►
From YouTube: IETF-CBOR-20230823-1400
Description
CBOR meeting session at IETF
2023/08/23 1400
https://datatracker.ietf.org/meeting//proceedings/
A
B
C
Yeah,
the
the
new
interface
has
a
much
better
use
of
screen
real
estate,
so
I
was
trying
to
use
it
as
much
as
possible.
It
didn't
have
all
the
chair
functions,
but
it
was
really
nice
as
a
participant.
D
D
A
I
know
nothing,
nothing
of
whether
Anders
plans
to
be
here
or
not.
A
So
let's
go
ahead
and
get
started
Carson
the
slides
are
yours,
go
for
it.
C
Thank
you.
So
there
are
two
items
on
the
agenda
today:
the
diagnostic
mutation
and
the
deterministic
encoding
discussion
I
have
one
slide
that
actually
looks
at
other
documents.
C
C
So
we
are
still
behind
comfortably
behind
that
document,
but
we
probably
should
should
do
what
we
discussed
in
117
and
then
actually
do
a
publication
requested.
Yes,
on
this
document,.
A
And
the
time
yeah
I'm,
that's
one
of
the
things
I'm
behind
on,
but
I've
been
meaning
to
get
time
tag
into
production
into
publication
requested.
So.
C
Yeah,
thank
you.
So,
on
the
sibo
package
side
I
think
we
we
still
have
validation
by
implementation
on
our
schedule
and
the
use
case
dnsc
board.
So
as
vacation
time
is
slowly
going
to
an
end
in
Germany
I.
Think
Martin
will
also
have
more
data
to
look
at
so
I
think
we
should
look
at
that
at
one
of
the
next
interims.
C
So
that's
the
not
this
time
slide.
So,
let's
just
quickly
recap
with
what
most
of
you
probably
know.
Sibo
is
a
representation
format
and
it
comes
with
two
languages:
the
diagnostic
notation,
which
allows
us
to
look
at
c-bar
without
having
to
decode
a
binary
or
X
and
cddl,
which
allows
us
to
describe
data
models.
And
originally
we
said
the
diagnostic
notation
is-
is
really
a
second
class
citizen.
C
We
don't
even
Define
a
grammar
for
it
and
so
on,
and
it
seems
that
there
are
now
more
applications
where
it
makes
sense
to
do
that.
So
the
document-
the
edn
literal
document
that
originally
just
was
meant
to
provide
a
way
to
do
domain-specific
literals
in
diagnostic
notation
that
actually
has
acquired
an
ABN,
F
Syntax
for
Diagnostic,
notation
and
and
some
more
information
about
Iran
versus
cddl.
C
These
two
languages
unfortunately
have
very
different
routes,
so
they
they
are
not
similar
in
places
where
they
could
be
similar,
so
that
I
think
that's
useful
to
to
document.
C
I
wrote
the
ABN
f
a
couple
of
months
ago
and
when
I
looked
at
it
again,
I
immediately
found
a
bug.
So
that's
already
on
the
GitHub
as
a
for
request
and
I
think
we
need
to
work
on
on
the
the
other
GitHub
issues
that
have
been
raised
and
and
then
maybe
go
into
reviews.
C
C
Of
course,
we
already
have
a
common
syntax,
but
as
as
Christian
wrote
in
issue
number
eight
that
that
has
a
problem
with
having
to
double
up
slashes
in
in
certain
cases.
C
So
it
would
be
nice
to
to
have
a
way
to
handle
that
one
would
be
to
have
something
like
like
the
slash
star
syntax
of
C
but
yeah.
There
are
a
few
problems
with
putting
this
into
ABN
F,
so
maybe
the
other
thing
that
has
been
asked
for
repeatedly
our
end
of
line
comments,
and
there
are
three
candidates
for
End
of
Line
comments.
C
The
the
hash
mark,
the
semicolon,
which
would
give
some
commonality
with
a
CDL
and
the
double
slash,
which
would
give
some
Community
with
JavaScript
and
looking
at
various
documents,
even
actual
rfcs
hash
mark
and
semicolon
have
been
seen
in
the
wild,
even
though
they
are
not
not
officially
part
of
the
syntax.
So
maybe
we
should
just
legalize
them.
C
The
the
double
slash
would
be
more
like
JavaScript,
but
it
clashes
with
the
existing
slash
syntax,
so
I
I'm
not
sure
we
want
to
do
that.
So
I
would
probably
propose
going
for
hash
and
semicolon,
or
just
one
of
them,
I'm
completely
open
here,
but
make
sure
we
have
one
form
of
End
of
Line
comments
and
then
it's
much
easier
to
to
add
information.
