►
From YouTube: Carvel Community Meeting - Nov. 11 2020
Description
Carvel's community meeting is held fortnightly on Wednesdays at 12PM ET / 9AM PT via Zoom (http://community.klt.rip)
Join our Google Group (https://groups.google.com/g/carvel-dev) to get updates on the project and invites to community meetings.
---
Agenda:
Announcements 📣
1. vendir v0.12.0 (release notes)
2. KubeCon office hours - November 20th 10-10:45 PST
Carvel Tech Forum 💻
1. Dead Reckoning on imgpkg: Where are we now? Where are we going?
- kbld package/unpackage:
https://github.com/k14s/kbld/blob/develop/docs/packaging.md
- Docs: https://github.com/k14s/imgpkg/tree/develop/docs
- Releases: https://github.com/k14s/imgpkg/releases
- Prior / Other Current Art
https://github.com/containers/skopeo
https://github.com/deislabs/oras
https://github.com/aws/containers-roadmap/issues/43#issuecomment-691091513
C
C
C
C
C
D
Because
carmel
follows
the
philosophy
of
having
purpose-built
tools
that
solve
one
issue
it
felt
like
we
were
overloading
k-build,
because
cable
had
other
responsibilities
like
resolving
image.
References
from
kate's
manifest
it
just
felt
like
we
were
putting
too
much
into
k-build
that
it
wouldn't
be
able
to
solve
both
problems
well
or
as
well
as
they
could
be
solved.
D
D
Right
now
we
have
the
push,
pull
and
copy
commands,
so
they
pretty
much
do
what
they
sound
like
push.
You
can
bundle
up
a
set
of
configuration
on
your
file
system.
It
will
throw
that
into
a
tarball
and
add
it
as
an
image
layer.
So
then,
you
can
distribute
your
configuration
through
a
docker
registry.
D
D
So
then,
you're
able
to
then
use
the
config
that
you've
just
unpacked
in
tools
like
k,
build
ytt
cap
and
then
copy
does
what
it
sounds
like
as
well
like
it'll
copy
the
bundle
image
and
any
images
that
the
config
references
to
either
another
registry
or,
if
you're,
doing
offline
relocation
to
a
tarball,
which
you
can
then
unpack
on.
Whatever
network
is
internal
using
image
package
as
well.
D
C
C
B
C
D
D
C
C
D
Yeah
so
right
now
the
mvp
has
those
three
commands
so
you're
able
to
create
a
bundle,
pull
a
bundle
and
relocate
a
bundle.
It
also
works
with,
like
generic
images,
so
you
can
use
the
tool
to
just
do
a
basic
image.
Relocation
either
register
to
register
via
a
toggle,
that's
sort
of
the
scope
of
it.
Now,
there's
not
too
many.
C
D
Yeah,
it
does
a
couple
different
validations.
I
think,
when
you're
pushing
a
bundle,
there's
certain
conditions
that
need
to
be
satisfied
about
the
bundle
which
then
provide
the
consumers
a
consistent
interface,
so
any
automation
they
have
set
up
is
guaranteed.
Certain
files
will
be
there,
so
they
don't
really
need
to
worry
about
it.
D
C
D
C
C
Okay,
there's
a
couple
of
couple
different
directions:
we
could
kind
of
take
this
conversation.
One
is
for
folks
in
the
room
that
maybe
have
a
sense
of
sort
of
where
we're
at
in
the
industry
in
terms
of
packaging,
just
like
I,
we
saw
recently
there's
a
kubernetes
enhancement
proposal
around
something
like
that.
There's
some
prior
art
around
some
pad
packaging
stuff.
C
So
that's
that's
an
option.
The
other
thing
we
could
get
into
is
what
are
some
of
the
things
that
we're
pretty
clear
about
what
we
might
tackle
next
or
like
a
future
road
map.
So
let's
take
that
first
question.
First,
has
anybody
done
any
research
researcher
is
looked
at
all
or
has
any
thoughts
about
like
where,
where
we're
at,
as
especially
focusing
on
kubernetes
like
where
we're
at
in
terms
of
packaging
and
distribution.
