►
From YouTube: OCI Weekly Discussion - 2022-10-13
A
A
A
A
B
B
B
Me
neither
but
I
think
Amy
is
going
to
have
a
virtual
option
for
us,
so
I
think
we're
going
to
be
about
50
50..
B
A
B
F
Yeah,
hopefully
this
is
mostly
non-controversial.
You
said:
Famous
Last,
Words
Microsoft
isn't
going
to
use
these
anymore
for
new
images.
It
sounds
like,
and
we
think
that
no
one
else
is
using
them
for
anything
or
should
use
them
for
anything.
So
I
think
we
should
making
it
make
an
explicit
note
that
these
are
not
something
you
should
use
unless
you're
Microsoft
Circa
2021.
C
Back
to
the
URLs
field,
but
I've
I've
definitely
seen
people
use
this
in
experiment
where
they're
like
constructing
images
that
don't
actually
exist
and
like
pointing
to
GCS
buckets
for
yeah.
You.
F
F
Blahs
I
could
read
out
the
actual
thing
that
say
that
they
have
to
get
them
from
the
source
and
I
feel
like
if
you
are,
if
you
are
not
Microsoft
and
you're
doing
that,
I
feel
like
we
would
have
heard
of
you
by
now,
if
you're,
using
URLs
and
and
non-distributable
as
a
trigger
for
you
go
look
at
the
urls,
that's
like
a
different
thing,
but.
C
B
C
F
Brandon
I
know
that
you
had
a
change
for
sparse
indexes.
I,
don't
remember,
seeing
anything
for
sparse
images
is
sparse
images,
also
something
that
we
support
or
like
documented.
B
So
the
index
wasn't
even
me,
it
was
somebody
else,
but
take
the
credit
man
and
yeah
yeah
I'll.
Take
all
that
credit
for
that,
but
the
change
that
we
made
was
to
actually
document
that
Registries
can
reject
it.
It
wasn't
saying
that
Registries
can't
accept
it,
because
Registries
could
always
accept
it,
and
so
it
was
interesting.
Having
us
come
back
and
say:
well,
we
can't
say
that
people
can
recheck
this
stuff
because
we
never
said
they
couldn't
so
now
we
need
to
go
back
and
say
Here's
the
return
code.
F
Yeah
I,
don't
think
we
in
this
group
have
enough
Mojo
to
just
make
the
change
on
our
own
anyway,
but
I'm
hearing
open
a
PR
ping,
some
red
hat
folks,
ask
your
friends
check
your
Halloween
candy
for
any
non-distributable
layers
and
we'll
see
what
happens.
B
Yeah
I've
the
reason
I'm
asking
Steve
about
the
legal
implications
of
it
is
because
I've
implemented
this
in
reg
client
in
the
past
that
you
can
modify
an
image
to
take
the
foreign
layers
out
and
just
copy
the
blobs
over
like
a
regular
copy
command
and
that
works
unofficially
I'm.
Just
looking
for
the
legal
obligation
there
is
somebody
causing
all
kinds
of
end
user
license
agreement
issues
if
they
were
to
actually
use
that
feature.
B
Oh
never
but
yeah
that
well
it
wasn't
distributable
after
I
was
non-distributed
after
I
was
done
with
it,
but
yeah
so
features
are
there
that
could
be
used
for
this
stuff
and
I'm.
Just
wondering
in
terms
of
other
people
want
to
go
down
a
path.
The
the
tweet
from
Justin
that
I
highlight
in
the
comments.
Was
him
basically
saying:
hey
we've.
We
turned
this
off
for
everything,
but
Microsoft
images
and
Docker
hub.
No
one
complained.
B
F
F
Okay,
great
the
other
topic
I
have
on
the
list,
is
potentially
more
discussion,
so
I'm
happy
to
move
on
and
I'll
open
a
PR
that
we
can
all
argue
about.
Asynchronously
offline.
F
D
About
companies
that
want
to
keep
their
layers
proprietary.
D
D
You
know
this
stuff
with
you,
but
we
have
it
and
we
would
like
to
be
able
to
use
it.
