►
From YouTube: OCI Weekly Discussion - 2022-09-29
A
Hello
yeah:
we
are
going
to
be
tracking
towards
being
able
to
do
a
hybrid
meeting
for
Monday
of
kubecon,
so
nice.
C
A
A
A
A
A
I'm
kind
of
like
the
most
of
the
time
we
tend
to
be
able,
just
like
you,
know,
hey
here's
what's
going
on,
but
happy
to
be
able
to
kind
of
recalibrate
around
that.
B
A
D
That's
awesome,
I,
don't
know
we
will
be
helpful
for
people
who
are
virtual
or
yes.
A
C
A
Now
I
just
have
a
polycom
set
up,
so
people
can,
like
you
know,
call
in,
like
you
know,
Zoom
wise,
but
I'm,
perfectly
capable
being
able
to
put
in
like
audio
visual.
D
I
think,
if
you
get
I
think
fiddle
and
folks
like
that
who
Brandon
will
you
be
there
as
well.
B
A
B
A
B
A
B
D
A
B
A
I
will
send
out,
we
will
send
out,
invites
and
all
of
that
we'll
put
it
on
the
oci
Dev
calendar.
Oh.
A
B
E
A
Book
your
travel
to
be
able
to
come
in
for
a
meeting
early
on
Monday.
There.
E
Yeah
yeah,
as
long
as
it
doesn't
conflict
with
the
kubernetes
community
meeting
I.
A
Think
I'm
this
is
two
hours
I
think
you
should
be
able
to
make
this
one
work.
Hopefully,
but
I
will.
E
A
E
E
Yeah
I
mean
first
item
I'd
put
it
on.
There
is
an
update
on
what
has
happened
with
the
work
group
on
artifacts
and
follow
through
right.
We,
we
obviously
need
to
get
it
from
being
a
release
candidate
to
being
a
ga,
and
that
requires
the
you
know
applying
of
the
new
specifications
that
the
various
clients
and
and
servers
that
the
Registries
that
are,
you
know
required.
A
A
I've
got
a
space
in
here
and
then
from
there.
You
all
can
respond
on
that
email.
I
will
put
out
a
call
for
like
what.
What
do
you
want
to
talk
about
and
I've
got
that
written
down
for,
like
the
update
on
artifacts
how
to
move
to
GA
like
how
do
we
do
this
anyways
cool,
so
yeah
yeah
at
a
whole
month
to
be
able
to
plan
it's
gonna,
be
fun.
B
Let's
see
so
other
than
the
rc2
thing
we've
got
going
out.
I
know
I
wanted
to
work
on
those
pictures.
Anna.
You
have
been
working
on
those
on
your
PR
trying
to
get
those
pictures
looking
nice
with
all
the
links
going
to
the
right
place
and
then
I
came
through
and
said
actually.
E
D
B
Yeah
and
that's
why
I
went
along
with
you've
got
my
approval
for
now
and
let's
see
if
we
can
get
this
Merchant
all
throw
more
craziness
in
there
later
on,
see.
E
If
I
can
do
this
Brandon,
you
said
release
candidate
too.
We
got
some
some
kind
of
schedule
here
or
something
you're
thinking
about.
B
No
we've
already
the
vote's
already
open
for
it.
D
Just
a
bit,
Mike
I
missed
the
change
for
subject
rename.
So
what
happened
is
when
you
took
the
hash?
The
tree
did
not
include
that
one
change
so.
D
It's
just
kind
of
pulling
that
in
and
I
spoke
to
Winston
Vincent
said
just.
E
B
B
D
B
B
This
should
actually
be
an
arrow
point,
this
way
from
there
there,
but
the
reason
you
have
this
error
going
up
is
because
we've
got
the
subject
and
the
subject
from
either
the
image
or
the
artifact
can
go
to
any
manifest,
and
so
it
gets
really
awkward
to
draw
all
these
lines
all
over
the
place.
For
some
of
these
things,
the
thought
process
I've
got
is:
does
it
make
sense
to
put
a
box
above
these
boxes
of
saying?
Okay,
all
three
of
these
at
top
are
manifest.
D
B
I
got
the
feeling
this
was
more
than
this
was
like
someone
trying
to
understand
is
like
a
registry
maintainer
or
something
like
that,
how
all
these
objects
relate
to
each
other
if
they're,
making
some
kind
of
anti-relationship
database
or
something
like
that
so
I
feel
like
this
is
important
for
their
use
cases,
but
some
of
these
get
a
little
awkward
for
that.
