►
From YouTube: ASDF WG Interim Meeting, 2021-06-02
Description
ASDF WG Interim Meeting, 2021-06-02
A
Be
nice
and
professional
and
the
ipr
guidelines
and
network
replies,
you
see
it
here.
You
have
read
it
before
next
time.
Please.
A
So
this
is
the
agenda
for
today
yeah
they
are.
I
didn't
have
the
final
titles
for
the
presentations
you
will
you
probably
copy
from
the
I
can
update
the
the
in
a
moment.
The
titles
are.
B
D
Five
but
then
I
was
mistaken.
A
Oh
yeah,
okay,
so
so,
let's
take
it
in
the
sequence
of
of
so,
let's
think
of
the
sequence
of
the
what's
in
the
in
the
main
slide
set.
So
first
there
is
your
iana's
on
sdf
quality
of
round
trips.
A
Then
there
is
and
present
stuff
from
kersten
on
attaching
semantics
and
then
at
the
end
there
is
a
relationship
to
innocences
by
ari.
E
A
Paste
the
right
agenda
into
the
final
minutes.
Is
there
any
additional?
A
Yeah
good
any
other
things,
people
want
to
bring
up
today,
nope,
okay,
moving
on
rock
clips
next
slide.
Please.
A
I
need
to
I
need
to
trim
this
down.
I
guess
so.
This
is
where
we
are
today.
We've
had
three
virtually
in
trims.
This
is
the
first
one
we
have
had
our
ipf
meetings
at
109
and
110
with
hackathons.
A
A
No
great,
if
not,
I
guess
we
will,
we
will
plan
for
interims.
I
think
we
can
take
that
at
the
end
of
this
meeting
when
we're
done
without
presentations
or
do
it
over
email.
A
One
thing
that
we
haven't
brought
up
here
is:
if
even
if
we're
not
meeting,
I
guess
there
is
no
point
of
having
a
hackathon
in
the
week
before.
How
do
people
view
that.
B
There
was
a
question
in
the
w3c
web
of
things
meeting
earlier
about
that,
and
I
I
kind
of
pro
to
sort
of
guess
that
we
might
have
some
of
the
workflow.
I'm
sorry
processing,
sdf
processing
prototypes
to
work
on,
but
I
wasn't
sure
so.
I
think
if,
if
there
is
anyone
that
wants
to
work
on
prototypes
now's
the
time
to
plan
it,
but
I
I
haven't
heard
of
anything
coming
up.
A
So
but
that
would
be
more
of
the
1dm
side
right
or.
B
A
A
B
Implemented,
maybe
the
language
itself
and
not
one
dm
patterns,
yeah.
A
Are
people
open
to
doing
that?
I
kind
of
have
a
bit
of
a
problem
that
week
to
do
things
so
so
it's
I
might.
C
B
F
A
F
A
Maybe
or
you're
in
the
car,
so
maybe
you
can't
talk
so
yeah
all
right.
Let's
make
that
a!
Maybe
I'm
very
much
interested
in
working
on
the
tools
chains,
but
maybe
not
that
weak.
A
A
Let's
see
if
we
have
some,
I
mean
that
that
I
mean
we
also
need
to
plan
the
interim,
so
it
could
be
that
we
bash
our
heads
together
and
figure
out,
because
I
what
one
thing
which
we
haven't
done,
but
we
also
need
to
discuss,
is
of
course
the
the
we
should
say
the
final
parts
of
this
of
making
sdf
draft
an
rfc
and
because
I
think
we
have
so
you
know,
hand
wave
that
that's
it
will
kind
of
happen.
A
A
So
I
mean
for
the
minutes:
let's,
let's.
Let's
agree
that
we
don't
meet
at
no
objections
to
not
meeting
at
one
one
one
and
maybe
undoing
some
kind
of
sdf
processing
hackathon
activities
in
the
week
before
there,
and
then
we
will
sort
that
out
over.
A
Any
yeah
so
asdf
outreach.
Thank
you
move
on
next
slide.
Please
sorry.
My
previous
meeting
was
rather
intense,
so
I
kind
of
burned
all
my
brain
cycles
on
that
anyway,
asdf
outreach
and
so
far
beyond
1dm
I
mean
we
have
presented
this
to
the
msc
group
in
homeless,
backworks
and
there's
a
discussion
there
on
how
to
use
it
if
to
use
it
and
how
to
use
it
and-
and
that
is
being
discussed
this
iso
ic
jtc
one
se41
proposal
it
as
far
as
I
know
it
was.
A
It
was
approved
in
the
vote
with
pretty
actually
pretty
strong
approval,
pushback
from
korea
and
japan.
But
apart
from
that
strong
approval
and
now
erston
is
cleaning
up
and
they
had
some
kind
of
second
round
of
doing
objections
and
handling
those
questions.
I
I
provided
some
input
to
them
for
that,
but
I
don't
know
when
that
has
been
done,
but
but
it
seems
like
sdf
is
on
that
way.
Whatever
things,
what
more
is
happening?
There.
B
So
things
yeah
with
web
of
things,
they're
very
interested
in
harmonizing
the
design
patterns
between
sdf
and
the
the
new
thing
model
and
the
more
sort
of
offline
formats,
where
thing
description,
as
you
know,
is
more
of
a
instance
describing
hypermedia
control,
now
they're
looking
at
more
modeling,
and
so
they
have
needs
to
do
things
like
we
do
with
references,
and
you
know
other
kinds
of
validation
and
all
of
that
so
they're
very
interested
in
harmonizing
with
what
we
do
and
I
think
it
would
be
productive
to
to
set
up
a
meeting
and
and
have
a
joint
session
at
some
point
yeah.
B
I
think
they've
asked
me
to
set
that
up
and-
and
I
think
we
should-
we
should
consider
doing
that,
because
they
have
a
lot
of
similar
questions
and
issues
and
in
fact,
almost
every
everything
that
we
have
itself
an
issue.
