►
From YouTube: CBOR WG Interim Meeting, 2020-07-01
Description
CBOR WG Interim Meeting, 2020-07-01
A
A
Okay.
Today
we
have
sorry
for
the
lateness
of
the
agenda,
but
it
came
kind
of
at
last.
Second,
so
I
posted
the
link.
Let
me
read
post
a
link
to
the
minutes
on
the
agenda.
Today
we
have
several
items:
I
see
that
Jim
I
guess
you
have
added
this
item,
which
is
the
status
of
the
support
this
document
from
Barry,
but
I,
don't
see
Barry
in
the
meeting,
so
I
don't
know
I,
guess,
yeah
I,
don't
think
anybody
has
update
on
that
yeah.
A
Of
course,
of
course,
yeah
I
guess
we'll
try
to
contact
him
and
see
how
how
this
is
moving
then
see
were
tied
for
date,
status,
Mike
she's,
not
here
yet
I
think
he
he
replied
that
he
would
be
joining.
So
maybe,
let's
move
that
for
later,
then
we
have
discussion.
I,
don't
know
what
order
we
want
to
take.
A
C
This
draft
has
been
stuck
for
four,
almost
five
more
than
three
years
now,
because
my
esteemed
co-author
put
a
lot
of
stuff
in
the
was
was
very
innovative
and
and
novel,
and
all
that,
but
maybe
was
going
beyond
what
we
typically
do
in
a
tag,
a
document
and
and
a
so
an
object.
Identifiers
are
kind
of
very
basic,
so
maybe
we
shouldn't
try
to
attach
too
much
stuff
to
them.
C
So
I
took
the
document
from
three
years
ago
and
refocused
their
tool
to
doing
exactly
that,
be
finding
add
a
couple
of
tags
for
a
s
and
one
object.
Identifiers
and
I
removed
all
the
other
things
that
are
also
interesting,
but
maybe
should
go
into
a
different
document
with
those
tags.
I
kept
one
interesting
feature
which
is
tag
factoring,
so
you
can
put
an
ID
tag
on
an
array
or
a
map,
and
then
the
ID
tag
applies
to
all
elements
of
the
array
or
to
the
key.
All
the
keys
of
the
map.
C
I
think
that
that
may
be
a
useful
abbreviation,
but
I
removed
all
the
more
complicated
forms
of
using
tags,
text,
X
and,
and
all
that
and
I
think
we
need
to
be
a
little
bit
critical,
whether
I
didn't
throw
that
baby
with
the
bathwater
there.
But
I
think
that,
what's
currently
in
the
draft
is
a
is
a
pretty
good
set
of
functionality.
That
will
say
say
so.
If
99%
of
the
problems
on.
C
C
D
C
C
E
E
Working
call
comments
that
came
in
blank,
which
was
a
wedding
suggestion
by
Jim
that
I
adapted
most
of
the
others
were
saying,
be
clear
that
there's
no
times
with
me
so
I
now
say
that
there's
no
signs
around
them.
You
make
a
second
snot
supply
and
I
see
that
Paxton
was
funding
the
third
you're
doing
about
Angela
Merkel's
day.
C
B
E
A
B
B
B
C
C
Feature
should
actually
do
for
us
managing
those
namespaces
and
making
sure
a
specification
can
can
use
names
defined
in
other
specifications
without
having
to
provide
the
full
name
on
every
use,
and
that's
why
my
text
has
had
a
distinction
between
symbols,
which
are
the
names
that
are
that
we
are
using
within
acidity
respect
and
various
kinds
of
global
or
external
names
or
references
that
there
is
also
difference
between
names
and
references.
So
a
reference
can
be
something
that
that
still
needs
to
be
resolved
until
you
actually
have.
C
C
So
if
you
look
at
SDF,
we
have
a
global
namespace
in
there
too,
but
you
essentially
use
the
the
curie
construct
in
SDF
to
reach
into
that
global
namespace,
which
is
one
way
of
doing
that.
This
not
necessary.
The
one
I
would
suggest
for
Silla
yeah,
but
there
needs
to
be
a
link
between
the
the
global
names
and
the
names
that
are
actually
used
in
a
specification.
