►
From YouTube: IETF109-ASDF-20201117-0900
Description
ASDF meeting session at IETF109
2020/11/17 0900
https://datatracker.ietf.org/meeting/109/proceedings/
A
We're
gonna
so
welcome
to
the
asdf
working
group
and
we'll
just
go
through
things
here
there
we
go
so
note.
Well,
you've
probably
seen
this
already,
but
if
you
haven't
then
please
take
a
look
at
at
the
note.
Well
ask
some
questions
and
there's
quite
a
number
of
people
that
you
can
come
to
friendly
ads
working
group
chairs.
Please
be
nice
and
professional,
we're
a
pretty
small
group.
So
if
you
have
something
to
interject,
then
to
the
extent
of
politeness,
please
just
do
that.
A
You
can
just
turn
on
the
mic
or
you
can
go
go
in
the
queue
if
you
aren't
ready.
If
you
aren't
trying
to
speak
to
the
point,
but
you
have
another
point
then
put
yourself
in
the
queue
and
we'll
get
to
you,
but
if
you
would
just
want
to
interact
on
one
on
a
particular
point,
then
I
would
say:
don't
stand
on
the
queue
at
this
point.
A
So
there's
some
more
details
of
the
note.
Well,
you
wanna
said
know:
what's
going
on
be
read:
bcp,
nine
and
the
other
ones
there.
I'm
sure
you've
all
seen
it
so
today
we're
gonna
go
through
a
bunch
of
things.
Specifically
status
update
on
the
group
and
then
we're
going
to
get
right
into
the
sdf
1.1
features
and
the
results
from
the
hackathon.
Is
there
any
other
items
that
someone
thinks
needs
to
go
into
the
agenda
at
this
point
now's
the
time
to
ask.
A
Hearing
none
we're
gonna
move
on,
so
there
is
a
kodi
imd
for
which
is
an
ether
pad
for
notes.
Ahri
is
ca,
is
adding
things
please
feel
free
to
add,
correct
and
copy
and
paste
as
you
wish
into
it
to
things
we
are
using
meet
echo.
If
you
aren't
on
it,
then
you
probably
can't
hear
me,
but
you
may
be
on
the
audio
queue.
A
There
is
a
jabber
room
that
you
can
connect
directly
to,
or
you
can
also
click
on
the
tab
with
the
two
red,
well
they're
red
on
my
screen
balloons
to
there
and
the
blue
sheet
is
managed
by
meat
echo.
So
you
don't
need
to
do
anything
and
we
have
a
note
taker
already,
so
our
working
group
was
in
fact
chartered.
Finally,
in
october
my
name
is
michael
richardson
and
my
colleague
is
nicholas
whiddell.
We
are
the
co-chairs
of
this
working
group,
so
hello
and
thank
you
for
coming.
A
We
had
a
bit
of
a
hallway
meeting,
as
the
working
group
was
installing
the
pro
pro
forming
process
and
we
did
a
bit
of
a
youtube
sdf
1.0
tutorial.
So
basically,
the
results
of
that
are
on
youtube.
If
you
want
to
know
a
lot
about
sdf,
then
and
how
it
came
to
be,
then
please
go
there
and
we'll
be
doing
a
little
bit.
A
A
B
First,
we
tried
to
address
this
by
selecting
one
solution,
but
that
did
not
work,
and
then
we
came
up
with
the
solution
to
actually,
instead
of
trying
to
select
one,
we
actually
introduced
a
new
layer
that
facilitates
translation
between
the
different
data
models
via
a
neutral
format.
That
was
that
could
help
with
this
translation,
and
this
proved
very
successful.
B
So
this
slide
was
used
a
bit
old,
but
it's
now
an
itf
as
this
working
group,
and
so
in
addition
to
the
work
in
the
asdf
group,
there's
also
the
work
in
the
one
data
model.
Alliation
group
there's
a
link
to
it
at
the
bottom.
Here
please
go
there.
If
you
want
more
information,
because
they
are
doing
a
couple
of
complementary
things
to
the
stuff
in
acf.
So
while
the
format
is
defined,
specified
and
developed
in
iatf,
there
is
work
on
tool
chains.
