►
From YouTube: Charts Chat 20180418
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
Yeah
there's
this
progress
which
are
linked
and
they
don't
want
to
put
the
stable
image
I
want
to
use
latest,
because
this
is
their
update
policy
and
I
find
that
kind
of
weird,
because
they
also
have
stable
an
interesting
gap.
I
think
and
I
think
we
should
require
a
stable
version.
Usually
usually
two
I
mean
a
release.
Facilities
I
want
to
have
a
latest
image
in
each
other.
A
A
A
B
A
I
would
say
it's
a
pain,
the
other
way
around,
because
then
you
don't
have
control.
For
example,
if
you
do
latest
and
you
push
with
dev
environment
and
everything
works
or
QA
environment
and
the
version
gets
incremented,
you
push
it
to
production,
and
then
you
get
a
different
version
than
what
you
tested
and
you
don't
realize
it
and
there's
no
pinning
versions
are
really
important
in
a
lot
of
environments,
and
so
we
don't.
We
have
a
policy
that
you
need
to
use
a
pin
version.
I.
A
B
C
C
Yeah
image
tags
should
use
stable
versions.
Okay
under
configuration
the
only
so
I'm
just
just
thinking
about
this
chart
here.
It's
log
DNA
agent,
which
I
believe
is
an
agent
for
a
SAS
service
and
and
so
the
only
thing
that
the
reason
why
they
may
want
to
use
latest
here
is,
if
you
know,
if
you're,
using
a
really
old
version
that
doesn't
doesn't
work
with.
You
know
the
latest
version
of
the
SAS.
That
might
be
a
problem
for
them.
That
said,
you
know
if
that
pod
number
restarts,
then
it's
always.
C
A
I
mean
I
can
understand
the
pros
and
cons.
We
have
a
policy
for
a
reason
like
I
still
want
a
pin
right
and
that's
our
default
policy
and
I.
Think
going
with
the
latest
is
gonna
end
up
creating
long
term
problems
for
us
just
if
anybody
actually
does
try
to
use
this
for
real
and
they
want
to
have
everything
locked
down.
Do
you
actually
trust
your
upstream?
Are
you
auditing?
How
does
this
all
work,
and,
at
least
in
enterprise
environments,
were
a
lot
more
picky
about
what
we
run?
A
Sometimes
we
actually
have
to
build
the
images
ourselves
and
some
of
the
enterprise
environments.
You
can't
even
grab
upstream
images
depending
on
your
policies
and
what
you've
got
to
go
under
and
what
regulations
you're
under
and
so
I
can
I'm
more
of
the
pinning
side
of
things
and
I
would
ask
them
to
pin
any
way,
because
we
don't
acquiesce
to
everybody
else's
policy.
They
come
to
our
policy.
C
Right,
yeah,
that
makes
sense
to
me
yeah
and
then,
and
also
the
the
thing
I
said
about
the
SAS
thing.
If
it
happens
that
someone's
on
an
old
version
that
doesn't
work
with
the
SAS
thing,
you
know
there
should
be
an
error
day
on
on
the
Southside
saying
you
know,
you
need
to
go
and
update
your
agent
and
that
should
be
a
straightforward
process,
so
say
I.
Think
I
think
we.
We
can
probably
stick
to
our
policy
here.
A
And
if
you're
a
SAS
service
right,
you're
gonna
have
people
who
are
pinning
you
can't
just
always
count
on
them
doing
latest
anyway,
and
so,
if
they're
saying
you
know,
if
that
ends
up
being
a
problem
just
to
speculate
for
a
moment,
they
don't
want
to
do
that.
Then
they're
gonna
run
into
other
issues,
because
if
they've
got
big
companies,
they're
gonna
enforce
pinning.
C
I
put
this
up
so
this
one
came
up
with
the
set
of
changes
to
the
Redis
chart.
The
stable
Redis
chart
to
add
cluster
support.
I
think
I,
don't
think
it's
configured
by
default,
if
you,
if
you
pump
out
a
replicas
in
it
and
it
enables
a
class
of
support,
but
it
does
it
in
a
in
a
fairly
different
way
to
the
existing
Redis.
That's
H
a
shot,
so
the
Redis
H
a
chart
uses
Sentinel,
which
is
you
know,
entirely
valid
bit
different
approach.
C
So
I
was
wondering
if
it
made
sense
to
rename
the
Redis
H
a
chart
to
read
a
sentinel
to
make
it
more
clear
what
the
strategy
for
replication
is
there
and
and
that's
similar
to
MongoDB,
for
example,
which
has
the
MongoDB
replica
set
name.
