►
From YouTube: OCI Weekly Discussion - 2022-04-21
A
A
Yeah,
no,
I
think
that's
good
and
we
can
turn
the
meeting
into
that.
I've
actually
had
a
little
bit
of
that
this
past
week
or
so
there's
a
few
things
that
are,
thankfully,
thrusting
thrusting
my
attention
back
into
the
low
level
container
runtime.
So
it's
it's
been
a
weird
couple
of
years
and
I've
had
my
hands
off
the
keyboard
a
lot.
So
I'm
happy
to
dive
back
in.
B
C
I
was
really
just
coming
into
lurk,
but
it
sounds
like
okay,
we
had
like
it.
You
might
need
me
for
something.
B
Unrelated
to
oci
stuff,
but
you've
got
the
cncf
calendar,
don't
you
yeah
yeah?
They
wanted
to
change
the
cncf
s,
tags
working
group,
one
supply
chain,
okay,.
C
Oh
yeah
hasn't
dropped
me
an
email
because
it's
gonna
be
way
easier
to
track,
but
yeah
like
happy
to
fix
that.
C
A
E
B
My
previous
meeting
was
myself
and
brandon,
and
so
it
was
a
brandon
and
brandon
being
so
now
we
have
the
michael
michael
meeting
here
and
we'll
just
de-confuse
it
by
just
using
your
last
name.
Instead,
I
don't
know
I.
D
A
No
real,
immediate,
there's,
no
real
meeting.
I
was
just
gonna.
Actually
it
sounded
like
there's
a
couple
issues
possibly
to
talk
through.
So
if
you
don't
have
a
particular
thing
to
raise,
then
it's
good.
A
I
think
I
think
I
think,
there's
I
think
the
the
reference
group
was
a
was
the
big
thing
to
get
it
like
into
a
working
group
and
then
there's
probably
just
like
general
tidiness
and
momentum
to
do
and
clean
up.
That
has
been
lingering
for
a
while,
and
some
of
that
actually
just
needs
me
maintainers,
so
maintainer
consensus,
I've
seen
more
maintainers
getting
involved
on
the
issues.
So
usually
this
meeting
was
just
like
if
maintainers
couldn't
couldn't
bubble
it
up
on
an
issue.
So.
B
D
Brandon
we're
making
these
kind
of
meetings
a
habit,
do
what
brandon
and
I
were
on
the
working
group
meeting
on
tuesday,
which
was
also
similarly
without
a
agenda.
A
A
They
tend
to
be
longer
and
like
more
boring
timelines
and
like
there'll,
be
flurries
of
activity
and
then
like
nothing,
it's
been
very
normal
over
the
life
cycle
of
it,
so
as
as
some
folks
jump
in
and
jump
off
that
they
might
feel
like
it's
too
boring
not
worth
their
time,
but
it's
just
like
in
the
life
cycle
of
it
kind
of
a
slow
moving.
F
F
Of
lots
of
ci
cd
test
bugs
as
we
roll
docker,
you
know,
doctor
shim
out.
A
Probably
probably
not
entirely,
I
mean
to
bike
shed
here,
but
I
am
wondering
since
it's
come
up
a
few
times
and
it's
it
is
kind
of
an
interesting
point
of
how
bound
together
the
image
spec
is
with
the
distribution
spec.
B
That
campus,
we
were
talking
about
the
outputs
of
the
working
group
and
should
it
be
a
champion
spec
or
another
spec
or
in
the
oci
artifact
repo,
and
I
think
the
push
I
was
hearing
was.
It
needs
to
be
a
new
spec
for
the
artifact
media
type.
B
A
B
What
do
we
want
to
do?
Go
alias
and
that
way
anybody
who's
got.
Existing
code
still
works,
but
the
alias
points
over
and
so
the
new
location
would
register
as
the
same
object.
B
The
what
shadowing
on
the
layout
foreshadowing
foreshadowing
I've
got
a
talk
accepted
for
that
one
good.
A
B
The
other
one
I've
been
thinking
about
is
the
image
reference
not
to
be
confused
with
the
reference
type
discussion
we're
having
over
on
the
working
group,
but
the
image
reference
being
like.
