►
From YouTube: IETF104-DTN-20190326-1610
Description
DTN meeting session at IETF104
2019/03/26 1610
https://datatracker.ietf.org/meeting/104/proceedings/
A
B
Do
sheets
are
going
through
the
room
as
usual,
we're
under
note
well,
which
says.
Essentially
you
know
everything
you
say
or
you
write
is
under
is
covered
under
those
rules.
Please
read
when
you
have
time
proposed
agenda
that
my
colleague
eric
has
put
out
the
call
for
proposal
I
guess
we
ought
to
get
some
convergence
layer
done.
B
C
D
B
E
B
Yeah,
we've
done
less
calls
I,
don't
I,
don't
know
if
you
remember
that
we
did,
but
we
did
and
well
conversions
layer,
TCP
conversion
layer
being
discussed.
Hopefully
the
last
time
this
week
will
progress
through
sending
to
the
ihe
for
our
ad
for
his
new
work.
You
know
just
as
an
introduction
to
his
new
tenure.
B
F
Okay,
I'm
Scott
Burleigh
I
work
at
Jet,
Propulsion
lab
in
Pasadena
California.
We
talking
about
minimal
TCP
convergence
layer,
adapter
I,
can
be
very
brief.
I've
got
exactly
one
slide.
Let
me
just
go
ahead
to
that
slide.
The
intent
of
MTC
PCL
is
to
provide
a
convergence,
their
adapter.
That
is
simple
enough,
that
we
can
agree
on
it
readily
and
provide
it
along
with
the
BP
sex
specification
and
the
BP
disc
specification
as
a
as
a
package,
so
that
we
can
start
moving
everything
together
onto
the
RFC
editors
queue.
F
To
that
end,
the
M
TCP
Convergys
layer,
adapter,
is
very
short.
It's
about
eight
pages.
You
can
review
it
in
about
ten
minutes.
I
think
the
we
talked
about
it
a
little
bit
after
the
last
IETF.
Since
then,
the
only
changes
are,
first
of
all
changing
the
the
name
from
simple
TCP
convergence
layer
to
minimal
TCP
convergence
there
that
seemed
to
be
more
acceptable
to
everybody.
F
It
was
and-
and
it
and
the
the
working
group
also
adopted
it.
So
the
name
now
is
has
got
to
she
can
working
group
in
it.
I
think,
instead
of
me
and
the
second
and
really
the
only
really
significant
change
was
adding
a
language
about
establishing
TLS
sessions.
In
the
same
way
that
TLS
needs
to
be
supported
in
the
TCP
convergence
layer
and
and
in
fact,
the
language
that
I
added
there
is
lifted
all
not
exactly
verbatim
but
we're
pretty
close.
F
That
had
already
been
compiled
and
I'm,
assuming
that's
all
gonna,
be
workable,
unacceptable,
TLS
language.
The
only
other
has
changed
that
I
made
in
the
document
as
I
was
going
through.
It
was
there's
one
point
where
describing
the
structure
of
the
EM
TCP
protocol
data
units,
the
the
something
had
come
up
in
another
context,
that
there
was
possible
ambiguity
in
in
saying
that
the
encapsulated
that
the
encapsulated
bundle
that
was
being
shipped
in
the
MPD
you
was
a
Seaborg
byte
string.
F
We
wanted
to
make
it
clear
that
it
was
a
simple
byte
string
of
defined
length
rather
than
an
undefined
indefinite
length,
sepal
string.
Those
are
the
only
changes.
The
revised,
updated
spec
was
posted
as
an
internet
draft
on
the
28th
of
February
and
at
this
point,
I
leave
it
to
the
working
group
as
to
whether
or
not
we
can
go
ahead
with
it.
C
H
H
I
was
looking
into
the
MVC
l
and
also
the
the
bundle
protocol,
v7
and
I
ran
into
just
a
rough
question
for
the
working
group.
As
far
as
the
policy
of
seaboard
tagging,
I
noticed
that
tags
are
mentioned
in
MVP
v7
and
not
really
in
the
m
TT
PCL
is
there?
Is
there
a
policy
as
far
as
like
tags,
just
shouldn't
be
present
at
all
in
DC
board
encodings?
H
F
I'm
not
sure
I've
answered
for
that.
The
intent
the
VP
b7
respect
in
the
M
TCP
spec
was
to
state
all
of
the
C
for
representational
elements
that
were
required,
and
else
so
there's
there's
I,
think
no
intent
to
say
that
you
can't
do
anything
else
and,
for
example,
a
an
extension
block,
but
I
I
think
both
specifications
are
silent
and
I
will
I
will
declare
intentionally
silent
on
that
question
unless
there
needs
to
be
something
said
about
it:
a
scope
that
Rick.
E
That's
being
transported,
whether
a
tag
is
there
or
not,
doesn't
manifestly
change
the
meaning
of
what
you
are
parsing
sure
if
you're
doing
a
wire,
dissector
or
trying
to
do
this
in
in
Fiji
a
then
it's
an
issue
but
I
think
if
you
want
to
tag
that
this
string
is
a
date,
that's
useful,
perhaps
for
your
domain,
so
I'm
not
sure
where
I'm
going
here
but
I'm
I'm,
not
sure
the
presence
of
tags
needs
to
be
specified.
Canonically
within
the
specification
and
I.
E
Think
the
c-more
spec
says
you
should
what
you
do
with
the
tags
is
your
business.
So
if
you
just
throw
them
away,
because
it's
irrelevant
to
your
team,
to
what
you're
doing
that's
fine
but
I
think
to
be
a
compliant
Seaborg
consumer,
you
must
handle
tags,
but
we'll
have
to
get
the
actual
sequence
back
to
determine
that
must
or
should
that.
F
I
G
I
I
E
E
I
F
I'm,
it
would
have
been
great
to
have
this
question
on
the
mailing
list
before
we
got
here.
Yeah
I
would
have
to
research
in
in
seaboard
to
give
you
a
better
answer,
but
I
will
say
that
I'm
perfectly
fine
with
the
removing
the
length
field
from
the
protocol
data
unit.
That's
if
it's,
if
it's
not
necessary
great.
That's
the
Rick.
E
With
chair
hat
on
sorry
to
a
drunk
you,
this
is
a
classic
mailing
list
conversation
and
can
I
recommend
we
do
this
properly
on
the
mailing
list.
So
there's
a
record
of
why
we
came
to
this
decision
before
we
get
asked
again
in
three
months
time,
probably
by
the
same
people
again,
because
we've
forgotten
yeah
so
definitely
take
this
onto
the
list
of
these
guys.
E
B
F
K
F
F
F
F
E
L
Sorry
I
mean
if
you
actually
want
to
request
a
port
and,
yes,
you
need
to
have
an
a
section
on
it:
I
I'm,
not
we
try
to
keep
down
the
port
usage
if
it's
really
I
mean,
and
we
shouldn't
be
spending
a
lot
of
ports
if
it
tells
necessarily
I,
don't
understand
enough
about
this
case
to
give
a
recommendation
from
standing
here.
But
do
you
have
other
methods
of
taking
route
or
is
it
local
configuration
for
which
convergence
that
you're
using
etc?
F
L
L
E
Responding
to
what
we've
just
heard,
I
I,
personally,
not
chair
hat
on
personal
hats
on
I'd-
be
perfectly
happy
to
see
some
text
in
the
document
that
says
we
don't
need
to
reserve
a
port
for
this.
It's
it's
a
minimal,
tcp,
CL
pre,
configure
it
in
some
way.
Out
of
that,
and
as
long
as
that
text
is
in
there,
then
fine
that's
proceed
there.
F
F
Would
I
guess
I
would
advocate
against
trying
amongst
the
two
protocols
on
the
same
port,
I,
don't
think,
there's
a
I,
don't
see
a
significant
operational
advantage
to
doing
that,
and
and
if
you're
not
going
to
do
that,
then
I
think
it's
unnecessary
to
require
implementations
to
be
able
to
distinguish
between
the
two
are
given
that
the
first
few
bytes
are
different.
How
much
do
you
have
to.
E
Do
yeah,
but
the
first
few
bytes
are
a
length
and
then
some
payload,
so
I
could
craft
something
that
you
know
looks
just
like
TCP
CL
header,
but
is
actually
a
specially
lengthened
packet.
I.
Think
that's
a
can
of
worms
personally
I,
wonder
about
sure
you
can
write
an
implementation
which
does
both
and
that's
great
but
I.
Think
it's
a.
M
M
F
E
Rick
jumping
in
again
what
I
could
hear
there
is,
what
yoga
is
proposing
is
is
perhaps
is
this
specification
and
I'm
gonna
need
to
check
the
TCP
CL
specification
is?
Is
there
a
unique
enough
header
or
preamble
on
the
connection
establishment
such
that
the
I
know,
there's
a
there's,
a
version
number
in
TCP
CL?
Then
we.
F
E
To
put
a
preamble
at
em,
TCP
and
ASCII,
or
something
on
the
front
end,
so
we
can
say:
Oh
someone's
pointed
their
web
browser
at
me
by
mistake
or
I'm.
Looking
at
the
room
for
feedback
on
this,
because
your
look
makes
a
point,
everyone
miss
configures
everything,
but
is
it
then
their
problem
rather
than
the
protocol
specs
problem,
III?
Think.
