►
From YouTube: OCI Weekly Discussion - 2022-05-26
D
Zoom
did
something
weird
on
a
recent
upgrade,
and
so
I
backed
it
out
and
I
checked
and
it
looked
like
it
was
good
on
the
newer
update.
So
I
thought
I
was
good
with
the
new
version,
but
who
knows.
B
D
B
Yeah,
just
to
just
to
consolidate
things
that
I
need
to
change.
B
So
is
it
not
putting
the
headings
or
putting
the
headings
but
with
some
detail
underneath.
A
I
I
was
mostly
going
off
at
brandon's,
I
mean
I
think
that
the
those
categories
are
too
vague
or
maybe
don't
cover
everything.
I
think
correct
me
if
I'm
wrong
brandon
you're
saying
put
like
a
mark
down
comment
with
like
examples.
D
No
you're
hearing
me,
okay,
so
yeah
the
thought
process
I
had
there
was
just
people
are
going
to
come
in
different
ways,
and
so
I
figured
just
in
case
someone
had
different
scenarios
not
to
have
a
hard
coded
list
of
bullets
they
had
to
go
through,
but
just
to
have
like
a
general
area
of
here's
ways,
you
can't
contribute
fill
in
whatever
makes
sense
for
you.
B
Okay,
so
not
necessarily
detailing
what
people
need
to
fill,
but
some
comment
about
you
know
what
are
possible
or
suggestions
on
what
work
you
can
highlight.
E
Okay
I'll
make
that
change,
then.
A
Does
anyone
have
a
good
idea
of
docker's
kind
of
like
approach
to
what
they
were
saying
on
tuesday
tianon
you
work
at
docker
now.
F
Yeah,
so
the
things
that
chris
and
tonis
are
talking
about
are
honestly
things
that
I
didn't
have
any
visibility
on
until
like
60
days
ago.
So
I'm
probably
not
the
best
person
to
talk
about
them,
but
they
are
planning
to
come
back
with
some
kind
of
proposal
f,
as
proposed.
So
okay.
A
Yeah,
personally,
you
know,
we've
been
even
though
we've
been
making
progress
on
the
e.
I
I
don't
mind
if
they
put
something
up
and
it
works
and
works
for
everything,
I'm
mostly
confused
about
what
it
looks
like,
but
I
guess
we'll
we'll
wait
and
see.
F
Yeah
the
way,
the
way
I've
been
looking
at
their
their
recent
conversations
is
that,
even
if
all
this
ends
up
being
is
good
practice
for
explaining
why
e
is
the
best
option,
because
we're
going
to
have
to
do
that
to
all
the
other
maintainers.
This
is.
This
is
a
useful
exercise
right.
It's
it's
lower
stakes
than
that
final
conversation
with
the
maintainers
is
going
to
be
so
I
think
from
my
perspective,
it
it
seems
like
it's
been
really
productive
and
helpful.
To
have
someone
like
poking
a
little
stick
at
it
and
going
hey
can.
B
Is
are
folks
just
talking
in
the
pr
about.
A
Misha,
I
think
you
were
there
on
tuesday
right
yeah,
we
basically
there
there
was
just
it
seemed
like
there
was
just
going
to
be
ongoing
back
and
forth,
and
so,
even
though
you
know
it
makes
work
more
work
for
chris
and
then
I
think
the
only
way
we
can
really
evaluate
what
they're
saying
is
for
them
to
submit
a
proposal.
So
I
I
put
a
I
opened
a
new
issue
which
is
like
and
tagged
them
on
it
and
said
open
a
proposal
and
we'll
go
from
there,
because
I
don't.
A
I
don't
think
we
can
just
like
take
shots
at
proposal
e.
If
there's
not
an
alternative,
I
mean
you
can
take
shots
but
then
what's
you
have
to
kind
of
say?
What's
so
then.
A
Well,
that
yeah,
so
that's
sort
of
what
I'm
trying
to
get
at
is.