They
are
sort
of
considering
some
sort
of
feature
and
thing
model
that
is
going
to
require
the
same
kind
of
decision
to
be
made.
So
we
want
to
harmonize
all
of
that
plus.
B
We
want
to
make
sure
that,
at
the
other
level,
that
our
definitions
are
things
that
they
can
use
and
as
annotations
and
thing
descriptions,
but
that's
that's,
I
think,
even
secondary
to
like
the
the
basic
design
pattern.
B
So
that's
something
I
don't
have
a
time
frame
for
that,
because
they're
just
finishing
publishing
their
their
specs.
So
what
we're
looking
at
is
the
next
version.
So
until
now,
they've
been
sort
of
too
busy
to
engage,
but
after
they're
face-to-face,
maybe
during
their
face-to-face,
so
they're
having
a
face-to-face
the
same
time
as
etf
111,
basically
in
the
last
couple
weeks
of
the
of
june,
so
maybe
during
their
face-to-face
or
shortly
after
so
that
sort
of
late
june
early.
D
A
F
A
I
will
be
on
vacation
that
time
that
time,
but
but
at
some
point
yes,
I
mean
I
agree,
we
should
definitely
try
to
work
with
him
better.
Absolutely.
F
E
Yes,
would
anyone
have
any
contact
with
the
china's
a
I
I
ai,
I
think
it's
called
association
of
industrial
iot
or
something
like
this.
I
don't
know
the
acronym
very
well,
but
since
it
doesn't
ring
any
bell,
I
guess
the
answer
is
no.
A
B
C
B
Aii
and
it's
looks
like
china-
it's
a
government
organization,
143
members
by
invitation,
only
mission
to
and
unite
government
industry,
blah
blah
yeah
right.
So
it's
industrial
internet.
B
A
But
michael
richardson
do
you
have,
did
you
have
any
thing
in
mind
or.
E
Oh
there
you
are,
I
I
would
say
it's
this
a
case
of
broken
telephone,
where
someone
said,
ask
me
how
what
the
relationship
would
be,
and
I
like
I've,
never
heard
of
them,
and
so
I'm
just
spreading
the
the
signal.
I
think.
A
I
need
to
there.
There
was
something
like
that
being
in
the
session
where
our
iso
is
this
in
iso
iscj1,
where
sdf
proposal
was
proposed
to
prevent
presented,
there
was
something
that
sounds
like
this.
I
will
check
on
that,
but
I
mean
if,
if
they
have
an
interest
of
learning
more
about
us,
I
mean
we
were
happy
to
talk
to
them.
I
guess
at
least
figure
out
what
it
means,
what
they
are
about.
A
Great
one
thing,
maybe
more
to
mention
here
that
I
forgot
to
add:
I
guess
we
don't
have
any
progress
on
the
electronic
data
sheets
for
the
textile
manufacturing.
B
Not
heard
back,
but
I
I
will,
I
will
do
another
outreach
on
the
electronic
data
sheets.
Folks
are
busy,
you
know
doing
releasing
their
own
thing
and
they
should
be
done
with
that
by
now.
So
maybe
maybe
maybe
it's
a
good
time
and
then
the
technical
style
people
I
did
reach
out
to
them.
I
didn't
hear
anything
back,
but
I
can
send
them
another
one
I'll
just
keep
pinging
them.
Yeah,
okay,.
A
One
thing
that
was
that
should
have
been
here
as
well,
which
is,
I
mean,
outreach
very
close
outreach.
It
was
currently
that
there
was
a
last
core
meeting.
That
was,
I
think
we
can
bring
it
up
in
the
notes.
So
there
was
a
in
the
last
core
meeting.
There
was
kind
of
looking
at
sdf
and
coral
and
how
they
can
be
used
together
and
someone.
So
there
are,
if
you
were
not
attending
that
the
notes
are
interesting
in
this
aspect.
A
A
F
A
Yeah
yes,
moving
on
next
slide,
please.
A
So,
let's
see
here
sorry,
I
realized.
C
Actually,
no,
so
there
are
two
things
that
I
think
we
should
be
discussing.
First
of
all,
there
is
a
dash
of
six
out
that
differs
from
dash
of
five,
essentially
in
resolving
pr
30.
C
So
I
think
we
had
various
discussions
in
various
forum
about
how
to
do
this
and
came
up
with
various
candidate
syntaxes,
and
when
I
tried
to
write
this
up,
I
essentially
found
that
it's
probably
easiest
to
come
up
with
the
simplest
possible
syntax.
So
that's
what
ari
proposed
and
I
actually
sent
a
message
to
the
mailing
list.
So
if
you
want
to
pull
that
up.
C
So
that's
maybe
something
that
it
would
be
good
to
get
feedback
for
whether
ahri's
syntax
is
the
one
we
can
go
forward
with.
I
mean
it's,
it's
really
simple:
it's
just
adding
the
the
min
items
and
max
items,
qualities
to
the
sdf
object,
so
yeah.
We
we
should
check
whether
that
handles
the
multi-instance
cases.
C
We
have
been
talking.
A
A
A
A
B
C
If
you
look
at
the
the
email,
I
wrote
yesterday
shortly
for
midnight
that
points
to
two
sdf
documents
in
the
exploratory
repository
that
show
aries
approach
and
and
what
I
think
my
comments
led
you
to
do
on
on
your
approach
and
so
ahri's
approach
validates
us
doesn't,
and
we
should
check
whether
we
we
can
agree
with
going
for
the
very
simple
change.
G
B
A
G
Sure
all
right,
okay,
excellent!
So
one
more
comment
here
on
the
version
discussion.