How
do
you
expand
alpine
to
dot
io,
slash
output,
library,
alpine
colon
latest,
and
what
that
field
looks
like
when
you
take
it
different
places
been
seeing
the
request
for
things
like
when
we
put
on
the
file
system
with
the
image
layout.
B
Should
there
be
a
reference
for
how
that
looks,
and
so
might
be
another
area
to
look
at,
but
I've
definitely
seen
questions
of.
Where
is
it
standardized
that
this
thing
that
talks
to
docker
hub
gets
expanded
out?
You
know,
where
is
that
spec
specified,
and
it's
just
like
we
all
just
kind
of
know,
and
that's
not
the
best
answer
to
people.
D
I
just
I
mean
I
have
a
similar
question
about
redirects
for
blob
downloads.
It
seems
like
a
pretty
common
pattern
for
registries,
but
it's
it's
was
in
the
docker
spec
or
docker
api
docs,
but
not
carried
over
into
the
oci
distribution,
but.
A
Yeah-
and
I
mean
maybe
maybe
michael
if
you
want
to
open,
I
think
that
discussion
last,
I
saw
effectively
culminated
that
if
it's
already
teased
out
in
an
actual
like
ietf
rfc,
then
like,
let's
defer
to
those
right.
D
An
establishment
pattern,
I
think
client
for
client
behavior,
that
that
makes
sense.
The
distribution
spec,
though,
says
that
I
believe
it
says
registries
must
return.
200.
D
With
this
blob
data,
right,
which,
which
seems
like
a
little
bit
stronger
yeah
for
the
server
than
we
actually
want
to
specify.
A
D
B
B
E
I
feel
like
there
should
have
been
some
follow
up
with
what
those
inconsistencies
were,
so
it
could
have
been
addressed
in
the
spec.
That
was
a
long
time
ago,
though,.
E
I
think
they
created
a
fork
of
distribution
and
image
in
order
to.
E
E
A
Yeah,
it's
good,
but
right
before
the
the
flush
of
people
joined
or
just
saying
that,
like
the
in
the
life
of
the
group,
it's
you
know
certain
things
get
very
get
very
active
at
times
and
then
activity
falls
off,
even
though
there's
still
plenty
to
do
so.
A
A
But
in
in
these
last
you
know
a
couple
of
months
of
calls
here
have
been
pretty
quiet,
there's
a
different
conversation,
but
it's
not
been
just
like
packed
agenda
and
people
trying
to
get
things
talked
through.
So
it's
been
pretty
chill.
A
A
A
But
yeah
the
the
image
layout
is
was
was
made
as
a
thing
to
be
able
to
make
it
distribute
to
bull.
So
that
makes
sense,
but
the
descriptor
very
much
I
think,
could
move
over.
A
A
B
Shucks
yeah
I'm
looking
at
the
ad
gcr
one
so
yeah.
I
could
explain.
A
A
Yeah,
it
has
been,
it
has
been
tricky,
sometimes
to
keep
the
the
teams
in
sync
with
the
maintainers
now
also
in
sync,
with
the
code
review
code
owners,
I
think
we
should
actually
redo
the
code
owner.
So
it's
not
everybody's
individual
names,
but
the
whole
group
name.
E
So
the
folks
in
the
reference
type
working
group
are
almost
done
with
creating
a
proposal
and
would
like
to
bring
it
up
to
the
community.
H
I
thought
that
was
part
of
what
we
had
put
down
as
owners
is
that
there
was
at
least
one
maintainer
as
an
owner
of
the
working
group
as
well.
A
E
Can
anyone
step.
A
E
A
Can
you
put
a
link
to
that
in
the
in
the
that
hack
and
d
of
the
what
to
review.
B
I
think
we're
still
going
through
it
on
our
side,
but
all
through
the
we
had
it
from
last
time.
So
I'll
just
pull
back
up.
E
B
B
B
Yeah
and
the
discussion
that
I
was
having
with
a
few
people
earlier
this
week,
including
mike
brown,
the
other
mike
brown
there.
