
ACP
we're
thinking,
brewski
was
the
document.
I
was
specifying
that
also,
basically
now
brewski
is
really
only
the
bootstrap
part
and
the
certificate
renewal
is
simply
standard
EST.
B
B
Obviously
we
need
to
use
tunnel
mode
with
IPSec,
so
we
stumbled
over
that
because,
obviously
the
ACP
is
forwarding
packets
so
that
in
it's
not
only
all
locally
generated
packets,
then
I
did
add
a
lot
of
text
about
the
grasp,
specifically
also
in
the
section
of
why
we're
using
grasp
for
service
discovery
and
what
we
consider
there
to
be.
You
know
a
basic
property
of
the
ACP
that
is,
support,
service
discovery
with
grasp,
as
opposed
to
IP
multicast,
which
is
what
I've
seen
a
lot
of
community
is
always
coming
back
to
it.
B
You
know
prior
life
I've
been
doing
a
lot
of
multicast
and
people
say:
oh,
let's
do
that
for
service
discovery
on
it's
really
horrible,
so
yeah
you
can
read
up
on
that
and
they're.
Also,
a
bunch
of
you
know
optimization
the
future
we
can
do
with
with
ripple
next
slide.
So
this
was
basically
the
you
know:
magic
outcome.
B
B
So
this
is
basically
the
picture
showing
how
we're
forwarding
in
the
ACP
packets,
between
different
secure
channels,
once
as
datagrams
over,
for
example,
IPSec
for
just
the
forwarded
just
to
provide
the
ipv6
vrf
that
the
ACP
is
and
then
in
addition
to
that,
were
basically
for
grasp
itself,
building
end-to-end
TLS
connections,
and
that
is
because,
in
the
security
review,
the
question
was
brought
up.
What
happens
if
we
have
impacted
notes
in
the
ACP
that
cannot
be
trusted
anymore
and
you're.
Just
you
know,
passing
unencrypted
grass
messages
through
it.
B
Well,
the
outcome
would
have
been
that
if
we
would
have
kept
this,
which
is
what
we
started
out,
then
the
security
folks
would
later
come
back
and
tell
any
aasa
that
we're
building
that,
if
you're
relying
on
grasp,
you
still
need
to
come
up
with
an
application
layer
encryption,
because
you
can't
trust
the
intermediate
path.
All
that
much
right.
That's
basically
why
I
said.
Well,
let's,
let's
cut
the
short
here:
let's
move
to
TLS
also
for
the
unicast
in
the
ACP
for
grasp,
and
so
that's
the
picture
showing
how
that
all
interacts
next
slide.
B
We're
pool
routing
a
bunch
of
details
of
other
things,
so
in
the
whole
IP
an
IP
encapsulation
scheme,
the
need
for
more
addresses
came
up,
and
initially
we
had
an
addressing
scheme
in
the
ASEP
that
would
give
every
acp
node
up
to
eight
bit
worth
of
addresses.
I
changed
that
with
a
bit
so
that
the
addressing
can
now
assign
eight
or
sixteen
bits
to
every
individual
node
and
the
other
things
rip.
Olia
worked
a
lot
with
the
Michael
and
Pascal
on
figuring
out
all
the
details
for
ripple
implementation
next
slide
yeah.
B
So
then
all
the
the
ACP
interfaces
right
so
in
terms
of
you
have
secure
channels.
You
have
an
IPSec
channel,
okay,
how
you're
running
IP
over
it,
and
especially
if
you
have
a
LAN,
you
have
multiple
ACP
neighbors
and
you
want
to
make
it
look
like
a
lens.
I
was
looking
on
a
bunch
of
ITF
made
in
the
there
isn't
really
good
reference,
so
I
basically
chopped
up
a
section
explaining
how
you'd
basically
create
a
virtual
LAN
interface
on
top
of
multiple
point-to-point.
B
Acp
secure
channels
run
nd
on
it,
and
so
hopefully
that's
it
sufficiently
complete
to
allow
implementations.
Hey,
CP
Connect
was
also
extended,
the
section
just
to
showing
that
you
know
what
many
people
have
started
to
see
as
a
workaround
like
you
connect,
a
non
ACP
device
to
the
ACP
is
actually
very
good
long-term
option
if
you're,
really
thinking
about
composing
software
in
a
device
through
multiple
software
modules
that
you
connect
through.
You
know
internal
virtual
interfaces,
because
at
that
point
in
time
you
don't
need
inside
a
system
to
run.
B
You
know
another
crypto
layer
right
so
next
slide.
So
then,
I
added
a
section
about
Diagnostics,
which
is
usually
I,
was
funny
talking
to
other
folks
in
the
ITF
about
their
solution.
They
said,
oh,
we
forgot
about
Diagnostics
I
need
to
do
it
now,
so
these
are
all
informational
pieces,
also
the
whole
discussions
which
I
think
had
been
going
on
for
for
a
while,
but
we
never
had
good
text
for
it
right.
So
how
do
we
enable
and
disable
the
ACP
globally
on
a
node
on
a
per
interface
basis
right?
B
So
we
just
started
out
with
a
simple
example.
Well,
you
don't
want
to
do
ACP
discovery
on
an
ACP,
secure
channel,
obviously
that's
clear,
but
basically
now
there
are
all
these
details
about
brownfield
greenfield
rollout
and
the
global
and
interface
enabling
and
disabling
next
slide.
Okay
and
the
final
stuff
was,
you
know,
I
had
in
the
IDE
in
the
last
year,
I
hear
it's
always
discussion
with
folks.
That
said,
all
that
system
looks
nice,
but
there
is
one
or
two
things
that
we
don't
like
like.
B
We
already
have
some
existing
schemes
for
unique
identifiers
and
devices
with
which
we
can
create
addresses,
or
you
know
we
don't
like
this
encryption
or
that
so
was
basically
just
summarizing.
How
you
know
variations
of
the
ACP
could
easily
be
defined
proprietary
or
in
the
ITF
in
future.
Work
on
the
different
aspects
you
yeah
and
then
thanks
Bryan
30,
more
references
through
the
review.
So
ok,
that's
it
I.
Think.
A
B
F
Brian
Covington
two
points:
one
is
just
for
information.
Both
this
draft
and
the
previous
one
defines
some
grasp
objectives.
So
I
have
written
demo
code
for
both
for
all
three
objectives
concerned
and
prove
that
they
do
in
fact
work
and
produce
valid
results,
and
if
you
want
sample
messages
in
binary,
I
can
generate
them
for
you
either
of
the
authorship
teams.
If
you
want
to
include
such
examples.
G
F
All
right,
they're,
all
both
of
those
drafts
and
the
prefix
management
draft,
which
we
didn't
need
to
discuss
all
three
and
there
are
considerations,
define
grasp
objectives
and
Ayane
has
pointed
out
already
that
they
need
an
expert
reviewer
because
otherwise
they're
not
ever
going
to
let
those
drafts
can
progress.
So
we
do
need
at
least
one
Ayana
expert
appointed.
B
Drafts
will
sit
there
forever,
yeah,
maybe
just
in
conjunction
with
that.
So
we
had
some
discussions
about
how
to
make
the
grasp
discovery.
Flexible
and
you'll
see
another
draft
that
I
have
later
on
the
agenda.
That's
trying
to
pull
that
all
out
of
the
brewski
and
acp
things,
keep
the
grass
objectives
in
these
main
RFC's
as
simple
as
possible
and
then
make
the
flexibility
through
extensions.
So
maybe
that
helps
to
quicker
clothes
on
any
questions
related
to
those
objectives.
F
H
A
I
H
Also
other
important
change
in
the
document
was
about
the
security
considerations,
and
this
is
essentially
a
new
section,
and
the
one
important
change
from
the
previous
version
of
this
session
is
that
we
have
a
distinction
about
outsider
in
insider
attacks,
and
this
is
something
that
was
really
disgusted
by
the
outers
of
the
the
document.
So
the
bottom
line
on
this
distinction
is
that
when
you
have
an
outsider
tag
or
an
attacker
which
is
out
of
the
ACP,
we
have
similar
properties
to
attacks
from
other
systems.
H
Other
networking
systems
and
when
we
have
an
insider
attack,
is
some
note
inside
the
ACP
that
can
use
the
autonomic
distribution
functions
or
functions
by
the
aces
in
order
to
provide
to
perform
attacks.
So
here
the
important
thing
is
that,
as
an
atomic
distributed
environment,
the
surface
often
are
for
an
attacker
is
larger,
and
this
is
it's
the
most
important
information
about
this.
So
we
have
some
discussions.
H
We
had
some
discussions
about
if
this
text
is
clear
enough
as
a
distinction
between
these
two
kinds
of
attacks
and
as
a
new
text,
it's
very
important
that
we
have
some
reviews
and
some
feedback
from
from
the
group
next
slide,
and
so
the
next
steps
for
this
document.
We
believe
that
how
the
the
raise
concerns
and
the
reviews
and
all
the
change
required
are
performed.
So
we
are
asking
for
the
work,
move,
less
call.
H
A
F
Hi,
so
that
came
quickly,
I
wasn't
expecting
it
so
soon,
okay,
so
this
is
not
the
first
time
you've
seen
this
draft,
but
I
think
it's
getting
to
the
point
now
where
we
really
need
to
decide
whether
we're
going
to
go
forward
with
this
work
and
where
we
need
some
feedback
from
people,
who've
really
thought
about
how
they
might
use
it
or
implement
it
next
slide.
