►
From YouTube: Unimatrix Technical Board call 2021-06-14
B
A
So
did
we
agree
on
where
to
put
the
recordings
for
this?
Should
we
put
them
on
youtube.
C
D
I
think
so
it's
better
the
url
is
describing
meeting
minutes.
How
do
you
feel.
A
D
Yeah
you,
you
will
upload
the
recorded
video
to
the
youtube,
and
so,
however,
so
it's
easy
to
access
for
for
us.
D
If
you
describe
the
you
are
there
in
the
meeting
minutes
on
the
github,
I
keep
git
drop.
A
Okay,
dora
is
joined,
so
I
think
we
can
start
so.
This
meeting
is
to
mainly
to
find
the
consensus
around
what
we
want
to
do
in
uni
matrix
and
what
different
companies
are
interested
in.
So
first
point
is
just
to
get
the
consensus
about
which
components
each
company
think
is
important
and
what
should
be
included
in
unimatrix.
A
So
I
divided
it
up
a
bit
about
the
host
requirements
and
the
the
device
apis
and
which
programming
languages
also,
if
you
believe
that
we
need
an
orchestration
or
app
management
component,
so
maybe
I
can
start
from
axis
side
so,
regarding
the
host
requirements
so
far,
we
have
only
required
some
kernel
configurations
to
enable
to
run
containers,
so
those.
A
Requirements
should
be
maybe
more
specifically
like
they
do
for
docker.
They
have
a
script
that
checks.
If
you
are,
your
device
is
capable
of
running,
so
we
should
have
a
similar
thing
for
unimatrix
that
checks.
If
your
device
has
enabled
all
all
the
required
there
will
be.
I
mean
some
required
and
some
optional
configuration.
A
And
then,
regarding
the
programming
languages,
we
have
talked
about
c
plus
and
python,
and
that's
for
our
for
our
uses.
A
That's
enough
right
now
and
then
about
orchestration
or
app
management.
We
want
to
focus
on
ansible
because
that's
actu,
that's
what
we
are
looking
at
at
axis
right
now
and
we
think
ansible
is
it's.
It
solves
the
cases
that
we
want.
We
can
go
into
the
specific
use
cases
in
next
point
in
point
two:
what
we
to
solve
with
the
orchestration
or
maybe
we
should-
I
don't
know
if
we
should
say
orchestration,
because
that's
such
a
huge
term-
so
maybe
it's
better
to
say
app
management
here
or
container
management.
A
D
So,
according
to
hyper
members,
so,
as
I
mentioned
before
so
so,
ansible
introduction
is
a
very
huge
change
from
the
original
plan,
and
so
we
will
try
to
realize
the
container
orchestration.
D
So
is
it
possible
to
be
the
continental
orchestration
in
the.
D
D
Structured
so
maybe
when
ansible
is
introduced,
so
the
integrators
communicate
with
each
device
h
right,
yes,
for
example,
or.
D
So
so,
for
example,
so
now
hyper
hyper
is
investigating
the
cryo
cri
or
io
is
instead
of
control
d.
So
is
it
okay?
Is
it.
A
D
So
how
why
so
the
reason
why
ipro
is
investigating
the
cryo
is
the
container
d
is
the
insole
sizes
large
larger
zone
cryo?
Therefore,
so
we
are
investigating
cryo.
D
A
I
don't
know
if
cryo
is
implementing
the
container
d
grpc
api.
I
don't
think
so.
D
That's
all
so
so
as
well,
so
lancy
is
larger
than
the
shirang.
Therefore,
so
hyper
is
also
investigating
syrian
introduction.
D
So
so
maybe
so
I've
remembers
mentioned
in
the
past
meeting,
so
our
system
doesn't
have
the
debuff.
D
Therefore,
so
if
there
is
no
debuff
so
maybe.
D
Some
solution
is
needed
to
introduce
the
container
orchestration
data
by
using.
D
Rollout
did
you
remember
the
issue
about
the
debuffs
according
to
the
meeting
minutes,
so
several
companies
doesn't
have
the
dbs
on
the
project.
