►
From YouTube: Unimatrix Technical Board call 2021-08-04
A
B
B
B
So
I
do
the
homework
about
the
previous
tv
call
and
so
at
first
so
I
brought
a
stakeholder
accepted
device
broker
to
introduce
to
the
camera,
and
so,
however,
so
it
seems
so,
the
further
investigation
is
needed
for
confirming
the
behavior
on
the
camera
device
and
so
far
so
the
ipro
rnd
members
confirm
their
behavior
with
the
latest
reload.
B
However,
so
there
is
some
problem,
therefore
so
hansen,
please,
I
explain
the
problem
about
introducing
the
divas
broker.
Is
it
talking.
C
Oh
yes!
So
if
I
remember
that
from
the
last
tv
course
that
we
actually,
they
talked
about
the.
D
C
Broker
but
I'm
gonna
check
the
new
gitlab,
the
the
gita
and
the,
and
I
saw
that
the
bus
demons
so
so
I
I
quite
confused
because
is
my
point
of
view
that
first
brokers
and
the
first
demon
is:
is
it
different.
A
Yeah,
I
I'm
so
let
me
step
in.
I
made
a
a
proof
of
concept
implementation,
so
I
have
a
look
at
dbos
broker
and
the
problem
with
using
the
boss
broker
is
that
you
there's
the
dbos
broker.
Is
the
basic
debus
demon?
Yes,
but
it
it?
Doesn't?
A
It
can't
start
a
socket
itself,
so
it
relies
on
a
launcher
to
create
a
socket
and
set
up
other
some
basic
housekeeping
things,
and
that's
done
in
the
debuss
broker
case.
It's
done
by
a
dbus
launcher,
which
also
does
a
lot
of
other
things
like
integration
with
systemd
and
things
that
we
don't
want.
A
So
I
I
started
coding
a
minimal
launcher,
but
then
I
thought
maybe
I
can
just
use
the
old
debuss
daemon,
because
that
that
one
doesn't
have
integration
with
systemd,
and
so
so
I.
D
A
Just
took
the
old,
the
best
demon
and
made
a
container
that
starts,
he
was
demon,
and
so
I
I
made
a
small
proof
of
concept
and
it
was
was
able
to
run
the
bus
daemon
and
this
on
a
system
that
didn't
have
d
bus
and
then
lauren
could
attach
to
the
socket
and
then
a
client
could
talk
to
laura
over
this
socket.
So
the
proof
of
concept
was
successful.
However,
I
didn't
use
divas
broker.
I
used
the
old
implementation,
debus
team.
C
Yes,
but
after
this
meeting
so
we
will
go
to
the
conclusion
which
one
we
gonna
use
deeper
broker
or
the
first
demon.
A
A
B
B
So
I
confirmed
this
item
with
hyper
internal
members
and,
as
a
result,
the
get
set
resolution
in
rotation,
resize
and
movie
jpeg
encore
yb
to
rgb,
convert
three
image
capture
and
convert
from
browser
to
pic
packed,
pixel
and
multiple
input
supported
in
inference
interface,
so
maybe
their
general
requirements.
B
B
B
A
Objectives
yet
so,
but
if
if
this
is
important,
I
think
each
company
probably
should
list
what
apis,
they
believe
are
fundamental
for
the
sdk.
A
So
and
audio
could
be
one,
and
I
think
also
something
like
a
overlay
api
is:
let's
say
you
have
an
application
that
generates
metadata
like
bounding
boxes
and
things.
Maybe
you
want
to
create
an
overlay
on
the
video
with
that.
So
a
standardized
overlay,
I
think,
is
something
that
could
be
really
useful.
B
Okay,
after
now,
so
let's
build
the
consensus
about
the
apis,
okay
and
so
finally,
so
how
different
are
the
memory
consumption
about
the
container,
runtime,
container
d,
cryo
and
nancy
and
siran?
C
So
we
create
a
comparisons
between
the
conductivity
and
the
cryos
in
a
in
the
case
rest
cluster
and
in
here
on
on
the
continuity,
and
we
will
want
to
be
fair
with
when
we
do
the
comparisons.
C
So
we
create
a
example
that
we
run
the
container
d
and
tribal
both
using
the
currencies
continue
run
temp
and
the
next
is
that
we
want
to
compare
the
the
resource
consumption
between
the
one
c
and
c
ones,
but
we
cannot
config
to
make
the
container
be
working
with
u1
yet
because-
and
I
think
the
continuity
they
require-
a
intermediate
layers-
the
container
d
stream
to
communicate
with
the
low-level
container
110
like
a
c101c.
