►
From YouTube: IETF-CBOR-20220112-1500
Description
CBOR meeting session at IETF
2022/01/12 1500
https://datatracker.ietf.org/meeting//proceedings/
A
B
B
B
B
I
hope
everyone
had
a
good
new
year
and-
and
yes
again
thanks
marco
for
taking
notes,
let's
get
get
into
the
very
light
agenda
that
we
have
do.
We
have
any
working
group
document
status.
We
need
to
talk
about
right
now.
C
C
Go
ahead,
so
I
think
this
is
about
the
seabor
tags,
I'm
so
confused.
My
two
documents,
the
network
one's
already
done
the
file
magic.
So
I
think
it's.
I
think
that
we're
all
done,
I'm
just
a
little
bit
confused
as
to
whether
someone's
waiting
for
carson
and
I
or
we're
waiting
for
something
else.
Just
don't
know
waiting
for
write
up.
It
says
so.
Does
that
mean
it's
outside
of
my
of
our
problem?
It's
your
problem,
barry.
A
So
as
as
a
and
I
should
be
driving
this
as
a
shepherd,
my
latest
status
is
that
we've
had.
A
There
was
a
bit
back
and
forth
in
the
in
in
the
last
interim
before
before
new
year,
I
sent
you
the
authors,
one
of
your
sentences
on
on
something
that
came
up
there
and
I
think
that
the
status
there
would
have
been
that
all
all
those
latest
things
would
have
should
should
wind
up
in
a
document
that
we
can
have
a
second
working
group
last
call
on,
and
I
think
that
document
is
not
out
there
yet,
but
I
might
easily
be
mistaken.
A
So
the
last
update
that
happened
to
that
document
was
it
was
seven
published
on
on
the
15th
of
december,
and
we
had
the
interim
basically
on
that
day.
So
I'm.
C
A
C
So
this
was
the
looking
at
the
diff.
This
was
the
header
tag,
the
new,
the
additional
one
55
801
that
was
added,
and
so,
if
you
have
a
sentence,
then
I'll
go
look
at
the
minutes
and
and
figure
out
what
that
was.
Then
I
think
that
I
think
that
you're
right
that
it's
on
us
to
to
do
this
so
I'll
go.
Look
at
that
carson,
maybe
has
a
better
memory,
since
he
did
the
last
edit
pass.
A
No,
I
think
it
came,
it
would
have
come
by
mail,
because
I
wasn't
quite
sure
where
that
would
go.
C
A
D
A
And
the
other
is
text
that
was
sent
after
the
discussion
on
the
after
the
after
the
meeting.
So
if
you
want
to
take
that
up
for
the
working
group
last
call,
then
I
think
it
would
need
another
version
of
the
document.
D
Well,
not
the
document
status
of
this
working
group.
I
just
wanted
to
say
that
I
asked
the
cosi
chairs
and
ads
what's
going
on
with
1952
1953
1954
and
I
hope
we
are
getting
some
some
progress
there.
B
Yeah-
and
I
I
followed
up
on
that
because
at
one
point
I
was
a
shepherd
on
on
those
or
actually
the
responsible
a.d
before
I
gave
it
back
to
ben.
So
I
got
some
messages
about
it,
but
I
haven't
seen
any
activity
in
all
48
and
it's
been
in
all
48
for
way
too
long.
B
Okay,
so
then,
let's
go
ahead
and
move
on
to
the
other
item.
The
cddl
for
seabor
and
jason
carsten.
Is
that
you
yes.
D
So
the
the
discussion
we
have
had
in
in
various
places
was
that
it
would
be
good
to
be
able
to
write
a
single
piece
of
cdl
that
would
describe
both
a
json
and
a
seba
rendition
of
some
information
and,
of
course,
the
the
problem
with
that.
Is
that
often
it's
not
really
exactly
the
same.
D
Information,
you
would
have
there,
so
it
can
be
a
bit
difficult
to
describe,
which
part
is
is
actually
just
jason
versus
sibo
and
which
part
is
actually
the
part
that
describes
the
subtle
differences
and,
unfortunately,
chrome
doesn't
show
me
the
window.
I
want
to
share.
D
Why
is
it
doing
that
to
me
anyway?
If
you
look
into
appendix
a
of
of
the
the
document
that
I
sent
around
there,
you
can
see
one
way
of
doing
this
using
the
feature
feature
that
we
we
have
standardized
in
rfc
9165
and,
of
course,
that
does
require
some
tool
support.
So
I'm
currently
writing
the
tool
support
into
the
the
second
system,
cdl
tool
and
yeah.