C
So
that
would
be
my
proposal
going
forward
on
the
trailing
commas.
Json
famously
creates
an
editing
problem.
Each
time
you
you
Edge,
align
or
delete
a
line
at
the
end
of
a
Json
array
or
a
map
you
forget
to
add
or
delete
the
comma
and-
and
that's
just
a
lot
of
wasted
time
so
enabling
training
commas
has
been
requested.
A
lot
and
I
have
written
a
PR
which,
as
with
the
other
ABN
F,
still
needs
to
be
tested,
but
maybe
people
can
can
have
a
look
at
that.
C
So
these
These
are
essentially
the
the
user.
Visible
features,
there's
also
a
maintenance
issue
for
some
reason:
hex
floating
Point
didn't
make
it
into
diagnostic
notation
and
that
that
is
often
useful
when
working
with
exact
floating
Point
numbers,
so
I
propose
adding
that.
So
that
would
also
be
a
user
visible
feature,
but
not
one
that
we
inherit
from
from
Json
groups,
and
there
are
also
a
couple
of
ABN
F
fixes
that
are
out
there
as
for
requests
and
that
that
you
just
just
need
to
be
checked
and
applied.
C
And
finally,
there's
the
editorial
issue
that
that
we
never
mentioned
that
that
the
basic
integer
format
without
dots
or
exponents
gives
you
integers
in
diagnostic
notation
and
anything
with
plots
or
exponents
gives
you
a
floating
Point
values
that
that's
implicit.
If
you
look
at
appendix
a
of
RC
7049-
that's
already
in
there,
but
we
not
never
actually
thought
about
writing
it
down,
because
it
was
so
obvious
to
many
of
us
and
that's
of
course,
not
so
great.
C
Okay,
so
we
lost
Christian
I,
think
so
what
I
think
we
oops
I
think
we
need
to
do
here
is
get
some
reviews
on
on
the
GitHub
issues
and
pull
requests.
I
didn't
flag
anyone's
specifically,
but
please
put
in
your
reviews.
If
you
find
something
and
and
understand
it
and
like
it
or
don't
like
it,.
A
E
Thank
you,
I
am
in
favor
of
adding
and
of
defining
the
use
for
End
of
Line
comments
of
both
pound
and
semicolon,
because
one
or
the
other
are
very
widely
used
in
a
variety
of
rfcs
and
and
it
seems
harmless
to
enable
both
and
perhaps
to
make
a
recommendation
for
the
use
of
one
over
the
other,
and
for
that
Carson
and
Barry
and
Marco
I
are
Christian.
I
would
suggest
you
guys
chime
in
I'm,
strongly
opposed
to
adding
double
slash,
because
it
just
adds
too
much
ambiguity
and
too
many
contents.
E
So
that's
that's
that
one.
Thank
you
and
thank
you
for
noticing.
Hex
voting
point.
We
needed
to
be
explicit
about.
C
Okay,
so
I
go
to
the
detronistic
encoding
documents,
so
the
the
mailing
list
has
pretty
much
exploded
in
the
last
two
months
with
discussion
about
the
gymnastic
encoding.
When
we
originally
put
this
into
RFC
7049,
we
thought
well,
nobody
is
going
to
use
that,
but
it's
so
easy
to
do.
Why
don't
we
put
it
in?
C
So
that's
why
it's
in
there
and
now
we
have
increased
attention
to
this-
probably
at
least
partially
caused
by
the
the
need
to
do
privacy,
oriented
data
formats
where
parts
of
data
structure
are
blanked
out
and
replaced
by
hashes
and
and
that's
difficult
to
do.
If
you
don't
have
deterministic
encoding,
but
of
course,
all
the
old
use
cases,
some
of
which
we
may
or
may
not
like
very
much
are
there
as
well.
C
So
that's
what's
going
on
and-
and
this
has
been
push
forward
by
by
wolf,
mcnelly's
deterministic
sibo
document,
a
number
of
other
documents
are
there.
Unders
has
written
another
document
which
does
one
thing
well.
It
provides
a
lot
of
examples
which
we
will
have
to
put
into
the
other
documents
or
with
the
ones
that
we
want
to
go
forward
with.
C
But
it's
not
quite
clear
what
this
document
does
and
it
doesn't
help
that
that
there
have
been
like
22
revisions
of
this
in
the
last
couple
of
weeks,
so
I
really
have
to
look
up
which
revision
is
currently
active
to
find
out
what
what's
in
there
anyway.