C
So
so,
but
basically
to
look
at
this
table
that
you
can
see
that
a
system
using
the
continuity
with
the
run
sys
compared
with
the
cryos.
We
don't
want
to
see
that
it's
consume
more
results
than
is
about
the
41
megabytes.
C
And
with
the
car,
the
difference
with
the
system
using
the
the
cryos
with
the
low
level
control
runtime
run
cmc1
without
a
system
using
the
run
c,
which
we
took
more
results
than
the
c1
system
and
it's
around
the
12
megabytes.
So
in
total,
a
system
with
the
container
d
and
ram
sees
compared
with
the
system
using
the
cryo
and
silent.
It
will
took
the
overall
more
than
50
megabytes.
C
So
here
is
our
first
comparisons
between
the
continuity
and
cryo.
So
so
we
can
have
the
first
overviews
the
difference
in
the
resource,
consumption
between
the
content,
d
and
files
and-
and
I
think
we
can
have
the
better
choice
when
we
can
try
to
use
the
cryo
instead
of
continuity,
to
reduce
the
resource.
Consumptions.
C
A
A
Yeah,
so
I
think
that
should
be
the
question
which,
which
I
tried,
so
I
I
think
the
formal
decision
we
should
take
is
which
ecosystem
we
are
targeting
and
from
our
at
least
from
access
side.
A
We
are
not
so
much
interested
in
running
kubernetes
on
the
edge
we
are,
so
we
want
to
be
able
to
use
the
api
directly.
A
B
Also
alfredo
excel,
so
our
purpose
is
to
use
the
oci
image
in
common
between
cloud
site
and
the
edge
and
so
between
each
companies.
B
Therefore,
so
even
if
so,
our
api
is
different,
so
maybe
the
common
oci
image
and
can
be
shared,
therefore,
so
so
far
so
we
are
discussing
so
to
unify
the
api.
B
The
several
api
should
be
introduced
to
the
animatrix.
How
do
you
feel.
A
No,
I
don't
think
I
think
we
should
target
one
apis
yeah
I
do
otherwise
the
choices
are
two,
so
there
should
be
one
api
and
that
could
be
implemented
by
different
binaries.
So,
for
example,
container
d
also
supports
cri.
If
you
have
a
plug-in,
so
the
question
is:
if
so,
the
question
is
which
api
we
are.
A
A
I
don't
think
it's
good
to.
It
doesn't
promote
interoperability
if,
if
we
have
many
apis.
E
Because
cri
only
is
really
targeting
only
kubernetes
so
but
if
we
choose
c
cri,
we
are
limiting
to
kubernetes
orchestration
container
d
supports
both
kubernetes
and
the
docker
orchestration,
so
it
depends
on
which
communities
are
we
targeting
with
unimetrics.
E
A
E
E
D
D
E
Opinion
now
it's
hard
to
make
a
decision
between
cri
and
container
d
without
having
a
good
feeling
about
which,
which
orchestration
methods
we
want
to
support
with
unimatrix
and
axis
today.
We
feel
that
kubernetes
might
be
limiting
just
support
at
this
stage,
since
we
could
have
a
lot
of
opportunities
with
partners
that
also
by
using
other
orchestration
tools.
A
E
But
kubernetes
is
a
long
term,
as
I
see
it
and
in
short
or
term
maybe
you
need
support
travels
as.
E
A
E
A
E
E
A
Docker
is
quite
well
accepted
ecosystem
for
managing
containers,
so
that's
even
if
you're
not
using
the
docker
daemon
itself
on
the
edge
you're
using
the
api
like
that
is
container
d8.
E
Yes,
for
inter
mobility
purposes,
container
b,
of
course,
is
a
better
choice
for
the
reference
implementation,
but
for
a
specific
company
implementation.
E
A
Maybe
this
is
even
something
for
the
steering
committee
to
discuss
with
it's
about
yeah,
it's
more
of
a
business
decision.
Maybe.
A
E
Yeah,
maybe
we
should
add
something
about
which
kind
of
partners
we
would
attract
and
which
ecosystems
we
are
using.
A
Yeah
for
our
partners,
I
would
say
using
docker
even
internally,
is
even
if
we
have
looked
a
lot
at
kubernetes.
F
E
A
E
Yeah,
it
depends
on
if
we
are
talking
about
being
interoperable
with
other
ecosystems
or
we
are
trying
to
build
an
ecosystem.
So
that's
a
different
way
to
to
see
it.
Do
we
need
to
support
the
orchestration
mechanism
that
our
partners
are
using
or
we
or
could
we
mandate
that
you
need
to
use
this,
I'm
not
sure
which
way
to
go.