B
Yes,
next
slide,
so
just
a
few
words
around
the
asdf
work,
as
michael
said,
we
just
kicked
off.
This
means
that
we
were,
we
are,
would
like
to
reach
out
to
additional
organizations.
We
start
out
with
the
with
the
organizations
that
are
in
in
one
dm
of
course,
but
we
are
super
happy
to
bring
in
others,
and
obviously
the
value
of
asdf
increases,
what
the
more
that
other
organizations
join
and
we
can
create
the
interoperability
between
all
the
larger
set
of
organizations.
B
A
B
Of
course
it
fits
since
one
of
the
groups
going
into
this
work
was
the
ipso
group
in
ms
backworks
and
obviously
the
models
are
very
much
related,
so
it
is
it's
a
pretty
good
fit,
but
then,
of
course
it's.
It
is
also
a
kind
of
you
need
to
do
some
translation
and
and
on
the
iso
jtc
one
sc-41.
It's
actually
will
be
on
discuss
tomorrow.
So
I
don't
have
a
it's
a
new
york.
C
C
Okay,
so
I
think
I'm
supposed
to
to
speak
to
these,
the
next
few
slides,
I'm
kristen
berman.
I
got
dragged
into
the
1dm
work
about
a
year
ago
and
I
have
not
invented
anything.
That's
in
there.
I've
just
tried
to
make
it
a
document
that
people
can
use
that
ultimately
will
be
able
to
get
standardized.
C
So
let's
talk
about
this,
this
weird
numbering
scheme
here
sdf
1.1,
usually
in
the
idf-
we
we
number
things
dash,
0,
0.01
and
so
on,
and
what
we
are
trying
to
do
here
is
have
a
few
of
these
dash
something
versions
be
qualified
as
implementation
drafts,
so
they
they
kind
of
represent.
That's
the
the
objective.
C
They
represent
working
group
consensus
at
the
time,
with
a
full
knowledge
that
we
will
learn
new
things
when
using
it,
and
therefore
we
will
need
to
make
further
iterations,
so
fda,
sdf
1.1
will
be
a
03
or
02
or
whatever
happens
that
that
we
actually
have
had
something
like
like
a
a
small
working
group.
Last
call
for-
and
we
agree,
we
want
to
publish
this
well,
not
really
publish.
C
So
we
have
collected
a
number
of
open
issues
in
the
github
that
doesn't
mean
we
will
not
look
at
any
other
open
issues
anymore
after
we
have
closed
these,
but
these
are
the
ones
that
apparently
receive
the
the
largest
amount
of
interest
from
people
who
wanted
to
convert
their
ecosystem,
specific
data
models
into
something
that
can
be
expressed
in
in
sdf.
I
I'm
going
to
go
through
these
issues
in
in
more
detail
later
issue.
Number
three
really
is
about
something
that
already
is
in
sdf
1.0,
but
we
hadn't
really
tested
enough.
C
Yet
so
we
had
to
do
the
the
validation
and
one
tiny
thing
came
up
that
we
number
that
we
term
number
seven.
So
how
do
you
use
these
references?
Can
you
override?
Can
you
augment
them
issue?
Number
four:
is
we
punted
on
supporting
compound
data
composition
for
1.0,
and
I
think
we
now
understand
what
we
can
do
there,
so
that
was
one
of
the
results
of
the
hackathon
and
then
the
third
big
area
is
one
of
expressing
not
not
a
conjunction
like
here
but
disjunction.
C
So
at
the
hackathon
we
we
mostly
tested
out
the
the
view
we
have
of
issue
three.
We
came
up
with
a
convention
on
using
the
queries.
That
is
a
little
bit
surprising,
come
to
that
in
a
minute,
and
we
discussed
how
we
can
manage
these
namespaces
in
a
way
that
supports
the
progress
of
a
model
from
an
ecosystem
model
to
a
converged
model
and
well
the
various
transitions,
while
in
those
stages
and
on
issue
four.
Well,
we
use
the
the
discussion
about
compound
types
or
aggregation
to
also
fix
a
few
words
in
in
1.0.
C
So
with
that,
let's
next
slide
jump
into
issue
number
three
and
a
little
bit
issue
number
seven
next
slide,
so
we
we
can
instead
of
writing
something
into
the
model.