C
So
this
is
maybe
a
useful
query
for
things
that
should
go
into
one
of
the
other
documents,
and
this
has
also
submitted
that
as
a
document
to
the
independent
stream,
which
I
think
is
not
good,
because
we
generally
don't
take
documents
out
of
the
middle
of
working
with
proceedings
and
make
independent
stream
rfcs
out
of
them.
A
So
in
interjecting
there,
one
of
the
things
that
Elliot
has
independent
stream
editor
has
asked
is
for
the
chairs
to
weigh
in
and
I've
held
back
so
far,
because
I
wanted
to
have
this
discussion.
But
my
view
is
the
same
thing
that
it
may
be
appropriate
for
Anders
document
to
go
to
the
independent
stream
at
some
point,
but
not
now
we're
still
having
active
discussion
about
what
to
do
with
this
whole
issue
and
I
think
that
our
input
to
the
independent
stream
editor
should
be
to
hold
off
until
until
that
discussion
completes.
A
Okay,
so
I
I
will
pass
that
on
to
Elliot
in,
in
the
thread
that
he
started
on
this,
which
has
not
been
copied
to
the
working
group
mailing
list
and
I'll.
Let
the
working
group
mailing
list
know
what
the
status
is.
C
Okay,
so,
given
that
we,
we
have
had
this
in
the
document
for
about
10
years
and
and
have
implementations
out
there
and
so
on,
I
thought
it
might
be
a
good
idea
to
actually
write
up
some
of
the
knowledge
that
that
we
have
Acquired
and
that
doesn't
necessarily
have
gone
into
the
update
rc8949.
So
I
wrote
this
background
document.
C
I
haven't
updated
this
since
originally
submitted
that
that's
probably
next
to
do
but
I
hope
to
have
an
informational
document
already
soon
that
that
provides
all
this
background
information
I
mean
people
talk
about
sub-normals
and
and
have
very
little
idea
of
why
they
are
relevant
to
deterministic,
encoding
and
so
on.
So
this
is
really
the
kind
of
information
that
that
I
was
trying
to
capture
in
in
this
document.
C
So
at
some
point
we
should
decide
whether
this
is
useful
as
an
Information
Network
group
document,
but
yeah
we
can
do
this
later.
We
can
just
collect
more
information
in
that
the
other
document
I
wrote
was
trying
to
get
some
structure
into
this
by
splitting
up
what
what
Wolf's
document
did
into
two
parts.
One
part
that
deals
with
encoding
and
one
part
that
deals
with
more
application,
specific
or
wolf,
has
suggested
saying
domain
a
specific
issue,
so
one
is
really
about
c-bar
and
the
other
one
is
really
about.
C
C
So
at
the
bottom
of
the
slide,
we
have
the
the
sibo
Baseline,
which
is
applying
to
all
sibo
instances,
of
course,
and
then
we
add
something
which
is
the
core
deterministic
encoding
requirements
from
RFC
8949,
which
has
a
few
holes
which
I
tried
to
show
on
this
slide.
C
So
we
we
put
another
thing
on
top
of
that
which
is
the
the
common
deterministic
encoding
profile
or
CDE
for
short,
that
closes
those
holes
and
provides
us
with
a
common
profile.
So
the
basic
idea
is
that
there
are
some
things
that
will
be
common
to
the
majority.
Probably
not
all,
but
the
majority
of
deterministic
encoding
applications
that
use
C1
and
that's
CDE,
and
on
top
of
that
we
have
domain
profiles,
and
these
are
because
CDE
already
handles
the
the
bytes
of
the
representation
from
it.
C
The
the
domain
profiles
really
are
about
mapping
data
from
the
application
data
model,
which
is
often
influenced
strongly
by
the
the
platform
data
models.
So
when,
when
you
only
have
in
64
and
and
you're
in
64,
then
you
probably
have
a
certain
view
of
of
the
data
that
that
you,
you
will
send
around
and
and
it's
good
to
capture
this
in
a
domain
specific
profile.
C
But
the
assumption
is
that
we
will
have
several
of
those,
not
hundreds,
probably
not
even
tens,
but
but
probably
there
will
be
a
single
digit
number
of
domain
profile
set
some
point
so
to
provide
an
alternative
visualization
here
with
CDE
in
the
bottom
and
and
dcbo
as
an
example
profile
on
the
top
you
can
see
on
on
in
the
blue
cloud
in
the
bottom
Cloud
that
in
this,
at
the
serialization
level,
we
often
have
multiple
representations
for
the
same
data
item.
C
So
the
data
item
in
the
middle
cloud
in
the
green
cloud
is
a
four
a
number
four,
an
integer
four,
and
we
can
represent
this
in
in
the
pressure
form,
which
is
just
a
single
by
hex04,
or
we
can
represent
it
in
a
longer
form
for
some
reason.