N
F
N
F
D
Ed
brain
just
a
comment
which
was
I
think
the
value
of
having
both
of
them
is
that
the
minimum
TCP
CL
stays
minimum
and
so
I
think.
While
there
are
many
features
that
could
go
into
it,
having
something
that
is
really
only
that
which
is
necessary
to
show
what
we
need
to
for
bundle
protocol
is
the
rubric
for
determining
what
goes
into
M
TCP
I.
B
B
H
Okay,
so
I
have
a
very
few
number
of
slides
and
the
summary
version
is
that
this
is
very
close,
so
the
first
couple
of
slides
can
be
skipped
for
the
next.
No
changes
have
been
made
to
the
background.
This
is
just
for
a
copy
of
some
of
the
earlier
presentation
point
is
to
update
what
already
exists
and
to
make
what
already
exists
in
PC,
PC,
lv3,
simpler
and
more
robust,
where
it
was
possible
and
add
extensions
for
the
benefit
of
the
protocol.
Can
you
get
in
the
next
slide?
Please?
H
So
still
the
goals
are
to
not
change
the
scope
for
the
workflow.
The
the
CL
works
in
before
much
the
same
way
as
it
did
in
v3.
There's
still
the
same
phases
of
negotiation,
there's
still
the
same
message,
headers
and
message
type
codes,
and
so
ideally
this
would
make
it
easy
for
an
implementation
to
be
adapted
to
move
to
v4
or
to
be
cross
compliant
with
v3
and
v4
next
slide,
please.
So.
H
The
last
edits
in
the
draft
based
on
feedback
from
George
and
the
other
IETF
first,
was
to
completely
remove
the
transfer
in
a
message
which
originally
was
present.
Only
because
of
compatibility
issues
in
TCP,
CL,
v3
and
I'll
talk
a
little
bit
about
this
and
a
couple
of
later
slides
about
PDU
sighs.
H
So
the
extension
items
which
were
present
in
the
transfer
unit
message
have
just
been
moved
into
the
first
transfer
segment
message,
which
is
identified
by
a
start
date.
So
this
simplifies
the
state
machine
if
reduces
some
PDU
overhead
and
I'll
talk
about
the
transfer
total
length.
But
this
is
is
also
moved
into
an
extension
item
for
a
simplification
of
the
sequencing
and
reduction
in
the
message
size.
The
total
extension
item
length
has
been
reduced
from
64
bits
32
bit.
H
H
The
default
and
minimum
session
timeout
was
clarified
to
be
compatible
with
the
recommended
values
from
the
CL
v3.
So
there's
no
change
from
v3
here
and
then
reply
bit
was
added
to
the
session
termination,
but
just
to
avoid
a
feedback
loop,
so
the
whichever
entity,
whichever
peer
of
the
session,
sends
the
first
session
termination.
The
reply
bit
has
not
said
when
the
the
reply
message
is
sent.
H
The
reply
bit
is
set
so
there's
a
clear
sequencing
of
the
session
termination
and
in
a
valid
session
there's
only
ever
going
to
be
one
such
and
termination
message
coming
out
of
each
peer.
One
of
them
is
going
to
be
initiation
and
one
of
them
is
going
to
be
the
reply
and
that's
it
and
then
the
session
termination
encoding
was
simplified
to
avoid
variability
of
the
message
content
now,
instead
of
the
type
code
being
present
or
not
present,
that
the
code
is
always
present
and
there's
just
an
unknown
type
code.
H
So
the
took
the
transfer
total
length
was
moved
into
an
expansion
item
and
the
reason
for
this
is
for
a
couple
of
things.
One
is
to
allow
that
minimum
number
to
be
a
bare
minimum.
If
you
do
want
to
do
single
segment
transfers,
then
the
transfer
that
the
total
length
becomes
redundant
because
it's
all
in
the
one
segment
and
segment
already
has
a
length
and
also
moving
this
moving.
This
out
of
that
separate,
transferring
it
message
into
an
extension
item.
H
Although
there
is
some
overhead
with
the
extension
item,
the
overhead
of
the
extension
item
is
actually
two
out.
That's
smaller
than
the
overhead
of
transferring
it
message.
So
there's
no
overall
penalty
for
doing
what's
been
done
here
and
this.
This
simplifies
the
sequencing
and
it
actually
brings
the
TCP
seal
or
into
more
kind
of
logical
line
without
b3,
with
structure
and
even
when
the
transfer.
H
H
The
example
agent
that
was
done
in
Python
has
been
updated
for
the
draft
revision,
11
Lester,
sequencing
and
there's
some
other
embellishments
have
been
made
just
to
show
what
can
be
done
with
the
v4
protocol.
It
now
does
graceful
session
termination,
and
this
was
something
that,
as
a
proof
of
concept
as
a
kind
of
a
justification
for
some
of
these
behaviors
I
wanted
to
do
this,
because
there
have
been
four
questions
coming
in
about
the
initialization
and
in
termination
sequencing.
H
So
this
just
gives
some
some
concrete
evidence
that,
yes,
this
is
a
workable
behavior
and
it
does
provide
actual
value
to
the
protocol
and
also
is
a
proof
of
concept,
because
it
was
something
that
is
discussed,
but
it
wasn't
really
looked
at
I.
Don't
think
at
all.
Is
the
agent
actually
has
a
feedback
algorithm
to
have
adjustable
segment
size
to
hit?
In
this
case,
the
target
metric
is
time
to
acknowledge
that
the
difference
between
the
time
when
the
segment
is
being
encoded
and
when
the
acknowledgment
comes
back
for
that
segment
and
operation.
H
I
think
this
is
the
kind
of
thing
of
it.
An
agent
really
would
look.
We
would
be
looking
for
want
to
work
with
it.
You
want
to
guarantee,
via
your
urea
unacknowledged
akkad
data
is
gonna,
be
no
more
than
X
number
of
seconds
old
on
a
whatever
your
screen.
You
never
give
and
I
also
implemented
just
this
minimum.
The
programmatic
generator
to
exercise
text
refers
to
demo
agent
X
by
please
Vikram.
H
One
other
thing
that
was
done
again
these
days
justification
I,
for
they
were
to
call
itself
points
up,
I
put
together
of
Wireshark
technically
and
I.
Really,
your
exercises
I
think
all
in
the
behavior
before
convergence
layer
it
supersedes
the
stock
Wireshark
I
expected.
That
was
what's
called
bundle
and
it
was
the
TCP
Co
b3
sohave
years,
but
this
handles
all
the
messaging
your
handles
actually
Kiowas
yep
within
sessions
and
reassembling
of
individual
transfer
segments
into
the
total
transfer
data,
so
I
think
interoperability
and,
for
example,
of
a
pack
working
code.
H
H
It
was
just
a
picture
of
some
of
the
dissected
CLV
for
messages.
I
mean
nothing
should
stand
out
it
yet
I
sector
like
wires
from
before,
but
again
just
matter.
My
justification
for
is
a
realizable
thing,
and
is
it
a
reasonable
thing
to
be
organized
array?
This
and
I
mean
nothing
image
here.
Nothing
in
the
corridors
are
actually
too
complex
in
the
cutting
up.
H
There
was
a
comment
about
reducing
the
individual
item
extension
length.
Basically,
this
is
it
worth
what
it
is
that
question,
but
to
the
working
group.
Is
it
worth
saving
two
octets
in
an
optional
PD
you
to
produce
the
besides
the
extension
items,
I
I,
don't
see
any
strong,
real,
strong,
locking
reason
for
or
against
other
than
reducing
the
extension
items
by
two
octets.
H
If
a
if
an
individual
extension
needed
to
encode
data
longer
than
the
16-bit
size
would
allow
it
to
do,
then
the
extension
would
have
to
define.
Well,
if
you
want
to
encode
this,
then
you
have
to
have
multiple
items
and
they
get
stitched
back
together.
Somehow
because
items
can
be
in
multiples
and
repeating
sequence,
so
it's
not
stopping
anything
large
from
happening,
but
it
makes
it
more
complex
on
the
extension
to
do
so,
but
even
rossik.
Okay,
sorry.
H
The
yeah,
the
the
my
concept
originally
of
extensions,
was
I,
had
put
more
more
of
the
negotiated
data
into
extensions,
so
the
iya,
some
of
the
key
IDs
and
some
of
the
individual
frames,
small
parameters
and
even
he
IDs
being
related
to
URLs,
have
processing
limits
that
are
in
the
order
of
I
expect
thousands
of
look.
That's
not
a
single.
E
H
This
is
the
last
one
of
them.
The
current
status
is
I.
Do
have
some
that
last
so,
like
those
copy
and
paste
naming
big
things
face,
I
did
I'd
see
that
there
were.
There
were
actual
reason,
codes,
numeric
reason,
codes
included
an
Diana
tabled,
but
not
in
the
actual
requirement.
Body
tables
I
think
it
makes
sense.
They
include
the
reason
codes
in
both
just
because
it's
you
don't
have
to
jump
around
nothing
different,
but
some
documents
to
look
up
if
you're
trying
to
implement
this
thing
here.
But
it's
not.
B
B
Okay,
so
I
guess
we're
going
to
so
Brian.
If
you
could
liken
in
one
week
he
get
the
final
version,
so
we
can
forward
it
to
the
ihe
with
the
bundle
of
the
of
the
other
specs
Edie.
You
are
identified
as
the
document
shepherd,
so
we
would
be
looking
for.