So
it's
very
clear
what
what
replication
strategy
is
being
used
there
I,
don't
think
we've
ever
renamed
a
chart
before
no,
so
that
might
be.
E
C
E
C
B
B
B
A
E
A
A
A
quick
question:
what's
I
mean
aside
from
the
short
namespace
name,
that's
used,
is
there
a
big?
What's
the
end
value
to
users,
to
like?
What's
the
reason
in
changing
the
short
a
little
bit
with
stable
the
name,
you
do
the
way
it
sits
space
or
directory
like
we're,
asking
people
to
jump
through
hoops,
we're
doing
a
bunch
of
work.
So
what's
the
end
value
for
people
doing
that.
E
E
With
a
lot
of
implementations
of
Redis
H
a
where
it
was
an
AJ
to
begin
with,
where
you
know,
people
misunderstood
whether
it
was
you
know
the
active,
active
type
stuff
for
the
charting
stuff
work,
so
I
think
there's
enough
confusion
on
on
the
uptake
side.
It's
gonna
be
painful.
For
anybody
and
everybody
who's
been
using
a
chart.
B
E
D
C
How
do
you
go
and
pick
the
right
one
that
fits
your
needs
right,
because
sentiment,
Sentinel
might
be
something
you
need,
or
you
might
want
something
more
kubernetes
native,
so
I,
say:
yeah
I!
Think
that's
where
I
see
the
about
you
and
renaming
it,
because
otherwise,
you
know
you're
gonna
have
to
dig
into
the
Redis
a
chart
to
figure
out
what
it's
actually
doing.
E
It's
a
number
yeah
I
wish
we
had
like
you
know
in
some
future,
where
were
tribes
have
taken
over
the
planet?
I
use
a
chart.
You
know
Advisory
Board,
where
we
could
reach
out
to
end-users
and
say
hey
we're
about
to
change
this
stuff.
What
what
do
you
think
we
don't
have
anything
like
that?
It
would
take
a
long
time
to
get
there,
but
I
definitely
think
it's
something
we.
E
E
C
G
Like
it's,
it
gets
messy,
it's
probably
better,
to
start
with
straightening
change
links.
Then
you
know
where
everything
yeah.
A
But
even
when
it
comes
back
to
this
red
sha-1
right
so
I
like
to
have
a
change
logged,
but
I
also
could
the
readme
and
all
the
documentation
be
changed
so
that,
because
we're
really
just
talking
about
that
directory
name,
that's
all
we're
talking
about
changing,
because
everything
else
the
patterns
he
uses,
how
it
uses
that
can
all
be
updated
in
the
read
mean
everywhere
else,
and
if
I'm
searching
for
it
on
Kubb,
apps
and
I
search
for
something
right,
it
should
be
able
to
pull
information
from
that
readme.
That
gives
all
those
extra
details.
A
C
I
think
the
issue
is,
if
you
did
helm,
Redis,
helm,
sorry
home
search
today,
I,
don't
think
it
would
show
up
and
kind
of
add
a
glance
just
being
able
to
see
Redis
and
actually
I
like
helm,
says
no,
it
does
work
so
I
guess
it
is
searching
over
to
read
me
all
the
description
or
something
but
yeah.
If
I
don't
know,
if
I
did
a
helm
search
Redis,
it's
not
clear
to
me
that
that
charts
using
Sentinel.
C
Okay,
yeah
I,
think
I
think
we
have
tools
like
tools
like
you,
apps
hub.
It's
a
little
easier
to
see.
What's
going
on,
it's
you
can
you
have
to
read
me
but
but
yeah?
Definitely
the
helm,
search
cases
there's
no
way
to
just
you
know
fewer
read
me
or
I.
Guess
you
can
inspect
inspector
chart,
but
it
I
don't
think
that
I
think
does
I
show
you
to
read
me
if
you
inspect
a
shop.
E
A
C
G
And
you
can
you
can
switch
a
like
your
release
to
a
new
chart
anyway,
country
I.
Think.
E
G
C
E
G
It
mostly
constraining
Albert
Johnson,
a
few
other
things.
E
E
Issue
and
and
I
think
we've
already
lost
this
particular
chart.
Maintainer
and
church
rebuilt
user.
They
decided
to
post
their
own
charts
for
various
reasons,
including,
but
we
move
slowly
on
things
that
they
need
to
get
their
products
out
the
door
and
enhance
their
users.