C
So
what
the
the
common
deterministic
encoding
does
is
essentially
creating
a
bijection
between
the
data
model
items
and
the
civilization
items
that
we
prefer
to
actually
see
and
that
are
actually
then
required
in
deterministic
encoding.
C
The
the
figure
may
look
nice,
but
it
completely
hides
the
fact
that
sibo
is
an
extensible
format.
So
it's
really
hard
to
maintain
the
the
beauty
and
simplicity
of
this
figure.
Once
you
add
extensions,
so
that's
why
Community
medicine
encoding
requires
a
little
bit
of
work
beyond
what
section
421,
actually
not
411,
defines
in
the
the
sibo
standard,
and
then
the
next
step
is
taking
this
set
of
generic
data
model
items
so
see.
Sibo
has
a
generic
data
model.
C
Actually
it
has
several
of
them,
but
I'm
not
going
to
run
into
the
details
right
now,
and
this
is
what
sibo
can
encode
and
the
community
gymnastic
encoding
creates
a
bijection
between
what
we
want
to
see
on
the
wire
and
what
we
see
in
our
heads
in
the
model,
and
the
next
step
is
to
go
to
the
domain
and
look
what
what
what
the
domain
like
to
do
with
its
data
to
get
a
deterministic
serialization
and
there
we
have
two
major
items
we
have
identified
so
far.
C
One
is
exclusion,
so
the
domain
data
model
May
simply
say
you
cannot
represent
minus
10
to
the
power
of
19..
In
this
domain
data
model,
which
is
perfectly
fine
if
the
domain
doesn't
need
that
it
does
create
some
implementation
complexity,
because
it's
outside
the
the
range
of
N64
and
un64.
So
you
you
have
to
to
somehow
manage
negative
integers
that
are
more
negative
than
you
can
express
using
in
64..
C
So
you
can
exclude
that
of
the
other
function
that
the
domain
profile
could
have
is
reduction,
so
it
takes
several
data
items
that
might
be
handed
down
by
an
application
and
Maps
them
to
a
single
generic
data
model
item
so,
for
instance,
the
number
four
which
is
supposed
to
be
an
integer
number
and
the
number
of
4.0,
which
is
supposed
to
be
a
floating
Point
number.
C
These
are
different
at
the
generic
data
model
level.
These
are
different
at
the
application
Level
at
least
in
most
platforms.
Javascript
is
the
one
important
exception
and
what
the
reduction
does
is.
It
looks
for
numbers
that
are
for
floating
Point
numbers
that
are
equivalent
to
an
integer
number
and
converts
them
to
integers
in
certain
specific
cases.
C
So
this
is
really
the
the
layering
that
has
come
out
of
the
discussion
and
I'm
really
glad
we
had
this
discussion,
because
the
the
structure
of
section
4.2
is
really
a
demonstration
that
we
didn't
understand
all
this
when
we
actually
wrote
up
the
the
crop
data
okay.
So
what
this
work
does
or
is
supposed
to
do
from
my
point
of
view,
so
my
objectives
are:
we
want
to
facilitate
wider
use
of
the
gymnastic
serialization
for
a
number
of
use
cases.
The
the
background.
C
The
document
has
a
number
of
them
both
sent
a
longer
list
of
use
cases
to
to
the
list
to
the
mailing
list
today.
So
there
are
a
lot
of
cases
where
it
helps
to
have
the
gymnastic
civilization,
and
we
want
to
facilitate
that
without
somehow
devaluing
the
the
unconstrained
use
of
sibo
that
that
we
have
in
many
other
places.
C
So
we
would
use
CDE
as
a
common
diffusing
recording
profile
as
a
common
set
of
encoding
decisions,
which
is
essentially
identical
with
section
four
two
one
and
a
few
tweaks,
and
that
can
be
implemented
in
a
generic
encoder
decoder.
So
much
of
the
work
that
is
needed
to
get
deterministic
civilization
out
there
would
only
need
to
be
done
once
and
then,
on
top
of
that
we
can
create
one
or
more
simple
domain
profiles,
so
we
don't
need
10
or
20
Page
documents.
We
need
one
page
documents
and
wolves.
C
Dc
board
would
be
the
first
instances
of
that
if
he
thinks
that
that
is
what
we
should
do.
Otherwise
we
could
also
use
another
one
like
like
under
DC
ball.
If
that
stabilizes.
C
So
the
the
documents
that
I
think
should
come
out
of
the
working
group
is
the
informational
document.
