►
From YouTube: OCI Weekly Discussion - 2023-06-29
D
Got
concerns
but
more
around
if
we
GA
this,
and
so
my
comment
here
was
going
to
be
that
I'm,
okay,
releasing
an
RC
as
long
as
folks
understand
that
that
is
going
out
there,
because
we
don't
have
implementations
that
have
implemented
this,
and
so
we
need
feedback
from
implementations
to
understand.
Is
this
something
we
want
to
consider?
Ga
it's
not
something
that
we
have
run
through
tested
with
a
lot
of
implementations
and
said
this
is
the
gold
standard?
We
think
it's
going
to
go
forward.
B
C
D
There's
been
a
lot
of
pushback
Lately
from
folks
saying:
hey,
you
had
an
RC
and
then
you
changed
it
and
it's
kind
of
the
feedback
we've
been
giving
is
well.
Implementation
showed
up
and
they
did
things
we
weren't
expecting,
and
so
we
had
to
adapt
from
that,
and
so
I
want
to
make
sure
we're
saying
that
we're
cutting
RC
here
we
don't
have
find
Reglan,
just
got
it
this
past
week.
I
don't
have
anybody
to
test
against
because
nobody's
implemented
this
yet
so
well,.
C
I'll
I'll
test
your
stuff
for
sure.
I
will
do
that
and
if
you
have
an
AWS
account,
I
can
maybe
even
have
you
test
against
it,
but
yeah
I
think
that's
there's.
You
know.
We've
talked
about
this
in
the
past
and
I.
Think
honestly,
we've
had
so
much
changing
in
the
guard
and
so
many
new
interest
people
and
you
know
we've.
We
talked
about
this
way
back.
C
It's
you
know,
there's
a
bit
of
a
chicken
and
egg
thing
for
some
of
us
where
we
can't
release
without
a
spec,
and
if
we
do,
then
you
know,
and
then
you
know,
I
know
our
our
friends
at
Azure
did
release
and
then
they
got
caught
with
the
fact
that
we
made
changes.
So
you
know
either
way
it's
it's
a
little
stuck
so
yeah
I
understand
that
I
just
I.
C
Think
fundamentally,
if
we
cut
an
RC
and
then
we
test
it
and
then
we
get
toward
a
release,
it's
what
we've
been
doing
and
I
guess
since
I
have
the
mic
and
I
won't
shut
up,
at
least
for
30
more
seconds.
I
will
say
that
you
know
my
opinion
has
always
been
standards.
Aren't
implementations
so
we
have
to
yes,
we
have
to
sanity
check
it.
We
have
to
make
sure
it's
going
to
work.
I
think
with
this
particular
release.
What
we're
looking
for
is
to
make
sure
nobody
has
objections
to
the
conceptual.
C
You
know
design
implications
of
this
spec
right
of
this
minor
release.
Are
we
all
happy
with
it?
Do
we
think
it
hits?
The
right
notes.
Are
people
in
Conflict
still
about
some
of
the
features
or
how
they're
implemented
or
how
they're
specified,
and
then
that
leads
to
implementation?
C
So
I
do
think
you
can
decouple
the
release
of
a
specification
standard
and
its
implementation,
so
I
I'm
not
particularly
pressed
about
having
a
ton
of
testing
or
even
the
compliance
Suite
aligned
before
a
spec
is
released,
but
you
know
I'm
not
going
to
stand
in
the
way
of
saying
well,
let's
cut
an
RC
and
give
it
another
two
weeks
or
something:
that's
that's!
Not
a
hill
I'll
die
on.
D
Yeah
I
think
people
looking
at
RCS.
They
tend
to
see
those
coming
out
and
they
say
well
that
means
that
somebody
has
tested
and
development.
They
think
it's
good
to
go
and
they're
pushing
out,
for
you
know,
consumer
feedback
and
in
a
typical
development
cycle
that
would
be
make
sense
in
a
spec
development
cycle.
C
E
Just
to
add
more
confidence,
the
conformance
tests
about
a
month
or
two
ago
we
are
testing
against
zot's
latest
RC,
so
even
like
today,
so
anytime,
a
PR
opens
it's
running
the
conformance.
So
if
you
can
consider
the
conformance
test
as
a
client
kind
of
and
then
zada's
implementation,
so
I,
don't
think
anyone's
here
to
speak
to
that.
But
that's
one
example.