There
were
some
comments
from
the
registry
maintainer
side
that
I
want
to
make
sure
that
we
get
covered,
because
I
think
a
lot
of
what
we've
been
pushing
for
so
far
has
been
from
the
end
user
side,
building
the
client
side
half
of
this,
and
we
can
recommend
whatever
we
want
to
recommend.
B
B
B
B
Some
of
their
concerns
were
yeah.
That's
nice
to
do
if
you
want
to
do
sorting
filtering,
but
they
weren't
necessarily
sure
that
they
wanted
to
support
something
like
these
queries
against
all
annotations.
B
They
wanted
to
limit
and
say
yeah
you
can
filter,
but
we
only
want
you
to
filter
on
our
three
annotations.
That
way,
we
can
put
it
into
a
database
somewhere
and
index
off
of
it
and
implement
those
queries,
and
so
that's
where
we
need
to
figure
out
what
makes
the
most
sense
for
them
and
how
we
can
keep
both
the
users
happy
and
them
happy,
because
if
you
tell
the
users
they
can't
filter
on
all
the
fields
that
raises
other
concerns.
B
B
B
Url
encoded,
instead
of
base64
the
whole.
Actually,
this
whole
thing
is
url
encoded,
and
if
there
are
multiple
filters,
there
will
be
multiple
field
equals
on
your
url
parameter.
B
But
just
how
do
you
know
when
you're
parsing
the
string
when
to
stop
reading
the
field
start
reading
the
operator
and
then
start
reading
the
values?
B
I
saw
the
little
equal
equal
and
then
someone
said
well,
not
equal,
it's
just
equal,
not
equal,
and
that
made
all
the
other
things
fall
out
with
it
and
easy
enough
to
parse
syntax.
It's
just
not
pretty,
and
so,
when
things
aren't
pretty,
I
always
say
if
someone's
got
something
pretty,
please
I'm
very
happy
to
take
options.
A
B
That's
that's
a
little
bit
more
bike
shedding
the
the
bigger
concern
they
had
was
just
the
field
itself
that
we're
talking
about
is
any
annotation.
If
we
go
back
up
to
where
an
artifact
is
any
annotation
that
gets
thrown
on,
the
artifact
could
potentially
be
queried
against,
and
there's
some
hesitation
from
that
on
the
registry
side
of.
Do
we
want
to
allow
that.
C
E
Now
I
think
we
did
bring
up
the
fact
that
a
client
can
filter,
but
that
would
be
very
inefficient,
which
is
why
the
which
is
why
the
proposal
here
has
the
server-side
filtering.
B
B
B
This
is
exactly
what
we
need
to
be
able
to
do
for
a
long
term
for
adding
that
reference
api
later
on,
so
making
sure
that
getting
this
added
is
something
that
the
image
spec
maintainers
are
going
to
be
comfortable
with
that's
important,
artifact
spec
we're
assuming
we're
going
to
open
another
spec
in
another
repo
and
get
that
pushed
in
so
they'll.
I'm
assuming
goes
up
to
the
oci
talk
or
something
like
that
to
get
a
new
repo
created
and
approved
and
go
through
that
spec
creation
process.
B
H
So,
on
nation's
topic
of
presenting
to
the
main
seniors,
I
see
on
the
original
working
group
proposal
that
derek
mcgowan
was
listed
as
an
owner
there
and
it
looks
like
he's,
been
participating,
at
least
on
this
most
recent
proposal
he's
a
maintainer
of
the
distributions
back.
B
E
B
A
I
think
so
I
mean
like
this
that
the
I
guess
the
thing
here
is
that
you're
it's
it's.
This
is
actually,
I
guess,
yeah,
I
think,
we'll
fall
over
there.
I
could
get
into
bike
showing
this
yeah.
B
A
A
B
Good,
okay
yeah,
so
that
is
helpful.
At
least
lets
me
know
where
I
need
to
start
putting
put
my
changes.
I've
started
working
a
little
bit
on
what
the
change
inspect
is
going
to
look
like,
which
we've
got
an
old
pr
setting
from
way
back
out.
I
think
dan
had
been
the
one
that
proposed
it
a
while
back
for
adding
a
reference
type
to
all
the
different
objects.
E
No,
I'm
excited
to
see
this
land.
B
Here,
let
me
do
it
this
way
put
in
the
other
monitor,
so
that
we're
both
reading
together,
yeah.
