►
From YouTube: IETF100-MODERN-20171115-0930
Description
MODERN meeting session at IETF100
2017/11/15 0930
https://datatracker.ietf.org/meeting/100/proceedings/
A
E
D
A
D
D
G
D
I
guess
anybody
who
needs
to
make
a
comment:
please
do
it
through
the
meat,
echo
I,
guess
if
we
can't
get
a
summons
jabber
room,
okay,
so
the
agenda
is,
we
have
just
an
hour
here
through
a
couple
of
very
interesting
topics:
I
think
that
we'll
we'll
go
through
and
understand
that
some
offline
work
has
been
going
on
here.
That
should
be
interesting
to
see
any
any
comments
on
the
agenda
or
agenda
bashing.
That
needs
to
be
done.
G
Maybe
a
little
rubberneck
my
way
through
this-
that's
probably
simplest,
so
yeah
we're
gonna
talk
today,
a
little
bit
about
Terry
Terry
for
those
of
you
who
have
not
attended
modern
meetings
in
the
past
and
I.
Think
I
see
faces
that
look
kind
of
familiar
to
me.
Looking
around
this
room,
Terry
is
a
candidate
protocol
for
adoption
here
at
modern.
G
To
try
to.
Can
you
go
to
the
next
line,
to
try
to
really
figure
out
a
way
to
normalize
three
functions
that
we
use
in
the
administration
of
numbers?
Chris
went
and
I
actually
spent
some
time
since
we
got
here
in
Singapore
hanging
out
and
talking
about
this
and
I
know.
This
has
helped
clarify
for
me
a
bit
the
distinction
between
these
three
interfaces
and
good
ways
to
describe
them,
but
it's
they're
pretty
much
allow
three
kinds
of
operations
to
happen.
G
One
of
the
most
allocation
operations.
An
allocation
operation
is
a
request
in
number
administration
where,
yes,
some
entity
wants
to
get
a
number
assigned,
probably
or
own
number
block
allocated,
but
they
can't
do
it
themselves.
They
lack
the
authority
to
do
it
themselves.
They
kind
of
have
to
ask
somebody
else,
so
an
example
of
this
would
be,
you
know,
an
end
user
who's
going
into
a
you
know,
an
ATT
store,
that's
going
to
be
using
some
interface
type
interface
could
acquire
a
number
from
AT&T
from
a
block.
G
Ntt
would
end
up
creating
a
new
Terry
record
that
they
would
sign
for
that
particular
new
number.
That's
been
allocated
and
assigns
in
this
case
all
these
operations
end
up
operating
on
Terry
records,
that's
going
to
the
core
concept
behind
Terry
now.
The
second
is
operations,
that
is
the
management
operation,
and
this
is
an
operation.
G
That's
basically
about
pushing
a
record
there,
Terry's
services
that
exist
and
if
what
you
want
to
do
is
to
you
have
and
put
it
into
a
service,
you
use
management
that
could
be
updating,
modifying
an
existing
record
in
that
service
as
well
or
it
could
be
provisioning.
Something
fresh
that
hadn't
been
in
that
tarry
service
before
and
finally,
there's
retrieval.
Retrieval
is
a
search
like
operation
where
you
want
to
get
a
Terry
record.
A
Terry
service
exists
that
has
records
in
it.
You
say:
okay
within
these,
given
search
criteria.
G
Give
me
one
or
more
records
you've
got
that
I
should
see
for
that.
If,
in
fact,
I'm
authorized
to
receive
them,
that's
a
basic
idea.
We
think
that
most
of
the
things
you
could
conceivably
want
to
do
with
numbers
come
down
to
those
three
basic
operations
and
that's
kind
of
the
core
idea
behind
Terry
to
have
a
way
that
we
can
define
records
that
are
is
completely
valid
for
all
three
of
those
use
cases,
and
this
is
what's
so
different.
G
G
That
is
a
respect
in
which
again
I
think
we're
doing
something
slightly
different
than
what
we
were
doing
back
in
the
day.
I'm
good
I'm,
just
teasing.
You
know
some
discs
Kaiser,
but
but
I
mean
no
I,
think
it
does
explain
kind
of
what
the
motivation
was
for
revisiting
things.
This
way
next
slide,
and
you
know
so
we
have
a
picture.
There
are
these
clients
there
are
these
services.
You
know.
G
There's
a
concept
that
there
are
people
who
are
authorities
authorities
tend
to
have
credentials,
we
think
of
them
now
pretty
much
like.
There
are
potentials
like
there,
the
cuento
we
talk
about
in
stir
and
in
Acme
these
days
and
the
authorities
use
these
credentials
to
sign
records
and
they
provision
these
records
up
to
the
service
through
something
like
the
management
interface
I
just
described,
and
then
all
kinds
of
clients
can
use
the
retrieval
interface
to
access
these
services
and
get
the
records
back.
G
And
ideally,
these
clients
should
be
able
to
see
these
signatures
that
the
original
authorities
generated
when
they
created
these
records
and
use
those
to
decide
whether
or
not
they
trust
them,
and
this
is
kind
of
another
thing
that
I
think
is
very
different
about
this
model
than
the
way
we
thought
about
enum
or
kind
of
anything
that
was
down
to
like.