You
know
the
Shepherd
right
up
cool.
Thank
you
very
much.
Are
we
Scott?
Are
we
ready
for
sending
the
your
document
to
for
publication.
E
F
E
B
D
D
We
did
not,
and
we
chose
in
the
working
group
not
to
make
those
context
part
of
the
security
specification
itself,
but
to
publish
them
as
a
separate
document
next
slide.
So
from
the
ITF
103,
there
was
a
single
question
that
came
out
of
a
review
from
a
different
standards
body
called
the
CCS.
Yes,
that
review
came
back
and
said
your
specification
of
cipher
Suites
is
different.
D
Notionally
from
how
we
like
to
use
security
associations.
Would
it
be
a
good
idea
IETF
if
we
were
to
create
a
security
associations
block
and
if
you
did,
what
would
that
look
like
what
value
would
it
have
in
the
in
the
security
suite?
We
looked
at
it
and
we
drafted
it
to
see
what
it
would
look
like
and
we
presented
it
to
the
working
group
and
the
working
group
resoundingly
said.
It's
a
terrible
idea.
D
D
The
compromise
was
to
change
some
of
the
terminology
so
that
it
was
not
confusing
and
in
the
security
document
instead
of
using
the
term
cipher
suite
ID,
because
cipher
suite
has
a
particular
meaning
to
many
people
to
say,
security
context
and
a
security
context
could
certainly
be
used
as
a
cipher
suite
identifier,
but
could
also
be
used
in
a
different
way,
particularly
to
carry
an
association
or
Association
information
other
than
that.
There
were
no
changes
with
the
security
documents
next
slide
and
then
in
following
through
well.
D
So
there
were
no
other
significant
changes
and
we
just
went
back
and
to
apply
that
change.
We
removed
the
term
cipher
suite
in
those
areas.
We
replace
it
with
security
context
instead
of
cipher
suite
ID
parameters
results
their
security
context,
ID,
params
and
results,
and
then
because
we
were
only
prototyping
notionally
with
the
security,
Association
block
would
look
like.
D
Obviously
we
took
all
of
that
out
in
the
oh,
nine
and,
and
those
were,
the
changes
was
reacting
to
that
final
decision
in
103,
so
VP
Seco,
nine
as
it
exists
now,
has
incorporated
the
comments
from
the
security
area,
which
was
simply
to
mandate
that
any
encryption
is
assigned
a
Shin
using
an
AE
ad
cipher
suite
and
to
clarify
that
integrity
in
this
case
means
signature
on
open
text
and
then
from
the
CCSD
s
perspective.
We
had
gone
back
and
they
gave,
and
actually
this
might
be
on
the
next
slide.
So
next
slide,
please.
D
Yes,
they
went
back
in
and
did
one
more
review
of
BP
Seco
nine
and
their
exact
word
was
oh,
this
looks
very
good,
so
we
take
that
as
a
positive
sign.
They
had
just
these
comments.
Coming
back.
Three
of
them
were
minor
clarifications
in
the
text
which
we
will
happily
make
within
the
next
week
and
they
simply
said
to
clarify
a
little
bit
of
the
language
when
you
say
ensure
integrity
services,
don't
ensure
that
data
isn't
changed.
It
just
ensures
that
if
it
is
changed
you
can
you
can
detect
it
so
little
things
like
that.
D
One
of
them
is
to
say:
why
do
you
allow
security
source
and
context
parameters
to
be
optional,
because
there
are
cases
where
you
will
not
need
them,
and
in
those
cases
we
don't
have
to
go
through
the
overhead
of
carrying
empty
fields,
and
the
other
is
when
you
say
that
authenticated
encryption
is
used
in
a
b
c
b.
There
should
be
a
Showell
statement
in
there,
not
just
a
note,
I
think
the
author
missed
it
actually
earlier
in
the
document.
D
There
is
a
must
associated
with
the
use
of
this
suite,
so
I
think
that
the
spec
actually
already
does
what
it's
being
asked
to
do
so.
I
I
don't
have
a
full
overview
that
full
up
overview
of
BP
SEC
onine,
because
we've
gone
through
it
in
this
working
group
several
times
it
has
gone
through
the
last
call
and
what
I
have
done
for
the
past
two
IETF
working
groups
is
simply
to
describe
the
changes
that
were
made
coming
out
of
reviews
and
how
we
have
responded
to
them
other
than
those
three
comments
from
CCS.
D
Yes,
to
clarify
some
of
the
terminology
there.
There
are
no
other
open
comments
about
the
specification
that
we
are
aware
of.
So
a
question
that
I
had
before.
Moving
on
to
the
I
apologize
for
the
title
slide,
the
interoperability
security
context,
it
is
to
come
back
and
say:
are
there
any
other
comments
associated
with
BP
SEC?
We
have
not
seen
them
online.
We
do
have
evidence
that
there
are
reference
implementations
that
people
have
made
for
this
without
issue.
But
at
this
point
we
believe
that,
aside
from
those
three
nets,
we
are
done.
E
Of
the
open
questions
as
Rick
Taylor
here
with
my
chair
hat
on,
is
the
working
group
asked
aides
to
room
well,
so
the
SEC
IDs
asked
for
the
interoperability
that
the
cipher
Suites
document
to
be
written
and
to
be
maintained
as
a
separate
document
so
that
it
could
be
updated
independently
to
the
to
the
main
document
of
the
main
document
is
sat
in
last
call
at
the
moment.
It's
ready
to
go
and
it's
going
forwards.
E
D
There
is
another
document
called
the
interoperability
security
context,
which
had
been
a
personal
draft
and
then
was
submitted
as
a
working
group,
and
the
working
group
had
decided
that
we
needed
to
put
that
together
with
the
BP
SEC
documents,
so
that
the
reviewers
could
review
both
a
protocol
and
an
example
of
how
you
would
define
and
use
a
security
context,
and-
and
that
second
is,
is
these
suites
here?
Essentially
the
only
changes
from
that
context?
D
Is
we
changed
the
terminology
from
cipher
suite
to
security
context,
with
the
rather
notable
exception
of
the
title
of
this
slide?
We
updated
the
references
to
point
to
the
latest
release
of
BP,
biz
and
BP
sac,
and
we
submitted
it
as
a
working
group
document
and
I.
Don't
know
if
I
have
another
slide
behind
this.
That
speaks
what
I
thought.
I
did
and
I
was
wrong
yeah.
D
The
H
max
sha
for
integrity,
a
s4
for
confidentiality
that
they
also
have
the
benefit
of
being
well
represented,
also
in
C
bar
with
cozy,
and
we
were
able
to
point
to
some
references
there
in
terms
of
how
those
items
would
be
seemed
or
encoded
which
we'd
like
to
have
the
c4
encodings.
So
we
think
that
those
are
good
interoperability,
suites
and
we
hadn't
heard
any
concern
from
the
reviews.
The
personal
draft
reviews
that
we
have
had,
but
we
would
need
some
review
at
the
working
group
level
of
those
security
contents.
E
So
continuing
from
where
I
left
off
the
Ed's
personal
document
is
now
a
working
group
document
that
was
accepted
after
the
last
IETF
in
Bangkok,
the
and
I
press
the
button
very
late,
but
it
is
now
a
working
group
documents.
Of
course
the
BP
SEC
itself
has
been
hanging
in
last
call
ready
to
go
and
we're
now
going
to
send
it
to
the
ad.
E
So
really,
this
document
was
intended
to
go
with
it,
but
it
has
not
been
last
called
yet
do
we
does
the
working
group
believe
this
document
is
ready
for
last
call,
and
that
is
a
formal
question.
We
will
ask
it
on
the
mailing
list
as
well,
but
I'm
judging
the
room.
Our
intention
is
to
last
call
this
get
the
comments
and
then
move
it
on
as
fast
as
possible.
So
people
need
to
be
prompt
if
they
have
issues
but
we'll
get
that
question
to
the
mailing
list
on
the
mailing
list
as
soon
as
possible.
Q
So
hello,
I'm
zhang
zhehan,
so
I
will
be
presenting
the
work
I've
done
on
interfacing
the
synchronous
in
synchronous
network
management
protocols
next
slide
please.
So
our
network
management
is
administrating
and
monitoring
different
set
of
nodes
on
a
network
protocol
users
are
usually
very
specific
for
network.
So,
on
my
opinion,
work
we
would
have
an
account
we
would
have.
An
protocol
would
have
a
low
delay
high
bandwidth
and
we
will
make
use
of
these
whereas
and
conferences
on
dgn.
We
don't
have
all
of
these
prerequisites
so
another
we
have
to
do,
have
different
constraints.
Q
So
on
an
IP
network,
for
instance,
we
have
what
we
call
the
pool
mechanism
on
the
left,
so
we
would
send
a
request
and
it
that
I
would
be
sent
back
directly
under
reply
and
each
end.
The
way
is
completely
different.
We
would
send
comments
to
have
some
ultimate
and
some
automation
with
incense,
for
instance,
packages
on
the
right.
The
graph
a
report
will
be
sent
every
10
minutes,
and
so
this
would
be
called
the
pressure
mechanism,
so
both
of
this
approach
every
optimal
for
the
only
networks,
but
they
wouldn't
work
on
the
other
one.
Q
So
if
we
wants
it,
for
instance,
to
manage
some
notes
on
a
TN
from
notes
presence
on
an
IP
network,
we
would
either
use
one
or
the
other
they
wouldn't
be
very
efficient.