For
example,
an
artifact
can
point
to
an
image
index
with
the
subject
field.
It
can
also
point
to
itself.
B
Yeah
each
one
of
these
things
are
saying:
hey
you
can
have
zero
one
of
these.
You
can
have
zero
one
of
these.
You
can
have
zero
one
of
these,
but
you
can't
have
one
of
all
three
and
there's
no
grouping
of
that
kind
of
stuff,
and
so
I'm
I'm
leaning
toward
actually
sitting
down
trying
to
make
a
different
picture.
I
just
figured
I
would
mention
that
that's
going
through
my
head
in
case
people
have
ways
they'd
like
to
see
things
drawn
differently.
E
B
Yeah
I,
don't
know
what
a
point
is
going
to
look
like
coming
from,
like
the
image
manifest
up
to
the
parent,
manifest
I'll
see
what
this
thing
does
with
it,
because
it's
got
they've
got
the
little
dot
language.
Whatever
this
thing
is:
there's
a
definition
for
that,
and
so
I
can
start
grouping
these
things.
I
just
don't
know
what
it
looks
like
when
I
start
drawing
arrows
from
a
subgroup
to
the
parent
group,
we'll
figure
that
out.
B
D
I
I
need
a
I
need
to
update
the
tests
for
the
artifact
schema
as
well
as
need
one
more.
D
My
understanding
was
that
those
tests
should
catch
those
kind
of
invalid
examples,
because
I
know
that
it
runs
through
the
example
files
on
the
MD
files.
Looking
for
examples,
I
just
need
to
verify
whether
once
the
schema
testing
is
those
kind
of
things
are
also
caught.
So
a
little
bit
and
there's
an
action
item
on
you.
B
E
D
The
one
positive
thing
I
like
about
that
schema
validator
is,
if
the
examples
are
broken,
it'll
actually
catch
it.
So
if
somebody's
missing
something
small
and
they
make
a
PR
and
it
runs
through
the
pr
test
also,
so
it
just
gives
a
sanity
check
in
some
way,
but
it
is
an
overhead
to
maintain
yeah
you're
right
about
that.
No.
B
And
distribution
itself
needs
to
get
updated
with
our
own
validation.
It's
not
compliance.
What
do
we
call
it
over?
There
conformance.
C
B
E
B
E
B
E
B
Yeah,
send
it
I
I
should
ask,
since
we
have
people
here
for
this,
I
was
going
through
and
cleaning
up
some
of
the
working
group
stuff.
We
already
archived
the
working
group
repo,
so
that
was
good.
That
was
I,
think
locky
knocked
that
one
out
I
started
going
through
the
oci
playground
and
I
should
probably
pull
this
up
on
a
tab.
B
Another
tab
another
day
and
this-
and
there
is
a
method
to
me,
asking
this
question
right
now.
C
B
D
So
in
the
last
call
I
think
they
we
can.
Potentially
we
were
trying
to
get
a
CLI
out
that
would
work
with
oci
distribution
here,
because
the
spec
was
not
merged
and
I
think
Monday
onwards,
with
once
the
spec
files
are
updated.
We
could
potentially,
but
I
mean
it's
fine
to
Archive.
These
I
just
need
some
time
before
coupon
to
maybe
just
keep
the
demos
around
the
distribution.
One
has
a
has
a
hacked
implementation
for
showing
that
end
to
end
with
the
New
Path,
and
things
like
that.
So
yeah.
B
Power
off
on
distribution,
I
went
to
a
bunch
of
the
other
ones,
I
don't
know
if
we
did
anything
like
a
container
registry
but
yeah
anything
people
are
still
working
on.
I
want
to
hold
off
archiving
just
in.
D
Case
yeah
so
auras
and
and
go
possibly
I'll
work
with
shui
who's
on
the
Olas
maintainers,
who
can
kind
of
like
maybe
close
this
once
he's
done
with
it,
but
I
know
he's
working
on
these
two.
D
B
D
Have
a
bunch
of
like
implementations
on
how
to
do
recursive
copy
and
things
like
that,
so
it's
basically
the
library
implementing
it
to
use
the
artifact
manifest.
The
other
thing
is
the
tagged
fallback
is
not
implemented,
so
they
need
to
get
that
one
as
well.