We
can
write
a
reference
into
the
model
and
that
reference
points
either
to
a
different
place
within
the
same
model
or
to
a
different
model
and
import
some
some
spec.
C
C
Let's
say
we
have
defined
a
data
type
that
we
call
length,
and
that
is
obviously
a
number.
The
length
cannot
be
less
than
zero.
The
si
unit
is
meter.
Oh,
this
is
using
the
old
name,
sorry
and
there's
also
a
description.
Well
that
probably
needs
to
be
longer
than
here
in
this
example
and
then
later
on.
C
If
we
have
the
the
cable
length
measurement
from
the
tdr
to
the
r
unit
in
a
router,
then
this
can
give
us
something
that
is
like
a
length,
but
the
tdr
unit
really
cannot
show
anything
about
cables
of
at
least
five
centimeters.
So
we
actually
modify
this
description
a
little
bit.
So
this
is
one
example
where
we
had
to
decide
whether
we
we
want
to
allow
this
kind
of
overriding.
C
So
this
is
actually
following
the
lisk
of
principle
here,
because
the
the
second
one
is
the
subset
of
the
first
one.
But
of
course,
when
you
can
override
anything
you,
you
you're
not
forced
to
follow
the
disk
of
principle
next
slide
so
in
in
practice,
this
would
occur
in
an
sdf
property
declaration,
and
so
sdf
data
is
just
a
place
where
you
can
define
things
that
can
be
referenced
or
data
types
that
can
be
referenced,
and
this
is
not
by
itself
declaring
affordance.
The
affordance
is
declared
in
the
sdf
property.
C
So
when
we
do
this
between
different
specifications,
then
of
course
we
have
to
manage
a
namespace
that
is
much
larger
and
we
decided
to
have
a
global
namespace
that
is
based
on
uis.
These
uis
are
not
necessarily
meant
to
be
resolvable,
but
they
are
names
and
using
a
namespace
table
and
the
default
namespace.
C
We
can
say
the
the
length
that
is
being
declared
up
here
in
the
yellow
part.
The
defined
actually
not
declared
that
can
be
referenced
in
the
white
part,
using
an
sdf
ref
that
points
to
that
namespace
and
interestingly,
the
namespace
names
can
differ.
The
interesting
part
about
the
namespace
is
not
its
namespace
name,
which
is
a
local
meta
in
the
specification
but
the
the
url.
C
C
C
C
A
couple
dozen
examples
that
make
use
of
this
in
one
or
the
other
way,
but
of
course,
so
we
should
use
all
time.
We
have
to
experiment
further
with
this
and
find
out
whether
we
maybe
also
want
to
use
this
in
different
contexts
than
just
referencing
data
types
and
michael
says
in
the
chat.
Please
step
up.
If
you
have
questions
about
this
and
christian
asked,
the
semantics
of
curries
are
still
plain
concatenation
and
the
answer
is
yes.
C
So
the
path
part
is
what
gives
you
the
the
uniqueness
and
the
part
behind
hash
is
a
fragment
identifier,
a
json
fragment
identifier
that
takes
the
form
of
a
of
an
rfc,
6901,
pointer,
json
pointer.
That
says
up
there
in
in
that
document,
look
under
sdf
data
length
and
you
you
find
it
there.
I
find
myself
I'm
I'm
so
used
to
big
blue
button
as
a
conferencing
system
that
are
pointing
to
places,
and
echo
doesn't
support
that.
C
C
C
A
C
C
To
okay,
thank
you
for
the
questions
issue
number
four.
So
this
is
about
composition
or
aggregation,
as
I
have
been
calling
that
compound
types.
We
have
many
many
terms
for
this.
It
essentially
means
that
we
are
constructing
something
out
of
multiple
parts
and
we
punted
on
that
for
for
sdf
1
0,
but
we
couldn't
completely
punt
it.
So
we
had
something
awkward
for
composing
parameter
lists,
and
this
is
on
on
the
way
out
now
that
we
have
real
composition
and
yeah.
C
So
this
is
an
example
in
in
the
new
syntax,
where
we
have
two
properties.