D
D
E
C
Oh
yeah,
sorry
I
forgot
my
hand
what's
up
yeah,
no
just
to
close
off
on
that
that
thread.
I
I
think
you
nailed
it
right
to
the
the
idea
that
someone
will
run
this
run
an
RC
and
know
that
it's
been
tested.
That
sounds
like
an
implementation
to
me
and
I.
Think
that
is
a
really
important
aspect
of
this.
C
Is
that
specifications
aren't
implementations
and
again
this
is
you
know,
maybe
my
my
Unix
graveyard
showing
but
like
if
you,
if
you
had
to
have
software
in
the
wild
before
you
release
the
specification,
we
wouldn't
have
specifications
that
standardize
anything
so
I
completely
appreciate.
Being
conservative
and
I
absolutely
think.
Testing
is
good.
I.
Think
Josh
brings
up
a
really
good
point.
There
is
an
implementation
to
test
and,
and
we
can
even
do
compliance.
So
you
know
if
I
had
a
vote.
C
I'd
say:
let's,
let's
go:
let's
Push
It
Forward,
you
know
that's
what
micro
releases
are
for
eign,
but
that's
just
you
know,
that's
my
opinion.
It
has
been
I
I'm
eager
to
get
specifications
released
so
that
we
can
release
the
registry
changes
against
it.
E
C
C
Trailings,
that
might
be
what
I
meant
Josh.
That
might
be
what
I
was
trying
to
say
is
that
if,
without
a
spec,
we
can't
release
SO
waiting
for
you
waiting
for
people
to
be
able
to
test
against
us
as
an
implementation
will
never
happen,
which
is
why
we're
going
into
July
and
we
still
haven't
released.
We
we
have
an
implementation
in
ECR,
just
like
Azure
does
in
ACR,
since
goodness
since
last
year.
Since
about
this
time
you
know
maybe
the
fall
of
last
year.
We
can't
release
it
so
to
Josh's
point.
C
You
know
and
I
guess
maybe
to
my
initial
point.
If
we're
conceptually
good
with
this,
if
we
spear,
if
we
feel
like
the
spirit,
animal
in
the
room
has
stopped
grumbling,
maybe
we
should
really
consider
you
know.
Is
this
1.1
right
now?
Can
we
just
cut
it,
and-
and
you
know,
and
then
yes,
there
could
be
bugs,
but
that's
what
that
is
what
patch
releases
are
for
right.
We
have
a
1.1.1
that
we
could
cut
if
we
need
to.
C
Hopefully
we
never
do,
but
but
there's
no
breaking
change
there
right
and
I
think
to
Brandon's
Point.
That's
what
you're
concerned
about
is
having
what,
if
there's
more
breaking
changes
right
and
so
I
think
I.
Think
if
we
focus
on
that
that
high
level
and
and
make
sure
that
you
know
we're
sort
of
aligned
on
what
the
content
of
this
thing
is.
I
think
that'll
be
good.
B
D
Right
yeah
to
your
question
of
Sasha,
it
was
a
bit
of
both
Josh
is
asking.
Can
we
do
an
RC
and
I
do
about
that?
I
said
I
would
be
comfortable,
giving
a
thumbs
up
to
an
RC
with
the
acknowledgment
that
we
haven't
tested
at
all
yet
and
that
putting
the
RC
out
there
is
getting
us
to
a
place.
We
can
start
testing
it.
It's
not
the
typical
release
cycle,
most
people,
think
of.
B
F
See
I
think
at
this
point
in
time.
Emit
spec
has
no
concerns
to
Katana
RC.
Tomorrow,
distribution
I
haven't
been
following
Josh
and
we
can,
and
Brandon
can
maybe
talk
about
that
post.
That
I
mean
from
earlier
conversations
like
an
RFC
period,
for
we
cut
a
we
got
a
candidate
for
a
ga
and
maybe
give
a
time
for
to
three
to
four
weeks
like
what
we
did
last
time
might
be
good
enough
for
operators
to
come
and
save.
D
G
Sort
of
naughty
Josh
I
think
you
need
to
vote
to
approve
the
pr
you
pushed.
F
I
had
a
process,
question
or
image:
spec
John,
bull,
approved
on
the
PR,
but
didn't
send
out
an
email,
but
the
process
calls
out.
You
need
an
email
response.
Should
we
update
the
process
to
say
PR
approval
or
email.