C
Yeah
I'm
not
sure
that
would
be
they're
useful,
but
this
there's
no
way
to
actually
stop
people
from
defining
names
and
we
don't
have
an
armed
guard
sitting
behind
every
person,
so
people
will
define
names
and
the
obvious
problem
here
is
that
we
not
only
have
different
people
writing
different
modules.
We
also
have
the
same
module
going
through
revisions,
so
we
will
have
several
definitions
of
the
same
name
in
in
somewhere
and
we
will
need
to
be
able
to
disambiguate,
which
one
we
actually
mean.
C
C
Yes,
I
think
the
young
people
are
trying
to
do
something
similar,
except
they
also
have
embraced
semantic
versioning.
So
the
version
is
part
of
a
name,
and
they
also
have
dates
on
these
things,
and
so
on
and
I
need
to
understand
how
exactly
all
these
parts
play
together
in
being
what
yang
is
definitely
something
we
should
be
looking
at,
because
they
have
just
gone
through
the
process
of
fixing
up
there.
They
are
naming
versioning
and
publication
process
and
and
all
this
stuff
so.
B
C
D
D
At
least
pieing
itself
doesn't
exactly
always
know
how
to
pull
in
the
like.
You
can't
specify
the
module
the
most
recent
version
of
this
module
as
far
as
I've
ever
been
able
to
feat
see,
so
you
sometimes
wind
up
building
some
tooling
around
the
tooling
so
that
you
can
get
the
right
right
thing
with
the
right
name
to.
C
C
We
have
package
managers
that
actually
do
the
work
for
us
before
we
can
use
naming
schemes
like
that,
and
that's
probably
the
kind
of
software
that
you
have
been
talking
about
now
and
the
other
end
of
the
spectrum
is
to
have
a
reference
system.
So
we
have
an
expression
that
says,
give
me
the
newest
foo,
and
then
we
have
a
resolution
process
that
actually
resolves
that
reference
into
a
specific
module
module.
C
C
I
would
expect
that
that
we
can
do
something
like
such
a
reference
system.
We
still
should
think
about
this
a
lot
because
it's
easy
to
fall
into
the
trap
of
applying
your
experience
from
managing
pieces
of
software
to
managing
specifications,
and
that
is
different
and
not
all
intuitions
actually
apply.
F
We
are
not
talking
about
fights,
yeah,
I,
think
specifications
is
a
good
thing
and
and
I
think
yang
people
think
about
as
fires,
mostly
so
there
is,
this
date,
tag
added
to
them,
sorry
about
calling
it
a
check,
but
actually
it
is
so
so,
and
that
is
to
disambiguate
revisions,
but
that's
it.
It
is
not
used
in
yang
itself
anymore.
It's
only
used
in
the
specification
of
the
module
on
the
specification
text
around
the
module
effectively
and
we
need
something
that
is
not
only
for
files.
That
is
I,
think
very
important.
F
B
F
The
early
days
when
creating
CDD,
we
there's
an
order
to
the
routes,
so
so
at
the
start,
I
always
thought
about
maybe
concrete
to
fires.
And,
of
course
the
order
is
clear,
but
yeah
just
uploaded
blobs
of
data
that
resides
memory
and
never
RFI.
They
still
have
to
be
ordered
correctly
and
now
you
don't
have
a
fire
system
univ
to
specifications
and
they
have
to
kinda,
be
safe,
descriptive
in
their
own
way,
and
that
is
the
reason
why
we
think
this
route
rule
has
to
be
at
the
beginning.
F
If
to
route
rules,
I
could
to
specifications
that
are
relying
much
other
and
you
can
date
them.
And
now
you
have
no
fire
in
their
surnames,
no
namespace
problem
because
they
know
each
other
so
to
speak
or
their
build
with
awareness
of
each
other,
but
they
still
have
to
be
in
the
correct
order
and
is
I
think
something
that
is
also
to
the
module
design
here.
But
maybe
correct
me
if
I'm
wrong.
C
C
Of
order
gets
a
little
bit
less
important,
a
message
with
a
reading
list
a
couple
of
minutes
ago.
That
goes
into
this
in
a
little
bit
more
detail.
So
we
cannot
completely
ignore
all
of
this
kind
of
problem
that
we
are
having
where
we
are
building
our
specifications
out
of
files
and
have
to
be
careful
putting
them
into
the
right
order.
I'm,
not
sure
that
we
will
have
that
once
we
have
a
module
system,
yeah.