F
D
F
If
you've
heard
from
folks
that
would
potentially
benefit
from
this
feature
sticking
around,
then,
whenever
we
open
the
pr
to
get
rid
of
them
or
discuss
getting
rid
of
them,
if
you
could
ping
them
in
and
say
like
this
might
be
going
away.
Is
that
okay.
D
They
probably
will
not
reply,
because
the
way
that
they've,
like
interacted
with
me,
is
basically
through,
like
an
open
source
tool
that
they're
using
internally
to
go
and
look
at
their
layers,
but
I
can
also
think
of
a
situation
where,
like
my
own
company
who's
working
with
the
some,
you
know,
customers
that
would
like
to
that
are
typically
very
careful
about
where
their
data
goes
and
how
their
data
is.
Shared
would
be
interested
in
this
feature.
C
D
Yeah,
it's
it's
one
of
those
things
where
it's
like.
Well,
the
spec
says
non-distributable.
So
if
you
wanted
to
build
a
client
that
reads
that
data,
then
you
could
to
take
advantage
of
it,
but
I
suppose
it
can
be
encoded
in
some
other
way.
D
Yeah
so
I'm
I
mean
yeah.
There
are
ways
around
it,
but
it
would
be
nice
if
it
was
in
a
spec,
something
that
is
reasonably
stable.
D
Well,
so
consider
like
a
layer
that
is
on
an
internal
site
which
they
want
to
share
within
their
network,
but
do
not
want
it
to
go
outside
to
another.
Like
organizations
Network
and
they're
Distributing
the
clients
they're
sharing
the
client
like
in
their
like
whole
organization,
then
they
want
some
ability
to
tell
the
other
organization:
hey
I,
don't
want
you
I,
don't
want
you
pushing
like
failure
to
push
this
image
to
your
registry,
because
our
registry
stuff,
that's
in
our
registry,
doesn't
go
outside.
F
Rather
than
like
invent
a
scenario
where
non-distributable
layers
might
be
helpful,
I
like
I,
think
to
John's
Point
non-distributable
layers
are
maybe
a
third
of
the
solution
to
the
thing
you're,
describing
in
real
life
and
and
I
think
that's
a
good
reason
to
get
rid
of
them
right,
like
I,
think
or
you
know,
deprecate
them
for
future
use.
F
If
you
want,
if
you
want
layers
that
can
never
Escape
your
network,
you
can
already
run
a
registry
entirely
within
your
network
and
solve
it
that
way,
or
you
know,
this
is
sort
of
a
hybrid
solution
that
is
very
I,
think
Microsoft
use
case
specific,
so
I
would
rather
not
proliferate
more
uses
of
them
when
the
only
one
we
know
of
is
going
away.
D
D
F
Yeah,
the
non-descript
non-distributable
layers
are
not
a
security
feature
like
it
tells
you
where
to
go,
get
them
and
anyone
who
can
see
where
to
go
get
them
can
go,
get
them
so
they're,
not
like
like
an
access,
control
or
firewall
thing.
We
have
Registries
that
are
off
capable
and
private
Registries
for
that
yeah.
So
I'm,
gonna,
I'm
gonna
take
the
action
item
to
go
file,
an
issue
file
or
sorry
file,
a
PR
with
specific
language.
We
can
argue
over
and
then
ping.
F
All
right,
the
next,
the
next
thing
I
was
talking
to
some
wasm
folks.
The
world
of
wasm
is
crazy,
y'all.
They
they
are
packaging,
their
wasm
I'm,
going
to
call
them.
Binaries
was
in
binaries
in
oci
images,
which
is
good
they're
using
they're,
reusing,
existing
Technologies
to
package
and
distribute,
and
they
should
be
rewarded
Unfortunately.
They
are
only
building
single
platform,
images
and
they're
lying
about
their
OS
and
architecture
being
Linux
amd64.
In
all
cases.
F
F
F
So
we
talked
about
it
and
the
current
guidance
for
picking
an
OS
and
architecture
is
very
go-centric.
It's
explicitly
like
we.
We
recommend
the
ones
that
go
says
to
use
for,
go
OS
and
go
Arch.
F
The
equivalent
for
go
OS
and
go
Arch
would
be
osjs
and
Arch
wasm,
but
that
doesn't
make
a
lot
of
sense
in
wasm
land,
because
JS
JavaScript
is
not
like
involved
at
all
in
in
modern
wasm
usage.
