►
From YouTube: CBOR WG Interim Meeting, 2021-05-05
Description
CBOR WG Interim Meeting, 2021-05-05
A
A
So
do
we
have
anyone
around
who
has
read
the
document
and
would
be
willing
to
phrase
their
notes
in
a
way
that
it
could
become
a.
B
Review
well
I
heard
from
hank
who
has
a
standing
conflict
at
this
slot
in
in
the
week
that
he
has
looked
at
the
document.
He
hasn't
written
up
his
review
yet,
but
is
optimistic
that
he
can
do
this
within
a
week.
A
That
would
be
excellent
and,
as
I
seem
to
have
pronounced
it,
a
bit
badly
it's
about
ctdl
control.
So
the
document
with
the
new
operators
like
that
and
cat.
C
I
can
review
it,
I
have
very
little
knowledge
or
opinion
about
it,
but
I
can
read
it
and
see
if
I
understand
it.
C
B
Yeah,
it's
really
putting
together
three
things.
One
is
the
abnf
thing
which
is
kind
of
a
no-brainer
in
principle,
but
it
has
a
few
details
that
we
had
to
get
right
before
this
actually
could
be
used.
One
is
the
the
infrastructure
for
that
which
has
this
dot
cat
and
dot
dead
and
also
the
dot
plus,
which
which
is
just
speculative
right
now,
but
we
know
that
that
some
models
are
out
there.
B
That
could
have
used
that
and
the
third
one
is
the
the
feature
extension
the
feature
control,
which
turns
out
to
be
extremely
useful
in
the
sdf
work.
So
I
think
we
really
have
uncovered
a
good
way
to
handle
specification
or
model
evolution
here,
and
this
is
certainly
going
to
move
forward
and
and
become
a
more
more
capable
thing
over
time.
But
right
now,
the
the
very,
very
simple
thing
that
the
the
feature
control
offers
already
solves.
Eighty
percent
of
the
problem
that
we
were
having.
A
Is
there
any
particular
tooling
out
there?
That's
using
that
already
that
I
could
work
into
the
list
of
implementations
that
are
out
there.
B
Well,
everything
is
implemented
in
the
classical
ctdl
tool
and
I
know
that
that
andrew
weiss
has
a
partial
implementation,
but
I
think
he
recently
hadn't
hasn't
had
that
much
time
to
complete
things
and
actually
adding
abnf
is
is
a
lot
of
work.
If
you
don't
have
the
the
abf
infrastructure
in
your
platform
already,
and
it's
not
like
people
have
been
lining
up
for
implementing
a
b
and
f
in
various
platforms.
A
C
A
B
Yeah
so
I
have,
I
wrote
this
implementation
about
four
years
ago,
when
we
were
discussing
ways
to
to
efficiently
represent
the
web
of
things,
thing
descriptions
and
since
they
are
using
json
id,
the
the
those
thing
descriptions
tend
to
be
quite
chatty
now,
normally
a
thing
description
may
not
need
to
be
actually
on
the
constraint
device.
So
the
the
pressure
to
to
get
this
compact
isn't
that
strong.
So
after
in
some
initial
discussion,
we
we
didn't
really
work
on
that
for
a
while,
but
yeah.
B
So
I
I
dug
this
out
again
now
and
the
the
decompressor
is
complete.
I
think
so.
That
certainly
needs
to
be
validated.
B
The
compressor
is
rather
minded
so
right
now
it
does
a
perfect
job
on
sharing
items,
but
it's
rather
simple
minded
on
prefix
and
suffix
compression,
because
yeah
I'm,
I
haven't
done
the
the
formal
analysis,
but
I
would
expect
that
there's
significant
theoretical
complexity
to
that,
so
we
will
rarely
see
an
implementation
that
does
a
really
great
job
for
that.
B
If
it's
a
generic
implementation,
but
an
implementation
that
actually
implements
a
specific
application,
data
model
may
be
able
to
to
exploit
the
functionality
of
of
the
packed
format,
and
maybe
at
some
point
I
will
add
something
to
the
implementation.
That
makes
it
easier
for
an
application
to
control
how
how
cbo
pack
actually
works,
but
that's
not
there
yet
so
right
now
you
get
a
generic
compression
and
that
is
right
now
limited
to
just
a
prefix
sharing
of
strings.