Should
we
have
this
like
big
flat
file
database
in
the
sky?
G
So
the
way
we've
defined
these
operations,
they
consist
of
a
a
request
to
nd
response.
You
know,
requests
and
responses
have
concepts
like
there's
a
source,
a
subject.
There
are
restrictions
if
you're
doing
we
retrieve
requests
like
give
me
all
the
numbers
in
this
area
code
or
something
like
that,
is
an
example
of
this
kind
of
restriction.
You
might
ask
for
we
have
a
concept
of
response
codes,
but
this
is
all
defined
in
the
base.
Carey
specification
in
a
way,
that's
very
abstract.
G
H
G
There'll
be
a
binding,
a
way
to
take
a
given
protocol
and
kind
of
show
how
the
semantics
of
a
request
and
response
map
on
to
that
protocols
primitives
so
for
HTTP,
which
is
likely
to
be
the
binding
we're
going
to
use
for
most
of
the
things
that
we're
interested
in
here
it
has
its
own
response
codes.
It
has
own
ways
to
define
requests
and
responses.
You
encode
the
subject
into
this
restful
aspect
of
the
URL
that
you're
actually
referring
to,
and
things
like
that.
It
is
our
job
once
we
all.
G
This
is
interesting
enough
for
us
to
do
it
to
then
go.
Take
that
next
step
and
say:
okay,
here's
how
you
Mack
the
semantics
we've
described
here
on
to
something
concrete
like
HTTP,
but
this
again
going
back
to
the
many
failings
of
enum
and
I
have
as
much
responsibility
for
them
as
anyone
does.
You
know,
we
didn't
really
think
about
things
that
way.
We
really
thought
about.
Let's
take
this
protocol
and
it
has
these
people
in
spearhead,
like
cram
as.
G
Can
into
it
and
like
overload
anything
we
need
to
so
in
order
to
avoid
that,
we've
focused
this
effort
initially
on
just
let's
figure
out
what
the
semantics
should
be,
and
then
we
can
figure
out
how
to
map
them
onto
protocols
to
do
the
job
right.
This
that's
a
pretty
much
as
much
of
a
refresh
of
what
terry
is
is
we're
gonna
do
I,
guess:
I
have
one
more
refresh
on
Terry
slide.
I've
already
talked
about
how
they're
collected
services,
it's
important
to
say
when
we
talk
about
a
service.
Don't
imagine
this
like.
G
There
is
one
service
in
the
sky
right,
I
know,
I
showed
a
picture
where
there's
a
service
and
all
these
authorities
are
pushing
records
into
it.
Services
could
be
radically
distributed,
and
this
is
something
Chris
is
going
to
be
talking
about
right
right
after
I'm
done,
which
is
ways
that
we're
looking
at
doing
peer-to-peer
and
distributed
services.
Various
kinds
be
aware
as
well
that
we're
not
considering
records
is
something
that
are
monolithic
or
unique.
There
could
be
in
a
service
multiple
records
that
cover
a
single
TM.
G
There
could
be
a
record
in
a
service
that
covers
a
block
of
numbers,
one
that
covers
just
a
single
number
and
when
you
query
with
the
retrieval
interface
to
find
records
for
that
number,
you
might
get
end
up
getting
back
both
of
those
records.
If
the
service
policy
thinks
that
you
should
and
you'll
trust
those
records
based
on
how
well
you
trust
the
authorities
that
generated
them,
and
it's
a
very
open
and
flexible
model
that
way
and
yeah,
we
have
a
whole
ontology
in
the
modern
framework.
G
Let's
you
know
how
we
think
this
maps
onto
government
entities
on
communication
service
providers
onto
registries
registrar's,
but
really
we
see
this
as
again-
something
that's
applicable
to
a
very
wide
range
of
scenarios,
some
of
which
look
very
much
like
number
administration.
Today,
some
would
look
like
things
we
think
are
plausibly
ways
we
we
might
want
to
take
this
in
the
future,
and
that
indeed,
was
the
exercise
that,
in
part,
heading
inspired
by
having
FCC
workshops
asking
this
kind
of
blue-sky
about
ways
to
do
this,
and
we've
tried
to
think
about.
G
You
know
what
seems
most
flexible
for
what
are
the
likely
future
or
compatibility
scenarios
next
slide
all
right.
So
a
lot
of
people
heard
me
talk
about
this
for
a
long
time
and
you
know
were
to
varying
degrees,
either
bored
by
it
or
saying,
oh,
my
god,
you're
trying
to
figure
out
a
way
to
make
it
so,
like
end-users
own
telephone
numbers
and
carriers
own
nothing,
we
must
be
stopped
in
any
cost,
and
rather
than
kind
of
perpetuate
that
fixation
on
those
particular
use
cases.
We
tried
to
find
a
use
case.
G
What
numbers
are
invalid
in
a
numbering
plan
and
what
numbers
happen
in
a
given
moment
to
be
allocated
or
assigned,
and
so
we
decided
to
say,
okay,
a
good
way
to
kick
the
tires
on
the
overall
data
model
and
system
that
we've
been
very
abstractly
constructing
here.
Let's
then
take
a
very
concrete
example
and
show
how
we
could
apply
that
to
this,
and
so
we
produced
internet
draft
about
this
and
again,
our
hope
is
that
this
could
be
used
to
share
information
between
carriers
to
help
them
decide
what
kinds
of
calls.
