►
From YouTube: OCI Weekly Discussion - 2021-08-04
Description
Recording of the weekly OCI developer's call from 4 Aug 2021; agenda/notes here: https://hackmd.io/El8Dd2xrTlCaCG59ns5cwg#August-4-2021
B
Let's
see
looks
like
we
have
a
good
crowd
already,
so
I
just
posted
the
hack
md
in
the
chat.
We
have
a
few
few
agenda
items,
but
you
know
if
you
want
to
chat
quickly
about
oss,
north
america.
A
Good
fun
realistically
like
so,
we
have
space
reserved
for
tracking
for,
like
100
or
so,
and
I'm
thinking-
that's
probably
going
to
be
enough
based
on
like
how
many
people
are
here
and
we'll
be
looking
at
september,
30th
at
open
source
summit,
north
america
and
really
my
question
for
this
group
is
as
far
as
structure
and
content
and
things
do
we
want
to
be
able
to
do
a
cfp.
Do
we
want
to
just
kind
of
like
figure
things
out
in
here?
A
The
cfp
call
for
proposals
like
do
like
an
actual
like
full
agenda
and
do
like
you
know,
here's
real
talks
and
all
of
that
or
we
can
look
at
being
able
to
do
like
more
kind
of
discussions
in
the
room.
B
Yeah,
I
mean
my
my
perspective,
which
means
obviously,
hopefully
there
are
others.
So
this
is
just
me,
but
you
know
conferences
give
us
plenty
of
chances
for
kind
of
formal
talks
on
various
topics.
So
I'm
assuming
this
group,
you
know,
I
think
a
majority
of
the
time
could
easily
be
reserved.
For
you
know
almost
an
unconference
type,
you
know
spend
the
first
few
minutes.
People
write
ideas,
people
vote.
What
do
they
want
to
talk
about
that?
That's
my
that's
my
thought.
A
C
A
Right
now
we
have
it
set
for
in
person.
Obviously
we
can
work
on
being
able
to
pivot
to
virtual
as
needed
or
being
able
to
do
like
a
hybrid
piece
in
here
again
as
needed
right
now,
I'm
tracking
for
being
able
to
say
we
have
a
room.
We
have
coffee
go
to
town.
B
Yeah
I
mean,
I
think
this.
The
reason
it's
set
up
this
way
is
because
there's
an
actual
conference
that
is
attempting
to
be
in
person
that
week
so
clearly
that
could
change
or
who
knows
I,
I
don't.
Obviously
I
have
no
no
role,
no
role
in
the
conference
organization,
but
I
guess
what
I'm
saying
is
like.
I
think
nobody
knows
what
the
end
of
september
looks
like.
So
we're
just
forging
ahead.
B
D
A
C
E
I
think
I
think
the
the
the
gods
are
not
not
with
us
for
september.
A
B
Sounds
good
and
yeah
anisha
to
your
point
I
mean
obviously
we'll
we'll
attempt
to
the
best
of
our
ability
to
make
allowance
in
case
there.
It
ends
up
being
some
kind
of
hybrid
thing,
but
mike
brown
will
be
taking
pictures
and
like
doing
live,
live
stream
of
a
whiteboard
or
something.
B
We'll
make
it
work,
my
phone
will
travel
right,
yeah
cool
all
right,
so
I
don't
swell
the
second
there's
two
agenda
items
listed
here.
B
H
H
B
B
H
The
if
I
can
summarize
the
issue,
I
think
the
latest
status
on
that
issue
was
someone
should
recommend.
The
specs
should
recommend
that
registries
and
clients
enforce
some
maximum
size
on
a
manifest,
because
for
safety,
for
you
know
to
to
reduce
the
ability
to
get
dosed
by
it
that
the
spec
probably
shouldn't
state
a
it,
could
recommend,
but
shouldn't
require
a
value
like
I
think
in
the
issue.
It
was
like.