E
If
you
go
to
someone
who
would
like
to
incorporate
cameras
into
existing
big
systems,
they
might
already
have
chosen
between
docker
and
kubernetes,
and
they
want
to
integrate
our
products
or
any
of
our
member
companies
products
in
the
system.
Then
you
can't
mandate
orchestration
in
that
they
don't
currently
use.
A
E
A
E
But
there
are
different
use
cases,
of
course,
so
when
using
ansible
and
container
d
we
are
saying
we
don't
need
to
do
load,
balancing
and
other
stuff
that
kubernetes
is
doing.
So
we
from
our
point
of
view,
ansible
and
container
d,
would
solve
the
issues
we
have
at
this
point
in
time
to
distribute
the
applications.
E
B
Container
d,
it
can
be
supported,
you
know,
cri
and
you
see
a
cryo
can
be
supported,
cri
and
therefore
so,
and
the
deployment
api
should
be
cri
and
so
unification.
B
B
So
especially
so
I
pro
strongly.
B
For
sure
so
hyper
is
going
to
investigate
by
using
the
k3s.
E
Then
it's
clearly
if
you
already
have
selected
the
kubernetes
diaper
and
we
are
using,
we
are
targeting
partners
as
well
because
doing
docker.
We
need
to
support
both,
but
that's
a
clear
result.
I
think,
but
that
also
means
that
the
reference
implementation
at
unimatrix
would
have
to
be
container
d,
but
you
can
swap
container
b
to
cri
when
you
make
your
product
images,
but
then
we
don't
have
a
common
container
management.
E
Really
a
standard
or
anything
it's
it
depends
he
always
doing
interoperability
or
with
standardizing
an
ecosystem.
That's
a
governance
question.
So
what
about
a.
A
B
So,
as
I
mentioned,
so,
the
deployment
to
api
should
be
cri,
and
so
so
and
in
addition,
so
the
difference
about
the
difference.
So
we
prepare,
we
will
prepare
both
of
them.
A
B
So
basically
so
so
it's
orchestration
point
of
view,
so
the
kubernetes
is
expounded
widely.
In
addition,
so
if
we
introduced
the
kubernetes
so.
B
A
B
So
therefore,
so
in
this
meeting,
so
we
expect
you
to
explain
the
a
way
of
creating
the
security
by
using
the
ssh.
A
Yeah
we
have
a
presentation.
So,
okay,
I
don't
know
if
we
we
can.
Maybe
we
can
do
that.
One
now
then
scream.
Are
you
ready.
B
If
it's
reasonable
so
abroad,
members
may,
I
might
change
their
thinking.
G
G
Yes,
yes,
okay
yeah!
This
is
this
is
a
kind
of
high
level
overview
of
how
we
want
to
use
ansible
to
do
some
kind
of
management
of
of
of
a
server.
I
have
mentioned
server
as
a
generic
term.
It
can
be
as
a
camera
too,
but
the
concept
more
or
less
I'm
presenting
here
is
is
quite
as
an
overview.
G
So
typically
the
ansible
is
installed
on
something
like
a
workstation.
G
G
We
made
this
presentation
as
a
generic
one
and
the
example
is
taken
from
how
we
say
access
some
kind
of
a
virtual
machine
from
either
an
issue
or
or
an
aws,
more
or
less.
G
The
the
process
remains
same
so
in
in
any
of
the
environments.
You
you
create
a
key
pair
using
ssh
key
keygen,
and
so
you
could
you
create
a
creepkeeper
on
your
workstation
and
you
copy
the
workstation
public
key
on
to
on
to
the
servers
that
you
want
to
access
and
yeah.
The
copy
here
can
be
in
can
always
be
debated,
how
you
are
actually
copying
it.
G
G
The
the
key
has
to
be
on
the
server
the
public
key
on
the
server,
and
then
we
use
the
ssh
to
login
to
the
server
with
the
private
key
that
we
have,
which
is
definitely
private,
and
it
has
to
remain
on
the
workstation
and
and
protecting
that
private
key
on
the
workstation
definitely
is
a
responsibility
of
whoever
is
trying
to
manage
the
server
and
the
example
command
that
I
have
given
how
ansible
tries
to
do
is
ansible
with
minus
key
file
as
a
private
key
that
that
is
available
on
the
workstation
with
minus
I
option
as
the
servers
list.
G
So
maybe
the
I
is
a
is
a
short
form
for
inventory
in
this.
In
this
case,
the
servers
list
of
server
one
two
three
can
be
created
like
a
list
of
a
list
of
cameras
or
list
of
servers
that
you
want
to
maintain
and
using
the
same
private
key,
because
we
have
copied
our
public
key
on
to
these
servers.