D
F
I'm,
so
it's
the
other
way
around
for
John
Bull
on
the
image
back
so.
G
F
Like
RC
dot,
one
versus
rc1,
that
seems
to
be
a
an
issue
that
came
up
and
we
had
to
reset
the
word.
I
mean
I,
reset
it
just
to
be
clear,
but
there's
no
issue
tracking
this
anywhere
there's
a
discussion,
but
it's
not
being
captured.
It's
captured
as
a
part
of
another
issue,
so
I
was
hoping
we
could
formalize
this
and
I
agree
that
for
future
RCS
we
would.
This
is
the
process
we
would
recommend
because
there's
no
release
like
Clarity
on
the
versioning
similar
for
RC's.
Yet
we.
D
D
All
right
so
otherwise
other
than
everybody's,
staring
at
me
saying
when
is
he
going
to
hit
the
approve
button
and
I'll
I'll
get
to
that?
Probably
today
on
image,
spec
and
likely
tomorrow
in
distribution,
maybe
I'd
get
that
out
earlier
I'll
give
it
a
thumbs
up.
I
just
need
to
spend
some
more
time
doing
my
final
pass
through
everything
other
than
that.
D
C
C
Got
it
yeah,
we
I
I,
think
we're
still
rolling
out,
but
I'll
check
on
that.
If
it's
not
now
it's
going
to
be
within
a
week
or
so
we'll
have
a
you
know
a
fully
up-to-date
Festival
environment.
D
And
I
haven't
looked
at
distribution
distribution,
I
know
they
had
had
something
in
progress
and
they
said
something
on
their
PR
a
while
back
that
said,
hey
we're
gonna
have
to
restart
this
one
because
it
changes
from
image.
Spec
and
everybody
started
crying
earlier.
Github
repos
so
sadly
feel
bad
for
that
side
and
then,
like
I,
say
I've
got
on
Richland
I
was
trying
to
look
for
comparable
commands
and
or
us
cosine
crane.
D
That's
what
I
was
looking
for
was
a
copy
that
pulled
in
not
just
the
Manifest
and
all
the
child
manifests
and
the
blobs
to
do
like
a
full
copy
of
an
image
one
repository
to
another,
but
pulling
in
all
the
referrers
to
each
of
those
manifests.
And
then
any
content
from
those
refers.
Recursely
gone
through.
F
Yeah
I
know
I
know
that
a
lot
of
support
for
that,
but
not
the
index,
one
right.
That's
the
one
that
they're
still
working
on
okay,
it's
kind
of
like
a
chicken
egg
problem,
because
without
getting
RC,
they
don't
have
a
library
that
they
can
pin
to
the
laughter
pin
domain.
So
that
was
the
other
like
we
can
still
try
it
out
and
just
burden
of
the
library
itself.
D
Thank
you,
parent
copy,
okay,
I've
added
stuff
to
the
docker.
One
people
probably
won't
agree
with
that.
Putting
annotations
over
there
so
I'm
sure
I
made
somebody
upset
with
that
one
and
then
I'll
pull
in
things
from
and
respect
early
just
to
do
local
test.
D
I
think,
given
the
checks,
do
we
have
enough
majority?
Do
you
even
need
me
for.
F
G
D
Got
another
couple
hours
I
think
here,
so
there
you
go,
I
will
I'll
get
in
respect
on
here
soon.
D
They're
they're
actual
PRS
and
stuff
going
in
for
me
on
this.
One
I
dropped
a
couple
of
them
in
there
today.
D
G
D
Expect
we've
got
a
we've,
got
a
link
over
to
the
original
label
schema
and
our
annotations
for
how
we
mapped
the
annotations
from
label
schema
to
the
oci,
annotations
and
label.
Schema
has
not
been
maintained
over
the
years,
so
they
are
kind
of
the
repo
while
back
their
certificate.
Tls
certificate
on
the
https
site
expired
a
long
time
ago,
and
so
things
like
that,
you
need
to
clean
up
all
right.
E
D
E
G
I
G
Group,
the
where
the
windows
team
has
an
HD
show
they
do
run.
C
supports
differently,
building
things
in
into
one
ship
and
the
runtime
spec
doesn't
really
have
any
guidelines
around
gems.
You
know
processes
that
would
host
a
run
C.
