►
From YouTube: T2TRG Interim Meeting, 2021-11-29
Description
T2TRG Interim Meeting, 2021-11-29
A
Okay,
then
perhaps
perhaps
we
we
get
get
going
so
sebastian,
I
don't
know
if
you
had
anything
prepared
for
the
previous
meeting,
you
would
like
to
get
going
directly
or
is
it
easier
to
start
with
an
overview
of
what
is
the
latest
in
the
td
and
tm
happening
of
things?
B
Well,
I
have
some
slides,
which
gives
a
quick
overview
about
why
thing
model
and
what
kind
of
features
we
have
so
far.
B
I
can
quickly
go
over
the
slides
and
after
that
we
can
make
discussion
on
relation
to
sdf.
So
we
have
some
experience
on
plugfest
about
this.
So
typical
we
have
a
lot
of
sdf
samples
which
are
transformed
to
thing
models
or
two
to
a
single
description,
so
yeah.
It's.
I
put
my
proposal
to
start
with
with
this
approach,
yeah.
So
to
give
you
a
quick
update
about
what's
going
on
at
w3c,
and
then
we
can
move
on
to
comparison
and
questions
and
so
on.
C
B
Yeah
does
it
looks
good.
E
B
What
do
you
see
now?
I
should
have
seen
your
movie
okay,
then.
Maybe
let's
go
this
way,
I'm
not
going
to
presentation
mode,
okay,
well
right!
So
yeah.
Thank
you
for
having
me
here
and
giving
the
chance
to
give
you
the
latest
about
the
thing
model.
B
So
this
are
some
slides
which
I
use
for
my
lecture.
So
so
they
are,
I
would
say,
a
very
detailed
full
with
a
lot
of
examples,
but
maybe
it's
quite
quite
nice
to
have
all
the
good
feeling
for
what
you
can
use
some
approaches
here,
and
so
that's
the
reason
why
I
have
so
many
examples
here
in
this
slide.
So
right
just
to
give
you
a
quick
motivation.
Why
was
this
now
introduced?
B
Also
in
web
of
things,
the
scene
model
approach
and,
as
you
know,
we
have
two
different
life
cycle
aspects
of
iot
equipment
and
and
starting
from
from
designing
iot
devices
until
it's
installed
by
the
customer
and
deployed
by
the
customer,
you
have
different
phase,
which
different
kind
of
yeah,
I
would
say,
meter
information
are
available
and,
as
you
know,
the
thing
description
is
a
description
of
a
device
that
is
really
active.
B
It's
online
or
accessible
you
prov,
you're,
getting
all
the
detailed
information,
that's
inscription
how
to
yeah
to
get
to
specific
information
from
the
device,
and
it
also
means
you
have
all
the
metadata
like
ip
address
and
also
the
security
meter
data
so
that
you
know
what
kind
of
security
mode
you
must
apply
to
have
a
very
successful
connection
to
uit
equipment
and
yeah,
and
the
situation
is
sometimes
that
you
have
things
designed,
which
maybe
you
have
a
different
setup
regarding.
B
I
don't
know
protocols
or
security
modes
and
so
on
and
and
of
course,
you
have
sometimes
also
not
all
the
information
required
to
some
specific
point
like
how
is
the
ip
address
which
is
in
the
future,
used
yeah.
So,
as
you
know,
ip
trust
typical
provided
in
the
thing
description.
B
However,
when
you
buy
this
equipment
here
like
in
here,
then
your
typical
ip
address
is
not
known
yet
because
it's
then
just
known
when
it's
installed
by
the
customer
and
you're
assigned
an
ip
address
right
and
the
question
is,
if
there's
a
way
how
we
can
design
already,
I
would
say
only
the
capabilities
of
the
thing.
B
Without
going
into
the
details
of
communication,
data
metadata
and
and
serial
key
security
meter
data-
and
that
was,
for
example,
one
of
the
motivation
here
to
have
something
which
giving
us
the
opportunity
for
doing
so
and
another
aspect
which
we
like
to
have-
and
this
is
what,
for
example,
also
sdf
having
on
board
is
the
opportunity
to
adapt
already
existing
definition
or
basic
definition
from
other
models
yeah.
So,
let's,
let's
assume
we
have
some
different
kind
of
lamps
here,
one
of
them.
You
can
change
the
colors
and
you
can
dim
and
all
the
stuff.
B
And
then
you
have
basic
lamps,
which
is
providing
basic
information
like
the
status
or
you
have
the
functionality
to
switch
on
and
off,
and-
and
this
is
typical,
but
the
basic
functionality
which
you
always
have
also
also
in
the
smart
lamps
here,
which
you
have
more
functionality
yeah
and
the
question
is
how
you
can
simply
adapt
existing
definition
from
some
basic
thing:
models
or
thin
definitions
and
adapted
to
do
not
again
this
redefinition
right.