A
So
I
mean
it's
possible
to
to
replace
divas
with
with
a
simple
socket
communication,
but
it's
of
course,
some
work
involved.
A
Well,
it
is
like
that
right
now,
but
of
course
it
can
be,
it
can
be
changed.
A
E
There
it's
important
to
add-
I
think
in
this
to
this
discussion,
that
there
are
two
different
or
two
sides
of
this
d
bus
dependency.
There
there's
one
dependency
on
this
on
the
dbus
system,
bus
itself
for
session
initialization
to
to
create
the
connection
between
client
application
and
the
lyric
service.
E
But
then
there
is
a
different
side
of
that
of
the
d
bus
dependency
in
terms
of
a
dependency
to
the
sd
bus
library,
for
point-to-point,
communication
for
for
serializing
and
deserializing
data
on
the
point-to-point
socket
communication
between
the
client
and
service.
E
So
there
are
two
different
flavors
of
that
dependency
and
so
that
I
think
that's
a
an
important
aspect
to
that
discussion.
D
Yeah
for
sure,
so,
according
to
eyebrows
investigation,
so
it's
a
little
difficult
to
rewrite
the
serialize
and
the
serial
rights
without
debuss.
E
Yeah,
I
think
that
that
is
a
bit
harder
to
to
work
around.
E
E
F
E
E
Yeah,
it's
yeah
it
in
general,
it
it
is
definitely
a
an
alternative,
but
the
the
hard
part
here
is
that
and
that
we
we
trans
over
the
sock,
the
unique
socket
link.
E
We
transfer,
file,
descriptors
and
dbas-
supports
that
very
nicely,
but
there
is
no
concept
of
of
file
descriptors
in
protobuf,
yeah,
you're
right,
so
that
that
must
be
in
case
such
a
mechanism
is
used.
E
I
mean
there
is
one
alternative.
If
I
have
thought
a
lot
about
this
and
the
I
think
the
the
way
if
we
would
go
for
a
protobuf
replacement.
E
In
that
case,
we
actually
need
to
implement
a
protobuf
based
rpc
in
the
same
way
as
grpc
does
so
there
is,
I
mean
there
is
a
kind
of
a
plug-in
interface,
the
protobuf,
where
you
implement
your
own
transport
protocol
or
whatever
we
should
call
it
the
exactly
the
same
way
as
sgrpc
does.
I
think
that
is
the
correct
way
to
do
it,
but
it's
a
substantial
amount
of
work.
A
A
On
on
that
on
larad
itself,
while
the
other
one
is
a
dependency
on
your
whole
system,
so
I
think
your
biggest
problem
is
probably
that
you
need
to
run
the
bus
demon
right.
E
G
So
it
does
appear
that
dbus
can
be
run
inside
a
docker
container.
I
haven't
actually
tried
to
do
this
myself,
but
I
mean
that's
an
easy
way
to
drop
it
in.
E
Yeah
we
do
it,
we
we
do
it
often
like
that
we
access
we
access
the
host
system,
bus
from
a
container.
E
I
think
we
have
actually,
I
think
we
have
done
that
too.
I
think
we
have
run
la
rod
with
with
a
d
bus
broker
inside
a
container
actually.
E
It's
the
the
only
thing
you
need
to
be
able
to
to
access
is
the
socket
the
socket
files
that
is
under
slash
var.
E
Yeah,
I
think
I
mean
it's,
it
could
be,
it
could
be
used,
but
their
protobuf
itself
does
not
have
any
concept
for
file
descriptors.
It
does
not
care
about
file
scriptures,
so
to
speak.
E
D
A
sword
happy
so
ipro
member
is
about
to
investigate
the
flat
buffer
after
now,
and
so
is
it
no
mean.
E
Either
I
I
think,
what,
in
order
to
be
able
to
transfer
file
descriptors
between
processes,
you
need
to
build
a
transport
mechanism
that
is
based
on
unix
domain,
sockets,
that
the
only
option
more
or
less.
E
E
Yeah
yeah
I
mean
it
could
be,
it
could
be
done,
but
it
it
it's.
It
can't
be
done
with
a
stock
grpc
and
I
I
actually
saw
I
read.