So
I
think
we
need
to
come
up
with
a
strategy
for
facilitating
people
that
are
shipping
their
own
software.
We
think
we
feel
that
use
case.
E
A
Yet
this
kind
of
comes
to
the
whole
problem
of
the
queue
is
getting
it's
getting
pretty
busy
for
pull
requests.
It's
a
lot
of
work
to
maintain
a
few
effect.
If
you
go
look
at
the
contributions
for
the
charts
repo,
there
are
a
lot
of
ridiculous.
The
contributions
are
really
up
right.
So,
even
though
we
continue
to
merge
more
and
more
and
more
stuff,
more
and
more
things
are
are
coming
in,
I
mean,
go,
go
look
at
the
graph
here.
A
I'll
drop,
the
link
in
for
contributors
and
you're
gonna
see
why
we've
got
the
problem
right.
There
I
mean
look
at
this
spike
for
contributions
that
we're
getting
right.
It's
it's!
It's
it's
big
and
ya,
go
and
then
go
look
at
the
backlog
and
look
at
how
many
more
charts
people
want
to
contribute
that
we
can't
even
review
this
morning.
I
was
going
through
things.
I
was
like
updated.
Readme
I'm,
like
okay
I'll,
go
review
this
one
real,
quick
and
I
realized.
A
That
was
just
the
last
commit
on
a
new
chart
that
somebody
wanted
to
contribute.
So
many
our
repo
model
right
now
is
having
trouble
scaling
with
the
number
of
people
who
review
things
or
the
number
of
people
who
have
owned
things
and
the
way
the
whole
thing
works
and
that's
a
concern
we're
having
trouble
keeping
up
with
everything
being
in
here
with
the
way
this
all
works.
You.
G
Know
that's
I
I
noticed
we've
been
contributing
some
things
and
running
off
local
build
copies.
You
know,
we've
got
a
repo
which
is
managing
our
to
cost
us
we've
got
some
of
the
stuff
that
hasn't
been
merged.
It
could
just
local
jobs
where
they're
pulling
out
of
that,
but
yeah.
So
they
understand
is
another
reason
why
I'm
playing
for
the
membership.
E
E
D
E
E
You
know
you're
on
some
gradient
here
of
like
trying
to
keep
quality
versus
just
letting
everybody
commit
whatever
they
want,
and
they
went
more
towards
the
you
know
you
can
commit
it,
and
then
they
had
the
Stars
mechanism
and
those
kinds
of
things
that
allowed
the
good
ones
to
bubble
up
to
the
top.
And
so
they
kind
of
just
have
a
different
approach
from
the
get-go.
E
G
E
Certainly
am
a
hundred
ex
more
passive
in
my
review
than
that
right,
hardest
and
I.
Just
I
can't
scale
to
you
know
every
boat
he's
done
an
amazing
job
of
it,
but
I
would
like
for
us
to
be
able
to
say
see,
is
green,
go
and
just
leave
it
at
that
with
with,
with
a
light,
look
at
it
just
to
make
sure
that
there's
nothing,
obviously,
mining
Bitcoin
in
the
middle.
That
would
be
my
ideal
state,
and
then
we
could
just
kind
of
just
be
making
sure
that
there's
nothing
insane
going
in.
E
E
D
A
So
maybe
yeah
I,
like
where
you're
going
with
get
more
things
in,
we
probably
need
to
push
for
more
people
to
get
owners
files.
So
they
can
approve
the
stuff
themselves
and
then
get
notified
with
the
reviewers
on
hey.
Come
review
this
because
some
of
the
charts
that
are
doing
that
I've
just
stopped
looking
at
because
it's
taken
care
of
it's
always
the
new
additions
and
then
the
church
that
don't
have
owners
files
or
the
ones
that
only
have
one
owner
that
there's
a
problem.
G
E
Everybody,
the
problem
is,
they
have
to
be
also
kubernetes
members,
so
that
limits
we
know
in
and
say
who's
committed,
the
most
or
who's
out
of
the
most
lines
of
code
to
this
chart
and
add
them
to
owners,
and
then
we
have
to
go,
find
them
chase
them.
Get
them
to
apply
for
membership,
which
is
has
been
harder
than
I
would
have
thought.
F
F
Think
it
may
be
worth
setting
the
expectation
that
hey
just
because
you
submitted
her
or
PR
does
not
mean
that
you
should
set
your
business
on
that
PR
getting
in
within
a
specific
time
frame,
because
this
is
not
just
about
your
business,
I
mean
yeah.