G
You
know
containers,
but
yet
there's
some
spec
API
guidelines
for
testing,
but
it's
only
for
testing
the
runtime
spec
itself
only
has
operations
no
actual
apis
in
the
runtime
spec
right,
but
yet
people
working
on
building
their
own
version
run
say
wanting
it
to
work
inside
of
container
D
for
certain
activities
or
certain
types
of
VMS
on
certain
platforms.
They're
asking.
What
do
we
do
right?
How
do
we
make
sure
that
our
tools
run?
You
know
it
was
runtime
spec
compliant.
My
I
pointed
to
run
C
and
you
know,
but
then
they
said
well.
G
J
I
think
that's
a
common
point
of
confusion
because,
from
the
perspective
of
container
D,
the
shim
is
the
runtime
run.
C.
The
binary
is
an
implementation
detail.
It's
an
internal
implementation
detail,
it
could
be
there.
It
couldn't
be
so
like
Microsoft
is
actually
doing
the
idiomatic
thing
with
run
HCS,
where
the
Run
HCS
shim
builds
the
entirety
of
HCS
shim
into
the
binary.
It's
all
single
process,
that's
what
makes
sense
on
their
platform
and
I.
J
Think
that's
just
as
kosher,
but
maybe
where
things
get
confusing
is
the
fact
that
lots
of
people
don't
want
to
re-implement
all
of
the
Machinery.
That's
actually
needed
to
be
a
be
a
complete
oci
runtime
like
managing
the
I
o
streams
and
handling
reattach
events
and
whatnot,
and
so
instead
they
have
reused
the
Run
C
interface,
where
that
is
the
combination
of
command
line,
flags
and
environment
variables
that
the
shim
uses
to
coordinate
the
Run
C
binary,
so
I
think.
J
Maybe
the
confusion
is
that
we
have
this
proliferation
of
runtimes
like
yukai
c-run
sysbox,
that
all
are
run
C
compatible
and
they
reuse
the
Run
C
shim,
even
if
I
think
C
run,
has
done
the
right
thing
lately,
because
they
have
a
bit
of
a
bit
of
code
entry
to
build
their
own
shim.
That
swaps
run
C
for
C
rod
and
then
names
the
binary
differently,
which
is.
B
G
G
Well,
it's
an
interesting
problem.
You
know
how
do
we
do
we
want
to
as
a
specification
organization
start
to
create
Gem
apis,
and
on
top
of
that,
I
can
tell
you
the
container
D
V2
shim
is
going
to
be
replaced.
Okay
at
some
point
in
time
as
we
as
we
get
the
new
Sandbox
for
support
that
we're,
adding
in
which
is
going
to
be
a
much
more
enhanced
shim,
but
but
yeah
still
still
shown
in
itself
in
an
API.
J
I
mean
I,
guess
I
would
Advocate
I
mean
this
is
from
the
from
the
Moby
side.
But
having
had
to
do
decent
amount
of
thinking
about
this
lately,
I
mean
I
would
Advocate
to
keep
it
in
internal
implementation
detail
and
if
anything,
the
issue
is.
B
J
Specification
ends
and
I
think
the
issue
is
more
that
it
probably
some
work
should
be
looked
at
actually
taking
the
Run
C
sham
and
factoring
it
out
into
a
generic
library
for
go
library
for
creating
your
own
oci.
Runtime
is
probably
more
what
people
want
and,
and
we
see
that
happening
with
yukai,
actually,
where
yukai
now
has
like
a
live.
Something
I,
don't
remember
what
it's
called
it
might
be
like
live
runtime
or
something.
I
B
J
Like
naming
naming
choices
aside,
yukai
is
kind
of
turning
into
a
generic
toolkit
and
library
for
building
your
own
container
runtime.
Where
what
happens
on
the
other
side
of
the
you
know,
what
we
think
of
at
container
D
is
the
shim
boundary.
Is
your
own
code
without
having
to
jump
through
hoops
to
be
compatible
with
the
Run
C
shim,
and
that
probably
is
really
what
people
are
asking
for
on
the
go
side
when
they're
trying
to
build
their
own
runtime
yeah.
I
J
I
The
okay,
so
this
is
really
all
a
considerity
discussion,
but
there
is
like
the
Run
C
sham
is
built
on
top
of
library
that
anybody
can
use
that
is
generic.
The
runzi
stuff
has
Linux
specific
stuff
because,
like
it
needs
to
do
Process,
Management
and
all
a
bunch
of
Linux
specific
things,
that's
kind
of
outside,
of
like
baking
in
multiple
different.