Please,
okay,
would
just
look
at
the
overview
implementation
model.
F
This
hasn't
changed.
Autonomic
service
agents
will
call
a
library
which
they
see
is
the
API,
possibly
through
an
extended
function.
Library,
if
we
add
more
sophistication
later
on
and
that
API
library
will
have
to
call
the
core
of
grasp
through
some
sort
of
inter
process
communication
method
depending
on
the
operating
system,
because
some
of
the
core
functions,
of
course,
have
to
run
in
their
own
process
and
in
their
own
memory
space.
So
next,
please
there's
a
couple
of
data
structures
which
application
programmers,
because
ASI
writers
are
basically
application.
Programmers
will
need
to
understand.
F
One
of
them
is
The
Grocer
objective,
which
has
the
name.
It
has
a
property
whether
it's
synchronization
or
negotiation
project
--iv,
and
it
has
a
loop
count
as
everybody
ought
to
know,
and
it
has
a
value
and
I
just
remind
you.
The
value
of
a
gross
projective
is
anything
that
you
can
express
in
Siebel,
which,
if
you're
thinking
in
terms
of
an
object-oriented
language,
is
basically
pretty
much
any
format
of
object
that
you
want.
F
If
you
are
in
C
well,
you
have
to
think
of
the
same
thing
at
a
lower
level,
but
there's
still
basically
no
restriction,
because
C
ball
can
do
pretty
much
anything.
The
other
thing
that
programmer
needs
to
know
is
the
concept
of
an
AS,
a
locator
which
is
either
your
own
locator
or
the
locator
of
appear.
You're.
Communicating
with
the
actual
locator
component
of
an
is
a
locator
is
typically
an
ipv6
address,
so
it
could
be
theoretically
an
ipv4
address
or
a
URL
or
fully
qualified
domain
name.
F
You
need
to
know
the
protocol
that
applies
typically
TCP
or
UDP,
but
doesn't
necessarily
have
to
be,
and
you
need
to
know
the
port
if
the
port
is
relevant
and
there
are
some
other
less
important
properties.
So
the
AP,
those
the
the
main
I
atoms
that
are
foster
around
in
the
API
next
please-
and
this
is
just
a
first
of
a
couple
of
pages
of
summary,
of
the
calls
that
we
propose
register
and
ASI
asi.
So
the
grasp
knows
it's
there
register
and
objective
so
that
grasp
knows
this.
F
A
si
can
manipulate
that
objective
and
discovery,
which
means
discover
another
remote,
a
si
that
handles
the
same
objective.
There's
one
we
tacked
on
it's
a
fairly
late
stage,
call
send
invalid,
which
you
can
send
at
any
time
to
say.
I
didn't
understand
that,
but
that's
not
really
very
interesting.
Next,
please,
then,
for
negotiation.
F
Negotiation
is
symmetrical
as
defined,
but
it
has
to
start
as
a
client-server
interchange,
because
one
side
has
to
tell
the
other
side.
I
want
to
negotiate.
So
there's
a
request.
Negotiate,
which
is
our
an
SI,
tells
another
one.
It
wants
to
negotiate
and
there
is
a
listen
negotiate,
which
is
how,
in
a
si
says,
I
am
willing
to
receive
request,
negotiates,
but
once
you've
matched
a
request
to
listen,
everything
else
is
symmetrical
and
there
are
three
options
negotiate
step,
which
is
where
you
send
an
objective
to
your
peer
saying.
F
This
is
what
I
want
next
and
if
Pierre
can
reply
saying
this
is
what
I'll
give
you
and
negotiate
wait.
If
you
want
to
insert
a
delay
in
the
process
and
then
negotiate
when
you
want
to
end
the
negotiation
and
the
result
of
the
negotiation
is
success,
then
you
have
nothing
to
say,
but
if
the
Gulf,
the
result
of
the
negotiation
is
failure,
you
give
a
reason.
Next,
please
synchronization
is
similar
but
simpler.
F
You
can
synchronize
an
object,
meaning
I
want
the
value
of
this
objective
or,
you
can
say,
listen
synchronize,
which
means
I
am
willing
to
hand
out
the
value
of
this
objective.
To
anybody
who
wants
it
and
flood
is
the
broadcast,
while
the
multicast
version
of
this,
where
you
simply
send
the
objective
out
to
all
nodes
and
get
flood,
is
an
additional
call
that
someone
can
use
if
they
want
to
find
out
it's
an
objective
has
been
flooded,
though
the
way
it's
defined.
F
If
you
do
synchronize
and
the
value
has
been
flooded,
the
synchronized
call
will
return
the
flooded
value
to
you.
Okay.
Next,
what
we've
done
recently
we've
updated
everything
to
match
the
approved
version
of
grasp.
We
have
added
a
note
that
a
simple
node
might
not
want
this
API.
This
simple
node
might
have
grasped
built
into
its
main
line
code,
so
this
is
not
compulsory
and
we
added
send
invalid
next,
please,
a
few
things
are
missing.
F
F
There
are
two
approaches
to
handling
the
asynchronous
nature
of
grasp.
One
is
to
assume
a
threaded
environment,
and
this
means
that
quite
a
few
of
the
calls,
because
they
can
run
in
their
own
thread,
can
be
blocking
calls.
Good
example
is
discover
which
blocks
until
you
get
back
a
discovery
result
which
can
of
course
take
some
time.
F
The
other
implementation
model
is
to
not
assume
you
have
a
threaded
operating
system
thread
in
language,
in
which
case
you
have
little
choice.
You
have
to
use
an
event,
loop
type
structure,
which
basically
means
that
you
can't
have
blocking
cause
in
the
API,
because
otherwise
you
will
block
your
event
loop
on
each
on
each
of
these
calls
there's
a
model
proposed
in
the
API
for
dealing
with.
F
That
is,
if
you
call
something
in
this
mode,
which
would
otherwise
be
a
blocking
call
they
all
simply
if
it's
not
got
a
nice
answer,
for
you
will
simply
give
you
a
return
code,
saying
no
reply
and
when
you
finally
get
a
callback,
you
will
get
back
a
genuine
answer.
So
you
just
repeat
your
call
in
the
event
loop
until
you
get
back
an
answer.
J
B
So
we're
never
doing
it
that
way,
so
no
I
mean
some
of
the
things
that
you
know
through
the
also
the
security
review
and
other
stuff.
That
I
think
will
impact
the
API
would
be
to
expose.
You
know
the
security
properties
of
the
underlying
a
transport
that
you
have.
Whether
for
example,
you
need
fragmentation
or
you
know,
message
size
limitations
or
you
have
the
security.
So
I
think
there
are
some
some
good
things
to
add
on
that
side.
Thanks.
A
Yeah
any
more
from
the
room,
those
people
sandwich.
This
is
important
of
all.
We
should
repay
all
that
makes
sense
to
you
for
your
guys,
I
mean
for
me,
I
I,
think
you
know
this
is
very
important
work.
It
actually
increases
the
reusability
of
grasp
and
that
will
makes
the
ASA
much
either
factor
III
carry
I,
do
have
a
question
for
you
regarding
to
this
document,
because
that's
our
API
and
that's
our
API
between
the
functions
within
our
devices.
A
K
Good
question
I
want
to
come
back
to
you
on
that
one
because
I
don't
have
an
answer
straight
away:
okay,
traditionally
we'd
like
to
think
about
things
on
the
wire,
but
then
that's
Mount,
Rainier
right,
but
in
a
situation
like
this,
if
the
interoperability
within
the
the
two
functions
is
not
appropriate,
this
is
going
to
impact
implementation.
So
right
now,
I,
don't
know
all
right.
Yeah.
F
L
L
H
A
So
I
would
like
to
get
the
sense
from
the
room.
People
think
this
work
is,
you
know,
make
sense,
and
we
should
you
know
as
working
group
care
about
it,
whether
that's
adopt
as
informational
document
or
a
stand
tract
wind.
We
can
discuss
that
later,
yeah,
okay,
so
people
who
think
this
work
is
makes
sense
and
the
working
culture
care
about
it.
Please
home
now.
A
A
K
It
it
struck
me
that
that
that
call
really
just
you
know,
showed
a
whole
heap
of
apathy
within
the
room
regarding
this
topic,
so
I
I,
don't
think,
there's
any
judge
of
consensus
there.
This
can
can
go
forward.
So
please
take
that
to
the
mailing
list
and
I'll
be
watching
carefully
sure
on
that,
it's
a
working
group.
You
need
people
to
to
be
engaged
and
to
work
on
these
things.
Yeah.
Yes,
yes,
so.
G
F
G
There
should
be
a
separate
module
to
handle
the
distribution
function.
Next,
please!
So
you
will
see
this
in
the
API
view.
Just
as
brand
mentioned,
there
could
be
some
extended
function
based
on
clasp.
So
that's
the
information
distribution
position
fits
into
the
Animas
architecture.
Reference
model.
Sorry,
no
architecture,
okay!
Next,
please,
so
the
current
draft
describes
some
very
basic
patterns.
G
G
G
G
G
We
call
it
a
selection
criteria
to
let
the
module
knows
in
what
conditions
they
should
decide
to
send
information
to
the
neighbor
or
not,
and
the
third
one
is
is
in
a
real
deployment.
The
devices
will
get
on
land
incrementally,
so
some
information
just
cannot
distribute
to
other,
not
in
in
just
one
time.
