►
From YouTube: OCI Weekly Discussion - 2022-01-13
Description
agenda/notes: https://hackmd.io/El8Dd2xrTlCaCG59ns5cwg?view#January-13-2022
A
It's
crickets
all
around
now
it's
fine!
This
happens
like
not,
everybody
is
like
getting
rolling
yet
and
it
sounds
like
the
reference
working
group
is
going
pretty
well
so.
C
A
C
A
D
Figure
I'll
give
people
a
couple
minutes
to
show
up
on
the
one
I
threw
out
there,
the
pr
880
over
an
image
spec,
it's
it's
missing
out
there.
I
think
there
was
some
agreement
that
we
liked
it.
It
was
just
hung
up
on
a
broken
ci
job,
and
I
think
that
wasn't
anything
to
do
with
this
pr.
It
was
just
ci
itself.
F
Vincent
and
I
and
john
johnson
went
through
and
fixed
a
bunch
of
things,
so
if
you
have
prs
that
are
waiting,
it's
worthwhile
to
go
and
rebase
them,
I
also
replaced
esc,
which
is
a
great
project
and
cool.
I
replaced
it
with
go
embed
to
get
rid
of
a
dependency,
but
we
didn't
merge
that
one.
F
Oh
yeah,
you
already!
You
already
know
about
this
one
brandon,
so
I
think
you
commented
on
it.
It
was
a
while
back,
but
yes
yeah.
I
should
be
running
so
and
that's
my
ci
report.
F
F
F
F
D
D
Yeah,
I
believe
so
yeah.
There
was
a
little
bit
of
back
and
forth
with
vincent
on
whether
we
wanted
to
say
the
runtime
requirements
was
like
a
platform
requirements.
D
B
G
D
Mine
was
880
handling
multiple
index
entries,
and
you
would
comment
in
there
that
you
wanted
to
add
a
little
clarification
that
there'll
be
a
platform
requirement
on
the
platform
field
and
I
was
intentionally
leaving
what
the
requirements
were
out
of
the
pr,
because
I
figured
some
people
would
have
requirements
on
maybe
an
annotation
or
something
like
that
for
their
one
specific
use
case.
I
didn't
want
to
lock
us
in.
G
Okay:
okay:
I've
got
that.
D
The
thought
process
in
there
you
might
have
some
runtimes
that
look
at
an
index
of
a
whole
bunch
of
images
out
there,
and
one
runtime
might
know
that
it's
able
to
pick
a
later
image
in
that
index
list
or
later
manifest
for
whatever
it's
doing,
and
so
it
may
not
look
at
the
platform.
Specifically,
you
might
look
at
it
and
be
able
to
figure
out
hey.
This
is
some
other
field
that
I
was
looking
for.
G
Yeah;
okay,
thank
you
for
that.
I'm
sorry
for
waiting
for
my
brain
comes
up
to
speed.
Okay,
so
yeah,
sorry,
sergey
the
the.
G
G
That
figuring
out,
you
know
a
way
forward
on
whether
things
are
extensions
or
whether
they're
just
updates
to
the
api
itself,
because
I
think
that's
one
of
the
the
the
last
last
time
we
were
like
really
processing
it
and
even
comparing
other
works
like
I
was
pointed
to
the
fact
that,
like
openshift
even
has
completely
like
broken
out
their
own
api
to
work
around
some
of
these
features,
rather
than
making
them
part
of
the
registry
themselves,
so
they
can
have
a
route
handler
and
that's
it's
effectively.
G
People
are
already
like
now
using
that
in
production
of
like
just
making
their
own
routes.
This
is
the
same
thing
that
I
think
are
kind
of
asking
about
with
features
with
acr
and
otherwise
like
do
you
just
make
your
own
route
completely
and
iron
it
out,
or
can
we
iterate
on
it
inside
the
oci,
like
within
the
registry
api
itself,
because
at
which
point?
Is
it
just
an
extension
or
like
an
actual
update
to
the
api?
The
distribution
api?