If,
if
the,
if
the
whole
thing
is
like
go
away
working
group,
then
I
think
we
have
to
ignore
it,
but
if
they
actually
want
to
provide
an
alternative,
then
I
think
that's.
F
I
think
one
thing
that
chris
has
mentioned
to
me
is
that
he's
specifically
concerned
that,
because
we
have
a
fallback
of
just
using
tags,
that
everyone's
going
to
start
using
the
fallback
and
there's
no
real
incentive
for
registries
to
implement
anything
else,
because
everyone's
already
using
the
fallback
and
they
don't
have
to
make
any
changes.
So
why
should
they
there's
there's
no
forcing
function
on
registries
to
actually
make
a
change?
D
But
at
the
same
time,
we've
also
heard
from
them
that
it's
no
good,
if
you
propose
something,
doesn't
work
with
existing
registries,
because
then
the
registry
is
going
to
upgrade
and
you'll
never
get
the
feature
out
there
and
there's
a
dead.
You
know
lame
duck
going
out
the
door
yep
and
we've
also
heard
that
they
don't
like
our
proposal,
because
all
the
tags
are
going
to
be
miserable
for
registry's
implement.
D
They
don't
want
to
see
a
whole
bunch
of
tags
all
over
the
registry
server
and
users
querying
lots
of
manifest
to
pull
the
annotations
up
and
that
kind
of
stuff.
So
that's
kind
of
why
I
want
to
see
them
actually
propose
what
they
want
to
propose,
because
I
feel
like
we're
seeing
arguing
a
lot
of
different
sides
of
the
same
issue
and
it's
like
we
need
to
narrow
down
what
we're
actually
discussing.
B
So
that's
that's
interesting
to
me
that
on
the
registry
side
they
don't
want
to
see
so
many
tags,
but
at
the
same
time
their
garbage
collection
mechanism
is
based
on
tags
kind
of
okay.
Is
there
another
mechanism.
D
So
the
mechanism
most
registries
use
will
look
at
the
tags.
Hub
extends
out
a
little
bit.
They
keep
track
of
multiple
revisions,
and
so,
if
you
override
a
tag
or
do
something
like
that,
they
still
keep
around
old
versions
of
the
tag
to
know
when
you're
pulling
by
like
a
pen
digest.
It
doesn't
immediately
go
away
when
you
deserve
write
the
tag
kind
of
stuff.
So
at
least
that's
the
description.
I've
seen
it
mirrors
what
I've
seen
in
practice
of
actually
querying
hub.
That
stuff
sticks
around
even
when
it's
not
tagged.
D
Their
big
concern
and
one
that
I
actually
share
and
the
reason
that
made
some
of
the
implementation
decisions
on
my
side
of
trying
to
do
this
is,
if
you
have
all
these
tags,
one
the
tag
api.
Their
concern
is
not
super
efficient.
You
gotta
query
a
tag
listing
to
see
all
the
tags
and
parse
through
that,
and
that
can
get
a
very
long
list
and
might
be
paged,
and
the
second
concern
I
heard
from
them
was
okay,
you
find
20
manifests
that
all
or
20
tags
that
all
look
like
something
you
might
want
to
query.
D
C
G
I
think
josh's
request
to
submit
a
proposal
makes
the
conversation
much
more
easier
right
and
given
that
we
literally
have
docker
gcr,
ecr
and
acr
at
the
table
of
this
discussion,
if
we
can
get
everybody
to
say
yes,
we
will
adopt
this,
then
maybe
that's
that's
a
great
start,
maybe
having
one
more
person
from
coe
would
also
kind
of
like
make
these
things
easier.
If,
if
he
can
get
a
consensus
of
moving
forward.
G
B
D
B
B
So
now
I
ask
my
question
again
about
versioning:
wouldn't
wouldn't
like
adding
a
new
version,
give
like
solid
separation
line
between
this
is
the
old
way
of
doing
it,
and
this
is
the
new
way
of
doing
it.