G
You
can
use
ansible
with
with
the
key
file
as
a
private
key
and
use
the
module
called
ping,
the
ping
module.
Here,
though,
the
name
looks
like
the
standard
icmp,
the
ping
module
of
ansible
does
a
bit
more
than
that
it
tries
to
establish
the
ssh
connection
and
and
kind
of
logs
you
logs
you
into
the
into
the
machine.
G
So
that's
how
the
ssh
communication
happens
and
just
to
emphasize
that,
from
from
from
different
examples
that
we
have
seen
most
of
the
most
of
the
say,
admins
who
tries
to
do
the
ansible
way
of
managing
servers
on
production
servers?
Typically,
once
you
do
the
bootstrapping,
you
create
a
user,
something
like
an
ansible
user,
with
the
possibility
to
kind
of
escalate
it
privileges
to
do
some,
some
management
level
tasks.
G
Otherwise
it
always
has
some
no
basic
privileges
or
basic
permissions
of
what
what
it
can
do
as
an
as
a
user.
G
So
in
in
such
cases,
they
the
root
can
be
disabled,
and
the
communication
to
the
server
can
happen
only
with
a
specific
user
which
we
have
created.
So
so
that's
that's.
How
this
ansible
uses
ssh
communication
to
the
servers
and
it
have
it
can
be
done
through
the
root
or
it
can
be
done
through
a
specific
user
whose,
whose
private
key
are
we
using
to
access
or
reach
that
particular
server?
G
And-
and
once
you
have
access
to
this
you
can
you
can
do
pretty
much
all
the
standard
management
commands
that
you
can
that
you
can
do
even
even
copying
this
into
the
authorized
keys.
There
is
a
module
called
authorized
keys
in
ansible
module,
so
even
even
this
also
can
be
done
through
ansible.
G
Yeah
and
and
as
as
we
have
discussed
before,
ansible
has
a
very
wide
array
of
modules
for
different
use
cases,
depending
on
how
we
want
to
bootstrap
or
set
up
our
cameras
or
servers.
G
Probably
we
have
to
make
a
detailed,
step-by-step
process
of
how
we
want
to
set
this
up
as
bootstrap
and
then
write
playbooks
and
use
appropriate
ansible
modules
in
achieving
that
particular
use
case.
B
That's
all
also,
we
will
try
to
know
the
way
of
saving
the
security
after
connecting
the
camera,
so
in
this
case,
so
if
the
integrator
is
a
good
guy,
so
it's
no
problem,
however,
so
if
integrator
is
bad
guy,
so
maybe
the
user
can
enter
the
camera
and
so,
for
example,
so
our
expectation
is,
for
example,
so
to
disable
the
our
yeah
I'll.
Do
the
user
yeah
like
that.
B
Who
you
would
like
to
know
so
so
I
know
actually
allow
the
end
users
to
access
the
to
the
camera
by
vr
ssh
and
therefore
so
after
the
user
access
to
the
camera.
So
how
does
arches
save
the
security
in
the
camera?
B
So
I
pro
prohibit
the
end
user.
B
To
access
by
using
the
telemet,
the
reason
is,
the
security
after
entering
the
camera
can
be
saved.
Therefore,
so
apple's
company
security
policy
prohibits
to
access
the
camera
vrssh
there
also.
I
would
like
to
know
the
way
of
making
the
security
in
the
camera
by
axis.
G
I
think
I
think
the
ssh
is
not
enabled
by
default
in
in
any
of
our
cameras,
but
frederic
correct
me.
If,
if,
if
I'm
wrong,
we
each,
we
should
be
in
some
kind
of
developer
mode
to
be
able
to
do
an
ssh
into
the
camera
on
on
the
field
or
it
is
left
to
the
left
to
the
users
once
we
kind
of
give
the
camera
to
them.
A
I
think
it
is
possible
to
enable
it
even
on,
but
the
root
user
is
not
really
root.
It's
just
a
name.
That's
the
intention
anyway.
So.
A
G
F
F
If
you,
if
we
don't
install
python
in
the
camera,
assembler
only
can
use
the
rain
module.
G
I'm
sorry,
are
you
trying
to
install
ansible
on
on
the
camera?
We
are
not
talking
about
installing
ansible
on
the
camera.
F
I
think
I
mean
assembler
has
a
two
kind
of
node:
a
control
node
and
managing
node
control.
Node
control
node
is
our
workstation
or
our
service
and
the
manager
node
is
our
camera
right.
F
F
So
we
we
need
to
so
we
need
to
install
a
sample
in
control
node,
but
in
managing
node.
