►
From YouTube: Mesos Containerization WG 06 15 2017
Description
Agenda/Notes:
Volume ownership/permission (Ilya)
[Jie] different types of volumes in Mesos
Sandbox
Host volume
Docker volume (external, dvdi based)
Persistent volume (local, or CSI based in the future)
[jpeach] We need to call out the semantics we want in the doc.
[jpeach] alternative: require that all containers that want to access the same volume are under the same user
[jie] setgid bits? Do we really need that?
[jpeach] should the ownership of the volume be allowed to be changed
[yanx] similar to jpeach’s comment, we can allow framework to specify the ownership of the persistent volume and the access control is up to the operator.
Action items:
Jie will schedule a meeting to discuss the semantics we want to support with relevant stakeholders
Status update
Planning spreadsheet
A
All
right
so
I
think
it's
being
recorded.
Alright,
hello,
everyone,
so
this
arm
make
those
containerized
containerization
working
group
today
is
June
15.
We
start
with
their
time
and
just
just
just
a
reminder,
the
meeting
is
recorded,
so
the
today's
topic
is
going
to
be
on.
Let
me
share
my
screen,
so
today's
topic
is
going
to
be
on
alia,
it's
going
to
show
us
in
a
leader
discussion
on
the
volume
group
and
ownership
cleanup
work.
So
let
me
show
this.
B
A
C
Okay,
can
you
see
my
screen?
Yes,
okay,
perfect
so
volume
ownership
and
mazes,
so
we
have
in
mazes.
We
have
persistent
volumes
and
there
is
one
problem
with
with
them
right
now,
when
the
agent
is
trying
to
mount
the
persistent
volume
that
is
or
is
not
already
owned
by
another
already
used,
then
it
sets
the
ownership
of
that
volume
to
match
its
of
the
sandbox
directory.
C
This
is
this
is
completely
fine,
but
even
the
volume
is
currently
in
use
by
other
containers
and
supposedly
those
other
container
containers
are
running
as
different
users.
Then
tasks
in
the
newly
launched
container
will
fail
because
they
will
fail
to
read
or
write
from
the
persistent
volume
because
of
incompatible
ownership
and
permissions.
C
C
If
they
are
running
aren't
running
as
a
non
root
user,
then
still
they
will
have
access
to
the
to
the
persistent
volume
that
they
are
sharing
because
they're
running
in
same
user
and
the
bully
is
owned
by
the
user.
The
other
case
is
when
all
containers
are
running
as
the
mix
of
various
users.
Those
can
be
route
or
non
route,
so
root
container
containers
that
are
running
as
much
as
root
users
are
fine
again
because
they
have
root
permissions,
permissions
and
containers
that
are
not
non
root.
A
A
Like
I,
mean
I,
think
emails
us
right
now,
like
we
have
a
lot
of
type
of
volumes,
one
is
called
persistent
longing
where
you
are,
you
can
actually
purchase
the
data
on
a
local
machine
and
and
some
of
the
wrong
like
host
volume,
which
is
the
ownership,
is
totally
out
of
band
because
he
likes.
For
example,
a
container
say:
hey
I
want
to
buy
amount,
a
host
falling
into
my
container,
and
the
permission
of
that
Vong
is
totally
up
to
the
operator
can
pick
the
host
for
a
me.
A
We
potentially
cannot
change
that
so
host
falling
one
thing
per
system
volume,
and
we
also
have
like
doctor
volume,
which
is
kind
of
external
loading
in
the
future
we
need
to
have.
We
will
have
like
a
persistent
volume,
backed
by
CSI,
like
kind
of
ears
volumes,
so
I
think
like
they're
like
multiple
problem.
Another
dimension
with
this
problem
is
darker.
Container
itself
might
have
a
different
me
a
little
only.
We
have
sorry.
So
let
me
let
me
try
to
be
sort
at
the
another
one.
A
So
on
so
right
now
the
solution
is
we
always
unclick
the
host
user,
ID
and
GID,
so
that
we
make
sure
that
the
processing
is
identical
container
can
access
to
that
sandbox.