So
we
are
not
doing
suffix
sharing.
We
are
not
looking
at
arrays
and
maps.
B
Yet
maps
are
particularly
interesting
because
it's
a
n-dimensional
problem
to
to
find
a
good
set
of
prefixes
for
a
map
because
they
aren't
ordered
so
prefixes
actually
aren't
prefixes
so
that
that
will
require
some
innovation
to
do
it
in
a
generic
way,
but
it
still
can
be
used
very
efficiently
for
specific.
A
And
any
any
user,
any
any
user
voices
or
applications
that
that
when
it's
the
the
particular
application
implementation
has
been
out
just
for
what
was
it
half
a
day
or
so
yes,
so
I
I.
I
think
that
this
will
need
a
bit
of
playing
around
by
all
involved
and
I'd
invite
everyone
to
do
it
so,
especially
with
their
own
kind
of.
So.
A
D
Suit
would
have
used
this
had
had
it
been
done
sooner,
but
I
doubt
we'll
go
back
and
open
it
up,
since
we're
at
working
group
last
call
with
the
spec
that
would
have
used
it.
A
Do
you
have
access
to
some
kind
of
small
corpus
of
suit
documents
that
you
could
just
throw
into
the
existing
compressor
and
kind
of
see
how
well
it
fares?
Even
though
there
is
something
else
going
in
there
or
is
there
a
way
for
to
to
kind
of
undo
the
optimizations
that
suit
does
and
throw
that
at
the
at
the
existing
simple
compressor.
D
B
Structure
yeah,
as
soon
as
you
do
something
at
the
application
level,
it
becomes
hard
for
for
the
for
a
generic
packer
exactly
to
unravel
that
and
and
provide
its
stuff.
A
I
was
thinking
a
bit
along
the
lines
of.
If
there
is
anything
left
ensued,
where
you
basically
can
opt
out
of
of
what
you
did
in
in
what
you
did
to
make
the
compression
happen,
then
then
this
then
this
could
be
applied
again.
No,
I
don't
know
well
that
fits
at
all.
No
that
doesn't
fit
what
we
have
at
this
point.
A
All
right
that
would
also
be
up.
That
would
also
be
a
good
pointing
time
to
to
have
another
look
at
the
document,
and
I
definitely
plan
on
on
having
one
more
look
myself.
A
B
Any
more
comments
on
this
for
for
here
and
for
now
for
for
the
way
forward.
I
think
the
the
important
observation
is
that
the
the
actual
pact
format
is
the
the
interesting
part
and
the
part
that
probably
can
stay
stable
for
a
long
time,
while
the
specific
part
that
does
the
table
set
up,
in
particular
the
tag
51
that
that
is
currently
proposed
in
in
the
document
that
might
have
more
frequent
variations.
B
A
The
question
of
the
of
the
scope
of
the
tables
is
that
pinned
down
in
the
current
document
already
or
is
there
some,
because
this
will
come
back
again
when
we
talk
later
about
the
records
proposal?
Is
it
can?
Can
the
table
be
altered
inside
the
document,
or
is
that
always
something
that
independent
of
whether
it's
done
through
tech,
51
or
any
other
tag
done
from
the
outside.
B
What's
inside
that
definition
and
that's
different
for
influences
from
the
way
the
the
record
proposal
works,
where
you
have
components
in
in
the
data
item
that
influence
all
components
that
that
are
topologically
after
that
component
and
that
topological
aspect
is
a
little
bit
complicated
because
maps
are
not
ordered
in
sibo.
So
you
cannot
influence
one
part
of
a
map
from
another
part
of
the
map,
but
you
can
influence
anything
that
follows
that
map.
B
So
this
linear
approach,
which
also
is,
is
in
the
existing
tag,
what
was
the
25,
but
but
that's
even
worse,
because
it
doesn't
define
the
partial
ordering
that
linear
approach
is
in
contrast
to
the
hierarchical
approach,
and
I
think
in
the
in
the
long
run,
we
will
find
applications
for
both.
But
I
think
the
hierarchical
approach
is
going
to
be
the
normal
one
and
the
other
one
is
going
to
be
a
bit
more
specialized.
A
Okay,
that
that
that
seems
to
be
a
good
point
to
to
continue
the
discussion
to
the
to
the
record
proposal,
which
is
on,
which
is
using
that
that
linear
approach
carson.