B
G
They
are
invalid
or
allocated
next
slide,
so
the
approach
we
have
in
Terry
valid
basically
takes
the
fundamental
record
concept
and
I'll
show
you
a
record
in
Terry
in
a
moment
and
adds
a
new
element
to
it,
and
settlement
is
called
allocated
and
we
have
three
values
for
it
in
the
way
that
we've
defined
it.
The
most
important
thing
to
understand:
I,
guess
about
invalid
numbers
to
start
is
what
it
says
at
the
bottom.
The
idea
behind
these
records
is
to
treat
them
as
a
kind
of
whitelist.
G
If
there
is
a
record
that
an
authority
has
signed
for
a
number
range
in
North
America,
then,
is
a
valid
number
age.
If
you
interrogate
the
set
of
signed
records
for
a
particular
numbering
plan,
and
the
number
is
not
falling
with
any
valid
record,
it
is
an
invalid
number.
So
in
other
words,
this
assumes
that
we
would
build
an
allocation
map
that
is
exhaustive
for
the
blocks
that
have
been
assigned
for
this
particular
numbering
plan.
G
But
the
basic
idea
was
yes
to
break
this
into
these
three
categories,
where
you
can
either
say
for
a
given
record.
That
would
show
in
number
block-
and
we
have
kind
of
two
flavors
of
Records
I'm
talking
about
you'll,
see
me
use
this
R
and
T,
so
think
about
our
records
is
representing
number
ranges.
Prefixes
blocks
like
a
10,000
block,
a
1000
block,
T
Records
tend
to
be
for
individual
numbers,
so
we
want
to
be
able
to
say
either.
Yes,
this
is
a
number
range
that
has
been
allocated.
G
That's
not
saying
anything
about
whether
individual
numbers
within
it
have
been
assigned
or
to
who
merely
that
it
has
been
allocated.
This
is
the
kind
of
thing
that
a
high-level
registry,
like
the
north
american
numbering
plan,
for
example,
could
generate
as
a
record
and
say
we,
as
the
authority
know
this
particular
NPA
and
XX
or
as
a
pooling
Authority.
This
particular
thousand
block
has
been
allocated
assigned
to
me
and
something
much
more
narrow
assigned
mean
for
this
particular
number.
G
You
know
so
someone
is
at
least
using
it
or
is
least
capable
of
using
it
is
not
merely
allocated,
but
there
is
some
entity,
some
customer,
perhaps
but
carrier
to
which
it
has
been
assigned
and
no
means.
This
is
a
valid
block
that
the
new
numbers
are
currently
assigned
to
it
there,
and
there
are
NPA's,
for
example,
that
we
know
are
valid
NPA's
in
the
north
american
numbering
plan,
but
are
not
yet
in
use,
and
so
this
would
be
for
things
like
that.
Yes,.
C
E
This
may
well
be
out
of
scope,
but
I'm
curious.
It
seems
to
fit
roughly
into
my
the
notion,
as
describing
high-level
nature
sort
of
say
over
numbers,
so
Bayes
and
another
set
of
kind
of
numbers,
namely,
which
exist
but
should
not
be
used.
For
example,
555
numbers
on
an
example
of
numbers
that
are
minded
they
are
allocated
in
some
sense,
I
mean
in
a
sense
that
they're
they
have
a
known
status,
we're
known
statuses.
They
don't
really
exist,
my
nail
reserved
for
in
whatever
purposes
so
I
wonder
and
I.
Don't.
E
And
so
when
you
just
need
to
me
come
on
my
sense
would
be
that
being
more
explicit
as
to
what
that
mean,
because
notion
allocated
it's
kind
of
I
just
went
as
you
know,
is
kind
of
a
little
on
the
vague
side.
So
having
a
little
bit
more
of
what
is
the
property
of
a
number
and
then
you
can
indeed
visa
alla
khair,
because
what
do
you
mean
by
allocated?
Is
you
expect
you
shouldn't
be
surprised
if
you
receive
a
call
from
that
number
yeah.
E
Implication
from
a
practical
perspective:
it's
not
just
some
and
having
more
information
as
to
what
that
is
so,
for
example,
test
numbers
would
be
one
example
about
God.
In
some
circumstances,
you
want
to
know
that,
yes,
this
is
indeed
a
number,
and
if
you
see
it
internally,
it's
probably
mind
there
may
be
a
good
thing
if
you
see
the
externally
at
pop.
Obviously
it's
a
bogus
number,
because
nobody
should
nobody
has
ever
been
given
Bo's
test
numbers,
so
I
would
want
to
think
about
as
to
audio
a
more
descriptive.
E
Why
Robin
and
having
yet
another
kind
of
query
that
you
can
have
a
in
rhesus
man
if
you
map
and
broadly
as
in
yes
I,
should
be
able
to
receive
calls
from
that
number,
no
I
shouldn't
and
then
here's
some
additional
information
as
to
what
that
number
happens
to
be
mmm-hmm,
that's
fine,
but
that
might
be
an
I
fit
into
back
without
cleaning
thing,
the
other
one
since
I'm
you're
going
to
cover
that
I'm.
Tell
me
the
other
issue.