We
want
to
get
T
ours
and
quick
as
we
can,
but
especially
when
someone
like
this,
we
go
hey.
This
needs
more
discussion
before
we
decide
whether
this
is
legit
or
not.
That
discussion
has
to
be
given
some
room
that
happens
and
I.
F
E
That's
totally
fair
thing,
though
we
should
have
had
the
discussion
on
the
issue.
Just
like
any.
You
know
open-source
project
we
shouldn't
have
to
have
the
person
join
this
meeting
they're
in
at
different
time
zone,
so
maybe
completely
inconvenient
for
them
to
to
be
here.
We
have
an
open
issue
where
we
can
have
a
thread
to
discuss
it.
So
if
it's
possible,
we
should
just
act
exactly
this
in
those
threads
yeah.
C
And
I
don't
mean
to
to
move
the
discussion
away
from
the
issue
when
I
suggested
getting
on
a
core.
Just
sounded
like
that
would
be
easier
to
hash
out.
I
did
forgot
to
take
into
account
that
time
I
was
in
a
different
time
zone
entirely
better,
but
yeah
I
agree
in
general.
We
should
try
to
keep
discussion
to
the
issues.
A
So,
but
here's
a
thing
right:
we
keep
coming
back
to
charts
and
folks
think
about
it.
The
way
we
think
about
NPM
right,
I'll,
just
use
NPM
as
the
model,
where
you've
got
this
one
giant
monolith
that
just
holds
everything
right.
It's
just
one
massive
repository
and
other
systems
like
apt
and
there's
lots
of
other
systems
use
a
distributed
model
where
there
is
a
central
repo
I
mean
even
I
was
poking
at
ruts
model,
because
rust
has
an
index
that
you
download
and
it's
it's
similar
to
the
model
we
use.
A
But
of
course
you
can
have
other
you
know
with
these
models
that
do
this.
You
can
have
multiple
registries
and
indexes
just
like
docker,
there's
docker
hub,
then
there's
Quay,
there's
GC,
are
lots
of
people
have
their
own
on-prem?
It's
kind
of
this
distributed
model
for
different
registries
and
helm
is
actually
built
on
that
model,
but
I
know
a
lot
of
folks
businesses
and
things
like
that
tend
to
think
of
here's,
the
central
repo
get
it
here.
A
Maybe
there's
a
way
we
can
say
host
your
own,
here's
how
you
can
do
it
and
here's
how
maybe
you
can
make
that
more
discoverable
and
then
they'll
be
more
interested
in
hosting
their
own,
and
people
will
know
how
I
can
add
it.
I
don't
know
how
to
solve
this
problem,
but
that's
that
might
be
a
way
to
go.
E
We
don't
have
great
policy,
so
you
might
have
a
change
that
you
think
is
innocuous
I'm
totally
fine,
within
the
bounds
of
you
know
what
the
repo
should
expect
and
that
might
cause
a
hang-up.
So
things
like
that
just
kind
of
come
out
of
nowhere.
So
that's
where
you
know
I've
been
relatively
lenient
for
things
that
we
don't
already
have
documented
somewhere
to.
E
G
F
So
I
wonder
if
there's
a
possibility
where
we
could
say:
hey
here's
a
place
where
you
can
add
a
description
of
your
repo
and
then
you
can
point
your
home
plan
and
say
like
hello,
repo,
ad
or
home
repo
list,
and
it
goes
out
for
this.
You
know
kind
of
listing
of
repositories
and
says
here
all
the
repos
you
can
add
which
one
you
want
to
add.
D
G
E
G
E
G
C
C
You
know
I
think
one
one
point
about
the
curation
is
I,
think
I
think
it
is
good
to
have
best
practices
and
and
have
some
sort
of
you
know
being
sure
that
that
the
charts
in
the
stable
repo
generally
follow
the
best
practices.
I
think
that's
valuable
and
you
know
comparing
it
to
some
other
package
managers
out
there,
like
NPM,
like
like
there's
a
slightly
different
use
cases
in
that
what
you're
installing
is,
is
you
know
the
the
upstream
software
directly
and
it's
not
a
set
of
manifests,
are
kind
of
generic.
C
You
know
kubernetes
related
and
may
have
certain
best
practices
applied
to
them.
So
a
lot
of
the
best
practices
that
we
have
are
based
on.