But
this
is
like
not
ideal
like
I
wish.
I
could
have
a
better
solution
for
that,
but
I,
don't
know
what
evac
the
solution
is.
Yeah
just
give
more
context
on
this
problem,
so
we
have
multiple
problems
to
solve
here
and
then
the
goal
of
this
meeting
is
trying
to
brainstorm.
What
are
the
potential
solution
that
we
can
solve
those
issues
and
see?
C
A
C
Yeah
so
two
possible
solutions.
One
of
the
possible
solutions
is
using
the
old
good
old
UNIX
groups.
There
is
a
thing:
that's
called
a
supplementary
group.
It's
actually
a
list,
a
list
of
groups
that
can
be
assigned
to
to
the
process.
Besides
the
primary
the
process,
primary
group-
and
we
can
use
it
this
way
so.
C
So
we
can
extend
the
container
launching
for
with
a
list
of
subprime
simplement
supplementary
groups,
supplementary
gifts
that
will
be
passed
to
the
launcher
and
the
launcher
will
simply
set
all
those
leads
for
the
process
that
it's
running
inside
a
container.
In
addition
to
all
the
supplementary
gifts
that
it's
already
setting-
and
this
will
allow
the
process
in
the
container
to
access
the
volume
without
change,
without
changing
its
you
eat
or
get
so,
we
can
have
any
primer
you
and,
if
or
any
give
you
eat
and
primarily
I.
A
A
D
Ahead
anywhere,
yeah
I
should
yeah
in
in
the
problem
section
of
traffic
document,
I
think
it
might
be
worthwhile
and
specifying
the
similarity
that
we
want
Bobby
beam
columns
so,
for
example,
which
what
we'd,
like
semantics,
do
one
degree
on
different
users
to
be
able
to
write
files
diffusers
to
definitely
able
to
delete
each
other's
files
in
there
just
making
those
file
systems
tonight.
Ever
since
explicitly
one
puts
the
solutions
into
a
better
context.
A
Yeah
I
think
that's
a
good
point
like
we
should
call
out
like
what
exactly
the
semantics.
We
want
to
support
that
like,
for
example,
you
want
to
allow
those
four
containers
to
share
the
same
typical
pain,
cystic
ammonium
or
you
on
multiple,
inherently
to
be
able
to
come
by
with
the
file
systems
in
that
like
like.
What's
what
the
girl
eventually
achieve
for
different
type
of
warnings
that
I
mentioned
right.
B
A
I
think
so
again,
I
think
the
I
think
the
my
understanding
of
the
solution
here
is
all
the
files
created
by
the
container.
It
will
be
owned.
The
primary
group
and
primary
unity
is
going
to
be
your
leverage
as
a
primary
GI
DMV
ID
of
that
process,
but
at
the
same
time
we
also
make
sure
that
this
proposal
should
also
make
sure
that
at
a
sacramental
group
to
that
total
file
to
all
the
files
so
that
all
these
files
can
be
accessed
as
a
reply.
Well,
we
haven't
decided
yet.
A
So
that's
one
way
to
do
that.
I
can
recursively
change
the
ownership,
but
recursively
adds
its
additional
giv
to
all
the
files.
Another
approach
is
like
you:
don't
do
that
you,
you
only
do
the
top-level
directory
so
that
on
to
make
sure
that
at
least
that
the
other
container
has
access
to
that
volume
or
not
necessary
and
the
files.
D
D
D
Sorry
fit
again,
so
I
don't
get
that
part.
So
you
have
listen,
I.
Think
in
this
document.
That
is
the
implicit
assumption
that
tasks
that
are
accessing
the
volume
are
running
as
different
users.
Other
different,
you
are
dates
mm-hmm,
but
that
it
just
seems
to
be
just
an
assumption
that
are
there's
no
requirement
for
that
to
control.
A
F
A
Problem:
okay,
oh
I,
see
so
the
yeah
okay
I
see
got
it
so
you're
saying
like
if
you
required
that
it
have
that
requirement
that
if
all
the
containers
that
want
to
actually
the
same
person
among
it
has
to
be
under
the
same
user
day,
you
don't
have
to
solve
that
issues
because
it
works
right.