E
E
Could
do
recently
reassign
we
when
was
reassigned,
because
we
I
would
that
you
would
want
to
use
be
what
people
want
to
be
able
to
do.
Is
they
say?
I
have
gotten
permission,
I'm
bad
person
to
call
that
number
whatever
in
July
and
but
number
of
us
reassigned
in
October
through
somebody,
which
means
somebody
else
now
has
it,
which
probably
means
I
no
longer
have
permission
to
call
that
number,
because
it
has
probably
when
we
assigned
to
somebody
else.
So
that
is
again.
E
This
would
doesn't
reveal
anything
of
any
significant
privacy
concern,
but
it
tells
me
immediately
that
whatever
permission
I
had
as
a
bank,
for
example,
to
call
that
number
is
no
longer
relevant.
So
again
you
might
want
instead
of
having
ten
different
queries
we
might
want
to
think
about
is
to
whether
they
all
fall
roughly
into
this
type
of
public
information
necessary
to
determine
whether
a
number
should
we
call
should
be
an
originator
of
a
call.
It
could
be
an
originator
of
a
call
and
what
else
do
I
need
to
know
about
it?
Yeah.
G
So
I
mean
certainly
I'd
say
in
the
design
of
this.
We
came
up
with
three
values.
It
could
be
that
there
are
more
values
that
are
useful
for
the
two
things
you
just
described,
I
would
say:
I
can't
imagine
for
Assigned
in
particular
having
an
assigned
date,
be
something
that
is
available
as
like
an
optional
field
attached
to
that
right.
So
yes,
this
allocated
to
sign-
and
it
was
assigned
on
the
following
date-
seems
like
a
useful
thing
to
be
able
to
add
to
that
so
I.
G
You
happy
to
put
that
in
that
the
no
thing
it's
interesting
I
mean
you
know
when
we
talk
about
like
an
area
code
not
in
use.
Yet
if
you
actually
read
the
draft
I
go
through
actually
in
some
detail,
things
like
easily
recognizable
codes.
I
go
through.
You
know
all
the
cases
where
you'd
want
to
use.
No,
because
these
things
have
or
what's
been
reserved
like
we
have
hold
whole
sections
of
you
know
NPA's
that
are
reserved
now
write
to
that.
The
that
general
blob
is
what
I
wanted
to
flag
with
that.
G
Yes,
this
is
a
special
thing
and
you
should
treat
it
especially
because
of
that,
but
I
hope
for
things
like
ERC's
are
for
unassigned
blocks
or
even
blocks
that
were
assigned
and
then
reclaimed,
which
there
are
a
couple
of
those
right
that
that
should
just
fall
under
no
and
that's
there's
really
no
other
reason
for
no
to
exist.
Otherwise
we
just
not
have
a
record
at
all
right,
like
the
reason
why
that
we
distinguish
between
no
and
having
no
record
is
because
these
things
it's
possible
in
the
future.
They
could
be
used.
F
Think
the
historical
thing
is
an
interesting
thing:
I've
been
sort
of
vaguely
thinking
about
it
and
whether
or
not
it
should
be
explicitly
part
of
the
record
or
or
is
it
just
a
database
thing?
That's
crea
Buhl's
through
some
separate
thing,
but
knowing
all
the
historical
you
know,
porting
or
whatever,
and
all
the
signatures
associated
with
that
and
you
know
any
potential
activity
that
you
know
like.
We
talked
about
air
conditions
or
service
provider
to
tries
to
allocate
a
number
in
service
provider.
One
or
you
know
some
of
those
types
of
things
like
do.
G
I
could
definitely
imagine
having
a
whole
historical
kind
of
extension
for
these
records
in
general.
That
would
point
you
to
either
either
an
index
to
prior
records
that
this
is
tied
to
right
and
the
date
in
which
you
know
this
record
was
changed
in
this
way,
I
mean
so
we
we.
Obviously,
this
whole
model
is
designed
to
be
extremely
extensible
in
those
kinds
of
ways.
G
These
JSON
objects
right
so
right
effectively,
so
I
mean
yeah,
it's
exactly
getting
into
the
weeds
of
what
those
are
that
we
need
to
do
right
to
turn
this
into
something
real
and
not
just
theoretical
right.
F
E
Sorry,
I
really
important
because
you
know
extensible
what
all
about
is
that
if
you
want
to
add
more
information,
kind
of
as
additional
background
that
you
may
not
care
about
in
some
cases,
but
it
may
give
you
better
Diagnostics
that,
like
you,
got
a
test
number
some
way
or
it
showed
up
in
a
place
that
it
wasn't
supposed
to.
That,
tells
you
something
interesting
what
it
is,
but
I
and
so
having
a
more
fine
finer
granularity
there.
But
in
terms
of
use
case,
what
we
seem
to
be
this
seems
to
be
information.
E
E
G
E
Has
this
number
meaning
if
you
receive
a
call
from
it
in
itself,
it's
at
least
it
passes
that
check,
and
so
maybe
focusing
on
what
is
B
and
pump
use
case
kind
of
level,
and
so
that
we
don't
stuff
stuff
in
there.
That
is
likely
to
be
seen
as
confidential
or
proprietary,
which
seems
to
do,
but
I
want
to
have
all
the
information
that
I
can
put
there
as
opposed
to
having
17
different
areas.