How
you
know
things
are
configured
within
the
cluster
and
not
necessary
to
do
with
the
actual
application
right,
so
how
storage
processes
are
used,
how
our
back
is
installed
and
stuff
like
that
and
I
still
I,
think
I
think
there
is
a
lot
of
value
in
having
a
gate
there,
just
to
check
that
those
things
are
are
being
done
in
a
in
a
correct
and
and
also
standard
way.
E
But
the
hard
thing
for
me
is
that
I
can't
keep
tabs
anymore
like
when
we
were
at
50
charts
and
seeing
these
patterns
come
up.
I
could
keep
track
of
them
now.
I
can't
tell
if
I'm,
able
or
auerbach
not
true
or
our
back
do
the
thing
I
just
I,
don't
know
how
to
keep
track
of
all
that
stuff
without
spending.
You
know
an
hour,
/
PR,
making
sure
that
all
this
stuff,
that's
that's
at
especially
new
charts,
yeah.
E
C
So
that's
kind
of
why
we
started
putting
together
the
review
guidelines,
doc,
which,
basically
because,
before
what
we,
what
we
used
to
do
is
you
know?
Whenever
did
we
see
someone
trying
to
configure
our
back?
We
point
to
another
chart
that
we
knew
follow
the
best
practices,
but
if
we
can,
if
we
can
just
point
this,
a
concise
stalk
which
just
shows
the
different
patterns
and
the
best
practices
for
this,
but
for
the
pattern,
then
it's
much
easier
way.
Just
to
refresh
you
know
your
memory
and
and
check
out
it's
actually.
E
B
E
If
you,
if
you
find
something,
that's
wrong
in
a
in
PR
that
you're
gonna,
you
know
point
out
to
someone
and
say:
can
you
fix
this?
Can
you
make
sure
it's
added
here
and
then
at
least
I
can
try
to
make
sure
that
I've
got
this
open
all
I'm
doing
my
chart,
reviews
or
see
if
there's
things
there,
okay,.
B
C
So
I
was
gonna,
say:
there's
a
couple
good
discussion
points
from
this,
but
I
did
want
to
just
gain
overall
consensus
on
what
people
is
it.
It
sounds
like
you
know.
The
people
who
are
interested
in
this
have
commented
and
raised
their
forts,
but
I
think
we
need
to
make
a
decision
on
on
what
we
actually
want
to
put
into
guidelines
here.
C
So
I'm
I
think
I
agree
with
fake
what
you
said
about
the
notes
being
the
place
to
to
make
this
obvious.
Matt
Fisher
raised
a
good
point
that
if
a
parent
chart
includes
a
child
chart
that
is
collecting
collecting
data,
then
you
won't
see
the
notes
for
that
child
shot.
So
you
may
not
know,
but
I
think
that
I
guess
it
this
I
guess
the
only
way
to
really
fix
that
right
now
is
to
see
if
there
are
any
parent
charts.
C
E
E
A
Yeah
yeah
and
when
I
think
short
term
part
of
me,
so
how
much
designed
around
the
apt
model
very
specifically
around
the
Debian
apt
model,
and
so
there's
probably
a
lot.
We
can
gain
from
that
right.
So,
if
I
look
at
a
blue
tube,
there's
that
there's
some
repos
that
are
automatically
added
that
are
used
by
apt
out-of-the-box,
but
for
a
lot
of
things,
go
build
your
own.
A
In
fact,
Launchpad
helps
you
create
your
own
app
repos
and
tells
you
what
to
do
to
go
at
it,
and
then
people
can
add
that
to
their
own
Docs
and
I,
don't
know
how
many
times
what
I'm
gonna
move
to
I
do
the
same
thing.
I
go
add
something
now.
If
I'm
gonna
go
into
the
store
interface
that
doesn't
work
and
now
there's
a
new
thing
and
a
boon
to
called
what
is
the
new
store
thing?
You've
got
snaps
that
gets
around,
that.
A
That
goes
back
to
that
central
rather
than
decentralized
model,
and
so
this
is
kind
of
the
hole
that
the
search
ability
and
discoverability
I
think
if
there
was
a
way
to
have
a
searchable
thing
kind
of
like
what
Kubb
apps
has
done
and
even
holy.
Even
if
you
go
over
to
the
CNC
up,
landscape
landscape
got
CNC
F,
dot,
IO,
where
you
have
this
ability
to
visualize
and
find
things,
but
you
could
have
multiple
sources
submit
to
it,
and
then
people
could
maintain
their
own
repositories
to
find
stuff
right.
A
I,
think
that
could
be
an
option
and
if
somebody
says
go
install
my
own
I
mean
even
so
I'm
playing
around
with
the
service
catalog.