Do
you
want
to
say
a
few
words
or
shall
I
just
introduce
the
topic
in
first
place?
Please
go
ahead.
A
So
this
is
the
point
this
is
coming
from
is
a
is
a
request
for
is
apparent
asses.
This
is
a
request
for
tiger
location
for
usage,
that
is
by
and
large,
looking
like,
looking
like,
like
csv
or
looking
like,
like
useful
for
all
kinds
of
data
that
are
that
have
repeated
rows
in
a
in
an
existing
grouping.
A
So
the
the
current
proposal
by-
and
I
don't
know
whether
I
pronounce
the
the
name
right
is-
it
probably
is
to
set
up
an
initial
set
of
of
keys
like
basically
like,
like
column
names
and
then
in
the
same
in
the
same
tagged
item.
I'm
sorry,
no
in
in
following
items
in
the
same
array
to
reference
using
a
using
a
number
of
defined
tags,
those
keys
and
then
just
have
the
values.
A
So
if
you're
thinking
of
a
often
say
a
cinema
document,
then
you
will
that's
that
has
the
same
structure
over
all
records
or
as
it
it
can
be
used
here,
even
over
a
large
proportion
of
its
records.
It
could
say
that
the
regular
record
in
kind
of
the
first
kind
of
record
consists
of
a
name
and
a
value
and
a
t
and
then
go
on
just
having
an
array
tagged
with
this
tag.
A
Two
six,
eight,
a
zero
and
items
just
the
name,
just
the
value,
just
the
t
and
rep
and
and
continue
through
with
that.
A
A
Carsten.
How?
How
intensely
has
you
have
you
been
in
contact
with
chris
on
this
so
far
or
was
the
well.
B
Since
the
end
of
last
year,
we
had
an
on
and
off
discussion,
and
it
took
me
a
while
to
pick
up
what
he
was
really
trying
to
do,
because
I
have
specific
things
I
have
been
working
on
that
they
kind
of
shaped
my
my
perception
of
this.
So
you
know
there
is
rfc
8742,
which
is
for
homogeneous
arrays.
B
So
that
was
one
piece
of
misunderstanding
and,
of
course
the
the
other
piece
of
misunderstanding
was,
since
I
was
working
on
on
sibo
effect
at
the
while
at
the
time
I
I
thought
this
was
about
compression
only
about
a
more
compact
representation,
but
it
seems
what
what
chris
really
has
in
mind
here
is
also
a
semantic
component.
So
you
could
have
some
some
additional
semantics
for
each
of
those
tags.
He
doesn't
show
how
to
define
that
semantics.
B
Maybe
that
semantics
actually
can
be
completely
derived
from
the
record
definition
that
is
in
in
the
tag
105,
but
I
think
that's
fundamentally
different
from
the
the
simple
packing
aspect,
and
I
think
that
that's
also
something
that,
in
the
long
run,
will
will
shape
some
of
the
tag
definition
discussions,
because
there
are
a
number
of
common
data
structures
that
people
are
using
and
records
are
among
those
that,
of
course,
can
be
represented
in
sibo
in
some
way.
B
But
you
you
don't
say
that
what
you
just
represented
actually
is
a
record.
So
getting
this
semantic
uplift
in
into
the
tags
is
always
one
thing
that
comes
with
the
enhanced
representation
aspect,
so
in
the
the
long
run.
What
what
I
I
would
like
to
achieve
is
that
we
have
a
little
library
of
of
standard
information
model
level
types
with
tags
that
tell
you
how
to
represent
this
at
the
the
sibo
data
model
level,
so
people
who
come
around
and
say:
oh
I
have
these
records
they.
B
They
know
that
they
can
use
this
existing
structure.
They
don't
have
to
invent
anything
and
people
who
implement
more
comprehensive,
sibo
implementations
can
put
in
direct
support
for
the
the
most
important
types
of
those
applications.
Don't
have
to
do
that
again.
B
So
when
you,
when
you
look
at
language
specific
formats
like
like
the
birch
stuff
that
that
comes
with
erlang
or
the
the
haskell
30
stuff
and
so
on,
they
all
have
done
this
kind
of
work,
but
focused
on
their
language,
of
course,
and
yeah,
depending
on
on
the
the
network
affinity
of
the
language.