One
is
called
simple
one.
This
could
be
the
name
of
the
device
and
compound
one
which
might
be
the
the
location
of
the
device
and
element.
One
is
the
gps
coordinates
and
element.
2.
Is
the
street
address
so
whatever
so
we
have
a
component
property,
and
the
this,
of
course
could
also
be
modeled
as
separate
properties,
element
1
and
element
two.
C
C
We
decided
to
use
the
same
syntax,
json,
schema,
alt
uses
for
json
objects
and
that's
maybe
a
little
bit
confusing,
because
we
are
not
saying
that
this
property,
then,
is
represented
as
a
json
object,
because
we
are
at
the
information
model
level.
We
are
not
at
the
level
where
we
would
make
representation
decisions
like
that,
but
because
we
are
using
a
data
modeling
language
to
represent
information
models,
it's
pretty
normal
that
we
we
use.
C
One
way
of
representing
things
as
as
a
stand
in
for
the
information
model,
property
that
are
trying
to
represent
here
and
well
adjacent
object
is
a
good
way
to
group
things
together
into
a
bigger
thing.
Json
schema
org
also
has
a
required,
and
that
thing
is
called
property.
Sorry
in
in
the
compound,
the
word
property
means
at
least
three
things
here,
so
this
can
be
used
to
say
that
certain
of
the
the
map
keys,
key
value
pairs
are
required
and
others
are
optional.
C
So
that's
where
we
came
out
next
slide
and
now
that
we
have
it.
We
can
also
use
this
to
build
the
parameter
lists
for
actions
and
events.
So
actions
have
input
data
and
output
data
and
in
1.0
we
essentially
built
the
input,
data
and
the
output
data
as
arrays
of
references
to
sdf
data,
which
means
we
could
point
to
data.
But
we
couldn't
say
what
these
data
actually
meant.
C
So
we
can't
say
whether
the
the
intensity
value
that
we
are
pointing
to
is
a
red,
green
or
a
blue
intensity
value
and
with
a
new
way
of
structuring
things,
we
have
to
write
a
little
bit
more
json,
but
now
we
can
actually
put
in
names
here
so,
for
instance,
the
input
data
is
the
amount
of
squeezing
you
get
from
the
hug
and
the
output
data
is
the
amount
of
comfort
you
get
now.
If
you
look
at
this,
you
may
think.
C
Oh,
there
was
a
data
type
up
there
in
1.0
and
that
already
had
the
information
and
that
works
as
long
as
all
your
your
input.
Data
elements
have
sufficiently
different
data
types,
but
we
ran
into
examples
like
the
red
green
blue
light
where
the
data
types
are
actually
the
same.
So
the
data
type
reference
in
1.0
didn't
tell
you
what
it
was
and
then
now
we
get
this
in
in
the
name
like
squeeze
or
comfort
here.
C
C
C
C
C
Is
that
as
it's
usual,
when
you
just
give
text
strings,
then
then
you
think
you
know
what
these
things
mean,
but
it
actually
turns
out
that
that's
not
always
very
clear
and
just
from
a
single
text
string
that
looks
like
an
identifier
in
a
programming
language,
and
there
are
also
funny
little
things
like,
like
typos
and
and
those
text
strings
which
cause
immense
pain
when
you
repair
them
and
so
on.
So
it's
not
necessarily
exactly
the
the
way
you
want
to
do
things
next
slide.
C
C
C
C
C
There's
also
the
question
whether
the
the
numbers
that
are
given
there
really
are
information,
model
level
content
or
are
not.
So
there
are
some
industry
standards
where,
where
it's
not
hard
to
say
well,
most
rgb
devices
have
something
that's
going
from
zero
to
255.
So
so,
let's
go
with
that,
but
there
are
other
things
where
really
different
ecosystems
have
different
numbers
and
we
have
to
work
about
mapping
files
to
handle
those.
And
if
you
look
over
into
the
issue
behind
this,
then
you
see
one
example
from
the
microcluster.
C
C
C
The
the
alternative
to
that
is
to
say
a
choice
is
a
set
of
named
alternatives,
so
we
we
actually
want
to
attach
semantic
information
to
each
branch.