The
service
catalog
has
their
own
thing.
You
go
at
it
as
a
repo
and
then
you
can
install
service
catalog
stuff
right
because
they
understand
there's
this
decentralized
model.
We
can
do
that
if
we
were
able
to
up
the
discoverability
somewhere
and
people
could
have
more
that
distributed
model
where
they
manage
it
themselves.
This
comes
back
to
this
thing.
With
packages
right
packages
doesn't
metadata
repository.
A
If
we
had
something
like
that,
then
the
charts
could
be
like
the
aboon
two
ones
that
just
shipped
with
it
rather
than
everything,
and
then
people
could
and
we
could
be
slower.
We
could
be
meticulous,
we
could
be
a
pain
in
the
butt
and
then
other
people
could
submit
theirs
elsewhere,
because
the
discoverability
was
still
there
right.
E
A
E
Wait,
wait
we've
been
hosted
right,
I,
just
I
would
just
want
the
software
that
can
actually
do
the
use
case,
which
would
be
to
aggregate
things
and,
and
I
personally
would
be
fine
with
you
know
blessed
freebo's
get
lab.
You
know,
apps
code,
these
ones
that
we've
seen
that
the
charts
and
they're
relatively
mature
being
sub
repositories
where
we
provide
them
as
part
of
stable,
which
is
the
main
stable
stuff
that
comes
out
of
our
repo
plus
the
additional
Blessed
repose
and
aggregated
through
some.
You
know,
chart
Museum
or
whatever
endpoint.
E
C
From
a
UI
side
of
things,
I
think
I'd
be
happy
to
add
more
repos
to
keep
Apps
hub
and
that
would
definitely
increase
the
visibility
you
there
and
again.
It
would
I
think
it's
it's
with
monocular!
Isn't
really
a
nice
way
to
add
use
the
provided
repos
yet
so
it
would
have
to
be
done
by
by
us
hosting
that
service
we'd
have
to
manually,
go
and
configure
those
repos
that
it's
in
this
thing.
But
you
know,
if
there's
a
small
set
of
curated
once
we
can
definitely
go
and
do
that
and
what.
A
If
they
wanted
to
go
more
than
a
small
set
of
curated,
what
would
the
direction
have
to
be
for
Kubb
apps?
Well,
first
of
all,
if
it
went
beyond
that
to
a
broader
scope,
the
same
way
that
package
just
has
they're
just
all
over
the
place,
is
that
something
that
Kubb
apps
might
be
interested
in
I.
Think.
C
A
C
And
actually
that
already
exists
today
with
the
incubator
repo.
So
if
you,
if
you
go
to
any
of
the
incubator
charts
and
keep
up
sahabat,
it
shows
you
the
command
to
add
the
repository
before
installing
the
chart.
Yes,
it's
just
that
right
now,
all
those
repositories
are
indexed
manually,
configured
manually
in
the
indexer.
E
E
E
A
One
of
the
things
you
have
to
do
is
we
have
to
find
out
who
will
pay
for
hosting
all
of
that
stuff
because
chart
museum
while
it
runs
as
a
service,
it
stores
everything
in
object,
storage,
basically
right,
s3,
whatever
will
actually
have
to
find
somebody
to
pay
for
that
hosting
long-term
as
it
grows
and
grows
and
grows
and
grows
and
grows.
So
that's
probably
one
of
those
things
we'd
have
to
chase
down.
Let's
the
Kemosabe,
it's
pretty
cheap.
E
A
Yeah
I,
wouldn't
mind,
I,
wouldn't
mind
going
that
multiple
thing
way
and
then
have
a
nice
UI.
You
know
have
Kubb
apps
make
it
discoverable,
have
this
nice
hosting
option,
because
we
end
up
making
the
ability
to
store
that
stuff
far
more
scalable
than
it
is
today
and
I
like
that,
and
then,
of
course
we
could
do
things
like
for
people.
Who've
got
is
so
one
of
the
things
that
I've
been
delinquent
on
I'll.
Take
that
it's
taking
our
CI
process
and
moving
it
into
something
anybody
can
stick
in
a
containerized
workflow.
A
A
A
B
B
B
A
C
So
the
other
thing
we
may
want
to
do
is
is
pay
close
attention
to
what's
happening
with
repos
and
home
free,
because
I
think
they're
talking
about
a
built
in
push
model,
and
we
may
want
to
just
see
how
that
how
that
works
for
us,
especially
considering
this
way,
if
we
want
to
make
if
we're
gonna,
have
an
easy
way
for
people
to
to
just
add
things,
but
but
still
centrally,
haven't
so
centrally
located
borden.