G
F
Yeah
and
all
that
comment
was,
as
I
took,
your
feedback
vincent
and
the
other
feedback
we
heard
in
this
group
and
summarized
it
into
that
comment,
because
there
was
a
couple
of
like
inconsistencies
that
I
think
in
when
asked
about
one
specific
example.
We
said
one
way
and
then
give
it
another
example.
We
said
the
other
way,
and
so
all
this
comment
did
was
look
at
everything
all
together
and
say:
here's
how
we
can
make
this
proposal
consistent
as
an
extension
proposal.
So
hopefully
we
can
take
the
spec
applied
like
rebase.
F
This
comment
on
top
of
the
spec
and
then
we
should
be
in
in
pretty
much
agreement.
What
I
did
want
to
get
get
some
gauge
on.
Did
I
miss
anything
in
that
comment?
Were
there
still
like
concerns
when
we
see
it
all
together,
vincent?
D
Yeah,
I
was
gonna,
give
vince
a
second
to
load
his
brain
there.
The
the
only
thing
I
was
thinking
of
was
that
we
were
changing
it
from
namespace
component
to
namespace
component
operation
and
makes
sense
I'm
good
either
way
on
that
one.
My
original
thought
process
was
probably
leaning
more
towards
most.
D
This
is
going
to
be
like
http
level,
and
so
it
would
be
a
git
to
put
a
delete
or
something
like
that,
and
that
was
how
you're
going
to
define
whether
or
not
you're
creating
an
object
deleting
an
object
listing
an
object,
something
like
that,
but
if
it
makes
more
sense
to
an
operation
there,
because
we're
going
to
have
more
use
cases
that
can't
easily
be
handled
by
http.
That's
fine
as
well.
C
F
D
F
D
Back
and
forth,
even
we
go
back
and
forth
even
within
those
ci
spec
in
other
places
like
if
you
do
a
tag,
delete
you're
actually
going
down
and
putting
the
tag
name
in
there
and
doing
delete
operation.
But
you
can
do
a
tag
list
and
that's
slash
list
after
it.
So
we're
inconsistent,
even
amongst
ourselves.
Normally.
F
I
was
because
the
reason
we
did
tag
list
was
to
like
a
fix
like
the
like
name,
because
you
can
have
any
arbitrary
name
in
between
there
and
if
you
have
tag
list
on
the
end
it
it
fixes
the
like
parsing,
but
you
can
just
do
it
with
a
tag,
and
that
was
a
we
just
never
got
that
change
in
so
don't
use
tags
list
as
an
example.
That
was
a
poor
one
operation,
I
agree,
is
a
bad
word.
Functionality
would
be,
is
kind
of
what
I'm
going
for.
F
That's,
that's,
not
a
very
good
word
either,
but
it's
like
you
have
some
component
and
then
some
sub
some
collected.
F
F
F
F
Manifest
compatible
types
and
then
list
those
right.
You
could
say,
like
here's,
all
the
compatible
types
for
this
manifest.
C
G
Oh
no.
What
no,
I
don't
know
about
to
make
a
comment,
but
it
has
nothing
to
do
with
what
you're
talking
about.
G
I
think
I
think
my
brain
is
still
just
trying
to
part
process
like
making
sure
since,
since
we
don't
do
much
with
like
acl,
speak
otherwise
making
sure
that
whatever
these
paths
are
that
are
kind
of
like
embedded
and
discoverable,
whether
it's
registry
level
or
repository
level,
that
not
too
much
is
like
re-created
so
like.
If
you
have
some
extension,
that's
dealing
with
the
digests
or
you
know
whatever
it
is,
manifests
or
digests
or
blobs,
or
something
like
that.
Those
paths
aren't
just
being
completely
recreated
again,
I
don't.
G
I
don't
want
too
much
I'd
hate
to
see
too
much
redundancy
forced
upon
registry
implementers,
because
now
something
is
nested,
but
as
long
as
there's
enough
in
place
that
they
can
apply
any
acls
that
are
not
really
specified
how
to
imply
how
to
how
to
apply
them.
But
that's
that's
really!
The
conversation
comes
down
to
like
sure
you
could
mess
up
the
apis.
However,
you
want
to
that.
The
spec
allows
you
to
mess
around
with
the
api.