So
going
back
and
forth
with
them
a
little
bit.
They
would
like
to
do
better,
but
they
don't
want
to
use
the
OS
of
JS
to
match
go.
F
Instead,
they
would
like
the
OS
to
be
wesi
wasi,
which
is
sort
of
the
interface
between
wasm
code
and
platform
code
like
an
OS,
would
kind
of
be
and
basically
explicitly
diverge
from
the
go
recommendation.
When
the
architecture
is
wasm,
they
also
have
a
couple
of
variants
that
they
might
like
to
push
which
I
think
we
should
just
add
to
the
list
of
variants
in
the
spec
and
say
when
the
arc
is
wasm,
you
can
use
the
variance
span
or
slight
or
made
up
thing
in
the
future.
F
B
C
I
convinced
Phil
SDS
to
stop
doing
validation
with
manifest
tool
around
the
go.
Os
go
Arch
stuff
because
it
was
Ill
like
would
reject
anything
that
wasn't
a
valid
combination
according
to
the
go
compiler
but
I
convinced
him
to
stop
doing
that,
and
that's
a
the
only
place
I
would
say
this
is
a
problem
shouldn't
be
anymore
as
far
as
I'm
aware,
but
yeah,
it's
definitely
possible.
Other
people
are
doing
this.
F
The
spec
says
supported
I
I
should
look
at
the
actual
text,
but,
like
basically
supported
values,
are
those
supported
by
the
go
tool,
notably
absent
from
that
is
the
go
tool
today,
which
made
tomorrow,
add
50
new
combinations
of
go
OS
and
go
Arch
that
you
may
not
learn
about
so
I
think
I.
Think
in
general.
It's
a
good
idea
to
be.
You
know
not
do
that
validation
and
just
expect
anything,
but
yeah,
sometimes
folks
are
going
to
be
overly
strict
to
their
detriment.
F
I
think
that
I
talked
about
this
with
John
yesterday
a
bit,
and
he
convinced
me
there
is
a
case
where
we
might
need
to
be
careful,
John
I'm,
going
to
plagiarize
you.
Let
me
know
if,
if
I'm
about
to
lie
so
go
does
not
have
good
wazzy
support
today.
Wazzy.
Is
this
like
interface
layer,
but
beneath
wasm?
That's
new
and
go
doesn't
know
anything
about.
F
If
we
decide
that
in
OC
Island,
the
OS
is
wazzy
and
then
in
the
future,
go
learns
more
about
wazzy
and
says
that
their
OS
is
actually
blah
blah,
not
wazzy
or
something
you
know,
some
made
up
giraffes
or
some
other
made
up
string,
then
we'll
be
in
a
case
where
we
might
want
to
support.
We
might
want
to
remove
our
special
case
of
wasm
and
say
you
should
use
the
Go
Value,
but
it's
different
than
the
one
we
had
before.
So
we
may
I
think
there's
a
way
out
of
that
corner.
F
We've
painted
ourselves
into
in
that
case,
but
I
first
of
all,
I,
don't
think
it's
likely
I
think
based
on
like
reading
some
issues,
they
would
probably
go
with
wazzy,
especially
if
we
had
the
precedent
and
we
can
paint
ourselves
out
of
that
corner.
If
we
want
to
by
saying
it
can
be
either
wazzy
or
giraffes
whatever
the
go
value
is.
B
F
Creates
an
OS
of
Wazi
that
is
not
wasm
yeah
or
that
is
not
compatible
with
our
idea.
It
wasn't
today,
yeah
I,
think
I
think
that's
also
possible.
I
mean
in
in
this
theoretical
sense,
not
in
the
like
practical
sense,
yeah,
it's
sort
of
hard
to
hard
to
decide
right,
I,
think
that's
roughly
on
par
with
them,
redefining
the
semantics
of
you
know,
risk
v64.
F
F
I'm
also
fine,
if
the
outcome
of
this
is
this
seems
fine,
let's
go
build
some
images
with
this
combination
of
stuff
and
not
specify
it
for
a
little
bit,
but
like
at
least
just
wreck,
you
know
like
off
the
Record
recommend
that
they
build
their
images
that
way,
and
then
with
the
idea
that
eventually
we
would
specify
it
that
way.
