►
From YouTube: Carvel Office Hours - April 22, 2021
Description
Carvel Office Hours - April 22, 2021
We meet every 2nd and 4th Thursday of the month at 11:30am PT / 2:30pm ET. We hope you can join us live at the next one!
This week's office hours was a working group session centered around imgpkg-003 Rename images when copying bundles #22. Further details can be found in the agenda: https://hackmd.io/5Bh2IXwTShSrA0YdBY4AGg
A
Hi
everyone
welcome
to
this
week's
edition
of
carmel
office
hours.
It's
a
special
edition
where
we're
actually
taking
the
full
hour
to
do
a
just
sort
of
a
working
group
type
of
session
which
we'll
go
over
in
a
minute.
Just
a
reminder
to
please
adhere
and
abide
by
reading,
abide
by
our
code
of
conduct
listed
out
here
in
the
agenda
when
you're
attending
the
cargo
office
hours.
These
meetings
are
being
recorded
and
uploaded
to
our
youtube
playlist
within
48
hours
of
them
being
recorded.
A
If
you're
watching
this
from
home,
we
would
love
to
have
you
join
us?
Live
we
meet
every
second
and
fourth
thursday
of
the
month
at
11,
30
a.m,
pacific
time,
2,
30
p.m,
eastern
time,
and
if
you
aren't
able
to
make
it
to
the
office
hours,
we
do
also
meet
every
monday
for
carnival
community
meeting,
and
that
is
at
the
same
time,
11
30
am
pacific
time,
2
30
p.m.
B
B
So
this
is
for
image
package.
Sorry,
I
forgot
to
say
that
to
start
off
so
basically
image
package
when
it
copies
images
from
one
registry
to
another
registry,
it
copies
all
the
images
associated
with
the
bundle
into
the
same
repository
inside
the
new
registry.
So
all
the
images
are
part
of
the
same
repository
and
we
realize
that
that
causes
some
problems
to
to
our
users.
B
So
what
this
proposal
is
trying
to
address
is
precisely
that
giving
the
ability
to
the
users
that
are
caught
doing
this
copy
to
select
how
they
want
their
images
to
be
placed
on
the
destination
registry.
So
so
this
is
broad
strokes.
What
this
proposal
is
about,
and
we
we've
been
doing
some
feedback
on
it
for
more
than
a
month
now,
and
there
are
two
there
were
like
one
major
feedback
portion
of
it.
B
So
this
is
like
this
wasn't
part
of
feedback
that
we
rewrote.
Basically
the
majority
of
the
proposal
itself
to
try
to
address
this
and
see
if
it
makes
if
it
is
more
coherent
right
now.
So
that's
the
first
part
that
I'd
like
to
get
some
feedback
on
and
then
the
second
part
that
I'd
like
to
talk
is
a
different
proposal
of
how
we
are
doing
strategies.
B
But
before
we
do
that,
I'm
not
sure
if
everybody
had
time
to
read
through
the
proposal.
So
I
am
so
sorry
for
the
people
in
youtube
right
now,
but
I'm
gonna
propose
that
we
use
10
minutes
of
our
time
now
to
read
through
the
proposal.
So
everybody
has
some
time.
So
this
is
not
prime
entertainment
moment
right
now,
but
so
use
10
minutes
to
read
through
the
proposal.
The
proposal
is
very
big:
it's
not
going
to
go.
It's
not
I'm
not
sure.
B
C
B
B
So
there
are
like
some
major
changes
here
where,
like
there's
a
new
image,
to
try
to
depict
a
little
bit
better,
what
it,
what
each,
what
we're
going
to
refer
to
in
each
of
the
terms,
and
then
this
part
in
here,
is
also
a
little
bit
different.
Where
we
have
like
some
some
more
information
about
which,
which
term
are
we
using
in
each
context
and
roughly
the
proposal
in
terms
of
the
proposal
in
itself?.
B
B
So
and
then
we
have
some
explanation
on
on
some
of
the
copy
strategies
that
we
think
that
might
be
interesting.
So
we
depicted
four
of
them
here.
It
doesn't
mean
that
we're
going
to
implement
them
all
right
away
or
that
we're
going
to
implement
them
all
at
all,
but
we're
going
to
implement
the
ones
that
we
believe
are
more
that
are
more
used,
because
in
the
end
you
could
use
always
the
overrides
to
accomplish
the
other.
