►
From YouTube: CBOR WG Interim Meeting, 2019-10-09
Description
CBOR WG Interim Meeting, 2019-10-09
A
A
So
Caston
was
telling
us
the
status
of
this,
for
a
tag
is
announced
with
approval,
announcement
be
sent
and
we
grace
so
it's
waiting
on
a
new
version
and
then
Alexi
will
have
to
press
the
button
I
guess
to
to
send
it
to
the
earth,
editor
and
tip
or
sequence
status.
Actually,
when
I
set
the
agenda
with
announcement
to
be
sent
now,
the
announcement
has
been
sent
and
it's
in
the
RFC
editor.
B
A
B
A
A
B
B
B
B
C
B
B
The
the
get
rid
of
fellows
terminology
thing
I
think
we
are
done
here.
So
if
you
find
anything
where
we
still
talk
about
the
enclosed
data
item
following
a
tag,
please
tell
me,
but
I
think
we
are
done.
We
are
so
I
think
this
can
be
closed
and
that's
why
I
did
close
it
now,
on
the
other
issues
that
there
are
some
that
now
have
a
label
has
PR
and
let's
look
into
the
pull
requests.
B
B
You
may
have
too
little
data,
which
even
also
might
not
be
a
problem.
Of
course
you
cannot
cannot
derive
the
decoded
the
Seaboard
data
item
with
too
little
data,
but
it
may
not
be
fatal,
but
it
may
just
mean
we
have
to
wait
for
more
input.
Byte,
so
I
have
to
obtain
more
input
bytes
by
reading
them
from
some
sort
of
thing.
So
it's
pretty
much
trivial
to
do
test
vectors
for
too
much
data.
B
In
the
pseudocode
and
based
on
that,
we
can
now
go
into
examples
and
the
examples
are
structured
in
exactly
the
same
way
so
yeah.
There
are
no
examples
for
the
too
much
data
case
because
they
are
trivial
to
generate
the
examples
for
too
little
data
end
of
input
in
the
head
at
input
in
a
string,
a
little
length
string
and
of
input
in
a
mapper
array,
indefinite
length
strings
that
are
not
closed
and
indefinite
length
maps
and
arrays
not
closed.
B
This
is
the
complete
list
of
all
the
first
types
with
results:
the
additional
information
values
there
are
reserved
to
bite
and
coatings
of
simple
types.
I
have
added
one
example
here:
kind
3
is
indefinite
length,
string
chunks
that
are
not
after
the
correct
type.
So
this
says
it's
an
indefinite
length
string,
but
what
India
is
a
Goomba
another
string,
so
these
needs
need
to
be
checked
and
cause
syntax
errors.
B
You
cannot
have
indefinite
length
string,
chunks
that
are
indefinitely
strange,
runs
into
and
so
a5f
with
another
5f
in
it.
That's
not
allowed
and
that's
actually
something
that
my
implementations
don't
check.
So
I
have
to
fix
that.
So
thank
you
for
the
test.
Vectors
Lorentz
kind
for
is
a
break,
stop
code
that
is
occurring
in
the
wrong
place
like
completely
outside
often
indefinite
length
item
or
in
it
definitely
array
or
at
a
place
where
map
has
a
value
position.
B
I
sent
a
script
that
shows
how
to
extract
the
actual
byte
strings
from
the
text
of
the
section,
and
my
version
of
the
script
actually
runs
it
against
my
to
receive
our
implementations
and
they're
actually
all
invalid,
except
for
for
the
two
where
my
implementation
doesn't
excuse
me
err,
they
are
all
not
well-formed,
except
for
the
two,
where
my
implementation
doesn't
find
that
they
are
not
well
formed,
so
the
only
way
I
would
handle.
This
is
design
this
to
Paul,
who
might
have
another
reading
of
the
English
language.
D
Well,
this
looks
really
good
to
me.
I
I
think
it's
yeah
III
have
a
in
in
general
from
looking
it
over.
It
looks
just
fine
him.
It
looks
really
like
an
improvement.
You
know
your
your
way
of
extracting
the
the
data
that
seems
cool
and
useful
and
definitely
more
useful
than
the
tables,
and
the
descriptions
are
make
sense
to
me
as
well.
D
Think
that's
application,
C
bar
right,
and
if,
if
it's
not
a
sequence,
then
you
you,
you
definitely
know
when
you
have
too
little
data.
If
it
is
a
sequence,
you
don't
know,
and
the
text
doesn't
say
that.
B
B
So
there
are
other
cases
of
too
much
data
than
the
simple
sequence
that
case
on
the
too
little
data
side.