If
you
look
there,
there
are
some
some.
D
Definitions,
generics
json
only
and
sibo
only
which
use
a
dot
feature,
json
and
dot
features
sibo
and
then
the
combination,
which
is
just
a
type
choice
between
adjacent,
only
and
and
a
sibo.
D
Only
so
this
how
this
would
be
used
can
be
seen
in
this
example.
So
if
somebody
else
can
share
that,
that
would
be
nice
because
I
don't
know
I
I
made
it.
I
made
it
almost.
D
Okay,
so
what
we
are
seeing
here
is
the
the
cda
definition
for
an
rc
8396
claim
set
a
cwt
cut
claim
set,
and
that,
of
course
was
designed
to
to
replace
jwt
claim
set
or
to
be
to
to
be
used
like
a
jw
t
claims
it
and
for
the
seven
claims
that
are
defined
in
in
the
cod
definition.
D
Six
actually
look
very
similar
to
what
we
have
in
in
jwt,
and
the
seventh
is
one
of
those
examples
where,
where
things
are
just
not
working
out,
so
what
we
have,
for
instance,
is
an
issuer.
D
The
issuer
is
identified
by
the
the
string,
iss
and
jwt
it's
identified
by
the
number
one
in
a
cwt
and
this
this
switch
jc
can
be
used
to
switch
in
the
the
right
part
of
the
definition.
So
this
works
very
well.
D
D
D
So
the
the
three
lines
here
are
really
what
what
I
expect
more
and
more
documents
to
simply
use
in
their
cdl,
and
then
they
can
write
a
single
piece
of
cddl
for
both
sibo
and
json
and,
for
instance,
cinema
might
have
used
that
it's
easy
to
write
the
cinema
spec
in
a
way
that
that
makes
use
of
this
jc
switch
and
new
documents
can
can
go
ahead
with
that,
and
essentially
I
would
like
to
to
poll
the
sibo
community
if
this
is
really
what
we
should
be
doing,
how
we
should
be
using
cddl
to
describe
both
json
and
sibo
in
a
single
specification
or
whether
we
we.
C
Karsten,
if
I
can
ask
the
point
here,
is
to
my
less
insufficiently
caffeinated
brain.
The
point
here
is
that
when
we're
doing
json,
we
want
a
string
and
when
we're
doing
c4,
we
wanted
integer
key
that
that's
the.
C
Yes,
yeah
and,
and
we
don't
want
to
mix
and
match
them,
we
don't
want
to
have
you
know
a
string
here
and
then
a
in
c
bar.
We
don't
want
to
have
a
string
and
then
an
integer
key,
and
then
you
know
whatever
back
and
forth
and
that's
why
there's
this
at
the
bottom.
Those
highlighted
lines
essentially
says
you
do
a
or
you
do
b
and
that's
what
your
your
business
is.
D
Okay,
maybe
we
don't
need
an
answer
right
now,
but
I
think
it's
important
that
if
we
go
this
direction,
we
do
this
consciously
because
that
will
be
used
in
a
lot
of
specifications.
From
now
on.
A
And
the
chat
rates
reads
positive
with
ira
michael
and
me
saying
that
yeah
this
this
sounds
good.
D
D
Of
course,
this
simple
example
doesn't
actually
cause
all
the
problems,
so,
for
instance,
in
json
we
have
the
problem
that
the
number
range
is
limited
if
we
are
using,
I
json,
and
that
sometimes
leads
to
interesting
gymnastics
and
yeah.
All
that
is
probably
requires
some
additional
functionality,
but
we
might
also
have
to
put
in
a
few
more
control
operators,
so
we
can
turn
a
number
into
a
string
and
things
like
that,
but
these
are
already
find
extension
points,
so
we
can
go
there.
A
Brief
question
looking
forward
to
the
cddl2
feed
proposed
features
of
mac
of
actually
matching
data
that
is
kind
of
assigning
matches
to
two
names.
A
Is
this
in
conflict
with,
or
is
there
some
envisioned
way
of
how
these
labels
could
then
say
that
this
value
that
is
somewhere
deep
in
this
byte
string
is
actually
equivalent
to
that
value
in
the
in
in
the
numeric
form
or
say
if,
if,
if
one,
if,
if
the
c
version
said
it's
b64
base,
64
encoding
of
this
and
the
keyboard
is
just
its
binary
data,
for
that,
could
we
match
this
and
that
in
those
annotations
it's
not
something
to
be
resolved
here,
but
it
would
give
me
confidence
in
this
approach.