I
read
through
a
lot
of
a
lot
of
support
issues
related
to
this
matter
on
grpc,
and
it's
pretty
clear
that
that
the
guys
that
are
governing
that
the
the
grpc
project-
they
are
not
at
all
interested
in
adding
support
for
this
kind
of
features.
E
E
D
Standpoint
standpoint,
so
maybe
we
have
to
discuss
that
thing
internally.
E
It's
interesting
to
reflect
a
bit
on
what
what
they
have
been
doing
inside
of
android,
for
example,
and
what
they
are
using,
because
I
I
think
that
I
think
that
that
k,
the
problem
that
they
are
so
that
that
is
solved
in
android.
E
For
example,
that's
that's
very
similar
to
to
the
problem
that
we
try
to
solve
here.
We
we
want
isolated
applications
being
able
to
call
services
that
runs
on
the
host
operating
system
and
they
have
a.
They
have
a
mechanism
that
that
is
called
a
binder
to
do
that
and
the
exactly
this
feature
of
file.
Scripted
transfer
is
an
important
important
feature
of
binder.
A
Should
we
should
we
investigate
running
devos
in
a
container?
I
think
that's
could
be
a
way
forward.
A
H
I
tried
to
find
another
rpc
library
we
now
we
use
the
anti
open,
open
source,
ipc
library
if
we
have,
if
you
guys
want
to
learn
about
that,
rpc
depend
ipc.
H
H
No,
it's
similar
to
the
dfc,
but
it's
oh
it's
prior
by
the
several
methods
to
communication
mounted
to
multi
and
multi-tool.
Venture
multi
venture
point
this
better.
I
have
to
time
I
have
time
I
need
more
time
to
learn
if
it
supports
the
file
descript.
E
Yeah,
I
just
I
would,
I
think,
from
from
my
point
of
view,
I
think
grpc
is
fantastic
for
for
network
based
tcp
ip
based
rpc
and
protobuf
is
a
really
nice
idea,
I'd
of
language
for
this
for
building
an
idl.
E
It's
saying
it's,
I
think
from
my
point
of
view:
it's
it
solves.
We
have
two
different
problems
here.
It
could
be,
of
course,
be
extremely
relevant
in
in
the
context
of
unimatrix
to
do
machine
to
machine
rpc,
and
for
that
I
think
grpc
is
fantastic,
but
there
is
also
what
what
we
primarily
is
discussing
here
when
when
we
are
talking
about
ip
cameras
we
are,
we
are
talking
about
device
internal
apis,
and
for
that
I
don't
think
grpc
is
the
best.
E
Is
the
best
solution
to
call
into
the
host
into
services
running
on
the
hosts
operating
system.
C
E
Yeah
it
could,
it
can
certainly
be.
It
can
certainly
be
used
in
some
cases,
but
it's
in
other
cases
like
for
for
data
intensive
applications
where
you
need
to
transfer
large
amounts
of
data
like
video
data
image,
data
in
and
out
of
the
container
between
the
containerized
application
and
the
service
running
outside
of
the
container.
I
think
it's
not
very
well
suited
trying
to
trying
to
solve
that
kind
of
problem
with
grpc
is
to
me
a
bit
like
trying
to
screw
with
a
hammer.
G
John,
can
I
probe
a
little
bit
on
the
need
to
actually
move
file
descriptors
around,
as
opposed
to,
for
example,
pointers
to
shared
memory.
E
Yeah
it's
one.
One
reason
is
that
I
I
mean
the
the
the
main
reason
to
that
is
is
performance.
E
There
are
other
reasons
as
well
as
it's
a
bit
easier
to
handle
to
handle
separation
between
different
clients,
for
example
from
security
perspective
by
not
by
not
exposing
the
files
in
the
file
system
that,
but
the
main,
the
main
reason
is,
is
performance
and
we
use
we
make
use
in
layer
of
of
so-called
dma
buffs,
which
is
a
linux
kernel
concept
for
sharing
data
between,
for
example,
one
driver.