10
gigs
might
seem
like
a
huge
manifest
today,
but
might
seem
quietly
small
in
10
years.
H
If
this,
if
people
in
general
nod
at
this
fact,
I
will
gladly
take
this
to
a
change
to
the
spec
to
recommend
this
enforcement
without
a
record
without
a
specific
recommendation
for
what
that
value
should
be,
and
we
can
sort
of
discuss
it
in
more
concrete
terms
where
people
nodding
turns
into
people
approving
turns
into
changing
the
spec.
H
Just
manifest
blobs,
it's
a
separate
issue,
we
might
consider
to
say
there
is
like
registry-
should
enforce
some
maximum
block
size,
but
I
think
the
specific
issue
is.
I
can
already
today
push
a
completely
valid
manifest
that
is
arbitrarily
large
bytes
by
the
question
in
the
in
the
chat
was
size
limit.
How
bytes
are
entries
total
byte
size
of
the
manifest?
So
I
can
push
it.
I
can
push
it
completely
because
annotations
are
like
magic
bag
of
bytes.
H
I
could
push
a
completely
valid,
manifest
that's
100,
gigabytes,
large
and,
if
you're
fine,
with
accepting
that
you
should
at
least
know
that
you're
fine
with
accepting
that.
H
F
H
This
size,
where
the
the
minimum
is
something
we
would
recommend,
based
on
current
survey
of
current
implementations,
maximums
the
minimum
of
all
of
those
maximum
sizes.
F
Right
so
we
want
so
we
want
to
push
registries
up
as
reasonable
as
possible
and
push
clients
down
to
a
reasonable
place
and
so
yeah
yeah.
It
depends
on
who's
perspective
if
it's
a
minimum
or
a
maximum,
and
I
think
that
we
talked
about
four
megs
as
being
the
current
problem
like
limit
of
docker
distribution
or
distribution
distribution.
The
current
practical
limit
that
we.
H
Have
today
yeah
if
you
operate
a
registry
or
a
client
and
are
surprised
that
a
four
megabyte
manifest
is
to
be
considered
normal,
speak
soon
or
speak
on
the
pr
that
I
will
send
to
make
this
a
recommendation
as
a
size.
I
think
four
megs
is
a
perfectly
reasonable
size,
but
you
might
disagree.
E
It
feels
right
to
have
some
some
table
of
what
those
limits
are
currently,
I'm
just
not
sure
if
it
makes
sense
to
to
make
them
rules
or
restrictions.
H
C
H
Yeah,
I
think
that's
a
good
idea
for
for
informing
the
value
that
we
put
into
the
recommendation
in
the
spec.
I
don't
want
the
spec
to
include
a
table
that
falls
out
of
date
quickly,
but
in
the
discussion
of
the
of
the
thing
we
put
into
the
spec,
we
should
include
empirical
data
collected
from
a
survey
of
registries.
H
Is
there
a
way
for
the
client
to
query
the
implementations
max
size
before
attempting
a
push?
There
is
not,
I
think
this
actually
came
up.
Maybe
it
came
up
in
that
issue
in
that
issue,
but
I
don't
think
we
want
to
add
a
new
endpoint
to
the
distribution
spec
to
get
capabilities,
because
that
seems
I
mean.
H
Maybe
we
do
and
if
we
do
there's,
maybe
you
know
a
half
a
dozen
other
things
that
a
registry
could
say
they
support
up
to
this
many
layers
in
an
image
or
up
to
this
many
manifests
in
a
index
or
something,
but
that
that
feels
like
a
bigger
amount
of
work
than
what
we
want.
At
least
I
think
so.
I'm
curious
what
other
people
think
as
far
as
if
you
operate
a
registry
and
are
fine
with
having
a
capabilities.
Endpoint.
F
What
do
you
think
I
would
lean
towards
just
maybe
I
should
in
terms
of
the
response
code,
it's
like
whatever
it
is
like
419.
It
says.
Oh,
this
is
too
large
yeah.