We
want
to
label
the
graph
that
we
set
up
here
and
by
putting
in
names
we
can
also
put
in
other
descriptive
information
like
like
rdf
references
and
so
on,
and
my
proposal
for
today
is
to
always
go
for
named
alternatives.
C
C
We
can
probably
use
the
branch
names
as
default
labels,
but
you
can
add
labels
in
any
data
type
in
sdf
and
we
could
also
add
semantic
tags
that
that
ground
this
in
in
some
terminology
next
slide-
and
this
is
the
the
example
that
that
mixed
any
of
and
sdf
enum-
and
we
now
have
a
single
choice
that
has
both
kinds
of
alternatives
integrated.
So
we
don't
need
the
unnecessary
nesting.
C
A
D
F
My
comment
well,
my
first
comment
was
that
this
also
kind
of
resolves
the
dilemma
we
had
between
one
of
and
any
of
in
jsonhema.org,
where
there's
really
an
ambiguity
about
they
both
really
kind
of
mean
the
same
thing.
So
this
sort
of
takes
that
choice,
away
and
and
just
presents
an
alternative
that
that
has
one
way
of
expressing
things
and
not
two.
F
So
that's
pretty
good,
and
also
I'd
like
to
just
comment
that
in
in
looking
at
trying
to
express
the
things
from
different
ecosystems
like
bluetooth
and
zigbee,
where
they've
done
bit
fields
and
things
like
that
that
that
are
combined
or
rather
value
ranges
that
are
combined.
This
is
and
really
for
the
enum
and
for
the
other
choices.
This
is
really
a
natural
pattern.
F
F
D
Yeah
well
again,
comment
now.
I
I
noticed
I
had
a
bit
hard
time
following
the
discussion
and
taking
notes
at
the
same
time,
maybe
a
bit
of
background
like
so
we're
now,
combining
the
enum
and
and
the
choice.
Part
of
me.
I
don't
quite
remember
our
discussions
we
had
before
I
went
for
my
leave,
so
my
brain
has
lost
a
few
months
of
discussion,
but
can
you
go
back
to
what
we
had
earlier
in
slides.
D
D
D
F
And
so,
if
I
was
standing
at
the
mic
in
a
room,
you
probably
would
have
seen
me,
but
what
I
wanted
to
say
was
that
the
actual
modeled
intention
of
this
pattern,
if
you
had
out
of
band
things
you
would,
you
would
have
probably
left
the
setting
at
zero
to
255
and
would
have
created
two
exclusive
out-of-band
states.
And
so,
if
I
was
rebuilding
this
from
scratch,
I
would
create
probably
two
resources,
one
that
had
the
the
settings
and
maybe
three
settings,
one
which
would
be
value.
F
The
other
would
be
minimum
and
the
other
would
be
set
to
previous
and
when
they
set
it
at
value,
it
would
use
the
scalar.
That
was
the
value
that
would
have
the
full
range,
but
since
they
wanted
to
pack
things
they
they
took
this
other
alternative.
But
you
see
that
this
pattern
would
be
useful
in
modeling
the
intended
pattern,
with
the
range
that
you
wanted.
It
could
be
0
to
255
and
there
could
be
two
other
two
other
things
and
you
would
be
free
to
map
these
in
a
mapping
file
differently.
F
If
you
wanted
to
and
implement
them
in
any
way
you
wanted.
As
long
as
you
had
the
property
of
these
being
exclusive
and
the
fact
that
you
expressed
it
as
a
choice,
does
basically
say
that
you
want
it
exclusive,
and
this
is
what
I
mean
by
saying:
it's:
a
natural
pattern
for
implementing
what
the
higher
level
intention.
The
semantic
intention
behind
the
weirdness
in
the.
D
D
Okay,
thanks
now
now
I
understand,
I
think
yeah.
This
seems
like
a
useful
pattern,
especially
when
you're
trying
to
represent
ecosystem
specific
things
as
accurately
as
as
possible,
probably
for
the
consolidated
models.
We
would
through
that
quite
a
different
way
and
then
have
it
easier
mapping
to
other
systems
but
ecosystem
specific
models.
Now
now
I
I
understand
the
use
case
here.