So
if
the
nod
can't
detect
whether
they're
our
new
neighbors
come
in,
then
it
can
send.
G
A
G
G
So
here,
I'd
do
a
summary
of
the
current
distribution
function
is
most
of
the
lira
guarding
to
pushing
some
information.
Why
not
just
push
it
to
the
others,
but
how
about
one
not
just
to
collect
something
poor
for
something?
I
think
this
should
be
also
in
the
scope
of
information
distribution,
so
in
that
ATF
and
other
give
a
nice
presentation
about
integrating
pops-up
mode
into
the
a
nice.
F
G
Next,
please,
okay,
so
in
in
my
opinion,
this
is
a
more
complete
picture
for
the
information
distribution
function
and
there
are
several
extensions
need
to
be
done.
The
first
one,
the
Select,
still
flooding
and
unsolicited
synchronization
pops
up
and
the
park
transport,
so
I
think,
with
these
more
complete
capabilities,
the
N
I
can
provide
a
kind
of
full
function
for
the
uglier.
Is
us
and
the
self-contained
mechanism
can
reduce
the
code
space
and,
more
importantly,
it
means
we
don't
need
any
other
pieces
of
protocols
to
be
combined
together
to
provide
some
function.
F
G
B
B
Think
it's
a
good
start
with
the
you
know:
requirements
it's
not
really
I
think
conclusive
as
a
proposed
solution
and
I
think
what
might
be
you
know
already
in
in
scope
of
the
Charter,
is
trying
to
figure
out
if
we're
missing
in
grasp
itself
core
things
that
need
to
amend
it
to
support
this,
but
I
would
hold
my
you
know
my
judgment,
also
after
Arthur's
presentation,
because
I
think
it
goes
in
the
same
direction
with
the
different
terminology
and
everything.
So
maybe
more
discussion
after
author's
presentation
might
be,
might
be
useful.
Okay,.
I
Thanks
hi,
Otto,
eka
and
so
I
find
it
extremely
useful
in
the
sense
that
we
have
in
the
reference
model,
the
information
distribution.
So
somehow
we
need
to
do
it.
However,
indeed
I
do
not
care
really
where
it's
done
in
the
sense.
We
need
to
implement
it
somehow,
whether
it's
grass
doing
it.
Okay,
fine,
let's
do
it
in
grass.
If
not,
then
we
do
it.
We
need
to
do
it
on
the
top
right,
and
these
are
the
issues
we
are
trying
to
discuss
since
the
last
time.
G
A
Now
it
seems
most
of
this
document
is
focus
on
the
requirements
and
also
actually
you
own.
You
haven't
really
covered
the
pops
up
and
packed
rest,
but
within
the
current
form,
yet
I'm
saying
he
may
be
the
best
way
would
be
first
to
make
a
requirements.
Document
and
I
know
earlier
and
I
said
you
know,
I
mean
want.
You
know,
I
integrate
document
to
do
both
requirements
and
they
the
solutions
but
actually
giving.
F
A
Believe,
that's
mostly,
you
match
the
current
charter,
because
that's
still
given
the
some
function,
which
can
be
reusable
for
most
of
the
unknown
caves
network,
functions
whether
we
do
that
in
the
pure
grasp
extension
way
or
make
it
us.
You
know
some
general
autonomic
service
agent
that
had
another
method
that
far
either
way.
Yeah
I
agree.
G
F
J
J
So
this
is
just
the
agenda
of
the
talk
start
with
what
we
want
to
achieve
with
this
document
then
go
through
an
overview
of
the
structure.
The
recent
changes,
the
main
points
for
attention
of
this
presentation
and
then
discussion
next
step.
So
what
we
want
to
achieve
with
draft
the
main
idea
is
first
movie
target
our
targeted
user
and
rid
of
this
draft
out
developer
of
the
ASI
autonomic
functions
in
animal
working
group.
J
We
have
been
working
a
lot
on
the
protocols
and
the
NI
drastic
cetera,
which
had
all
protocols
to
make
it
run,
but
there
will
be.
We
expect
a
lot
of
people
and
ecosystem
of
developers,
developing
autonomic
functions,
big
service
agents
in
the
anime
terminology,
and
so
we
want
to
address
their
needs
because
they
will
not
necessarily
be
experts
in
the
animal
protocols.
J
They
will
probably
be
experts
in
the
functions
the
autonomic
functions
they
would
like
to
design
and
order
would
like
to
attempt
to
operate
so
essentially,
Brian
previously
presented
the
draft
on
the
API
in
which
is
a
first
step
towards
extracting
the
cause
of
grasp
towards
a
developer
community,
but
it
will
not
be
enough.
So
we
want
also
to
in
this
draft
expose
practices
guidelines
how
developers
what
they
need
to
focus
on
what
they
need
to
concentrate
when
they
are
developing
their
codes
next
slide.
J
So
this
is
the
current
structure
of
the
draft.
There
is
first
an
explanation
or
details
on
the
logical
structure
of
ERISA,
as
defined
in
animal
then
important
part
on
the
interaction
with
it
within
the
autonomic
infrastructure,
the
Ani,
a
chapter
on
design
of
grass
tibs,
the
relationship
on
the
A's
are
structured
and
all
they
need
to
capture
the
grasp
objective
and
and
all
them
a
section
on
the
lifecycle
of
a
deployment
and
installation
of
phaser
section
and
coordination
which
is
still
empty
today
and
to
be
discussed
and
robustness.
Section
and
security
considerations.
J
J
So
first
the
ASR
runs
within
the
ACP
and
they
use
grass
for
the
communication,
but
there
are
also
a
number
of
hospital
needs
to
be
covered,
so
assumption
is
that
ASA
essentially
should
be
a
multi-threaded
software,
but
you
can
also
rely
on
a
more
even
group
approach,
but
you
have
to
include
you
have
to
incorporate
that
needs
to
be
a
certain
set
of
processes
running
in
the
ASA
by
default.
It
must
be
self
monitoring
and
self
restarting
in
case
of
failure
in
the
operation
of
the
ASA.
J
It
should
be
able
to
identify
it
by
itself
and
try
two
different
ways
of
restarting
the
process:
its
own
process.
There
should
be
threads
for
ending
the
different
processes
related,
for
instance,
to
grasp
all
the
fluid
equalization
negotiation
of
objectives,
but
also
for
order
housekeeping
functionality
to
manage
order
potentially
na
autonomic
devices.
Next
slide,
the
grass
objectives
follow
the
Grubbs
rules.
This
is
seems
logical,
but
grasp
currently
provides
no
transactional
integrity,
so
it
will
be
up
to
the
ASA
developers
to
end
out
happens
in
terms
of
logs
and
atomicity
of
all
the
operations.
J
J
So
this
is
a
section
on
the
lifecycle.
This
is
a
first
time
in
this
draft,
so
since
we
imagine
that
the
enema
network
will
be
continuously
evolving
and
reconfigured,
there
needs
to
be
a
systematic
lifecycle
in
order
to
install
in
sunshade
and
operate
at
the
different
AIDS
are
running
so
description
of
a
short
life
cycle
and
aspect
related
to
this
life
cycle
for
for
developers.
G
F
G
You
mission
in
general,
the
SS
you
support
multi
thread,
but
in
some
very
constrained
devices
such
as
the
LT,
not
maybe
you
can
also
have
some
description
of
very
tiny
ASA.
It
just
has
one
thread
and
one
very
specific
objective.
So
in
that
case
the
either
could
be
a
very
simple
and
don't
need
any
multi-threading.
G
The
other
first
come.
The
second
comment
is,
in
you
know
in
fue
scenarios
there's
a
concept
of
macro
service
which,
in
my
opinion,
is
very
similar
to
the
ASA
concept
in
anima.
So
maybe
there
are
some
explanation
between
the
concepts,
similar
concepts
in,
in
my
point
of
view,
I
think
the
they
are
essentially
the
same.
Maybe
we
can
have
some
discussion
on
that.
Sorry.
D
J
So
I'll
try
to
comment
on
your
comments,
the
first
one,
which
was
on
the
material.
It's
a
what
we
are
currently
in
the
draft
are
Iran's
own
assumption
and
there
is
also
the
introduction
of
the
event
loop
approach
so
for
sure,
depending
on
the
requirements
of
constraints
of
the
different
aiza
and
environments
in
which
that
will
run.
This
is
not
to
prescribe
any
monetary
implementation
to
the
developer,
but
just
to
give
guidelines
about
what
they
should
really
care
about
when
developing
their
own
functions.
J
It
sure,
if
it
really
constraints,
they
will
concentrate
on
having
efficient
light
code
etc.
So
they
can
rely
on
maybe
a
single
thread
or
event
oriented
approach.
But
it
should
be
aware
that
behind
that
grasp,
maybe
have
a
synchronously
so
that
you
should
not
lock
the
process
of
grass
because
you
can
not
only
in
a
serial
manner
and
the
second
comment
on
nav
I
think
this
is
a
more
general
discussion.
We
need
to
have
about
this
draft,
but
more
but
animatronic
functions.
If
we
expect
this
to
be
software,
just
software
pieces
of
software.
J
Yes,
there
are
some
other
contexts
like
nav
micro
services,
but
any
kind
of
embedded
code.
We
should
get
best
current
practices
and
see
what
are
the
specificities
of
autonomic
networks
we
need
to
describe
in
those
drafts,
for
instance,
like
coordination,
the
life
cycle
may
be
specific.
