►
From YouTube: IETF102-CORE-20180716-1550
Description
CORE meeting session at IETF102
2018/07/16 1550
https://datatracker.ietf.org/meeting/102/proceedings/
B
A
Okay,
so
good
afternoon
welcome
to
the
Colt
working
group.
We
have
two
meetings
this
week.
The
chairs
try
to
be
frugal
on
requesting
triangle.
That
probably
was
a
mistake,
so
we
have
a
two
hours
Club
today
and
a
one-hour
lunch
on
Thursday.
Now
the
good
thing
is
that
nobody
is
coming
after
us
in
the
one
hour
slot
in
Thursday.
So
unless
you
care
about
Reds-
and
some
of
us
do
care
about
right,
maybe
we
can
just
have
an
informal
hallway
discussion
in
this
room.
A
As
usual,
we
assume
people
have
read
the
draft,
so
we
will
do
very
little
introduction
here
and
very
much
discussion
of
detail
points
that
we
need
to
reach
closure
on
and
we
are
running
under
somebody,
our
principles
and
since
it's
Monday
I'm
showing
this
slide
for
a
couple
of
seconds.
So
there
are
some
policies
that
apply
here
and
among
those
are,
you
are
being
recorded
number
one
and
number
two.
A
If
you
know
about
a
patent
claim
that
what
we
don't
a
technology
you
want
to
talk
about
here,
you
either
have
to
tell
us-
or
you
can
choose
not
to
talk
about
the
technology,
whether
the
second
part,
knowing
about
a
patent
claim
and
not
talking
about
that
it's
frowned
upon.
But
it's
not
something
we
can
completely
outlaw
here
in
the
procedure.
So
if
you
want
to
read
the
details
of
the
procedure,
there
should
be
somewhere
here,
but
they
aren't
great.
A
A
We
had
a
number
of
last-minute
moves
here
and
even
this
list
isn't
quite
fine
Schmidt,
because
we
cannot
have
the
coma
I
call
crunch
Podge
right
at
the
starch,
because
some
people
are
stuck
in
another
meeting,
so
we
actually
will
after
corn
or
the
traction
will
start
with
Oscar.
Is
there
a
year?
I
hope
you
will
be
within
two
minutes
and
then
we
go
into
a
recess
rectory
and
then
we
can
either
with
an
AmeriCorps
grant
as
the
time
permits
and
on
Thursday.
A
We
have
lots
of
small
items
and
you
can
see
we
have
started
measuring
the
time
here
in
minutes
and
starting
and
stopping
at
1834,
and
things
like
that.
That's
not
going
to
work
in
any
way,
but
okay.
So
there
are
a
number
of
housekeeping
drafts,
for
instance,
that
we
just
need
to
take
off.
There
are
some
other
working
room
drafts
that
always
receive
short
shrift.
It's
Michael
here
not
yet,
and
if
we
miraculously
do
have
time
at
the
end,
we
can
talk
about
a
congestion
control
draft
that
is
being.
A
A
First
of
all,
we
have
a
relationship
with
the
NASD
workgroup,
which
is
meeting
on
Thursday
morning
and
I
understand
one
draft
that
we
will
touch
on
today,
where
we
discussed
on
Thursday
as
well,
and
of
course
there
are
dozens
of
other
humilated
working
groups.
Meeting
this
week
are
always
sending
agenda.
I,
don't
want
to
highlight
all
of
them
to
highlight
psst
yeah.
A
Also
one
thing
I
want
to
mention
is
that
we
have
a
number
of
customers
in
other
SDOs
of
those
tenets
developments,
organizations
and
we
try
to
keep
in
contact
with
them,
and
one
of
those
is
LCS
and
we
actually
will
have
a
coordination
call
with
them.
This
is
usually
an
affair
for
very
few
people.
We
already
have
12
seats
in
the
room
where
we
do
this
or
if
you
want
to
join,
please
ask
the
the
chairs.
A
We
definitely
want
to
accommodate
everybody
who
wants
to
do
that,
but
we
may
not
be
able
to
okay
quick
status
on
two
documents
that
are
in
is
reprocessing,
so
one
of
those
is
links
Jason.
That
is
actually
something
that
started
in
2012
with
zero
variants
added
2015
units,
so
you
would
think
by
2018.
A
It
might
be
time
to
finally
complete
this,
so
this
document
already
left
the
working
group
and
was
submitted
to
the
ASG
more
than
a
year
ago,
but
we
got
a
lot
of
feedback
from
from
the
other
web
related
groups
in
the
IDF,
so
we
decided
to
do
some
refocusing
and
that
refocusing
is
still
ongoing
and
one
reason
for
this
is
that
we
are
finishing
the
work
on
the
resource
directory
right
now
and
we
want
to
make
sure
that
everything
fits
together.
So
we
hope
that
the
people
interested
in
this
will
meet
in
the
hallways.
A
A
The
other
document
that
is
in
is
3
right
now
is
cocoa,
and
here
alexei,
who
just
came
in
had
an
interesting
idea.
Assigning
transport
area
ad
is
the
responsible
ad
and
that
that's
great,
because
we
got
really
good
feedback.
Now,
unfortunately,
the
authors
have
to
fight
back
to
work
on
their
feet.
Very
calluses
sitting
in
the
beggar
or
their
college
has
already
generated
a
per
request,
but
we
have
to
generate
a
new
version.
A
Maybe
we
can
do
this
this
week,
and
so
then
the
normal
process
should
continue,
and,
as
I
mentioned
of
course,
continuing
work
on
congestion
control
may
be
aware
that
there
is
continuing
work
on
the
TCPS
congestion
control
cheese.
It
was
invented
in
the
70s
and
it's
congestion
control
in
the
late
eighties.
So
we
don't
expect
current
to
have
one
final
congestion
control
solution,
and
indeed
there
is
another
draft
out
there,
and
if
we
have
time
on
Thursday,
we
will
discuss
that.
C
D
Little
mints
so
I
think
you're
talking
about
the
right
document,
and
the
point
is
I,
guess
that
the
updates
that
incorporate
your
feedback
should
be
already
in
a
request-
that's
out
there
and
should
be
the
basis
for
the
next
version
of
the
document
which
will
be
0
for
so
we
hope
we
can.
Finally,
it's
the
moment
it
more
explicit
response.
D
A
E
Okay,
so
we're
now
in
version
13
and
the
main
changes
since,
since
the
last
meeting
is
the
clarifications
on
the
further
details
that
was
based
on
the
icy
review
and
another
post
last
call
reviews,
so
in
particular,
in
this
new
appendix
D.
This
was
requested
be
as
the
overview
of
the
security
properties
of
the
protocol.
E
If
you
look
at
the
wiki,
you
find
how
we
have
addressed
the
different
review
comments
so
more
in
detail.
The
observe
option
was
previously
an
entirely
outer
option,
meaning
that
it's
unprotected
for
the
purpose
of
the
proxy
being
able
to
perform
operation,
perform
operations
related
to
observe,
and
now
we
have
an
increased
protection.
So
it's
additionally
an
inner
option
which
enables
the
endpoints
to
verify
each
other's
intent
in
particular
registration,
so
that
that
has
the
the
effect
that
it
simplifies
the
specification.
E
We
don't
need
to
consider
so
much
different
proxy
operations,
but
it
takes
less
support
for
for
proxy
processing.
So
so
that's
one
change
that
we
made.
Another
change
is
that
we
have
separated
out
the
observed,
processing
and
also
the
block-wise
processing
from
the
processing
in
general,
which
simplifies
the
description
I
was
appreciated
by
the
implementers.
E
E
We
also
have
a
new
subsection
on
your
host
port
processing
and
your
a
host
port.
That's
essentially
those
options
which
that's
the
only
after
only
options
remaining
all
the
others
are
end
to
end,
and
we
explained
in
this
option
in
this
section
why
this
option
is,
is
only
outer
only
and
we're
targeting
the
default
case
where
your
eye
host
port
coincides
with
its
default
values
and
therefore
are
omitted,
which
is,
of
course
favorable
both
from
a
privacy
point
of
view
and
from
an
overhead
point
of
view.
E
Then
we
already
got
comments
from
Martin
Thompson
on
HTTP
processing.
Those
were
included
in
an
intermediary
version.
We
have
expanded
on
the
coop
to
co-op
forwarding
proxy
section,
bringing
in
information
about
forwarding
from
other
parts
of
of
the
documents
of
compiling
that,
in
one
section
we
added
one
new
parameter
to
the
security
context.
It's
called
ID
context
and
it's
also
added
to
the
key
derivation,
and
this
parameter
already
existed
as
a
message
field
that
was
transported
in
an
Oscar
message.
E
So
we
transport
context
information
which
is
used
by
the
server
to
decide
to
retrieve
the
security
context,
and
this
this
is
used
both
in
grupo
score
and
in
sixties,
minimal
security.
What
the
change
was
that
we
now
introduce
this
as
parameter
also
in
the
security
context
and
key
derivation,
so
this
can
be
applied
in
a
common
way
for
those
two
applications.