B
B
Yeah,
so
the
curating
of
those
of
that
tank
collection,
which,
as
as
you
said,
is,
has
already
become
pretty
large
because
people
have
been
using
zero
for
all
kinds
of
applications.
The
curating
of
that
tank
collection
in
the
end
will
define
how
useful
cbo
is
for
for
writing
protocols
in
specific
application
domains.
A
I'm
still
trying
to
understand
a
bit
what
you
said
about
the
the
records
being
being
defined
by
the
application
and
having
their
semantics.
Does
this
relate
to
how
we
recently
talked
about
the
the
enums,
where
you
have
something
that
is
defined
by
the
application,
but
it
doesn't
really
tell
zebra
what
it
is.
A
Would
would
that
be
some
style
of
these
of
these
record
types
or
is
this
or
are
they
more
really,
I
mean,
or
is
the
expectation
that
people
will
really
define
their
own
record
types
as
dedicated
texts
here,
in
which
case
the
question
is:
do
they?
Why
do
they
need
those
introductory
kind
of
column,
name,
descriptions
in
first
place.
B
Well,
the
the
column
names
are
useful
if
you
are
sending
this
to
an
implementation
that
doesn't
have
the
the
record
definition
predefined.
So
you
you
still
can
represent
this
as
a
classical
json
structure
or
as
a
csv,
or
things
like
that.
So
the
the
sending
that
definition
with
the
data
is
is
useful
to
be
more
schema-less
in
effect,
but
at
the
same
time
adding
some
some
application.
Semantics
may
make
it
easier
for
for
specific
applications
that
that
have
specific
support
for
records
to
make
use
of
that.
A
That
that
sounds
to
me
a
bit
like
like
there's
a
lot
of
of
kind
of
soul
searching
to
go
on
in
in
in
what
what
is
really
the
application.
This
is.
This
is
trying
to
this
is
trying
to
target
and
maybe
also
see
whether
this
can
really
be
supported
by
a
by
a
single
solution,
because,
if
it's,
it
seems
to
me,
like
others,.
C
Some
of
it
would
be
gratuitous.
I
guess
the
question
is
whether
or
not
the
data
survives
the
round-tripping
and
not
the
metadata,
and
I
think
that's
that's.
Those
are
that's
an
important
distinction,
but
you
know
not
the
only
it's
not
the
only
consideration.
C
I
don't
know
if
you
guys
heard
the
the
covet
uk
covid
data,
which
was
round
tripped
in
and
out
of
csv
and
xls
and
whatever,
and
resulted
in
losses,
precision
right
and
it
was
absolutely
completely
incredible
what
they
were
doing,
why
it
was
like
you
know,
but
but
it's
a
good
example
of
what
happens
if
you
don't
survive
round-tripping
with
the
metadata,
but
I
I
just
think
that
in
that
case
they
lost
precision
but
in
other
and
they
exceeded.
Let
me
see
they
that's
the
case.
They
exceeded
the
number
of
columns.
C
That's
a
problem
right,
so
that's
not
a
problem
that
we
would
have
had.
If
from
that,
we
would
have
lost
metadata
about
something,
but
we
wouldn't
have.
You
know,
lost
data.
C
Yeah,
that's
the
point
I
think
is
that
there
and
and
and
so
okay,
so
you
round
tripped
it
and
then
you
lost
this
metadata.
C
So
a
you
could
go
back,
you
could
say:
oh
that's
a
problem,
but
that
just
means
I
shouldn't
go
in
and
out
of
this
program
I
need
to
just
go
in
and
the
other
program
needs
to
take
it.
You
know
some
other
thing
or
I
need
to
improve
the
program,
but
the
med
but
the
metadata.
So
it
would
be
unfortunate
if
the
metadata
disappeared
and
you
never
got.
You
never
saved
it
right.
You
overwrote
the
file,
but
on
the
other
hand
it
wouldn't
be
catastrophic,
is
what
I'm
trying
to
say.
C
Metadata
so
a
thing
that
that
occurs
to
me,
that
would
benefit
from
the
metadata
and
wouldn't
be
application.
Specific
is
something
like
a
graphing
or
a
statistical
program
which
is
just
going
to
use
the
metadata,
probably
to
label
the
axes
and
stuff
like
this,
but
otherwise
there's
no
specific
knowledge
of.
What's
in
it
right,
yep.