The
fact
that
they
are
running
in
close
control
loops
may
be
different
from
other
environments.
So
this
is
what
we
need
to
capture
and
not
not
to
do
in
vendor
wheel
if
there
are
very
good
practices
for
software
the
roadman.
This
is
not
the
goal
of
this
document.
B
So,
let's
Eckert
so
I'm,
not
even
sure
if
I
should,
if
this
is
a
comment
from
an
individual
or
as
a
working
group
chair
right,
so
I
think
the
first
thing
is
that
I'd
be
worried
that
this
would
be
considered
to
be
outside
of
the
current
Charter
scope
right,
given
how
it
deals
with
ASAS.
Secondly,
obviously
you
know
I
I
think
this
is
this
is
interesting
and
should
be
done,
but
we'll
need
to
figure
out.
You
know
how
and
and
when
we
do
it,
you
know
with
recharge
ring
or
so.
B
Thirdly,
the
the
question
is
always
a
little
bit
what
the
ITF
likes
to
do
and
there
seems
to
be
some.
You
know.
Distinction
between
things
that
you
know
are
done
on
interface
is
called
api's
versus
interfaces
that
are
called
you
know,
yang
and
so
I
think
we
discussed
also
in
some
of
your
prior
drafts
that
this
may
be
potentially
easier,
be
accepted
by
the
ITF.
B
If
it's
seen
as,
for
example,
starting
on
dependencies
between
you,
know,
different
data
models
that
are
representing
what
a
particular
mod
a
si
is
doing
right
as
opposed
to
starting
it
from
the
typical
software
development
perspective,
right
thinking
more
about
the
data
modeling
aspect
of
these
different
ASAS
and
how
they
interact
with
each
other
right.
Maybe
there's
also
another
approach
to
think
about
how
to
get
more
acceptance
in
the
idea
because
they
seem
to
be.
You
know
specific
things
they
like
and
they
don't
like.
J
F
Point
is
a
good
one.
We
do
need
to
think
about
that,
because
if
we
propagate
one
religion
and
if
he
propagates
a
different
religion,
then
you
know
whoever
I
don't
know
end
up
in
trouble.
So
I
think
we
really
should
take
that
point
rather
seriously.
I'm,
not
particularly
worried
about
the
problem
of
very
simple
ASAS.
As
I
said
on
the
API
discussion,
they
might
not
even
need
the
API.
F
F
I
think
we're
probably
in
this
draft,
focusing
on
what
you
might
think
of
as
high
end
ASA's,
which
are
relatively
sophisticated,
which
really
are
application
programs
and
if,
if
somebody
decides
in
the
end
that
it's
out
of
ITF
scope,
though
we'll
publish
it
as
an
independent
submission,
RFC
I
guess,
but
you
know,
I
think
we
are
genuinely
worrying
about
having
an
ecosystem
built
on
top
of
the
infrastructure.
That's
why
this
dropped
a
little
bit
difficult
to
position.
F
L
J
J
We
need
to
better
understand
in
a
group
what
we
think
those
either
will
be
in
terms
of
software's
I'm,
a
piece
of
software
and
then,
if
we
think
that
I
will
be
really
standalone
and
we
these
big
pieces
of
software,
that
maybe
they
will
not
be
suitable
for
microsoft
services,
but
you
cannot
answer
that
can
be
implemented
in
micro-services
approach.
I
just
want
to
make
the
distinction
is
that
yeah
for
the
working
group.
We
need
to
understand
what
we
think
si
will
be
designed
like
and
then
that
would
be
supported
by
different
software
patterns.
J
As
you
mention
not,
you
know
to
say
that
every
ADA
should
be
implemented
as
a
micro
service.
I.
Don't
think
this
is
really
a
mistake,
but
to
say
that
if
they
are
implementing
the
micro
service,
we
need
to
understand
what
will
be
the
impact
on
the
design
of
of
the
SI
and
what
guidelines
we
should
provide
to
the
ADA
developers.
That
I
mean
maybe
they
will
not
understand
anything
about
micro
service
but
to
understand
that
if
they
want
to
achieve
some
level
of
parallelism
or
distributivity
I
need
to
understand
those
needs.
A
She
without
my
work
with
your
head,
actually
I'm
cause
of
this
document,
but
I
would
like
to
take
that
his
cousin,
wider
than
the
document
you
know
to
be.
You
know
how
this
working
group,
should
you
know
care
about,
say
si,
because
you
know
for
me,
you
know
we're
only
so
far
we're
only
defying
the
autonomic
network
infrastructure
at
the
components,
but
to
be
able
to
make
them
useful.
A
There
should
definitely
be
our
own
service
agent,
but
actually,
if
those
agent,
you
know
very
specific
for
very
narrow
scenarios,
I
believe
that's
out
of
the
scope
for
the
this
working
group,
but
yeah.
If
you
know
there
are
two
categories,
why
is
you
know
this
working
group
should
working
should
care
about?
You
know
how
to
make
the
a
autonomic
nerve
network
infrastructure
and
its
components
to
be
easier
for
the
you
know
out
for
the
implementations
or
for
the
operators
to
use
I
believe
this
document
is
one
of
those
efforts.
A
Second,
if
you
know
there
are
some
generic
aces,
which
you
know
can
be
used
for
most
of
scenarios
or
covered
a
lot,
multiple
scenarios
that
should
be,
you
know,
considered
assumption
between
the
you
know,
exchanging
of
eternal
nominal
network
infrastructure
and
aces.
So
such
such
asa
may
be
concerned
by
the
working
group,
and
actually
you
know
giving
my
example
even
the
SME
and
the
broski
could
be
also
be
concerned.
As
you
know,
one
type
of
asa
as
well.
M
Problem
of
coordination
of
all
of
them
is
going
to
be
fundamental,
it's
not
something
which
will
be
easy
to
add
after.
Therefore,
the
coordination
and
orchestration
of
this
Asia
in
some
sense
have
to
be
added
to
the
problem
from
beginning
like
like
a
northern
ecosystem
for
this,
although
aces
and
I
think
see
that
how
this
is
done,
your
father,
but
it
should
be
in
my
day.
Second,
in
order
to
be
practical,
this
is
I
should
have
a
level
of
recursiveness,
in
other
words,
to
build
some
answer
from
others.
J
F
M
J
Slated
to
come
back
on
your
comments,
the
first
one
on
coordination,
so
there
is
currently
a
section
in
this
draft
and
I
mean
we
have
in
the
plans
to
try
to
fill
it.
There
was
previously
in
anima
dividual
draft
on
coordination,
so
we
welcome
anyone
willing
to
contribute
to
this.
I.
Also
think
coordination
is
important,
I,
don't
know
to
which
extent
it
should
be
in
this
draft
or
in
another
item
for
the
working
group.
J
But
not
not
this
session,
but
the
aspect
about
deploying
and
configuring
ASA
through
the
aspect
of
intent,
but
I
think
we
had
already
discussion
on
that.
It's
a
bit
posed
for
the
moment.
We
are
discussing
how
to
address
this
word,
but
I
think
addressing
a
more
general
approach
to
configure
groups
of
ASA
can
be
linked
to
the
discussion
on
policy
or
our
Indians,
not
necessarily
only
this,
but
it
has
been
discussed
in
the
bus
on
this
aspect.
J
D
A
F
Okay,
I
I
think
this
is
the
last
time
you'll
see
me
today
next
slide,
please
so
gonna
talk
briefly
about
this
new
new
work
motivation.
What
the
proposal
is
and
discussion
of
some
issues
and
next
steps.
Next,
please
so
gross
was
designed
around
the
notion
of
an
individual
technical
objective,
which
is
a
relatively
small
data
structure
that
can
be
sent
in
a
message
and
we
set
the
message
size
limit
to
2k.
So
it's
no!
It's
not
an
enormous
piece
of
data.
F
However,
as
we
heard
from
one
or
two
other
talks,
a
an
autonomic
network
may
need
to
distribute
larger
data
objects,
think
of
them
as
files
or
whatever
you
like,
but
things
that
are
clearly
not
going
to
fit
into
a
single
message,
though
there
are
many
ways
to
do
that
right.
There's
this
thing
called
HTTP.
This
thing
called
FTP
and
you
know
we
know
how
to
do
this.
G
F
A
Would
like
to
make
a
comment
here
personally
actually
likes
this
motivation
very
much,
because
that
is
very
beginning
of
enema.
We
said,
if
that's
possible,
we
would
like
to
have
grasped
as
C
as
a
single
single
loading
protocols,
actually
zero,
a
you
know
in
the
real
world
we
actually
he
and
I
owed
his
generous.
We
already
you
can't
grasp
as
the
owning
signaling
protocols
and
we
are
they
even
use
grasp
modified
grasp.
A
F
Okay,
so
the
basic
proposal
is
to
used
cross
promotion,
which
remember,
is
a
backwards
and
forwards
process
with
a
lot
any
number
of
steps
theoretically
possible,
but
use
that
negotiation
mechanism
to
transfer
your
bulk
data
one
block
at
a
time,
so
the
communication
model
is
slightly
asymmetric
as
you'll
see
compared
to
a
traditional
cross.
Pronouciation.
F
The
goal
is
simplicity,
not
performance,
I
should
emphasize.
If
you
want
high
performance
bar
better
transfer.
This
is
not
the
solution.