G
G
D
Yeah,
the
big
the
biggest
key
for
me
is
the
underscore
as
long
as
we
specify
that
I
think
that
clarifies
a
lot
on
the
the
acl
part,
though,
since
this
is
only
gonna
be
used
by
newer
clients.
Newer
tools
are
knowing
they're
gonna,
be
looking
for
their
stuff.
D
If
anything
changed
in
the
auth,
that
would
be
able
to
be
handled
because
you've
got
a
newer
component.
That
would
know
how
to
do
its
newer.
Auth
does
its
thing
I,
where
I'm
going
with,
that,
is
right.
Now
we've
got
git
input
as
the
two
scopes.
You
can
request
right
now
on
pushing
and
pulling
manifest
and
blob
stuff
like
that,
if
they
want
to
add
additional
scopes
within
this
extension
area
for
doing
other
operations.
G
But
so
I
get
that
and
that's
fine,
so
if
we
venture
down
the
path
of
just
saying
like
if
you
see
an
underscore
something
you
know
the
beginning,
beginning
word
object
like
the
first
character
is
an
underscore,
then
assume
this
is
some
kind
of
extension,
then
figure
out
who
who
owns
this
extension
and
that's
that
was
most
of
the
original
original
thought
for
this.
This
pr
was.
G
To
discover
if
there
is
an
extension
and
where
the
docs
about
it
or
who
owns
it
like,
where
can
I
learn
more
about
what
the
paths
I
should
expect
on
this
art
so
like
before,
getting
into
like
the
whole
namespace
component
operation,
or
you
know
like
that
kind
of
pattern
that
like
who
owns
it?
Where
do
I
find
information
about
it?
So,
if
you
encounter
a
path,
that's
a
underscore
something.
G
Then
what
what
can
I
expect
to
tell
me
how
to
interact
with
it
or
if
I
just
know
what
I'm
doing
when
I
see
underscore
oci
blah
blah
and
it
must
be
documented
somewhere?
That's
so
it
because
when
it
comes
down
to
it,
there's
no
need
to
actually
have
extensions,
because
you're
allowed
to
mess
with
the
api
already
or
you're
allowed
to,
like
add
stuff,
that's
not
in
the
specs.
G
D
C
People
try
to
write
something
they're
trying
to
look
at
for
existing.
It's
like
stack
overflow
is
so
prevent
relevant,
like
how
other
people
do
this
couple
taste.
So
we
can
give
them
that
pattern.
Then
it
just
makes
them
more
productive
and
to
brandon's
point
it
gives
them.
You
know
we
don't
collaborate
on
each
other
and
you
point
about
if
I
see
an
extension
somehow,
because
I'm
I'm
looking
at
some,
you
know
network
traffic
like
oh,
that's,
not
in
the
oci
spec.
Where
would
I
find
that?
That's
a
valid
question?
C
D
Pattern,
do
we
as
oci
even
really
care
too
much
about
component
operation,
that
sort
of
stuff,
or
is
it
more
once
it's
gone
to
the
underscore
name,
space
and
we've
said
okay
that
belongs
to
such
and
such
they
can
kind
of
do
what
they
want
to
here's
general
guidance
on
how
he
typically
thinks
people
should
extend
it,
but
you
know
once
you're
under
that
underscore
area.
That's
your
name
space!
If
you
want
to
do
something
crazy,
do
something
crazy!
That's
your
problem!.
B
Yeah
I
mean
the
way
I
read.
The
comment
is
underscore
oci
ext
discover
is
a
guidance
and
the
default
extension
for
discovery
is
implemented
in
that
way.
Whether
we
force
other
extension
authors
to
implement
that
ways,
I
don't
think
that's
what
we
are
saying,
but
this
is
the
guidance
is
how
the
comment
is
written
at
this
point.
B
B
So
we
know
that
underscore
ci
is
where
things
might
be
or
if
the
specification
drives
it
doesn't
affect
any
of
these
and
it
doesn't
clobber
any
custom
implementation
from
other
operators.
That's
that's
the
motivation
of
at
least
kind
of
parking
and
saying
that
underscore
something
is
how
you
can
evolve,
but
you
can
find
other
extensions
if
you
want
to.