So
what
we
did
was
thinking.
We
could
combine
the
two
mechanisms
and
exactly
so
so
our
goal
was
to
interface
in
synchronization
protocol
or
amp
and
network
infection
protocol
negev.
So
we
wanted
to
create
a
gateway
on
a
node
that
would
be
on
an
IP
network
and
under
the
chain.
Q
So
I
didn't
a
phase
between
the
two
the
we
wanted
to
for
that
we
had
to
translate
the
data
models
that
are
used
by
HTTP,
and
so
the
ATM
and
by
Anakin's
at
a
young
data
model
and
finally,
we
are
proposed
nian
would
be
from
an
account
manager
to
be
able
to
send
controls
to
an
amp
agent
to
fetch
data
from
this
agent
and
then,
who
gets
read
and
from
we
can't
measure
all
of
this
data
in
excited.
So
generally
architecture
is
what's
presented
here.
We
would
have
on
an
IP
network.
Q
Q
With
some
specific
role
they
have
been
set
before
next
I'd
listen,
so
the
protocol
stack
that
we
used
is
present
here
and
message
would
be
sent
from
an
approved
user
interface
B.
Since
you
given
to
the
manager
we
sent
on
the
IP
network,
we
see
the
make
of
agents
on
the
Gateway.
That
would
be
in
a
face
read
by
an
in
conduct
role.
That
would
translate
it
into
amp
sense
given
to
an
integer
if
it's
to
give
it
to
the
on
the
DTM
and
we
receive
only
agent
and
other
way
around.
Q
It
would
do
the
exact
sorry
and
for
this,
because
we
have
these
synchronous
and
asynchronous
mechanisms.
We
need
to
have
something
to
get
that,
sir
evenly,
so
we
have
a
cache
on
the
Gateway
meaning
to
on
one
side
to
register.
If
you
want
to
have
a
value,
but
the
value
is
not
available,
we
would
register
that
this
value
is
knitting
and
on
the
other
side,
when
values
I
sent
from
the
impatience,
then
we
would
store
all
this
value
on
the
Gateway.
Q
So
the
Gateway
itself
has
this
architecture.
Net
messages
would
be
sent
on
the
left
to
the
ancients.
Given
to
the
controller,
we
would
extract
it
using
the
information
using
transition
function
then
translated
into
and
messages
that
would
be
given
to
the
manager
in
between.
We
also
cache
the
value
and
on
the
other
side,
when
data
sent
by
the
MTA
agents
they
would
be
given
to
the
manager
and
the
control
we
extract.
All
of
these
dates
are
putting
into
the
cache
and
but
initially
send
messages
to
the
nakum
managers.
Q
So
translation
approach
should
be
using
is
present
in
here
on
the
left.
You
have
an
extract
of
the
ATM
for
amp
agents,
so
we
have
the
control
here,
which
is
called
gin,
rpts
or
general
reports.
These
are
two
parameters
that
you
can
see
on
the
red
one
to
our
collections
of
our
ISO
recently,
fire
that
identifies
any
element
on
the
right.
We
translated
this
control
into
our
PC
into
in
Mein
Kampf,
and
so
this
obviously
has
also
two
parameters
which
here
leaf
lists
which
correspond
to
a
collection
in
account.
Q
Q
This
utility
is
use,
as
an
account
manager
said
so,
for
instance,
if
I
want
to
send
general
reports
comment,
we
use
it
here,
T's
syntax,
so
the
young
that
emoji
would
be
read
and
parsed
by
this
it's
routine
and
we
can.
It
would
be
all
those
who
use
this.
So
we
have
here
two
parameters.
The
first
would
be
an
air
I
of
value
say,
for
instance,
you
want
to
generate
a
report
that
would
sense
the
number
of
visual
diary
and
the
agent.
Q
So
to
do
so,
we
give
as
an
ID
what's
value
one
generate
here,
the
namespace
of
it,
and
it's
also
needd.
So
next
time
only
defined
data
and
its
name
is
BR.
So
all
of
this
is
good
as
one
pointer.
The
second
point
models
are
the
end
points
of
the
manager
word,
which
the
agent
have
to
send
messages.
So
here
we
have
two
of
them.
We
have
IPIN
4.5
in
a
pin,
6.5
and
next
slide.
Please.
So
this
command
sent
by
the
income
manager,
which
translates
it
as
an
RPC
XML
RPC.
Q
So
you
have
it
on.
You
have
the
parameters
in
the
red
rectangle.
So
this
point
is
a
PC
will
be
received
on
the
Gateway.
The
Gateway
will
then
radiate
sparsit's
and
transform
it
into
the
dictionary
you
have
on
the
right.
So
this
one
is
an
internal
dictionary
that
we
use
in
the
Gateway.
So
you
can
see
the
area
diptych
all
the
specific
parameters,
so
you
have,
we
can
see
it's
an
amp
agent,
it's
a
control.
The
name
is
it's
from
sorry.
It's
the
ATM
is
an
pigeons
control.
The
name
is
general
Hertz.
Q
Then
you
have
the
parameters.
First
parameter
is
again
an
RI
with
his
name
space
at
the
type,
and
everything
second
parameter
is
the
lists
that
we
sent
before
and
this
this
structure
is
used
in
two
with
a
function
that
will
send
the
control
from
the
IMP
manager
and
that
Delta
G
is
a
recruit's
completely
to
recur
CBG.
So
if
you
have
parameters,
we
have
partners
in
the
air.
I
that
needs
to
have
parameters
also
itself
and
we
could
recursively
at
them.
Q
When
this
message
will
be
sent
to
the
manager
to
the
agents
report
to
be
sent
back
to
the
manager
on
the
Gateway,
so
you
have
here
an
example
of
this
such
report.
The
value
sent
would
be
the
ten.
This
will
be
stored
on
the
database.
When
then,
we
have
created
yang
module,
miss
no
in
Moodle,
with
several
commands
to
fetch
this
data
from
any
managers.
So
we
have
here,
gets
in
key
value,
even
32
to
as
forums
and
integer.
It
has
two
parameters.
Q
The
first
one
value
F
is
the
name
of
the
value
we
want
to
fetch
with
namespace
type
and
name.
Second
parameter
is
two
to
decide
how
all
hunan
the
data
to
be.
So
if
we
want
you,
if
we
give
a
small
burden,
the
data
to
all
would
be
considered
still
under
cache,
and
if
we
give
along
a
very
high
number,
then
we
can
have
more
data
from
this
guide
from
those
cache.
So
it's
the
time
it's
the
time
in
seconds.
Q
For
instance,
if
we
had
a
data
that
is
less
than
the
South
second,
all
on
the
cache,
then
the
APIs
you
apply
on
the
bottom
left
would
be
sent
back
to
the
neck
of
manager
with
the
value
10
and
the
three
timestamp,
which
are
no
means.
The
value
was
requested
on
the
Gateway.
The
moment
was
persist
on
agents.
In
the
moment
it
was
sent
back
on
the
Gateway
on
the
right.
Q
If
we
didn't
have
any
value
in
the
cache
or
if
the
value
were
considered
tails
at
all,
we
would
then
wait
for
a
value
value
to
arrive
and
the
recent
value
will
be
sent
as
a
notification.
An
application
like
a
mitigation
to
do
to
the
manager.
So
instead
of
I
mean
a
reply.
That
would
be
a
scary
request
and
any
require
bit
sense.
That
would
be
it
just
like
a
push
mechanism.
The
message
will
be
sent
directly
without
any
requests,
except
so
I
would
do
a
small
demonstration
I.
Q
E
R
P
Q
I'm
using
the
here,
the
mitre
simulator,
so
from
a
technical
detail,
kids,
it
combines
the
core
simulator
for
networks
and
it
also
uses
the
iron
sword
in
Ness.
Implementation
and
I
have
a
very
simple
scenario,
with
three
notes
so
on
the
Left,
it's
under
network
IP
network
on
the
right
is
a
summation
of
the
TN.
So
you
have
the
notes
that
moves
is
another
new
TN
in
the
middle.
You
have
the
gate.
Wing
I
have
to
create
some
directions
and
port
forwarding,
because
for
networking
it
has
to
use
the
normal
I
P
Network.
Q
Q
Q
So
on
the
bottom
right,
you
also
have
the
NIP
health
agents,
sorry,
the
agent
which
is
working
so
now
I'm
connecting
to
the
Gateway,
but
as
if
he
was
connected
to
the
agents.
I'm
loading,
the
young
data
module,
which
is
on
page
ins
and
then
I
will
add
ER
so
tempest
rolled
on
the
Epogen
to
send
regularly
time
some
information
like
it's
gateway.
Q
Q
Q
So
we
have
here
I'm
agent
control
regions,
Jinnah
reports,
so
I'm
first
parameter,
would
be
a
collection
of
two
elements.
We
will
ask
for
the
number
of
its
type
is
rule
on
the
aim,
agents
or
not
the
next
year.
The
second
parameter
would
be
the
number
of
control
the
happy
one
on
the
end
agents,
and
this
would
be
the
first
parameter.
Q
Second
parameter
I,
give
it
simply
an
empty
list,
so
it
would
be
four
sent
who
asked
for
which
managers
and
shoes
we
leave
it
empty.
Here
there
was
a
mistake
in
the
name
of
the
elite,
not
on
purpose.