Now
that
what
you're
saying
that's,
what
I'm
saying:
okay,
yeah,
that's
one
potential
solution,
yep
yep,
yeah
I
think
like
we
should
like
call
out
all
the
options.
This
is
just
one
option.
I
think
we
have
some
other
options.
A
All
right
so
I
think
we
should
keep
moving
on
this
approach
and
then
do
more
thorough
discussion
on
that
option
and
then
move
to
the
next
option
and
then
I
would
finally
discuss
if
there's
any
other
options
to
to
solve
those
issues,
and
then
we
can
do
more
discussion
there.
Does
that
make
sense
to
you
guys
yeah,
because
I
know
this
approach,
that
kubernetes
took
to
two
of
the
the
ownership
issues.
So
I
think
we
should
be
started
here
and
try
to
reevaluate,
whereas
no.
A
C
C
A
Doesn't
care
actually
I
see,
but
you
do
have
the
restriction
like,
for
example,
for
different
volumes.
You
don't
want
them
to
be
assigned
to
the
same
GID,
because
otherwise,
then,
potentially
some
tasks
don't
are
not
authorized
to
access
to
a
certain
volume
can
basically
add
a
feeling
that
it's
not
intending
to
so
so
you
give
me
one.
I
generate
one
g
ID
for
each
volume,
yeah
I,
don't
know
like
either
each
money
on
each
role.
D
E
A
A
Not
necessarily
I
mean
it's
just
an
ID
on
any
of
you.
You
part,
don't
need
to
do
that,
like
you,
don't
have
to
map
92
an
actual
stream
in
for
the
container.
All
you
need
to
do
is
for
the
process
on
launching
kind
of
container
the
first
process
you
make
sure
that
we
have
is
supplementary
group
in
each
group
list.
A
C
C
A
But,
for
example,
I
say:
I
have
two
containers
once
it
is
using
volume
and
like
the
first
container,
a
desk
like
say
stones.
That's
what
I
asked
you
to
be
things
like
a
thousand
and
then
the
second
contains
a
thousand
and
one
on.
So
you
send
key
ID
on
that
body
to
be
on
a
thousand
and
thousand
once
all
the
files
created
by
the
I
know
the.
A
A
C
A
A
If
this
is
the
right,
some
extra
knowledge
for
us
because,
like
like,
if
you
want
the
file
created
by
a
tab,
to
be
accessible
by
an
tab,
you
can
certainly
make
sure
that,
on
like
theater
user
mattress
or
their
some
of
their
group
matches
things
like
that.
So
there's
multiple
ways
you
can
achieve
that
and
I.
Don't
know
if
that's
okay,
a.
D
D
G
A
I
think
yeah
I
mean
two
subjects.
It'll
damage
like
a
daddy
idea
is
pretty
important
because,
like
I
say,
the
task
finishes
the
next
time.
I
want
to
make
sure
that
gid,
like
the
prevention
group
ID,
is
the
same
as
we
used
before
so
I'm.
So
done.
That
means
like
dynamically
generate
and
I
need,
doesn't
make
sense
to
make
sure
that,
like
a
56
to
mapping
that
I
can
be
used,
edgy
I
next
time,
I
create
a
container
I,
say
the
task
crashes
in
the
relaunch
by
the
framework.
A
Yeah,
it
will
have
to
be
persisted.
It
has
to
be
persistent
where
or
like.
Oh,
you
derived
that
from
some
somewhere
and
then
make
sure
it's
the
same
for
same
volume
all
the
time
it
cannot
be
like
on
the
fly
generate
ID
and
they
can
be
I
mean
it
should
not
be
different
in
different
runs,
okay,
so,
okay,
so
let's
go
move
forward
and
see
what's
left
there
in
the
dock,.
C
This
looks
very
nice
and
very
good
to
us,
but
the
problem
is
that
effective
permissions
and
the
default
ACL
that
are
inherited
for
newly
created
objects
in
the
volume
that
logic
is
kind
of
complicated
and
well.