If
you
want
to
use
the
new
way,
then
you
can
it's
up
to
you
whether
you
want
to
maintain
clients
that
do
both.
D
D
B
Yeah,
no,
I
I
mean
this
may
be
my
you
know,
ignorance
speaking,
but
there
are
already
existing
clients
that
query
like
index.docker.com
v2.
B
So
those
clients
can
continue
to
query
that
if
a
new
client
is
made,
you
can
either
say
like.
Okay,
we're
no
longer
going
to
query
v2
we're
only
going
to
query
v2
ev3
or
you
can
have
a
client
that
says
we
can
query
both
v2
and
v3.
I
mean
it's
kind
of
similar
to
the
way
that
so
in
python
there
is
a
mod.
There
is
a
module
called
six
which
queries
which,
which
kind
of
does
the
backwards
compatibility
for
you.
B
If
you're
using
python
3,
then
you
can
use
6
to
convert
that
to
python
2.7.
So
I
mean
I'm,
I'm
I
get
the
feeling
that.
D
Well,
let
me
throw
a
specific
example
out:
am
I
okay
that
might
drive
the
conversation
a
little
better?
There's
like
debian
version,
six
images
sitting
on
a
docker
hub
today
are
still
schema
v1,
so
I
still
have
to
support
schema
v1
a
lot
of
my
code.
D
If
we
say
the
solution
is
going
to
be
a
hard
version,
update
version
3
there
there
is
no
backward
compatible
solution.
I
think
docker
already
came
back
and
said:
no,
you
need
something
that's
going
to
work
today
and
so
let's
say
we
build
something
that
has
a
solution
that
works
today
and
so
someone
updates
the
debian
version
10
or
whatever
the
next
release
is.
I
should
be.
I
should
know
these
version
numbers
better.
D
They
push
the
latest
one
out
and
they
include
the
s
bomb.
They
include
the
signatures
on
that
image
today,
with
some
kind
of
backward
compatible
solution
and
the
only
solution
we've
come
up
with
so
far
involves
tags,
and
so
there's
going
to
be
a
you
have
to
do
a
tag
listing
query.
All
the
tags
out
there
find
your
details
find
your
spam,
find
your
signature
find
all
the
data
when
you
introduce
the
version
3
or,
however,
we
do
the
upgrade
to
the
newest
release
of
whatever
the
new
feature
is.
B
B
D
Yeah,
so
you've
got
blobs
and
you've
got
manifest
and
they
get
treated
differently
by
the
registry,
and
so
blobs
are
opaque.
We
don't
look
inside
or
we
shouldn't.
I
think
right.
Some
registries
do
if
they,
if
they
know
what
the
blob's
supposed
to
be,
but
we
usually
treat
the
blob
as
opaque
and
it's
just
strange
bites,
but
the
manifest
the
registry
is
looking
at
because
it's
going
inquiring
saying
how
are
this?
How
is
this
manifest
associated
with
what
blobs
and
so
okay.
B
D
Yeah,
well,
that's
something
that
all
registries
do
today,
they'll
see.
Okay,
this
is
an
index
manifest
versus
this
is
an
image
manifest
and
we're
going
to
introduce
later
on
this,
an
artifact
manifests-
and
that
says
here
all
the
blobs
that
I'm
associated
with
or
hear
the
other
manifest
that
I
point
to,
and
so
it
needs
to
know
how
to
parse
that
manifest
and
how
to
look
up
all
the
data
and
if
we
start
putting
new
data
in
that
manifest
that
they
haven't
looked
at
before.
G
G
Small
and
big
have
v2
as
a
kind
of
like
a
point
in
their
configuration,
english
controllers
and
whatnot
that
this
is
where
we
jump
off
and
say
this
is
everything
under
slash?
V2
is
docker
if
we
introduce
something
like
slash
v3,
that
is
a
a
pretty
significant
change
is
what
was
discussed
at
that
point
in
time
for
multiple
operators.