I
don't
know
about
an
actual
api,
but
I
don't
know
because.
H
God,
oh
so
so
then
this
would
be
a
yeah.
This
is
a
distribution
spec
issue,
not
an
image
spec
issue:
it's
not
about
the
validity
of
a
10
gig
image
index,
manifest
it's
about
the
ability
to
push
or
pull
those
which
is
distributions
by
gotcha
yeah,
and
we
should
we
should
say
when
it's
over
that
size,
you
should
respond
with
419
or
whatever
too
big.
H
H
I
think
that's
an
issue
and
as
far
as
sort
of
trying
to
feel
out
the
size
limits
on
things,
I'm
probably
not
going
to
go
much
further
than
4meg,
because
I
really
don't
want
to
page
someone
the
other
issue.
The
other
item
in
the
agenda
is,
I
wrote
a
little
tool
to
test
whether
registries
accept
the
data
field.
H
Field,
so
if
it
blows
up,
I
don't
want
to
page
somebody
anyway,
yeah
the
other.
The
other
item
was
here's.
A
handy
tool
to
generate
arbitrarily
large
manifests
and
push
them
to
places
using
the
data
field.
H
Oh
awesome:
well,
I
should
have
tried
harder.
Did
you
use
a
bicep?
No,
I
literally
just
kept
adding
zeros
until
I
felt
like
the
number
was
too
large
and
I
didn't.
H
Maybe
that's
how
they
set
the
value
themselves,
so
so
john,
even
though
gcr
does
support
10
megs,
you
would
still
recommend
that
registry
shouldn't
should
enforce
a
limit
somewhere
around
four
megs.
F
Depends
on
if
you're
asking
me
the
client,
maintainer
or
me
the
registry
operator
or
me,
the
sky
person.
I
would
say
that
registry
should
support
whatever
they
can
support
and
return
an
error
if
they
can't
and
they
should
support
at
least
four
meg
or
whatever.
H
We
decide
on
that
makes
sense.
That's
a
good
yeah.
I
think
that's
a
good
compromise
that
gives
maximum
specificity
and
generality.
H
In
on
the
topic
of
the
data
thing
of
the
data
field,
as
proposed
in
image,
spec
number
826
most,
I
have
yet
to
find
a
registry
implementation
that
doesn't
accept
the
data
field,
as
described
in
that
proposal.
If
you
know
of
one
or
can
think
of
one
or
want
to
test
it
against
one
and
see
if
it
works
or
doesn't
work,
that
would
be
useful
information.
H
But
I
think
there
was
some
concern
in
the
data
field
proposal
that
this
might
be
it
non-uniformly,
supported
across
all
registries
and
so
far,
I've.
Yet
to
find
data
to
that
effect.
H
Yeah
yeah
yeah,
exactly
I
mean
they,
the
spec
says
this
should
work.
We
think
it
works.
I
have
now
data
that
it
proves
that
it
works,
but
if
we
find
someone
that
doesn't
maybe
that's
enough
to
hang
up
the
data
proposal
but
the
ones
that
I've
tried
account
for
all
of
the
ones
I
could
think
of
and
some
I
couldn't
think
of
that
other
people
tried.
So
I
think
we
are
good
on
the
data
field
being
safe
to
use
for
clients
and
servers.
H
D
How
can
I
avoid
paging,
john
and
testing
this
data
limit.
H
I
think
that
john
would
say
that
they
have
all
the
safeguards
in
place
that
it
won't
actually
page
you
no
matter
what,
but
it
will
return
the
correct
response
and
not
actually
cause
problems.
I
can't
say
the
same
for
every
posse,
yeah
exactly
push
to
ghcr
and
you
won't
be
a
problem
or
push
to
you
know
and.
F
Yeah
page
someone
else
not
me
if,
if
it
pages
me,
it
should
have
paged
me
and
it's
a
bug
and
we
will
fix
it.