It
would
be
worthwhile,
maybe
having
a
small
design
team
meeting
or
something
on
that
to
really
get
back
to
all
the
things
we
have
now
learned
and
explored
on
the
versioning
and
related
to
the
number
29
here,
but
then
also
the
whole
data
mobile
version
in
discussion
at
large.
So
I
had
some
good
discussions
with
klaus
over
the
last
few
weeks
and
it
would
be
interesting
to
go
back
to
that
topic.
G
A
A
Yes,
great,
thank
you.
Alright.
That
was
a
good
point.
It's
good
also
not
to
bring
that
discussion
up
here
further
because
I
think
it
gets
so
messy
yeah
with
that,
then
I
think
we
can
move
on
to
the
to
the
presentations
here
today.
A
C
Having
been
done
before
we
publish
the
the
first
specification
is
an
rfc.
A
Very
good
point,
and
I
I
think
the
we
can
of
course
discuss
it
now.
I
I
must
say
that
I
haven't,
maybe
maybe
it
would
actually
yeah.
A
So
I
I
I
don't
say
that
I
I
I
personally
had
a
sort
of
exact,
perfect
vision
of
what
that
what
the
rfc
would
be
and
what
is
the
way
for
because,
as
we
look
on
on
things
like
the
stuff
that
ari
is
presenting
at
the
end
here
on
relationships
and
new
statuses
and
so
on,
that
is
really
cool,
stuff
and,
and
some
of
it
might
be
in
an
rfc
in
some
ways,
maybe
future
material.
But
still,
I
think
we
will
continue
to
build
on
all
stuff.
We
can
include.
C
Well,
since
many
of
us
actually
have
a
vision
for
the
timing
of
the
the
submitting
the
first
rfc,
maybe
it's
a
good
idea
to
exchange
those
versions.
A
C
Yeah,
so
we
can't
really
discuss
this
now
without
some
preparation,
but
maybe
for
the
next
interim
that
that
would
be
one
interesting
question
to
to
discuss.
A
I
think
yeah,
so
what
I
think
would
be
good
is
that
I
mean
like
kirsten,
if
you,
if
you,
if
you
have
it
in
your
head,
I
mean
please
I
mean
if
you
could
just
write
that
as
a
kind
of
a
feature
list
and
a
list
for
later
features
of
sort
of
post,
rfc
stuff
from
your
perspective
and
others
can
add
to
that
or
look
at
that
or
do
their
own
lists.
C
A
A
But
yes
had
had
I
had
more
time.
I
would
have
spent
lab
to
spend
some
time
on
that
before
this
meeting.
But
yes.
A
Let
let's
let's
point
to
this
a
bit.
I
think
it's
a
very
good
thing.
I
think
we
really
need
to
set
that
down
and
then
prepare
material
for
next
interview
on
on
this.
How
to
what
I
mean?
How,
when
do
we
conclude?
What
is
what
is
done?
A
The
exact
objective,
because
also,
I
guess,
if
we
want
to
do
more
things
that
go
beyond
our
charter-
I'm
not
saying
that
we
are
there
already,
but
we
were
chartered
to
do
the
sdf
spec
and
if
you
want
to
do
a
lot
of
other
things,
it's
it's.
Of
course
we
need
to
do
that.
Work
as
well.
A
A
So,
as
we
were
preparing
for
this
meeting
carson
reached
out-
and
as
I
understand
it,
this
is
a.
A
D
To
do
about
it
sure
thank
you.
Yeah
hi,
I'm
jana,
I'm
currently
working
on
an
sdlv,
yank
converter
as
part
of
my
master
thesis
and
I'm
a
student
of
carson.
For
those
of
you
who
don't
know
me.
Yeah
yang
is
a
modeling
language
for
data
center
over
network
management
protocols.
D
For
those
of
you
who
don't
know
and
while
working
on
my
converter,
I
ran
into
another
issue
which
is
round
trips
this
time
more
precisely
without
additional
measure
measures,
models
that
are
round
tripped,
which
means
converted
to
sdf
and
then
back
to
the
original
format,
will
be
completely
different
from
the
original
that
they
came
from.
Why
is.
G
D
The
reason
for
that
is
that
convergence
conversion
does
not
work
in
an
injective
way
where
one
structure
from
language
x,
in
my
case
yang,
is
converted
to
a
distinct
structure
in
sdf.
Instead,
multiple
structures
from
language
x,
in
my
case
yang,
are
converted
to
one
structure
in
sdf.
For
example,
yang's
choice,
structure
and
yang's
built
in
type
union
would
be
converted
to
sdf
choice.
D
So
when
you
want
to
go
back
when
you
want
to
do
a
round
trip,
it's
not
possible
to
know
anymore,
where
that
sdf
choice
came
from,
whether
it
should
be
translated
back
to
yang
choice
or
to
the
type
union
yeah.
So
the
solution
would
be
a
new
sdf
quality
by
the
way
cast
made
me
aware
of
the
possibility
to
introduce
a
new
sdf
quality.
So
it's
not
entirely
my
idea,
and
you
also
asked
me
to
suggest
how
that
could
look
like
next
slide.
Please.
D
D
This
is
what
it
could
look
like
for
the
example.
I
constructed
a
dummy
module,
so
this
is
not
taken
from
any
real
yang
module,
and
this
yang
module
contains
a
definition
of
a
derived
type
which
is
derived
from
the
union
built
in
type
and
is
a
union
over
the
type
string
and
decimal
64..
So
an
element
of
this
example
type
def
could
either
be
of
type
string
or
type
decimal
64..
D
This
is
as
of
now
translated
to
an
sdf
data
definition
example
typedef
containing
the
sdf
choice
between
the
type
string
and
the
type
number,
and
now,
if
you
were
to
go
back
without
any
additional
sdf
origin
quality,
you
wouldn't
know
whether
this
sdf
choice
was
originally
a
union
or
a
yang
choice.
So
this
is
where
the
origin
quality
steps
in
in
the
info
block.