I
would
like
to
say
it,
but
I
didn't
write
it,
because
it's
nice,
so
I
can
show
you
that
we.
If
there
is
a
mistake
in
the
type
check
or
in
the
name,
then
Ike,
we
can
still
check
it
using
the
idiom
and
prevents
the
message
from
these
things.
Q
Q
Yes,
so
here
I
actually
had
to
cut
the
video,
because
it
took
two
minutes
so
two
round-trips
for
the
data
to
arrive
under
the
manager
on
the
Gateway.
Here
we
could
see
we
had
more
data.
There
is
presence.
So
far
already
reports
has
been
rested
on
the
Gateway
and
if
you
use
the
special
yang
comment
that
we
created,
if
we
always
explain
before,
then
we
can
ask
for
in
Santa
too.
Q
So
we
will
ask
first
parameter
for
which
we
ask
we
see
which
value
once
you
have
so
here,
the
agent
EDG
number
of
control
run
on
the
agent
and
I
will
write
here
is
not
a
cache
timeout
to
show
you
the.
If
you
write
a
small
cache
timeout,
then
there
is
no
enough
read
more
reason
that
we
send
soldiers
in
a
prettier
place
sense.
But
if
you
increase
this
tigers
timeout,
then
we
get
the
most
recent
send
data
again
with
this
three
timestamp
I
explained
before
on
the
soroban.
Q
Q
So
in
conclusion,
we
be
able
to
send
any
control,
define
an
idiom
using
account.
We
can
also
have
in
these
controls.
We
can
have
several
parameters
for
you,
so
parameters
in
parameters
and
so
on,
and
we
can
retrieve
any
data
from
the
Gateway
directly
using
income
and
also
if
some
dates
new
data
is
arriving.
Then
notifications
are
sent
to
do
an
account
manager
and
exactly's.
So
there's
two
futures,
obviously
so
currently
focus
is
on
the
supporting
detail,
all
templates
from
NP,
which
are
heavily
use
it
in
some
used
in
some
idioms.
Q
So
we
want
to
be
able
to
get
the
reports
and
retrieving
letting
as
make
it
available
for
the
neck
of
Niger.
Also,
we
reduce
the
scope
to
having
only
one
gateway
here.
If
you
had
several
gates
aware
that
could
be
connected
all
together,
but
so
connected
to
all
through
the
same
GTM,
then
we
need
to
have
some
synchronization
between
all
the
gateways,
which
was
not
the
focus
of
this
work.
And
yes,
so
we
have
also
an
irrational
to
calculate
the
cash
that
how
long
we
have
to
wait.
Q
We
have
default
values
for
that
also
if
the
data
is
in
the
setting,
acceptance
range
arrive
at
the
right
sound
and
we
directly
without
asking
for
a
new
value.
This
issue
here
is
that
we
have
to
get
the
value
because
it
would
be.
Then
you
work
for
optimization,
and
it
is
not
the
focus
again
on
sweaty.
D
Is
that
is
not
the
cleanest
way
to
interface
with
something,
as
maybe
you
would
understand,
but
there's
other
work
going
on
to
create
either
an
RPC
interface
or
other
kinds
of
interfaces.
That
would
be
more
amenable
to
that.
So
the
other
part
of
the
optimization
is
some
of
the
stack
that
was
shown
had
things
in
there
because
of
the
necessity
of
working
with
an
existing
reference
implementation.
K
Q
Haven't
I
know
about
it.
I
know
that
the
conf
is
usually
is
supposed
to
have
some
commits
on
the
database
and
everything,
and
so
I
have
some
small
support
that,
if
you
comment
is
done
at
the
time
and
all
that
the
second
all
they
are
waiting
for
it
to
be
done.
This
is
not
as
developed
as
in
a
real
net
conf
database.
This
is
definitely
something
to
you
days
in
the
development
now.
E
A
Rick,
Taylor
I
actually
have
a
question
for
Erik
Voigt
who's
just
asked
who
I
happen
to
know
as
finally
got
his
yang
push
drafts
through
net
Kampf,
which
is
as
I
understand
it
and
I
am
NOT
a
net
complex,
but
a
it's
neck
off
notifications
on
steroids.
Really
isn't
it
do
you
have
any
advice
for
the
guys
working
on
this
on
whether
it's
that
push-pull
thing
you
started
your
presentation
with
suddenly
made
me
think
of
the
work
that
Eric's
been
working
on
and
whether
that
interface
could
be
more
elegant,
I.
K
D
Comment
just
at
Berean
again,
but
it
so
agreed,
but
in
the
example
that
was
presented
here,
there
was
a
command
to
say:
go
generate
this
value.
You
know
every
X
amount
of
seconds
or
minutes
or
days
for
this
amount
of
time
and
then
a
separate
query
had
to
come
in
and
say
that's
been
running
for
a
while.
Now
I
would
like
to
see
if
that
data
has
changed,
and
then
you
get
the
data
cache
timeout
which
says
hey.
Do
you
have
a
new
version
of
this?
D
But
if
you
had
something
like
what
Eric
has
been
working
on,
you
could
say
generate
this
data
every
you
know
10
days,
but
whenever
the
value
changes
send
it
to
me
and
then
that
would
also
help
with
that
that
that
that
push
inside
of
the
you
talk,
we
talk
about
push
and
pull,
but
you
don't
always
know
when
to
pull,
and
this
would
help
you
understand
when
to
pull.
Yes.
E
So
Rick
again
says:
I
think
we
summarizing
from
my
own
since
you're
saying
the
gateways
data
store,
that's
subscribing
to
yank
push
notifications
on
the
gateway
data
store,
which
is
being
updated
by
the
amp
management
system.
But
your
downstream
net
kampf
system
can
say:
oh
hey,
the
amp
subsystem
has
changed
the
data
store
in
the
gateway
and
therefore
I
am
now
receiving
notifications
of
updates
yeah.
K
What
fun
what
great
discussions?
Yes
in
this
case,
if
you're
doing
that
I
would
suggest
use
not
subscribing
to
data
scores.
You
subscribe
to
events,
we
actually
built
the
protocol
so
that,
when
you
log
in
it
will
give
you
the
history
of
events
even
previously
created
events
as
a
replay
capabilities,
they
have
the
ability,
even
if
you
lose
connectivity,
where
it
will
actually
give
you
a
new
start
from
a
last
event
that
was
sent.
So
it's
been
designed
for
filling
in
gaps.
Yeah.
L
B
Magnus,
while
it
is
going
to
the
mic,
it's
it's
not
a
new
question.
You're
asking
it's
been,
it's
been
discussed
so.
D
So
the
the
the
short
answer
is
there:
there
is
no
mechanism
to
say,
don't
overdo
a
quality
of
service,
that's
in
a
DTM,
because
there
it's
not
always
clear
that
you
understand
what
that
breaking
point
is
the
way
that
amp
talks
about
it
in
terms
of
DTN
is
it
will
create
bundles.
It
will
so
the
way
that
amp
addresses
it
as
a
protocol.
Separate
from
this
demonstration
is
to
say
that
we
try
and
have
very
compact
encodings.
B
Mark
here
I
have
a
question
for
the
speaker,
so
my
understanding
is,
you
have
somewhat
implemented
EMP
or
parts
of
it
to
do
your
translation.
E
But
mark
raises
a
very
important
point,
which
is:
although
the
ANP
work
officially
comes,
post
are
sort
of
phase
one
DTN
charter
items.
I
know
we
have
a
current
charter
item
on
the
architecture
of
management
in
DTN,
but
there's
been
a
lot
of
interest
around
what
Ed's
been
doing
and
his
team
have
been
doing
a
lot
of
work
and
it's
a
proposal
that
we
should
look
at
a
MP
and
work
on
this
as
part
of
kind
of
phase
2
of
the
DTN
working
group.
E
D
So
hi,
I'm,
ed,
again
and
well.
The
two
things
are
working
were
security
and
network
management.
We
had
talked
about
security
and
we
had
started
talking
through
network
management.
The
I
did
not
have
a
video
and
was
not
clear
to
me
that
I
could
share
my
screen.
So
what
I
have
is
a
series
of
screenshots
and
I
apologize
for
that,
but
either
John
or
I
could
could.
If
anyone
wants
to
stop
at
the
end
of
the
meeting
and
get
a
live
demo,
neither
of
our
laptops
we
can.
We
can
show
you
what
is
going
on.
D
We
we
have
a
set
of
specifications,
they
are
personal
internet
drafts
right
now
for
an
asynchronous
management
architecture,
data
model
and
binary
encoding
and
protocol,
and
we've
broken
those
out
as
three.
So
we
can
reason
about
them
separately
within
ion,
which
is
a
an
open
source
with
an
ion
open
source,
which
is
an
open
source
implementation
of
those
protocols.
There
is
a
fully
compliant
amp
manager
and
amp
agent.
We
are
talking
with
folks
who
maintain
DT
and
two
who
are
discussing
who
will
be
building
an
amp
agent.
D
The
thought
is
that
if
you
build
one
amp
manager,
then
it
can
manage
any
agents
regardless
of
whether
they're
in
ion
or
DTN
2,
but
we
will
have
at
least
two
and
then
other
space
agencies.
Khari
DLR
are
also
considering
their
own
amp
agents,
and
there
are
at
least
two
different
efforts
currently
to
build
alternative
interfaces
into
an
amp
manager,
one
being
a
web
interface
and
other
being
something
that's
driven
out
of
sequel
and
sequel
tables.