B
The
other
strategies
and
the
strategy
that
you
want,
but
this
having
this
particular
strategy
that
fits
your
your
needs
it
much
easier
because
then
you
can
just
provide
the
strategy
name
and
go
for
it.
So
one
of
the
strategy
is
the
current
strategy
of
image
package,
where
every
oci
image
is
copied
into
the
same
repository
as
the
bundle.
So
if
you
say
this
is
the
repository
that
we're
going
to
copy
the
bundle
to
it
copies
all
the
images
into
that
repository.
B
The
other
and
then
there's
some
information
about
it
about
how
this
would
work,
so
the
other
one
is
copy.
All
the
oci
images
to
the
new
registry
and
all
paths
are
relative
to
the
bundle
namespace
and
it
keeps
the
image
named
space.
So
this
is
a
very,
very
big
name
because
in
the
end,
it
does
a
lot
of
things,
so
it
copies
the
the
oci
images
to
all
the
place.
B
What
happens
is
that
all
the
images
on
this
particular
strategy
will
be
copied
underneath
this
namespace
here
my
name
and
it
will
keep
all
the
namespaces
that
are
from
the
the
image,
the
original
image
so,
for
example,
an
example
here,
the
origin,
dot,
io,
slash,
sound,
slash
path,
slash.
My
image
will
be
copied
into
the
bundle,
namespace
plus
register
plus
registry,
plus
the
original
namespace
from
the
image,
and
also
the
the
image
name.
B
So
the
next
strategy
is
copying.
The
oci
images
to
a
new
registry
and
all
the
paths
are
relative
to
the
bundle
namespace,
so
it
is
similar
to
the
above.
The
prior
run
that
we
just
talked
about.
The
only
difference
is
that
the
namespace
of
the
oci
images
is
not
kept,
so
everything
will
go
underneath
the
same
namespace
as
the
bundle.
So
again,
the
same
example
here
my
name
bundle,
so
my
private
register.io,
my
name
bundle.
B
D
B
The
relative
to
the
registry,
so
this
one
here
is
even
though
you're
copying
your
bundle
to
this
particular
namespace
inside
your
registry.
It
doesn't
image
package.
Doesn't
when
copying
the
other
images
it
just
keeps
the
original
namespace.
So
that's,
basically,
the
more
direct
translation
you're,
just
replacing
the
registry.
B
So
these
are
more
or
less
the
strategies
that
we
thought
that
could
be
interesting
to
implement
and
we'll
go
through
them
a
little
bit
more,
not
go
through
them,
but
just
there
there
are
some
implementation
like
order
that
we
think
we
should
implement
them
on,
but
I'm
not
sure
if
we
should
tackle
that
right
now.
So
there's
like
some
information
about
the
overrides,
how
the
overrides
work,
the
the
the
part
that
eventually
might
cause
some
confusion.
B
If
we
add
a
slash
in
the
beginning,
it
means
that
it
is
here
this
doc
image
will
be
copied
to
the
registry.io
slash
here,
while
if
you
do
not
like
in
this
case,
you
do
not
provide
a
slash
in
the
beginning.
It
will
assume
that
the
namespace
that
you
want
to
put
the
image
in
is
the
same
as
the
bundle.
So
it
starts
on
the
bundle
namespace.
B
So
this
is
more
or
less
to
mimic
what
you
have
in
the
command
line.
It's
the
same
behavior
as
in
the
file
system,
so
this
this
part
might
might
need
a
little
bit.
I
added
here
a
little
bit
because
this
note
just
to
make
sure
that
we
had
some
documentation
around
this,
just
because
this
might
be
a
nuanced
use
case.
B
So
there
are
some
matchers,
and
these
matchers
that
are
here
are
the
same
ones
that
we
have
in
k
build
we're
trying
to
keep
a
try
to
keep
the
the
same
naming
so
that
people
when
they
go
through
tool
to
tool
they
they
they
know
what
to
expect
with
each
matcher.
So
you
have
an
exact
matcher
that
you
put
an
image
on
it,
and
then
we
have
a
matcher
that
matches
on
registry
and
or
repository
depending
on
what
you
provide.
B
B
There's
an
advanced
matching
scenario:
where
we'll
allow
the
user
to
use
a
starlark
function,
our
ytt
start
alert
functions
to
do
the
same
strategy
or
to
implement
their
own
strategies
instead
of
having
to
be
implemented
via
code
and
then
there's
some
functions
that
are
helpful
helpers
for
that.
