►
From YouTube: OCI Weekly Discussion - 2022-08-25
B
That's
all
the
way
from
Wyoming
yeah,
it's
I,
don't
think
it
represents
any
specific
thing
other
than
it's
one
of
those
hats
that
if
you
buy
it,
some
of
the
proceeds
go
to
the
national
parks.
A
B
C
A
I
will
say:
Mike.
You
have
been
extremely
proactive
on
this
one
I
appreciate
all
your
feedback,
usually
I'm,
the
one
that's
in
the
last
minute.
A
C
E
A
C
A
A
A
A
E
A
That
jumped
in
and
over
on
image,
spec
I
thought
we
had
one
more
commit
over
there.
I
wonder
if
I'm
missing
one
oops
wrong
button
there
reload
this
page,
see
if
I'm
missing
one,
because
I
thought
we
had
yeah
there's
the
other
one.
I
was
like
V
bats.
I
know
you
gave
me
some
feedback.
So
let's
see
there,
we
go
so
I
just
updated
this
from
vbats
earlier
I've
got
the
CBP
I'm.
G
F
At
the
last
minute
that
was
until
your
slack
message,
I
completely
forgot,
I
didn't
realize
that
that
is
the
25th.
A
That's
the
way
it
happens,
isn't
it
excellent
guess?
That's
all
so.
I
got
those
in
there,
the
other
one
that
happened
earlier
in
the
week.
I'd
just
change
the
ice
cream
to
generic
manifest,
and
example
s
bomb
instead
of
a
flavor,
it's
I'm,
hoping
those
make
more
sense
people
going
into
the
future
that
haven't
seen
all
the
working
group
stuff.
A
That
was
it
since
last
week
for
what
we
changed
in
the
image
spec.
So
a
lot
more
has
been
changing
here
since
the
beginning
of
this
whole
thing,
but
anybody's
been
following
along.
You
didn't
miss
much
over
here
and
then
distribution
spec.
Over
on
this
side,
we
did
change
a
fair
bit
of
stuff.
We
went
back
to
the
original
specs
and
conformance
over.
Here
we
got
rid
of
the
separate
upgrade
requirements
in
the
separate
client
implementation.
A
Those
files
got
deleted
and
then
they
all
got
folded
into
the
spec.nd.
So
we
got
a
lot
of
lines
just
got
merged
into
here.
We've
got
I
think
a
bunch
of
various
nits
and
cleanups
and
whatnot
throughout
the
file
and,
let's
see
a
whole
bunch
of
clean
up
definitions
thanks
so
much
on
those
Mike
Mike
on
this
one
is
this:
do
we
want
to
talk
about
this
real
quick
while
we're
here
the
image
index.
C
Not
clicked
on
the
Run
wood,
this
follows
from
the
definition
from
the
way
it
was
specified
or
currently
it's
specified
in
the
image
you
know
index
MD
or
not
not
in
the
AMD.
Sorry,
there's
a
list
of
definitions
and
in
the
image
specification.
E
C
C
Got
it
so
they
I
think
they
believed
at
that
point
that
that
was
generic
right
as
opposed
and
then
I
guess
on
the
other
ones.
We
had
things
like
when
you're
looking
at
the
image.
The
list
of
you
know
the
list
of
blobs,
for
example
they're
they're
in
a
particular
order,
and
so
when
they,
when
that's
defined,
it's
explained,
you
know
what
kind
of
index
that
is
so
in
order
list
in
order
you
know
order
list
or
index
that
kind
of
thing
I,
just
so
really
all
I
was
doing,
was
normalizing.
A
C
C
Right
so
my
suggested
changes
here
just
to
basically
to
pick
it
up
and
bring
it
over.
Then,
if
we're
going
to
put
it
here
and
then,
if
you
want
to
change
it
in
in
the
in
the
owning
specification,
then
we
you
know
change
it
there
and
then
change
it
here
and
another
po.