E
So
that's
some
changes
in
this
version
and
from
from
our
point
of
view,
authors,
we,
our
next
steps,
is
we.
We
have
still
some
some
more
review
comments
based
on
this
new
version
from
from
Jim
which
we
like
to
include
and
publish
in
the
new
version,
and
that
version
we'd
like
to
interrupt
test
as
soon
as
possible.
E
A
The
next
step
is
for
Eric
to
review
the
draft
after
the
ITF,
and
then
we
know
where
we
are,
but
of
course,
as
a
working
group,
we
also
want
to
make
sure
that
the
changes
we
are
doing
during
is
reprocessing
what
the
working
group
wants.
So
it's
really
important
that
that
we
stay
on
the
ball.
You
listen.
F
So
yes
I,
we
talked
quickly
yesterday
worth
authors
and
I.
Had
a
quick
look
in
the
morning.
I
didn't
I
think
it's
lots
of
good
changes,
but
you
know
you
also
moved
some
sections
around.
So
it's
not
very
easy
to
diff
and
see
what
exactly
changed,
because
you
know
so
it
looks
like
a
lot
of
stuff
will
delete
a
lot
of
stuff
so
anyway,.
G
F
I
think
probably
the
best
ways
for
me
to
do:
IT
applause
Cole,
just
to
emphasize
that
there
were
changes
to
specific
sections.
You
know
HTTP
description
and
proxies
access,
security
considerations,
and
just
you
know
this
can
all
happen
in
parallel.
Why
Eric
and
I
are
looking
at
so
I'll?
Try
to
start
it
like
this
week,
so
this
will
be
two
weeks,
but
I
think
that
that
shouldn't
delay
the
document
much
further
than
that.
Also
give
the
working
group
opportunity
to
double
check
that
changes
well
done
correctly.
A
H
H
So
essentially
we
replace
the
old
pure
listener
terminology
with
side
server.
Then,
following
all
the
ascore
security
context,
a
zero
mention
now
we're
taking
advantage
of
the
ID
context
defined
in
the
first
place
as
part
of
the
of
the
context
in
the
menos
core
document.
So
now
we
are
just
using
it
here
to
convey
the
the
group
identifier
of
your
score
group
next
slide.
Please,
for
the
co-ceo
objects,
are
just
a
minor,
odd
equation
as
to
how
the
external
IAD
is
formatted.
H
H
Now
we
are
pointing
at
exact
phases
and
steps
of
the
message
producing
in
DOS
enough
score,
so
saying
excitedness
that
that
changes
that
escaped
that
doesn't
happen
anymore
and
so
on.
So
it
should
be
much
easier
to
read
and
follow,
and
in
this
respect,
also
by
the
way
following
another
thing,
we
checked
at
the
land
of
meeting
the
the
group
security
context
now
is
retrieved
only
based
on
the
group
identifier.
H
Next
slide,
please
and
back
to
possible
collisional
group
identifiers.
Again
we
discussed
about
that
in
particular
in
a
pending,
see
when
we
discuss
it
possible
encoding
of
group
identifiers
also
based
on
a
good
discussion
in
the
list.
We
asked
about
that.
Thank
you
very
much
for
feedback.
So
before
we
were
kind
of
seeing
possible.
Collisions
of
group
identify
this
a
big
deal
while
well
in
the
security
considerations.
We
discussed
the
essenti
/
security
at
all
is
not
a
big
deal
in
the
first
place,
also
from
the
from
the
processing
point
of
view.
H
So
you
might
have
some
kind
of
a
beauty
worth
trying
to
retrieve
security
context.
It
might
just
go
for
a
trial
and
error
until
you
pick
up
the
right
one,
so
discouraging
a
collision,
possibly
through
a
convenient
group,
ID,
sighs
sure
good
thing
to
do
as
a
good
practice
to
discourage
a
collision.
But
it's
not
the
end
of
the
world,
it
happens
and
fire.
There
was
some
minor
adjustment
in
in
the
appendix
discussing
well,
the
join
in
part
and
the
possible
retrieval
of
public
is
from
the
group
manager.
H
According
to
another
draft,
India's
working
group
related
to
message
form
and
to
to
perform
this
process,
but
a
further
more
extensive
list
of
updates
is
in
internet'
appendix
G
ones,
who,
by
the
way
next
slide.
Please
implementation
speaking.
We
plan
also
to
start
working
on
a
version
of
this
for
californium
building
on
the
current
L
of
score
and
implementation
already
available.
H
Otherwise,
we
had
already
available
two
implementations,
one
for
other
means
C
and
wanting
contiguous
for
200
platform,
though
they
are
aligned
to
an
altered
version
of
the
draft,
essentially
the
one
before
adoption,
so
they
possibly
like
to
be,
of
course,
aligned
next
slide.
Please
yeah,
just
a
quick
final
remark:
there's
of
course,
a
related
activity
to
this
document
in
India
is
working
group
describing
how
to
join
your
score
group
using
using
ace
for
the
concern
of
this
document.
H
A
One
is,
we
are
finishing
about
documents
right
now
that
do
something
with
groups
and
in
particular,
the
resource
directory
has
a
way
to
handle
groups,
and
one
thing
that
is
entirely
unclear
to
me
at
this
point
is:
does
this
fit
together
in
any
way
is,
is
something
we
done,
and
here
should
we
fix
one
to
work
better
with
the
other?
If
you
have
an
opinion
on
that.
I
J
Jim
shot
I
do
not
believe
that
there
is
any
overlap
in
what
these
documents
are
doing,
but
I
also
have
wondered
for
many
times
who
is
actually
managing
group.
It
groups
in
the
Rd,
because,
basically,
what
you're
looking
at
is,
if
you
have
a
third
party
managing
the
groups
in
the
Rd
that
would
basic
be
the
same
party.
Would
it
be
managing
groups
here
so
talking
about
those
two
things
as
being
the
same
as
interesting,
but
the
entity
is
actually
doing
that
management.
Right
now
is
in
ace,
not
here.
So.
A
If
you
speaker
from
the
stock
or
volunteer
to
collaborate
well
great,
so
we
have
one
person
who
is
not
only
interested
but
actually
wants
to
do
work
on
this.
We
had
other
people
like
German,
Michael
and
a
few
other
nodding
heads
here.
So
what
should
we
take
down
in
the
minutes,
as
the
people
who
investor
to
make
progress
on
this.
A
Okay,
so
that
was
one
question.
The
other
question
is
more
on
a
procedural
side,
so
it's
most
likely
that
the
three
documents
that
work
together
to
do
this
will,
at
some
point,
be
a
cluster
in
the
RFC
editor
queue.
So
what
is
the
the
relative
speed
with
which
we
are
finishing
these?
What
is
what
is
our
projection
when,
when
we
are
done
with
it,
jim
has
no
I've.
J
J
H
J
A
E
E
E
A
E
A
L
A
So
did
the
plan
would
be
I
mean
it's
summer
anyway.
So
there's
nothing.
Nothing
really
happening
right
now,
but
the
plan
would
be
to
get
the
oscar
version
to
a
near
approved
state
or
hopefully
to
an
approved
state
and
then
based
on
that
run
another
Interop
and
get
a
score
to
come
towards
the
last
one.
Okay,
great
wonderful,
so
anything
else.
We
want
to
do
on
a
scroll.
G
E
E
A
G
E
A
look
at
slides
feel
free
to
come
back
with
comments
or
questions
so,
but
on
the
message
sizes,
we
are
now
doing
half
the
message
size
of
stls
for
our
public
keys
and
the
third
of
the
total
number
of
met
bytes,
for
we
share
he's
authenticated
diffie-hellman.
So
that's
the
type
of
estimate
optimizations.
We.
E
Would
ask
John
if
he
has
any
views
on
that,
but
I
we
haven't
had
any
come
much
comments
on
this
version,
so
it
probably
wait
a
little
but
from
an
optimization
point
of
view
and
from
a
security
point
of
view,
I
think
we
are
close,
I
mean
the
security
is
not
necessary
for
implementation,
but
for
finalists
is
finalizing.
The
protocol
I
think
we're
close
okay.
A
A
Okay,
thank
you.
Okay,
that
should
finish
the
oscar
cloud,
but
we
actually
10
minutes
ahead
of
you
know
slightly
confused
in
general
right
now,
oops,
so
the
next
one
on
the
agenda
is
re.
No,
we
cannot
do
Cochran
yet
because
when
one
person
is
missing,
so
that
would
be
a
little
bit
later.
So
let's
talk
about
resource
directory
and
resource
directory
DNSSEC.
You
don't
think
you
wanted
to
talk.
I
Okay,
I
want
to
talk
a
little
bit
about
the
resource
directory
and
about
all
the
modifications
and
improvements
which
have
been
done
with
respect
to
the
former
two
versions.
Actually,
what
we
have
been
done
and
what
we
should
look
at
in
detail,
is
that
there
is
an
appendix
B
and
D
down
in
the
resource
directory,
which
explains
in
more
detail
how
you
go
from
reference
to
a
target
URI
and
how
you
can
use
that
and
how
that
distort
in
the
RT
and
how
you
can
actually
do
the
lookup.