F
C
C
B
C
I
already
agree
to
you:
I
know
what
you
mean
and
I'm
the
word
export
is
really
a
technical
term.
That
really
should
be
called
contribute.
So
specification
contributes
to
the
global
namespace,
and
it
does
this
in
such
a
way
that
you
and
you
don't
run
into
random,
concrete
walls
that
people
have
erected
in
in
good
intention,
but
not
knowing
that
the
world
was
going
to
rotate
under
them.
D
F
So
so
I
am
ia
human
readable
annotations.
That
I
thought
to
get
confuse
myself
I
put
in
to
see
the
specification
nowadays
is
actually
literally
depends.
So
I
depend
on
the
contribution
of
some
other
thing.
Otherwise,
I
will
not
function
properly.
It
will
function,
though
mostly
that
is
a
little
bit
weird.
F
So
it's
not
really
a
dependency
dependencies
I
think
are
things
that
do
not
run
or
do
not
work
at
all
with
note,
something
that's
contributed
by
the
contribution
is
maybe
it's
more
like
in
arguments
I
note
that
we
can
call
it
augment,
but
it's
more
like
an
overview,
so
you
only
add
things:
it
have
to
be
safe
suspicions
in
the
first
place,
otherwise
it
will
not
be
a
valuable,
viable
module
if
it's,
if
it's
importing,
sorry
it's
depending
but
actually
currency.
So.
C
It's
very
much
like
the
song.
It
concept
in
CD,
where
you
have
something
that
anyone
can
contribute.
Yes,
the
specification
doesn't
really
tell
you
which
of
these
contributions
are
good
or
bad
or
should
be
used
or
should
not
be
used
that
that's
something
that
needs
to
be
done
separate
from
the
specification
yeah.
F
I
thought
I
should
also
contradict
myself
right
now
again,
because
I
see
the
hair
specifications,
its
parts
that
are
right,
of
course,
not
safe
sufficient,
sometimes
sometimes
they
are
only
purely
contributing
CDI
specifications,
they're,
not
a
full
spec
by
their
own
right
and
I,
highlight
that.
So
these
are
not
there.
These
are
a
brute
CDI
as
I
call
them,
mostly
so
as
a
specific
kind
of
dependency,
so
that
they
need
something
to
contribute
to.
Otherwise
they
do
not
really
work.
C
F
F
C
Yeah,
so
people
should
be
able
to
pick
up
those
additional
preloads
and
usually
went
when
we
do
a
tank
definition.
Then
we
put
a
couple
of
lines
of
CD
DL
into
their
tank
definition
and
that's
really
something
that
would
be
selectable.
So
you
can
say
he
is
the
the
CD
DL
for
the
take
definition
and
I
can
just
import
that
without
a
lot
of
ceremony.
F
Yeah,
you
know
well
constraint
if
you
are
constrained,
if
you,
if
you
do
going
with
the
Jason
preload,
simply
some
types
are
not
allowed
there,
so
you
cannot
just
use
a
contribution
that
is
a
Jason
or
something
builds
on
that
and
then
plug
it
into
a
sorry,
otherwise,
the
the
co-op
reroute
and
plug
it
into
a
CD
aspect.
That's
for
Jason
that
simply
does
not
work.
C
F
There's
all
these
three
things
like
that,
the
namespaces
slash
some,
you
think
versioning,
the
the
module
T
here
that
we
have
with
contribution
and
its
complement
and
then
the
feature
self
is
always
its
are
connected.
Somehow
it
audio
always
related
closely
related
to
each
other.
So
it's
it's.
Why
the
independent
topics
and
they
can
be
defined
independent,
it
may
be
even
they
strongly
relate.
F
B
B
C
A
D
A
A
E
Add
a
sentence,
possibly
even
in
an
abstract
ability
that
saying
that
this
represents
on
their
dates
and
not
the
particular
point
in
time
and
Jim
got
some
text
to
replace
the
text
that
I
wrote
about
decision-making
for
access,
control
and
I'm
fine,
using
something
based
on
Jim's
text.
So
I
will
publish
another
draft
today,
probably
within
a
couple
hours.
At
that
point,
the
chairs
can
perfect
I.
Think
you.