H
I'm,
just
coming
back
to
I
think
we
can
take
this.
Probably
one
one
option
is
to
kind
of
break
this
up
a
little
bit,
because
what
we're
trying
to
do
is
add
the
artifact
manifest
to
the
existing
specification,
making
the
manifest
definition.
Generic
might
be
beyond
what
we
want
to
have
within
the
scope.
So
do
you
think
we
can
get
the
artifact
manifest
in
and
then
maybe
open
a
PR
to
kind
of
unify
with
making
generic
manifests
across
everything?
D
C
There's
two
parts
of
it
right:
the
distribution
specification
attempts
to
already
be
generic
with
respect
to
the
various
types
of
manifests,
so
it's
valid
what
he
was
trying
to
do
here.
The
problem
is
that
it's
for
a
specific
term
image
index
and
that
was
already
defined
in
another
specification,
so
I'm
trying
I'm
trying
not
to
step
on
on
that.
You
know
on
that.
Spec
yeah.
H
A
I
was
looking
at
the
order
of
operations.
Do
we
want
to
change
our
PR?
That's
got
the
definition
over
in
distribution
or
do
we
want
to
say
for
the
one
day
one
week
that's
going
to
take
just
get
a
separate
PR
to
go
through
here.
Merge
this
as
is
right
now
and
then
a
separate
PR
over
the
image,
spec
and
and
I'll
say,
is
one
of
the
image,
respect,
maintainers,
I'm
comfortable
being
out
of
sync
right
now.
F
Well,
it
was
just
skimming
over
the
image
spec
API
image,
spec
pr934-
to
see
if
we
actually
added
in
any
of
this
stuff-
in
that
example,
where
the
the
like
that
that
that
yeah
there.
If
and
and
we
don't-
actually
modify
any
of
this
so
high
level,
components
of
the
spec
include
the
so.
The
current
934
doesn't
even
touch
this.
F
Is
is
a
punch
list
being
kept
somewhere
of
like
these
kind
of
follow-ups?
You.
F
C
F
E
E
A
A
Sound
good,
so
if
we
go
that
direction
Mike,
are
you
comfortable
marking
this
one
as
a
result
over
here
for
now.
A
C
C
I
think
that
was
fair.
Sometimes
it's
content,
sometimes
a
subject.
I
actually
I
didn't
like
it
at
first,
but
the
more
I
thought
about
it
yeah.
This
is
good.
It's
it's.
What
was
meant
originally
by
the
word
artifact
in
this
spec
right.
A
C
A
C
A
A
All
right
and
then
yes,
some
other
definitions
we
pulled
in
just
to
make
sure
we're
defining
all
of
our
terms
refers,
refers
and
refers.
People
complain
about
that
and
actually
understand
why
the
little
the
little
in.
F
The
English
I
thought:
that's
almost
why
when
I
was
reading
through
and
I
see
lucky
or
whoever
has
already
commented
on
it,
but
like
the
field
refers
versus
referrs
like
would
it
be
and
I
get
it
that
there's
already
a
lot
of
work
and
back
behind
this,
but
like
would
refers
as
a
field,
be
almost
intuitive
to
say
refers
to
like
yes,
in
a
definition
or
a
string
of
like
defining
what
the
field
is.
C
F
I
know
that
sounds
like
a
big
change.
It's
just
like
in
if,
if
that's
what
it
means,
because
actually,
when
I
read
through
it
at
first
I
was
thinking,
it
was
a
list
of
refer
errors
rather
than
this
thing
refers
to
something
like
that,
like
the
directionality
of
it.
So.
E
F
C
D
A
A
E
I
F
D
C
I
A
A
A
A
F
F
Or
something
no
annotated
denoted
right,
isn't
it
yeah
exactly
yeah
yeah?
Maybe
that's
fine
I
just
actually
had
to
scroll
down
and
scroll
back
up
on
a
solid
capitalized.