For
me,
it
looks
hard
to
manage
because
those
people
to
atl
the
inherited
ACL
they
take
into
account
the
mode
of
the
file
when
it
when
it's
created
and
when
the
mode
change
changes.
The
ACL
does
not
change
so
I
think
James
speech.
You
know
a
lot
about
those
ACLs.
Maybe
you
can
tell
us
about
that.
D
From
so
it's
a
co
so
wacky
and
terrible
and
unique
modes
ACLs
for
this
application.
Basically
exactly
what
you
were
doing.
Activity
on
Windows,
so
I
think
the
windows
people
they're
going
to
want
to
district
put
a
CEO
on
the
volume
they'll
work
for
them.
I,
don't
know
that
the
default
a
CL
inheritance
on
on
Linux.
So
when
you
look
at
this,
do
you
know
if
they
default?
A
CEO
is
actually
inherited
to
child
causes?
Objects.
A
C
B
A
D
A
So
I
think
the
first
priority
is
trying
to
figure
out
I'm
trying
to
in
a
dog,
specify
all
the
problem
we
are
trying
to
solve,
and
these
are
dope.
I
propose
solutions
and
what
are
the
limitations?
And
then
we
pick
one
that
we
want
to
go
forward.
So
that
kind
of
my
suggestion,
I
kind
of
agree,
Kate
James
peach,
that
we
should
have
a
province
payment
very
clearly,
mythographers.
A
C
C
G
D
The
reasoning
behind
it
is
that
with
mode
bits
are
kind
of
wonky
and
they
don't
really
behave
in
the
same
ways.
People
want
apples
to
behave
so
anytime.
You
go
down
the
path
of
trying
to
combine
system
with
mo
bits
and
apples.
You
end
up
with
these
and
of
corner
cases
of
row,
which
one
means
what
and
how
do
you
combine
the
meat
Institute
so
I
do
agree,
though,
here
that
I
think
in
the
Linux
world,
apples
aren't
actually
widely
used
in
practice
and
the
complexity
is
kind
of
daunting
and
I.
A
A
So,
what's
left
on,
the
dark
is
for
tickets
that
are
in
the
relevant
physics
section.
Capital
e,
the
second
one,
so
48
93
is
what
we've
done.
As
you
know,
previous
attempts
not
for
you
know,
changing
the
ownership
of
the
following
for
each
task
or
other.
To
specifically
have
you
specify
the
the
owner
and
the
groups
and
basically
all
of
the
relevant
gets
when
you
were
actually
creating
the
system
bullying.
A
But
of
course,
that's
only
applicable
to
this
is
system
boolean,
so
I
don't
have
a
versatile
we're
one
of
the
nominated
to
be
but
come
with.
That
is,
and
we
would
so
in
India
I
already
have
a
code
and,
and
he
had
a
divine
box
and
I
think
G
and
and
myself,
and
they
actually
have
you
know,
talk
about
them,
but
the
problem
eventually
that
kind
of
stalled.
This
was
that
we
don't
have
a
mechanism
for
the
operation.
A
This
kind
of
a
resource
operation,
so
I
think
this
prize
very
closely
to
suji's
proposal
on
the
resource
providers,
but
the
problem
with
precision
bullying
is
I.
Usually,
cook
creamer
can
can
request
this
UID
this
GID
and
this
number
of
this
list
of
sublimity
GID
to
be
said.
However,
I
want
what
it
said
that
doesn't
work.
What
is
that
fails
to
do
that?
Then?
A
The
the
agent
currently
lacks
this
mechanism
is
to
self.
Neither
subversion
has
failed.
So
so
that's
what
I
hope
with
the
resource
provider
work
on.
Alongside
with
it,
we
can
standardize
on
this
feedback.
Aesthetic
update
for
all
of
these
operations
and
with
that
then
I
think
we
can
then
continue.
This
to
you
know,
have
the
boolean
creator
to
specify
the
Uwharrie
GID,
and
let
me
reschedule
so
so
the
API
is
going
to
be
so
excited.
Remember:
traffic.