The
next
slide,
because
this
this
little
help
with
this
I,
think
Thanks.
G
So
yes,
first
of
all,
this
is
just
what
it
looks
like
right:
here's
a
valid
block
allocated
equals
yes,
so
this
is
just
to
show
fundamentally
what
kinds
of
things
that
dot
dot
is
for
all
the
rest
of
the
junk
there
are
signatures
and
so
on
they
go
into
every
record,
but
go
to
next
one
yeah
I
mean
the
idea
that
we
have
these
in
to
record
categories
there
we
go,
you
know
so
so.
Basically,
there
is
a
distinction
between
this
individual
assignment,
stuff
and
TN
ranges.
G
G
If
you're,
not
you
don't
right,
so
I
think
the
data
model
has
it
right
to
make
it
work
that
way,
I
haven't
made
them
hierarchical,
just
separate,
you
know,
I
do
want
to
make
sure
people,
people
get
that
that
you
know
like
they're
people
who
feel
strongly.
We
should
make
this
hierarchical.
I've
heard
that
argument,
but
well
we
want
to
make
sure
that
within
each
range
you
know
the
T
records
are
all
embedded
in
the
our
records
under
the
you
know,
I
I
think
having
it
be.
G
F
G
You
do
in
the
sense
that
we're
defining
records
and
the
records
we're
defining,
have
a
certain
atomicity
to
them.
These
aren't
just
like
you
know,
sequel
yields
right
and
because
again,
because
different
people
sign
those
they're
they're,
the
records
have
this
like
cryptographic,
wrapper
around
them
and
some
specific
Authority
generates
them
so
because
of
that
I
I
tend
to
not
actually
think
about
them
like
they're
hierarchical.
This
is
this.
Is
this
is
a
very
important
design
decision
now,
so
we
need
that
anything
I
care
about
it.
G
E
A
mine,
John
Doe,
has
a
number
and
remember
before
and
after
happens
to
be
not
yet
assigned,
but
they
got
ball
block
I'm
assigned
to
recur.
You
at
that
point
within
portability
is
another
angle.
Affordability
is
number
one,
but
leaving
aside
for
a
moment
I,
so
the
the
notion
would
be
that
a
single
number
could
have
a
very
high-level
description
as
part
of
a
range
and
a
more
detail
to
further
some
subset
could
have
a
detailed
description.
E
My
other
concern
again
just
because,
like
you
said
we
have
a
number
of
different
publishers
of
this
information
could
be
elect
or
entity
could
be
minor.
General
and
number
management
entities
that
are
done
to
carrier,
I
or
some
third
parties
is
I
would
actually
argue
that
having
an
explicit
wife
or
somebody
decide
that
I
believe
that
these
numbers
are
like
I
said
invalid,
as
in
you
should
not
receive
numbers
from
that
is
actually
a
good
thing.
I
thought
you
should
explain
if
I
have
on
good
authority,
unassigned.
G
E
E
G
I
Brian
Rosen:
do
you
really
want
to
make
it
this
North,
American
specific
and,
in
particular,
not
have
free
fixes?
Oh
they're.
G
G
So
we
yeah
we
so
modern
does
is
differently
stir
and
trust
me.
This
causes
all
kinds
of
problems
so
like
we
were
fixing
that
next
slide.
Interestingly
as
well,
we
talked
a
bit
about
queries
and
propagation
in
this.
This
is
stuff,
though
I
think
I
need
to
beef
it
up
more
in
the
document.
This
just
comes
down
to
the
way
that
you
use
the
management
interface
to
make
sure
that
you
propagate
out
this
information
between
Terry's
services.
Some
of
this
is
relevant
in
particular,
to
distributed
registries.
G
That
Chris
is
going
to
talk
about
momentarily,
but
I
guess
the
important
thing
to
say
about
that.
Is
it's
easy
to
kind
of
read
the
way
the
drafts
is
written
now
like
we're
talking
about
taking
a
nrt
T,
taking
a
network
dip
like
every
time.
A
call
comes
in
to
look
and
check
to
see
if
this
number
is
valid
or
not,
that's
not
actually
what
it
what
it
means.
G
Logically,
we
talked
about
that
operation,
but
we
imagine
that
actually
they're
going
to
be
local
databases
of
these
records
that
are
being
collected,
probably
through
the
management
interface.
That
will
you
know
you:
could
you
can
imagine
a
regulatory
authority
and
national
authority
bulk
publishing
out
these
records
so
that
they're
locally
provisioned
in
databases
where
they're
needed
for
people
to
manage
this
in
real
time
next
slide?
G
So
do
we
need?
We
need
energy
and
discussion.
We
exit
some
discussion
today.
Discussions
good
I,
think
there's
some
some
takeaways
on
this,
like
I,
want
to
get
into
more
kind
of
the
yes
unassigned
dimension
of
this.
That
is
useful
and
yeah
so
and
there's
some
things
about
the
terminology
and
getting
that
better
aligned
that
we
still
need
to
do,
but
you
know
again
provided
people
think
like
this
seems
practical
enough
that
we
might
want
to
actually
build
it.
Like
is
this.
G
This
is
always
kind
of
my
ask
when
I
get
up
here
and
talk
about
Terry
right.