B
So
there
was
the
request:
how
we
can
simply
reuse
existing
knowledge
and
capabilities
in
new
thing
models,
yeah
and
yeah,
and
the
answer
is
that
we
have
to
find
the
thing
model
approach
so
maybe
to
to
give
you
just
a
background
about
the
history.
So
in
the
1.0
version
of
the
thing
description
we
had
had
already
this
approach
compiled
in
there.
However,
it
was
called
differently.
We
called
this
thing
description
template
that
time.
B
However,
since
td
1.1,
we
have
it
called
zing
model,
because
the
term
makes
more
sense
than
this
template
approach
and
yeah.
B
We
have
also
more
or
less
a
formal
definition
here,
so
I
think
the
script
I
think
model
is
a
description
for
a
class
of
things
that
have
the
same
capabilities,
describes
the
properties
actually
events
and
common
metadata
that
are
shared
for
entire
group
of
things
compared
to
I
think
description
model
does
not
contain
enough
information
to
identify
or
interact
with
your
thing
instance
as
well,
especially
last
sentence
make
very
clear
that
the
thing
model
is,
I
would
say,
a
uncompleted
thing
description
since
typically,
you
will
not
find
very
specific
communication
metadata
in
there
to
make
interaction
with
the
thing
right.
B
This
is
a
very
small
example
of
a
thing
description
on
our
lamp,
so
you
find
your
simple
interaction:
affordance,
that
the
lamp
is
providing
here.
So
you
find
a
property
status
and
action
toggle
and
event
overheating,
and,
as
you
can
see
here,
there's
no
additional
information
provided
about
communication,
metadata
yeah,
so
typical
we
would
only.
I
would
also
have
forms
in
here
to
get
information
what
I
have
to
do
to
set
up
the
the
network
interface
and
to
make
the
right
request
here
for
the
status,
as
example.
B
The
same
is
also
true
for
security
meter
data
and
so
on
so
yeah-
and
this
is
a
valid
thing-
this
testing
model
here
just
a
comparison
to
a
thin
description.
I
like
this
kind
of
illustration
here,
so
the
model
a
typical
kind
of
stamp.
We
are
missing
information.
B
Typical
in
there
and
when
we
want
to
create
a
valid
thing
description
out
of
the
thing
with
the
model,
then
all
the
missing
information
have
to
provided
there
to
to
the
thing
model
to
have
done
by
the
thing
description,
yeah,
and
this
also
shows
a
thing.
What
is
also
kind
of,
I
would
say,
class
definition
of
things
yeah,
where
you
can
instantiate
multiple
thing:
descriptions
based
on
the
number
of
devices
you
created
for
this
specific
class
system.
B
B
So
we
are
using
this
value
here
to
identify
what
is
the
thing
model
and
yeah,
and
then
we
have
a
couple
of
design
tools
to
yeah
design
thing
models
with
other
thing
models:
here's
books
on
the
use
case
if
you
like
to
extend
existing
thing
model.
So,
let's
assume
we
have
here
a
basic
thing
model
definition
like
this
on
off
and
then
we
would
use
the
links
container
to
say:
okay,
we
are
pointing
to
existing
thing
model
here
and
and
say
simple
extended.
B
That
means
that
the
instances
of
the
thing
model
so
when
it
comes
to
the
thing
description,
build
and
also
in
have
india
the
property
on
off.
In
addition
to
this
dim
property
here,
and
then
we
have
also
this
reference.
This
is
actually
the
same
approach.
What
you
have
in
sdf
as
well,
so
you
make
a
simple
reference
to
to
a
place
of
a
existing
definition
somewhere.
B
So
you
simply
say
where's
the
location
of
this
document
and
then
you're
using
json
pointer
to
say
please
copy
or
clone
everything
what
is
defined
somewhere
else
or
like
in
here.
So
you
say
you
want
to
reuse
the
definition
of
the
dim
definition
here
in
this
new
property,
which
is
called
dimming
and
and
what
is
also
possible
that
you
are
able
to
override
existing
definition.
B
So
that
means
that
you
simply
override
the
maximum
here
and
you
add,
extend
all
the
additional
values
like
the
unit
term
here,
so
you
you're,
providing
additional
information
here,
and
we
have
also
a
kind
of
restriction
of
this
overriding
approach,
and
this
is
explained
in
this
slide
that
you
should
be
also
careful
about
this
overwriting
topic,
because
here's
the
request
that
all
the
instances
of
the
new
definition
should
be
valid
also
to
the
old
definition
yeah.