So
it's
it's
non-trivial
is
what
I
can
say
if
you
wanted
to
add
something
like
slash
e3.
G
So
if
we
can
get
reference
types
working
on
the
slash
v2,
that
would
be
really
really
great,
but
I
I
don't
think
we
should
not
not
discuss
whether
v3
is
is
an
option.
That's
what
one
of
the
proposals
did,
but
I
would
strongly
encourage
if
we
can
stick
under
v2,
because
blobs
manifest
apis
and
all
those
apis
don't
need
to
change
and
introducing
one
new
api
doesn't
probably
require
a
full
major
version
bump
from
both
routing
as
well
as
schemas
is
probably
one
thing
that
might
help
the
discussion.
G
D
Yeah
as
much
as
we
all
throw
up
our
hands
and
scream
and
say
no
as
soon
as
someone
says,
to
change
your
version
number
and
I'm
definitely
one
of
them.
I
think
the
underlying
question
that
nisha's
getting
at,
though,
can
still
be
solved.
The
way
that
you
and
I
are
looking
at
sanjay
of
hey,
there's
a
new
api
available,
something
changed.
Okay,
we
we
realized
that
the
registry
has
new
capabilities
and
then
it's
just
a
question
of
now
is
there
that
need
for
re-indexing
and
that
kind
of
question.
D
B
Yeah
so,
at
least
from
my
point
of
view,
what
I
am
saying
is
that
folks
would
like
to
be
able
to.
B
You
know,
make
a
very
clear
distinction
between
the
changes
that
we
are
adding
with
this
and
that
and
whatever
is
existing,
it
feels
to
me,
like
you
know
in
that,
if
you
think
about
it
that
way,
proposal
e
makes
sense,
but
without
the
upgrade
path.
B
So
it's
I
mean
it.
It
feels
to
me
like
that's
that's
what
the
the
folks
at
docker
want
is.
You
know,
allow
their
allow
folks
who
are
who
have
adopted
docker,
to
continue
using
docker
the
way
that
they've
used
it
before
and
any
new
adopters
will
not
stomp
on
any
of
those
users.
If
they're
using
the
same
registry.
D
What
do
those
features
look
like,
and
I
think
that's
where
we've
been
going,
because
so
many
of
the
questions
on
the
reference
types
have
been
saying:
hey,
it's
tags
are
horrible.
We
don't
like
that
solution
and
I'm
I'm
in
that
camp,
but
they
get
us
at
least
partway
there
today,
and
so
then,
the
next
step
is:
where
does
it
look
like?
If
we
go
the
next
step
go
ahead,
saji.
G
B
A
I
would
I
would
like
to
move
on
getting
brandon
and
sanjay
into
image
spec.
So
if
we
can
get
nisha's
thing
and
then
I
I
can
re-nominate
or
someone
can
nominate
and
then
you
know,
I
can
click
a
green
button
and
I
don't
know
if
we're
all
on
the
same
page.
B
Yeah,
I
I'm
I'm
fine
with
that.
Do
you
want
me
to
make
the
change
now.
A
If
you
do
it
today,
just
ping
me-
and
I
don't
even
have
I
don't
even
have
rights
on
that
repo.
So,
okay.
B
Okay,
bye-bye
bye,
I
might
as
well
leave
too.
D
G
H
H
Work
on
it,
I
mean
like
not
working
it
on
the
plane
is
also
like
you
know,
a
time-honored
tradition.
I
understand
yeah.
I
am
currently
tracking
towards
being
at
open
source
summit
north
america
in
austin,
as
well
so
hoping
to
see
some
of
you.
There.
E
H
E
In
june
amy
correct,
oh
boy,
I
know
faster
than
you
might
expect,
who
knew
I'm
still
trying
to
submit
talk
for
kubecon,
north
america,
yes,
and
that
deadline
is
actually
extended
as
well.
Yes,.
E
H
H
Okay,
I,
I
guess
I'll
also
yeah,
actually
use.