We
can
go
test
it
and
make
sure
it
works
and
see
where
it
doesn't
and
stuff
like
that,
but
with
the
idea
that
we
would
eventually
specify
that
so
that
everyone
understands.
F
I,
don't
know
where
that
reputation
comes
from,
but
yeah
I
guess
I
was
a
little
also
surprised
to
see
that
go
is
so
far
behind
on
wasm
stuff,
like
the
you
know,
rust
and
typescript
seem
to
be
really
miles
ahead,
but
yeah
I
don't
know.
Speaking
of
trailing
specs.
F
Technical
is
better;
they
they
have
wazzy
support.
Wazzy
is
still
like.
You
know
on
the
way.
It's
not
it's
not
here
yet,
but
I
think
sort
of
building
up
some
more
specified
scaffolding
around.
It
will
help
them
too.
Things
like
this
perfect.
F
Okay,
I'm
I'm,
hearing
I
should
go
back
to
them
and
say
oci
folks,
nodded
politely
at
me,
while
I
talked
you
should
go,
make
your
images
this
way
and
then
eventually
I'll
send
a
PR
to
actually
specify
that
cool.
Thank
you
for
your
polite
knobs.
Everyone.
F
C
F
No
because
they're
not
like
you,
know,
CPU
architecture
variants,
they
are
dialects
of
wazzy
or
dialects
of
that.
That
interface.
C
C
F
Yeah
I
think
I
think
the
variant
thing
is
probably
the
least
urgent
to
specify
or
the
least
urgent
to
sort
of
Chase
down,
and
maybe
we
don't
specify
it
for
a
little
while,
until
this
answer
or
like
this
question
is
answered
and
maybe
the
variance
become
compatible
with
each
other,
and
now
we
don't
have
to
have
variants
at
all
sure,
but
yeah,
that's
that's
I
did
not
I
did
not
scroll.
All
the
way
down
to
this
page
could
work
for
scrolling
all
the
way
down
to
the
end
of
the
page.
B
Every
patch
response
comes
back
and
says:
here's
a
location
to
send
your
next
request
to
every
time
you
send
your
one
patch
request,
and
it
says:
okay,
send
your
next
chunk
over
this
next
URL
and
in
there
it's
embedding
things.
There's
a
uuid,
the
docker
embeds
into
it.
There's
also
some
state
that
gets
embed
into
there.
If
you're
interrupted,
you
don't
know
what
the
new
state
is
and
suddenly
you
can't
continue
and
recover
from
an
interrupted
Connection
in
there.
At
that
point,
at
least
that's
what
I
believe
I'm
running
into
that
may.
C
B
Resume
it's
so
it's
not
enough
for
me
to
resume
I,
don't
think,
because
I
tried
that
didn't
work,
but
it's
also
not
to
find
another
CI
and
in
the
docker
spec
they're
telling
you
to
go
in
there
and
start
tweaking
removing
parameters
stuff
like
that.
Just
querying
the
uuid
and
not
passing
the
State
field
at
all
and
then
there's
the
eye
spec.
B
F
B
F
B
C
B
B
C
B
C
B
I
I
did
see
the
right
response.
That
I
was
expecting
to
see
which
was
ghcr,
came
back
and
said:
hey
you
have
an
invalid
range.
Here's
the
correct
range
use
this
instead
and
that's
that's
what
I'd
like
to
see
if
I
sent
a
location
and
the
location
says,
Hey
I've
already
got
a
chunk
of
this.
Here's
where
you're
supposed
to
be
starting
from
great
I
can
work
with
that.
B
B
Next,
one
on
my
list,
just
kind
of
chugging
on
through
them
here
feel
free
to
slow
me
down
if
I'm
going
too
quick
on
any
of
these,
our
netlify
image
apparently
needs
to
be
updated.
B
A
A
B
A
E
Yeah
I
think
it's
probably
just
clicking
a
thing,
but
I'm
not
sure
it
might
just
be
like
Amy
for
those
access.
A
B
It's
been
done
thanks,
Josh
last
one
on
my
list
was
the
platform
config
949.
B
This
is
a
fun
one.
This
ought
to
get
some
conversation
going.