D
So
so
there
is,
there
is
some
sense
of
what
is
implementable
and
how
easier
hard
is
it
to
implement
and
what
features
would
we
get
out
of
it?
What
I
wanted
to
talk
through
and
we
can
go
to
the
next
slide-
is
what
we're
doing
under
ion
as
a
reference
implementation.
So
just
a
few
reminders
of
the
things
that
are
presented
here
are
based
on
either
personal
internet
draft
or
something
that
you
could
and
all
the
software
certainly
can
be
downloaded
off
of
SourceForge
and
looked
at
just
as
a
piece
of
terminology.
D
I
know
that
I
see
a
lot
of
new
faces
in
the
room.
I
know
we
had
talked
about
network
management,
asynchronous
management.
In
the
past,
we
draw
a
distinction
between
the
architecture
and
asynchronous
management
architecture
AMA,
which
talks
about
why
we
need
time-based
and
state-based
and
other
kinds
of
functions.
And
then
we
have
a
data
model
and
the
data
model
says
here
are
the
nine
types
of
data
that
we
need.
D
Tables
reports
strongly
type
variables
and
so
on,
and
this
is
how
you
would
use
them
and
why
they
should
exist
and
why
it's
not
duplicative
with
other
elements
of
a
model.
And
then
we
talk
about
adjacent
encoding
of
that
and
how
you
could
represent.
We
call
them
a
DMS,
but
we
could
think
of
them
as
mids
or
yang
modules
in
the
function
that
they
serve
a
JSON
template
for
their
specification.
D
So
we
could
write
tools
to
ingest
and
operate
on
them
and
then
the
ANP
is
simply
AC,
boring
coding
of
those
models
separate
from
a
string
encoding
or
a
JSON
encoding,
so
that
we
could
have
more
efficient
messages
on
the
wire
within
ion.
There
is
a
network
management
subdirectory
under
the
main,
open
source
directory
called
nm,
and
that's
where
all
of
the
implementations
live
in
Doc's,
a
DMS
is
where
Jason
ADM
files
live
and
in
contrib
camp
Python
is
where
a
particular
code
generator
is
now
show
you
some
screenshots
for
that.
D
When
you
build
and
install
ion,
you
get
a
manager
and
an
agent
that
you
can
run
next
slide.
If
you
want
to
start
the
manager,
you
start
it
by
typing
in
nm
Manager
and
as
a
courtesy
for
bootstrapping.
Just
in
this
reference
implementation,
you
give
it
you
tell
it
what
a
ID
it
should
have.
It
will
come
up.
D
It
will
read
any
persisted
definitions
from
the
I
on
simple
data
recorder
that
is
there
and
then
it
gives
you
a
glorious,
text-based
interface,
for
which
you
can
type
text
to
your
heart's
content,
and
while
it
is
functional
and
we
can
walk
through
it,
if
anyone
is
is
so
moved
to
continue
to
create
alternative
interfaces
into
this,
we
would
we
would
very
much
like
to
help
and
support
and
encourage
that
with
next
slide.
The
agent
similarly
is
started
up
on
the
command
line.
D
Again,
once
once
ion
is
running,
you've
done
make
install
you
get
animation
and
a
manager.
The
agent
takes
it
Sony,
ID
and
a
default
manager.
In
the
previous
presentation,
there
is
a
discussion
of
the
receive
managers.
If
you
don't
specify
a
receive
manager,
it
goes
to
a
default
manager
which
you
can
specify
here.
It
also
comes
in
sets
off
and
then,
when
it
does,
it
sends
a
registration
message
to
its
default
manager.
That
says
hello,
I
exist,
it's
not
required.
D
You
can
go
in
and
tell
the
manager
that
agents
exist,
but
it's
a
courtesy
next
slide.
So
what
can
you
manage
on
the
agent
when
you
go
into
this
text-based
interface?
You
see
the
agents
that
have
come
up.
These
are
the
ones
that
this
is
the
auto
registered
one.
You
can
forget
about
the
agents
and-
and
you
can
print
reports
or
persist,
reports
that
come
in
and
then
you
can
really
just
do
commanding,
which
is
mostly
what
you
want
to
do
anyway
and
that's
under
build
control.
D
There's
the
ability
to
send
raw
hex
to
to
the
agents
as
well,
which
is
something
that
we
just
have,
is
sort
of
an
expert
interface
so
that
we
can
capture
and
replay
things
next
slide.
So
what
does
ion
support
it
right
now?
What
we
have
done
is
generated
JSON,
encoded
ATMs,
which
are
also
published
as
personal
drafts
for
the
agent
ADM
itself,
which
is
this
is
the
ADM
all
trolls
and
the
data,
and
so
on
that
every
a
MP
agent
should
be
able
to
understand,
and
then
we
have
done
that.
D
Similarly,
for
a
generic
bundle
protocol
agent,
a
generic
LTP
agent
and
a
generic
VP
SEC
/sb
SP
agent,
and
then
we
then
created
a
set
of
specific
ATMs
for
ion
itself.
If
you
have
ever
used
ion,
you
know
that
there
are
multiple,
not
too
many,
but
a
couple
of
administrative
functions
that
you
use
to
control.
Ip,
n
admin
ion
second
min
BP
admin,
and
you
can
control
ion
by
SSH
into
the
machine
that
it
is
running
on
and
running
these
commands.
D
But
in
this
way
we
took
essentially
the
man
pages
for
all
these
created
ATMs
Auto
generated
the
code.
So
now
you
can
control
multiple
instances
of
ion
on
different
agents
over
just
a
bundle
interface.
You
know
using
the
N
P
protocol
next
slide,
so
on
so
I
just
go
back
one.
So
in
general,
that's
not
a
very
large
amount
of
data.
It's
it's
somewhat
large
for
for
us,
where
we
are
now
with
these
protocols,
but
it's
158
data
definitions
about
94
controls,
22
tables
and
so
on
next
slide.
D
We
have
tables
here's
an
excerpt
of
tables
for
a
generic
BP
agent.
One
of
the
tables
we
call
them
table
templates
is,
for
example,
alt
docks.
If
you
are
configuring
ion
and
you
would
like
to
get
a
general
dump
of
what
the
end
points,
what
are
the
index
out?
Docs
protocols
supported
schemes,
and
so
on.
You
can
generate
that
information
in
a
table
format
and
get
it
pretty
quickly,
and
it's
it's
just
a
nice
way
to
parse
that
remotely
next
slide.
D
So
here
would
be
an
example
of
a
poorly
ASCII
formatted
table
result
coming
out
of
the
text
interface,
which
is,
if
you
just
bring
up
a
loopback
interface
and
one
of
the
loopback
interfaces
that
is
bundled
with
ion.
You
will
get
in
this
case.
I
used
s
TCP,
but
this
would
be
the
out
ducts
that
are
Auto
configured
on
ion
or
s
TCP
next
slide.
And
then
you
also
have
a
report
and
report
templates.
They
are
a
bit
like
tables.
D
The
difference
is
that
in
our
model,
users
can
generate
custom
reports
as
well
that
pull
in
data
from
various
other
ATMs
as
they
need
to,
and
you
could
parameterize
reports
and
and
that
bit
of
associative
look
up
was
pretty
motivating
to
the
use
cases
that
we
have
so
the
generic
BP
ADM
has
two
reports.
One
is
a
full
report,
which
is
all
the
stuff
that
we
know
about.
That
is
specific
to
the
agent
at
the
overall
node
for
the
agent
and
then
a
specific
end.
D
Point
report
that
says
for
every
endpoint
here
are
a
couple
of
pieces
of
information
that
we
track
separately
next
slide.
We
can
certainly
request
a
report.
If
a
report
comes
in,
we
can
print
it
to
the
screen.
It
has
a
lot
of.
It
has
43
data
items
in
it,
but
a
data
item
just
like
a
report
can
also
be
parameterized.
There
are
cases
where
it
is
helpful
and
useful
to
do
that
in
cases
where
it
is
not
an
example
of
where
we
use
it,
and
the
available
storage
here
is
a
configuration
issue.
D
I
don't
actually
have
that
much
storage
on
my
machine,
they're
tremendous
well
to
some
unit
of
measurement,
I'm
sure
that
I
do.
But
if
we
look,
for
example,
bundles
by
priority,
that's
a
piece
of
data,
but
you
give
it
a
mask,
and
you
could
say
give
me
the
bundles
by
priority
for
all
priorities
or
just
for
low
priority
or
medium
priority
or
high
priority
in
the
JSON
ATM.
When
you
specify
this,
you
can
specify
I
would
like
this
data
item
with
this
particular
hard-coded
parameter
or
with
a
parameter
that
flows
in
from
somewhere
else.
D
We
particularly
like
that,
because
we
don't
always
understand
how
many
different
types
of
things
we
will
have,
so
we
don't
always
size
a
static
array.
We
also
don't
want
to
have
to
do
one
spin
of
understanding
how
to
match
to
an
array
index
a
particular
associative
lookup
and-
and
we
like
the
idea
of
being
able
to
put
the
associative
lookup
in
the
request
itself
next
slide.
D
So
if
we
look
at
then
with
the
JSON
files
themselves
are
so
in
this
case,
there's
one
called
the
SP
SP
in
iti
think
VP
SEC
agent,
and
this
is
how
you
define
a
control
right
now,
we've
tried
to
keep
it
simple
as
simple
as
possible.