A
Although
if
the
distinction
is
made
between
different
kinds
of
records
that
just
happen
to
have
the
same
name,
then
that
might
end
up
either
in
unfortunate
coins
coins,
coincidence
of
x's
or
very
practical
coincidence
for
xs,
depending
on
how
it's
used.
C
B
Okay,
so
I
think
that
that's
one
aspect
of
this
and
the
other
aspect,
of
course,
is
the
the
linear
aspect.
So
in
in
the
proposal
that
I
immediately
came
up
with
that
that
the
csv
like
proposal
there,
you
would
have
a
root,
provide
the
structure
and
the
yeah.
Thank
you
and
the
the
data
that
is
inside
that
structure
inside
the
data
item
gets
the
structure,
but
not
necessarily
other
places.
B
Of
course
doing
it
in
a
linear
way
means
that
you
can
encode
things
and
when
you
come
along
to
something
that
that
needs
this,
you
just
put
in
the
definition
and
then
you
can
can
use
it.
So
you
can
do
this
in
a
more
dynamic
way.
You
don't
have
to
ascertain
that
the
the
whole
thing
uses
the
same
structure
that
you
don't
have
to
ascertain
that
this
is
homogeneous.
B
On
the
other
hand,
for
for
someone
decoding
it,
it
may
actually
be
valuable
to
know
upfront
that
that
the
whole
thing
is
homogeneous.
So
there
are
two
different
kinds
of
applications,
and
I'm
not
sure
we
we
have.
B
We
always
have
a
strong
preference
here,
but
if
we
want
to
support
the
linear
way,
then
then
we
need
to
do
one
thing
which
is
defining
that
partial
ordering.
B
So
people
know
which
structures
that
that
follow
the
definitional
instance.
The
definitional
item
can
actually
make
use
of
that
data
safely.
B
Right
but
that
area
might
be
in
a
map,
and
so
you
can
build
more
complex
constructs
where
it
suddenly
does
become
important
to
think
about
the
partial
order.
A
But
might
it
not
suffice
to
only
define
that
order
within
within
the
top
level
structure
and
there
say
okay
yeah
we
can
have.
I
mean
the
proposal.
I've
read
from
chris
so
far
sounds
like
it's
more
local
code,
so
you
can
define
something
inside
the
area
and
that's
valid
later
on,
but
it
won't
be
valid
outside
that
area.
A
A
B
Yeah,
so
there
is
this
section
in
8949
where
we
say
it's
not
a
good
idea
to
rely
on
the
actual
encoding
sequence.
For
that,
like
tag
25
does,
and
so
this
is
not
recommended,
but
the
text
in
8949
doesn't
tell
you
what
actually
is
recommended
if
you
are
going
for
the
linear
the
case.
So
writing
that
up
at
some
point
is
probably
a
good
idea.
If
you
want
to
support
the
linear
approach
at
all.
B
Question
so
one
is
of
course,
the
the
notable
tag
documents,
but
that's
a
grab
bag
and
then
there's
so
much
stuff
in
there
that
it
won't
be
published
after
another
couple
of
years
or
so
so,
maybe
that's
not
the
best
place
to
to
do
something
that
other
specifications
want
to
essentially
normatively
reference.
Even
if
those
specifications
are
just
specifications
that
are
attached
to
a
tag.
Registration.
A
Yeah
and
I'm
I'm
not
sure
they
would
look
there,
because
if,
as
long
as
this
document
is
largely
for
for
this
use,
use
that
and
for
that
use
use
that
that's
not
intuitively
something
that
where
you
would
look
for
guidance
for
doing
yet.
Another.
A
A
B
So
I
think
this
would
be
a
little
bit
out
of
scope
for
them.
Okay,.
B
B
Yeah,
so
what
my
personal
view
was
that
the
the
sibo
working
group
does
essentially
two
things
and
one
is
curating
their
tank
space
and
and
making
sure
that
that
the
batteries
that
are
included
with
c-ball
actually
take
you
to
all
the
places
where
you
need
them
to
take.
B
You,
and
the
other
thing
is
the
cddl
language
and
and
making
sure
that
this
covers
all
the
bases
that
we
can
get
a
cgi
2.0
going,
but
also
that
we
curate
the
control
operator
space
that
is
defined
in
there
and
that's
maybe
an
interesting
observation
that
we
have
done.