There
is
a
reference
to
the
format
name,
which
is
yang
and
to
the
rc
of
yang.
D
This
isn't
really
needed
for
my
converter,
but
I
thought
it
would
be
nice
background
information
and
then
in
the
sef
data
element.
There's
an
origin
statement
containing
they
said
typedef
and
type
and
typedef
defines
that
the
statement,
a
statement,
one
of
the
statements
used
was
typedef
and
type
defines
that
another
statement
used
was
type
and
that
statement
has
the
argument
union,
the
typed.
D
If
statement,
does
not
have
an
argument
yeah
since
I
already
need
this
in
my
converter,
as
it
is
right
now,
I
implemented
it
kind
of
temporarily
via
the
common
quality
description,
where
I
kind
of
added
a
code
that
indicates
the
statement
argument,
but
I
basically
did
it
in
the
same
way
as
in
this
sdf
quality,
where
there's
a
statement
and
an
argument
defined
so
the
way
I
presented
it
here
would
already
work
for
me,
but
it
isn't
really
elegant
to
do
it
over
descriptions.
D
That's
why,
for
my
purposes,
it
would
be
really
nice
if
there
was
like
more
elegant,
so
to
speak,
official
way
of
representing
origins
and
yeah.
To
sum
up,
this
would
be
a
working
solution
for
me
and
I
would
be
glad
to
hear
what
you
think
of
my
suggestion.
E
So
is
there
other
things
other
than
yang?
That
would
have
the
specific
union
versus
choice
issue,
and
would
they
refer
to
it
as
union?
I
I'm
completely
ignorant
here
so.
D
E
Mostly,
I'm
just
concerned
about
the
the
the
word
union
might
be
confusing
some
potentially
in
some
other
place
that
I
have
no
clue
but
either.
So
that's
all
I'm
asking
about
is
whether
this
is
the
right
concept,
the
right
right
right,
the
right.
It's
the
right
concept
of
whether
it's
the
right
way
to
to
explain
the
to
the
concept,
and
maybe
someone
else
has
a
better
knowledge
than
me.
G
Well,
knowledge,
first
of
all,
thanks
for
bringing
this
up,
I
think
this
is
a
important
issue.
I
mean
I
have
a
couple
of
questions
about
your
your
design
here.
First
of
all,
would
everything
in
the
origin
block
be
kind?
G
Let's
call
it
origin
specific
vocabulary,
so
there
would
not
be
like
name
conflicts
between
different
origins
or
how
do
you
envision
that
to
work.
D
So
you
mean
that
what
is
defined
via
the
statement
and
arguments
has
to
be,
you
know
official
or
what
do
you
mean.
G
E
D
C
I
think
the
interesting
case
is
when,
when
we
have
a
model
from
ecosystem
a
and
that
is
converted
into
sdf-
and
it
has
some
of
these-
these
origin
qualities
and
then
somebody
translates
it
into
b
ecosystem
b,
for
instance
in
yang
and
starts
working
with
it
and
and
making
sure
that
we
have
the
right
unions
and
choices
and
whatever
in
place
and
then
translates
this
back
into
sdf.
C
D
D
I
actually
had
the
idea
first
to
put
inside
the
the
the
origin
statement
in
the
for
the
definition
and
the
common
qualities
to
put
in
each
one,
the
format,
name
and
format
reference,
so
yang
and
rc,
etc.
C
Yeah,
let's
not
design
the
mechanism.
I
think
that
that
the
name
spacing
mechanism
we
have
in
sdf
actually
might
might
be
good
enough
to
to
handle
this,
but
I
think
in
in
general
it
will
be
important
to
have
this
kind
of
coordination
and
yeah.
We
also
want
to
make
sure
that
that
a
converter
of
ecosystem
x
doesn't
mistake,
statements
that
are
made
in
in
the
origin,
qualities
of
ecosystem
y
as
something
that
that
actually
is
from
ecosystem
x.
C
B
B
That
I
prepared
were
where
I
tried
to
look
at
having
extension
vocabularies
that
had
specific
origins
already.
One
of
them
was
the
web
of
things
or
actually
the
the
different
vocabularies
that
are
used
for
protocol
bindings
in
the
web
of
things,
the
http
vocabulary,
the
mqtt
vocabulary,
etc.
B
So
I
I
think
what
what
we
need
to
you
know.
I
think
what
carson
said
is
you
know,
plus
one
that
that
our
namespace
mechanism
probably
can
deal
with
conflicts,
but
what
we
need
to
do
is
to
have
a
robust
design
for
a
common
way
of
adding
extension,
vocabularies
or
or,
if
that's,
not
exactly
the
right
thing,
something
that
accomplishes
that
those
common
requirements
that
we
have.
B
I
think
extension
vocabularies
make
sense,
but
there's
another
way
to
think
of
it
in
terms
of
mapping
files.
Well,
maybe
it's
not
a
whole
bunch
different,
really
mapping
vocabularies
there's
the
idea
of
whether
it's
in
the
file
or
out
of
the
file,
and
if
we
manage
the
conflicts
correctly,
the
processing
wouldn't
care
whether
it's
in
the
file
or
a
separate
file
with
json
pointer
references.
It
would
still
work
the
same
way.
B
I
was
thinking
of
it
as
a
back
mapping
to
you
know
in
terms
of
a
translation,
so
forward.
Mapping
would
be
to
map
into
a
target
instance
format,
but
the
back
back
mapping
would
be
to
you
know,
to
annotate,
based
on
the
original
definition
format
and
I've
seen
other
use
cases
where
this
would
be
a
good
thing.
B
So
I
definitely
would
totally
support
this
being
one
of
the
one
of
the
things
that
we
do
with
annotations
also
definitely
see
it
as
annotations
that
don't
change
the
semantics
of
the
sdf
file,
but
augment
it
and
provide
ways
of
integrating
integration
points
for
other
to
make
sdf
integrate
into
other
ecosystems
more
easily.