It
says
Oh
nicely
explained
with
examples.
I
For
example,
it
tells
you
where
the
see
the
16
Authority
part
where
that
is
used
and
how
that
is
prefix
to
the
relative
reference.
It
tells
you
about
RFC,
80
to
88
and
66
19
and
its
difference
and
how
we
have
made
rules
to
make
sure
that
we
do
not
go
into
conflict,
and
it
tells
you
also
about
the
host
relation.
How
that
fits
into
the
whole
thing.
I
I
You
then
construct
you
do
a
new
sort
of
resolution
from
the
relative
reference
which
you
see,
the
team
is
in
the
brackets
and
that
man
goes
to
the
go
up,
so
he
can
to
the
authority
scheme
before
and
then
you
get
the
team,
which
then
is
the
target
you're
right
that
you
go
produce,
and
so
the
whole
thing
is
going
through
in
the
appendix
B
and
explains
where
it
is
loops.
Also,
you
see,
for
example,
we
have
your
relative
anchor
and
also
that
is
resolved
against
the
base.
Uri
and
you
can.
I
Short-Term
memory
yeah,
and
so
you
see
how
their
anchor
is
constructed
there.
The
other
point
is
that
the
link
context,
which
is
used
in
the
relation
attributes,
how
that
is
also
formed
exactly
because
there
were
some
conflicts
between
the
two
drafts
about
how
that
is
done,
the
further.
We
have
to
be
sourced.
Look
what
and
use
the
resource
good
always
returns
an
absolute
target
and
the
absolute
target
is
constructed
from
the
registration
base
URI,
which
actually
is
the
authority
and
the
scheme
and
the
authority
from
which
we
have
them.
Post
fix
the
relative
reference.
I
Please
beat
the
appendix
well,
we
have
a
registration
base
URI,
and
that
is
known.
That's
the
base
year
of
mr.
velman
court
I
should
not
say
strip
not
constructed,
and
that
one
is
used
to
than
to
get
the
absolute
URI,
which
you
see
this.
The
further
the
base
URI
when
you
do
with
a
third
party
registration,
it's
the
one
which
can
be
defined
with
the
base
is
basis
parameter
in
the
in
the
registration,
so
that
one
is
stored
in
the
resource
directory,
the
registration
base
URI
and
then
the
look
look
up.
I
I
The
anger
is
used
as
okay
in
1606
90
I
can
see
it
better.
Thank
you.
It's
useless
and
busier
in
which
the
relative
target
is
down,
but
in
82
a
TV
actor.
The
anchor
is
not
actually
used
for
the
resolution,
so
that's
indifference
and
also
in
the
16th
a
result
anchor.
The
context
is
the
target.
Ori
is
the
past
trip,
though
otherwise
the
cover
the
complex
is
given
by
the
base
URL.
I
So
what
we
have
been
trying
to
do
is
to
make
sure
that
those
conflicts
do
not
appear
in
the
resource
directory
and
that
we
have
some
rules
to
make
sure
that
you
do
not
cut
into
a
problem.
Chief
yep
then
other
improvements
to
the
Rd
text.
So
we
have
this
domain,
which
was
very
confusing
because
it
used
elsewhere
as
well,
quite
a
lot
and
we
called
it
sector,
but
we
maintained
the
Deus
parameter
because
of
backward
compatibility.
The
cones
mr.
Colt
context,
which
is
not
in
context
because
that's
something
else.
I
I
The
simple
registration
has
been
not
changed,
we
borrowed
it
and
also
the
examples,
etc
have
been
done
more
correctly,
the
lookup
we
show
how
they
resolve
references,
I,
actually
returned,
and
the
LT
that
is
the
lifetime
is
not
exposed
in
the
lookup
anymore
because
it
gives
ambiguous
results
is
at
the
lifetime
as
we
specified,
or
is
it
the
lifetime
which
is
still
left
or
is
the
lifetime
which
has
been
confused,
consumed
left
that
out
and
the
update
of
the
registration
has
been
I?
Think.
I
Doot-Doot
pennies
thanks
for
the
reviews,
especially
Jim,
very
much
appreciated
them.
We
still
have
refused,
probably
which
come
in
which
we
shoot
reacts.
We
still
have
some
ambiguous
texts
left
and
right,
so
we
are
working
on
that,
but
we
singing
that
you
are
really
because
we
think
that
the
most
structural
things
have
been
solved
and
that
it's
more
about
clarifications
and
getting
the
terminology
correct,
etc,
and
we
think
that
this
typically
working
coconuts
gold
coins.
J
J
J
A
We
actually
did
start
a
little
bit
of
interesting.
It
will
continue
at
this
week
now.
What
happened
to
me
was
I.
Had
a
student
write
this
thing
and
he
intended
something
in
that
I
didn't
take
seriously
because
seven
hundred
lines
of
code,
but
it's
the
implementation.
So
if
you
want
to
provide
a
resource
directory
implementation,
it's
not
that
complicated.
Oh
I've.
A
A
We
have
some
four
or
five
implementations
we
could
carry
together.
One
thing
that
we
forgot
to
put
on
the
slides
is
that
there
is
this
project
called
F
Interop
that
is
happening,
put
together
online
interrupts
of
IDF
protocols,
and
so
maybe
we
can
set
up
something
with
them
in
a
September
timeframe
to
actually
do
this
intro.
So
what
would
that
be?
A
convenient
time
frame
for
people.
N
N
Okay,
pretty
small
number,
so
you
might
be
interested
to
know
what
we're
doing.
But
first
we
could
talk
about
the.
Why
simply
the
Corps
Charter
states
that
we
should
interoperate
with
DNS
discovery
in
some
manner,
and
so
it
seems
to
me
that
at
least
to
the
level
of
restful
api
is,
we
should
be
able
to
be
able
to
export
those
from
the
Rd
into
DNS,
and
this
is
called
DNS
based
service
discovery.
N
The
use
case
is
I
think
the
primary
one
will
be
in
mixed
environments,
where
you
have
primarily
HTTP
clients
talking
to
collapse
through
middle
box,
which
is
called
the
cross
proxy
I
worked
on
a
standard
several
years
ago,
which
is
actually
the
genesis
of
this
proposal
was.
It
could
be
smart
energy
profile
that
is
now
in
I,
Triple,
E
standard
and
it's
beginning
to
be
deployed
in
parts
in
Hawaiian
California
for
solar
energy.
So
we
do
anticipate
that
these
there's
any
more
of
these
environments
over
time.
N
The
granularity
of
DNS
SD
is
larger,
so
it
allows
you
to
basically
search
for
services
services
such
as
who
provides
temperature
who
provides
light,
and
so
once
you
have
a
collection
of
instances
that
support
that
service,
you
can
then
drill
into
them
and
find
out.
What
are
you
know?
This
is
one
that
you're
interested
in
and
it
also
could
potentially
be
used
for
the
chicken
and
egg
problem
of
locating
our
DS
in
the
first
place.
N
N
So,
if
you're
looking
for
printers,
for
example,
the
DNS
database
that
you
are
querying
will
have
several
pointer
records
that
have
pointer
records
named
ITP,
for
example,
which
is
the
internet
printing
protocol
you'll
get
back
when
you
do
the
query
for
IP
P
pointer
you'll
get
back
a
list
of
instances
which
will
then
populate
view.
Mac
users
are
very
familiar
with
this
when
they're,
adding
printers
and
so
forth.
N
Those
instance
names
then
correspond
to
a
serve.
Ns
are
being
a
text
record
pair.
The
SRV
record
will
give
you
the
host
name
and
the
port
where
that
service
can
be
located.
The
text
record
will
give
you
additional
information
about
that
service
in
the
form
of
key
equals
value
pairs.
So
if
it's
a
color
printer,
for
example,
you
would
see
information
about
that
in
the
text
record.
N
So
the
SRV
record
provides
you
with
a
hostname.
You
would
the
hostname
finally
to
an
A
or
quite
a
record,
and
what
I
should
say
is
that
the
service
type
that
you're
searching
for
when
you
do
the
Pointer
search
actually
is
a
stand-in
for
the
protocol
that
you're
using
to
communicate
with
that
service.
So
it's
a
bit
like
a
scheme
in
a
URI
in
that
sense
that
it
encodes
a
lot
of
information.
N
In
a
simple
token,
so
you
know
a
priori
the
kinds
of
services
that
you're
searching
for,
and
you
have
the
protocol
that
will
that
will
just
you
know,
talk
to
those
services
again,
I
think
we're
used
to
talking
about
resources
and
rests
I.
Think
if
you
squint,
you
can
see
that
anything
that
exposes
a
restful
api
basically
fits
the
definition
of
a
service
and
so
can
be
exported
into
dns
SD.
N
Directory
but
potentially
in
the
future,
we
could
be
exploring
resource
directory
information
to
other
types
of
directories
as
well.
So
that's
TBD
instance,
name
as
we
define
it
here
is
thought
to
be
global,
unique
across
all
devices
of
this
type.