Is
this
a
concrete
enough
thing
we're
talking
about
that?
It's
worth,
building
the
HTTP
binding
for
it
and
saying
here's
a
system
that
would
actually
do
this
and
I
think
we're
pretty
close
on
this
now.
Actually,
I
think
this
is
something
that
it
might
actually
be
useful
to
people
today
so
like
have
something
that
does
this,
which
is,
is
perhaps
not
been
true.
Some
of
the
things
we've
discussed
in
the
past.
J
G
So
short
answer
is,
we
assume
stir
in
the
sense
of
so
these
authorities
here
that
are
capable
of
generating
the
records
to
get
published.
They
have
to
have
credentials
that
they
sign
the
records
with
records.
Have
signatures
built
into
them.
So
we
assume
that
they
have
credentials
that
actually
have
a
scope
of
authority
for
number
ranges
in
them,
because
stir
can
assume
that
that's
the
answer.
G
On
that,
like
should
we
do
this?
Is
this
worth
pushing
this
forward
like
I'm
trying
to
make
a
heading
says?
Yes,
so
heading
says
we
should
actually
do
the
work
of
getting
the
protocol
interface
for
this,
because
we
we
just
you
know
I,
don't
want
to
do
that
unless
people
think
this
is
useful
and
awkward
and
that
there's
an
something
with
enough
meat
on
it.
That's
worth
doing
anyway,
you
know.
E
True
eases
a
I
think
it
can
actually
be
at
least
two
use
cases
that
we
already
identified
both
in
restores
type
of
environment
and
the
second
one
is.
It
is
sufficiently
simple
and
sufficiently
we'd
only.
It
offers
kind
of
a
first
good
test
case
to
see
if
this
whole
thing
works
and
I
know
it
works
as
it.
It's
not
too
complicated
to
do
this,
you
know-
and
does
me
international
stuff
and
all
about
and
having
something,
and
it's
sufficiently
simple
that
this
is
something
that
a
hackathon
could
do.
I'm
and.
B
B
F
F
It's
just
the
overview.
Actually
we
can
go
to
the
next
slide,
so
I
made
this
picture.
Hopefully
this
makes
sense
to
people
trying
to
map
the
roles
that
we
defined
register
on
register
R,
the
idea
of
an
administrator
and
the
services
that
would
have
to
be
available
in
maybe
a
real
deployment
of
this
and.
F
Part
where
you
know
records
are
being
passed
back
and
forth.
As
part
of
that
was
part
of
the
distributed
registry,
you
have
the
local
records
at
each
thing.
You
have
the
Terry
server.
You
have
the
database,
the
local
database,
where
the
records
are
stored
and
then
the
drip
node,
which
basically
has
the
interface
for
once
you
update
or
when
you
want
to
update
record
and
you
pass
it
out.
F
F
F
F
So
I'm
sort
of
making
the
assumption
and
I
think
we're
in
agreement
on
this,
that
all
these
transactions
are
registry
based
and
what
that
means
is,
if
I
think
the
the
major
example
here
is
the
creation
of
a
number
block
or
the
request
for
a
number
block,
if
you're
not
via
the
administrator,
doing
that,
if
you're
a
service
provider
doing
that,
you
need
to
have
some
sort
of
pending
status
or
request
status
for
that
number
block.
And
then
an
authority
would
have
to
come
back
and
then
do
another
registry
transaction.
F
G
I
mean
that's
of
John
here
again,
though,
the
one
qualification
of
that
is
course
again,
the
so
creating
any
Terry
record
requires
you
have
this
credential
right,
and
so
it
really
comes
down
to
the
fact
that
if
you
give
10
people
a
credential
with
the
same
scope
of
authority
this,
this
is
what
creates
this
very
interesting
problem
for
doing.
This
is
a
distributed
registry.
So
you
imagine
that
there's
a
block
of
10,000
numbers
and
you
give
10
carriers
the
ability
to
authority
over
that
block.
They
can
create
records
for
numbers
in
that
block.
G
G
F
I
think
that's
based
on
policy
and
we
probably
want
something:
that's
independent
of
policy,
but
there
will
be
a
hierarchy,
I
think
of
levels.
You
know,
the
simple
model
is
a
service
provider.
You
know,
has
an
OC
n
that
represents
their
ability
to
request
numbers
and
other
and
all
those
things,
but
they
don't
have
the
ultimate
say
to
like
okay
I'm,
going
to
allocate
this
block
and
just
take
it
away
and
I'm
gonna
allocate
every
number.
F
Or
something
like
that,
so
there
has
to
be
some
two
way
transaction.
There
can
go
to
the
next
slide,
so
yeah
that
sort
of
leads
into
what
I
was
just
saying.
What
there's
a
I
think
I
started,
calling
it
one
step
in
two
steps
and
maybe
there's
n
steps,
but
this
idea
that
you
know
there's
different
transactions,
and
so
we
can
go
to
the
next
slide
and.
B
G
F
F
So
yeah
I
talked
about
a
two-way:
it
was
more
of
an
allocate
assigned.
Maybe
you
have
the
credentials
to
allocate
within
that
thousand
block
or
ten
thousand
blocks.