I
don't
mind
if
you
page
me,
but
it's
probably
not
me
getting
paid
honestly
yeah
yeah
yeah.
D
So
the
only
like
registry
ish
provider
that
I
worked
with
is
bundle.bar,
and
I
mean
I've
kind
of
taken
liberties
poking
josh
about
why
I'm
not
able
to
why
I'm
not
able
to
upload
blobs,
because
there's
like
a
data
limit
of
this
that
and
the
other-
and
that
usually
happens
when
I
try
to
upload
s-bombs
because
they
do
tend
to
be
large.
D
But
if
the
limit
is
just
for
manifest,
I
don't
really
I
mean
I
cannot.
I
don't
have
the
imagination
to
think
like
where,
in
the
manifest
can
people
add
metadata
about
the
s-bombs
and
what
kind
of
metadata
would
they
add
like?
I
don't
I
don't
have
I
mean
yeah.
I
cannot
think
about
anything
like
that,
but
I
have
seen
the
community
just
put
all
kinds
of
fields
that
to
cover
all
different
kinds
of
use,
cases
that
I've
never
thought
about.
D
G
D
Yeah
I'm
trying
to
avoid
that
yeah
putting.
G
D
D
Yeah
yeah
so
yeah,
that's
that's,
definitely
a
possibility
that
people
will
use
the
data
field
to
stick
a
very
giant
s-bomb
yeah.
H
You
mean
a
size
minimum,
not
a
size
limit
right
right,
a
minimum
recommended
a
minimum
recommended
maximum
manifest
size.
D
H
But
it's
not
a
minimum
size.
We're
not
saying
all
manifest
must
be
four
megs
or
larger.
We
can
pad
it
with
zeros
to
to
make
it
up
to
the
to
the
maximum
size.
No
clearly
there
is
some
wording
help
needed
here,
and
your
input
will
be
useful.
G
H
G
F
Click
right!
Yes,
that
that's
the
question
how
long
I
will
hit
merge
as
soon
as
enough
people
have
approved
it.
So.
H
I
don't
think,
there's
a
reason
not
to
try
it
out
before
it's
approved.
I
don't
know
how
long
it
will
take
to
get
approved.
It's
taken
a
while
to
get
it's
been
proposed
for
a
while
already,
so
I
assume
it's
not
going
to
get
approved
in
the
next
by
the
end
of
this
sentence,
but
you
know
I
don't.
I
don't
actually
remember
what
the
remaining
hang-ups
were
on
it.
H
I
think
it
was
in
general,
like
registry
support
and
the
size
limits
and
registry
support
check,
as
far
as
I
know
and
size
limits
tbd,
but
also
it's
it's
unrelated,
because
you
can
already
push
without
data.
You
can
push
a
four
and
a
half
megabyte
manifest,
so
we
should
have
that
limit
regardless
of
data.
E
H
D
Yeah,
I
I
was
going
to
ask:
what's
the
status
on
the
tob.
A
I
had
a
question
about
one
of
them
being
like
a
full
like
lgtm,
but
I
think
we're
good
to
go
here.
B
B
I
think
you
pointed
out,
you
know
it'd
be
nice
to
see
the
outcome
of
you
know
those
none
of
them
were,
you
know,
life-changing
updates,
but
there
were
there
were
kind
of
good
clarifications,
so
I
don't
know
if
steve
is
going
to
like
re-spin
that
before
merge
it'd
be
nice.
I
think
that
would
be
the
right
thing.
I
mean
we
have
the
lgtms,
but
if
he
can
provide
the
updates,
then
I
think
yeah.
I
think
we're
ready
for
that
version
to
so
I
think.
A
Part
of
the
issue
is
that
there's
a
lot
of
different
like
kind
of
feedback
kind
of
wandering
around
out
there
and-
and
I
think
the
last
pr
on
this
one
actually
had
a
pretty
good.