You
have
the
name.
You
have
the
parameters
that
it
takes.
You
have
a
description.
There
is
some
other
metadata
associated
with
the
a
with
the
ADM
itself,
I'm
sure
there
may
be
over
time
some
of
the
things
we
would
like
to
add
when
we
define
a
control.
D
We
are
all
ears
for
that.
But
again,
the
BCB
rule
takes
5
parameters
when
sending
from
source
to
destination.
This
target
use
a
b
c
b
with
this
cipher
suite
security
context.
At
this
point
and
keenan
next
slide,
BP
endpoint
report,
parameterised
report,
you're
able
to
come
in
and
and
say,
I
have
a
report,
and
the
report
itself
takes
a
parameter
in
its
definition
and
the
report.
D
Template
will
have
three
pieces
of
data
in
it
and
each
of
those
three
pieces
of
data
also
happened
to
be
parameterised
listed
somewhere
else
in
the
ADM
and
I
would
like
to
flow
the
endpoint
ID
parameter
into
the
definition
of
the
ones
that
contain
the
report.
So
so
what
is
interesting
about
that
is.
The
ADM
has
a
way
of
specifying
whether
and
how
you
flow
parameters
into
contained
objects,
and
then
you
don't
have
to
worry
about
that
when
you
auto-generate
the
code
or
when
you
specify
it
in
some
other
way,
eggs.
E
Like
it
can
I
just
jump
in
there
because
I
know
from
my
conversations
because
I'm
pretty
dumb
when
it
comes
to
these
things,
the
parameterization
is,
is
a
very
strange
concept
and
you
wonder,
while
this
effort
has
gone
in
the
reason
is
to
make
sure
that
pretty
much
all
these
requests
are
one
round
trip,
because,
unlike
you
know,
this
is
it.
This
is
a
DTN.
It's
gotta
be
one
round
trip,
because
you
could
have
four
hours
of
latency
on
this
I'm,
particularly
ok,
guys
for
the
neck
home.
E
E
T
D
I,
it
is
also
a
point
that
we
have
talked
about
in
the
AMA
document,
which
is
sort
of
our
architecture
and
reasons
and
rationale
for
why
we
made
some
design
decisions,
although
we're
certainly
happy
to
revisit
any
of
them
and
and
talk
through
them.
Next
slide,
then,
if
you,
if
you
have
these
Jason
scripts
or
these
Jason
files,
we
package
a
Python
script
with
the
I
on
open
source.
It
lives
in
NM,
/,
contribs
camp
Python.
D
D
If
you
had
previously
auto-generated
and
hand-tooled
some
dot,
H
and
C
files,
you
can
give
those
as
parameters
back
and
it
will
carry
your
changes
through
when
you
do
updates
and
that
sort
of
code
round-tripping
and
what
it
will
produce
in
an
output
directory
of
your
choosing,
our
v
dot
h,
d--
files
that
can
be
compiled
into
ion
to
help
with
its
adoption
next
slide.
So
it
will
create
an
agent
directory
with
three
files
that
get
compiled
into
an
agent.
D
It
will
create
a
manager
directory
with
a
file
that
gets
compiled
into
a
manager
and
then
a
shared
dot
H
file
that
can
go
between
them.
What's
interesting
is
is
that
the
manager
file
does
not
need
any
user
edits.
It
has
all
of
the
information
in
it,
as
is,
and
you
don't
ever
need
to
look
at
it
again.
The
agent
has
one
file
that
is
non
user,
editable
and
then
two
files
add
a
check,
see
where
you
go
in
and
put
implementation
information
next
slide
and
here's
an
example
of
what
that
looks
like
the.
D
We
have
some
API
parameters
for
extracting
and
validating
parameters,
and
Aion
already
has
an
eye
on
secondment
function
that
performs
this
purpose,
and
we
just
pull
the
parameters
out
and
call
the
function
with
some
error.
Checking
next
slide.
In
addition
to,
if
you
define
a
function,
you
have
to
give
implementation
for
the
function.
If
you
describe
a
data
value,
you
must
find
a
way
to
sample
that
data
value
we
had,
for
example,
in
the
BP
endpoint
report.
One
data
item
was
whether
the
endpoint
was
active.
D
This
is
the
kind
of
stuff
that
is
auto-generated
for
a
defined
piece
of
data.
If
you
then
go
to
the
next
slide,
you'll
see
an
implementation
of
how
to
go,
get
and
return
that
piece
of
data
the
last
one
and
I
don't
have
an
example
for
it
here,
although
it
is,
there
is
a
table.
If
you
define
a
table
template,
you
need
a
function
that
will
populate
and
send
back
the
table,
but
otherwise
for
reports
for
macros
for
variables
for
time-based
rules
for
state-based
rules,
there
is
no
user
action
associated
with
that.
D
You
implement
functions,
you
implement
data
collection,
you
implement
tables,
and
then
everything
else
is
auto-generated
and
compiled
in
for
you,
and
it
has
made
the
adoption
of
these
ATMs
in
Aion
go
quite
quickly
and
next
slide.
That
might
be
it
so
so
that
is
where
we
are.
There
is
a
reference
implementation
in
ion.
D
There
are
scripts
that
will
parse
JSON
ADM
files,
the
formats
for
all
of
these,
the
a
MP
and
the
C
Boer
encoding,
the
ADM
and
the
JSON
template
the
overall
architecture
and
the
design
decisions
and
all
of
the
ATM's
that
we
have
talked
about,
which
I
think
total
is
about.
15
documents
are
are
available
and
they
are
published
as
I
think
personal,
personal
IDs,
but
I
think
you
can
get
to
them
in
some
way
from
from
the
DTN
page.
D
E
J
Hello,
my
name
is
Martin
Milkha
I'm
from
ex-works,
and
we
are
working
on
hackathon.
We
did
some
work.
We
would
like
to
present
to
you.
We
do
have
implementation
of
pi
d
TN,
which
is
a
python
implementation
of
some
basic
DT
and
stuff,
and
although
we
are
communicating
with
Microvision,
so
this
is
integration,
work
and
microfiche.
Tn
is
developed
by
D,
free
tea
and
network
from
Felix
and
Felix
Walter
and
Marie's.
Money
like
this.
J
So
basically
is
the
presentation
we
shot
on
the
hackathon.
Our
use
case
is
very
simple:
we
do
have
two
ground
stations
and
we
would
like
to
communicate
with
each
other.
They
do
not
have
any
any
other
means
of
communication,
just
a
satellite
because
they
might
be
they
might
be
located
on
different
planets.
Actually,
we
we
are
not
planning
to
locate
them
outside
of
Earth
in
2020,
so
the
youth
is
the
actually
ground
station.
One
get
some
sensory
data
like
temperature
or
similar
encodes
them
into
the
bundle.
J
Then,
eventually,
when
the
satellite
is
in
sight,
it
will
send
it
over
DT
n
bundle.
Protocol.
Seven
satellite
will
not
see
the
canal
station
right
away,
so
we
will
have
to
store
this
bundle
somewhere
and
eventually,
as
it
continues
on
the
orbit,
and
we
will
see
the
connotation
to
it-
will
send
this
bundle.
J
We
are
still
discussing
whether
we
are
going
to
beacon
approach
or
whether
we
are
going
to
use,
or
that
would
be
opportunistic
routing
or
deterministic
that's
under
discussion,
but
for
the
purpose
of
this
presentation
of
what
be
implemented
is
that
we
have
one
docker
container
per
object
that
would
be
ground
station.
One
and
satellite
and
ground
stations
are
running.
Parity,
N
and
satellite
is
running
Micronesian
and
we
have
some
demo
which
we
can,
which
you
can
see
on
the
laptop.
If
you
are
interested,
it
is
really
works
like
slide.
Please
so.
J
We
are
touching
free
free
drafts.
Here,
bundle
protocol
version,
seven.
Then
there
is
TCP
CL
conversion
is
where
we
are
working
with
the
old
version
version.
Well,
not
the
one
discussed
today,
the
previous
version
for
now,
and
we
are
also
using
IP
nd,
actually
there's
a
plan
to
use
IP
nd
code
has
been
written,
but
it's
not
used.
Yet
we
identified
some
issues
there.
J
For
example,
the
eye
pain
in
draft
is
expert,
but
we
would
like
to
see
that
up
to
date,
because
it
would
simplify
testing
of
our
approach
because
as
far
as
I
know-
and
please
correct
me-
if
I'm
wrong
of
we
are
not
aware
of
any
other
way
how
to
how
to
network
discovery.
Name,
sorry,
neighbor
discovery
could
be
done
except
this
already
expired
protocol.
J
So
that
was
the
issue
we
were
trying
to
address
on
the
hackathon
and
our
solution
was
that,
let's
check,
if
the
document
is
still
up
to
date
now
I
mean
IP
and
IDO
and
bring
it
back
to
life.
Maybe
up
to
date
means
like
okay.
It
is
expired
that
something
we
know,
but
maybe
the
information
contained
inside
is
still
valid.
So
that's
what
we
are
trying
to
to
confirm
or
reject
next
slide
please.
J
So
this
is
just
an
information
flight
for
the
hackathon
purposes,
being
basically
improved
our
implementation,
and
you
can
see
our
code
if
you
are
interested
it's
publicly
available
by
DT
ending
in
good
luck.