D
Might
might
be
useful,
but
I
think
that
that
actually
describing
the
the
transformations
that
are
needed
to
make
json
and
and
sibo
renditions
equivalent
are
going
a
little
bit
beyond
what
we
are
currently
doing
in
c
bar.
So
once
we
add
transformations
to
zebra,
we
probably
need
more
expressibility.
D
D
I
have
one
late
guest
brendan:
do
you
want
to
speak
up
quickly.
E
Okay,
so
I
have
been
working
on
a
draft
in
in
the
suit
working
group
where
I
have
realized.
I
have
a
need
for
a
generic
set
of
sibor
parse
errors
and
perhaps
a
more
specific
set
over
top
of
that
for
content,
validation,
errors
that
sort
of
things
where
you
might
compare
against
a
a
schema
and
find
mismatches
or
missing
things.
E
As
a
result
of
that,
I'm
wondering
if
there's
a
need
for
that
in
general
or
if
that's
something,
that's
very
specific
to
the
use
cases
I'm
looking
at.
So
I'm
wondering
if
there's
a
a
desire
to
to
have
that
kind
of
work
in
the
seaboard
working
group,
and
if
so,
if
anyone
would
like
to
work
with
me
on
that.
E
Yeah,
certainly
so
the
the
first
one
obviously
is
going
to
be
an
overrun.
Where
you
run
out
of
the
expected
length
of
data
that
you've
got.
E
Then
there
are
type
mismatches,
key
mismatches,
integer,
decode,
overflows,
integer,
encoding
errors
and
because
seabor
has
a
number
of
features
that
I
would
call
probably
optional,
particularly
the
floating
point
support
and
the
and
some
of
the
other
simple
types,
there's
also
a
question
about
unimplemented
objects.
C
I
think
that
there's
yes,
I
would
be
interested
in
helping
you
to
do
this.
I
think
that
there
are
two
broad
categories
of
errors.
Ones
are
for.
The
first
examples
you
gave
is
there
aren't
enough
bites?
I
can't
complete
this.
This
is
broken.
This
is
invalid,
seabore
and
the
other
ones
are
largely.
C
Broad
categories
of
from
the
applications
point
of
view
they
may
be
irrelevant,
but
from
the
person
reviewing
the
error,
it
may
actually
be
pretty
important
to
know
yeah.
This
is
actually
valid
seabor,
but
this
device
doesn't
understand
it.
So
what
do
we
do?
Yeah
so
yeah.
C
You're
gonna
cast
me
and
I
I
guess
you
have
a
document
you're
thinking
about
writing.
Yeah,
that's
right!.
D
E
Absolutely
but
beyond
that,
there's
also
another
access
to
this
problem,
which
is
that
the
type
of
parser
you're
using
is
going
to
dictate
a
certain
amount
of
this.
So,
for
example,
if
you're
using
a
pull
parser,
you
won't
know
about
well-formedness
until
you're
done
and
by
that
point
it's
too
late.
E
Okay,
well,
I
will
write
the
document
and
submit
it
to
the
civil
war
working
group
after
running
it
past,
you
michael
since
you've
volunteered-
and
I
guess
we'll
see
where
it
goes
from
there.
F
On
that
last
brendan,
which
I
think
is
and
thank
you
michael
for
chipping
in,
I
think
it
is
needed
when
you
said
whole
set
of
errors
for
cozy.
I
was
thinking
of
several
applications
from
global
platform
and
elsewhere
that
are
now
using
cosy
and
have
yet
another
level
of
errors
that
are
somewhere
in
validity.
I
guess
or
not
implemented,
and
whether
just
suggesting
that
it
would
be
nice
to
be
able
to
conceptualize
and
label
sets
of
errors
or
have
a
naming
convention
or
something.
E
It's
a
bit
like
it
might
need
to
be
a
hierarchical
structure
so
that
you
can
at
least
get
a
classification
of
what
the
top
level
that's
coming.
That
this
is
coming
from
is.
D
It's
interesting
to
see
how
http
has
handled
this,
so
they
have
these
error
classes
like
like
four
or
five
and
then
within
these
error
classes,
they
have
some
defined
error
codes
that
are
describing
specific
issues
that
that
the
the
protocol
processor
runs
into,
but
they
also
describe
issues
that
are
close
to
application
issues.
For
instance,
the
the
was
a
409
conflict
in
http
that
that's
that's
pretty
interesting,
because
it
means
something
about
your
application.
Your
application
didn't
like
the
the
data.
D
So
maybe
we
need
a
little
bit
like
the
structure
for
these
errors,
as
well,
as
I
said,
distinguishing
well-formedness
validity
and
acceptability
to
an
application
and
so
on
so,
but
the
the
one
point
I
I
really
wanted
to
make.