A
B
F
A
What
is
the
next
step
for
this?
We
haven't
defined
the
mapping
file
format
or
anything
like
that,
because
I
think
this
should
be
related
to
that.
A
I
mean
on
this
william,
you
have
sort
of.
I
guess
this
solves
your
use
case.
Do
you
need
anything
more
with
this
right
now
or.
D
A
Okay,
thanks:
let's
put
it
on
the
list
this
on
a
list
of
things
that
we
need
to
look
deeper
into
and
move
on,
because
we
have
actually
had
a
bunch
of
slides,
left
and
yeah.
D
Yeah,
so
thanks
for
your
attention.
A
Well,
thank
you
very
much
for
presenting
it.
Karsten
you're
you're
up
next.
C
Yeah
next
slide
please.
So
this
is
essentially
the
the
related
thing
in
another
color.
So
I
was
asked
to
prepare
something-
and
I
remembered
that
I
already
already
had
prepared
that
in
2018
for
g2trg
meeting
next
slide.
So
essentially
the
the
presentation
at
the
time
was
about
attaching
semantics.
So
how
do
you
attach
semantics
to
an
instance?
You
have-
and
I
called
this
semantic
style
at
the
time,
because
essentially
we
we
have
learned
ways
of
doing
that
using
xpath
and
and
accessory,
and
so
on
a
long
time
ago.
C
So
we
know
how
to
do
these
things,
and
maybe
it's
a
good
idea
to
to
just
use
that
knowledge
in
designing
this.
So
in
in
the
semantic
style
world,
a
style
can
be
specialized
for
a
single
instance.
C
So
you
essentially
have
resolved
pointers
that
point
to
specific
places
in
that
instance,
or
it
can
apply
to
a
class
of
instances
and
then,
of
course,
you
need
to
find
what
what
you
actually
want
to
influence
with
the
additional
semantics
using
some
some
selector
mechanism
next
slide
and
when
we
look
at
this
from
the
sdf
perspective.
C
Well,
first
of
all,
sdf
models
are
instances.
So
we
can
actually
use
json
pointers
as
a
selector
syntax.
So
this
is
all
very
simple
and
I
think
some
of
the
mapping
file
proposals
essentially
do
that.
But
of
course,
in
some
cases,
selectors
that
actually
apply
to
multiple
models
also
would
be
useful.
C
So
we
could
do
something
that
actually
provides
a
mapping
file
to
all
ipso
models
and
not
just
one
of
them.
So
this
is
a
very
different
instance
versus
class
discussion,
because
sdf
semantically,
of
course,
are
classes,
but
the
models
are
instances
of
of
the
metamodel
that
allows
those
classes
so
that
that's
weirdness
numbers
one
next
slide.
C
What
can
we
do?
We
can
have
transformations
and
augmentation.
I
think
we
we
already
had
half
that
discussion
already
on
the
previous
point,
so
the
the
attachment
of
styles
may
lead
to
a
completely
different
document
with
with
a
different
target
language
different
target
generic
data
model,
which
was
the
idea
in
in
diesel
and
accessority,
or
it
may
be
more
on
the
augmentation
side.
C
So
you
have
something
like
css,
which
gives
you
properties,
in
addition
to
the
attributes
that
are
already
in
the
instance,
and
then
we
of
course
have
in
model
augmentation
where
we
stay
in
the
generic
data
model.
So
we
could
theoretically
put
all
this
information
in
the
instance,
which
is
essentially
what
what
the
jana
was
showing
or
we
can
have
extra
model
augmentation,
where
we
essentially
put
this
into
a
more
general
data
model,
which
maybe
is
not
not
that
much
different
in
difference
in
practice.
C
Next
slide,
because
in
in
sdf,
essentially
in
model,
augmentation
would
trigger
existing
extension
points
like
using
more
qualities
than
than
we
have
defined
so
far
and
extra
model
augmentation
essentially
creates
new
extension
points,
but
these
need
to
live
within
what
we
have
in
terms
of
extension
points.
So
there's
not
really
that
much
of
a
difference,
and
we
just
should
make
sure
that
in
sdf
we
have
ways
to
accommodate
data
that
we
didn't
think
about
when,
when
we
wrote
the
standard
next
slide,.
C
So
in
in
essence,
what
what
a
simatic
style
is
it's
a
pair
of
selectors
with
the
effect
so
in
css
that
that
would
be
every
html
element
with
the
class
warning,
gets
the
color
red
and
yeah,
so
so
how
these
selectors
look
like
and
how
these
effects
look
like.
That's,
essentially
what
we
have
to
define
next
slide.
C
And
yeah,
so
we,
if
we
only
address
sentence
instances,
then
the
adjacent
pointer
is
all
we
need.
So
we
could
say
that
that
in
the
playground,
the
sdf
type
foo
is
augmented
by
by
some
additional
information,
and
I
I
didn't
space
specify
the
effect
here,
but
the
effect
probably
should
be
adding
qualities.
C
So
this
would
be
a
little
bit
like
the
merge
batch
stuff
that
that
we
already
did
so
that
that's
all
I
have
prepared-
and
I
think
now
I
want
to
move
quickly
to
ari.
Who
has
another
view
on
this.
G
Okay,
thanks
carson.
Yes,
I
think
slightly
different
angle,
but
I
see
a
lot
of
commonalities
on
what
you
were
presenting
here.
So
the
background
on
on
our
work
is
I've
been
discussing
with
that
many
of
colleagues
I,
including
klaus
on
how
could
we
extend
sdf
on
to
support
more
complex
relationships
and
then
also
potentially
have
instant
specific
information,
and
the
background
here
is
that
we've
been
looking
into
converting
sdf
into
different
ecosystems
like
dddl
or
pcua,
and
also
looking
into
modeling
some
slightly
more
complex
things
that
we
have
been
using
stf
so
far.