If
we
decide
to
agree
on
that,
the
CDE
part
of
the
the
senastrec
draft
that
I
wrote
and
the
domain
profile,
part
of
which
is
currently
in
the
same
document,
so
we
we
will
see
how
we
handle
that
Wolf's
proposal
is
slated
as
an
experimental
protocol
I'm,
not
even
sure
that
is
necessary.
So
I
think
this
could
be
go
through
as
a
standard
structure,
track
document
and
yeah.
C
The
question
is:
can
we
absorb
the
information
in
the
the
other
two
documents?
So
we
we
don't
need
more
documents
on
DC
bar.
We
just
have
a
single
one,
and
should
we
split
this
off
from
the
CDE
part,
or
should
we
use
this
as
as
an
example
as
a
motivation
for
the
CDE
part?
So
we
can
actually
make
our
points
in
a
document
that
has
bought
layers
in
it.
C
C
C
C
This
has
implemented
something
like
this,
so
this
would
be
a
little
bit
on
on
the
yeah
Innovative
side,
but
we
we
could
make
it
work
and
then
we
would
not
run
into
problems
when,
for
instance,
rust
finally
gets
its
f-128
type
done,
and
we
want
to
make
sure
that
we
have
that
in
the
deterministic
encoding
set
like
like
the
other
floating
Point
types
yeah,
and
then
maybe
we
can
do
some
other
I,
actually
754
stuff
like
how
exactly
we
handle
payloads
and
then
so.
C
Those
are
little
issues
that
may
need
to
be
discussed.
I,
don't
think
any
of
this
is
Big,
except
maybe
for
the
tag.
Five
integration
but
I
think
that
that
is
the
work
that
needs
to
be
done
to
finish.
The
the
CDE
part
of
the
document
and
that's
all
I
had
so
comments.
Welcome.
C
E
Certainly,
to
say
that
he
could
be
standard
track
and
and
I
like
your
middle
ground,
suggestion
of
using
some
of
wolves
profile
as
an
example
or
examples
in
the
CDE
standards
track
document
to
encourage
people
to
be
careful
and
rigorous
when
they
write
profiles
and
not
re-quote
large
bodies
of
80
45..
E
C
E
C
One
effect
of
that
is
that
people
look
at
this
and
say:
oh
this
civilization
format
only
has
20
pages
of
documentation
and
the
other
one.
It
says,
60
pages,
so
the
60
pages
one
is
likely
to
be
more
complicated
and
we'll
use
the
20,
Page,
documentation
and
I.
Think
splitting
off
informational
documents
is
a
good
thing
we
should
do
more
often-
and
in
this
case
it
came
pretty
naturally.
E
Thank
you,
I
I
think
it
is
a
good
direction
today
for
workroom
for
documents,
shorter
rfcs
that
are
concise
and
that
don't
constantly
restate
rate
the
old
thing
as
well
as
having
enter
interspersed
rationale,
are
a
lot
easier
to
read
and
to
implement
the
IEEE
seven.
Whatever
number
is
with
the
nans
ETC,
the
IEEE
float
15,
something
or
other
the
floating
Point
is
that
is
that
hard?
C
I
think
it's
mainly
hard
because
the
the
actually
Source
leaves
a
lot
to
implementation
on,
not
the
numbers,
and
we
we
probably
want
to
extract
something
that
that
is
still
useful
for
those
few
people
who
actually
want
to
use
Nan,
payloads
but
I.
Think
the
vast
majority
of
of
the
civil
applications
only
need
a
single
nand
value
so
that
that
is
likely
to
be
what
what
the
most
people
are.
Going
for
and
I
must
admit.
B
A
To
position
this
as
something
that
that
can
can
work
with
Anders
is
path.
A
I
think
it's
in
our
best
interest
to
try
to
do
that,
but
it
may
be
that
eventually,
as
I
said,
we
we
may
eventually
get
to
a
point
where
we
publish
what
we
publish
and
then
Anders
still
feels
he
needs
to
publish
something
on
the
independent
stream,
but
we'll
see
how
that
can
go
and
whether
we
can
whether
we
can
get
something
that
satisfies
his
needs.
C
A
Okay,
so
I
think
we're
good
on
all
of
that.
The
that's
our
agenda
is
there
anything
else
that
we
need
to
discuss
today.
C
A
A
A
Okay!
Well
thanks
everybody
for
coming
and
we
will
see
you
on
the
mailing
list.
A
Yes,
oh
wow,
I
guess
I
should
add
that
that
I
I
did
finally
update
the
shepherd
right
up
while
during
the
call
and
and
we're
going
on
time,
tag
bye.