We
also
need
install
person,
because
if
you
don't
install
person,
the
assemble
canal,
the
assembly
that
only
can
use
can
use
a
royal
order.
This
is
this
information.
F
G
I
mean
ideally,
I'm
not
sure
if
I,
if
I'm
in
sync
with
your
response,
it's
always
ansible
is
installed
on
the
workstations
and
the
only
connection
to
those
servers
can
be
through
like
a
local
connection
or
or
over
an
ssh
connection,
and
you
are
not.
I
don't
think
python
is
needed
on
those
on
those
any
of
those
nodes.
F
G
Http
connections
and
other
things
with
ansible
or
cameras,
and
we
are-
we
were
able
to
do
that.
So
I'm
I'm
surprised
that
you're
saying
cameras
also
need
python
for
ansible
to
work
on
the
on
the
on
the
cameras.
Definitely
on
workstation,
where
you're
installing
ansible.
That
is
needed,
but
I'm
surprised
if
you
say
it
is
needed
on
the
servers.
G
I
think
that
that's
the
whole
idea
of
using
ansible
on
workstations.
We
are
not
we
we
and,
and
that's
how
we
have.
We
have
kind
of
not
recommended
to
use
ansible
on
the
cameras,
because
when
you
install
ansible,
it
tries
to
bring
a
lot
of
baggage
and
in
with
that,
python
also
comes
in
as
more
or
less
the
the
base
layer
and
which
we,
which
none
of
us
would
would
really
be
wanting
to
do.
F
F
A
A
D
But
I'm
connecting
by
using
my
phone,
maybe
I
just
list
five
file
apis.
The
first
one
is
the
video
is
a
video.
I
o
model.
It's
review
capture
apis,
maybe
the
we
should
support
the
video
capture
class
and
several
public
member
functions
like
the
open
release
is
opened.
D
Grab
like
this
kind
of
functions,
but
not
all
of
the
opencv
definition
apis.
We
should
support.
We
just
need
to
open
a
radio
file.
D
Video
capture
properties
maybe
use
this
apis
to
set
the
height
of
the
frame
and
the
radius
of
the
frames,
and
I
think
it's
optional
and
the
last
two
of
them
is
image
right,
use
this
api
to
save
an
image
to
maybe
a
gpg
file
and
the
last
one
is
resize.
D
This
function,
resize
the
image
source
down
to
the
specified
size.
D
This
these
apis,
I
think
we
should
support
at
first.
A
Okay,
yeah,
so
all
those
functions,
basically
in
the
video
io
library
and
the
core
right.
What
about
additional
additional
libraries
stuff
like
object,
detection
or
edge
detections
or
other
algorithms
and.
D
D
A
For
processing
of
images
or
like
so
that's
my
question.
D
From
from
my
point
of
view,
we
only
support
our
own
load
back
inference
back
end.
Yes,
will
not
support
the
opencv
standard
image
process.
A
Apis,
okay,
thank
you
for
your
input,
so
we're
a
little
bit
over
time
now.
So
maybe
we
can
continue
the
discussion.
I
think
next
time
regarding
opencv
I'll
bring
up
a
docker
file
which
builds
opencv.
We
can
more
have
a
look
at
specific
which
which
build
flex
to
enable,
because
this
enables
different
features
that
ultimately
goes
into
the
library
and,
of
course,
the
more
you
enable
the
bigger
it
gets,
but
so
I'll
prepare
some
presentation
for
the
next
talk
about.
B
So
in,
in
addition,
so
please
add
the
agenda
about
the
security
after
entering
the
camera.
A
B
Also,
please
at
the
agenda
so
about
the
security.
B
Oh,
no,
no!
No!
No!
There!
Silly
ramsan's
presentation
is
just
a
connection
about
a
ssh
right,
so
we
would
like
to
know
the
a
way
of
making
the
security
in
the
camera,
so
the
internal
camera
part
is
revealed
to
the
end
users.
B
Therefore,
so
how
does
access
save
the
security
in
the
camera,
for
example,
redress
the
directory
limitation,
like
that.
A
A
B
18Th
cb
call
is
planned.
E
The
same
day,
the
cc
call,
but
that's
we
do
have
at
the
same
time,
let's
see.
D
A
Small
vacation,
so
can
we
say
either
the
17th
or
the
week
after
the
25th.
B
So
I
I
will
not
receive
the
invitation
from
json
yet,
therefore,
so
if
the
time
period
is
different
between
so
tv
caller
and
civico,
so
it's
okay.
Every
time
it's
different,
but.
E
Freddie,
your
sebastian
was
to
have
it
on
the
17th.
Instead,
either.
E
If
I
could
make
a
reservation
that
day.