Yes,
you
know
when
you
have
too
little
data,
and
the
whole
point
of
the
comment
here
is
to
say
well
that
may
not
be
a
fatal
error
here,
because
it
might
just
be
necessary
to
wait
for
more
data,
so
you
no
longer
have
too
little
data,
but.
D
Don't
you
know
whether
you're,
which
which
kind
of
mode
you're
operating
in
so
yeah
so
I
mean
it
seems
like
you
could
say?
If
you
know
that
you
have
all
the
data,
all
the
data
has
been
read
and
you
still
have
too
little
data.
Then
you
know
you've
got
a
problem
if
you're
operating
in
a
mode
where
you
are
reading
data
n,
you
don't
know
whether
you've
read
all
the
data
or
not,
then
you
don't
know.
B
B
B
B
B
B
B
So
basically,
the
idea
is
to
make
quite
explicit
that
the
deterministic
encoding
is
a
subset
of
superset,
depending
on
how
you
define
a
set
of
preferred
theorization.
So
a
did
does
more
than
professor
ization
does,
but
if
you
have
deterministic
encoding,
you
always
have
preferred
the
serialization,
at
least
with
the
encoding
requirements
that
are
defined
in
this
document,
and
this
also
adds
some
text
about
the
floating
point
case
that
was
brought
up
by
Peter
hawk
sir.
B
B
B
If
there
are
no
comments,
then
I
ship,
this
to
Paul
number
one
one,
six
I
think
we
have
discussed
this
a
lot
and
I
finally
wrote
the
text
for
this.
This
is
actually
supposed
to
address
one
issue,
one
one,
three,
ninety
and
ninety
one,
and
so
first
of
all,
I
put
in
a
definition
for
expected
input.
Maybe
we
actually
want
to
use
this
a
little
more
but
I
put
in
the
definition
now,
as
usual,
we
are
using
expecting
it
in
English,
meaning
as
well.
B
Specifically,
we
are
using
it
to
describe
application-layer
validity
so
something
that
the
application
things
is
expected
as
its
input.
That
is,
ultimately
the
goal
of
all
three
bill.
Addition
steps
here,
well-formed,
village
and
and
expected.
So
this
addresses
the
three
layers
of
error,
checking
and.
B
Then
we
are
not
longer
talking
about
strict
mode.
We
are
talking
about
validity
checking,
so
the
first
reference
goes
to
a
little
checking.
Now
the
section
is
called
validity.
Checking
now
and
of
course,
the
the
text
has
changed
a
bit,
because
we
will
no
longer
making
all
these
wonderful
promises,
how
our
strict
mode
is
going
to
save
the
world,
but
we're
focusing
on
validity
checking.
B
B
So
this
needs
a
good
reading
as
well,
oh
yeah,
and
at
the
end
of
the
security
considerations,
this
discusses
how
a
generic
decoder
cannot
actually
do.
The
third
level
check,
in
particular,
since
the
set
of
defined
take
numbers
evolves.
You
may
be
in
a
situation
where
the
application
uses
a
tag
number
that
the
generic
decoder
of
know
about
yet,
so
it's
typically
not
an
error.
If
the
application
gets
the
tag
that
the
generic
Takada
know
about
just
means
all
the
checking
needs
to
want
to
be
done
in
the
application
level.
B
A
D
It
looks
pretty
close,
but
I
I
want
to
read
it
in
the
full
context,
but
yeah
a
little
bit
of
time
to
think
about
it.
I'll
just
add.
A
B
B
Hey
that
was
one
one,
six
one
one
seven
is
kind
of
a
specialty
item,
because
it's
about
this
Jason
Sasebo
conversion
number
range
issue
that
Peter
brought
up
explicit
and
are
proposing
a
broad
range.
That
is
the
I
Jason
range,
but
also
suggesting
that
there
are
various
other
ranges
that
applications
might
want
to
choose.
And
then
it.
B
C
A
B
B
B
But
there
is
no
validity
checking
of
the
embedded
see
what
item,
even
though
this
could
be
offered
as
an
option
by
generic
decoder.
So
this
is
one
point.
This
is
an
extra
technical
change
and
for
tank
36,
where
we
wondered
whether
that
is
why
that
is
stricter,
then
the
civil
item,
even
though
it's
much
more
work
to
implement
it,
I
put
in
sentence
that
says
a
valid
checking.
This
particularly
onerous
and
my
dear
bar
therefore
not
be
offered.
So
this
is
not
a
technical
change.