So
procedure
looks
like
this:
the
client
we
have
to
make
it
a
little
bit
asymmetric.
Think
of
one
of
the
sides
as
a
client,
the
other
side
of
the
server
the
client
discovers
the
server
using
standard
grasp
discovery
and
response.
F
Then
the
client
requests
the
file
that
it
wants
by
sending
a
negotiation
request
and
the
value
of
the
objective
that
it
sends
is
the
file
name.
If
you
think
of
this
jewelry
as
file
transfer
mechanism,
the
server
sends
back
the
first
data
block,
which
is
simply
a
negotiation
step
using
the
negotiate
message
and
the
client
acknowledges
it
with
another
negotiation
step
in
return.
F
So
you
see
the
asymmetry
here
that
on
the
message
from
the
server
to
the
client,
the
value
of
the
objective
is
actually
the
data
block
of
the
file
on
the
message
from
the
client
back
to
the
server.
The
value
of
the
objective
is
well
I
suggest
the
string
back,
but
it
could
be
anything
you
like
as
long
as
it's
understood
to
mean
acknowledge,
and
you
repeat,
steps
3
&
4
until
you
reach
the
end
of
the
file.
F
F
L
D
F
F
A
unique
objective
name
I've
been
using
it
routinely
to
move
files
between
a
Windows
machine
and
Linux
machine,
because
it's
easier
than
using
a
thumb
drive
for
pushing
them
up
into
a
github
and
pulling
them
down
again
and
the
client
plus
the
server
coming
to
about
250
lines
of
215
lines
of
Python.
So
it
really
is
very
simple
and
straightforward.
B
Question
so
I
have
a
couple
of
essays
between
two
notes:
they're
doing
some
things
and
now
I'm
bringing
up
these
two
ASAS
that
wanted
with
the
bulk
transfer
I
am
still
a
little
bit
confused
about
the
multiplexing
of
brass
three.
Would
there
be
any
risk
of
head
of
line
blocking
and.
F
F
B
F
B
So
you
know
whether
or
not
ASA's
would
like
to
use
some
existing
file
transfer
standard,
or
you
know
we
come
up
with
a
new
one
like
this
I'm,
not
sure
how
to
make
a
choice
between
that
right.
So
your
justification
is
a
simple.
You
know
some
applications
may
want
to
use
it.
So
I
think
it
would
be
good
to
have
more
brainstorming
on
the
justifications,
Pro
and
German.
F
G
F
G
G
F
Well,
this
is
a
first
draft.
I
mean
I,
decided
to
implement
it
to
see
if
this
syntax
for
locally
defined
objectives
worked
right.
If
you'd
like
me
to
put
in
an
ionic
considerations
and
use
it,
you
know
it
when
I
can,
but
this
this
is.
You
know
this
is
very
preliminary
test
that
I
did
so
a
we
could
write
it
as
a
general,
a
general
purpose
thing,
yeah,
absolutely
I
simply
didn't
do
so
for
the
first
draft.
A
F
If
we
take
his
fraud
in
the
working
group,
I
would
assume
we
try
to
do
a
standard
version.
Worse,
I
just
built
a
toy.
You
know
which
no
took
about
an
hour
by
the
way
it
so
it
was
very
easy
work
and
did
it
this
way.
Okay,
somebody
put
up
the
next
slide.
Okay,
so
we
got
some
feedback
on
this
room
from
Joel
helpin
actually,
and
he
raised
three
quite
interesting
questions
which
is
really
which
are
really
questions
about
how
general
we
want
to
make
this
right.
F
F
More
seriously,
Joel
made
the
argument
that
we
really
should
add
in
a
position
indicator
or
a
block
number
indicator,
because
there's
a
risk
of
getting
out
of
step
now.
I
maintain
there
isn't
because
actually
grasp
negotiation
is
a
strict
and
shake
my
turn.
Your
turn,
my
turn
your
turn,
so
I
don't
believe.
F
There
is
any
way
we
can
get
out
of
step,
because
if
a
block
doesn't
arrive
or
if
an
acknowledge
doesn't
arrive,
you
know
that
you've
lost
the
transmission,
but
Joel
made
the
argument
that
that's
not
good
enough
and
you
should
include
the
block
number,
which
would
of
course
be
a
trivial
change
to
the
protocol.
I
described.
F
The
third
one
is
more
difficult
because
he
said
well,
you
know
you
can't
assume
things
are
going
to
work
in
in
real
life,
and
so
you
need
to
be
able
to
resume
a
file
transfer
which
you
could
do
if
you
get
out
of
the
block
number
you'd
have
to
add
a
little
bit
more
protocol
machinery.
If
you
are
going
to
resume
a
failed
transfer
rather
than
restarting
it
and
then
I
thought
well,
am
I
really
going
to
implement.
F
You
know
a
whole
of
FTP
functionality
in
this,
because
if
you
put
those
three
things
together,
then
you
know
well,
why
not
also
implement
passive
mode
and
you
could
do
practically
anything.
So
the
question
is
whether
we
would
want
if
we
take
this
work
forward,
whether
we
want
to
keep
it
very
simple
and,
except
that
it's
inefficient
and
and
in
in
failure
conditions,
it
would
be
really
clunky
and
slow
or
do
we
want
to
make
it
a
fully-featured
protocol
with
all
the
bells
and
whistles
that
we
know
about.
F
H
F
G
H
F
B
Hope
nobody
expects
me
to
answer
that
question
right
now,
sir
I
was
trying
to
talk
about
something
else
the
so
you
know
if
we
keep
this
information,
it
sounds
like
a
good.
You
know
recommendation
for
you
know
how
to
use
grass
to
do
something
that
you
know
many
grass
applications
would
want
to
do
and
I
think
that
should
still
be.
You
know
in
charger
for
the
grasp
charger
item
but,
of
course
that's
something
to
be
discussed
with
the
ad
he's.
B
I
know
no
I
mean
the
the
only
thing
you
know
since
since
we
started
grasp,
and
then
you
know
at
least
for
the
ACP
or
so
arrived
at
the
conclusion
that
were,
you
know,
going
to
not
doing
the
unicast
at
all
with
datagrams,
but
just
with
you
know,
reliable
transport
connections,
hopefully
even
secure.
The
question
is
a
little
bit.
You
know,
is
the
segment
limitation
in
grasp
as
what
you
find.
It's
still,
you
know
viable
beneficial,
or
should
we
rather,
you
know
even
think
about
updating
that
and
saying
hey?
B
B
B
F
B
F
B
K
F
We
have
this
set
of
problems
that
being
talked
about
for
distribution
right
and
we've
stopped
talking
about
intent,
because
we
don't
know
what
it
is,
but
the
fact
is
once
we
do
know
what
intent
is.
We
will
need
to
distribute
it,
so
we
need
something
right,
I'm,
not
saying
we
need
this
I'm
just
saying.
Well,
we
could
do
this
that
if
we
don't
have
this
we'll
have
to
use
some
other
solution
for
the
problem
which
you
know
could
be
to
TCP
as
still
assert.
F
F
K
M
G
Will
download
image
for
4h,
so
in
that
case,
if
grasp
and
provides
such
capability,
then
no
other
configuration
for
the
dedicated
transport
FTP
or
TFTP,
and
that's
one,
the
other
one,
maybe
a
longer.
The
ASA
itself
could
also
be
delivered
if
it
is
a
reservist,
some
kind
of
small
software
and
it
could
be
dynamically
deployed
on
devices.
So
it's
again.
F
K
To
the
mic,
please,
just
paraphrasing,
he
just
described
to
simple
use
cases.
One
was
about
configuration
and
the
other
was
a
was
about
dynamically,
a
signing
function
through
basically
an
image.
So
you
know
that
that
has
a
nice
play
scenario
in
IOT
and
the
you
know
the
things
that
are
happening
in
suit
and
so
on
and
so
forth,
but
okay
I'll
leather
discussion
to
continue.
Thank
you.
L
F
N
Good
morning,
everyone
I'm
machine
and
Otto
is
also
here,
so
we
both
can
answer
your
questions
if
you
do
have
any.
So
it's
a
little
bit
of
surprise.
That
being
already
mentioned,
that
information
distribution
is
very
important
for
the
autonomic
networking
topics.
So
here's
a
little
bit
more
details
and
our
thinking
and
idea
about
this
part.
We
are
kind
of
proposed
and
our
alternative
and
mechanism
for
how
to
do
the
information
distribution
next
slide,
please.
N
So
this
work
is
actually
for
the
of
course,
an
autonomic
networking
infrastructure
extension
and
try
to
propose
and
and
useable
components
for
this
working
group.
Next
slide.
Please,
okay
and
this
work
is
also
following
our
last
presentation
about
pubs
and
subtopic.
So
here
we
and
this
time
we
proposed
and
to
include
a
kind
of
event
service
into
the
autonomic
networking
as
a
reusable
component.
A
N
N
Currently,
according
to
our
investigation
study,
we
found
that
the
communication
model
in
the
current
autonomic
networking
infrastructures
is
kind
of
a
tide
that
we
cope.
The
client
and
server
model,
which
means
that
our
our
autonomic
function
our
standard.
We
are
directly
send
a
message
to
a
destination
node,
so,
for
example,
we
can
and
from
the
grasp
we
can
use
these
for
the
dynamic
peer
discovery
and
stay
synchronization
and
north
of
parameters
set
in
negotiation.