C
C
I
just
don't
have
an
example
where
I
actually
would
like
to
to
use
this.
So
I
I
didn't
set
up
such
an
example,
but
if
we
really
want
to
to
benefit
from
this
implicit
ordering
in
the
array,
then
we
could
do
that,
but
we
also
need
to
warn
people
that
having
this
implicit
numbering,
of
course,
is
very
brittle.
So
when
they
change
the
the
set
of
entries,
the
numbers,
change
and
all
hell
breaks
loose.
C
D
C
So
let
me
reply
to
that
before
michael
gets
next.
Basically,
there
are
two
cases
where
what
what
you're
saying
makes
sense,
one
is
an
ecosystem,
actually
decides
to
do
all
their
development
in
sdf.
They
no
longer
use
their
own
weird
way
to
write
models.
They
write
all
their
models
in
sdf
and
for
that
ecosystem
it
becomes
very
honorous
to
actually
do
their
numbering
in
a
separate
file
and
set
up
all
the
back
and
forth
pointers,
and
so
on.
C
You
may
have
this
situation
where
there's
just
an
industry
standard
way
of
giving
hours
the
number,
zero
and
g
the
number
one
and
b
the
number
two
or
something
like
that,
and
there
is
no
point
in
doing
all
the
same
ecosystem,
specific
mapping
for
all
ecosystems,
because
all
ecosystems
use
the
same
mapping
and
that
would
also
be
a
place
where
we
would
want
to
put
ids
into
the
conv
into
the
model.
In
this
case
the
converged
model,
and
then
we
wouldn't
have
a
need
to
to
generate
tons
of
mapping
files.
D
Okay.
This
is
the
default
case.
I
don't
need
any
extra
piece
of
code
of
doing
mapping.
Let's
do
one
one
on
one
can
push
the
same
values
across
I.
I
think
that
would
be
a
valuable
thing
to
enable,
but
again
probably
we'll
we'll
see
with
a
few
examples
of
mapping
files
having
those
or
not
having
those
if
it
makes
a
big
difference.
But
my
hunch
would
be
that
that
would
be
a
useful
thing
to
have
kind
of
default
value
for
the
enum.
F
So
what
I
wanted
to
say
was
that
I
think
first,
we,
we
kind
of
probably
resolved
that
just
relying
on
an
array
index,
which
is
an
artifact
of
the
chosen
representation,
which
is
already
an
approximation,
is
probably
not
a
good
idea.
But
fortunately
we
have
a
facility
called
default
that
already
exists
in
json
schema
that
we
can
use
to
specify
these.
That
really
has
a
well-defined
way
of
behaving
that
probably
suits
our
requirements
as
well
as
another
one
called
constant.
C
A
A
F
F
F
F
A
different
system
of
accounting
for
maybe
multiple
numbering
sets
in
line
or
something
like
that
might
be
something.
If
we
have
more
more
use
cases,
we
might
find
that's
a
better
solution,
but
yeah
I
I
guess
I
get
your
point
on
default.
It's
something
I
just
used
in
the
examples,
but
yeah
we
don't
want
it
to
be
put
in.
If
you
don't
specify
anything,
that's
not
the
desired.
D
Maybe
good
good
way
forward
like
if
we
could,
like
gonna,
do
a
three-way.
I
mean
three
different
ecosystems
example
of
of
some
existing
minimum
and
see
how
how
would
it
look
like?
I
think
that
the
choice
pattern,
what
we
see
here,
I
I
don't
see
any
obvious
problems
with
that.
I
mean
one
potential
thing
to
think
about,
would
be
this
array
approach
but
good
point
on
the
brittleness.
So
maybe
this
approach
is
is
better
for
it
and
then
see
how
how
it
looks
like,
but
I
think
this
is
going
to
good
direction.
C
So
if
you
have
a
tool
that
generates
this,
that
tool,
of
course,
can
do
the
numbering
and
insert
zero
one,
two
three,
four:
five
as
labels
and
and
two
the
choice.
It's
only
when
you
do
this
manually
that
it
becomes
a
little
tedious,
but
it's
not
that
tedious
and
it
also
saves
you
from
the
brittleness.