Two
protocols
c
and
and
cddl
that
actually
are
so
extensible
that
it's
useful
or
even
important,
to
actually
curate
that
that
those
the
space
that
those
extension
points
provide.
A
Publishing
publishing
guidance
for
how
how
to
extend
the
space
probably
stands
a
good
chance
of
falling
into
the
curating
point
of
the
of
the
agenda
of
the
on
the
charter.
B
A
A
B
Cinema
packs
are
sequences
of
records
which
are
not
homogeneous,
but
there
are
only
going
to
be
a
small
number
like
a
dozen
or
so
different
records
in
in
many
different
kinds
of
records
in
many
sentiment
packs
so
essentially
saying
ahead.
These
are
the
kinds
of
records
we
have.
I
was
saying
within
the
the
sequence
of
records.
Oh
here's
a
new
one
and
we're
going
to
use
this
henceforth
that
that
might
actually
fit
cinema.
It
might
have
fit
cinematic
very
well.
A
Yeah,
and
probably
many
of
them
might
also
have
a
case
which
really
didn't
come
up
here
yet
in
in
this
discussion,
where
they
have
some
of
the
shared
elements,
and
one
more
so
there
could
be
in.
If
we're
looking
at
the
potential
representations.
Again,
when
going
with
the
explicit
form,
they
could
be
really
used
for
using
just
one
item
or
three
items
in
a
single
record
and
then
just
kind
of
using
names
as
they
come
along.
A
A
B
Yeah,
so
I
think
that
that's
a
good
idea
throwing
cinna
against
this
proposal
and
see
what
happens.
A
A
A
Later,
michael,
if
you're
talking,
we
don't
hear
you.
A
C
Session
hi,
sorry,
you
were,
I
received
a
phone
call
as
you
were
pinging
me
hi.
So
there
was
some
discussion
about
whether
we
would
like
to
have
a
way
to
indicate
an
ip
in
a
full
ipv6
address,
meaning
all
128
bits,
plus
a
prefix
that
you
would
use
to
describe
that
network
that
it's
on,
and
so
that's
more
akin
to
what
you
feed
into
ifconfig,
for
instance,
as
opposed
to
how
you
describe
the
network-
and
I
don't
know
whether
that's
a
useful
extension
or
what
the
opinion
of
others
are.
C
I
know
carsten
has
made
a
suggestion
as
to
a
syntax
and
ole
tron
thought
that
that's
what
that
was
include
was
included,
and
I
guess
I
feel
I'm
looking
for
guidance
as
to
what
the
working
group.
B
Feels
yeah,
so
the
the
sequence
of
events
was
that
that
ola
tuan
looked
at
this
and
said
how
about
this
and
yeah.
I
kind
of
agreed
that
having
this
interface
configurations,
where
you
have
an
address,
plus
the
number
of
bits
to
take
out
of
the
address
to
derive
the
the
prefix
of
the
network,
I
agree
that
this
is
a
very
common
application.
So
when,
when
we
address
addresses
and
prefixes,
we
might
as
well
address
the
the
interface
configuration
kind
of
combination.
C
Yeah,
I
think
that's
a
good
word
for
it
interface
configuration
the
thing.
Is
that
there's
some
other
stuff
that
that
usually
you
wind
up
with
the
interface
configuration
that
would
include
the
default
route,
maybe
name
servers,
and
so
what
I'm
trying
to
say
kind
of
hint
here
is
that
maybe
the
person
who
has
that
problem
that
they
want
to
store
interface,
configurations
or
communicate
them
has
some
other
data.
They
want
to
communicate
as
well,
and
so
maybe
it's
really
a
higher
level
concept
than
what
we're
describing
here
is.
C
C
Is
this
we
would
lose
the
distinction
in
in
this
thing?
Yes,
so
so
at
this
point
my
preference
would
be
to
not
solve
the
problem
and
simply
say
this
is
something
we
do
not
address
and
someone
else
should
address
with
some
other
tag
and
but
I'm
I
don't
feel
strongly
one
way
or
the
other
I
could
write
put
the
text
in.
I
just
don't
know
that
I
could
do
a
good
job
of
explaining
it
clearly.
A
So
if,
if
someone
were
to
put
all
these
routing
information,
routing
information
and
dns
and
so
on
into
into
their
own
format,
do
we
have
everything
there
so
that
they
don't
have
to
reinvent
things
that
we
already
described
here.