So
just
gonna
go
scroll
past
that
there
are
some
considerations
about
the
nested
bundles,
but
this
didn't
change
at
all.
So
I'm
just
gonna
skip
that
because
we
are
a
little
bit
short
in,
and
I
think
that
the
majority
of
the
changes
are
that.
B
Oh
there's
this
part,
the
replication
considerations.
So
we
were
talking
about
the
scenario
where
you
can
replicate
two
registries,
so
registry
a
can
be
replicated
to
registry
b
without
like
as
an
internal
process.
So,
for
example,
if
you're
using
arbor
you
can,
you
can
replicate
to
regis
a
registry
from
another
one,
and
we
have
some
considerations
here
about
that
scenario
as
well,
because
in
some
situations
it's
possible
that
you
can
even
change
the
the
the
the
name
spacing
where
you
put
the
images
on
and
so
on.
B
B
B
From
this
point
on,
we
assume-
or
we
force
you
to
when
you
copy
you'll
copy
everything
to
the
same
registry,
to
try
to
ensure
that
you
don't
have
any
problem
with
authentication
and
so
on,
and
also.
We
believe
that
it
doesn't
make
much
sense
for
you
to
be
able
to
when
you're
copying,
to
spread
through
multiple
registries,
the
images.
So
that's
the
idea
here.
B
B
D
D
F
So
that's
terrible
of
course,
because
as
soon
as
you
move
your
entire
project,
one
directory
up
or
whatever
all
your
load
paths
are,
are
garbage
right
because
they
no
longer
point
to
anything
right,
and
so,
if
we
think
of
this,
as
you
know,
registry
apis
kind
of
provide
us
a
way
to
you
know
effectively
organize
things
like
you
know
like
we
do
on
a
file
system
right.
So
to
me,
the
default
behavior
seems
to
be.
F
More
correct
to
be
relative
to
bundle,
so
you've
actually
been
the
same
name
space,
because
if
I
write
one
copy
config
and
then
I
run
it
against
one,
you
know
this
bundle
over
here
that
I'm
trying
to
copy
in
one
place.
I
should
be
able
to
take
that
copy
config
and
ideally
run
it.
You
know
against
a
different
location,
so
that
would
be
kind
of
baseline
right
now
there
is
going
to
be
some
degenerate
cases
of
like
hey.
F
I
want
to
really
really
have
this
image
at
the
root
of
the
registry
at
this
particular
location.
Right,
and
so
that's
where,
like
the
you
know,
the
effect
of
the
non-relative
specification
comes
in
but
yeah.
So,
ultimately
I
guess
I'm
suggesting
that
we
don't
look
at
it
as
you
always
want
to
start
with
an
absolute
path.
You
want
to
probably
be
looking
at
it
from
the
common
cases,
always
do
the
relative
path
and
then
absolute
part
is
you
know,
that's
the
kind
of
a
unique
use
case.
C
B
C
Yeah,
I
do,
I
do
think
of
these
things
as
like.
They
are
appointed
to
these
proposals,
like
our
point
in
time,
understanding,
but
they
do
capture
like
what
are
our
motivations
for
doing
things
sometimes,
and
that
seems
like
a
durable
driver
be
helpful
when
we're
trying
to
remember
now.
Why
did
we
do
that?
D
For
the
names
of
these
strategies,
can
we
like,
like
we,
are
introducing
this
concept
of
relative
and
absolute
repository
paths?
Like?
Can
we
lean
more
into
that
in
those
names
in
like
like
relative
repo,
or
something
like
that
like,
instead
of
because
it's
not
very
clear,
like
what's
being
kept
relative
to
what
and
stuff
with
just
like
relative
to
bundle,
key
name.
D
D
B
D
D
F
Not
be
like
a
fully
relative
for
alternative
naming,
I
think
I
at
least
personally
find
it
useful
to
see
all
strategy
naming
kind
of
laid
out
to
compare
to
like
the
previous
set
of
names
versus
trying
to
tweak
individually,
because
I
think
they
need
to
make
sense
within
you
know
a
set
of
strategies
right
like
they
should
be
consistent
between
themselves
right.
So,
if
we're
kind
of
proposing
a
set
of
different
names,
it
is
for
me.
I
would
want
to
see
how
this
I
don't
know.
B
D
One
other
question
I
had
is:
are
repo
matchers,
using
the
same
logic
as
like
the
slash
verse,
no
slash
like
if
my
image
repo
matcher
has
a
slash
in
front
of
it.