So
maybe
that's
just
one
transaction,
because
you
can
just
sign
it
and
and
assign
it,
and
then
I
also
was
thinking
about
Eric,
and
you
know
how
do
we
handle
those
I
have
some
other
thoughts
in
the
next
slides.
But
you
know
what
what
does
happen
when
somebody
tries
to
allocate
something
in
a
one-way
transaction
and
that's
false,
and
we
have
to
consider
those
cases.
G
B
B
F
I
guess
in
my
sort
of
HTTP
crud
style
thought
process
I
was
trying
to
map
these
things
and
I
think
for
the
most
part,
it
works
as
far
as
I
can
tell,
but
acquisition
mostly
corresponds
with
create
operations
or
put
in
HTTP
next
slide,
and
that
management
is
mostly
about
modifying
records,
existing
records
or
adding
information
to
them
so
and
correct
me.
If
I'm
wrong
on
this
and
your
mind,
this
corresponds
with
mostly
update
or
post
operations
and
then
next
slide,
and
then
retrieval
obviously,
is
just
get
retrieving
information.
F
F
Right
now,
drip
does
say
that
you
passed
the
payload
in
the
vote,
so
somebody
could
like
authority,
could
vote
no
just
intentionally
because
you're
doing
something
that
you're
not
supposed
to
be
able
to
do,
rather
than
so
just
some
thoughts
that
we
could
actually
take
advantage
of
the
the
process
itself,
but
that
would
be
specific
to
a
distributed
registry.
It
wouldn't
obviously
work
on
a
on
a
traditional
one
and
then
next
slide
yep
any
hands
comment.
B
E
That
this
is
we,
if
you
get
the
confirmation
from
your
registrar,
that
you've
got
your
domain
iam.
Generally
speaking,
if
that
didn't
have
my
lesson
and
some
error
happened
right
then
the
registrar
needs
to
and
you
have
a
referral
to,
I
can
pending,
because
your
registrar
went,
rogue
and
I.
Don't
think
we
want
to
deal
with
that.
Kiel.
B
F
F
E
G
G
Right,
the
other
you
were
Adam,
alright,
so
post
post
put
and
get
so.
If
your,
if
work,
if
we're
in
so
an
allocation,
we're
saying
there's
a
record
that
doesn't
exist
yet
and
I
want
you,
the
server
I'm
talking
to
to
make
a
new
location
with
this
thing
in
it
is
that
post,
that's
usually
put
that's
usually
put
so
it's
because
it's
probably
a
location
that
doesn't
doesn't
exist
at
all
previously.
K
F
I
mean
there
is
some
different
philosophies,
but
in
general
that
I'm
aware,
but
is
there's
a
new
record
post
is
like
an
operation.
That's
general
purpose
like
I
want
to
do.
X,
Y,
Z
things
to
the
data.
You
know
it's
used
in
lots
of
different
operation,
different
scenarios,
whether
it's
just
modifying
a
record
or
actually
want
to
like
start
some
action
working
or
whatever,
because
post
you
put
the
payload
in
the
requests,
and
you
generally
get
a
response
with
a
payload
back
confirming
that.
But
anyway,
the.
G
F
H
So
I
wanted
to
give-
and
I
can
be
very
brief
with
this.
Quite
frankly,
an
update
on
nationwide
number
portability
so
go
to
the
next
slide
in
2016
seems
to
be
a
theme
here.
There
was
a
we
put
a
draft
use
case
in
related
to
something
called
nationwide
number
portability.
This
is
something
that
was
going
on
in
the
US.
The
FCC
was
looking
to
to
do
this.
Why
don't
you
move
on
to
the
next
slide
bunch
of
work
done
in
2016
things
kind
of
went
dormant
for
a
while,
but
there's
some
recent
activity.
H
It's
back,
so
the
FCC
put
out
a
notice
proposed
rulemaking
and
a
notice
of
inquiry
related
to
the
work
that
was
done
on
nationwide
number
portability.
So
to
put
the
two
things
together,
there
was
a
draft
that
was
a
use
case,
because
the
one
proposed
solution
for
nationwide
number
portability
had
a
brand
new
numbering
resource.
Basically
in
the
US,
a
new
area
code
that
would
be
used
as
a
routing
number
and
would
help
facilitate
deployment
of
nationwide
number
portability
it
they've
brought
this
back
up.
H
H
It
also
calls
for
a
parallel
IP
PSTN,
creating
a
new
PSTN,
so
you
have
kind
of
the
old
TDM
PSTN
and
a
parallel
IP
PSTN,
and
this
is
where
the
non-geographic
numbers
live.
This
is
where
they're
routed
so
because
the
numbers
are
on
this,
IP
PSTN
gives
the
opportunity
to
do
things
differently
than
we
do
today
gives
the
opportunity
to
conserve
numbers,
no
block
assignments,
more
relevant
telephone,
related
information
model,
Terry
new
bindings
and
encodings,
for
how
you
access
you
acquire
retrieve,
manage
the
numbers
distributed
registry
model.
H
They
referenced
it
themselves
in
their
in
their
order.
Let's
see
if
you
go
to
the
next
slide
so
quickly,
what
the
problem
is
is
lattice
the
country
is
divided
into
the
u.s.
is
divided
into
204
ladders.
These
are
little
geographic
areas,
some
of
them
big
that
they
they
were.