So
let
me
take
this
dimming
property,
which
is
copied
from
from
this.
B
This
example
here
and
you
see
the
maximum
is
overwritten
by
120
yeah.
So
now
is
the
danger
that
instances
can
be
higher
than
120
so
like
110,
and
this
will
be
then
valid
regarding
this
definition.
However,
it
will
be
not
valid
regarding
to
the
to
the
source
definition
or
to
the
to
the
origin
definition
anymore,
and
that's
the
reason
why
it's
not
allowing
to
this
in
that
array.
B
Then
yes,
for
example,
also
to
risk
that
you
have
some,
for
example,
some
floating
values
here
and-
and
if
this
is
the
case
like
20.5,
then
this
will
be
not
valid
regarding
this
integer,
which
expects
some
discrete
values
in
here
right
well
and
this
example
will
be
correct.
So
this
extends
this
existing
definition.
However,
overwrites
this
dim
approach
or
the
stim
definition.
B
However,
it
is
in
the
range
of
this
source
definition,
so
that
means
all
the
instances
that
are
created
here
out
of
this
new
definition.
Here
they
are
definitely
valid
to
the
to
the
source
definition.
B
Then
we
have
also
this
composition.
This
is
also
a
very
useful
approach
to
yeah
to
say,
I'm
like
to
create
a
new
system
based
on.
B
This
composition,
simple
by
this
links
container,
you
say:
okay,
this
new
system,
swatch
ventilation,
consists
of
ventilation,
thing
model
and
on
this
led
sting
model,
and
what
you
should
also
provide
here
is
a
instance
name.
So
this
is
a
way
to
give
a
kind
of
identifier
of
this
thing
model.
So
this
is,
for
example,
useful
for
doing
so.
B
If
you,
for
example,
have
a
system
which
uses
a
multiple
different
kind
of
sub
models,
multiple
same
ventilation
or
same
led
lamps,
and
then
you
can
provide
a
own
instance
name
for
each
model
of
the
led
to
make
a
differentiation
of
them
in
the
application
later
on.
B
Personally,
I'm
not
very
happy
about
the
instance
name
term.
Maybe
it's
still,
maybe
it's
getting
renamed
in
the
future,
but
at
the
moment
this
is
the
working
assumption
well
and
then
we
have
also
provided
a
algorithm.
How
you
can
generate
things
description
out
of
this
thing:
model
approaches
yeah.
I
don't
want
to
jump
over
each
step
here,
what
you
have
to
do
there,
but
there
are
some
replacing
process
needed,
and
then
you
have
to
replace
the
placeholder
identifier.
I
will
show
you
what
this
is
about
actually
and
so
on.
B
There's
some
very
simple
processing
steps:
how
to
generate
out
of
thing
models,
thing
description,
yeah!
This
is
the
example
which
also
shows
we
have
also
this
tm
required
what
you
also
have
in
sdf-
and
this
simple
says:
okay,
if
you
have
some
required
properties,
definitely
that
defined
then
this
should
be
definitely
showing
also
show
up
in
the
single
description
instance.
B
Then
right
and
as
you
can
see
here,
this
is
not
a
valid
thing
description
regarding
this
thing
model,
since
it's
expecting
two
interactions
here,
namely
the
property
status
and
the
action
toggle,
and
here
only
the
status
is
present,
and
here
we
have
a
valid
one
and
the
next
one
is
also
valid,
because
this
is
optional.
This
event-
and
this
also
means
it
can
be
simple-
also
reflected
by
this
in
description.
B
So
it
would
be
also
valid
to
have
it
here
well
and
then
we
have
the
placeholder
concept.
The
placeholder
concept
is
simple
approach
to
give
in
the
thing
model,
some
yeah,
I
would
say
some
places,
which
should
be
then
replaced
when
it
comes
to
generate
thing
description
instances.
B
So,
for
example,
here
you
want
to
give
a
specific
pattern
of
the
title,
then
you
say
you
can
say
your
thermostat
number
data,
so
the
tester
then
replaced
also.
What
you
can,
of
course,
do
is
to
provide
first
information
of
the
protocol
metadata,
so
you
can
provide
a
scheme
value
here.
However,
you
have
done
here
to
replace
this.
B
This
part,
with
the
ip
address
of
the
program
that
you're
going
to
use
and
so
on,
then
you
have
also
other
places
where
you
can
provide
this
placeholder
and
if
you
want
to
generate
thing
description
out
of
the
theme
model
with
the
placeholder,
then
you
should
have
this
kind
of
placeholder
map
can
be
any
kind
of
a
map
approach
here.
So
the
specification
does
not
force
to
use
this
in
a
json
format.
You
can
do
it
also
and
that's
different
format,
but
this
is
just
example
how
this
can
look
like.