I
G
B
G
G
H
G
What
does
this
organization
want
to
do
with
the
runtime
spec
and
it's
conformance
tools
for
runtimes
or
maybe
the
answer
is
just
let
bio
and
container
D
and
other
container
runtimes
go
off
and
work
on
these
show
apis
and
maybe
we'll
come
to
an
agreement
at
some
point
in
time.
I
think
it's
important
because
of
the
the
changes
that
are
going
to
have
have
to
happen
in
kublet
around
the
Pod
spec.
G
The
whole
concept
of
you
know
a
set
of
containers
being
executed
by
a
runtime,
and
how
do
you
do
that
and
being
compatible
across
the
various
container?
Runtimes
right,
I
think
that's
important.
It's.
What
oci
was
put
together
for
to
try
to
make
sure
that
containers
would
be
compatible
across
container
runtimes
and
now
we
need
to
move
but
I
think
towards
a
more
of
a
pod
view
of
that
kind
of
API
going
forward
and
not
just
pause.
Also,
these
other
types
of
runtimes.
D
I'm
not
complaining
I
was.
My
comment
was
originally
tortati
because
he
was
throwing
a
comment
there
in
the
in
the
chat
saying
he
had
some
derail
conversations
going,
but
derails
are
good.
A
I
I
just
posted
them
there,
so
I,
don't
know
whether
we
want
to
discuss
them
or
not,
but
it's
mainly
about
authentication
authorization
and
the
standardizing
on
the
response
codes
for
the
distribution
apis.
That's
something
that
we
figured
out
that
we
get
responses.
That
has
nothing
to
do
with.
Actually
what
the
error
is.
G
B
G
D
G
D
B
G
G
A
F
I
D
We've
got
a
handful
of
stakeholders
in
there
I
believe
I've
captured
the
names.
I
have
not
been
pushing
this
myself,
just
knowing
my
bandwidth
with
everything
else
going
on
with
these
releases
coming
out
it.
It
may
very
well
be
something
as
soon
as
we
get
the
1.1
out.
We
come
back
here
and
we're
like
okay,
we're
done
with
that.
What's
the
new
thing,
and
then
we
put
this
front
and
center.
D
In
addition
to
this,
the
other
thing
I
know
that
people
have
been
asking
for
is
where
we
had
pulled
out
the
artifact
manifest
a
while
back
the
thoughts
of
whatever
might
replace
that
in
the
future.
There
is
the
artifacts
working
group
that's
being
done
over
on
the
cncf.
What
is
that
the
tag,
application
or
something
like.
D
Tag
apps
orders
today,
runtime
I,
think
it's
apps,
so
they've
got
their
working
group
that
they're
working
on
getting
up
and
running
I,
think
they're
still
in
the
whole
governance
stage,
or
something
in
there
right
now
or
coming
up
with
their
application.
For
that,
but
I've
pointed
them
over
here
a
few
times
and
they've
said
they're
thinking
about
do
they
need
to
find
a
manifest
I'm
like
that
sounds
like
something
we
need
to
work
with
oci
on
and
not
do
the
server
on
the
cncf
side.
D
So
much
so
there'll
be
hopefully
some
collaboration
between
them
for
some
of
those
as
well.
A
Yeah
another
derailing
item.
Actually,
we
are
mostly
interested
in
kind
of
the
search
and
filtering
from
this
working
group.
There.
A
couple
of
things
that
we
kind
of
discovered
is
really
hard
to
filter
information.
So
we
need
to
download
a
lot
of
things
on
the
client
side
filter
it
there
that
impacts
performance.
A
There
are
things
that
actually
it's
hard
to
query
the
server
and
tell
the
server
I
just
need
this
thing,
like
don't
send
me
all
the
rest
of
the
garbage
that
was
kind
of
our
main
interest
there
in
that
working
groups.
I,
don't
know
whether
there
is
a
appetite
also
to
discuss
this,
or
should
we
kind
of
being
the
other
working
group.
I
know
that
I'm
monitoring
what's
happening
there.
I
cannot
attend
the
beatings
and
I'm
getting
planned.
D
They
want
to
start
going
out
to
talk
to
all
the
people
are
putting
things
out
there
Beyond
just
the
stuff
we
think
about
going
in
container
Registries
today,
they're
looking
at
any
package
repository,
npm
Debbie
and
pick
anyways
building
packages
out
there
and
how
they
put
this
into
an
oci
registry.