Next
slide,
please
yeah!
This
is
probably
the
most
interesting
flight
of
this
ball
presentation
of
which
would
be
the
feedback
to
DTN
workgroup
or
we
would
like
to
see
IP
and
it
draft
up-to-date
and
possibly
adopted
by
workgroup.
J
Unless
there
are
some
other
means
how
we
can
do
the
discovery,
because
what
we
are
doing
today
is
that
if
you
want
to
test
cooperation
of
multiple
multiple
implementations-
and
there
are
some,
for
example,
there
is
this.
As
far
as
I
know,
implementation
in
go
language
just
recently
showed
on
the
on
the
email
list,
and
there
is
Java
implementation,
VR
able
quite
easily
to
put
them
into
container
and
communicate
with
them,
but
they
do
not
know
easily
about
each
other,
because
this
IP
n
D
stuff
does
not
work.
J
C
bar
is
used,
so
if
the
workgroup
decides
that
the
IP
any
draft
should
be
brought
up
to
date,
then
probabilities
would
be
one
of
the
things
which
should
be
fixed.
It
should
be
C
or
most
probably
involve
drugs
to
be
consistent,
and
then
there
was
this
recommendation
from
Felix
that
IP
and
D
currently
obviously
depends
on
IP,
but
it
would
not
have
to
because
the
IP
requirement
is
not
kind
of
it
would
be
possible
to.
J
It
would
be
possible
to
extract
that,
so
we
would
have
something
like
not
IP
and
D,
but
D
T,
N
and
E,
which
still
can
name,
but
that
would
mean
I'd
be
VIP
whoa,
but
it
would
also
support
filter
clas,
which
currently
is
not
the
case.
Other
option
would
be
like
what
we
have
today
that
there
would
be
one
one
draft
pair
where
CLA
that's
also
an
option.
So
this
is
something
we
would
like
to
discuss
today
feel
free
to
express
your
opinion
and
that's
it.
Thank
you
for
your
attention.
Thank
you
very
much
guys.
E
Rick,
chair
hat
on
maybe
discovery,
is
an
active
charter
item
and
I
know
about
shortly
before
I
started
as
a
chair,
people
were
looking
at
neighbor
discovery.
I
was
a
proponent
for
I,
agreed
with
Felix
that
I
think
a
DTN
nd
using
clas
and
maybe
event-driven
would
be
a
way
to
go.
I
know
there
are
other
people
in
the
room
Ronald,
for
example,
who
wants
to
go
the
other
way
so
they're
doing
it.
If
you
want
to
volunteer
to
start
working
on
it,
it
would
be
very
welcome
your
way
if.
E
J
We
are
especially
interested
okay.
Well,
it
seems
we
are
in
kind
of
agreement
regarding
first
point
that
we
want
to
have
the
draft
up
to
date
and,
regarding
second
point
possibly
as
well
both
at
this
as
the
end
we
should
be
removed.
Regarding
third
point:
what
do
you
guys
think
I
mean?
Is
it
a
good
idea
to
remove
the
IP
tendency
in
IP
nd
graft
and
call
it
DT,
N
and
D
or
something
similar?
Do
you
think
it's
this.
B
B
J
Know
the
way,
oh,
the,
why
we
are
actually
interested
into
into
this
is
that
it
all
of
us
very
easily
to
test
our
application
level
changes
with
other
implementations.
So
that's
why
we
care
and
well
you
know,
so
you
made
a
point
actually
that
the
TCP
CL
is
gonna,
be
the
only
only
convert
as
layer
yeah.
That's
a
valid
point.
J
N
Hopefully,
have
some
people
in
the
queue
already
yeah
I
was
about
wrong
in
felt
I
was
about
to
say
that
the
only
converts
layer
data
that
we
have
so
far
is
IP
based,
so
I'm
not
saying
that
I,
PMD
or
DTM
E
and
E
should
be
the
only
game
in
town.
You
can
do
other
neighbor
discovery
protocols
later,
but
I
presented
this
in
Buenos
Aires
and
it
was
killed
within
five
minutes
which
year.
N
N
E
Bucher,
there's
nothing
to
stop
us
doing
both
I
see
real
value
in
refreshing,
IP,
neighbor
discovery,
making
progress
on
neighbor
discovery,
particularly
as
there's
interest
here
in
the
room.
Great.
Let's
do
it
if,
in
parallel,
people
want
to
have
a
long
esoteric
argument,
I'm
happy
to
join
the
longer
esoteric
argument
on
the
mailing
list,
and
hopefully
that
might
well
it
probably
won't
be
productive
at
all.
But
it'll
keep
us
all
busy.
I
can
do
one
and
somebody.
S
E
But,
however
Rick
here
there
are
other
stos
such
as
CC
SDS
I
can
see
Scott
standing
behind
you
in
the
queue
whom
who
will
standardize
through
their
own
body,
convergence,
layer,
adapters
and
they're.
Looking
to
the
IETF
for
for
guidance
on
given
the
IETF
owns,
bps
v7
and
is
driving
that
forwards,
then
I.
F
S
You
want
to
put
on
architecture
together
on
this
neighbor
discovery
so
that
you
can
slide
at
different
train
transports
underneath
it
where
IP
is
the
one
we're
doing
here,
and
somebody
else
may
do
it.
You
know
some
other
SDO
may
do
it
in
a
non
IP
fashion.
Great
we
can.
We
can
all
get
together
and
knock
our
lights
out
on
that
one.
But
out
of
this
standards,
body.
E
F
Substrate
for
neighbor
discovery,
but
I
don't
see
any
advantage
in
having
a
proliferation
of
vastly
different
neighbor
discovery
protocols
for
different
environments.
I
think
it
makes
much
more
sense
to
have
a
standard
data
exchange
of
neighbor
discovery,
information
and
and
separate
from
that
connected
through
and
through
an
interface
through
a
well
defined
interface.
The
convergence
layer,
connection,
establishment,
connection
management,
software
that
would
initially
just
do
IP,
but
could
in
the
future
do
something
like
Bluetooth.
G
B
Mark
here
so
it's
great
to
see
mark
as
a
chair
great,
to
see
people
interested
in
the
next
topics,
because
that
were
that,
are
you
know?
Neighborhood
scurry
is
one
and
addressing
architecture,
and
a
few
others
have
been.
You
know
sitting
there
waiting
for
you
know
the
ground.
The
basic
work
to
be
done,
which
is
we're
done
almost
I,
have
to
remind
that.
We
also
have
LTP
as
an
RC
that
was
discussed
here
right
somewhere.
We're
not
only.
J
B
J
One
short
note:
this
is
Martin
Bilka
again
I.
Just
want
short
note
from
what
Scott
said.
Oh,
we
had
a
lot
of
feedback
on
the
hackathon
people
coming
to
our
to
our
group.
Asking
like
I
can
be
used
this
in
IOT
world.
We
have
this
problem
that
actually
we
cannot
see
our
sensor
all
the
time,
so
it
there
seems
to
be
an
interest
from
industry
out
part
of
the
world
of
using
this,
because
right
now
they
do
use
some
proprietary
things.
J
J
B
So
I
took
a
few
I
took
a
few
notes
of
you
know.
Next
steps
so
BP
base
chairs
will
send
the
request
for
propagation.
Our
Shepherd
Fred
would
be
writing
up.
Finally,
TCP
L
TCP
CLV
for
Brian
will
post
last
version.
Quick
change
chairs
will
request,
SAP
application
add
bearing
Shepherd
will
submit
the
Shepherd
right
up.
Bp
sec
had
to
post
the
last
version
chairs
to
request
for
publication
scott
burly
to
write
the
do
the
write-up
em
minimal,
tcp
CL.
B
E
B
B
Anybody
I
have
someone
in
mind
the
same
Shepherd
that
does
the
BP
say
because
it's
kind
of
a
the
two
documents
Scott
do-
and
you
know
that
would
be
a
good
results
to
do
all
this
and
I
guess
probably
at
that
time
would
be
a
good
time
for
chairs
and
the
working
group
to
actually
revisit
that
were
milestones
and
we
actually
started
this
with.
You
know
a
topic
of
interest
which
is
neighbor
discovery
and
obviously
management
would
be
another.
You
know
clear
possible
mice
on
so
we'll
have
to.
B
O
E
It's
probably
worth
getting
through
those
last-minute
panic
comments.
As
we
get
the
documents
out
the
door
and
once
they're
safely
nestled
with
the
iesg,
then
we
can
start
our
discussion.
I'd
also
say
I.
Think
addressing
we've
got
to
start
tackling
addressing
fairly
soon
because
we're
pretending
you
don't
need
to
name
anything,
but
but
you
do.
Q
E
From
rick
one
last
thing
other
than
thank
you
is
a
big
thank
you
to
the
hackathon
guys
and
the
guys
showing
running
code.
You
go
to
a
lot
of
IETF
meetings
where
people
are
talking
about
the
status
of
drafts
and
then
everyone
types
on
their
laptops.
It's
really
nice
to
see
stuff
happening
in
the
working
group
involving
running
code
and
that
kind
of
stuff
I.
Personally,
really
like
that
kind
of
thing.
So
big,
thank
you
to
you
guys.
Otherwise
are
we
done
mark
we're
done?
Thank
you
very
much
guys.
Thank
you!
Blue
sheets.