B
So
there's
a
simple
lookup
of
this
identifier
and
it's
a
simple,
replacing
then
the
parts
with
the
value
that
is
provided
in
the
map.
So
what
I
p
addressed:
the
maximum
value,
the
observable
situation
and
so
on-
and
this
is,
I
think,
a
very
cool
approach
when
it
comes
to
mass
deployment
that
you
provide
a
specific
batch
of
placeholder
maps
for
specific.
B
Well-
and
this
is
the
last
slide,
so
we
have
also
a
tool
which
calls
editor,
which
helps
you
also
designing
thing
models,
and
it
also
shows
nicely
the
dependencies
of
the
thing
models.
B
So
if
you
have
to
find
some
dependencies
like
to
your
extension
models
or
you
have
a
composition
and
so
on,
then
you
can
simply
also
this
here
is
a
nice
graph
here,
but
in
general
it
is
a
tool
which
gives
you
opportunity
to
edit
your
file
in
the
text
form
or,
if
you
like
it
more,
I
would
say
track
and
drop
or
more
on
this
assistance
diet,
and
you
can
also
call
this
a
graphical
approach
here
and
design
your
thing
description
or
seeing
models,
if
you
like
it.
B
Well,
that's
the
last
slide
yeah.
Now
I'm
open
for
questions
and
discussions.
A
We're
very
good
thanks
a
lot
sebastian.
This
was
this
was
great
first
obvious
question:
can
we
get
a
copy
of
the
slides.
A
D
Can
a
thing
model
have
a
form
you
showed
your
example
without
forms,
but
what,
if
I
wanted
a
form
that
had
like
for,
I
think
about
bacnet,
for
example,
where
you
have
things
like
the
you
know:
object
id
and
things
like
that
that
are
fixed
on
the
device,
but
you
just
don't
have
the
ip
address
so
with
bacnet.
I'd
want
a
form
that
would
have
separate
properties
for
things
like
object
id
and
all
that
when
I
generated
a
td
I'd
want
to
generate
the
url
for
for
the
bacnet
from
that.
B
Yes
sure
so
there
is,
I
mean
not
this
limitation
on
on
what
is
not
allowed,
I
would
say
yeah,
so
you
can
of
course
also
have
forms
in
there,
and
so
you
can
simply
have
this
and
then
place
it
here
and
to
remove
this,
and
then
you
can
say.
Okay,
I
have
an
entry.
D
D
Yeah,
so
I
I
guess:
I'd
want
to
point
out
that
in
sdf
we
just
use
json
pointer,
so
we
don't
really,
but
that's
what
you
would
when
you
do,
the
transform
you
would.
You
would
put
placeholders
in
that
that
map
would
reference
the
json
pointers
that
are
in
what
we
call
the
mapping
file
or
mapping
information.
But
it
seems
like
that's
what
the
equivalent
in
sdf
would
be
just
the
use
of
json
pointers.
B
Yeah
I
mean
you
can
do
this
the
same
here
right.
You
can
also
use
the
tm
ref
right
and
you
can
also
copy
the
stuff
in
there
right
on
this
level
yeah,
but
this
would
also
work
and
what
was
also
work
is
that
you
have
something
like
that.
Did
you
say
you
have
a
placeholder
on
this
on
this
level?
You
had
to
say,
okay,
here,
that
you
need
some
forms
info
yeah,
which
provided
by
the
mapping
file.
Somehow
right
would
be
also
possible,
so
you
can.
B
Provide
as
much
many
information
you
would
like
to
have,
or
does
it
make
sense
to
to
to
to
be
generic
enough
and
the
only
thing
which
will
say
what
not
makes
sense
to
provide
in
the
same
model
this
which
this
goes
more
in
this
instance
specific
information
yeah
so
like
the
ip
address
or
or
ids
of
something
or
like
something
like
that.
So
if
this
is
a
more
or
less
not
not
required,
and
everything
is
possible
right.
D
So
also
that
was
my
final
comment
was
your
use
of
the
word
instance.
D
D
E
B
E
B
Yeah,
this
is
the
composition
thing
here
that
you,
you
have
six
existing
things,
yeah
like
lamps
and
engine,
I
don't
know,
and
you
want
to
create
out
of
them
a
new
system
right.
D
B
Exactly
yeah,
so
that's
the
question
if
the
instance
name
is
taking
over.
I
don't
know
in
the
in
the
title
of
the
thing
description.
I
don't
know.
D
The
context
is
we're
having
a
currently
having
a
discussion
in
the
sdf
about,
what's
an
instance
and-
and
so
this
seems
to
be
another
example
of
people
using
the
word
instance
to
describe
something
that
isn't
really
network
accessible.