So
this
was
a
place
to
kind
of
start
and
just
make
the
PRS
directly
into
a
restaurant
here
so,
but,
given
that
that's
resolved
I
think
this
will
go
away
in
the
next
week
or
two
okay
I'll
take
care
of
it.
I
mean
this
is
my
headaches.
B
E
The
on
the
same
line
of
thought
here
with
orez
needing
to
move
to
the
new
specs,
the
I
wanted
to
point
out
that
I
saw
a
helm.
It
was
using
a
a
feature
that
was
in
the
oci
image
specification
to
actually
use
an
image
with
the
different
type.
You
know
Helm
media
type,
which
was
allowed
originally
in
the
image
spec
right.
D
E
It
looks
like
November
of
last
year
the
pr
went
in
so
just
just
sort
of
a
heads
up
to
you
guys
it's
not
just
the
oras
guys
that
need
to
move
up
to
the
spec.
We
also
have
some
Legacy.
You
know
abuse
like
if
you
will,
or
you
know,
overuse
so
if
the
image
spec
right
so.
E
D
D
I
kind
of
like
talk,
spoke
to
the
health
maintainers
to
just
make
sure
hey.
This
is
coming.
Let's
make
sure
that
we
start
the
conversation
on
this,
because
it's
a
major
breaking
change
for
them,
because
well,
that's
used
to
use
continuity
under
the
covers,
but
continuity
doesn't
have
the
support
for
like
arbitrary,
manifest
types
and
things
like
that,
but
either
way
it
will
be.
It
will
take
time
right,
like
moving
the
helm
Community
to
yeah.
A
C
D
E
Notice
that
yep
yep
they're
they're
using
a
you
know
an
image
manifest
with
the
overloaded
type
right.
D
Helm
is
doing
the
config
media
type
right
now.
B
B
D
Don't
know
if
I'd
ever
Norse
I
don't
know
whether
you
can
use
the
helm
CLI
with
Docker
Hub.
It
works
with
most
of
the
other
registries,
but
darkerub
is
again
a
special
case.
Maybe
they
do
they're
still
filtering
I.
Don't
know
we
just
need
to
try
out
the
home
seller,
but
this
is
a
released
version
of
Helm.
So
they've
a
huge
I
mean
they've
gone
down
that
path,
as
that's
the
only
way
to
kind
of
store
the
charts.
Besides
the
image.
B
Yeah
I
saw
a
comment
from
Justin
said
coming
very
soon
when
asked
about
getting
that
filter
taken
off
on
HUB,
so
I
think
we're
close.
That
was
two
days
ago,
and
he
doesn't
he's
been
saying
it's
coming,
but
we'll
see
hopefully
release
your
means
really
soon.
C
E
D
E
E
Mentioned
that
she
was
already
submitting
to
the
to
the
day
day,
one
a
session.
B
Mike
I
know
you
had
comments
earlier
where
I
was
saying:
hey
you
and
I
should
get
together
and
talk
about
some
of
your
thoughts
on
other
things.
You
were
thinking
about
doing
with
the
artifact
manifest
questions
you
had
on
some
emerges.
If
you
want
to
chat
about
that,
I'm
happy
to
chat
about
those
as
well.
E
E
We
have
a
lot
of
free
users
stuff
like
that
right
and,
if
you're
adding
the
poll
for
an
artifact
or
three
different
artifacts,
possibly
those
those
are
going
to
be
pretty
expensive
requests,
because,
if
you're,
a
free
user,
you
only
have
100
right
if
you're
I'm,
sorry
free
and
honest,
if
you're
a
free
user
with
Docker
Hub,
you
only
get
200
pulls.
So
if
you
wanted
to
have
five,
you
know
containers
on
five
pods
and
you've
got
these
artifacts.
You
can
sort
of
seat
the
math
right.
E
You
know
if
you're
going
to
do
a
cluster
so
without
without
having
unlimited,
you
know
full
access
to
these
hubs,
so
from
from
that
perspective
and
from
performance
perspective,
be
able
to
run
fast.
Solutions
I
was
hoping
that
we
would
have
at
least
one
pattern
where
you
could
push
with
the
index
right
or
pre-push
your
artifacts
and
then
list
them
in
the
index.
So
then,
when
the
client
goes
to
pull
it,
they
would
be
able
to
say:
oh
okay,
I've
got
in
this
index.
These
images
for
the
platforms
and
I've
got
these
artifacts.
E
E
E
You
know
for
caching
purposes,
I
know,
but
but
I
I
I
saw
that
as
a
potential
and
I
wanted
to
talk
about
that
right.