N
We
found
that
and
possibly
in
some
times
the
destination
node
is
not
available
and
also
some
kind
of
requests
and
the
communications
is
not
instant.
For
example,
you
have
to
wait
for
a
while
so
for
those
kind
of
scenarios-
and
there
is
no
any
kind
of
solutions
covered
by
the
autonomic
networking
so
far.
N
Next,
please,
okay,
here
also
mentioned
by
Shen-
that
we
kind
of
need
a
very
strong
requirement
and
why
we
need
another
or
another
kinds
of
communication
models
in
not
anomic
networking
infrastructure
here,
and
we
did
kind
of
investigations
and
see
some
scenarios
or
examples
might
need
new
communication
model,
which
is
an
a
synchronous
communication.
So
here.
F
N
Four
examples:
the
first
one
and
it
is
that
the
replying
of
the
other
node
of
the
destination
takes
longer
time
than
the
direct
communication.
So,
for
example,
if
our
economic
function
wants
to
and
ask
the
other
node
to
do
some
database
search
or
lookup,
you
may
take
longer
time
than
the
simple
direct
communication
model.
So
the
second
one
is
that
what
if
several
nomicon
wants
to
share
some
common
information,
for
example
in
the
information
distribution
and
draft,
we
mention
that
the
intent
of
the
network
has
to
be
distributed
across
the
network.
N
So
in
that
case,
some
shell
to
information
or
interest
has
to
be
distributed
across
the
network
and
the
third
one
is
that
probably
the
autonomic
functions
in
the
network
may
have
their
local
views,
which
is
different
from
each
other.
In
that
case.
In
that
case,
those
autonomic
functions
have
to
collaborate
with
each
other
to
do
some
network
scheduling
or
to
do
to
make
some
to
make
some
decisions.
In
that
case,
with
local
view
of
the
network,
how
could
they
converge
your
common
view
of
the
network?
N
In
that
case,
they
have
to
constantly
exchange
and
synchronize
their
local
information
together.
In
this
case,
we
also
need
kind
of
an
information
storage,
for
example.
Putting
some
information
from
our
donors
mentioned
by
being
the
Circa
is
that
sometimes
some
information
has
to
be
calculated
and
before
it
can
be
used.
For
example,
we
are
the
sense
of
the
network
and
the
autonomic
function
has
to
come
up
with
kind
of
a
collective
decision.
What
is
the
average
of
the?
N
What
is
the
average
temperature?
What
is
the
average
value
of
the
whole
network?
In
that
case,
we
have
to
collect
the
data
and
do
some
calculations
and
put
it
somewhere.
Ok
next
slide.
So
in
general,
here
we
find
that
an
in
addition
to
the
direct
communication
model,
which
is
already
covered
by
the
current
autonomic
networking
and
reference
models.
N
How
about
we
introduce
some
kind
of
interact
and
assign
Qo'nos
communication
model,
which
is
even
service,
so
here,
in
addition
to
the
curve
in
the
synchronize
model,
we
propose
to
extend
with
even
service,
which
means
that
to
interacted
parties
will
be
decoupled
and
the
sender
can
publish
some
information
and
the
receiver
who
are
interested
in
such
information
can
just
simply
subscribe
that
information
and
the
event
service
module.
The
reusable
component
will
be
responsible
for
handling
that
information.
N
B
N
Yes,
you're
right
and
pubs
and
SAP
and
here
the
event
service
and,
in
our
opinion,
is
to
enabled
the
pubs
and
SAP
so
pubs
and
sub
should
have
a
kind
of
even
service
so
that
it
can
can
can
do
to
a
synchronous
communication.
So
we're
not
saying
that
this
is
replacement
of
pops
and
sub.
Actually,
it
is
for
pops
hands
up
and
your.
B
N
I
Also
answering
directly
so
usually
when
you
have
pops
up,
you
need
some
kind
of
decoupling
between
the
sender
and
the
receiver,
so
you
need
some
intermediate
party
to
essentially
store
messages
or
you
know,
backlog
them
or
whatever,
and
that's
essentially
our
way
of
saying.
Okay,
we
can
call
the
event
service.
We
can
call
pop
sub
service
I,
don't
care
it's
just
a
name,
but
in
implementation
component
to
enable
pops
up.
Yes,.
N
Okay,
here,
we
just
simply
propose
kind
of
alternative
ways
to
to
include
such
an
even
surface,
how
to
include
that
even
servicing
in
our
autonomic
networking.
So
the
first,
of
course
is
we
can
develop
such
an
event
service
as
a
an
individual
or
customized
and
Isis
and
included
the
two
different
application
autonomic
at
number
application
and
functions.
This
is
the
first
way
we
we
see.
The
second
way
which
we
prefer
is
that
we
can.
We
may
then
have
to
include
such
event,
service
component
or
module
into
our
autonomic
and
at
work
infrastructure.
N
In
that
case,
we
may
have
to
you
know,
include
define
in
a
common
way
that
we
should
have
such
a
module,
and
we
somehow
need
to
extend
that
as
brian
saying
that
we
may
somehow
associate
that
part
with
the
grass
api.
Is
that
so
that
it,
the
grass,
can
also
use
that
part
for
a
synchronous
communication.
N
So
here's
just
a
very
scratchy
view
that
what
if
we,
we
include
the
event
service
and
into
the
current
autonomic
networking
infrastructure.
So
here
we
added
the
and
yellow
box
over
there
and
at
the
api's
and
on
top
of
the
grass
and
and
risky
so
that
the
upper
layer,
autonomic
functions
and
asa
scan
or
share
that
reusable
components
for
all,
for
example,
as
being
said
for
information
distribution,
but
in
a
very
a
synchronous
way.
So
basically
that
module
has
four
jobs.
First
of
all,
listening
to
the
network
or
from
the
ACB
from
the
asp
domain.
N
What
can
they
even
published
in
a
second?
It
will
notify
the
other
animal
note.
For
example,
autonomic
functions
with
a
certain
event,
and
certainly
it
will
handle
the
local
even
publication
to
the
network.
The
the
fourth
one
is
very
important:
it
not
only
in
just
responsible
for
for
notification
and
and
subscription,
but
also
responsible
for
storing
dot
information
across
the
network.
So
two
major
jobs,
even
so
this
here
I
put
the
even
queues
over
there
and
most
of
the
information
storage.
N
Here's
a
kind
of
those
component
will
be
impact
by
our
current
oppose
about
a
bomb
put
too
much
time
on
that.
Okay,
we
actually
didn't
put
the
the
page
that
raised
the
question.
Should
this:
should
this
working
group
consider
such
a
topic,
but
this
is
what
we
would
like
to
ask
everybody
here
and
yes,.
A
F
B
Which
everybody
needs
right,
so
the
first
piece
of
the
cake-
and
we
currently
don't
have
the
scheme
whereby
you
know
the
network
gets
partition.
New
parts
come
on
and
it's
there,
but
in
the
end
we
still
want
to
have
you
know
the
flooding,
but
also
keeping
it
persistent
right.
So
maybe
that's
one
simple
thing
to
say
that
that
can
come
with
a
simpler
solution,
either
grass
extensions
simple.
You
know
grasp
ASAE
that
does
the
reef
flooding
or
so
and
then
the
other
much
cooler,
much
bigger
issue.
B
Is
that
you're
starting
to
have
a
lot
of
interesting,
a
si
datasets
that
not
everybody
in
the
network
needs
for
which
you
need
to
set
up
things
that
well
in
my
prior
life.
I
would
have
called
multicast
distribution,
trees,
right
and
that
would
be
cool,
but
I
think
we
need
to
find
a
lot
more
in
our
support
for
doing
this
right.
Yes,.
B
Terms
of
there
are
these,
you
know
there's,
unfortunately,
Alex
ran
out
so,
but
there
folks
been
working
on
distributed
data
store
also
with
yang
in
the
ITF
and
those
might
be.
You
know,
a
Jason
sees
that
we
could
work
with
to
establish
an
infrastructure
to
support
those
things
to
create
the
use
cases,
because
I
think
at
this
point
in
time,
obviously,
that
that
bigger
scope
would
be
much
less
arguable
to
be
in
charger.
N
Okay,
okay
I
see,
but
but
here's
what
we
see
is
for
the
even
service
is
kind
of
glued
for
four
other
components
you
mentioned
before,
for
example,
distributed
databases
and
some
kind
of
a
Zen
koan
synchronous
communication
requirements.
This
even
serves
glue
them
together.
This
is
what
we
are
saying,
but.
B
Just
not
having
a
you
know:
we
don't
have
a
structure
in
the
ITF
to
create
more
shareable
infrastructures
right.
If
you
look
at
how
flooding
has
been
reinvented
by
every
bloody
distributed
routing
protocol
right
I
mean
that's
insane,
but
that's
basically
the
ITF
right.
If
you
want
to
make
any
progress
and
change
that
it's
going
to
be
difficult
right.
That's
token,.
N
N
N
G
N
D
F
F
A
N
I
At
Wagga,
one
of
the
courses
so
also
answering
partly
to
journalists
remark.
So
if
you
remember
what
we,
when
we
introduced
first,
this
pops
up
as
a
need
so
to
say
in
Prague.
Actually
it
was
done
in
the
round
regarding
the
possible
reach
shattering
of
a
NEMA,
so
I
kind
of
agree
with
what
Tyler
said
before
that
in
the
general
sense
it
might
go
beyond
the
current
charter.
