►
From YouTube: IETF-CBOR-20210630-1400
Description
CBOR interim meeting
2021/06/30 1400
A
A
A
We
have
a
short
agenda
for
today
that
is
going
through
documents
that
have
changed
all
that
sorry
yeah
documents
that
have
changed
or
have
rece
have
had
received
promises
to
report
something
about
them.
On
network
address.
We
have
network
addresses
with
a
few
new
changes
in
there
file
magic
with
pending
changes
with
respect
to
to
implement
your
advice
and
someone,
as
it
looks
like
right
now.
Probably
me
will
report
from
the
coral
implemented
work
implementary
team
to
talk
about
how
we're
planning
to
use
siebel
pact
there.
A
B
A
Then,
let's
get
started
with
network
addresses.
This
has
two
new
additions
since
the
last
meeting,
one
being
that
there
is
now
cddl
in
place
that
helps
especially
writers
of
other
specifications,
to
include
a
network
address
and
there's
also
an
edition
about
addresses
with
prefix
the
the
thing
we've
talked
about
a
lot
in
the
context
of
describing
a
subnet,
but
it
might
be
useful
to
have
an
address
in
there
too.
A
B
But
I
just
dialed
in
so
I
have
no
idea
what
you
have
said,
but
yeah.
Essentially
we
had
that
discussion.
We
it's
not
a
slam
dunk,
because
it
does
increase
the
complexity
of
this
draft.
On
the
other
hand,
it's
something
that
that
I've
actually
used
a
lot
in
configuration
files,
so
I
expect
it
really
useful.
D
Yeah,
so
I
just
posted,
I
just
shared
the
two
things,
so
this
is
the
cddl
addition
to
the
draft
and-
and
this
is
the
one
of
the
two
parts
of
the
adding
the
thing
carson.
I
know
what
you
mean
by
by
this
sentence
here
down
here,
but
I
wondering
whether
it's
completely
understandable
to
people
whose
english
is
less
good
now.
D
B
D
E
D
D
D
So
basically,
I'm
saying
that
you
should,
if
you
are,
if
you
normally
would
be
processing
many
objects
in
a
file,
then
it's
probably
appropriate
to
put
a
seabor
sequence
in
otherwise
the
seaboard
tag
is
probably
less
invasive
if
because
it
doesn't,
doesn't
necessarily
require
any
changes
to
any
other
users
of
the
file.
On
the
other
hand,
if
you
are
the
first
creator
of
the
file
and
user
of
the
file,
then
it
may
be
quite
arbitrary
as
to
what
you
create.
A
How
how
about
this?
To
be
honest,
I'm
not
fully
sure
if
the,
if
the
addition
of
the,
if
the,
if
the
registration
for
the
tags
for
the
content
formats
are
in
there-
and
I
know
that
I
should
probably
given
that
I
requested
this.
I
should
probably
review
this
pull
request
within
this
week,
so
that
that's.
A
A
Is
this
purely
something
that
one
would
do
when
receiving
when
receiving
data
over
the
network
in
a
content-formatted
way
and
wanting
to
preserve
that
information
on
disk?
Or
is
this
something
that,
under
any
circumstances,
someone
would
do
in
in
their
own
file
format
because
they
are
registering
attack
for
it
anyway?
I.
D
I
I
would
say
that
that
you
asked
two
questions
that
and
I
don't
think
either
of
them
are
addressed.
Actually
the
first
thing
you
asked
is:
would
you
translate
content
format
or
mime
types
that
you
received
over
the
net
to
a
tag
to
put
onto
the
file
or
vice
versa?
Would
you
create
content
format
as
a
server
based
upon
tags
that
you
find
in
the
file?
I
would
say:
that's
entirely
appropriate,
but
would
be
specific.
D
You
know
to
co-app
or
http
and
would
be
completely
inappropriate,
for
I
don't
know
smtp
right,
so
that
really
belongs
in.
I
think
in
a
in
the
document
about
that
protocol
we
could
make
some
suggestions
about
co-app
or
pat
since
co-app's
already
published.
Http
is
already
published.