E
It
could
be
some
kind
of
like,
let's
say,
an
isp,
an
image
signal
processor
that
could
provide
data
in
the
form
of
an
image
that
data
is
then
exported
to
a
userspace
application
as
as
a
file
descriptor
that
refers
to
the
underlying
dma
buffer
inside
of
the
kernel
that
could
subsequently
be
imported
later
on
to
to
a
different
driver,
for
example,
one
that
handles
deep
learning,
accelerate
that
handles
the
deep
learning
accelerator
and
that
can
in
by
importing
the
file
descriptor.
G
I
see
so.
I
would
also
recommend
that
folks
take
a
look
at
the
work
that
got
done
on
the
ross
ros
robotic
operating
system
project
about
two
years
ago
for
ross
three,
you
know
they.
They
were
dealing
with
point
clouds
as
a
an
imaging
source
and
it
was
becoming
really
problematic
for
them.
So
they
wrote
a
shared
memory,
transport
mechanism,
that
kind
of
serves
the
same
purpose
spiritually.
But
you
know
I
don't
know
if
it
actually
provides
the
the
security
guarantees
that
your
dma
approach
does.
G
But
it
is
it's
it's
given
them
the
same
kind
of
efficiencies
that
that
you're
looking
for
here
and
I
believe
that
their
implementation
is
pretty
portable.
E
Could
you
send
a,
could
you
send
the
link
or
some
as
a
pointer
in
the
chat
sure?
Let
me
just
track
it
down
here
great.
G
A
You,
okay,
should
we
move
on
from
the
dbus
discussion
to
I'd
like
to
hear
more
from
hikvish
and
what
your,
what
your
view
on
the
components
of
unimatrix
are
and
also
around
orchestration.
H
Okay
for
the
requirements
of
components,
I
think
there
is
another
one.
We
need
to
consider
it's
about
the
app
format
if
we
want
to
use
the
oci
image,
but
we
need
to
consider
how
to
add
some
app
description
into
it.
For
example,
the
app
request
resources
requirement
the
ram
requirements
like
this
kind
of
information
and
yeah.
This
is
a
another
requirement
of
the
components
of
unimatrix
and
the
orchestration
use
cases
I
think
there
is
another
use.
Cases
is
to
active.
The
app
is
download
the
license
to
the
host
to
active
the
app.
H
A
A
Yeah
either
I
mean
either
there
is,
you
can
use
oci
and
you
can
write
your
own
things
in
oci
or
we
can
do
our
own
manifest
a
separate
manifest
that
describes
the
application
and
it's
and
the
things
like.
G
A
There's
two
two
ways
to
do
it.
I
guess,
but
maybe
the
best
one
is
to
use
our
own
manifest
and
rather
than
parasiting
on
oci.
H
Yes-
and
we
can
also
add
lab
to
the
oci
image,
but
you
know
we,
you
mean
we
don't
need
to
standardize
this
kind
of
information
in
the
uni
matrix.
A
H
A
So
regarding
app
management
and
so
on,
what
from
hikvision
side?
What
tools
are
you
looking
at
for
doing
things
like
that,
like
this
kind
of
use,
cases
of
deploying
mass
deploying
or
mass
configuration
and
the
distribution
of
keys
and
certificates
and
things
from
access
side,
we
we
are
looking
at
ansible
is
that
does
that
seem
suitable
for
your
cases.
H
Because
of
the
security
consideration,
maybe
some
cameras
will
not
open
the
ssh
port
by
default.
That's
my
main
concern
and
yeah.
H
D
The
end
user
can
access
to
the
internal
part
of
the
camera.
Therefore,
so
we
have
to
prepare
the
method,
avoiding
the
so
unknown
attack.
D
Presentation
like
that,
however,
so
the
if
the
trojan
horse
is
installed
so
in
such
a
case,
so
the
logic
is
not
established
and
in
addition,
so
if
the
integrator
is
a
bad
guy,
so
maybe
the
manager's
program
may
be
executed
in
the
camera,
so
frederickson.
So
how
do
you
keep
the
security
in
the
internal
camera.
A
Well,
you
don't
have
to
have
the
root.
Account
is
not
needed
for
the
ssh.