A
B
A
Its
specifically,
it
has
to
be
a
name
right,
cannot
be
the
numeric
and
you're
saying
like
they
think
this
model,
like.
Oh,
you
can
show
the
creation
and
the
you
can
show
the
permission.
Ownership
of
the
boning
that
on
someone
crate
and
whether
a
container
connects
to
the
bottom
clearly
depends
on
the
operator
completely
using
the
group
information.
A
Well,
because
the
best
is
that
your
cluster
should
be
homogenous
enough,
so
that,
and
which
is
true,
I
think
when
we
were
running
this
process
and
the
working
relations
at
work.
I've
worked
with
on
individually,
so
the
firmer
can
Qualis
be
certain
that
certain
users
and
IDs
art
on
that
on
that
on
all
of
the
boxes.
So
it's
definitely.
These
frameworks
choice
to
first
create
volumes
in
that
in
a
substructure
way
that
that
can
support
its
usage
pattern
and
then
launch
tasks
that
conform.
A
So
that
usage
pattern
so
that
you
do
not
run
into
violations
security
permissions.
So
that
is
certainly
probably
tonight.
It's
going
to
Olivia
alleviate
the
problem
with
you
know,
changing
jerseys
and
using
different
someone
for
Jenny
different
every
time,
but
what
I'm
not
certainly
is
outside
of
the
persistable
in
a
case,
if
you
have
all
these
volumes
external
and
front
post
and
what
they
would
them,
yeah
I
think
like
the
hosts.
A
A
We
don't
have
external
one,
because
that
can
be
deprecated
I'm,
not
deprecated,
but
like
it's
not
going
to
be
like
something
that
we're
gonna
support,
long
term
and
all
things
is
going
to
be
on
the
CSI
based
warnings
based
the
persistent
volume
so
as
long
to
figure
out
a
solution
for
position
volumes
that
I
think
I
should
cover
all
the
external
logging
cases.
Although,
as.
E
A
Through
the
CSI
supported
bowling,
they're
still
ultimately
created
by
the
framework,
so
I
think
if
the
same
thing
can
happen
for
the
even
of
the
framers
designate,
what
do
you
Rd
gdns
of
them,
3G
IDs
user-
to
give
to
these
volumes?
The
word
rent?
Yes,
so
just
FYI,
so
CSS
spec
right
now
does
not
have
diagnose
any
information
in
english
pesto
on
this
currently,
no
way
that
an
operator
or
a
framework
in
on
presence
on
to
be
under
a
particular
group.
So
there's
I
mean
you
can
pass
as
a
parameter.
A
B
A
Our
repo,
as
well
like
so
that
I
could
for
an
uncle
when
you
create
a
bony,
actually
say
hey.
This
is
a
user
in
a
group.
I
want
Gabon
into
the
end,
but
it's
going
to
be
tricky
because,
like
I
also
see
aside
from
the
time
when
I
can
not
just
it's
going
to
be
very
generalization,
handle
NFS
all
these
kind
of
like
a
different
storage,
backhands
or
different
stories.
I
can
have
different
way
of
getting
user
and
group.
So
I,
don't
know
how
to
I
wrote
that
so
definitely
I
guess
if
I'm
correct.
A
D
A
To
feedback
all
the
upper
operations,
so
in
essence
we
can
get
feedback
head.
So
that's
it
works.
Awesome,
I
know
you're
good
at
as
part
of
this
here
soldier,
a
binary
post
I
mean
even
if
that's
all
that
we
seem
to
have
discussion
already
survived
approach
and
I
think
we
need
to
do
a
thorough
evaluation.
We'd
like
to
start
by
things.
Really
thinking
like
like
a
refinery
up
or
maybe
I
can
work
with
you
and
again
in
accent
that
we
guys
can't
work.
A
Payment
I
could
work
I
think
we
should
agree
on
what
the
semantics
to
and
support
and
then
figure
out
a
solution
right
so
to
me
that
the
bar
has
to
be
set
higher.
So
for
so,
if
because
technology
aside,
I
mean
allowing
files
and
potentially
I,
don't
know
millions
event,
I
should
be
modified
upon
a
task
launcher,
container
creation
could
be,
could
be
scanned.