This
was
created
and
to
messenger.
H
Lex
carried
number
calls
within
the
ladder,
but
long
distance
had
to
carry
calls
outside
the
ladder
bottom
line.
Is
the
network's
been
designed
like
this
they've
the
vine
design?
The
switches
like
this?
It's
created
a
problem.
You
can't
port
a
number
out
of
one
of
these
ladders
into
another
ladder,
that's
sort
of
the
problem-
and
you
know
I
kind
of
call
this
because
of
there's
a
legacy
problem,
it's
sort
of
a
y2k
problem.
We
don't
even
know
where
the
problems
are.
H
If
you
were
to
try
and
break
this,
if
you're
trying
to
make
you
know,
move
numbers
from
one
geography
to
another
with
our
current
infrastructure.
So
if
you
go
to
the
next
slide,
so
the
solution
calls
for
a
parallel
IP
PSTN,
with
a
bridge
between
the
two
and
the
bridge.
Is
these
non-geographic
routing
numbers?
This
is
this
new
numbering
resource
that
in
essence,
lives
on
the
IP
PSTN
calls
to
the
non-geographic
routing
numbers
on
that
on
the
TDM
network
are
routed
to
the
IP
PSTN
for
call
handling.
H
The
reason
for
the
the
non-geographic
routing
number
is
any
TDM
switch
can
route
in
an
area
code.
There's
you
know
you.
We
can't
do
something
where
you
have
to
make
a
change
to
the
TDM
network,
but
routing
by
a
via
an
area
code.
Any
switch
can
do
that.
So,
if
you
go
to
the
next
slide,
so
here's
kind
of
high-level
vision
for
what
that
IP
PSTN
would
look
like.
H
You
see,
connections
to
the
PSTN,
the
TDM
PSTN
connections
from
these
things,
I
call
non-geographic
gateways
which
are
look
at
non-geographic
gateways
as
the
tandems
in
the
TDM
network.
I
I
didn't
go
into
it
before,
but
basically
there
are
two
what
they're
in
essence,
204
tandems
around
the
country
and
204
ladders,
and
in
order
to
get
numbers
you
have
to
connect
to
those
these.
This
would
be
sort
of
the
equivalent,
but
it
wouldn't
have
any
of
these
geographic
problems
that
exist
in
the
in
the
TDM
PSTN
and
just
go
to
the
next
slide.
H
So
to
summarize,
let's
see
on
the
TDM
PSTN
there's
kind
of
an
existing
process.
The
geographic
number
would
be
ported
to
the
to
the
non-geographic
routing
number.
On
the
IP
PSTN
new
processes.
We
have
all
kinds
of
new
processes:
mapping
the
non-geographic
routing
number
to
the
gateway,
non-geographic
gateway.
We
require
the
providers
would
acquire
the
non-geographic
routing
numbers.
The
gateway
providers
would
manage
the
the
routing
numbers
they
you
know
tell
what
the
address
is.
H
Networks
would
have
to
retrieve
this
information
and
yet
one
other
thing
I'd,
like
to
point
out:
service
providers
could
use
this
process
to
migrate,
their
customers
and
their
telephone
numbers
from
the
old
TDM
PSTN
to
the
new
IP
PSTN
I
like
to
say
cap,
the
TDM
PSTN
and
grow
the
IP
PSTN.
So
in
summary,
the
last
slide.
So
what
does
this
mean?
So
it
seems
that
the
u.s.
FCC
is
still
interested
in
future
of
numbering
and
number
administration
issues.
H
Modern
is
still
the
only
workgroup,
that's
that
is
working
on
global
standards
for
numbering
in
an
IP
environment
and
we'll
learn
more
in
the
future.
Oh
one,
more
thing,
I'd
add
which
I
found
out
after
putting
the
slide
together.
They've
FCC
has
a
Advisory
Committee
called
North
American
Numbering,
Council
they've,
recreated
that
just
just
yesterday
or
the
day
before
and
they've
added
a
nationwide
number
portability
workgroup
to
that
to
that
work.
So
yes,
in
fact
me
and
any
happened
to
be
on
the
so
again.
H
E
H
It's
like
yeah,
you
could
look
at
it
two
ways
it
could
just
identify
the
non-geographic
gateway,
but
period
likely.
The
non-geographic
gateway
providers
would
have
multiple
customers
behind
them,
but
they're
supporting
so
I
could
see
non-geographic
routing
numbers
per
carrier
that
they
support.
That's.
E
E
So
today,
in
its
annual
number
of
outage,
as
opposed
to
block,
but
it's
not
a
remapping
to,
because
you
could
imagine
and
we
mapping
for
an
individualized
number,
which
would
essentially
then
kind
of
change
for
portability
model
right,
you're,
adding
another
layover
into
action.
At
that
point,
right
right
and.
H
H
Those
yeah
yeah
and
by
the
way
one
of
the
other
things
I
propose
is
that
since
we
have
this
area
code
of
eight
million
numbers,
why
don't
you
just
give
out
non-geographic
numbers
to
consumers
as
well?
You
know
so
that
they
don't
have
to
eat
up
geographic
numbers
in
the
states
and
in
the
error,
codes.
I
didn't
mention
that
is
year,
but
that's
also
a
concept.