Yet
yeah.
F
D
D
F
D
B
D
Definition
that
uses
another
definition
and
then
we
can.
We
can
have
what
I've
been
calling
successive
refinement
or
you
can
have
several
layers.
I
can
define
what
what
a
bacnet
object
is
and
then
I
can
define
an
analog
input
object
and
then
I
can
define
a
a
a
temperature
measurement.
That's
an
analog
input
and
all
three
of
those
can
be
defined
by
saying
more
about
adding
more
constraints
like
units
and
things
like
that.
My
first
one
doesn't
have
any
units.
The
second
you
know
I
only
add
units
at
the
third
stage.
D
B
So
the
composition
is
the
very
newest
one
there.
This
was
a
request
from
bosch
to
have
this,
so
so
they
would.
They
had
a
lot
a
lot
of
use
cases,
and
I
see
also
this
a
lot
of
potential
of
of
use
cases
or
scenarios,
and
so
the
composition
is
very
the
last
one
and
at
the
moment
we
don't
have
further
feature
requests.
B
So
at
the
moment
I
would
say
it's
kind
of
stable
and
so
yeah,
I
would
say
it's
now
put
in
the
1.1
version
of
the
thing
description.
Maybe
what
is
awesome.
B
Makes
sense
to
say
here
we
will
maybe
in
a
new
charter,
move
the
section
what
we
have
as
a
thing
model
move
to
our
own
specification
document,
because
I
think
it
makes
sense
to
to
consider
the
thing
model
more
separately
from
the
theme
description
and
not
being
a
part
of
the
thing
description
itself,
because
it's
getting
more
and
more
intention
to
work
with
the
single
model
approach,
since
it's
quite
interesting
from
the
life
cycle
aspect
to
to
to
have
something
at
the
designing
approach
which
can
be
used
from
the
very
beginning
and
and
and
the
part
how
to
generate
things,
description
or
singing
models.
B
Then
yeah.
Another
aspect
to
be
considered.
But
at
the
moment
I
would
say
yeah
it's
to
call
to
to
test
the
thing
model
approach
as
it
is
to
provide
comments.
So
we
have
still
time
until
march
for
going
the
candidate
recommendation.
B
So
we
have
still
time
to
yeah
to
to
modify
or
optimize
approaches
what
you
what
I
have
shown
here.
But
at
the
moment,
from
the
point
of
view
of
features,
I
would
say
it's
everything
is
covered
now.
A
And
no,
no,
no,
I
was
wondering
like
like,
if
we,
for
example,
would
build
more
codes
to
get
going
between
stf
and
tds.
Now
is
the
moment
I
sorry
between
sdf
and
tm.
It's
like
now
it's
the
moment
to
look
at
the
spec
and
see
like
how
how
features
map
are
we
missing?
Something
should
something
maybe
that's
made
easier.
B
Yeah
I
mean
we
have
this
transformation
code,
I
mean
also
in
the
testing
folder.
We
have
a
plenty
of
examples
of
sdf
and
thing
model,
for
example,
again
has
a
couple
of
provided,
whereas
it's.
B
B
So
this
are,
for
example,
thing
models
that
I
used,
one
of
them
last
test
fest
and
here's,
this
converter
tool,
which
we
used
or
which
I
used
that
time.
So
a
student
of
mine
have
developed
it
as
a
tool
to
transform
sdf.
B
A
B
A
I
guess
we
don't
have
time
to
look
at
these,
but
thanks
a
lot
for
the
for
the
update
and
actually
a
colleague
of
mine
made
a
bit
of
a
demo.
Where
also
integrates
your.
I
think
it's
your
tool
with
the
other
translator
tools,
so
we
can
go
with
from
sdf
to
all
the
other
ecosystems
and
and
back
so
I
think
we
are
that
that
sounds
sounds
very
good
like
now.
Is
the
right
time
to
look
at
also
the
details
of
like
what?
What
are
we
potentially
missing
there.
A
A
D
With
sdf,
where
I
I
made
a
an
extension
to
sdf
that
that
uses
thing,
description
and
thing
model
elements
like
data,
schemas
and
forms,
and
so
basically
in
the
workflow,
you
can
say
what
elements
of
data
schema
you
want
to
bind
to
a
thing
using
a
mapping
file
and
also
what
form
you
can
bind
to
a
an
sdf
definition.
And
then,
when
you
generate
the
thing
model
or
thing
description,
it
uses
that
information,
as
as
hints
or
as
as
guidance
in
the
construction.