F
Everybody
really
wanted
the
catalog
when
we
implemented
it,
the
the
yeah
I
mean
the
catalog
is
obviously
a
namespace
that
doesn't
follow
this
we're
gonna
have
to
reserve
it.
So
I
mean,
I
guess,
summarizing
the
concern
here.
The
concern
here
is,
we
don't
have
a
registry
for
the
name
space
and
that
name
that
different
extensions
could
cover
each
other
and
be
mutually
incompatible.
D
I
I
got
the
feeling
we
are
doing
that
here
within
this
pr
saying
that
if
you
make
your
own
name
space,
you
need
to
come
here
and
reserve
it
with
us
just
to
keep
it
from
getting
clobbered
by
anybody
else.
F
Well,
oh
yeah
yeah!
I
I
guess,
as
the
pr
is
currently
yes,
we
can
absolutely
do
that.
I
mean,
I
think
I
think
just
like
hey.
If
you
don't
want
it
to
clobber,
you
better
try
to
register
it.
If
you
we
need
to
have
some
rules
about
that,
so
people
can't
name
a
name
squat
like
when
I
register
underscore
like
steve's
shady
crypto
extension
right.
F
It
should
have
a
plug,
it
should
have
some
context,
maybe
a
project
and
a
in
a
contact,
and
if
you
don't
register
it,
you
just
risk
being
mutually
compatible
in
the
future,
and
it's
not
like
a
requirement
right.
It's
like
a
here's,
your
risk.
That's
that's!.
G
That
that's
what
I
was
hoping
for
is,
like
you
know,
yeah
you
can
already.
You
can
already
mess
around
with
the
ap
api
or
add
on
to
it
and
that's
the
whole
point
of
the
compliance
check
is
like
cool.
You
mean
minimum
compliance.
Anything
else
you
did
that
hitting
in
the
spec.
I
don't
actually
care
about.
G
You
know
that's
a
tailored
solution,
and
so
then,
here,
if
you
could
come
in
and
actually
say
just
discover,
whatever
extensions
are
available,
like
oh
wow,
there's
all
this
other
stuff
and
you
get
this
json
body
back,
that's
like
and
here's
the
docs
for
our
readme
and
some
of
them
might
be
like
open
containers,
repo
docs.
That
means
it's
probably
been
standard,
standardized
and
accepted
upstream,
but
effectively
is
just
an
extension
of
the
you
know,
iteration
on
the
api
itself.
G
G
It's
almost
like
it's
almost
like
how
do
we
have
a
pattern
for
rfcs
or
you
know
whatever
docs
on
like?
Oh,
I
encountered
this
http
header.
What
what
is
it,
what
it
is,
what
it
do?
How
do
I
interact
with
it
if
it's,
if
it's
there
and
I
don't
do
anything
with
it,
I
don't
care.
G
Okay,
that's
enough
conversation,
I'm
not
looking
to
bike
shed
just
make
sure
we're
all
on
the
same
page.
Then,
if,
if
it
pairs
down
or
at
least
in
the
same
like
heart
of
it,
I've
got
a
pr
where
I
was
fixing
up
like
after
I'd
merged
into
this
111.
G
I
had
merged
tasia's
stuff,
but
there's
a
few
places
where
it
was
inconsistent
from
a
couple
of
files
which
is
fine
so,
like
my
local
working
branch
is
kind
of
a
mess
right
now
we
talked
about
it
and
then
every
you
know
unplugged
for
the
holiday.
So
it's
good
to
get
back
into
that.
G
Largely
is
there
anybody
knee-jerk
reaction
like
holly
against
this?
I
think
john.
At
some
point
you
you'd
said
what's
the
point,
but
given
this
discussion
is,
is
there
any
counterbalance
on
your
opinion
on
that
john.
H
I
don't
have
too
strong
of
an
opinion
at
this
point.
It
like,
ideally,
we
add
any
extensions
just
to
the
api.
If.
H
G
G
Piece
that
I
underscore
namespace
could
be
like
it's
yeah.
I
don't
know.