In
you
know
the
given
sector,
but
maybe
that
constraint
is
not
easily
met.
So
that's
something
for
us
to
visit
where
you're
using
RT
and
I
F,
which
are
already
defined
in
other
rfcs
or
proposals.
N
So
let's
talk
about
how
these
mappings
work
so
inss
becomes
the
instance
part
of
the
dns
SD
service
name,
as
I
said,
resource
type
may
have
to
be
converted
from
existing
resource
type
into
something
that
conforms
to
the
format.
That's
used
by
DNS,
SD,
probably
just
say
something
about
that
format.
Here:
service
types,
there's
a
registry
in
I
Anna,
the
guiding
RFC
is
1660
335.
These
tokens
are
they
basically
conform
to
the
LVH
syntax,
so
they
can
be
comprised
of
letters
digits
or
hyphens
and
they're
15
characters
in
length.
N
So
if
you
look
at
the
RT
registry
right
now,
some
of
those
resource
types
exceed
that
length
limit
and
currently
all
of
them
seem
to
have
periods.
So
we
can't
just
do
a
one
to
one.
You
know
can't
just
use
resource
types
as
they're
currently
defined
in
the
in
the
resource
type
registry.
So
an
agent
that's
going
off
and
doing
this
export,
maybe
there's
some
sort
of
heuristic
that
can
be
used,
replaced
the
periods
with
hyphens
or
something
along
those
lines.
So
that's
that's
some
work.
That's
you
know
needs
to
be
discussed
in
the
group.
N
So
here's
an
example,
it's
kind
of
color-coded
to
show
how
an
agent
might
do
this
mapping
so
the
first.
It's
basically
the
process
kicks
off
by
querying
the
resource
directory
or
you
know
using
another
method
to
basically
find
so
you
can
see
as
part
of
the
query
string,
we're
looking
for
this
exp
attribute.
So
again,
that's
set.
If
you
want
your
information
to
be
exported,
we
get
back
a
sample
record
here.
From
from
this
node
FD
FD
1
2,
3
4.
It
says
that
it's
a
got
a
temperature
sensor
available
at
that
path.
N
The
instance
name
is
set
to
indoor
temp,
and
so
then
you
can
see
what
the
resulting
resource
records
are
anywhere.
Where
you
see
an
ellipsis
there,
that's
just
a
stand-in
for
example.com
dot,
so
I
just
deleted
that
and
shorten
it
up
a
little
bit.
But
you
can
see
this
mapping
so
in
particular
what
began
life
as
the
resource
type
you
know,
has
to
be
sort
of
massage.
N
You
know
to
fit
into
the
resource
type
that
is
acceptable
to
or
the
service
type
that's
acceptable
to.
Dss
D
I
think
one
other
thing
comes
up
to,
and
that
is
whether
we
use
sort
of
like
a
hierarchy
and
search
like
a
particular
SDO
like
oh
I,
see
four
subtypes
or
whether
that's
just
one
long
flat
string
and
as
it
as
we
discussed
in
the
draft
that
can
be
done.
One
of
you
know
one
of
two
ways
so
I
think
one
advantage
of
doing
it.
N
This
way
where
the
main
type
is
oh
I,
see
or
some
other
sto
name,
is
that
that
token
could
potentially
be
used
to
express
more
about
your
protocol
bindings.
So,
for
example,
if
your
if
you've
got
multiple
bindings
in
your
definitions,
WebSockets
or
other,
if
you've
got
a
number
of
different
ways
to
express
the
binding
who
could
encode
it
in
the
in
the
type
and
then
encode
the
actual
function
that
you're
providing
in
the
subtype
so
I
think
that's
it.
Questions.
N
I
should
say
to
that
I
think
this
draft
is
this.
This
version
is
pretty
readable
from
here
out
I'm,
going
to
be
spending
a
lot
more
cycles
on
this
and
we'll
be
tracking
changes
and
so
forth
as
we
go
on.
So
the
other
thing
I
would
ask
is
that
you
follow
up
any
proposals
or
suggestions.
You
have
on
the
list
and
we'll
make
sure
we
get
those
incorporated
in
the
next
version,
and
then
the
final
thing
I
want
to
say
is
following
on
to
the
interrupt
for
the
resource
directory.
N
O
So
Dave
Taylor,
so
Stuart
and
I
were
having
a
little
side
conversation
over
here
that
maybe
he's
getting
up
to
come
into
the
same
thing
about
the
so.
The
a
question
in
the
bottom
part
is
around
the
area
where
you
have
you
know
our
temperature
and
so
on,
and
so
my
understanding
from
the
discussion
we're
having
is
that
that
string
right
now
must
be
ini
registered
in
order
to
be
usable
in
an
hour.
O
P
Now
we
spend
a
lot
of
time
debating
how
long
identifies
needed
to
be
and
came
to
the
conclusion
that
15
characters
was
long
enough,
that
it
wasn't
a
scarce
namespace.
So
we
won't
can
be
fighting
about
allocating
them,
but
short
enough
to
still
be
efficient
on
the
wire
and
cool
which
is
supposed
to
be
targeted
at
the
same
constrained
devices
that
DNS
SD
was
allows
arbitrary
length
and
and
very
verbose
resource
types
which
makes
the
package
much
bigger
than
DNS
SD
packets,
I,
sometimes
paraphrase.
O
Okay,
so
thank
you
for
the
clarification
Stewart,
so
I
think
that
the
discussion
is
actually
about
the
underscore.
Oh
I,
see
dot
field.
I
think
that
the
semantics
of
RT
values
is
that
a
piece
of
code
is
not
supposed
to
interpret
dots
as
being
hierarchy
or
something
in
there.
And
so
the
question
is,
if
underscore
oh
I
see,
must
be
IANA
registered.
O
O
A
First,
the
general
comment:
we
are
increasingly
seeing
situations
in
where
we
have
different
INR
registries
that
that
need
to
be
in
sync
on
some
registration
activity.
That's
going
on
and
first
day,
I
have
a
quick
slide
on
the
draft
about
proactive
registration
of
content
format,
numbers
for
media
types.
So
that's
a
similar
situation.
G
A
A
Now
one
question:
if
we
take
resource
types
which
are
defined
in
RFC,
66
90
and
turn
this
into
a
hierarchical
namespace
might
be
what's
the
hierarchy
and
and
as
they've
mentioned,
it's
not
just
a
COS
that
defined
these
things.
So
how
what's
our
perception
of
that
hierarchy?
What's
our
guidance
for
how
the
hierarchy
should
be
used
and
I?
Think
that
will
require
some
thinking
and
we
would
then
go
ahead
and
do
the
editor
to
260
690,
whether
we
do
this
in
links
JSON
or
in
a
separate
raft.
A
I,
don't
know
that
defines
how
exactly
we
plan
to
work
with
resource
types
in
the
future
and
that
might
also,
for
instance,
contain
guidance
that
they
have
to
fit
into
78
bits
and
so
on.
So
we
can
try
to
address
that
that
problem
there
as
well,
but
of
course
it's
already
there
in
registry,
so
we
cannot
solve
it,
but
we
can
just
invest.
N
Please
please
comment
on
the
mailing
list.
You
know
in
the
worst
case,
this
agent
or
whatever
it
does
that
that
does
the
you
know,
the
mechanical
exporting
from
the
Rd
into
DNS
sd
would
have
to
have.
You
know
heuristics
for
every
party
that
it
encounters
or
groups
of
artis,
which
would
be
you
know.
Obviously,
now
at
that
point
the
thing
starts,
looking
like
an
application
layer
gateway
which
is
nasty,
that's
we're
trying
to
get
rid
of.
P
So,
if
I
have
an
app
that
will
show
me
temperature,
it's
looking
for
temperature
sensors,
if
we
could
imagine
I
have
an
app
that
wants
to
find
all
devices
made
by
Samsung,
regardless
of
what
that
device
does.
It
might
be
a
washing
machine,
it
might
be
a
printer,
it
might
be
a
thermostat
I,
don't
care
if
it's
made
by
Samsung
I
want
to
find
it.
I
have
a
hard
time
imagining
what
that
app
would
do
that.
N
Just
gonna
before
Kirsten,
yes,
I
just
want
to
mention
that
we
may
have
an
additional
problem.
I
mean
in
the
early
days.
Co
app
was
only
bound
to
UDP
right,
but
now
we're
talking
about
TCP
bindings,
WebSocket
bindings,
all
the
rest
of
that.
So
all
that
has
to
be
expressed
in
the
service
instance
name,
because
at
the
end
of
the
day
you
have
to
have
a
complete
stack.
That's
going
to
talk
to
the
server
on
the
other
side,
so
an
additional
thing
we
might
want
to
think
about
is
deviating
is
potentially
expanding.
N
A
Q
Q
Originally,
it
was
I
think
proposed
by
Zack
and
it
was
used
to
distinguish
resources
within
the
same
end.
Point
that
had
the
same
resource
type,
so
I
have
some
temperature
resource
type
and
I
want
to
say
that
yeah
this
resource
is
for
indoors
and
the
other
one
is
for
outdoors
or
I
had
the
car
before
tire
pressure
sensors.