The
platform
in
the
config
Json
is
currently
I'm
going
to
share
a
window
on
this.
One
move
some
things
around
real
quick
window,
move
that
over
just
case,
someone
wants
to
look
at
it
and
you've
been
doing
terrible
job
sharing
screen
so
far,
so
the
config
itself
today
has
all
the
individual
variables
in
there
and
Sebastian
came
and
said:
hey
we've
already
got
a
go
variable
for
that.
We
just
use
this.
It's
instructed
automatically
expands,
since
we
didn't
give
it
a
name.
B
You
reference
everything
exact
same
way:
it
works.
It
makes
no
changes
to
the
spec.
It
makes
no
changes
to
the
Json
marshalling
or
on
marshalling
code.
The
only
thing
that
breaks
is
potentially
anybody
that
initialized
something
in
their
go
where
they
defined
it
and
actually
passed
the
values
in
there.
B
You
know
they
put
the
individual
variables
in
you
have
to
pass
them
like
that
or
Sebastian's
going
off
and
he's
made
a
whole
lot
of
PR,
so
other
projects
that
were
doing
it
with
small
changes,
see
if
I
can
pull
one
of
these
up
and
show
you
what
it
looks
like
pretty
simple:
all
he
did
was
pull
out
wherever
someone
was
setting
in
one
place
and
he
would
take
out
the
platform
setting
in
one
place
and
just
set
it
somewhere
else
or
if
someone
was
defying
the
OS
in
there.
He
just
defined
it.
B
So
a
couple
of
places
this
he
actually
did
it
the
opposite
direction,
because
he
had
the
whole
platform
variable
a
lot
of
places
there
saying
the
OS
and
he
just
pulled
it
out
from
being
said
individually
set
as
a
as
part
of
the
Constructor
here,
instead
of
setting
it
in
the
instructor,
he
actually
pulled
it
out
to
a
separate
place
down
below
instead
of
down
below.
So
it
is
a
breaking
change.
F
B
E
So
that
was
that
was
like
the
this
PR
was
a
response
to
a
PR
I
made
to
container
d,
where
container
D
was
copying
platform
Fields,
but
only
OS
and
architecture.
So
it
was
quietly
dropping
the
variant
and
I
noticed
because
I
had
a
V6
image
that
was
now
claiming
to
be
V7.
Due
to
the
platform
normalization.
E
F
E
Me
yeah,
we've
in
within
the
oci
I,
can't
remember
if
it
was
runtime
spec
or
image
spec,
but
we've
had
this
conversation
before
and
the
conclusion
was
that
the
spec
version
is
for
the
spec
itself
and
for
breaking
change
in
the
spec
itself,
not
for
the
go
code
and
that
it's
unfortunate
that
go
insists
that
the
version
number
applies
to
both
and
treats
it
as
if
it's
semantic
for
both
but
not
much.
You
can
do
there.
B
E
B
It
was
kind
of
a
breaking
if
I
thought
about
her
I
might
have
held
off
on
hitting
the
approve
button
right
away.
But
let's
see
anybody
else
here
that
John
you're
on
the
list
of
maintainers.
What
do
you
think.
C
B
C
B
C
A
B
There
was
one
actionable
item
out
there,
which
was
up
there
at
the
top
942,
that's
also
actual
by
me
to
go
through
and
do
a
quick
review
on
the
artifact
schema
stuff.
Sanjay,
don't
fly
off
the
road
I'll
cover
it
for
you
yeah,
so
he's
got
the
change
out
there
for
adding
a
bunch
of
schema.
Json
updates
artifact
go
test,
just
all
the
test
card
that
we
normally
need.
B
Whenever
we
add
a
change
like
that,
we,
through
this
game,
changing
without
throwing
the
test
tennis
now
we're
going
back
and
fixing
that
so
I
think
it
looks
good
to
me.
I
just
need
to
go
to
and
verify
it
and
I
just
wanted
to
point
out
here
to
get
another
set
of
eyes
on
there,
since
we
need
a
second
thumbs
up
on
it
before
you
merge
it.
B
B
Going
once
going
twice
sold
all
right
have
a
great
rest
of
the
week.
Everybody
and
yeah
get
ready
for
the
kubecon,
see
you
there
take
care
bye,
guys.