I
guess
I
don't
know
sriram,
you
look
more
advanceable,
but
it
doesn't
really.
A
F
G
F
Is
secure,
I
think
ssh
will
use
that
to
say
log
into
the
virtual
machine.
Probably
in
this
case
log
into
the
camera
and
ansible
also
does
does
the
same
so
ansible.
F
A
A
F
D
Oh
so
so
I
would
like
to
confirm
what
you
mentioned,
so
that
does
it
means
so
the
user
id
should
be
implemented
into
the
firmware
of
the
camera
right
user
id.
A
D
I
I'm
sorry
but
so
c
ramsan's
boss
is
very
small
and
so
yeah,
okay,
okay,
I
can
see
I
can
hear
you.
F
Okay,
can
you
hear
me
now
yeah,
I
can
hear
you.
F
Yeah,
I
think,
if,
if
you
are
using
a
ssh
with
with
some
kind
of
a
minus,
I
kind
of
options
and
you
give
the
pm
file
it
is
it
is,
it
is
being
it.
It
is
kind
of
with
a
trusted
user
rather
than
a
specific
user
id
credentials.
So
the
user
id
credential
like
username
and
password
is,
is
one
approach
and
using
the
the
pm
file
is
is
another
approach,
so
it
depends
on
how
you,
how
is
implemented
on
your
cameras,.
A
A
But
if
you
have
the
right
security
mechanisms
implemented,
it
it
says
should
be
secure
and
you
can
so
I
don't
think
just
by
using
ssh,
it's
necessarily
more
insecure.
Just
because
it's
a
well-known
port.
A
D
No,
for
example,
eyebrow
is
still
investigating
quantities
case.
Yes,
however,
so
we'll
try
to
introduce
up
armor
like
that,
so
I
don't
know
ansible
well,
however,
so
does
also
have
such
a
secure
function
to
keep.
A
D
A
Yeah,
so
that's
totally
orthogonal,
so
you
and
I
think
uni
matrix
should
not
specify.
Does
it's
up
to
each
vendor?
I
think
if
you
want
to
run
up
armor
or
any
other
external
framework
for
secured.
D
But
so
as
long
as
so
unimatrix
use
the
ssh
architecture,
so
maybe
we
should
provide
the
end
user.
The
reference
implementation
difference
implementation.
A
A
Just
as
long
I
mean
from
axis
side
we're
working
on
ansible
playbooks
from
our
end,
so.
A
That's
the
road.
We
will
probably
go
for
our
own
devices
and
if
you
have
alternative
for
app
management,
that
you
are
willing
to
contribute.
A
Okay,
we
have,
we
have
five
minutes
left
now,
so
is
there
anything
else,
you'd
like
to
add
about
either
the
components
or
the
the
use
cases
for
app
management?
D
Also
hypersync,
so
maybe
so
unimatrix
project
needs
to
attract
a
third
party,
because
the
party
is
a
content
vendor
and
so
therefore,
so.
D
D
So
unimatrix
project
should
provide,
for
example,
so.
D
Pre-Trained
models
and
samples
sample
programs,
and
so
that
the
third
party
can
develop
the
ai
application,
speedy
and
efficiency
efficiently
and
for
sure,
so
unimate
project
doesn't
have
enough
fund,
but
so,
for
example.
So
so
we
can.
We
can
have
the
website,
therefore,
so
the
oss
linkage
is
described
in
the
webpage,
for
example,.
D
So
if
we
attract
the
third
party,
they
may
provide
the
such
a
modules,
pre-streamed
model
and
sample
program
to
the
end
users.
D
Therefore,
that
is
so,
maybe
even
if
we
don't
have
enough
fund,
so
the
way
of
realizing
to
attract
a
third
party
is
existed.
So
what
do
you?
What
do
you
think
about
it?
So.
A
A
Okay,
I
think
our
time
is
up.
Let's
I,
I
think
we
will
have
a
regular
tv
calls
now
every
month,
so
I
will
call.
I
will
send
out
a
new
invitation
this
week
for
the
next
call
to
continue
discussions.
G
A
For
joining
today
and
nice,
to
hear
from
you
again
thank
you
very.