I
Then,
however,
later
we
have
seen
also
that
the
information
distribution
is
very
much
related
to
all
this,
and
indeed
we
can
limit
ourselves
to
the
information
distribution
in
the
sense
that
it's
communication
service
right.
So
we
used
rasp
as
an
example,
concrete
example:
it
exists
and
it's
chartered
and
so
on,
but
again
information
distribution
is
not
only
in
space.
It's
also
in
time,
and
that's
why
this
storage
somehow
becomes
necessary.
There
is
information,
it
must
be
postponed.
We
know
it
from
that
autonomous
networking.
I
By
the
way
we
call
it
ourselves
as
a
normal
networking,
and
to
that
you
know,
belongs
also
the
storage.
We
cannot
just
ignore
it,
so
how
we
integrate
it
in
which
point
and
for
which
exact
job
I
do
not
know
and
by
the
way,
I.
Don't
think.
It's
very
important
that
this
is
just
a
proposal
and
we
can
of
course
go
for
the
proposals,
but
somehow
we
cannot
ignore
the
storage,
whether
it's
a
distributed
storage
that
has
been
used
in
other
groups
or
somehow
else,
but
I
believe
it
will
be
inconsistent
and
incomplete.
I
If
you
only
provide
the
communication
service
as
not
a
normal
service
or
to
say
because
otherwise
it's
you
know,
we
push
all
this
complexity
to
the
developers
and
I
kind
of
agree
with
what
has
been
discussed
about
before
that.
We
also
need
to
consider
developers
gives
them
some
guidance,
and
you
know,
come
up
with
some
concrete
use
cases
here
and
not
just
talk
about
some
infrastructure
without
any
API.
Without
any
support,
it
will
be
unclear.
I
N
In
principle,
or
or
in
principle,
our
motivation
is
just
like
that
when
we
are
an
emerge
and
aims
to
build
some
kind
of
autonomic
networking
infrastructure
for
the
autonomic
functions
of
the
network,
so
what
we
see
here
is
that,
in
addition
to
the
direct
communication
model,
we
also
see
that
autonomic
functions
may
need
a
synchronous
communication
model
which
we
hope
it
can
be
supported
by
the
autonomic
networking
infrastructure
in
a
very
native
way.
This
is
just
our
simple
motivation
here.
B
Slide:
okay,
so
this
this
basically
is
a
proposal
for
for
work.
That
I
would
like
to
see
going
standard
strike
as
part
of
the
existing
charter
items,
so
an
additional
milestone
that
kind
of
crosses
over
between
a
CPA
and
I
and
primarily
grasp,
and
so
this
goes
back
a
little
bit
to
the
roots
of
you
know
what
I
wanted
to
see.
Grass
do,
and
so
you
know
we
had
discussion
right.
Why
do
we
not
use
DNS
SD
for
service
discovery
and
especially
in
brewski,
we
kind
of
discuss
and
argue
that
long
and
hard?
B
And
it's
still
in
the
brewski
draft
for
use
outside
of
the
ACP
and
of
course
we
concluded
on
unrest
inside
the
ACP
and,
ultimately
to
me
it's
just
you
know
a
new
transport
for
service
discovery.
Granted
grasp
does
a
lot
more,
but
it
is
also
a
transport
for
service
discovery.
But
then
we
really
didn't
finish.
B
All
these
type
of
services,
for
which
you
know
service
name,
definitions,
already
exist
in
DNS
SD
that
we
would
like
to
use
across
the
ACP
as
well.
So
what
is
then
missing
in
the
existing
grass
definitions
for
DNS
SD
type
services
right
so
right
now
we
haven't
defined
anything.
How
grasp
could
use
existing
IANA
service
names,
so
we
would
have
to
come
up
with
our
own
service
names
and
if
you
track
the
mailing
list
for
the
last
few
years,
we've
always
kind
of
let
let
us
use
this
name.
B
This
name
this
name
so
kind
of
duplicating
the
effort.
I
think
we
shouldn't
need
to
reinvent
service
names
right.
We
also
have
no
definitions
for
how
multiple
grasp
objectives
could
share
the
common
attributes
like,
for
example,
the
attributes
we
have
in
DNS
SD
priority
wait
any
other
type
of
other
key
value
pairs
that
we
have
in
DNS
SD
and
then
also
how
to
do
distance
based
server
service
selection.
So,
for
example,
registrar's.
B
It's
not
possible
without
you
know
a
specification,
so
we
try
to
put
a
subset
of
these
things
into
a
CP,
&
brewski
and
then
basically
at
least
the
brewski
author
said
well,
can
you
just
go
out-
and
you
know
try
to
do
this
more
uniquely
across
the
different
service
that
we
have
in
a
separate
document
and
even
for
a
CP,
growing,
large
and
larger
I
felt
it
would
be
ideal
to
really
have
this
in
a
separate
drop.
So
that's
a
history.
B
Let's
try
to
look
at
what
we're
doing
proposing
to
do
here
next
slide,
so
separate
document
write,
remove
all
the
complex
service
discovery,
details
from
ACP
and
brewski
and
then
ultimately,
this
document
would
meant
to
update
Bruschi
ACP
in
a
backward
compatible
way,
meaning
that
what
we've
defined
as
the
grass
objectives
for
the
service
discovery
in
brewski
an
ACP
would
be
amended
with
these.
You
know
all
optional
things
to
basically
introduce
the
policies
that
we
need
and
then,
of
course
also
new
services
across
the
ACP
using
grasp
would
be
able
to
use
those.
B
That's
already
where
you
know,
service
names
like
EST
and
brewski
exists,
syslog
and
all
the
others,
and
then,
of
course,
the
registration
procedures
would
basically
say
well.
We
would
welcome
you
to
still
register
these
in
the
grasp
registries
for
objectives
if
you're
doing
something
new,
but
if
you're
just
building
a
proxy
from
dns
sd4
mdns
into
acp,
for
example.
Then
of
course
you
don't
even
know
what
service
names
are
there.
B
So
it
would
be
hard
to
register
them
all
in
grasp,
then
the
payload,
what
I'm
proposing
here
in
the
in
the
draft
is
not
to
extend
grasp
headers
for
anything
we're
doing,
but
instead
come
up
with
the
trick
on
how
we
have
a
common
definition
of
these
things
in
all
different
objectives,
in
a
way
that
it's
visible
cross
objectives.
That's
a
little
bit
of
a
heck
and
I'd
love
feedback
about
that
I'm,
not
particular
to
how
we
do
it.
But
we
need
to
do
it
some
way,
and
you
know
the
proposal.
B
So
here
we
have
a
new
element
that
says
here
was
the
original
hop
count
that
the
sender
said
when
he
sent
it
so
now,
I
see
in
my
grasp
header.
It's
been
reduced,
15
hops,
so
this
announces
15
hops
away
from
me
and
another
announcer
is
30
hops
away
from
me.
Err
I'm
going
to
maybe
you
know,
for
whatever
I
want
to
do
pick
the
closest
one
or
you
know,
maybe
the
furthest.
One
next
slide,
all
right,
so
DNS
SD
in
half
a
slide.
So
in
DNS
SD,
basically
there
is
an
intriguing
encoding
of
services.
B
So
let
me
quickly
show
on
the
beginning,
you
see
something
like
you
have
a
service
and
that's
basically
pointing
to
a
bunch
of
service
instances
and
service.
Instances
are
names
that
are
primarily
meant
for
human
consumption
and
selection
right
so,
for
example,
you're
looking
at
the
printer
service
you're
getting
back
and
that's
called
browsing
a
list
of
service
instance
names
that
are
nicely
human,
readable
names
like
my
printer,
your
printer
and
their
printer,
and
obviously
by
looking
at
you
know
your
user
interface.
B
You
know
which
printer
you're
selecting
so
now,
if
I'm,
basically
not
autonomic
service
agents,
I
wonder
how
important
this
form
of
selection
of
one
out
of
multiple
service
instance
is
going
to
be,
as
opposed
to
a
selection
based
on
more
formal
parameters
of
a
service
instance,
and
these
more
formal
aspects
of
a
service
instance
are
embodied
in
DNS
SD
in
two
parts.
B
For
once,
there
is
the
serve
record,
which
has
the
priority
and
wait,
which
are
simple
numerical
things,
to
select
a
service
instance,
and
then
there
are
optional
service,
specific
parameters,
and
you
know
if
you're
trying
to
use
a
service,
maybe
you
want
to
look
up
at
the
service
parameter
like
I,
don't
care
if
it's
euro,
my
printer
right
as
long
as
I
can
see
it.
Hopefully,
I
can
pick
up
the
print,
but
I'd
love
a
color
printer
I'd
love
a
printer
with
a3
and
all
those
things
would
be
in
these
key
value
pairs
right.
B
You
see
that
in
the
surf
record
there
is
also
the
host
name,
so
the
service
resolution
from
the
service
name
over
the
service
instance
name
goes
into
a
host
name
and
only
that
host
name
goes
to
the
actual
locators
off.
You
know
the
transport
ports
on
which
you
can
consume
the
service
and
these
host
names.
Well,
you
know,
ACP,
we
don't
have
you
know,
amend
a
mandate
to
support
DNS
or
host
names
at
all.
B
So
host
names
in
our
case
would
also
be
something
that,
by
default,
we
wouldn't
use
I,
also
defined
a
way
to
introduce
host
name
resolution
through
grasp,
as
part
of
this
I've
been
in
the
last
slide,
but
I
would
also
think
in
general
is
optional.