E
F
E
E
A
E
E
A
E
B
D
F
Yeah,
that
is
a
subtlety
added
to
the
things
that
contributes
so
a
module
which
contributes,
amongst
other
things,
dedicated
features
that
are
clearly
highlighted,
and
that
is
fine,
I
think
that
should
be
exposed.
But
then
again
it
is
a
feature
notation
then
and
I
assume.
That
is
already
if
you
go
through
the
spec,
but
you
have
to
do
in
any
case,
I
assume
as
you
process
it,
you
will
know
all
the
features
so
I'm,
not
sure.
C
Yeah
I
think
the
whole
thing
about
features
is
what
can
you
actually
do
with
that
then?
So
in
the
CDR
tool,
I
was
going
to
put
in
a
feature.
Excuse
me
for
being
able
to
select
which
features
are
enabled,
which
means
we
assume
they
are
present,
which
features
are
disabled,
which
means
we
assume
they
are
not
present,
and
then
there
are
other
features
there
that
are
not
not
influenced
by
this
command-line
interface.
So
they
don't.
B
C
C
C
B
C
And
that's
actually
it's
something
that
will
happen
a
lot
that
we
will
have
a
specification.
That
only
has
a
JSON
version,
our
only
Sieber
version
initially,
and
then
we
add
the
other
one
in
as
well.
And
then
you
need
to
handle
both
the
fact
that
parts
of
the
specification
may
not
have
been
updated
yet
and
the
fact
that
implementations
actually
may
provide
only
one
of
the
top.
F
C
Yeah
and
so
far
we
at
the
language
level
we
managed
to
avoid
features
except
we
have
one
very,
very
obvious
feature.
So
we
are
extending
the
set
of
control
operators
and
the
simple
effect
that
is
education
uses.
A
particular
control
operator
means
that
it
requires
that
feature
from
the
compiler
and
we
probably
want
to
refine
that
a
little
bit.
C
C
And
note
that
the
validator
auction-
actually
there
are
different
kinds
of
processing
models.
We
don't
have
actually
embraced
all
these
processing
models
so
being
able
to
perform
a
particular
part
of
the
processing
model,
so,
for
instance,
putting
in
default
values
that
that
will
be
one
feature
of
CDL
processor
and
some
will
have
it
and
others
won't.
C
It
should
be
no
surprise
that
that
I
really
love
defining
terms,
because
those
terms
allow
us
to
be
specific
about
the
concepts
we
are
using.
So
it
may
be
useful
to
just
spend
some
some
time
defining
those
terms.
So,
of
course,
we
are
probably
going
to
discard
half
of
them
at
some
point
because
they
turn
out
not
to
be
so
useful,
and
so
once
we
have
done
that
a
little
we
might
want
to
identify
invariance,
so
the
properties
of
the
solution
that
are
going
to
stay
or
that
are
least
desirable.
C
B
B
A
D
C
D
C
A
D
A
A
C
A
A
A
A
A
Let's
see
which
is
yeah,
so
that's
gonna
be
the
same
day
as
our
interim,
the
15th
of
July.
So
we
discussed
this
a
little
bit.
Last
interim,
we
were
saying
that
we
will
continue
this
detailed
discussion
now,
it's
approaching
so
again.
If
you
more
detail
about
what
agenda
items
you
would
like
to
have,
then
let
us
know
so.
C
One
thing
that
that
might
come
up
I
already
mentioned
that
there
is
an
interest
in
getting
a
specification
done
for
JSON
path
and
that
specification
would
probably
be
applicable
for
SIBO
as
well.
Let
me
send
in
some
form
so
the
first
meeting
on
Monday
morning
is
going
to
the
to
be
the
dispatch
working
group
and
that
might
actually
have
an
idea
on
how
to
handle
this
standardization
either
added
to
the
SDF
buff
or
handle
it
separately.
E
C
For
XML
and
that's
that's
a
monster
and
most
people
who
want
to
do
something
with
JSON
or
SIBO
need
a
similar
query
language
and
there's
something
called
JSON
pointer
which,
which
is
nice
and-
and
we
already
tried
once
to
extend
that
for
SIBO,
but
notice
that
this
is
not
as
easy
as
we
thought.
So
we
didn't
complete
that
yet
and
JSON
path
is
somewhere
between
the
simplicity
of
Jason,
pointer
and
and
curing
equivalent
complexity
of
XPath,
and
we
still
have
to
find
out
where
exactly
that.
That
is.