Is
we
generally
standardize
error
numbers
when
we
accept
when
we
expect
the
state
machines
to
react
differently
to
different
errors?
D
So
if
I'm
writing
an
http
client,
I
know
that
when
I
get
a
409,
then
then
I
have
a
specific
situation
that
my
client
might
want
to
react
differently
to
a
409.
D
I
mean
if,
if
the
the
c-bar
is
not
well-formed,
that's
difficult
to
react
to,
because
a
reasonable
implementation
always
sends
well
formed.
So
what
what
does
it
mean
to
get
this?
This
error
message?
D
So
I
think
it
would
be
good
to
have
an
idea
of
what
what
the
implementations
are
going
to
do
with
these
error
codes.
E
So
I
think
that
suit
may
be
an
interesting
case
because
of
this,
the
the
the
problem
that
suit
faces
is
that
it's
not
a
conventional
two-party
communication
pattern.
You,
you
have
a
minimum
of
three
parties
involved
where
one
of
them
is
a
relay,
but
also
responsible
for
running
that
state
machine.
So
the
relay
ending
up
being
part
of
the
the
pattern
does
actually
produce
some
different
results.
E
So,
for
example,
if
you
have
a
sebor,
that's
not
well
formed,
it
needs
to
know
that
the
seabore
isn't
well
formed,
because
this
may
have
been
a
assigned
document.
And
if
you
have
a
seaboard,
that's
not
well
formed
inside
a
signed
document.
That
tells
you
a
lot
about
whoever
wrote
that
document
and
that
you
need
to
find
out
what's
happened
there
and
it
could
potentially
produce
a
an
automated
response
in
in
fully
automated
systems.
E
C
It's
also
a
case
for
I
think
it's
up
for
being
able
to
get
and
decode
the
sever
that
is
well
formed,
even
though
it
ends
short
or
other
things,
so
that
you
can
actually
tell
like
you
may
need
to
to
look
at
it
to
find
out
who
screwed
it
up
and
if
the
tools
don't
let
us
get,
the
libraries
refused
to
do
that
because
they
they
take
more
of
a
nanny
approach
to
things.
Then
I
think
that's
a
problem
for
that.
C
B
C
C
But
the
more
important
was
was
essentially
a
beauty
contest
as
to
whether
or
not
the
seymour
tagging
of
things
that
are
not
sebor
should
say,
seabor
or
something
else
and
carson
had
expressed
a
view
that
the
seaboard
indicated
that
the
encoding
of
the
tag
was
seabor,
even
if
the
stuff
afterwards
was
not,
and
I
at
least
felt
that
felt
seemed
wrong,
and
I
think
that
ira
also
felt
that
was
wrong.
So
I'm
wondering
if
there's
some
other
opinions
in
the
here
or
whether
the
question
was
well
formed.
D
D
A
I
don't
quite
understand
what
how
that
is
related
to
whether
the
okay,
so
yeah,
sorry,
I
I
see
now
so
through
an.
A
C
Since
the
purpose
of
the
this
byte
string
is
in
some
sense
because
we
have
to
put
something
there,
we
might
as
well
put
something
there
that
clues
a
human
to
that
who
might
be
looking
at
it
to
what's
there.
The
human
is
likely
to
key
off
of
the
string
and
not
have
you
know
unless
they
have
a
good
hex
dump
and
a
good
memory.
They
may
not
recognize
the
rest,
but
either
way
would
clue
into
them
that
you
should
probably
apply
a
sea
bar
decoder
to
the
first
12
bytes.
C
C
So
the
only
real
question
is
whether
there's
an
extra
clue,
that's
worth
communicating
to
the
human
or
here
or
not,
that's
really
it,
and
I
think
it's
it's.
I
think
it.
It
does
fall
into
a
beauty
contest
here
and
I
don't
have
a
strong
feeling,
one
way
or
the
other.
D
Yeah,
bringing
up
the
human
really
is
is
nice
example
for
for
the
problem
that
I'm
having
an
attacker
in
a
security
context
will
have
lots
of
joy
if,
if
they
can
write
a
tag,
that
means
something
different
than
what
a
human
sees
there.
D
Yeah,
but
I
I
certainly
can
live
with
see
box
or
whatever
was
proposed
here,
I'm
just
pointing
out
that
there
are
little
effects
that
you
may
not
want
to
ask.
C
B
Hearing
nothing,
I
will
again
thank
everybody
for
joining
and
michael
for
taking
notes
and
christian,
of
course,
as
well,
and
one
last
call
anything
else.