I
know
last
time
we
were
talking
about
whether
or
not
the
version
bump.
I
will
resist
a
version
bump
as
much
as
I
can
an
image
effect.
We
make
these
changes
just
because
keeping
that
version
the
same
is
how
I
get
my
backward
compatibility
with
all
the
existing
clients.
B
B
Another
one
that
they
come
up
with
and
it's
been
mentioned
a
little
bit.
I
think
this
was
aws
mike
brown,
mentioning
it
was
whether
a
data
field
made
more
sense
than
promoting
up
the
annotations,
since
we've
already
approved
the
data
field,
we're
talking
about
in
the
response
api
that
we
would
pull
up
all
the
annotations
from
whatever
object
we
had
in
there,
so
that,
if
you
need
to
do
client-side
filtering,
you
could
but
also
just
giving
the
client
all
the
data
they
need.
B
When
they're
looking
at
this,
so
the
comma
came
back,
does
it
make
sense
to
pull
up
all
the
annotations,
or
should
we
just
put
a
data
field
here
for
the
manifest
we're
pointing
to,
and
I
can
see
both
sides
of
that.
The
one
thing
that
I
thought
of
after
the
meeting
was
that
what,
if
that
manifest
itself,
has
a
data
field
that
we're
pointing
to
they
might
want
to
pull
up
everything
into
that
manifest
and
make
that
you
know
get
as
close
to
that
four
mega
limit
as
they
can
to
make
it
only
one
pull?
B
And
so
then,
if
you
then
rebase
encode,
all
of
that
and
put
that
up
here
as
another
data
field,
you're
now
pushing
a
bunch
of
limits.
So
I'm
feeling
like
annotations,
might
be
a
better
level
for
this
one.
But
I
am
open
to
people
saying
otherwise
bless
you
vincent
hope,
you're
getting
better.
E
E
B
B
So,
instead
of
going
up
here
and
saying,
give
me
all
the
annotations
on
this
thing.
Let
me
pull
all
these
up.
You
could
also
say
in
here.
Let
me
base64
this
entire
manifest,
which
includes
all
the
annotations,
the
media
type.
On
this
thing,
all
the
other
details
in
here
just
the
whole
thing.
Base64.
The
whole
thing
drop
that
in
as
a
data
field
over
here
in
the
response
to
the
api
and
that
kind
of
works
it.
B
I
think
it
still
solves
what
you're
looking
for
nisha,
but
the
flip
side
that
I'm
worried
about
is
okay,
yeah,
that's
great
until
you've
got
these
config
and
layers,
and
these
layers
themselves
might
have
a
few
data
types
dispersed
within
there.
The
convey
itself
might
be
injected
as
a
data
type,
so
you've
pulled
back
a
lot
more
data
than
I
think
some
people
are
initially
hoping
for.
B
B
So
yeah
any
thoughts
people
have
on
whether
or
not
to
pull
all
the
annotations
up.
Otherwise,
sorry
now
my
voice
is
cracking
here
in
bensco.
Hope
you
didn't
give
me
anything.
B
B
So
in
terms
of
we've
got
a
few
things
we
think
we
can
do
now,
and
so
would
it
make
sense
to
get
these
initial
things
out
there
into
the
specs
today,
so
that
people
can
start
using
this
for
references
today
put
their
things
up
there
using
the
oci
artifact
definition,
which
puts
them
in
an
image
spec
start
using.
B
If
there
are
any
concerns
of
no,
you
need
to
go
through
and
do
the
whole
distribution
distribution
before
you
push
up
any
changes
to
image.
Spec
I'll
be
disappointed,
but
let
me
know
now,
if
that's
the
case,
that'll
be
good
misha.
Is
that
thumbs
up
to
make
everything
one
massive
change?
Instead
of
breaking
it
up?
No.
B
B
C
The
notes
about
the
slack
being
broken
for
oci
yeah,
we
know
about
it,
there's
something
weird
going
on
chris
knows
about
it,
but
if,
if
you
need
any
help,
just
send
them
my
way.
Oh
I'm
happy
to
be
able
to
like
invite
people
in
it's
fine.