So
if
you
think
of
it,
you
know
grasp
service
resolution
in
many
cases,
because
we're
using
it
from
automated
entities
ASAS
in
the
network,
could
be
a
lot
simpler
than
in
many
cases,
DNS
SD.
That
was
meant
for
human
consumption
next
slide.
B
So
here
is
the
wonderful
sea
bore
CD,
DL
formatting
of
a
service
record
which
in
grasp
we
would
have
as
part
of
the
objective
value.
So
this
is
basically
the
element
you're
inserting
and
pretty
much
everything
is
optional
there.
So,
first
of
all
private
parameters,
which
is
basically
whenever
we
want
to
extend
something-
and
it's
not
standardized
anymore,
then
we've
basically
got
the
message
type
and
that's
basically,
where
we're
saying:
okay,
we're
basically
going
to
describe
a
service
or
we're
requesting
a
description
of
the
service,
or
we
want
to
rated
or
request
enumeration.
B
Those
are
pretty
much
the
four
simple
operations
that
I
think
we're
going
to
get
away
with.
Then
the
service
right,
that's
a
service
name,
it's
duplicated
here
it
would
also
be
in
the
objective
name.
It
needs
to
be
in
the
objective
name,
because
that's
where
you
know
grasp
needs
to
have
something
that
you're
announcing
or
that
you're
looking
for
in
M
discovery
when
you're
looking
for
a
service
right.
So
it's
duplicated
in
here,
primarily
for
the
point.
B
When
you
have
an
announcement
of
potentially
multiple
services,
then
the
instance
name
right,
so
my
kitchen,
printer
optional,
then
I
also
edit
the
domain
so
functionally.
You
know
service
discovery
about
across
grasp.
One
simple
way
to
understand
it:
it's
like
humongous
large
network
version
of
mdns,
like
it's
flooded
in
the
network,
so
by
default
the
domain
would
be
dot
local
and
that's
kind
of
the
grasp
instance
you're
using
like
the
ACP,
but
it
could
be
other
names
as
well.
In
that
case,
we
need
to
add
more
functionality
on
how
to
do
that.
Priority.
B
Wait
as
in
DNS
SD,
key
value
pairs
set
of
key
value
pairs,
exactly
like
DNS
SD,
then
I
added
a
new
parameter
called
range,
which
is
basically
waiting
the
service
selection
between
priority
and
wait
versus
the
distance
right.
So,
whether
you're
more
interested
in
distance
based
service
selection
or
more
priority
weight,
I'm
not
going
to
control
what
the
algorithm
is
read
up
on
the
on
the
slide,
so,
hopefully
a
flexible
way
to
allow
you
to
define
in
your
service
definition
whether
it
should
be
priority
weight,
end
or
a
distance
based
selection
and
then
see
locator.
B
So
grasp
has
locators
and
the
see
locators
are
simply
extending
locators
with
one
other
element,
which
is
a
bloody
small
string
telling
you
in
which
context
this
locator
lives,
because
in
the
classical
Ani
environment
we
already
have
two
context.
One
is
the
ACP
that
we
usually
assume,
but
for
many
ASAS
it
would
be
very
valuable
to
also
have
locators
across
the
data
plane,
because
the
data
plane
may
have
higher
performance
or
you
may,
for
other
reasons,
in
the
Asaf
knowledge
about
the
data
plane,
locators
of
a
particular
service
instance,
not
only
the
ACP.
B
So
that's
basically
now
context
locator
so
that
both
are
possible
and
yeah.
That's,
basically
the
service
structure,
and
that
allows
us
to
express
everything
that
we
can
express.
Well,
they're,
small.
You
know
not
that
often
use
details
in
DNS
SD
that
we
haven't
covered
here,
but,
let's
say
90%
of
what
DNS
SD
can
do.
We
can
express
here
also
in
the
by
direction
way
you
could
build
a
gateway.
You
know
converting
one
into
the
other
next
slide.
So
these
are
the
message.
Exchanges
right,
the
most
simple
one
is
the
grasp.
Em
flood
right.
B
That's
an
unsolicited
announcement
of
objectives,
service
instances,
message
type
is
described
right,
you're,
describing
a
service
if
you're
doing
a
unicast
is
already
the
other
flooding
options.
M
discovery,
you
find
an
objective,
that's
a
describe
request.
It
defines
that
most
of
the
the
attributes
in
the
describe
requests
are
invalid.
The
only
thing
of
interest
are
the
service
may
be
service
instance
name
that
you're
looking
for,
and
it's
actually
very
interesting
paradigm
for
service
discovery,
because
it
does.
The
discovery
stops
at.
The
first
instance
is
found
right.
B
So
you
can't
have
all
the
policies
in
terms
of
priority.
Wait
in
this
discovery
in
grasp,
but
in
a
large
network
it's
a
lot
more
efficient
because
you're
not
flooding
something
right.
If
you're,
basically
having
a
lot
of
distributed
service.
Instances
and
the
policy
is
always
don't
bother
the
whole
bloody
network,
with
the
fact
that
you
are
there
and
offering
a
service
which
is
really
meant
for
your
vicinity.
Mdm
discovery
very
efficiently
supports
that
right.
B
So
in
that
respect,
of
course,
grasp
and
the
schemes
that
we
have
is
a
lot
better
than
simple,
let's
say:
mdns
based
flooding
alone.
That
would
be
M
discovery.
Then,
of
course,
the
same
type
of
things
can
be
done
equally
through
unicast
negotiations
when
you're
just
talking
about
service
and
negotiation
discovery
announcement
between
instances
that
have
already
found
themselves,
like
oh
I,
know
it
here
is
a
big.
You
know,
Gateway
for
all
type
of
service
to
the
knock
and
then
enumerate
enumerate
equal
requests.
B
Name
you're
interested
in
into
its
parameters,
including
locators
right
and
the
backward
compatibility,
is
simply
by
the
fact
that
the
way
that
the
service
element
is
defined
means
that,
if
you
don't
have
the
service
element
but
you're
just
having
what
is
currently
in
the
ACP
Draft
and
what
I
would
also
like
to
see
in
the
brewski
Draft,
which
is
just
you
know,
it's
a
service
served
or
brewski
served
on
est.
No
other
parameters.
It
simply
means
that
is
a
describe.
The
locator
to
use
is
from
the
grasp,
header
and
you're
done
with
it
right.
B
So
that's
the
full
backward
compatibility
as
soon
as
you
add
the
service
element
to
the
objective
value,
you're,
basically
supposed
to
look
into
all
these.
You
know
elements,
priority
weight
and
so
on,
so
that
that's
kind
of
the
gist
of
the
proposal
next
slide
yeah.
So
here's
the
encoding
on
how
to
put
the
service
element
in
a
way
into
thousands
of
different
objectives
in
a
way
that
it
can
be
easily
detected.
B
And
if
you
don't
care
about
the
stuff,
you
don't
need
to
have
to
do
it
and
we
don't
need
to
come
up
with
new
grass
patter
extensions
for
quote
Ayana
registered
common
objective
value
elements
which
this
effectively
is
so
basically
what
you're
doing
is
the
objective
value
needs
to
be
met.
A
map
is
what
is
easily
extensible,
which
means
it's
a
set
of
key
value
pairs
and
we're
simply
going
to
define
one
P
which
is
going
to
be.
B
You
know
a
magic
string
at
our
FC,
something
something
which
would
be
the
number
of
this
thing,
and
then
inside
of
that
we
again
have
a
map
which
is
a
set
of
key
value
strings,
and
each
of
the
keys
would
have
to
be
an
ini,
registered
reused
service
element
and
the
one
the
two
ones
that
were
defining
here
in
this
draft
is
the
sender
loop
count,
which
is
basically
what
you
know.
The
sender
announces
in
the
EM
flood
and
then
the
other
one
is
the
service
element
which
I
was
just
describing.
B
So
those
are
the
two
and
then,
if
in
the
future,
somebody
comes
and
says
hey
across
different
grass
objectives.
There
is
this
one
function,
this
one.
You
know
data
structure
that
I
want
to
reuse.
Well,
you
know
get
yourself
an
iron
allocation
put
it
in
there.
So
that's,
basically
the
the
simple
encoding
that
gets
us
rid
of
having
to
extend
grasp
headers
with
some
standard
payload
things
which
we
didn't
do
and
I
think
this
should
hopefully
work
equally
well
next
slide
yeah.
So
this
was
basically
just
the
last
option.
B
B
Yes,
I'll
also
try
to
give
a
variant
of
this
presentation
on
Wednesday.
The
DNS
SD
working
group
asked
me
to
do
that,
but
I
probably
need
to
focus
less
on
these
wonderful
details
of
the
encoding,
but
rather
more
on
what
the
heck
is
anime,
and
why
is
it
doing
something
new
right
so
will
be
different.
I'll
have
to
put
this
slide
together.
First.
A
Believe
all
the
plantations
of
today,
so
thanks
for
all
those
fronts,
front
enters
to
tape.
The
perfect
timings
we
have
two
minutes
left,
so
we
have
done
all
those
plantations
and
actually
still
continue
to
calling
new
work
item.
New
proposals,
but
before
you
submit,
is
a
new.
Please
check
the
current
agenda,
although
it's
extendable,
but
we
would
like
to
give
any
priority
to
the
content
in
the
development
at
work
items.
So,
thank
you
all
see
you
all
in
London.