They've
got
some
grandiose,
Visions.
G
D
Yeah,
but
it
makes
sense
on
the
oci
side,
one
of
the
things
that's
been
on
my
back
burner
for
a
long
while
to
really
flesh
out
the
tag
response
tag,
listing
response
to
have
a
lot
more
than
just
tags,
but
to
pull
back
full
descriptors
untag,
manifest
whatever
content
is
in
there,
so
that
people
make
judgments
and
then
maybe
add
some
more
capabilities
onto
that.
That
says,
don't
give
me
all
tags,
we
give
me
tags
matching
these
criteria
might
be
something
you're.
Looking
for.
A
Yeah
that
also
can
can
serve
some
purposes.
One
interesting
thing
that
you
mentioned
is
untucked
stuff
right
now,
especially
with
how
much
content
will
be
pushing
retargeting
and
so
on.
A
lot
of
things
will
be
kind
of
just
referenced
by
digests.
How
do
we
find
those
things.
D
So
we're
not
just
accepting
I
think
it's
just
json's
comeback
right
now,
so
application
Json
will
be
the
normal
tag
listing
except
header,
but
to
say
in
addition
to
that,
to
more
more
functional
response
that
we
would
be
looking
for.
So
it
would
have
a
different
accept
header
in
there
and
then
the
server
would
know
I'm
getting
this
request
from
a
client.
It
accepts
a
whole
lot
more
than
what
the
or
it
accepts
the
new
format
and
so
I'll
push
the
new
format
to
the
new
client.
D
D
D
D
D
H
D
Today,
people
use
it
today
already
it's
just
it's
like
how
we
add
extra
Fields
into
our
manifest.
Today
you
can
add
an
extra
field
into
Json
you're,
just
looking
for
the
one
bit,
you
know,
of
course,
so
the
reason
I
bring
that
up
is
that
the
tag
listing
is
a
much
smaller
subset
of
what
they're
looking
for
doing
over
an
artifacts
working
group.
A
The
most
frequent
kind
of
use
case
that
we
get
for
anything
that
we
are
doing
is
like
give
me
the
latest
from
this
artifact
type,
because
as
we,
you
remember,
we
discussed
in
the
past
that
actually
we
do
not
override
the
tax
we
just
and
then
we
don't
delete
old
ones
and
because
now
you
go
and
actually
update
the
content
on
certain
intervals
or
when
something
happens,
and
you
don't
care
what
is
actually
added
us
before
that,
then
you
want
to
go
and
say,
look
give
me
the
latest.
A
G
List
when
we
talked
about
doing
a
pub
sub
API
Todd,
basically
you
would
hey.
This
is
what
I
want.
This
is
the
namespace
I
I
need
it
for,
and
then
you
would.
You
know-
and
you
could
say
this
is
the
last
time
I
cached.
This
information
in
the
API
would
return
back
contacts
to
get
you
up
to
right,
set.
A
G
B
A
G
Yeah,
that's
sort
of
the
point
here
right
yeah
we
currently
we
only
have
push-pull
apis.
We
don't
have
an
API
for
Hub
content.
If
you
will
right-
and
this
is
jumping
in
that
space-
we're,
for
example,
Docker
Hub
is
team,
and
so
on.
Yeah
Docker
currently
has
an
API
for
logging
on,
and
you
know
querying
this
confirmative
information
for
your
your
particular
repo
as
well
as
them
off,
and
we
just
haven't
touched
that
API.
This
is
the
doctrine
of
API.
G
A
A
Don't
care
about
any
image
now
I
need
to
know
what
kind
of
artifact
was
pushed
because
I'm,
not
a
vulnerability
scanner
I'm
something
else
and
I
need
to
act
on
I,
don't
know
baseball
right,
I
need
to
check
the,
for
example,
the
licenses
For
Less
bomb
and
and
so
on.
Yeah.
G
B
G
G
G
G
B
B
B
G
A
Do
you,
when
you
actually
authenticate
with
this
the
response
says:
oh
you
don't
have
this
permissions
on
this
Repository
right
or
right
majority
say
that,
but
it's
not
in
a
standard
way
because,
like
especially
getting
the
catalog
is,
is
a
nightmare
right.