A
And
it
is
like
john
johnson
was
the
one
that
that
has
we
can
address
these
in
follow-up
prs
I'll,
just
drop
the
comment
in
here
as
well.
So
like
we're
all
kind
of
on
the
same
page,.
D
Yeah
I've
I
read
through
the
pr
and
I
had
nothing
really
to
contribute
because
yeah
I
had,
I
haven't
seen
anything
like
that's
been
implemented
or
implemented
successfully,
and
so
I
I
have
no
idea
whether
this
is
good
or
not.
F
Yeah
I
mean
we
also,
we
don't
have
to
merge
this.
I
mean
we
could
start
going
through
the
motions
as
if
this
were
merged
and
then
decide
if
it
is
good,
but
I
don't
know
if
they're
like
officially,
we
have
to
merge
something
before
some
process
can
begin.
D
I
think
that's
okay,
to
I
mean
that's
my
opinion.
I
think
it's
okay
to
merge
it
and
then
update
it
later.
If
we
find
that
you
know
something's
not
working.
C
C
I
I
reviewed
this
a
couple
times
and
I
think
my
biggest,
I
guess
like
concern
as
a
community
member,
if
you
want
to,
if
you
want
to
call
it
a
concern,
is
that,
like
the
structure
of
working
groups
like
this,
these
documents,
for
example,
provide
enough
structure
to
ask
to
say,
like
this
is
here
to
support
you
and
to
help
foster
like
collaboration,
as
opposed
to,
like
the
other
stance
that
where
it
can
take
where
it's
like.
C
We
are
like
the
rulers,
and
these
are
the
rules
you
must
follow
and
like
and
we're
watching
you
and
we're
going
to
dismantle
your
working
group
if
you
mess
up
so
like.
I
wanted
to
kind
of
make
sure
that
that
that
the
feel
of
the
document
didn't
go
in
like
that
direction
and
it's
not
to
say
anything
about
oci
but
like
there
tends
to
be
a
lot
of
like
I
don't
know.
I
thought.
C
D
I
think
it
would
be
a
good
idea
to
start
at
least
deciding
what
initial
working
groups
we
would
spin
off
at
this
time,
and
maybe,
as
vanessa
said
like
see
how
much
of
you
know
how
many
people
want
a
certain
thing
versus
another
thing
and
find
some
like
overlap
or
common
ground.
B
You
know
join
with
some
other
folks
with
similar
interests
and
do
that
so
that
the
t.o.b
at
least
the
current
to
be
hopefully
would
be.
You
know
nothing
like
what
vanessa
was
just
talking
about.
I
mean
we,
we
all
have
enough
work
in
our
own
plate
to
be
micromanaging,
a
bunch
of
groups
and
trying
to
tell
them
what
to
do
or
when
to
start
or
stop
like
the.
The
whole
idea
was
to
provide
enough
of
a
framework
that
people
with
common
interests
that
revolve
around
the
oci
specs
can
just
go.
B
So
you
know,
I
think,
like
john
is
saying,
maybe
including
me:
maybe
I've
been
too
hung
up
on,
like
you
know
getting
this
done,
and
then
you
know
having
people
propose
some
groups.
You
know
you
know
we
can.
We
can
say
now's
the
time
like
propose
some
group.
So,
let's
you
know
get
things
rolling
so
and
back
to
the
pr
itself,
like
I
think,
there's
just
like.
B
If
I
scroll
back
through
there's
a
bunch
of
kind
of
knit
like
things
that
maybe
we
can
just
summarize
for
steve
and
then
there's
a
few
kind
of
open
questions
like
I,
I
think,
if
we
just
summarize
that
maybe
that
helps
get
it,
you
know
kind
of
here's,
the
last
one,
two
three
things
and
and
that
kind
of
resolves
current
feedback.
Yeah,
that's
gonna,.
B
D
So
phil
you're
suggesting
that
somebody
could
like
currently
suggest
a
working
group
and
ask
if
they
could.