We
can
make
some
suggestions,
but
the
specific
thing
that
I
understood
from
the
content
format
was
that
for
all
of
the
things
that
we
have
a
content
format,
we
essentially
already
have
a
tag
allocated,
and
so
you
don't
need
to
go,
find
a
new.
D
D
So
so
what
all
has
said
in
this
is
that
if,
if
there's
a
content
collapse,
content
format
number
section
blah
blah
blah
there
and
that,
then
you
don't
need
to
allocate
a
new
number.
That's
all
that
really,
I
think,
was
the
point
of
of
we
said
so
far.
We
didn't
say
anything
about
whether
you
should
translate
back
and
forth
on
or
off
disk
right.
So
if
you
think
that
we
should
say
that,
then
that
would
be
and
to
me
a
different
discussion,
because
it
has
nothing
to
do
with.
D
Obviously
it's
hard
to
translate
to
and
from
content
format
if
there
isn't
a
content
format.
But
but
you
know,
if
you,
if
you
did,
if
you
did
do
a
new
thing
and
you
got
a
co-op,
a
seabor
tag
and
you
got
a
mime
type
and
a
content
format,
then
you're
all
set
with
all
of
those
things.
If
you
wanted
to
translate
back
and
forth.
E
B
So
we
still
have
this,
this
old
experiment
that
we
never
completed
into
a
proposal
where
all
media
types
get
a
content
format
number
which
is
easy,
because
there
aren't
that
many
media
types
and
that's
really
something
that
at
some
point
we
need
to
do
that
doesn't
help
with
with
weird
parameters.
But
it
helps
with
with
the
basic
problem
that
just
about
any
media
type
out
there
will
have
a
content
format
number
then
so,
there's
even
a
range
reserved
in
the
content.
D
B
Because
the
the
the
way
this
range
well
you're
right,
somebody
could
just
sit
there
and
and
watch
the
content,
format,
number
registry
and
register
the
the
other
things
but
yeah.
I
think
it
would
be
better
if
this
were
simply
automatic
and
not
relying
on
an
external
entity.
We
actually
don't
have
a
control
over
ssitf.
D
B
B
B
B
D
Oh
okay,
so
all
right,
I,
I
think
it
muddies
the
water
for
this
document,
because
I
think
it
muddies
what
is
a
simple
bcp
which
actually
doesn't
have,
and
it
is
a
bcp
I
think,
with
what
is
in
fact
a
revision
to
an
iana
process,
which
I
think
needs
to
be
a
document
of
a
maybe
it's
also
a
vcp,
but
I
think
it
needs
to
be
documented
in
different
in
a
different
stream.
Almost.
D
F
D
I
mean
it
needs
to
get
a
different
kind
of
reviews.
What
I
really
mean,
I'm
I
I
know
that
that
you
know
most
of
our
process.
Documents
actually
are
bcps,
but
because
they're
standards
track,
but
I'm
saying
that
they're
kind
of
a
bcp
in
a
different
sense
and
that
that
this
document
is
just
here's
a
good
way
to
do
something
right
and
it
doesn't
really
establish
a
standard
necessarily
and
then
the
other
side
is
when
you're
doing
this
process.
F
Mean
yeah
I
mean
it,
so
I
suppose
the
the
the
trigger
the
the
the
pointer
would
be
that
if
this
formally
updates
the
document
that
created
the
registry,
then
I
think
that's
enough
of
a
pointer
to
to
make
it
work
and
yeah
I
mean
bcps,
get
the
same
review
as
standards
track
documents.
So
I'm
not
I'm
not
concerned
about
the
level
of
review.
F
B
B
B
D
B
B
D
E
D
To
always
ask
for
content
format
as
well,
and
they
would
get
one
in
this
in
this.
They
would
get
a
content
format
from
co-app
and
therefore
they
would
get
a
a
tag
from
this
range
as
well.
But
the
point
was
that
they
would
always
ask
for
a
content
format
when
they
did
a
mime
type
registration
and
therefore
we're
updating
the
mime
type
template.