G
So
if
you
go
to
next
slide
there,
you
have
it's
a
bit
of
explanation,
what
we
mean
by
these
related
relationships.
So,
as
you
know
today,
the
models
they
can
only
have
quite
simple
child
parent
relationships
or
composition
by
having
accidents,
properties
group
into
sdf
objects,
sdf
of
group
and
sdf
things,
which
can
be
further
grouped
into
further
sdf
things.
E
G
Can
indicate
that
okay,
this
is
a
flow
residue
called
floor
and
the
target
is
as
specific
another
model,
so
the
next
slide.
What
we
will
be
suggesting
here
is
to
add
a
a
similar
feature
to
sdf
to
support
that
kind
of
modeling.
It
could
be,
for
example,
called
sdk
relations,
and
the
kind
of
style
that
we
have
been
converging
towards
is
to
have
this
a
map
of
relationships
from
any
of
the
object
or
things
that
is
the
source
of
those
relations,
and
then
you
have
a
set
of
named
maps
there.
G
For
example,
here
you
have
a
relationship
called
contained
in
it
would
always
have
a
target
which
would
be
one,
for
example,
one
of
the
definitions
in
if
the
other
stf
files
are
in
the
same
sdf
file.
I
could
also
have
a
type
if
it
has
a
formal
type,
for
example,
from
an
external
ontological,
lucky
version
saref
here
as
an
example,
and
it
could
also
have
properties
so
further,
giving
giving
more
information
about
this
specific
relationship.
G
So
if
you
go
to
switch
to
the
next
slide
there,
you
see
an
example.
Sdf
object
called
room
sensor
that
would
have
at
least
one
property,
but
they
would
also
have
relationship
to
this
stf
object.
Rule
that
it
would
be.
This
sensor
would
be
part
of
using
the
sarif
vocabulary.
G
So
that's
profit
design
where
we
had
so
far.
If
you
go
on
our
next
slide,
we're
also
exploring
or
you're
using
occasionally
a
shorthand
design
where,
if
you
only
have
the
type
and
target
or
just
name
a
target,
you
could
simplify
the
design
a
little
and
just
have
key
value
pairs
where
the
type
or
name
is
the
key
and
the
the
target
is
the
is
the
value.
But
then
you,
you,
miss
the
capability
of
being
properties
and
and
more
details.
B
Type
is
really
rel
and
right
and
and
properties
is
really
target
attributes.
If
you
use
some
of
the
language
we
used
used
to
use
to
describe
type
links
of
the
hyperlink
type
indeed,
so
it
would
even
maybe
maybe
make
sense
to
use
rel
instead
of
type,
and
maybe
you
maybe
make
sense
to
use
target
attributes
instead
of
anyway,
we
could
argue
about
these
metrics,
but
the
pattern
makes
sense.
B
The
relations
simplified
pattern
here
without
the
target
attributes
would
be
more
limited,
but
it's
not
clear
what
it
optimizes
the
other
one
would
be
like
a
super
sad
yet
indeed,.
C
C
G
Okay
thanks
so
indeed,
maybe
it
makes
sense
completely
previous
slide
to
show
the
more
more
capable
definitions
but
yeah,
I
mean
my
good
point
I
think
is
especially
since
the
point
of
stfs,
as
the
format
is
to
be
friendly
for
developers
and
multiple
ecosystems
using
terminology
that
is
used
in
other
ecosystems.
G
So
it's
not
a
very
good
idea.
I
think
this
one
is
using
the
terminology
of
dddl
format,
but
we
maybe
we
should
explore
a
bit
what
the
other
ecosystems
are
using
and
then
agreeing.
D
B
G
Okay,
good
in
any
other
reflections
on
on
this
idea
in
general,
adding
relations
and-
and
I
guess
what
we'll
be
still
pressure-
testing
the
exact
details
of
the
format.
So,
but
if
the.
G
B
I
was
going
to
say
I'm
I'm
very
interested
in
essay,
wrap
integration
as
well
and
looking
at
sort
of
how
how
that
would
work.
And
so
I
might
have
some
examples
that
that
test
this
in
the
next
week
or
so.
C
So
I
think
it
would
be
really
useful
to
do.
We
will
use
terms
like
relation
and
type
and
so
on,
to
always
make
clear
on
what
level
of
the
metal
model
we
are
at
the
moment.
So
is
this
a
relation
between
the
the
class
that
this
sdf
type
actually
generates,
or
is
this
a
relation
that
needs
to
be
defined
for
every
instance.
G
Very
very,
very
good
point,
and
actually
now
I
realize
that
this
basic
example
actually
mix
mixes
the
two,
because
I
copied
this
from
another
place.
So
maybe
that
you've
got
the
previous
slide.
Michael
then.
The
I
guess
that
example
is
is
better.
It
only
uses
that,
let's
say
class
level
things,
but
but
indeed
whether
that
means
that
the
target
on
the
instance
level
can
be
only
of
that
class.
B
Well,
that's
interesting
because
if
you
look
at
saraps,
they've
they've
done
the
discipline
of
having
separate
relation
types
for
those
things
that
refer
to
saref
classes
versus
those
things
that
refer
to
sarep
types
that
are
defined
within
those
classes,
the
application
types
they
have
different
relations
that
you
use
for
instances
versus
classes,
but
that
you
know,
I
guess,
there's
some
question
about
that.
Good
point.
G
Good,
okay,
but
I
guess
what
I'm
hearing
and
norman
is
against
like
this,
a
general
idea,
so
we
are
definitely
gonna,
need
more
pressure
testing
and
seeing
how
different
how
it
works
with
different
ecosystems
on
the
types
and
names
of
the
qualities,
and
then
this
instance
versus
class
aspects
would
need
to
be
clarified.