I
mean
I
guess,
as
you
said,
you
don't
want
this
to
block
spinning
off
a
working
group
so
like,
for
example,
if
I
want
to
say
like
hey,
I
want
to
start
a
working
group
on
container
supply
chain.
Transparency
short
for
s-bomb.
I
suppose,
then,
is
that
it.
What
do
I
have
to
do.
B
Yeah,
so
I
think
there
might
even
be.
B
Where
is
it
so
so
when
we
first
started
this
discussion,
you
know
steve
who
couldn't
be
here
today,
wanted
to
propose
one
around
reference
types
and
so
issue
96
and
the
tob
repo
has
you
know
I,
I
think,
a
rough
format
that
probably
is
worthwhile
for
others
who
want
to
propose
so
open
an
issue.
Here's
our
objectives,
here's
what
we're
trying
to
accomplish!
Here's
the
people
involved
and
yeah,
so
you
I'll
paste
that
in
the
chat,
so
other
people
can
see
what
I'm
talking
about
so
yeah.
B
I
think
that
step
one
is
just
you
know,
it
does
not
need
to
be
extremely
detailed,
but
kind
of
this
format
is
helpful
because
it
lets
everyone
else,
know
kind
of
what
what
you're
trying
to
do,
and
it
may
be
a
good
place
to
rally
others
who
are
interested
in
the
same
topic
to
say.
Oh
yeah,
I'd
like
to
be
involved.
B
And
then
I
guess
we
should
read
this
charter
pr
to
see
what
that
means.
The
team,
I
think
the
tob
just
sort
of
says
yep
sounds
good,
go
forth
and
that
that's
yeah
step
one
is
just
effectively.
Somebody
proposes
it.
B
There's
there's
some
defined
scope
and
some
goals,
and
we
all
say:
okay,
yep,
sounds
good
and
then
from
there
I
think
there's
a
lot
of
hopefully
lightweight
process.
We
just
need
to
figure
out.
You
know.
How
often
do
you
want
to
come
back
to
this
call
and
say
here's
our
update
or
where
are
you
going
to
keep
status
and
kind
of
some
working
docs
on
what
you're
up
to
but
that's
effectively?
I
think
all
that's
needed
to
get
going.
B
B
D
Say,
for
example,
like
currently
there's
a
proposal
for
reference
types,
but
you
are
coming
from
the
perspective
of
a
specific
use
case
that
may
leverage
reference
types
do.
B
I
mean
I
think
that
comes
back
to
do
the
scope
and
goals
of
the
of
the
current
proposal,
fit
with
what
you
and
and
others
are
trying
to
accomplish,
and
if
not,
then
maybe
it
becomes
a
discussion
about
whether
there
needs
to
be
a
different
proposal.
But
I
think
that's
why
each
working
group
needs
to
have
kind
of
a
clear
scope
and
set
of
goals
so
that
you
can
decide.
D
B
Yeah,
nothing
else
on
the
agenda,
anything
that
someone
wanted
to
cover
that
hasn't
come
up.
C
Yeah,
I
feel
like
a
broken
record,
but
I
I
still
haven't
figured
out
what
could
be
the
next
step
to
deploying
the
rfc
jekyll
template
for
the
oci
docs.
I
did
ping
vincent,
but
I
think
he's
busy
because
he
hasn't
responded
yet.
So
I
I
think
just
someone
in
a
position
of
power
to
like
create
repos
and
stuff
probably
needs
to
help
me
out.
C
B
Yeah
so
cool,
I
mean,
if
nothing
happens
by
next
week.
Let
us
know
I
I,
the
oci
is
one
github
org
that
I'm
never
quite
sure
who
has
powers
to
do
what
so,
but
we
can
figure
that
out.
If
we
need
to
when
in
doubt
ask
chris
thanks,
yup.
B
B
Cool
all
right.
Well,
I
think
we're
sounds
like
we're
good
for
today.
So
yeah
talk
to
everyone
next
week
have
a
good
one.