C
Well,
they
would
of
course,
have
a
a
network
configuration
which
would
be
a
prefix.
They
would
then
have
a
may
have.
They
could
either
described
an
iid
to
put
on
the
lower
64
bits
or
they
could
describe
the
whole
120
bits
that
you
should
configure
they
dns
dns,
reverse
bv4
v6
tagged
as
appropriately
a
default
route.
You
know
you
could
have
a
default
route
listed,
a
number
of
them
in
v4
or
v6
off
versions,
so
you
could
have
a
bag
of
those
or
an
array
of
them.
C
So
there's
a
whole
bunch
of
ways.
But
to
me
it
just
sounds
like
it's
a
that's
it's
at
this
point.
It's
just
a
map
with
a
bunch
of
keys
and
for
all
I
know
it's
already
in
a
yang
con
yang
model
somewhere.
In
fact,
I
bet
you.
The
interface
description
is
already
in
a
yang
model
somewhere.
B
D
C
Such
as
in
an
rpki-
yes,
so
yes,
okay,
so
first
of
all,
they
would
be
describing
v4
prefixes
and
v6
prefixes
yeah.
So
they
don't
need
interface
descriptions,
so
they
wouldn't
care
about
trailing
bits
if
they
needed
to
indicate.
I
don't
know
the
address
of
their
rpki
server.
I
don't
think
we
actually
have
that,
because
we
don't
delegate
really,
then
that
would
be
just
an
address
there
as
numbers.
C
D
That
would
be,
if
you're,
actually
stating
the
who
can
originate
routes
for
that
address
block,
but
so
you've
got
all
the
way
to
the
row
up
as
opposed
to
sure.
D
Get
the
yeah,
but
yes,
as
long
as
you
could
tell
the
difference
between
the
ais
number
or
an
as
number
range.
Then
I
think
you're,
okay,.
C
Yeah,
so
we
don't
have
anything
specific
to
deal
with
the
as
number
tagged
or
as
number
range
okay,
but
I
think
that
that
would
fit
into
the
structure
of
the
thing
you
would
say:
they're
on
the
left
or
on
the
right
or
or
whatever
or
yeah.
Maybe
you
need
a
tag
for
as
number
ranges
that
would
be
different.
C
Oh
okay,
I
mean,
I
guess
I
can
see
how
ayanna
would
hand
them
out
to
rars
that
way,
but
but
anyway,
I
I
said:
there's
nothing
there,
there's
nothing
there
to
support
or
or
not,
support
as
numbers
or
as
number
ranges,
but
on
the
brief
on
the
the
ip
address
side,
I
think
it's
complete
perfect.
Thank
you.
D
When
I
read
the
document,
I
was
thinking
about
this
in
the
back
of
my
mind
and
I
didn't
catch
anything
on
the
address
block
side.
C
So
I
I'm
not
hearing
I'm
gonna
write
an
email
back
and
say
I'm
not
hearing
overwhelming
support
for
putting
this
in
now.
If
someone,
if
someone
thought
that
they
needed
to
do
this
in
three
years,
they
could
either
define
a
new
tag
or
they
could
define
a
new
variation
that
for
this-
and
I
I
think
this
is
not
something
that
we
should
do,
but
I
could
be
wrong
so
I'll
write
an
email
about
that.
B
Sounds
like
good
well
can't
be
as
disagreeing.
B
I
didn't
hear
that
sorry.
Well,
I
don't
agree
so
the
the
reason
I
don't
agree
is
it
immediately
came
to
the
mind
of
ola
because
he
knows
that
people
need
that,
and
I
also
had
the
little
bit
of
a
face
palm
here
by
that.
B
Essentially,
I
didn't
think
of
it
when
I
looked
at
this
for
the
first
time
and
yeah,
so
so,
waiting
for
the
problem
to
come
is
is
like
assuming
that
that
that
that
problem
may
never
come,
but
that
it
came
to
the
mind
of
all
this
immediately,
and
I
must
agree
that
something
that
people
want
to
do.
C
I'm
not
I'm
not
saying
I
will
not
doing
it.