That's.
E
B
E
B
B
Yeah,
but
what
you
need
to
know
is
for
any
one
of
these
things,
so
I
pull
down
the
Manifest
for
4f
for
FFC
you're,
going
to
query
the
registry,
no
matter
what
just
using
the
API
to
say,
I've
got
for
FFC.
Give
me
all
the
refers
to
that
and
all
everything
is
kind
of
subject.
So
you
query
the
refers
interface
if.
E
D
E
D
D
E
E
E
E
C
E
Yeah,
especially
if
you're
just
going
to
sign
the
platform,
images
and-
and
you
know
and
they're
going
to
already
have
the
response
you,
you
would
know
what
all
my
all
your
artifacts
are
up
front.
If
you
wanted
to
package
it
up,
let's
say
it's
a
banking
solution
right
and
you
don't
want
third
parties
adding
s-bombs
to
it.
After
the
fact
yeah.
D
E
E
Two
two
models
here:
right:
one
model
where
it's
a
public;
it's
a
Public,
Image
right,
a
public
index
and
and
you're
allowing
for
third
parties
to
attach
right,
artifacts
to
it,
and
then
another
model
is
where
it's
a
private
repo
with
a
private
index.
And
you
don't
allow
third
parties
to
add
anything
to
it.
Right.
E
B
E
B
A
E
Third
party
reference
to
that.
Somehow
that
says
yes,
this
is
a
valid
sign
yourself,
can't
sign
yourself,
although
I
will
I
would
I
could
argue
that
when
you're
pulling
down
that
that
Index
right,
it's
got
a
Shaw,
which
is
you
know,
somewhat
secure
and
you're
pulling
it
from
a
log
on
right,
and
you
should
be
the
only
one
that
has
access
to
your
private
repo
right.
So
it
should.
It
should
be
somewhat
secure
right.
Just.
D
E
B
Yeah
part
of
the
reason
we
shot
away
from
the
stuff
Docker
was
doing
was
knowing
that
people
get
very
attached
to
those
Digest,
and
so,
if
you
have
the
digest,
Upstream
that
says,
I'm
digest
ABCD
right
and
you
pull
it
down
to
your
environment,
and
you
want
to
add
your
own
signature
to
it.
Yep
they
didn't.
People
didn't
want
to
change
that
digest.
They
wanted
to
be
able
to
change
the
metadata
without
changing
the
digest.
Yeah.
B
E
B
And
that
that
does
go
back
to
where
I
was
looking
at.
What
do
we
get
by
doing
this?
If
we
start
putting
stuff
in
that
original
index?
B
There
are
two
two
challenges
there.
One
is
that
you
can't
easily
strip
stuff
out
so
somewhere
on
the
original
Upstream
says:
here's
the
s-bomb
and
you
have
something
internal
that
says
no
we're
really
the
s-bomb
originator,
there's
no
easy
way
to
take
out
the
original
one
that
you
might
not
like
as
much.
E
C
E
E
E
B
E
Or
something
like
that,
you
got
it
yep
that
would
work
yep.
That
indicates
it's.
Of
course
you
know
a
partial
shot
right
only
on
the
contents,
not
the
actual
index
itself.
C
D
Yeah,
that's
pretty
close
to
I
believe
the
initial
like
it's
it's
an
extension
on
proposal,
F,
which
is
the
partial
digest
or
like
partially
compute
the
portion
of
the
of
the
Manifest
that
and
then
use
that
digest
to
sign
it
in
some
way.
It
requires
a
lot
of
client
tooling,
but
doable.
E
B
E
D
So
maybe
I
can
share
something
we
so
from
from
ACR
side.
I.
Think
I
think
this
came
back
from
auras,
I
believe
and
unfortunately
you
know,
ECR
microphone
is
India,
so
we
discussed
using
the
created
field
to
kind
of
do
the
Sorting,
because
the
specification
kind
of
kept
out
the
Sorting
thing
it
kind
of
gave
us
a
little
bit
of
Freedom.
So.
D
So
that
way,
the
client
does
get
some
say
on
what
needs
to
come
about.
Hopefully,
tooling.
Does
the
right
thing
I.
C
E
D
Right
and
and
I
think
the
default
model
for
the
registry
is
nobody
trusts.
The
the
content
from
the
registry
is
it's
just
a
Content
delivery
mechanism.
All
verification
happens
on
the
client.