We
try
to
see
if
Humphrey
builds
in
native
support
for
that
somehow
yeah.
A
So
the
push
stuff
that
we're
working
on
I
think
Josh
from
char
museums,
kind
of
spearheading
that
looking
at
the
API
for
it,
because
it's
just
like
home,
has
downloaders
right
where
a
plug-in
can
implement
s3.
Just
like
there's
the
helm,
s3
plug-in,
you
could
say
s3
colon,
slash,
slash,
and
it
knows
how
to
pop
into
the
plug-in
to
do
it
to
fetch
it.
We
were
gonna.
A
A
We
were
looking
at
chart
museum,
hopefully
to
work
out
the
default
implementation
with
chart.
Okay,
I
wrote
the
proposal
and
that's
why
the
language
is
nebulous,
because
we
wanted
to
give
chart
museum
some
time
to
work
this
out
and
to
Josh's
credit
he's
working
out
through
all
those
hard
things
like.
How
do
we
handle
authentication
and
authorization
right?
A
The
the
real
hard
part
ends
up
being
index
files
for
those
in
demo.
Files
pulling
them
down.
I
did
a
bunch
of
testing
work
and
it
turns
out
that
the
way
we
handle
things
is
the
worst
case
scenario
right.
If
you
look
at
libraries,
you
could
use
the
way
we
parse
the
file
format.
We
chosen
everything
it
is
the
as
far
as
memory
goes,
the
worst
case
scenario
here:
I
can.
G
A
I
did
a
bunch
of
analysis
on
large
files
like
I
created
a
200
Maggie
yamo
file
because
I
figured
our
central
repo
is
gonna,
get
pretty
big
right
right.
There
could
be
a
problem
eventually
so
I
said
what's
a
terrible
case
scenario,
so
at
the
home
summit.
Somebody
else
did
this
the
folks
over
at
Jay
frog
and
they
found
well
we're
gonna
run
into
problems,
and
so
I
started
doing
stuff.
In
here
you
can
see
we
did
some
benchmarking
or
I
did
some
benchmarking
and
the
top
one.
A
The
worst
case
scenario
with
memory
is
the
library
we
use.
If
you
actually
just
switch
to
the
normal
go
gamma
library
it
uses,
you
know
to
go
from
the
normal
ammo
library
to
the
one
we
use
is
like
2.2
times
the
memory
well
and-
and
if
you
just
did
everything
in
JSON
and
use
some
pointers
and
stuff
in
there,
it's
actually
one-tenth
the
memory.
The
case
we
use
today
so
if,
instead
of
an
indexed
item,
oh
we
had
an
index,
dot,
JSON
and
then
just
use
the
default.
Json
parser
in
elm
be
one-tenth.
C
So
I
I
would
quite
like
to
make
a
decision
on
this
and
have
something
go
in
I
think
there
were
two
winners
yeah
and
both
having
kind
of
different
needs.
There's
a
Vespa
provision
to
chart
which
runs
us
where
the
NFS
server
in
cluster
and
mounts
that
on
a
on
a
read/write
once
piston
volume.
So
we've
closed
on,
put
any
BS
volume
appear
on
AWS
and.
C
That,
as
a
NFS
volume
to
b2b
users
read
write
many
elsewhere
and
it
uses
the
whole
storage
past
vintner
approach
to
do
that
and
then
the
other
one
is
a
NFS
client
provisional,
which
is
similar
to
the
server,
except
that
the
the
NFS
server
is
outside
the
cluster,
and
you
would
have
to
point
to
it
from
that
client
job,
so
I
think
both
of
about
of
different
use,
cases
and
I
think
both
should
go
in
I
was
just
curious.
If
anyone
had
any
thoughts
around
this,
you.
G
C
C
C
And
then
the
last
one
is
so
for
a
while.
We've
mentioned
that
we
officially
support
and
an
N
minus
one
of
kubernetes,
which,
where
n
is
you
version
the
highest
version
supported
by
most
of
the
providers,
which
I
believe
right
now
is
still
one
night,
even
though
110
is
out,
but
when
I
was
looking
around
for
this
I
noticed
that
it's
still
not
actually
written
anywhere
in
the
repo
yeah.
A
A
And-
and
we
want
to
get
there
so
I'm
totally,
ok
with
charts
to
doing
that,
but
I
would
say
it's
unofficial
support
from
us,
the
maintainer,
so
we've
already
talked
about
being
totally
overworked
right
and
not
able
to
scale.