C
So
most
languages
that
have
this
iota
kind
of
thing
where,
where
you
have
automatic
numbers,
numberings,
have
a
way
to
switch
to
a
position
in
the
numbering
scheme
in
the
middle
of
the
sequence,
which
is
easy
to
do.
If
you
can
invent
your
own
syntax,
it
just
becomes
very
ugly
if
you
have
to
express
it
in
json
by
the
way,
there
is
also
an
email
that
water,
no,
not
an
email.
A
C
C
C
C
I
mean
towards
that
that
extract
jason,
schema
all
stuff
and
populate
this
from
jason.
Schema
won't
have
a
big
problem
actually
doing
the
the
renaming
from
whatever
adjacent
schema
argues
to
what
we
use,
so
they
would
rename
both
any
of
and
and
sdf
into
choice,
and
they
would
have
to
invent
some
labeling
for
the
any
of
case.
C
D
D
D
D
D
D
Yeah
because
I
think
like
for
me,
the
main
reason
why
the
plane
enum
doesn't
work
is
that
it
it's
missing
even
a
way
to
add
description
like
if
that's
something
what
we
we
really
need,
and
I
don't
want
to
have
an
awkward
way
to
add
description,
because
I
think
that
should
be
part
of
it.
Almost
every
time.
E
C
B
B
D
Things
we
have
come
up
with
and
what
are
the
commonalities
there
for
the
for
these
different
ecosystems
yeah,
as
margaret
richardson
was
asking
like,
I
think
it's
a
key
that
we
become
multiple
and
then
we
see
how
we,
how
well
this
system
maps
to
each
and
if
we
see
something
that
will
work
with
an
ecosystem.
Okay,
then
we
have
to
come
back
to
the
drawing
board
and
to
that
extent
I
was
actually
interested
of
asking
like
everyone
here
in
the
session
today.
D
Few
ecosystems
that
we
are
ourselves
actively
working
on,
but
if
you
have
worked
on
some
ecosystems
with
these
kind
of
structures,
would
be
very
good
to
get
your
feedback
on.
I
mean
also
all
the
issues,
but
this
one
in
particular
is
perhaps
something
that
you
may
have
use
on
so
either
right
now,
or
you
know
later
on
on
the
list
or
in
one
of
the
rooms
just
wanted
to
get
that
kind
of.
If
anyone
has
right
now
in
the
comment
zone.
C
Okay,
so
with
respect
to
the
the
item
number
three
and
on
this
slide,
I
think
it
indeed
would
be
good
to
have
an
example
that
maps
to
multiple
ecosystems,
even
if
we
don't
standardize
the
mapping
file
exactly
at
this
point
in
time.
I
think
we
can
all
understand
the
various
syntaxes
that
we
have
been
using
for
for
mapping
files.
So
I
still
think
we
can
examine
this
as
part
of
examining
this
example
and
see
if
we
really
can
close
number
two
number
five
in
this
way.
C
Okay,
this
doesn't
mean
that
the
the
set
of
issues
is
is
closed
or
something
people
can
still
propose
new
issues
and
in
particular,
number
seven
probably
needs
another
set
of
examples
to
look
at.
So
that
would
be
another
thing
to
discuss,
and
my
point
of
view
is
that
we
really
should
go
to
a
somewhat
time-boxed
release
process.
C
I
I
don't
have
a
detailed
idea
of
what
a
good
rhythm
would
be,
but
I
think,
given
that
we
have
various
holidays
in
the
way,
it
may
be
a
good
idea
to
use
december
8th
as
the
date
for
final
draft
for
1.1,
which
most
likely
will
be
a
dash
or
two
and
then
do
something
like
like
a
comment
period
on
that
or
working
with
last
call
or.
However,
you
want
to
call
that,
and
then
we
use
the
the
comments
on
that
to
have
a
dash
03
that
is
sdf
1.1
under
the
christmas
tree.
B
C
C
C
So
I
think
we
we
have
done
that
to
a
certain
extent,
and
we
are
now
opening
up
the
domain,
the
the
subset
of
the
ecosystem
models
that
that
can
be
translated,
but
of
course,
we
also
have
to
keep
keep
track
of
the
ones
that
cannot
yet
be
translated
into
sdf-11
and
see
if
there
is
an
1.2
feature.