The
directory
scanned
and
modified
as
just
to
testify.
A
B
A
I
think
they're,
like
a
multiple
like
a
barrier
and
all
the
solutions
that
we
discussed
in
the
dog
and
we
can
discuss
later,
but
I
think
most
important
things.
We
figure
out.
That's
what
exactly
the
semantics
who
you
want:
okay,
different
type
of
volumes
like
host
volume,
like
what
does
the
nicest
I.
E
A
Alright,
so
I
documen
those
comments.
Thanks
for
the
comment
and
what's
for
the
next
step,
we
have
13
minutes
left.
My
key
needs
next
step
is
on
the
enacting.
A
peach
like
in
India
right,
let's
sit
together,
I
can
to
figure
out,
let's
work
together
on
and
talk
on
like
this,
it's
the
right,
semantics,
everyone
and
then
don't
but
pollution.
Yet
just
like
it
it
did.
You
see
what
we
want
like.
Let's
agree
on
that
first
I
feel
like
solution
to
do
that.
A
Work
I'll,
try
to
schedule
a
meeting
on
this
one
so
that,
like
we,
can
work
on
this
and
towards
a
solution
for
that.
Does
that
make
sense
to
you
guys?
Probably
anyone
else
are
interesting.
This
topic,
let
me
know
and
send
an
email
to
me
and
I'll,
create
a
meeting.
I.
Think,
like
the
goal
of
that
meeting,
is
still
trying
to
figure
out
what
the
right
semantics
should
be
for
this
problem,
which
type
of
water
and
then
which
we
agree
on
that,
let's
think
about
solutions,
not
good.
That's
my
action
either.
A
All
right,
so
we
have
like
10
minutes
left
I
do
want
to
go
through
the
planning
scrap
that
we
have
last
time.
So
on
me
decades,
I'm,
clearing
and
I
can
share
my
screen.
Oh,
you
already
did
sorry
about
that.
All
right,
so
hey
guys
see
my
screen
yeah
all
right,
so
I
think
we
have
10
minutes.
I.
Think
I
want
to
do
a
quick
status
update
on
these
things
and
like
this
is
kind
of
the
thing.
A
In
the
first
meeting
we
kind
of
like
lifts
all
the
things
that
we
want
to
do
and
on
their
urges,
urgency
and
the
approximation.
We
don't
have
a
target
release
yet,
but
if
you
do
have
a
target
release,
hopefully
the
owner,
please
put
the
target
release,
like
which
version
you
want
on
the
future,
to
reason,
and
so
people
can
plan
ahead
and
I
just
want
to
do
a
update
like
only
like
hi
party
either,
and
if
you
do
want
to
say
something
about
some
like
low
party
items.
A
B
A
A
B
A
A
Actually,
listen.
Are
you
interested
working
with
the
NX
on
this
one,
one
yeah
sure
like
tomorrow
we
can
talk
offline
about,
like
my,
how
we
coordinate
on
this
one,
okay,
yeah,
so
so
great,
so
tnx.
What's
the
party
for
this
one?
So
it's
like
last
how
we
kind
of
collab
the
thing
to
teach
you
which
means
like
it's,
not
super
nice.
You
have
I,
don't
have
a
mediate
on
to
you.
A
Feature
has
been
flying
next
six
months
and
we
want
to
roll
out
to
production.
That's
q1
easier
is
like
I
need
a
yesterday,
yeah
well
yeah,
yeah
I,
just
think
I'll
for
for
for
my
priority
to
one
some
more
appropriate
okay,
I'll
change
that
to
t1,
and
you
can
think
with
Jason
I
think
Gilbert
you
can.
Maybe
I
could
share
this
work.
Yes,
alright!
So
we
don't
have
too
much
time
left
so
I
wanna
go
through
like
a
t0
and
t1
if
possible,
like
what's
the
status
on
the
image.
E
B
G
B
A
Month
a
thing
like
that:
building
off
1.3
takes
like
a
month,
I
think,
which
is
something
that
once
in
6
for
sure,
okay,
so
1.4
is
something
that
you're
looking
for
yeah.