G
Okay,
very
good,
then
maybe
we
move
to
the
sdf
instances.
G
So
this
is
a
related,
but
but
a
different
topic,
so
here
also
we're
proposing
an
inequality
for
different
reasons,
so
we've
been
so
far,
mostly
working
on
on
a
class
converting
classes
from
between
different
ecosystems.
Of
course,
when
you
do
use
these
classes
in
any
live
system,
you
also
need
to
add
instant,
specific
information
to
be
actually
telling
more
details
about
the
specific
instance
of
a
class
you
are
creating.
G
So,
for
example,
if
you
create
the
web
of
things,
thing
description
or
choral
documents
from
sdf
models,
you
would
need
this
information
or,
if
you're
provisioning
devices
from
different
ecosystems
that
have
specific
features
or
already
set
for
those
specific
devices
would
be
useful.
Able
to
tell
about
this
easy
information
and
we
did
discuss
in
the
core
entering
as
nicholas
mentioned,
on
this
sdf
into
coral
and
there
you
can
see
more
examples
of
how
we
were
using
this
sdf
instance
keyword.
G
One
disclaimer,
this
feature
is,
is,
let's
say,
less
thought
true
and
unrefined
than
than
the
relationships
idea,
but
here,
for
example,
one
way
how
it
could
look
like
so,
if
you're
creating
an
instance
called
room
one
you
could,
that
is
instance
of
sdf
object
room
from
this
api
names
that
will
create
you
an
instance,
and
then
you
could,
for
example,
set
specific
properties
to
specific
values.
For,
for
that
instance,.
B
G
B
What
I
was
doing
was
using
sdf
thing
to
declare
instances
and
that's
that's
the
thing.
That's
different
from
what
we
originally
may
have
thought.
Sdf
thing
was
only
used
to
declare
classes
that
contained
other
classes
or
whatever,
but
I
thought
well,
why
not?
Why
not
use
this
sdf
thing
to
also
declare
instances
and
that's
what
I
did
for
the
the
plugfest
and
the
protocol,
binding
experiment
and
all
of
that
and
it
seemed
to
you
know
it
didn't
I
mean
it
seemed
to
make
sense,
but
this
is
an
interesting.
B
B
Does
that
mean
that
if
you
have
room
okay,
let
me
see
if
you
have
a
definition
for
object
room
then
that
all
that
already
has
a
room
name
defined
and
what
you're
doing
is
just
setting
the
value
of
that
property.
So
room
name
is
already
defined
in
room
right,
yep,
okay
right!
So
it
is
this
very
similar
pattern
and
then
you're
using
constant
yeah
right.
Okay,
great.
G
G
So
here,
and
then
also
one
consideration
here-
is
that,
like
how
far
do
this
domain?
Should
we
even
go,
because
we
do
have
you
know
ways
to
express
those
instances
like,
for
example,
thing
description
and
we
shouldn't
be
reinventing
a
team
description
just
for
the
sake
of
it,
because
the
key
question
is
like
if
you
would
go
from
different
ecosystems,
is
there
some
piece
of
information
that
would
be
useful
to
be
in
sdf
that
it
goes,
makes
it
easier
to
go
either
team
description,
coral
document
you
know
like
with
mlm
stack?
B
There
are
some
differences.
What
you
were
describing
that
thing
description
and
coral
are
both
unique
in
that
they
contain
addresses
the
other
formats,
don't
so
what
you're
talking
about
is
really
model
to
model
versus
model.
To
instance,
when
I
did
the
prototype
of
model,
to
instance,
with
basically
the
what
I
was
just
describing
an
sdf
thing
that
had
some
some,
you
know
these
instance
definitions
in
it
like
nailing
down
the
values
of
things,
and
all
of
that
I
also
use
the
sdf
extension
point
going
back
to
you
know.
B
D
B
Protocol
binding
part,
and
in
doing
that,
I
made
it
really
easy
to
create
a
thing
description,
so
I
did
not
see
that
at
all.
As
reinventing
thing
description,
I
saw
that
as
creating
a
very
simple
way
of
augmenting
an
sdf
instance
to
be
able
to
create
instances
of
thing
description
in
the
description
format.
D
B
B
It
could
just
be
something
you
do
in
memory
and
you
spit
out
the
thing
description,
but
I
found
it
really
interesting
to
make
an
extension
point
for
sdf
that
had
that
syntax
in
it
to
illustrate
the
process
and
then
also
to
make
it
sort
of
super
easy
to
just
wrap
up
a
thing
description.
I
will
use
the
tool
that
you're
the
in
the
intern
made
or
the
whatever
I'm
sorry
you're.
B
B
But
basically
that
was
how
that
worked,
and
I
saw
that
as
being
super
useful
and
also
you
know
if
you
wanted
to
have
other
instance
formats
that
that
do
have
addresses,
and
things
like
that,
you
could
define
extension
points
with
those
vocabularies,
that
that
makes
it
easier
to
translate
to
those
so
just
boot.
For.
C
Thought
so
generally,
every
instance
can
be
described
by
a
class
that
has
cardinality
one.
So
of
course,
you
can
use
something
like
sdf
to
describe
instances.
C
The
the
problem
really
is
that
the
point
of
classes
is
to
to
provide
rules
for
the
instances
so
who's
who's,
providing
the
meta
model.
For
the
instance
attributes
that
you
add
to
to
the
class-
and
I
think
that
that's
really
where
going
too
far
with
sdf
as
an
instance,
description
will
start
to
hurt.
B
Well,
another
way
to
think
about
that
would
be.
I
made
a
a
mapping
file
that
had
a
bunch
of
json
pointers
and
each
of
these
json
pointers
had
some
w3c
thing:
description,
syntax
in
it.
That
was
a
hint
for
how
you
generated
the
thing
description
from
that
sdf
and
still,
of
course,
it's
an
sdf
instance,
but
the
the
the
the
the
information
that
makes
it
an
instance
is
separate
and
it
could
be
separate
in
a
mapping
file.