I
F
F
E
I
C
A
All
right,
this
one's
still
left
unhandled
here
registry,
should
accept
and
manifest
with
the
refurbished
field
that
references,
something
doesn't
exist.
The
goal
here
for
people
don't
want
to
read
what
I
wrote.
If
you
push
a
signature,
we
want
to
make
sure
that
you
don't
have
to
have
the
image
there
in
advance
so
that
someone
could
push
the
image
after
the
signature
is
pushed
to
registry.
A
That
way,
the
registry
can
say
make
sure
that
only
signed
images
are
up
on
the
server
on
those
lines,
so
we're
just
sorting
out
the
order
of
operations,
and
so
the
wording
here
was
just
to
say:
hey
if
you're
missing
a
blob,
then
maybe
you're
going
to
reject
it.
But
if
you're
missing
the
reverse,
descriptor
Target
don't
fail
on
that
one
and
at
least
saying
it
should.
C
F
Well,
I
mean
have
we
in
the
conformance
tests?
Have
we
we've
already
seen
that
some
Registries
will
reject
other
fields,
and
even
if
we
have
buy-in
from
folks
for
that
to
change,
then
why
why
they
reject
it?
Who
knows,
but
on
this
particular
one
of
like
the
Nuance
of
if
a
object
doesn't
exist,
that
it
refers
to.
C
F
F
That
this
guy
chewed
around
on
a
lot
was
basically
came
down
to
how
am
I
supposed
to
garbage
collect
anything
if,
if
I
can't
see
that
it
refers
to
something
like
if
I
have
an
object
on
the
registry
and
nothing
refers
to
it
and
if
I
have
objects
that
refer
to
something
that
doesn't
exist,
you
know,
because
what,
if
I'm
pushing
a
bunch
of
what?
If
somebody
chooses
some
object,
maybe
not
an
s-bomb,
but
some
big
ass
object
that
they
don't
want
to
put
in
their
main
registry.
F
D
F
F
C
D
C
C
A
C
A
Well,
we
happen
to
find
garbage
collection
anywhere
in
here,
and
so
that's
always
been
left
up
as
an
exercise
for
the
registry
to
sort
out.
I
would
say
that
more
than
likely,
the
way
they
always
sort
this
out
is
either
you're
going
to
have
a
repo
that
says:
don't
you
see
anything
in
here
and
that's
how
you're
going
to
refer
to
external
stuff
or.
E
A
C
I
guess
I
guess
my
read
of
this
is
where
I
started
asking
those
other
questions
Brandon
now
that
I
remember
around
okay.
Well,
then
I
guess
I'm
going
to
need
to
push
my
artifacts
right
after
the
things
that
they
point
to
and
I'm
going
to
need
to
list
them,
probably
also
in
the
index
to
make
sure
that
they're
in
the
graph
and
they
won't
get
garbage
collected
until
after
the
entire.
You
know,
index
has
been
requested
to
be
the
lead.
C
C
A
H
There
is
a
scenario
where
we
want
to
push
the
signature
first
before
we
actually
push
the
image.
So
that's
kind
of
why
the
the
Gap,
like
you,
said
the
timing
of
pushing
the
referee
first
and
then
push
the
images
there.
If
we
make
this
like
a
mandatory
requirement,
even
for
GC,
it
makes
it
hard
to
implement
that
use
case.
C
C
C
A
I'm,
not
quite
sure,
I
follow
where
you're
going
with
that.
Well
you're,
like
we've
kind
of
gotten
to
the
point,
we
we've
got
a
lot
of
agreement
that
this
we
should
leave
it
as
I
showed
right
now
to
leave
Registries
for
the
flexibility.
What's
what's
the
future
follow-up
that
you're
looking
at.
C
C
C
C
F
A
F
F
A
J
K
C
J
C
C
D
K
J
J
C
That's
a
good
point:
it
doesn't
really
talk
about
the
the
time
value
of
this.