I'm
saying
that
I'll
write,
an
email
saying
that
you
think
it's
a
good
idea,
I'm
on
the
fence
on
this
okay
and
I'm
gonna,
ask
the
questions
I
just
asked:
wouldn't
there
also
be
some
additional
network
configuration?
C
What
is
the
use
case
for
this,
and
is
it
bigger
than
just
this
part
and
and
basically
I'm
going
to
say
you
know,
please
give
me
a
a
an
argument
for
why
we
should
do
it
now
or
why
it's
only
as
big
as
the
problem
that
we
have
right,
because
I
think
the
people
that
want
to
use
it
have
a
bigger
have
a
bigger
solution.
They
want
to
fit
it
into.
B
Yeah,
so
maybe
calling
it
interface.
Configuration
was
really
confusing
issues
here,
because
you
are
right.
An
interface
configuration
has
lots
more
things
in
it,
but
I
still
think
that
that
the
combination
of
an
address
with
a
prefix
that
address
is
in
is
a
rather
common
situation
and
why
we
are
allocating
one
plus
one
byte
tags.
C
Okay,
I
will
maybe
throw
the
email
at
I
know
the
lisp
working
group.
Maybe
they
have
some
caring
about
that.
I
suspect
that
occurs
to
me
that
they
may
have
some
thing
they
care
about
that
way
and
and
then
I'll
ask
the
question:
is
it?
Is
it
something
other
than
an
interface
configuration?
B
C
Okay,
I
I
take
your
point,
I'm
I'm
I
I
think
just
I
think
we
just
need
a
term
to
to
to
describe
it
right.
B
B
B
Yeah,
so
the
interesting
thing
is
that
these
tags
are
registered,
so
so
people
can
start
marching
ahead
using
what
what's
in
the
document
now,
and
if
we
add
this
address
blood
plus
prefix
thing,
then
it
will
make
it
slightly
more
useful
for
certain
application
but
yeah
it's
not
stopping
people
from
from
using
that
now.
A
One
one
thing
that
came
to
my
mind
on
the
on
the
context
of
these
things
are
probably
also
in
used
in
in
some
young
models.
Would
it
make
sense
to
cross-reference
those
from
the
from
the
description
just
to
kind
of
to
have
all
the
links
around
saying
that
there
are
other
contexts
in
which
this
is
used
and
they
are
compatible
or
not
compatible
because
they're
described
and
they're
described
here
and
there,
especially
when
people
come
for
this,
because
they
want
to
do
really
full
interface
configurations.
They
will
come
looking
at
this
so
yeah.
A
Here
I
have
a
former,
but
if
they
read
the
full
text,
they
might
find
that
there
is
something
that
helps
them
solve
their
larger
problem
and
in
passing
all
those
sources.
B
Yeah,
so
yang
comes
with
yang
plus
0,
so
that
there
is
a
defined
way
to
actually
represent
this
information
in
in
yang
there's
a
little
problem
as
a
really
sibo
tag
is
where
a
yang
data
specification
would
be
so
the
the
existing
specifications
that
are
not
describing
data
on
the
wire,
but
the
data
at
rest.
B
Yeah
transfer
to
to
make
make
use
of
them
in
a
protocol,
but
I
think
that
that's
pretty
obvious-
and
maybe
at
some
point
we
actually
should
as
as
was
discussed
when
the
asdf
working
group
was
started.
B
A
Together,
but
on
the
more
concrete
side,
is
there
other
other
specific
date,
data
descriptions
that
we
could
point
to
here
without
without
the
the
fully
generic
solution
to?
How
is
the
sea?
How
is
a
yang,
how
can
a
yang
item
and
and
the
zebra
tag
be
mapped
to
each
other
kind
of
coming?
Coming
from
the
practical,
practical
consideration
that
someone
comes
along,
wants
to
implement
their
ifconfig
on
cbr
finds
this
hops
on
it
and
could
have
a
useful
forward
useful
reference
to
to
the
whole
yank
topic.
A
Okay,
then,
any
other
comments
on
the
network
addresses
part
that
we
should
at
least
take
into
consideration,
arguing
that
we're
approaching
the
end
of
the.
A
Well
with
that,
thanks,
michael
for
taking
notes
I'll
upload
them
later,
when
the
video
is
ready
as
well
and
yeah,
thanks
for
the
good
productive
discussion
and
see
you
all
in
two
weeks
and
really
on
the
mailing
list.
Thank
you.