Q
This
is
which
one
is
it,
and
if
this
redefinition
for,
for
the
instance
from
in
SSD,
we
lost
this
capability
and
also
others
started
to
use
the
ins
attribute
for
their
purposes,
because
it
wasn't
so
well
behind
before
so.
Overall
I
think
we
have
to
rethink
a
bit
the
discovery
framework
that
we
started
to
build
with
the
link
attributes,
because
it's
a
bit
fragile
now
and
people
started
to
use
it.
There
was
no
definition.
We
lost
a
meaning,
so
yeah
and
there
was
not
much
feedback
on
the
mailing
list.
N
Q
So
that
was,
it
was
never
yet
formally
defined.
It
was
more
that's
back
in
the
days
when
we
implemented
it
and
used
it
and
remember
I
think
it
was
part
of
the
interface
draft
that
has
a
long
history,
a
lot
of
changes
to
it
and
so
on.
So
so
that
was
there
and
people
started
using
it
because
it
made
sense
its
solves
a
problem
that
is
out
there
and
now
it's
gone.
Q
I
I
think
it's
good
that
we
now
have
a
clear
definition
and
it
should
be
used
for
that
because
yeah,
it
makes
sense,
but
it
needs
a
gap.
So
it's
just
want
to
point
out
that
you
might
have
a
problem
there
because
kind
of
the
overall
how
to
use
web
linking
and
his
attributes
for
kind
of
the
the
the
80%
of
Discovery
cases.
How
to
do
that
so
I
know,
for
instance,
other
stos,
like
fair
hair,
they
looked
into
the
the
ins
attribute.
Now
it
has
a
different
meaning
yeah.
N
I
think,
in
a
situation
where
people
are,
you
know,
sort
of
using
something
before
it's
defined,
and
so
then
we
have
to
sort
of
you
know
post
facto
go
back
and
change.
You
know
make
the
definition
fit
with
the
with
the
use.
You
know.
The
point
is:
is
that
if
Rd
basically
had
a
section
and
defined
every
one
of
these
meta
attributes
that
we
anticipate
using,
then
there
would
be
like
one
source
but
I
think
right.
Now
we're
sort
of
defining
these
these
meta
data
attributes
as
we
go
along
or
find
a
need
for
it.
N
Q
A
N
Q
So
that's
that's
also
fine,
then
take
this
and
define
it.
How
this
looks
like
in
web,
linking
that's
all
I'm,
saying
and
also
I,
don't
think
we
have
a
problem
with
interoperability
with
people
what
they
implement
it
because
they
didn't
care
and
everyone
implemented
something
different.
So,
let's
throw
this
away
because
it
doesn't
help
interoperability
anyway,
so,
but
we
have
to
give
them
something,
so
this
would
be
the
right
way
that
it
becomes
interoperable.
Ok,
thanks.
P
P
If
I
query
a
resource
directory
for
temperature,
sensors
and
I
get
a
list
that
says
bedroom
kitchen
outdoors.
That
sounds
like
exactly
what
I
want.
I
mean
anybody
with
an
iPhone.
Here
you
tap
the
air
print
button
and
you
see
the
terminal
room
printer
and
you
see
the
registration
printer.
You
see
a
list
of
named
entities
that
describe
something
about
what
their
purpose
is.
So
how
is
indoor
temperature
outdoor
temperature
different
to
terminal
room,
printer
and
registration
printer?
They
both
sound,
like
instance,
names?
Maybe.
Q
Let's
go
to
the
car
example,
so
you'll
have
four
tire
pressure,
sensors
and
they
are
front
left
front
right
and
so
on
and
they
are
defined
within
this.
This
endpoint,
because
you
could
also
register
a
lot
of
cars
and
then
you
have
the
front
left
at
the
front
right
a
lot
of
times.
Also,
if
you
register
your
printers
and
you
have
a
lot
of
bedroom
printers
for
some
reason,
you
need
kind
of
okay
where
to
whom
does
this
printer
belong?
And
then
you
get
some
notion
what
this
meaning
is.
P
Q
Which
so
one
thing
is,
I
cannot
answer
this
completely,
how
people
implement
it
that
one
way
that
that
does
the
trick
is
that
you
have
it
scoped
within
the
EP,
and
the
EP
has
some
other
information
like
the
location
or
the
the
holder
of
the
car
and
in
this.
But
this
information
to
distinguish
the
the
oboe
group
of
est
Services
is
a
deaf
level.
That's
where
it's
annotated.
P
Q
So
it
was
written
down
like
the
ideas
of
that
were
written
down
in
the
early
versions
of
the
interface
craft
and
so
on,
and
people
started
using
that
and
they
got
a
good
idea
how
this
works.
It's
also
used
in
in
replicating
itself
in
in
some
other
ways,
and
but
it
really
points
down
to
the
problem
that
I
want
to
raise
that
yeah.
Q
P
So
it
sounds
to
me
if
I'm
understanding
this
right,
that
inclusion
is,
if
there's
a
standard
place
in
the
resource
directory
with
some
string
to
disambiguate
the
instances.
Then
maybe
you
construct
to
the
DNS
instance
name
from
that
standard
string,
plus
the
ins
yeah,
and
if
resource
directory
doesn't
have
a
way
of
differentiating
instances,
then
maybe
resource
directory
needs
to
add
that,
because
otherwise
it's
not
very
useful
yeah.
O
Dave
flavor
different
topic
model,
it's
related
to
one
tea
bag,
ocf
doesn't
use
it,
but
I
believe
that
RFC
66
90
allows
arbitrary.
You
are
eyes
in
an
r/t
value,
not
just
Iona
si
names.
Did
you
look
at
how
to
map
those
it's
in
are
T
values.
You
are
eyes
yep
allowed
in
66,
90
and
you're
talking
about
mapping
leap
format.
So
is
there
a
way
to
map
them
that
you
look
at
that,
did
not
you
should
at
least
say
something.
Even
if
you
say
sorry,
you
can't
net
those
okay.
N
Could
you
follow
that
up
on
the
list,
because
I'm
sure,
if
I
get
some
comments,
I'd
be
particularly
interested
to
know
whether
anybody
is
using
your
eyes
today
for
our
T's?
It
doesn't
seem
like
there's
much
in
the
RT
registry
other
than
what
oh
I
see
is
put
in
there.
I,
wouldn't
assume
that
there
may
be
people
working
on
it
that
having
gone
public
with
what
they're
doing
yet
the
point.
R
O
A
You
Carrie
so
I
think
it's
really
interesting,
because
we
are
now
seeing
a
number
of
things
that
are
happening
on
top
of
co-op
that
seem
to
flow
together
in
an
interesting
way.
So
we
had
this
semantic
interoperability
work
over
the
weekend.
We
have
the
resource
factory
where
we
have
this
DNS
SD
mapping,
and
it
seems
this
is
really
a
good
time
to
put
in
some
some
major
fence
posts
on
where
we
want
this
to
go.
A
J
A
Schema
or
something
like
that,
so
that's
netcrunch,
and
that
that's
a
family
of
documents
that
that
work
together
and
more
recently,
people
found
out
that
maybe
doing
some
of
this
over
HTTP
would
be
a
great
idea.
So
they
invented
another
bundle
of
documents
that
are
highly
related,
which
are
called
rest
country
and
what
they
are
doing
is
they're
using
HTTP
as
their
way
of
getting
data
from
A
to
B.
A
So
that's
what
what
happened
a
while
ago
and
then
we
started
something
that
we
originally
called
call
conch,
but
that
new
has
kind
of
gotten
lost
partying
all
this
over
into
the
constrained
space,
and
instead
of
using
XML
or
JSON,
using
C
board
to
represent
the
yang
using
coop
and
mapping
the
the
rest
conf
data
types
to
ones
that
that
work
well
in
coop
and
that's
the
document.
That's
called
a
commit
and
a
third
thing,
getting
rid
of
the
the
giant
identifiers
that
come
from
using
XML
or
JSON
based
yang
and
replacing
them
by
SIDS.
A
A
So
that's
the
the
state
of
the
whole
way
I
should
add
that
these
technologies,
of
course,
don't
have
to
be
used
in
exactly
those
bundles.
So,
of
course,
you
can
use
Yangtze
Bo
with
rest
country.
Lengthy
Bo
got
really
great
over
HTTP,
so
it's
not
as
clear-cut
as
this
nice
marketing
slide
here
is
suggesting,
but
for
people
to
understand
a
certain
kind
of
evolution
or
certain
kind
of
pairing
and
mixing
and
matching
of
things.
A
A
S
Thanks
Carsten
after
this
introduction,
I
think
that
I'm
going
to
go
pretty
fast
over
over
the
following
slides.
So
this
is
an
update
of
the
work
that
we've
been
doing
for
the
past
of
months
and
I
see
that
the
table
actually
did
not
go
really
well
here,
but
it
doesn't
matter
so.
Basically,
this
is
the
the
status
of
the
different
drafts
and
okay.
You
cannot
see
it,
but
the
main
drafts
are
the
core
Yanks
koryaks.