K
J
C
A
C
You
could
you
could
make
it
a
must
initially
accept
right
and
then
you
could
add
an
additional
sentence
saying
something
to
the
effect
you
know
make
garbage
collect
if
it
turns.
If
you
know,
if,
after
a
certain
period
of
time,
it's
no
long,
you
know
it
is
still
not
referring
to
an
existing
manifest.
A
C
A
J
K
A
Throw
a
scenario
out
at
you,
which
is
if
I
recursively
copy
an
image
between
two
places
and
it's
multi-platform
everything
else.
I
go
bottom
up,
so
I
go
all
the
way
down.
The
tree
start
copying
the
blobs
and
start
copying
the
image
manifest
then
copy
the
index
and
when
I
include
refers
in
there.
I'll
go
bottom
up
as
well,
and
so
I'll
copy
all
the
refers,
and
then
the
image
manifest
above
it.
H
A
D
C
And
it
and
it's
I,
think
it's
it's
definitely
important
to
figure
this
out
right.
We
in
in
the
past,
when
we,
when
we
did
the
directed
graph
ordering
right.
We
we
made
sure
that
you
didn't
have
any
dead
references
right
and-
and
it
seems
like
what
you're
telling
me
now
is
that
it
must
reject
a
manifest,
but
that
would
require
a
certain
sequence
on
the
pushes
always
right
which
is
doable,
but
you
just
have
to
decide
if
that,
if
that's
the
pattern,
you
want
to
use.
C
A
C
A
C
C
C
E
D
F
Kidding
keep
going
okay
if
you'll
re
refresh
that
page.
This
is
back
back
setting
it
at
this
point,
but
as
you're
talking
it
does
right
up
right
up
above,
oh
I,
guess
I
can
reset
your
cursor.
Damn.
F
Don't
know,
oh
just
oh
just
yeah
and
too
bad
too
bad.
We.
A
F
E
F
A
Okay,
let's
see
cover
decline,
implementations,
that's
good.
I
can't
hear
that
I
need
to
lower
case
some
letters
here
and
pushing
manifests
with
refers.
This
is
kind
of
like
a
whole
list
here
of
the
client
process,
what
they
do,
and
so,
if
the
client
sees
when
it's
trying
to
do
a
immature
and
rpac
manifest,
it
has
a
reverse
field
and
the
refers
API
gives
back
a
404.
This
is
kind
of
logical
workflow,
of
how
the
client
is
maintaining
that
image.
The
index
up
on
the
server
that
we're
looking
at.
A
So
I
don't
remember
everything
it
changed
in
here.
I
think
some
of
this
was
just
moving
the
client
field
over,
but
I
think
there
was
also
some
rephrasing
of
things
changing
stuff
from
like
the
index
to
refers
list
just
to
make
it
clear
that
that
was
kind
of
what
we
were.
What
this
tag
was
pointing
to.
A
F
F
If
somebody
doesn't
adopt
the
refers
API
and
if
there's
some
other
iteration
in
the
future
that
they
want
to
see
they
get
compliant
with
the
one
dot,
whatever
version,
but
is
it
I
just
I
guess
why
1.1
versus
a
1.0.,
whatever.
A
C
Well,
I
think
Vince.
We
don't
have
much
of
a
choice
because
of
the
1.1
we're
going
to
need
in
the
image
spec
for
the
new
Fields.
There's
it's
pretty
significant
and
you
got
a
new
manifest
type
that
we
have
to
support
right.
I,
don't
think
you're
really
gonna
have
a
choice
in
them
and
expect,
and
then
it
would
be
nice
I,
think
I,
I,
hate
to
say
nice,
but
I
think
we
should
keep
the
distribution
spec
at
the
same
point,
release
as
the
image
specs
since
they're
they're
sort
of
you
know
two
in
a
box.
C
F
G
E
A
Yeah,
it
is
a
digest
at
this
point.