D
We
have
proposed
extensions
for
sdf
as
a
the
mechanism
for
how
you
do
that,
but
also
the
content
in
terms
of
being
able
to
use
vocabulary
from
thing
description
or
like,
for
example,
I
use
the
modbus
vocabulary
to
generate
the
thing
description
from
that
it
has
to
come
from
somewhere
and
in
one
workflow
envisioned
workflow.
It
comes
from
the
sdf
file
as
an
extension.
A
Okay,
very
good,
I
mean
now
looking
at
the
time
of
this
being
super
interesting.
I
think
we
only
have
a
bit
more
than
15
minutes
till
the
end
of
the
session,
so
I
think
sebastian.
We
definitely
need
to
catch
up
later
later
again
on
on
more
of
this
so
yeah,
and
I
think
in
particular
I
want
one
part
that
we
didn't
have
a
no
chance
to
discuss.
Is
this:
how
do
we
go
from
sdf
into
yeah?
A
I
hate
to
use
the
word
instance,
but
you
know
something
like
you
would
use
a
td
to
describe
of,
and
I
guess,
if
you
go
for
td,
then
maybe
sdf
into
tm
and
then
tm
into
td
might
be
the
right
approach.
A
Then
the
question
is
like:
where
do
we
get
that
extra
piece
of
information
and
like
by
the
addresses
and
and
such
and
now
we've
been
discussing
it,
whether
it's
a
mapping,
file
or
something
else-
and
I
think
like
using
what
you
guys
have
learned
here
on
going
between
tm
and
td,
probably
makes
a
lot
of
sense.
B
Yeah
sure
yeah
I
mean,
and
as
I
mentioned,
I
am
still
everything
is
quite
open
in
in
terms
of
definition,
and
maybe
it's
not
very
practical
to
have
it
that
way
and
yeah.
I
would
be
happy
to
to
close
exchange
with
you
and
it
would
be
very
cool
that
we
are
definitely
looked
at
sdf
and
thing
model
are
working
very
smoothly
together,
yeah,
so
that
that
there
should
be
not
a
conflict
in
in
from
the
design
point
of
view.
B
So,
however,
I
am
the
the
student
which
I
asked
to
make
this
tool
and
you
make
also
it
is
comparison
and-
and
it
sees
this-
and
there
was
actually
not
really
a
plotting
point
in
anything
that
seems
to
not
to
walk
to
each
other.
So
everything
should
be
yeah
everything.
What
you
define
sdf
should
be
also
reflectable
in
the
thing
model
and
vice
versa.
Yeah
I
mean
unless
something
is
new
defined
in
sdf,
so
maybe
I'm
not
ever
aware
of
of
all
the
new
features.
B
What
you
have
maybe
designed
in
a
couple
of
weeks
or
months,
then
maybe
you
need
a
new
look
on
this,
but
my
experience
is-
and
I
don't
know
michael.
You
also
had
also
also
always
a
look
on
this
I've,
not
a
feeling,
there's
some
contradictionary
contradiction
or
some
problematic
definitions
somewhere,
which
may
be
resolved
some
problems
when,
when
you
like
to
convert
from
one
format
to
another.
D
No
there
there,
I
don't.
I
don't
think
there
are
any
any
show
stoppers
that
there
were
probably
some
things
that
you
know
that
we'd
like
to
align
like
read-only
for
readable
and
this
sort
of
thing,
and
also
you
know
to
to
really
nail
the
structure
part,
because
what
it's
all
about
composition,
I
don't
have
any
really
any
use
for
anything
that
does
isn't
composed.
That's
just
the
they're,
the
basic
elements,
but
they
have
to
get
composed
before
you
can
really
use
them
in
the
system.
D
So
I
think
those
that's
really
where
we
need
to
focus
on
how
we
build
systems
and
how
the
workflow
works.
When
we
build
systems
but
yeah,
I
agree
with
you.
There's
no
there's
no
main
show
stopper
it's
it's
just.
There
are
just
different
ways
of
doing
a
few
different
ways
to
do
things
and
we
need
to
sort
of
figure
out.
What's
you
know
whether
it
matters
and
where
it
matters
pick
a
good
way.
B
And
maybe
one
thing
which,
when
you
mention
how
to
generate
thing
description,
I
mean
one
thing
I
have
not
explained
to
you
is
how
to
generate
things
description
out
of
a
composition
definition
and
in
the
td
spec.
There
are
two
approaches
provided
for
doing
so.
One
approach
is
that
you
keep
the
structure
as
it
is
so
so
you
have
for
each
thing
model
then,
especially
the
thing
description
representation
of
that.
B
So
here
we
have
the
smart
ventilation
again,
and
then
we
have
the
model
for
the
ventilation
tm
and
the
led
tm
and
when
it
comes
to
generating
descriptions.