S
If
your
corset
and
kokum
I
and
we
have
received
so
there,
these
are
you
see
the
versions
of
the
respective
respective
drafts
and
most
so
the
first
three
are
now
I
mean
we
need
to
do
some
minor
changes,
because
we
had
some
input
on
the
mailing
list
which
we'll
do
in
one
week
after
the
IGF
and
then
they're
ready
for
what's
called
Moscow,
okay,
so
yeah
and
that's
the
cool
thing.
So
here
you
can
see
the
the
protocol
stacks,
which
Carsten
explains
so
you
can
have
them.
S
You
know
for
your
to
two
consults
after,
if
you
wish
and
the
net
compressed
convent
Corinth
and
what
we
have
today.
Well,
we
have
seed
registry,
an
example,
city
registry,
that
you
can
go
and
you
can
use
today
that
is
on
the
website,
combine
that
space.
There
are
already
some
modules
that
are
there,
so
you
can
just
read
them
if
you're
curious
to
go
and
see
how
it
how
it
works.
We
have
existing
implementations,
so
at
least
two
that
we
that
we
know
and
two
more
that
we
have
talked
to
some
people
privately.
S
So
in
go
long
and
in
C
and
on
the
interoperability
side
we
had
already
we
had
three
hackathons,
so
the
first
one
was
like
a
virtual
interrupt
where
we
we
did
some
basic
functionalities
between
the
private
private
implementations.
The
second
one
was
the
past
hackathon,
where
we
were
working
on
semantic
interoperability
over
using
combined
with
the
resource
directory
and
the
work
of
things,
the
things
descriptions
and
on
this
hackathon.
Actually,
we
started
so
in
the
meantime,
we
published
one
at
one
reference
implementation
to
the
F
interrupt
platform,
so
we
can
go
to
this
platform.
S
Anyone
here
can
go
to
this
platform.
It's
a
pretty
well
documented,
so
we
can
download
an
agent.
It
forms
like
a
VPN
network.
So
we
can
really
interact.
You
have
a
Co,
my
server
and
Akamai
quiet.
So
we
can
do
this
today
if
you
wish-
and
yes,
so.
This
is
what
we
did
and
during
this
hackathon,
actually,
what
we
wanted
to
achieve
was
to
to
have
to
start
an
open-source
implementation
of
core
called
client
and
server
starting
by
the
client,
and
this
is
what
we
started.
S
S
They
have
open
sourced
a
thing
that
is
called
a
yang
development
kit
ydk
that
actually
allows
you
to
generate
code
based
on
the
yang
module,
so
you
have
the
gang
module
and
then
you
can
generate
code
in
Python
in
CEO
Ringo
and
we
wanted
to
so
what
our
goal
is
and
if
this
will
continue
on
for
the
next
IDF
is
to
have
comma.
Oh
sorry,
Cole
conf,
a
cork
on
client
that
uses
this.
So
basically,
what
you
do
is
you
generate
your
code
and
then,
if
you
want
Usenet
code,
it's
already
there.
E
S
Are
adding
core
comp
so
it
works
out
of
the
box
and
you
have
all
the
nodes
that
we
took
or
on
the
theater
pad,
and
there
is
you
know,
a
github
repository
for
this
on
the
release
of
the
document.
So
a
young
young
seaboard
document
and
the
seed
document
were
reviewed
by
two
people
from
the
from
the
net
not
and
they
published
seven
discussions
and
some
suggestions.
Some
are
quite
minor
and
four
fix
them.
No
problem.
There
is
one,
that's
a
little
bit
more
significant
and
it
is
related
to
the
encoding
of
Delta
Delta
encoding.
S
So,
to
give
you
an
example,
what
we
have
today
in
the
Peter
that,
but
is
very
happy,
and
so
what
we
have
is
we
really
went
down
to
going
having
the
most
efficient
representation
that
we
can
have
on
the
wire
and
that's
what
is
it
today
and
it
is
the
upper
example
and
in
this
example,
is
you
know
when
we
have
when
we
have
the
requests?
We
have
already
the
seed
that
can
be
used
as
the
the
basis
as
the
the
point
to
which
all
delta
are
old.
S
Elders
are
calculated
so
if
here
in
this
example,
we
have
so
this
a
five
is
the
encoding
of
17:21.
So
that's
as
the
example
and
so
here
in
the
response,
the
delta
states
are
calculated
relative
to
this
entry
point,
so
so
in
order
to
get
absolute
Delta,
you
need
to
add
two
to
this.
To
this
entry
point,
and
one
of
the
one
of
the
inputs
that
will
be
received
is
that
that
could
be
a
little
bit
more
confusing
in
some
cases,
and
it
would
be
really
nice
to
have
it
in
explicitly
in
in
the
payload.
S
So
that
means
that,
even
though
we
have
it
here
in
the
request,
we
have
that
you
know
we
are
requesting
this
this
element.
We
can
also
include
it
once
again
in
the
payload.
So
this
way
we
remove
any
ambiguity
and
it
was
go
on
make
it
so
in
terms
of
the
cost
that
this
for
this
change
would
cost
us,
of
course,
when
introducing
text
radiations.
S
But
so
the
existing
case
is
more
compact,
but
it
would
require
potential
little
bit
more,
like
complexity
on
the
car
on
the
curves,
and
it
could,
you
know,
leads
to
some
ambiguities.
As
we
said.
The
other
point
is,
you
know,
easier
to
debug,
maybe
a
little
bit
more
straightforward
to
process,
given
that
we'll
be
losing
around
four
or
five
bytes.
Just
to
add
this
listing
explicitly
in
the
payload
arson,
yeah
customer.
A
Learn
from
the
floor,
I'd
like
to
fix
your
wording
a
little
bit.
There
was
no
ambiguity
in
the
existing
proposal,
but
we
missed
one
requirement,
which
is
the
requirement
that
you
can
rip
these
payloads
out
of
their
context
and
students
do
something
useful
with
them
and
I
think
this
is
a
pretty
important
requirement
and
it's
standing
against
I
have
computed
that
six
bytes
per
response,
most
real-world
applications.
A
It
competes
races
against
spending
six
bytes
per
response
on
something
that
that
is
redundant
when
you
do
have
the
context,
so
I
think
that's
a
an
important
or
difficult
decision
in
this
space.
If
you
look
at
this
specific
example,
you
have
on
the
slides.
The
first
thing
you
would
do
is
compress
those
those
ugly
gates
there
and
we
could
save
three
times
as.
G
A
By
doing
that,
and
then
you
can
spend
six
bytes
on
on
the
debugging
support,
but
I
think
in
the
IDF.
We
often
do
not
think
enough
about
deployment
issues,
and
this
is
one
place
where,
where
we
got
feedback
from
people
who
want
to
deploy
this
and
say
we
are
going
to
have
problems
with
our
tools.
If
you
do
it
this
way,
so
maybe
we
actually
have
to
bear
that
cognitive,
dissonance
of
wasting
six
bytes
and
just
make
it
simply
and
more
simpler
and
more
reliable
to
actually
roll
out
this
project.
M
Deborah,
having
worked
on
similar
problems
with
other
protocols,
I've
worked
on
where
you
you.
Basically,
your
uri
is
setting
a
context
and
you
have
want
to
take
a
chunk
of
the
payload
and
save
it
off
or
give
it
to
somebody
else
and
you've
lost
that
what
we
established
it
was
sort
of
copied
over
from
some
other
protocols.
We
invented
a
tag
that
you
could
stick
in
here
now.
M
I
know
this
message
with
signatures
and
stuff,
but
we're
assuming
we're
talking
about
some
I'm
taking
a
snippet,
so
signatures
in
the
encryption
out
of
the
picture,
so
I've
got
a
snippet
I
want
to
hand
somewhere
or
store
it
for
later
retrieval
or
whatever.
We
simply
add
a
new
attribute
in
there
called
self
borrowing
from
the
other
world
where
we
borrow
so
self
was
the
URI
of
the
context,
and
so,
when
you
need
it,
you
provide
the
context.
It's
not
always
present
so
on
the
wire.
T
A
T
A
S
S
A
A
I
mean
starting
from
a
Wireshark
detector,
to
lots
of
places
where
these
things
will
just
turn
up.
You
can
curl
and
get
some
some
data
somewhere
and
lost
the
context,
and
so
on.
It's
just
really
hard
to
make
sure
that
that
this
transformation
always
happens,
and
so
it
solve
some
of
the
problems,
but
so
yeah
I'd,
say
I
think
we
still
have
to
figure
out
some
details
there,
but
one
other
thing
that
might
result
from
this
change.
S
For
them
yeah,
that's
for
the
comma
document,
though,
to
see
what
would
be
more
yeah,
so
that
I
think
maybe
two
if
you
wish
to
discuss
on
the
mailing
list,
unless
you
feel
pretty
strong
against
it
or
unless
you
you
know,
you
have
other
ideas,
please
go
and
write
it
to
the
mailing
list
and
I
think
that
so
try
to
already
memories.