F
Well
it
so,
then,
if
you
scroll
down,
because
even
in
the
right
below
here
of
like
the
value
of
reference
and
then
the
examples
that
get
pasted-
oh
well,
it.
F
Oh
I
was
about
to
say
their
stuff's,
not
showing
on
yours
everywhere
that
it's
referred
to
as
this
like
you
know
reference
it.
It
doesn't
actually
show
what
it
is
or
a
description
of
like
like
it
should
kind
of
just
put
arbitrary
strings
in
there.
A
E
B
C
Some
of
those
references
were
you
know
things
like
you
know,
repository
slash,
you
know
shop
or
something
I
can't
remember.
F
F
G
G
D
G
B
C
C
A
C
You
want
to
make
that
refers
reference,
then,
as
opposed
to
the
other
references
that
could
be
either.
E
A
C
If
you're
gonna
do
that,
you
might
might
add,
you
know,
add
some
text
they're
explaining
that
you
know
we
may
be
adding
tags
later
so
that
when
they're
developing
their
code
in
the
registry
and
the
clients
they
can
keep
that
in
mind.
D
A
We
really
say
everything
should
be
at
200
or
404.
We
might
need
to
say
like
another
error
code
in
there
we
don't
want
a
404
to
come
back
if
it's
a
just
a
missing
ref,
something
that's
missing
that
doesn't
exist.
Yet
we
we
want
that
to
come
back
as
an
empty
list,
but
yeah
we
might
want
to
find
I,
don't
know
if
it's
a
500
or
some
other
reject
message
in
there
of
some
kind.
C
C
Make
the
the
refers
API
more
complex
if
it
supported
digest
I'm
sorry,
tags.
A
A
A
This
is,
if
you
push
something
up,
so
if
you
push
your
signature
up
in
advance
and
say
Here's,
my
signature
for
an
image
doesn't
exist.
Yet
the
registry
can
respond
by
saying:
hey,
there's
just
nothing
that
references
that
digest
that
you
gave
me,
but
once
that
image
is
finally
pushed
that
reference
exists,
then
you
need
to
return
the
signature
in
your
results.
So
the
order
or
pushing
shouldn't
affect
whether
you
get
that
signature
in
the
future.
A
C
Well,
it
seems
like
the
the
focus
of
the
work
group
was
on
the
the
use
pattern
where
I'm
pushing
my
you
know
a
priori,
pushing
my
image,
Index
right
and
and
all
the
manifests
in
Blood
and
then
I'm
going
to
add
artifacts
later
on.
It
seem
to
be
the
primary
case,
but
now
that
we've
got
an
image
index,
a
list
of
image.
C
Artifact
manifest
it
seemed
like
it
was.
You
know,
an
opportunity
here
to
be
able
to
push
all
of
my
images
first
and
then
push
my
artifacts
pointing
to
those
images
and
then
push
my
index
with,
and
you
know,
with
the
artifacts
listed
and
then
I
would
have
a
digest
with
a
full
graph
for
everything
I
needed
to
be
able
to
pull
all
the
artifacts
for
this
particular.
A
And
someone
someone
can
do
that,
we're
not
preventing
it
with
what
we're
defining
here.
The
reason
we
stayed
away
from
that
from
the
working
group
standpoint
was
we
are
looking
at
two
requirements.
One
is
the
ability
to
be
able
to
add
metadata
to
the
image
after
it's
been
pushed
by
the
originator,
and
the
second
was
if
you
want
to
copy
this
across
Registries,
but
you
don't
want
that
metadata.
How
can
you
strip
it
without
changing
the
Digest.
C
E
A
C
And
the
the
only
thing
that
was
I
think
missing
from
being
able
to
use
that
really
is
the
concept
of
well.
What,
if
I
need
the
artifact
to
reference
the
index
right,
that
index
wouldn't
be
there?