B
It
doesn't
have
to
be
in
those
extension
points,
the
other
things
you
know
the
schemas
and
all
of
that,
the
other
things
that
you
do
might
be
to
nail
down
the
constraints
of
values
and
that
could
also
be
in
the
separate
mapping
file.
So
I
don't,
I
don't
really
see
it
as
being
having
an
instance
needing
to
really
have
an
instance
format
for
sdf,
it's
more
like
how
you
would
incorporate
a
mapping
file
in
line,
but
your
mapping,
you
would
be
mapping
to
an
instance,
as
opposed
to
just
moving.
B
C
B
C
B
B
D
G
Yeah,
I
think,
there's
a
lot
of
commonalities
on
the
on
the
topic
of
like
map,
instances,
ecosystem
specific
features
and
what
jana
was
presenting
with
the
origins,
and
I
wonder
how
much
common
capabilities
can
can
we
build
that
would
be
serving
all
of
those
cases.
Well,.
B
A
The
expectations
of
an
instance
of
things,
and
so
on
so
maybe
being
clear
on
what
we
mean
by
an
sdf
instance
and
that
might
not
be
a
fully
fledged
run
time,
something
it's
actually
more
of
a
yeah
intermediate
instance.
Before
the
end
I
mean
I
don't
I
don't
know.
I
think
that
would
be
good
to
have
clear,
since
we
are
a
bit
branching
out
from
the
even
if
I
think
I
mean
this
is
valuable
stuff.
So
it's
not
not
that,
but
it's
just.
B
A
B
I
make
instances
of
things
that
is
everything
but
the
address
right.
So
I
make
an
instance
of
a
thing
that
might
be
a
bacnet
controller
that
has
everything
defined
and
all
grouped
and
all
the
different
affordances
and
all
the
different
constraints
and
numeric
sounds
and
and
all
of
that,
but
it
doesn't
have
the
address
and
all
it
really
needs
is
the
address.
So
it's
not
really
an
instance.
It's
a
it's
very
much
an
instance.
If
you
think
about
where
you
started
with
set
of
sdf
models,
but
it's
not
an
instance.
B
If
you're
thinking
about
a
network
snot
doesn't
have
an
address
yet
so
you
that's
the
template
from
which
you
make
maybe
50
of
these
in
a
building
or
something,
and
that
each
of
those
is
an
instance
in
your
in
your
you
know,
runtime
system,
so
yeah.
What
do
you
call
an
instance
is
is
probably
an
interesting
question
and
I
was
just
really
more
addressing
carsten's
question
of
what
is
the?
What
is
the
semantics
of
a
mapping
to
sdf
in
general?
B
A
I'm
afraid
we're
pretty
much
well
far
over
time,
we've
spent
almost
20
minutes.
I
think
this
discussion
is
something
we
need
to
have.
A
I
wonder
if
we
should
try
to
have
squeeze
in
another
interim
in
the
not
too
far
future
like
three
weeks
from
now
or
so,
would
people
be
up
for
that
or
yes,
please,.
A
So,
let's
sort
of
save
the
context
of
this
discussion
and
bring
it
up
without
the
the
kind
of
boilerplate
meeting
stuff
next
time,
let's
get
into
it,
because
I
think
this
is
good
stuff
that
we
should
have
and
great
I
don't.
I
didn't
want
to
kill
the
discussion.
It's
more
that
way.
I
think
we
are
out
of
time
when
people
are
chasing
me
here
on
my
side
that
I
need
to
end.
B
So
quickly
summarize,
we
could
prototype
this
selector
effect
sort
of
pattern,
not
so
much
as
a
hard
prototype,
but
to
test
these
three
use
cases
against
that
pattern
to
see
if
it
meets
those
requirements,
and
that
could
be
something
we
do
between
now
and
the
next
meeting.
A
You
know
when,
when
people
write
things
on
chat
because
you
usually
miss
it
yeah
good
points
there,
I
guess
yeah,
but
should
we
try
to
schedule
the
interim
three
weeks
from
now
wednesday
on
the
third.
A
Yeah,
I'm
afraid
I'm
so
my
outlook
is
not
perfect
in
showing
different
calendars.
For
some
reason,
I
don't
see
any
core
meeting
that
week
on
the
23rd,
but
then,
on
the
other
hand,.
E
I
know
karsten
still
has
a
conflict,
but
you
know
why
don't
you
I
don't
know
whatever
I.
I
have
meetings
back
to
back
the
whole
time
so.
H
D
E
A
I
have
a
keller
sink
problem
here.
Obviously,
so,
let's
try
to
fit
in
a
meeting
that
week
week,
25.
A
A
A
No,
actually,
not
our
house,
but
it's
something
very
far
from
here:
no
international
traveling,
so
carson
you
can
you
can't
make
it
on
wednesday.
B
B
B
Oh
pacific,
eight
to
nine
eight
to
nine
thirty.
Basically,
in
the
whole,
this
exact
range.
So
you
have
to
stand.
C
B
90
minutes.
Oh
no,
there
are
three
three
15-minute
ones
because
I'm
on
I'm,
I'm
literally
I'm
I'm
I'm
a
virtual
member
of
a
bunch
of
different
teams.
Oh
my
god!
That's
the
only
way
they
can.
They
have
to
make
it
work
with
agile.
They
still
are
stuck
on
having
to
have
the
agile.
So
this
is
like
the
hoops
I
have
to
jump
through.
To
do
agile.
A
I
don't
think
we
managed
to
set
the
time,
but
that's
what
yes,
let's,
let's
do
it
for
something
that
week
great
design
time
meeting.
Thank
you
very
much
everybody.
Thank
you.
I
really
apologize
for
messing
up
the
time
planning
here.