So
otherwise
you'll
just
go
ahead
and
you
know
and
do
the
change.
A
S
As
I
said,
the
young
and
see
bore
in
the
six,
a
drafter
are
ready
to
ship
after
the
site.
Yet
and
of
course,
after
we
have
been
apply
these
changes
and
with
application
in
rest,
confident
core
Cove,
we
had
two
reviews
from
net
net,
not,
and
yet
so
we'll
be
asking
for
last
call
last
call
same
thing
is
for
come
I,
and
one
of
the
things
is
that
we
would
love
to
have
one
or
two
reviews
from
core
where
people
are
more
knowledgeable
about
coop
and
about
you
know
these
things.
S
A
A
We
have
a
little
bit
of
testing
and
the
reviews
will
make
us
make
taking
care
change,
so
we
probably
have
to
do
another
round
of
interrupts
testing,
but
since
this
can
be
done
online
we
don't
we
scheduled
a
big
event
or
so
for
something
like
this,
so
my
hope
would
be
that
we
should
be
able
to
work
during
September
of
all
these
documents
after
the
Interop
has
been
done.
That
requires
a
little
bit
of
pain
from
the.
A
T
This
is
Hank
busy
be
coming
from
that
comfort
was
session
before
so
there
is
this
notion
of
that
cognitive
drafts,
they're
called
they're.
Basically,
they
transport
heart
ripped
out
of
the
generic
models,
and
you
can
stick
in
to
get
net
con
or
s
comped,
and
then
the
notion
cap
could
be
another
net
conf
notif
draft.
T
T
And
just
highlighting
what
they
do
to
be
modular
regarding
it
with
respect
to
transports
and
they
have
their
own
I,
don't
know
subgroup
inside
their
group.
That
is
ascending
to
do
some
aspiring
to
do
some
binary
stuff,
so
yeah.
I
think
I
think
the
chairs
aware
enough
that
con.
I
just
wanted
to
create
awareness
here.
Also
so
maybe.
T
A
A
Good
point
that
we
should
start
on
there's
also
another
document
in
in
the
Corcoran
Group
that
we
haven't
talked
about,
which
is
the
equivalent
of
how
it's
called
the
yang
library
yeah.
So
we
probably
want
to
do
that.
One
is
well.
We
may
not
need
to
do
it
exactly
on
the
same
timelines,
but
we
also
probably
should
should
try
to
align
this
so
anyway,
on
the
microf
would
quickly
say,
with
an
umbrella,
is
damaging
your
document.
The
network
management
data
store
architecture
nimda.
Is
that
changing
anything
in
your
document?
Yeah.
T
Yeah
but
again,
I'm
gonna
have
I,
did
the
libraries
or
the
capability
stuff
is
somehow
semantically
different
from
all
the
flavors
of
their
stories.
They
now
have
with
nmda,
which
is
applied,
running
startup
intended
something
something
a
lot
of
those.
So
that's
another
problem
escape
then
than
just
a
library
of
telling
you
hire
can
do
this,
which
is
pretty
much
I,
guess
a
lot
of
simpler
than
all
the
other
nmda
stuff.
That
is,
the
flavors
of
data
stores.
Yes,.
S
To
having
that
point,
I
think
that
this
could
be
the
place,
for
you
know
for
the
young
of
things.
These
kind
of
discussions
then,
can
carry
on
on
where
these
two
communities
meet
a
little
bit
more.
You
know
to
to
say:
okay.
Well,
there
is
kor
khon
and
there
are
the
other
things
that
come
from
the
from
the
net
come
from
from
from
from
the
yang
world,
so
that
you
know
that
we
were
discussing
in
any
more.
S
I
mean
we
had
a
site
meeting
to
to
IDF's
ago,
and
we
had
around
30
participants
from
net
mod
and
from
and
from
coal,
and
we
gathered
around
and
we
discussed
about
all
the
possibilities
that
come
I
at
that
time
opens,
and
you
know
the
user
yang
for
the
Internet
of
Things,
and
the
idea
was
of
course,
to
see
okay,
how
these
two
communities
can
work.
They
can
meet
so
that
you
know
we're
not
getting
purely
router
oriented
technology
and-
and
you
know
just
try
to
fit
in
and
then
they
move
on
their
ways.
S
And
you
know
they
just
ignore
completely
the
IOT
world,
and
so
that
you
know
there
is
this
place
of
discussion
and
once
there
is
core
count,
of
course,
then
yang
modules
will
start
being
suddenly.
There
will
be
very,
very,
very
easily
portable
and
usable
for
IOT
devices.
So
whenever
there
is
a
young
module
that
needs
to
be
developed
for
IOT,
then
that
would
be
also
some
kind
of
a
natural
place
to
be
to
be
discussed,
because
it's
people
that
know
ya
and
know
how
to
do
this
thing
and
people
that
understand
that
okay.
S
Well,
these
are
not
routers,
there
are
some.
You
know,
let's
be
a
little
bit
reasonable.
So
there
is
this
place
to
this
place
on
which
you
know
people
can
talk
about
and
can
design
these
models.
So
that
would
be.
This
is
an
idea
to
have
this
forum
this
this
group
eventually
the
yang
of
things,
and
it
seemed
that
it
was.
There
was
all
interest
at
that
time
and
right
now,
I
think
that
we
are
getting
to
the
point.
That's
the
techno
that,
right
now
the
protocols
will
be
there
and
we
can
start
this
discussion.
Yep.
A
So
if
the
the
you
know
thing
grows,
we
probably
don't
want
to
continue
to
do
this.
All
of
this
in
this
workgroup
in
it
Ria's
operations
and
management
at
some
point
in
time
and
no
longer
applications.
So
it
also
makes
sense
to
do
that,
but
that's
something
that
that
has
to
be
discussed
with
Eddie's
and
and
moved
into
above
and
so
on.
A
So
we
have
about
minus
five
seconds
in
in
the
current
segment,
but
I
would
like
to
quickly
bring
up
one
thing
on
Thursday
we
will
have
one
customer
of
coop
quickly
present
how
they
are
trying
to
use
coop
for
DDoS
detection
and
notification
of
ITER's
events,
and
today
maybe
Hank
can
say
a
few
words
about
how
co-op
might
be
used
for
ring.
Telemetry.
T
Hi
Hank
presenting
yeah.
This
is
a
short
introduction
about
concise,
yang
telemetry,
which
once
was
before
called
concise
yang
push,
but
the
term
force
went
away.
So
there
are
no
customized
data
stores
or
a
customer's
data.
Subscriptions
and
add
terminology
changes
their
lot.
So
I
I
used
an
term
that
is
basically
known
by
everyone.
We
are
creating
streams
here
so
excited.
T
Basically,
there
are
two
drafts
in
being,
basically
in
working
class
call
now
and
that
conch,
which
is
the
data
source
subscription
and
the
SUBSCRIBE
notifications.
Notifications
are
the
thing
that
you
get
typically
with
the
standard
yang
data
store
when
something
fails,
its
control
plane,
but
people
thought
like
well.
We
always
wanted
these
SNMP
traps
back.
Why
not
make
it
more
useful
for
data
where
you
change
and
then
they
created
subscribe
notifications,
so
you
can
now
build
your
data
model.
T
Ask
notifications
and
just
subscribe
to
them-
that's
kind
of
very
easy
because
we
defined
them
in
the
model
and
subscribe
to
them
at
the
end.
You
were
interval
or
unchanged
stuff,
and
and
that's
all
nice,
but
but
the
real
interesting
thing
here
is
to
make
use
of
all
the
existing
yang
notice
that
the
two
thousands
of
the
image
of
a
thousand
these
are
defined
here
and
the
IDF
and
then
subscribe
to
change
of
data.
T
Node
values
of
existing
data
store
yang
modules
and
that's
the
extra
power
here,
because
you
leverage
history
like
to
say
the
power
of
Yang
and
don't
we
have
to
redefine
anything
and
from
that
you
can
create
a
binary
telemetry
using
sip
or
other
yang
model
SIBO.
Basically,
and
if
you
close
look
at
the
comb
as
design
enables
you
to
do
kind
of
easy
coop
a
subscriptions,
for
example,
subscriptions
have
filters.
Sometimes
you
don't
want
the
next
at
least
maybe
yeah.
T
Sometimes
we
have
filters
and
you
don't
have
the
complete
data
module
value
changes,
but
just
a
subtree
of
it
and
such
which
can
be
quite
easily
identified
now
because
it's
said
so
we're
doing,
coop
observe,
probably
reliable
and
some
bunch
of
ion
and
said
and
Tara.
You
have
a
subscription
on
the
sub
three
filter,
expression
and
and
there's
nothing
as
easy,
simple
as
that
investor,
confident
conf
and
that's
prepared
for
something
of
this
constraint.
Note
environment
there's
something
more
to
it.
So
a
subscription
IDs
are
created
by
this
java.