So
if
you
do
it
only
for
one
for
one
device,
then
you
have
one
thing:
description
for
yeah
for
the
top
level
thing
model
yeah,
then
you
have
to
see
description
of
this
sub
system,
the
ventilation,
and
then
you
have
the
td
for
the
led
subsystem
and
often
they
are
linked
to
each
other
right.
B
B
There's
also,
another
approach
is
to
describe
what
I
call
self-contained
description,
where
you
have
all
the
information
provided
in
a
single
thing
description,
and
there
can
be,
however,
the
risk
of
name,
collision
or
name
clashes
and
to
avoiding
this
by
using
simple.
Then
the
instance
name
what
we
have
provided
in
the
links
container
of
the
composition
as
a
kind
of
prefix
of
of
the
of
the
properties
which
coming
from
the
subsystem
right.
B
So
these
are
the
property
from
the
ventilation
and
and
then
we
have
this
actions
from
the
led
and
we're
using
simple
the
instance
name
as
a
kind
of
prefix
and
underscore
yeah.
So
this
was
one
approach
or
one
idea
how
you
can
also
define
a
single
thing,
description
on
a
composed
definition
yeah.
So
this
is
the
instance
name,
and
this
is
then
used
as
a
prefix
when
it's
when
you
like
to
have
the
thing
description
in
the
in
the
single
file
right.
D
Well,
my
takeaway
is
still
that
we,
we
have
a
lot
of
different
ways
of
doing
things,
and
it
looks
like
what
that
you're
looking
at
a
few
different
ways
too.
I
don't
know
if
we're
looking
to
kind
of
settle
on
a
best
practice
or
to
allow
a
bunch
of
different
ways,
depending
on
the
use
case
any
anyway,
that's
kind
of
where
the
discussion
going.
B
D
A
Okay,
very
good
thanks
a
lot
sebastian.
That
was
a
good
update
and
good,
and
I
think
we
have
a
good
way
forward
here
to
explore
the
synergies
and
opportunities
there
together
good.
So
then
it's
no
more
final
questions
comments
there.
Let
me
do
spend
the
remaining
five
to
ten
minutes
on
the
on
the
last
topic,
the
connecting
ipso
devices
to
ddr
applications,
hacking
so
yeah.
A
So
here
the
idea
was
to
take
the
challenge
that
we
discussed
with
the
with
the
dddl
team
and
actually
did
some
practical
hacking
on
providing
the
device
is
described
with
sdf,
for
example,
an
ipso
device
into
the
ddl
platform
via
the
transformation
that
we
can
do
enabled
by
by
sdf,
and
here
we
would
need
to
do
a
a
bit
of
planning,
perhaps
to
kick
off
our
hacking,
hacking
activity
and
like
let's,
we
could
either
have
a
specific
hackathon
day
on
it
or
we
could
do
some
work,
work,
offline
and
gather
later
on,
but
maybe
the
first
things
would
be
to
document
a
little.
A
So
with
cars
kirsten?
You
were
actually
the
one
who
proposed
this
grand
challenge
originally.
Do
you
want
to
say
a
few
words
of
introduction?
What
what
do
you
have
in
mind
here.
C
Yeah,
basically,
the
idea
was
that
we
take
an
existing
sdf
model
and
if
so,
is
probably
a
good
one,
because
we
have
lots
of
experience
with
the
the
underlying
mechanisms
and
build
an
actual
dtga,
based
instance
of
that
in
in
the
asia
system.
C
C
So
you,
you
actually
can
look
like
an
episode
instance
through
a
converter
that
talks
to
the
dtd
instance,
and
that's,
of
course,
just
just
one
concrete
form
of
doing
what
we
we
want
to
do
in
this
semantic
converter
approach.
A
Hey
thanks
carsten
any
questions
comments
here
before
we
go
on
the
practical
way
forward.
F
A
Yeah,
well,
actually,
that's
that's.
That's
a
very
good
point.
Probably
I
would
start
the
easy
way
around
use,
something
like
the
you
know.
The
eclipse
slash
on
demo
live
with
them,
client
that
you
know
works
out
of
the
box
and-
and
you
can
have
you-
can
have
a
few
simulated
sensors
exposed
and
then
and
then
perhaps
put
that
on
on
the
lesson
live
with
the
management
server
and
through
the
management
server
apis.
A
You
would
expose
these
using
sdf
semantics
and
and
then
you
bring
the
sdf
models
to
there.
So
you
convert
the
aesthetic
models
to
ddl
and
then
from
dtdl
provide
a
a
binding
towards
deletion
apis,
and
you
would
then
need
to
build
on,
of
course,
on
the
population,
something
that
understands
the
stf
semantics,
but
that
should
be
pretty
much
the
same
content
as
we
have
in
the
mapping
files
for
the
ipsum
rules.