So
even
if
it's
out
of
order,
it's
still
up
to
the
client
to
kind
of
decide
if
the
signature
is
valid
and
within
the
SKU
and
all
those
kind
of
things
so
right.
E
D
B
E
E
E
D
E
E
B
I
have
opened
many
of
issues
on
registry
maintainers
that
are
not
happy
with
me.
Thank
you.
B
E
E
E
Without
a
a
DOT
One
requirement,
yeah
anyways
all
right,
oh
we're
good,
so
yeah
that
that
was
my
main
thing
was
I
was
looking
for
a
one
connection
fits
all
I
can
pull
everything
down.
Pack,
you
know
that's
been
packaged
up
kind
of
scenario
as
well.
E
B
I,
don't
think,
there's
anything
we
can
do
from
ocisa
that
someone
can't
do
that
right.
So
someone
could
take
our
spec
and
say
yeah.
All
that
stuff
is
great.
We're
still
going
to
do
a
different
way
and
Registries
are
going
to
allow
that
because
there's
nothing
against
the
spec
from
what
you're
saying
there
yep
my
my
hesitation
on
trying
to
formalize
that
is
I
just
I'm
still
coming
back
to
what's
that
extra
value.
B
If
we
still
support
the
refers,
because
as
long
as
you're
still
supporting
your
Furs,
you've
still
got
that
extra
API
call
in
there
and
I,
don't
think
we're
getting
enough
value
add
from
packing
it
I
think
the
value.
Add
you
get
from
packing
it
is
you
say
when
you
see
a
packed
thing,
you
just
don't
even
try
the
reverse.
Well,.
E
I
mean
okay,
so
let's
talk
about
the
value,
then
right
when
I
use
the
refers
to
get
a
list
of
manifests
right,
that's
list
of
Shaws,
and
if
those
shots
are
local,
I
don't
have
to
actually
pull
those
blocks
so,
depending
on
the
size
of
the
blobs
content,
whatever
it
is
right
when
I
call
the
refers,
API
I
may
gain
Advantage.
If
I've
already
pulled
those
artifacts
down
right.
If
there's
a
another
artifact
I'll
be
able
to
weigh
the
value
of
one
that
was
pre-built
and
already
pulled
over
right.
One.
B
E
B
B
B
It's
only
one
artifact,
but
it's
a
doozy
I
pack,
all
the
HTTP
traffic
from
all
the
requester
in
the
build.
B
B
Because
because
the
the
main
value
ad
I
I
got
from
packing,
it
was
that
when
it's
packed,
you
only
have
to
pull
the
one
index
and
the
downside
of
having
to
do
the
refers.
Is
you
have
a
second
API
to
pull
a
second
index,
and
so,
if
we
still
have
to
query
refers
we're
still
making
that
second
API
call
once
we've
gotten
to
that
point
of
the
second
API
call,
they
feel
equivalent
to
me.
I.
E
You
know
if,
if
you're
going
to
run
the
you're
gonna
on
the
same
connection
where
you
pull
the
Manifest
and
and
check
your
layers,
if
that
is
also
where
you're
gonna
have
an
async
task,
that's
doing
an
HTTP
header
for
refers,
then
yeah.
It
should
be
very
similar
in
performance
on
the
same
connection.
E
Yes,
I
I
see
that
I
see
where
you're
headed
on
that
that
question,
then,
from
that
perspective,
the
only
other
question
would
be
the
trust
Factor
right
and
what
takes
priority,
the
one
that
was
you
know,
pushed
by
the
actual
owner
of
the
image
or
the
the
s-bombs
that
you
know
were
attached
later.
I'm.
Sorry,
the
the
artifacts
that
were
attached
later.
B
That
way,
well,
the
the
expensive
part
comes
in
that
when
you
pull
the
s-bomb
now
you
need
to
query
the
reapers
on
the
s-bomb
and
so
now
you're
adding
extra
hops
there
for
each
one
of
these
things
you
pull
down,
and
there
is.
There
is
a
challenge
when
you
pull
the
index
you're
going
to
pull
the
reverse
or
the
index,
and
then
you
pull
your
image
manifest
and
you're
going
to
pull
the
reverse
damage
manifest.
B
C
E
Don't
gloss
over
that
that
sounds
expensive,
I
I
and
once
you
start
going
that
direction
right,
I
think
that's!
When
people
are
going
to
come
back
and
and
bulk
a
little
bit
and
say:
can
I
just
push
all
these
artifacts.