I
don't
know
that
we'll
ever
get
off
of
the
the
lead-in,
slash
v2,
but
I'm
sure
some
of
these
extensions
somebody
might
even
iterate
their
own
versions
of
an
extension
like
the
component
name
might
have
a
version
of
version
built
in
like
refers
referrers
v1,
referrers
v2.
If
you
actually,
if
you
made
a
formidable
change
and
it's
still
part
of
the
api.
F
So
I
just
made
a
comment
for
module
over
operation.
I
don't
know
if
it's
popular,
if
you
hate
it
download
it.
If
you
love
it,
give.
F
F
They're
going
to
avoid
having
conflicts
in
the
future
and
do
we
want
to
remind
people
that
they
can
do
whatever
they
want,
or
is
that
like
kind
of
implied
like
because
I
I
I
hear
that,
like
some
concern
that
people
people
can
do
whatever
they
want
and
still
kind
of
be
compliant?
But
I
don't
necessarily.
F
F
And
I'm
using
roughly
very
roughly
here
they
will
likely
be
compatible.
If
you
follow
this
like
and
and
I'll
I'll
say
it,
we
don't
know
it's
a
bad
idea
until
we
try
it.
I
don't
know
if
that
instills
confidence,
but
I
I
I
hope
that
we
can.
Let's
try
this.
I
I'm
saying
like
this
has
been
we've
been
talking
about
this
for
a
long
time.
Let's
give
this
a
shot
and
let
it
and
let
it
fake
and
if
it
doesn't
work
out,
we'll
we'll
try
again.
D
G
F
B
B
F
I
just
wanted
to
issue
another
ci
report.
The
cia
is
now
working
and
confirmed
recently
working,
but
clearly
it's
stable.
D
C
I
think
the
main
thing
is
just
for
people
that
have
busy
schedules
is
to
let
them
know
what's
coming,
so
they
could
plan
to
attend
the
meeting
or
not.
So
I
think
we
wound
up
with
the
right
people
here
today
for
these
two
things,
but
if
somebody
wanted
to
be
part
of
it,
they
didn't
know.
This
was
the
topic
to
cancel
some
other
meeting
show
up.
So
just
the
the
proposal
that
we
do
the
add
something
to
the
hack
talk
for
the
agenda
a
couple
days.
E
C
Yeah
master
becomes
the
stage
for
the
next
release.
Yeah
and
the
subject
of
change,
I
think,
is
one
of
the
things
we
had.
We
talked
about
before,
like
let's
at
least
get
it
in
maine,
so
that
there's
a
reference
point
there
and
if,
as
as
people
start
to
implement
it,
they
find
problems.
We
can
iterate
on
it
before
the
next
release
is
cut.
C
D
And
your
earlier
point,
steve-
probably
just
internally
admired
to
ourselves
that
a
day
or
two
before
the
meeting
to
probably
drop
a
little
slack
note
out
there
to
say
hey,
we
got
a
topic.
Dropping
the
hacking
deep
before
yeah
there's
a
slap.
C
Being
somewhat
different
but
tied
off
of
slack,
we
have
started
to
lose
some
of
our
older
conversations,
because
we
are
sitting
in
a
free
account
where
somebody
is
paying
to
run
the
service
for
us
developers
doing
it.
So
I
did
post
a
comment
to
chris
on
the
general
channels.
Like
is
this
something
we
might
want
to
pay
for,
so
that
we
can
get
the
features
of
the.
G
Pastry
they
are
paying
for
the
bottom
tier
of
slack,
just
not
the
full
featured
like
it's
it's.
When
we
first
started,
I
mean
like
I'm,
I'm
personally,
like
slack,
is
fine.
I
don't
think
it's
a
good
community
tool
because
it's
got
barriers
and
like
history
is
locked
in
a
box,
but.
G
The
first
thing
is
only
10
000
messages
for
an
entire
server,
so
across
all
channels,
after
10
000
messages,
it
just
falls
off
right.
So
we
are
on
like
some
lower
tier
paid
plan,
but
it
still
has
limitations,
they're,
just
apparently
adding
new
tiers
super
pro.