A
D
So
it
sounds
like
the
architecture
that
we
that
we're
envisioning
is
very
similar
to
the
architecture
that
we've
been
talking
about.
When
we
were
talking
about
the
ocf
to
lightweight
mtm
stuff,
that
the
semantic
bridge
translator
converter,
it
seems
like
a
similar
architecture,
you,
you
would
have
to
create
a
model
bridge
and
then
have
a
binding.
A
A
D
Maybe
it
could
even
be
thing
web
bing
web.
F
A
Yeah
because
I
think
yeah
team
web
has
a
lot
of
these
bindings
already
implemented,
and
if
we
can
generate
tds
out
of
out
of
this
model
all
of
these
models,
then
we
should
be
able
to
generate
enough
information
for
the
team
web
implementation
to
actually.
D
So
we
say
here
here,
you
know:
lightweight
m2m
is
sort
of
easier
because
you
have
this
other
sort
of
direct
path
of
doing
it
or
maybe
not,
but
but
also
that,
if
you
had
a,
I
had
to
adapt
to
a
lot
of
different
protocols.
This
is
easily
usable.
This
node
wad
engine
is
easily
usable
to
a
bacnet
and
modbus
and
mqtt,
and
whatever.
A
Indeed-
and
I
think
was
when
we
maybe
maybe
the
next
first
steps
would
be
set
up
a
bit
of
an
infrastructure
from
from
our
side
that
we
can
point
to
with
the
with
the
ddl
models,
and
then
they
have
a
have
a
chat
with
that
with
the
brian
and
and
his
team
on
like
where,
where,
where
do
we?
What
is
the
easiest
ways
from
from
the
ddl
or
or
mark
might
or
as
your
digital
twins
infrastructure
to
communicate?
A
And
whether
would
it
make
sense
to
have
the
node
water
in
between
you
know
not
what
could
provide
the
http
apis
or
whether
it's
go
to
directly
from
there
to
the
lesson
management
server
apis
and
in
particular
it
would
be
at
the
same
time
interesting
to
think
about
the
how
we
can
go
between
all
three.
You
know
ipso
ddl
and
ocf,
because
that's
maybe
could
be
guiding
our
architecture.
Our
decisions
here
to
think
about
three
ecosystems
at
once.
D
A
Indeed,
indeed,
and
actually
maybe
sebastian
do
you
happen
to
know
on
the
note,
what
has
there
been
any
development
on
the
light
with
mdm
site?
I
don't
know
it
has
a
co-op
implementation
or
from
the
beginning,
but
has
anyone
experiment
like
with
them
to
them.
B
I
think
there
should
be
no
problem
to
use
it
directly.
I
think
there
is
only
this
situation
that
that
you
always
get
the
full
payload
on
the
application
side.
So
if
you
requesting
a
property,
then
the
lightweight
mdm
has
our
own
predefined
payload
structure,
right
and-
and
this
is
a
full
path
to
the
application
so
which
I
think
it's
not
very
nice,
but
it
would
work
yeah.
So
then
you
have
to
work
on
the
that's
where.
D
I
have
worked
on
what
I
was
calling
a
semantic
api,
but
what
it
actually
is
is
a
way
of
of
extracting
things
from
the
payload
based
on
an
annotated,
json
schema
that
has
json
path,
pointers
into
it,
and
so
that
was
like
halfway
to
being
able
to
to
do
the
semantic.
So
what
you're
saying
is
node
lot
currently
does
the
protocol
translation,
but
we
would
have
to
build
the
semantic
translation
layer
on
top
of
that
at
least
the
payload
translation.
D
B
B
Yeah
yeah:
this
is
what
we
quite
often
have
this
discussion.
I
think
we
have
to
work
on
this
in
the
td
2.0,
so
to
have
a
nice
solution
for
this
discussion.
B
Unfortunately,
when
we
yeah,
unfortunately,
I
have
to
go
to
the
next
meeting
yeah,
it
was
fun.
I
will
send
you
the
the
slide
and
yeah.
Then,
let's
see
us
in
one
of
the
next
meetings
right.
A
A
Okay,
yeah.
Indeed
we
are
over
time,
so
I
guess
we
had.
This
is
a
good
time
to
conclude
also
or
for
the
rest
of
us,
but
and
let's
see,
I
guess-
we're
gonna
have
a
another
session
on
the
on
this
topic.
A
D
It's
a
reminder
about
one
dm
meeting
later
today
and
our
design
team
meeting
on
wednesday.
C
Yeah,
I
will
be
late
as
always.
I
hope
I
can
get
rid
of
this
regular
yang
zebra
thing
soon,
but
right
now
it's
conflicting.