T
T
I
am
entertaining
the
idea
of
populating
the
s
resource
with
keys
that
are
basically
the
subscription
that
is
created
by
the
server
and
then
you
post
a
subscription
RPC
input
with
a
ton
of
details
that
I'm
not
going
to
explain
you
right
now,
and
then
you
get
a
subscription
ID
back
and
just
look
at
s,
subscription,
ID
and
observe
that
so
it's
your
individual
created
subscription
on
datastore
no
value
changes
enter
our
binary
telemetry
on
change
or
periodically
it's
the
counter.
You
probably
don't
want
unchanged
subscriptions
on
Packer
consoles,
I,
guess
so.
T
There's
a
lot
of
flexibility
here,
especially
using
force
an
eye
patch,
and
we
are
also
considering
eliminating
the
complexity
of
XPath
expressions.
The
subtree
is
easy.
Xpath
is
hard
and
next
class.
Today's
I
think
somehow
Jason
translation,
just
natively
translated
into
C
war
and
it's
kind
of
I,
don't
know
a
little
bit
too
much.
Maybe
so.
We're
exploring
the
idea
of
using
a
specific
set
of
map
of
SIDS
with
conditions
to
substitute
the
very
complex
XPath
expression
that
is
now
being
able
to
expression,
see
where,
but
maybe
we
can
do
a
little
bit
simpler.
T
A
A
T
As
soon
as
it
is
easy
to
create
binary,
telemetry
streams
of
messages,
abundant
messages,
people
think
how
cool
this
is
event
event.
Data
now
I
would
use
it
for
security,
because
I
online
life
visit
the
security
events,
and
so
we
took
that
into
account
a
little
bit
and
saying
okay.
This
is
typical
operations
and
management,
but
please
expect
to
be
used
in
the
Shuji
area,
so
resilient
subscription
and
visibility
of
changes.
A
R
Yeah
so
in
in
person
how
to
did
like
this
draft
largely
has
been
done
for
a
while,
but
but
we
we
wanted
to
see
if
we
should
actually
integrate
some
of
the
other
things
that
we've
seen
the
industry
and
one
of
the
things
that
I
did
look
into
recently
was
the
OMA
lightweight
machine
to
machine
specifications.
They
actually
specify
a
couple
sub
branches
of
this
urin
called
OS
and
hopes
for
organization,
serial
numbers
and
organization
product
and
serial
number,
so
I
folded.
R
Those
in
because
I
actually
felt
that
it
would
be
important
that,
like
there's
one
place
that
actually
says
that
here
are
the
different
sub
branches
and
the
soon
taxes
would
actually
be
somewhat
compatible
and
and
there's
some
changes
that
I
had
to
do.
Because
of
that.
So
we
had
recently
introduced
this
org
sort
of
organization,
freeform
Device
Identifier,
which
used
:
and
and
these
things
from
OMA
used.
R
R
What
in
the
new
year,
what
is
in
the
new
version
and
just
to
highlight
sort
of
what
this
results
in
and
that
the
idea
is
that
that,
if
you
need
your
organization
needs
some
kind
of
serial
numbers
or
whatever
private
numbers,
then
then
we
have
a
branch
for
set
of
branches
to
do
that
and
there's
actually
three
levels
like
one
is
totally
free
format.
One
is
assuming
serial
numbers
in
an
organization
and
one
is
serial
number
and
and
product
type,
so
I
think
it's
a
reasonable
set
of
things
to
have
for
for
general
purpose
use.
R
But
there
are
some
some
questions
here
so
so.
First
of
all,
this
is
Leslie
my
own
opinion
that
we
should
actually
try
to
do
some
alignment
here
and
put
things
in
the
same
same
place.
It's
not
the
intent
to
hijack
any
other
work
and
I'm
gonna.
You
know
impact
other
people's
specifications
unless
they
they
want
us
to.
But
this
is
like
my
proposal
that
that
we
should
actually
do
this
and
asking
this
community
now.
R
Are
you
guys
here,
whether
that
makes
sense-
and
there
are
also
more
detailed
questions,
so
the
OMA
thing
actually
uses
identify
the
old
UI
identifiers
logically,
and
we
had
previously
used
the
pin
numbers,
which
was
the
eye
on
our
private
enterprise
numbers
and
and
there's
a
question
like
which
one
should
we
use
or
both?
And
should
we
change
this
like
what
the
guys
didn't
OMA
a
change
might
be
painful
for
them
or,
if
there's
no
usage
of
their
current
thing.
R
Yet
then
a
change
might
be
okay,
and
then
we
do
the
best
design
and
aligned
design,
but
I,
don't
know
like
totally
free
from
any
complications.
I
would
like
personally
rather
use,
pin
numbers
because
I
can
get
them
easily
without
any
cost
or
a
co2
is
actually
cost
real
money,
as
if
I
understood,
I,
to
fully
process
correctly
and
and
there's
some
some
details
that
that
we
may
not
have
to
go
into
right
now.
But
I'll
point
to
that
like
that.
R
Well,
the
versus
:
thing
and,
as
a
person
sign
think,
Ra
was
arguing
a
while
ago
to
me
that
that
the
person
sign
might
be
a
problem
from
a
sin
ml
perspective.
If
these
identifiers
are
used,
their
person
sign
is
used
in
in
the
oma
spec.
So
so
that
that's
the
proposal
in
front
of
you-
and
this
is
otherwise
done,
but
there's
no
way
the
widget
integrate
with
this
sort
of
stuff.
It's
the
question
and
please
comment.
V
I'm
here
I
can
speak
a
bit
on
what
we
do
in
OMA
and
basically
for
the
second
question
is
a
no-brainer
I
think
Ben
numbers
makes
more
sense
for
free,
and
why
not
do
you
do
you
have
an
understanding
of
how
much
uses
of
their
current
format
exists
today?
So
I
know
that
as
I
already
relay
before
with
you
at
least
for
us
ma,
is
the
most
common.
Your
annuity
may
I,
don't
know
all
the
members
I
mean
part
of
Campeche
knows
or
other
Orman
OMA
members
can
pitching,
which
is
the
most
common
either
way.
V
V
N
Carolyn
so
I'm
guessing
that
when
you
use
the
term
private
you're,
not
talking
about
privacy,
I'm,
just
wondering
whether
or
not
this
proposal
intersects
at
all.
With
this
mania,
that's
happening
sort
of
like
down
the
hall
about
not
using
hardware
identifiers
anywhere
for
any
purpose.
If
they
can
be
associated
with.
You
know
anything
in
the
real
world,
but.
R
That's
a
good
question
and
it's
been
discussed
previously
and
that
document
does
have
some
words
about
that.
Yeah
I.
Think
the
view
is
that
you
know
people
who
use
these
kinds
of
things
need
to
be
aware,
but
there
are
some
use.
It's
cases
where
it
actually
does
make
sense,
and-
and
we
want
to
enable
those
use
cases
and
and
documentation
like
if
I
in
my
own
network,
have
a
file
that
has
device
identifier.
It's
probably
okay,.
W
Or
akhirin,
and
so
maybe
I'll
give
a
bit
more
details
on
the
person
issue.
So
in
the
central
specification
we
recommend
debut
our
ends
as
one
convenient
way
of
keeping
the
names
prefixes
cuz.
They
can
easily
then
generate
global,
unique
names
and
so
far
the
Devi
or
hands
have
been
fully
compatible
on
their
character,
set
with
a
relatively
constrained
set
of
send
ml.
So
that's
why
I'm
a
bit
concerned
of
adding
person
sign
that
would
make
only
subset,
compatible
yeah.
R
And
just
for
clarification,
so
the
reason
that
is
being
added
is
that,
like
you
know,
if
I
import
this,
you
know
other
sub
identifiers
from
the
OMS
back
here
unchanged.
Then
that
draws
in
person
soon
box,
but
we
can
make
changes
and
I'm
suggesting
changes
on
other
aspects
and
it
sort
of
critically
depends
on.
If,
if
there's
usage
of
this,
then
we
can
make
much
changes.
But
if
there's
not,
we
should
make
the
best
design
possible
might
include
not
doing
that
person.
Son
son.
K
Hi
here,
cuz
I'm
just
kidding
awake
3:15
in
the
morning.
Thanks
for
waking
me
up,
it
is
more
or
less
like
I
mean
there
are
different
formats
which
we
are
using
for
the
end
point
name
and
we
are
converging
certain
things
like
what
Amy
mentioned
about
amia
and
other
things,
which
is
evolving
for
3gpp,
like
external
ID,
is
another
one
which
we
try
to
fall
back
on
nai,
which
is
again
referenced
here.
My
suggestion
is
like
if
we
can,
whatever
ietf,
if
we
can
bring
it
under
one
umbrella
and
generalize.
K
R
K
R
We
could
certainly
throw
out
the
addition
from
the
draft
and
says
publish
what
we
have
I
think
it
would
be
valuable
if
we
could
reach
an
aligned
proposal,
and
then
things
would
be
more
understandable
from
from
people's
point
of
view
going
forward,
so
we're
gonna,
try
and
report
back
and
then
revised
draft
accordingly,
and
hopefully
that
that
means
that
it's
going
to
be
a
perfect
self-contained
document
that
can
be
taken
to
work
new
Blasco.
Thank
you.
Okay,.