I,
wouldn't
know
what
the
digest
is,
and
you
have
that
recursion
problem,
so
it
it
made
sense
to
me
in
that
case
that
you
might
have
them
all
over
the
reverse.
If
it's
null,
it
means
it's
going
to
refer
to
the
index
in
which
it
is
included
or
the
manifesting
which
is.
C
It
is
included
in
that
case,
okay.
So
if
you
had
that
one
additional
use
right
where,
if
the
reverse
does
not
have,
is
not
set,
but
you
have
an
artifact
listed
in
the
index,
then
you
would
have
a
graph
right
in
those
artifacts
that
artifact
wouldn't
in
fact,
because
it
doesn't
refer
to
one
of
the
sub
you
know
manifests.
It
would
be
for
that.
This
right.
A
Index
I'm
going
to
say
that's
a
use
case
that
we
weren't
optimizing
for
and
so
I
don't
want
to
necessarily
drop
it
in
here.
Someone
could
do
if
they
wanted
to.
C
D
C
And
and
others
have
customers
that
are
taking
an
existing
set
of
images,
and
you
know
adding
extended
attributes
if
you
will
right
these
extended
artifacts
that
point
to
the
images
that
already
exist
in
the
registry,
I
think
there's
two
patterns
and
there's
different
sets.
A
bank,
for
example,
probably
doesn't
want
you
to
be
pulling
artifacts
that
talk
about
their
images,
I'm,
just
saying
you're
going
to
have
different
use
patterns
here
and
that
that's
what
I
was
I
was
trying
to
express
on
that
and
that.
A
Comment
yep
being
sensitive
time,
seeing
a
bunch
of
people
they
have
to
drop
off,
might
have
already
lost
a
bunch
of
them,
but
Sanjay
between
you
and
me,
and
anybody
else
that
wants
to
chime
and
Mike
too
do
we
want
to
handle
this
offline
before
next
week's
meeting
after
after
we
finish
up
today,
I'm
going
to
keep
going
for
a
little
bit,
but
I
just
want
to
be
sensitive.
Anybody
needs
to
drop
if
we
want
to
fix
this
stuff
and
get
it
merged
after
the
meeting.
A
A
But
but
I'll
keep
going
yeah,
let's
see
I
think
we
answered.
You've
asked
this
questionnaire
already
on
available
stuff.
Note
warning:
he
wants
red
flags
in
here.
A
A
H
A
And
the
commit
header
and
the
commit
message,
I
think
when
I
squash,
it
I
think
I
want
to
squash
it
does
that
too.
D
A
C
D
A
I
re-based
a
bunch
of
stuff,
so
let's
try
one
see
if
I
mess
it
up.
I've
got
a
backup
Branch,
just
in
case
I
do
but
from
the
playground
PR,
including
our
feedback
from
the
day
I'm
going
to
go
back
to
the
Upstream
main
that
we
got
out
there.
C
A
C
Yeah
some
sometimes
the
the
check
tools
like
the
best,
won't
accept
multiple
sign-offs,
but
we'll
accept
the
co-auth
yeah.
A
A
A
I
think
I
have,
oh,
you
guys
are
going
to
see
my
phone
do
not
screw
things
up
method.
A
A
D
A
A
A
H
H
D
H
D
A
E
A
D
A
C
A
H
Yep
this
call
is
recorded
and
whatever
we
say,
we
assume
it
to
be
binding.
So,
yes,
we
will
submit
the
the
follow-up
with
the
to-do
list.
I
think
that's
something
that
we
should
agree
on,
even
if
you
don't
put
it
on
the
in
the
actual
in
this
PR
or
comments
in
it.
H
I
think
we
can
immediately
open
an
issue
and
close
it
up
as
a
requirement
in
some
way.
Is
that
if
that
helps.
D
B
Yeah
right,
yeah
yeah,
all
right
I
got
a
drop,
but
thanks
thanks
for
all
the
work
on
this.