Is
it
going
to
match
the
full
thing
and
if
it
doesn't
have
a
slash,
is
it
just
going
to
match
at
the
end
of
the
image
url,
but
everything
before
that
is
the
same
or
does
the
matcher
not
follow
that
sort
of
relative
absolute.
D
B
B
F
F
F
F
And
they
just
you
know,
use
that
strategy.
They
don't
have
any
overrides.
Something
like
that,
and
you
know
some
time
ago
we
also
said
hey.
Well,
we
can
have
this
advanced
functionality
that
does
whatever
you
want
it
to
do
right
the
function.
You
know
you
can
match,
go
crazy.
How
you
want
to
match
it,
and
I
think,
kind
of
reading
through
some
of
the
matching
logic
that
we
have.
F
F
We
have
three
set
of
strategies
that
do
the
right
thing
for
you
and
we
maybe
even
extend
some
of
the
strategies
to
you
know
if
we
have
some,
you
know
demand
for
some
special
case
like
I
think
I
actually
added
a
comment
about
saying
like
maybe
we
should
have
a
strategy
that
uses
the
full
namespace
name,
but
that
replaces
slashes
with
dashes
or
something
like
that.
And
then,
if
somebody
does
want
to
artisanally
craft
some,
you
know
replacement
rules.
C
F
I'm
yeah,
I'm
saying
I
don't
know
if
I
got
what
you
just
said
right,
but
I'm
saying
instead
of
implementing
overrides
we'll
forget
about
it,
because
it
is
pretty.
I
don't
know
it's
not
crystal
clear
with
all
kinds
of
matches
and
whatever
what
what
would
match
what
right
and
instead
of
doing
that,
because
at
least
in
my
opinion
overrides
is
not
a
common
case
right.
F
F
Yeah,
what
would
you
wow
has
highlighted
this
stuff?
Yeah,
okay,
and
my
bet
is
that
most
people
will
never
write
a
function
like
that
they
just
use
a
strategy
and
the
strategy
does
the
smart
thing,
because
again,
if
they're
importing
all
kinds
a
lot
of
stuff
right,
there's
no
time
for
sitting
there
and
artisanally
crafting.
C
Overrides
yeah,
I
I
just
just
to
kind
of
properly
frame
it
in
my
own
head.
I
don't.
I
don't
see
that
somebody's
going
to
artisanally
craft
these
things,
but
what
they
might
do
is
write
a
script
that
takes
some
output
and
they
they
write
the
overrides
in
their
script,
because
the
strategy
that's
provided
by
the
tool
doesn't
give
it
to
them.
F
So
if
going
with
that,
then,
if
they're
writing
a
script,
then
I
would
argue
that
overrides
is
really
a
replacement
strategy.
As
in
like
a
forget
about
word
strategy,
I
didn't
mean
that
strategy.
I
mean
it's
just
a
should
be
a
dumb
replacement
and
it
should
not
bother
with
all
kinds
of
advanced
matching
of
like
only
match
by
the
repo
name
or
whatever.
You
should
just
be
like
here's,
the
full
string
use
you
know,
use
this
other
string.
C
C
Let's
see
what
that
first,
I'm
that's
not
the
word
you
used
but
like
at
the
very
least.
Let's
not
necessarily
include
that
as
it
has
to
go
with
these
other
strategies
like
we're.
Just
implementing
the
second
strategy
of
relative
to
registry
keep
name
space
whatever
that
ultimately
becomes
name-wise,
implement
that
and
then
see
we're
at.
Maybe
that's
good
enough
give
feedback
from
the
community
well.
F
You
know
I
haven't
really
taken
a
look,
detailed,
look
at
the
implementation
section,
I
don't
know.
Maybe
it's
changed.
Implementing
strategies
is
very
cheap
right.
We're
just
mangling
up
a
string
right,
it's
like
few
lines
of
code,
so
I
don't
actually
see
a
reason
why
we
wouldn't
want
to
just
implement
all
of
the
strategies
that
we
you
know
that
we
described
in
this
document
all
in
one
shot,
maybe
even
as
part
of
the
same
same
issue.
I
don't
know
like.
I
don't
think
it's
a
you
know.
It's
an
implementation,
complexity.
F
And
yeah
maybe
wait
at
that
point
and
then
see
if
we
need
to
add
the
function,
but
I
am,
I
guess
I
am
polling
for
here.