If
you
try
to
get
the
catalog
for
Docker
hub
for
ghcr
and
like
ACR,
it's
completely
different
approaches.
G
D
I
I
will
say
to
your
original
request:
hottie
you're,
mentioning
people
just
want
to
know
the
latest
and
I
was
watching
your
EU
talk
a
while
back
and
saw
your
was
saying,
give
me
the
most
recent
version
of
yes
bomb
or
something
like
that,
and
so
I
was
there
doing
that
client
side
on
the
tooling
the
year?
I
don't
know
if
that
was
like
sneaker
someone
else
that
had
that
whoever
you
were
whoever's
tool
you're
using
is
I
was
like
yeah
I
need
to
do
that
as
well
with
my
size,
but.
D
A
Yeah
it
it
works
for
demo,
I,
don't
know
how
that
will
perform
in
actually
AKs
cluster
that
needs
to
spin
up,
like
I,
don't
know:
10
000
pots
with
some
validation
and
if
you
have
like
even
30
of
those,
so
you
cannot
wait
30
seconds
to
to
pull
everything
on
the
each
individual
note
and
and
pass
it
the
client
side.
So.
D
Well,
the
reverse
response
is:
it
should
just
be
the
difference
on
your
first
response
between
getting
a
list
that
has
20
refers
or
one
refer,
but
that's
Json.
It's
like
20
layers
or
one
layer,
you're
not
pulling
I
mean
layers,
you'll
pull
all
the
layers,
but
in
an
index
only
pull
the
one
that
you
want
to
get
out
there,
and
so
you,
you
should
have
the
data
you
need
to
know
which
one
you
want
to
pull
from
those
annotations.
D
A
The
thing
is:
it's:
it's
not
deterministic
right,
I,
understand
that,
but
the
more
content
you
push
actually
the
longer
it
will
take.
So
if
you
have
little
content,
then
it
will
go,
but
we
are
starting
to
hit
like
some
timeouts
like
if
you
push
like
500
referrals,
for
whatever
reason
so
we're
trying
to
see
what
the
change
is
in
like
yeah.
If
I
have
10
referrals,
I'll
get
it
within
like
second
or
two.
A
If
I
have
500,
then
I'll
go
into
seven
eight
ten
seconds,
and
then
you
get
all
the
timeouts
there,
but
it's
not
deterministic
and
it's
not.
You
cannot
go
and
put
an
SLA
to
say
that
will
execute
in
this
amount
of
time,
and
we
don't
know
how
much
people
will
put
now.
We
are
telling
them
go.
Put
everything
in
the
registry
right
referral
types
you
can
go
and
write
them.
They'll
start
dumping
all
kind
of
stuff
there
right.
D
D
Don't
let
you
get
access
to
that
delete,
API,
Docker
out
being
a
big
example
out
there
that
I've
I've
hit
on
them
a
bunch
politely
in
the
issues
out
there
and
they've
turned
down
so
far
every
time,
but
they've
got
the
docker
specific
delete
API,
but
they
don't
Implement
their
CI
delete
API,
and
so
it's
it's
one
of
those
that
I've
seen
a
lot
of
Registries.
They
just
don't
don't
give
you
that
access.
D
The
when
we
do
it's
part
of
the
spec,
but
it's
just
not
allowed
from
access
rules
and
whatnot
or
Registries,
won't
implement
the
full
spec.
D
B
A
G
A
B
G
A
I
think
we
are
recognizing
that
actually,
we
should
have
some
delete
functionalities
because
there
are
other
problems
with
that.
Nobody,
it's
everybody,
is
scared
to
delete
images
from
the
registries
like
I
was
talking
with
some
of
like
the
teams
around
and
they
say:
oh,
we
don't
know
who
is
using
it
outside.
So
if
we
deleted
some
customer
will
yell
at
us
well,
this
is
not
supported
for
years
right.
Why
are
they
still
using
it?
There's.
A
D
I
mean
you're
talking
about
all
the
images.
I
was
just
chatting
there
saying
that
I
keep
going
out
there
pulling
the
Debian
sex
image.
I,
don't
use
it,
but
I
keep
pulling
it
because
it's
the
one
thing
I
can
remember
off
top
of
my
head.
That's
got
this
game
of
E1
and
so
I'm.
Just
constantly
testing
all
my
card
to
make
sure
schema,
V1
still
works
and
I'm
sure
docker's
like
who
on
Earth,
is
still
using
this
old
image.