Okay,
we
just
designed
we
just
we
just
discussed
the
bonding
Amish
eminent
group,
cleanup,
saying
Experion,
design
and
so
I
will
because
I
think
I,
like
J
teach
and
year
next,
is
that
okay
education
units
on
this?
Are
you
guys
interesting
this
one
yeah,
that's
good,
all
right,
cool
I
think
which
is
difficult
at
and
documentation
so
Gilbert.
G
F
F
A
A
A
E
A
B
E
A
Alright,
so
you
have
a
ticket,
also
yeah,
by
the
way,
make
sure
to
have
a
link.
So
of
all
these
features
the
owner
should
I
have
a
link
to
the
tickets.
Oh
my
god,
just
like
do,
link
here
to
a
hyperlink,
make
sure,
like
everyone
can
act
to
the
zero
to
understand
with
the
contact.
So
you
just
create
a
ticket,
but
you
haven't
started
yet
no.
A
A
No
okay,
it's
channeled
online
I
got
teeth,
not
I,
think
Oh
see
I
I
can't
speak
for
Chan
I'll
show
for
this
one,
so
I
think
he
sends
out
a
bunch
of
patches
I.
Think
it's
seeing
a
place
that
we
can
review
the
code,
but
I
do
want
to
do
some
changes,
especially
regarding
the
caching
and
store
I.
Don't
want
to
introduce
like
yeah
and
a
store
I
want
to
make
sure
that
we
are
moving
towards
a
unified
cache
in
directory
for
all
the
artifacts.
A
A
A
This
one
hasn't
been
started
yet,
but
that's
something
that
we
are
very
interested
in,
especially.
We
want
to
support
things
like
junky,
since
I
am
natures
container
like
dr.
Deema
inside
of
music
content,
so
I'm
kind
of
interesting
this
one
I
plan
to
spend
some
time
on
doing
mrs.
Cullen
packets
onto
to
tack
on
this
see.
If
I
can
you
spend.
D
A
I,
alright
I'll
link
a
ticket
I
think
it's
meaning,
like
I,
just
run
a
full-fledged
stalker,
DEMA
entire
metal
container,
and
then
things
like,
for
example,
like
I,
want
to
run
campaigns
build
out.
I
want
to
do
darker,
building
memories
of
container,
which
is
support
dad,
but
I
don't
want
to
use
the
host
out
Redeemer,
because
it's
a
great
color
abstraction
I
want
to
make
sure
that
all
the
path
launch
by
that
doctor,
you
don't
get
killed
when
the
mixed
container
teardown
think
like
that.
Okay.
A
That
true
I
think,
like
we
have
something
missing.
This
is
meaning
like
Ally
lxd
already
supports
that
you
can
run
like
a
demon,
king
pile,
a
LSD
container.
That's
so
that's
the
thing
we
want
to
do
any
as
well.
I
think
something
missing
is
like
cheating
these
places.
You
may
be
using
these
places.
Things
like
that.
So
I
already
check
it
out
this
one
like
what
exactly
the
piece
missing
right
now.
A
A
A
Well,
so
that's
two
weeks
away
right:
yes,
that's
in
two
weeks,
so
I
think.
If
yes,
Jason
Landover,
we
can
sit
together
chat
about
this
folks,
okay,
but
I
I
guess
I'll
have
to
let
you
in
low
not
now
but
closer
to
the
take
your
pick
yeah.
It
does
not
have
to
be
a
design
goal
cut.
Something
like
that.
It's
just
like
a
hey
I
want
to
discuss
something.
A
A
D
A
To
plant
around
all
exported
support,
sorry,
okay,
so
that's
great,
so
I
can
I
can
get
into
like
tending
agenda
items
and
the
depend
on
like
your
you
guys
time
or
something
like
that.
You
can
like
a
put
into
different
slots.
So
I
hear
like
a
capability
stuff
host
port
isolator
and
unix
and
gilbert
aces
like
people
really
close
focus
means
the
items
I
think.
Maybe
we
don't
have
time
to
go
through
all
of
them
next
time.