I
don't
believe
we
need
overrides
ever,
even
if
somebody
comes
in
and
says,
hey
I'd
like
to
replace
this
thing
with
that
thing,
there
wouldn't
be
any
reason
for
why
they
couldn't
write
a
little
function.
That's
a
few
lines!
Long
time
does.
That
argue
like.
F
B
So,
like
the
implementation,
where
I
think
we
came
up
in
time,
but
like
the
implementation
proposal
here,
it's
just
very
high
level
and
like
there's
a
first
part
that
is
some
pre
work.
That
needs
to
be
done
and
that's
what
I
call
mvp0
like
phase
zero
here
and
then
there's
an
mvp
one
that
I
believe
that
will
be
ready
to
deliver
this
to
users
that
implements
justice
strategies
and
the
overrides
is
part
of
the
phase
two
and
vp2.
B
That
is
just
like
that
part
that
if
we
we
can
agree
that
we
might
not
implement
this
mvp
ii
and
then
there
are
two
more
mvps
like
there's
like
a
final
mvp
that
I
called
it
here.
That
is
adding
this
the
the
more
complex
mapping,
that
is
the
ytt
function,
and
then
I
added
another
one
that
would
be
optional.
That's
like
add
extra
strategies,
so
in
the
end
like
that,
separation
is
already
built
in
and
we
can
stop
on
any
mvp
and
say
like
okay.
We
stop
here
and
see
like.
B
C
B
So
the
dry
run
in
itself.
It's
not
able
to
to
like
know
if
something
is
going
to
work
or
not.
The
only
thing
that
it
can
do
is
show
the
user.
What
what
image
package
will
try
to
do
so?
It
will
try
to
copy
this
image
to
this
place
and
it
will
have
this
name.
That's
that's
the
extent
of
what
we
can
do
with
the
current
apis
that
exist
in
the
oci
registry.
B
C
E
B
Defined
but
since
we
are
out
of
time
I'll,
just
just
read
through
through
this
hackmd,
just
and
eventually
we
can
discuss
this,
maybe
via
slack
or
something
and
try
to
get
some
feedback.
So
this
is
like
a
different
way
of
doing
the
strategy
where,
instead
of
you
putting
like
the
full
name
of
the
strategy,
you
say
like
I
want
to
keep
name
spaces
and
then,
for
example,
you
have
another
variable
that
says
like,
oh,
but
by
the
way,
prepend
the
bundle.
B
Okay.
So
like
this,
I
don't
think
it's
bundle,
name
that
we're
going
to
put
there,
but
maybe
like
prepend
prepend
bundle,
name
space
right
so
like
instead
of
having
the
the
strategies
that
keeps
the
bundle
names.
We
have
two
different
strategies:
one
to
keep
the
bundle
name
space,
plus
the
image
namespace
in
this
case
just
smushes,
the
two
with
the
flag.
So
that's
another
option,
but
maybe
I'll
create
a
slacking
thread,
a
thread
in
slack
to
for
us
to
maybe
discuss
this
a
little
bit.
C
B
Welcome
and
I
think,
like
some
other
parts
of
the
renaming
like
if
we,
if
we
really
consider
the
removal
of
the
the
overrides,
I
think
the
majority
of
the
problems
and
to
understand
like
in
this
naming
is
on
the
override
part.
So
if
we
really
decide
that
we
don't
want
the
overrides,
then
that
case
the
majority
of
the
problems
that
we
had
are
kind
of
solved
by
not
having
them.
A
A
If
you
have
any
feedback
you
wish
to
share
with
the
maintainers,
you
can
either
add
your
comments
on
the
proposal
directly
or
join
the
carvill
slack
channel
within
the
kubernetes
workspace
and
your
all
the
maintainers
are
on
that
workspace
and
you
can
provide
feedback
there
as
well.
We
certainly
welcome
the
engagement
and
to
hear
from
from
the
community
on
on
this
proposal
and
anything
else
you
wish
to
share
about
karvel
and
also,
if
you
want
to
join
us,
live
next
time.
A
We
do
meet
every
second
and
fourth
thursday
of
the
month
for
carnival
office
hours
at
11,
30
a.m.
Pacific
time,
2,
30,
p.m.
Eastern
time,
if
you
aren't
able
to
make
those
work,
we
do
meet
every
monday
for
a
community
meeting
at
the
same
time.
11
30
a.m,
pacific
time,
2
30
p.m.
Eastern
time-
and
we
hope
you
have
a
good
rest
of
your
week
and
we'll
see
you
soon
thanks.