A
B
D
D
D
B
E
D
D
D
I
don't
want
to
take
forever
on
this.
I'm
sorry,
I'm
not
really.
I
don't
really
care
one
way
or
the
other
I
just
I'm.
All
I'm
saying
is
that
it
seems
like
it's
a
single
coherent
thought
and
that
by
spreading
it
around
we're
making
it
in
different
documents,
we're
just
making
it
harder
for
people
to
know.
What's
going
on,
that's
what
I'm
trying
to
say.
B
A
D
F
F
F
A
A
Again,
we've
looked
into
siebel
pact
and
the
current
plan
is
for
coral
to
use
sibo
pact,
with
all
its
kind
of
with
all
semantics
and
and
the
mechanisms
like
defining
dictionaries
and
defining
text
with
the
limitations
that
we
will
need
to
profile
it
to
some
extent,
because
this
is
meant
seaboard
pack.
Coral
is
meant
to
be
read.
Ideally,
single
passed
by
very
constrained
devices
to
look
for
particular
links
in
a
document
that
they
receive
or
to
to
understand
and
populate
forms,
and
this
will
need
limitations
with
respect
to.
A
Where
can
where
can
attack?
Where
can
a
compressed?
Where
can
a
compression
tag
be
used
and
where
not,
and
can
can
can
prefix
compression
be
used
in
which
formants
and
all
those
things
right
now
we
are
moving
towards
removing
from
from
those
old
numeric
identifiers
towards
packed
seabor
without
any
constraints
applied,
but
we
will
look
into
the
choral
documents
that
we
get
out
of
it
and
the
mechanisms
we
use
to
later
decide.
A
One
thing
with
one
thing:
where
we'll
certainly
do
a
bit
of
experimentation
that
might
be
used
for
useful
for
simple
pact
in
more
general
is
where
to
get
the
dictionaries
from,
because
we
have
like
all
all
the
three
major
cases
in
in
choral
or
at
least
we're
considering
all
the
major
cases.
That
being
it
is
known
from
the
context,
for
example,
because
we're
using
a
specific
particular
particular
particular
content
format
that
just
has
these
pre-set
up.
A
It's
often
very
easy
to
set
up
something
that
will
when
unpacked
looks
like
terabytes
of
data
and
depending
on
depending
on
which
view
one
takes,
and
we
may
need
a
few
precautionary
measures
or
not,
because
if
it's
purely
a
syntactic
thing
then
the
parser
is
supposed
to
know
about
this
and
whatever
its
own
limitations
would
apply.
Sorry,
the
the
the
coral
processor
would
know
that
at
this
point
there
could
be
a
attack
and
that
could
lead
to
some
recursion.
A
B
So
I
I
don't
really
see
this
alternative
either
expanded
during
passing
or
hand
at
the
application
level.
So
between
them.
There
is
something
where,
where
the
parser
generates
a
dom
like
structure
that
actually
contain
references,
so
it
would
be
the
the
dom
access
functions
that
actually
look
inside
and
and
see.
Oh,
is
this
real
data,
or
is
it
a
reference
to
something
else,
and
that
would
enable
the
application
to
actually
maybe
use
an
a
variant
of
the
dom
access
function
that
exposes
whether
you
have
a
reference
or
an
actual
value.
B
A
B
B
A
B
Not
yeah,
and
we
already
have
the
the
problem
that
zebra
pact
can
have
cyclic
structures,
so
michael
made
a
proposal
to
make
that
impossible,
but
I
think
the
result
of
discussing
that
was
that
this
restricts
the
the
forms
of
compression
you
can
actually
get
from
packing
more
than
we
may
want
to.
So
anything
that
actually
follows
this.
B
B
B
Yeah,
so
the
assumption
would
be
that
we
we
can
come
up
with
a
rule
of
thumb
that
that
or
you
probably
need
much
less
than
40
steps
in
most
of
our
applications.
But
we
could
still
come
up
with
a
relatively
high
number
like
40
and
say
any
implementation
should
keep
account,
and
if
it
goes
up
to
40,
then
it
needs
to
stop
doing
its
strengths.