You
know
in
one
set
of
blobs
right
later
on
we'll
see:
yeah
can
I
zip
these
things
up
in
one
one:
master
artifact
right.
B
B
B
C
E
E
B
D
I
from
when,
when
we
started
implementing
the
cache,
it
was
pretty
much
the
ocl
layout
that
we
went
with
it
did
pretty
much
everything
to
be
able
to
store
artifacts
like
next
to
it.
So
there
is
no
I
mean
from
a
root
of
that
standpoint.
No,
we
didn't
think
about
it,
but
from
a
storage
and
a
charting
perspective,
it
is
in
the
same
model
as
the
image
layout
yeah.
C
E
E
B
E
Right
I
was
more
thinking
about
sidecar
type
containers
in
a
pod,
but
but
I
can
also
see
somebody
doing
a
you
know,
system
level
container
that
might
have
some
interest
or
even
outside
of
containers.
Yeah
you
could,
you
could
see,
you
might
want
to
run
an
application
that
they
can
check.
So
what
content.
D
B
D
One
one
use
case
that
we
there's
another
project
that
we're
kind
of
like
tying
these
two
together
in
that
the
goal
is
to
possibly
load
signatures
on
as
a
sidecar
and
use
this
kind
of
an
ocl
layout,
because
they
don't
want
to
bring
signature
from
the
registry
and
those
are
private
signatures
and
the
key.
D
Also
they
don't
want
to
bring
it
so
that
for
that,
basically
it's
the
ocl
layout
that
we
chose
okay,
we
can
carve
this
up
somehow
and
when
the
part
comes
up,
it
can
be
mounted
as
we
basically
call
it
like.
The
refers.
D
The
OCR
referral
store
is
what
we
call
it
and,
of
course,
referral
stores
can
have
all
these
and
the,
but
there
is
change
on
the
runtime
or
like
in
there's
change
in
the
like
the
verify
the
signature
verifier
to
be
able
to
go
and
look
into
this
directory
in
some
way,
which
is
pretty
much
a
layout.
It's
a
cache
layout
of
the
same
format.
So
it's
worked
and
that's
why
we
I
didn't
have
to
kind
of
like
race
that
separately,
but
it
is
a
sidecar
signature.
Loading
scenario.
E
E
B
When
you
get
into
the
runtime
Specter
you're,
reminding
me
of
this
old
issue
been
a
while,
where
you're
talking
about
different
ways,
you
could
potentially
inject
metadata
into
the
runtime
of
saying
this
is,
who
you
are
and.
E
B
Well,
I
think
the
problems
we
are
solving
were
getting
run
time
level
we
weren't
we
weren't
assuming
the
container
image.
That
was
writing
you
to
know
it's
our
signature.
We
assume
that
the
thing
that
was
going
to
run
the
container
image
was
going
to
verify
the
signature
before
even
saying
to
the
runtime
right.
E
E
E
E
B
E
B
B
Right
yeah,
but
like
I
say
that's,
clients
can
do
that
today
if
they
want
to,
but
for
us
do
it
on
ocim
I'm
working
through
all
the
mental
hoops
and
Logistics
in
there
of
how's.
How
do
I
writing
a
client
interface
in
there
handle
it
when
I
see
that
and
knowing
there's
still
the
option
for
a
registry,
Implement
reverse.
B
B
A
huge
pushback
we
got
from
Docker
was:
please
don't
hit
the
tag
listing
API,
knowing
how
slow
that
was
for
all
the
stuff,
and
so
we
we
distinctly
said.
Okay,
let's
use
a
well-known
tag,
we'll
accept
the
race
conditions
that
that
generates
us
just
knowing
that
we
can
query
that
manifest
for
that
index
directly
without
having
to
go
through
the
tag
listing
and
so
I
think
there
were
some
efforts
to
try
to
make
the
performance
a
little
better
in
this,
but
real
world.
We
we
still
want
to
see
it
yeah.
E
B
D
D
Is
the
concern
on
the
number
of
connections
or
is
it
the
concern
on
the
number
of
sdb
requests?
Because
my
understanding
is
for
Docker?
You
can
still
use
the
same
connection
underneath,
but
it
will
do
multiple
requests
and
they
rate
the
rate
limit
by
the
number
of
requests
right.
Yeah.
E
B
E
B
E
E
Itself,
does
it's
a
lot
of
traffic
to
just
to
make
a
connection.