So
that's
that's
another
part
of
the
reason
here
what
I
would
say
with
this,
though,
is
there's
actually
one
of
the
weird
problems
here:
right
capabilities
that
API
versions
dot?
Has
you
actually
have
to
have
tiller
to
do
this?
We
don't
have
a
way
to
detect
it,
otherwise,
right
so.
A
M
plaits
gonna
have
trouble
with
this
because
it
doesn't
have
that
information
unless
you're
running
and
tiller.
This
is
one
of
the
places
where
hometown,
let's
get
a
fall
on
its
face,
and
that's
the
reason
that
we
went
to
the
git
version.
We
probably
need
to
have
a
way
to
have
a
default
for
these,
which
would
be
probably
the
latest
which
would
be
apps,
v1
and
things
like
that,
and
then
we
probably
need
a
way
to
be
able
to
pass
in
a
file
to
say
what
else
there
is.
C
C
I'm,
just
I,
if,
if
we
can
I,
would
prefer
that
we
advise
people
not
to
use
capabilities
at
all,
because
because
of
these
issues,
and
also
it
tends
to
mean,
like
you,
look
at
this
tiara,
it's
it's
kind
of
hockey
and
hard
to
maintain
and
the
question
is
you
know
when,
when
do
you,
when
you
remove
extensions,
v1
beta
1?
What
do
you
have
that
in
there
forever
yeah.
A
And
the
only
reason
we
still
even
have
that
in
kubernetes
clusters
is
because
there's
an
API
policy
that
says
betas
sticker
alphas
go
out
the
next
release,
they're
just
on
betas,
stick
around
for
two
releases
or
six
months,
and
that's
it.
But
the
problem
is:
there's
a
bug
in
kubernetes
right
now
that
doesn't
let
them
remove
old,
API
versions
so
because
that's
not
getting
fixed,
we're
not
ripping
out
the
old
betas
and
stuff
like
that
until
that
gets
fixed
and
at
some
point
they're,
just
all
going
to
be
ripped
out
at
this
point.
A
F
A
B
F
F
C
G
G
A
Increasing
the
development
time
you
have
in
kubernetes,
and
so
one
of
the
things
on
that
they've
talked
about
is
if
they
can
get
all
tests,
green
you've
already
kubernetes.
Oh
this
ever
happen.
They're
gonna
give
people
extra
time.
If
you
develop
and
push
a
code
freeze
back
we'll
see,
that's
a
big
ask:
we've
never
had
alt
ice
cream
before
normally
I'm,
so
they're
trying
to
give
people
a
carrot
but
yeah
the
windows
getting
is
getting
way
too
short.
A
A
Thanks,
okay
and
then
yeah
I
think
we
could
start
recommending
people
do
a
change
log
too,
for
major
changes
and
we
can
probably
point
to
there's
a
site
that
says:
here's
how
you
should
do.
Your
change
log
always
create
a
change
log
yeah,
and
so
we
can
point
people
to
that
for
the
recommendation
on
how
to
write
it
and.
C
And
actually,
in
one
of
the
issues,
someone
linked
to
something
else
as
well
related
to
this,
it
was
the
analytics
in
charts
issue,
and
someone
pointed
to
actually
wasn't
yeah
well
I'll.
Try
to
find
it
and
I
can't
remember
what
it
was
called.
Unfortunately,
but
it
was
some
sort
of
I
think
it
was.
Our
semantic
release
is
what
it
was
and
I
think
it.
It
was
a
way
to
build
a
change
log
from
get
commits.
They
told.
C
A
Release
I
love
it
the
image
at
the
bottom,
its
bender,
holding
up
a
sign
that
says
until
all
humans,
interesting
all
right.
Well,
I'll,
post
this
up
here
shortly
to
the
web,
to
YouTube,
where
we
record
all
this
stuff
thanks
everyone
for
coming.
Is
there
anything
else
we
should
talk
about
Adnan
before
we
go,
I
think.
G
Matt
and
might
not
the
guest
today
is
having
yeah
I'm
gonna.
Do
it
today
gather
my
listed
few
hours
great.
D
G
Didn't
they
didn't?
They
was
up
that
thing.
They've
been
made
to
deploy
well
tonight,
yeah
they're
ready,
Linux
just
ready
to
do
the
room
there.
G
That
was
not
a
big
right
one
for
the
Sun.
It
would
clear
in
as
you.