That
would
help
with
that.
C
C
B
A
B
A
B
B
B
A
A
A
D
G
G
G
A
B
D
B
B
So
it
has
some
kind
of
limitation
on
the
general
general
applicability,
but
it's
definitely
so
that
it's
it's
it's
relevant
and
I
think
we
will,
as
we
reach
out
at
looking
at
additional
ecosystems,
that
already
do
their
modeling
information
modeling
in
uml,
like
opc
ua,
for
instance.
We
will
encounter
this
so
so
interested,
but
not
quite
sure
how
to
sort
of
map
things.
G
Together
that
it's
just,
I
was
just
just
generally
interested
in
what
your
views
works.
What
we
we
found,
you
all
know
a
richer
language
than
yang
in
some
dimensions,
and
hence
it
was
lossy
as
a
mapping.
So
we
we
lose
some
aspects
of
the
model.
We
have
to
also
orient
the
model
in
a
particular
way.
We
have
to
make
it
hierarchical
more
so
than
we
would
traditionally
do
with
a
networking
model.
We
tend
to
see
it
as
meshy,
and
you
know
you
sort
of
get
get
forced
into
bending
a
meshy
model
into
a
hierarchy.
G
So
we
we
may
have
overdone
that
as
well
and
we
may
have
been
able
to
go
for
a
forest
of
trees
rather
than
the
tree,
but
we
seem
to
have
gone
for
a
single
tree
and
hence
we've
distorted
our
modeling
somewhat
anyway.
So
I
was
just
interested
in
because
that
we
found
some
challenges.
Doing
the
mappings.
We've
got
a
tool
that
runs,
but
it's
not
perfect.
We
also
found
challenges
getting
people
the
right,
tooling,.
B
E
B
D
B
Looked
at
the
others,
bluetooth
and
so
on.
That
can
do
that.
That's
a
bit
more
challenging,
but
but
yes,
even
even
with
this
kind
of
limited
ecosystem
limited
application,
it's
it
still
generates.
There's
a
bunch
of
headaches
and,
as
you
say,
I
mean
there's
also
this
kind
of
abstraction
level
headaches
that
you
yeah
in
particular
when
you
need
to
address
multiple
receivers
so
to
speak,
yeah
yeah
so
very.
G
G
G
Yeah,
that's
fine.
It's
great
yeah
I've
got
some.
We've
got
a
we're
working
get
github,
so
I
can
send
you
links
to
that
and
there's
some
documentation
on
the
modeling
mechanisms
and
approaches
and
so
on
and
in
a
very
in
both
of
our
repositories.
Repositories
are
very
open,
so
we
know,
and
if
we
you
don't
even
need
a
password
to
get
the
documents
through
rnf
and
we
get.
Obviously
you
can.
You
can
see
so.
G
No,
that's,
we've
got
them,
we've
got
them.
I've
got
the
tooling
and
the
models
in
git
so
yeah,
it's
more.
The
documentation
explains
the
mapping
rationale
and
then
the
tooling
in
theory
supports
the
actual
mapping.
Mapping
definitions,
it's
not
perfect,
but
so
we
have
to
do
a
bit
of
hand
crafting
at
the
end
of
it.
C
G
I'll
try
and
get
a
bit
more
involved
in
this
because
it
does
seem
very
related
to
what
we've
been
trying
to
do
ourselves,
we're
looking
for
a
consistent
common
modeling
approach
across
multiple
bodies.
We
actually
work
with
meth
itt
and
I
think
for
others.
You
know
mfi,
ttl
and
f,
certainly
at
the
center
of
the
sort
of
relationship
around
uml.
So
so,
okay.
D
G
B
D
B
A
B
B
C
D
B
B
Calling
once
calling
twice,
then
I
think
we
are
unless
michael
starts
waving
here
and
I
think
we
are
done
for
the
day.
So
thank
you,
everybody
for
your
time
and
great
having
you
here.
I
hope
this
was
interesting
and
if
you
now
feel
that
you
will
be
want
to
hack
this,
please
let
us
know
so
we
can
include
you
in
the
hackathon.