C
E
D
C
F
So
truthfully,
I'm
not
100
familiar
with
the
tool
itself,
but
I
do
know
that
scopio
is
another
tool
that
sounds
kind
of
similar
or
hey.
I
I
know
that
you
might
have
more
perspective
on
how
similar
this
sounds
to
image
package
or
not,
but
it
might
be
something
that
people
can
check
out
as
far
as
whether
or
not
it
sort
of
alliance.
G
C
C
E
It
does
not
try
to
describe
what
to
do
with
the
bundle
itself
versus
cnab.
You
know,
has
the
certain
contracts
around
it
like
you
know,
oh
there's,
an
action
that
would
call
to
install
things
or
whatever
right
so
there's
the
executable
part
of
you
know
part
of
the
cnab
artifact
right,
which
you
know.
E
H
I
feel
like
the
problem
here
is
more
what
you
define
as
a
bundle
right,
because
in
cnab
a
bundle
is
a
group
of
things
that
you
are
going
to
install
that's
what
a
bundle
means
in
cnap,
so
it
totally
makes
sense
that
a
bundle
is
executable
and
is
installable
right
and
in
here
it
looks
like
that.
We're
talking
that
a
bundle
is
just
like
a
bunch
of
files.
H
E
My
my
argument,
I
would
say,
does
not
relate
to
the
naming.
I
just
use
bundle
in
both
of
those
cases.
It's
not
the
name,
it's
the
you
know
sure
call
it
something
else.
Foo
right.
I
don't
want
food
to
be
executable,
because
I
would
much
rather
have
full
have
you
know
appropriate
content
and
metadata
and
have
the
execution
be
done
by
something
else
that
maybe
is
more
controlled.
C
It's
an
interesting
footnote.
That's
a
good
good
point.
You
bring
up
that
bundle.
Bundle
is
used
in
in
a
number
of
contexts
to
indicate
something
that
is
executing
that,
like
maybe
we
can
explore
the
name
of
the
thing
that
you
move
around
I've
seen
terms
like
package,
which
feels
a
little
bit
more
like
a
a
noun
than
a
verb,
a
thing
that's
less
likely
to
or
it's
just
it's
a
collection
of
stuff.
C
D
D
To
allow
sort
of
relocation
of
like
catalogs
and
stuff,
potentially
so,
you've
got
some
set
of
apps
that
you
want
to
distribute.
You
can
create
a
bundle
that
references,
those
and
then
be
sure
that
when
you
relocate
the
catalog,
then
all
the
apps
also
come
along
with
it
and
then
also
starting
to
think
about
some
ideas
around
signing.
So
how
can
we
do
identity,
verification
of
a
bundle
to
make
sure
that,
like
vmware,
is
actually
the
one
that
distributed
a
certain
bundle?
E
C
So
to
the
difference
here
is
that
when
you
pull
your
you're
you're,
getting
the
bundle
alone,
not
the
images
downloaded
into
wherever
you're
pulling
into
and
so
you're
saying
it'd.
Be
it's
useful
to
be
able
to
target
a
tarball
where
there
isn't
a
registry
that
you
can
pull
from,
where
the
tarball
contains
the
configuration
and
the
images
identified
in
the
bundle
and
just
be
able
to
do
that,
pull
to
continue
the
workflow
without
having
to
first
copy
your
tarball
up
into
a
registry.
E
Yeah,
I'm
saying
that
the
cardboard
contains
all
the
information
that
technically
one
needs
and
so
for
people
who
do
have
a
registry.
They
can
freely
put
it
into
a
registry,
but
there
are
some
cases
where
image
package
may
be
useful
for
loading
up
the
data.
That's
just
saved
on
disk,
and
so
I
would
you
know
if
I
have
a
tarball
with
a
bundle,
content
or
just
a
regular
image
contents
right.
C
E
E
Looking
at
hey,
do
I
see
those
images
in
my
local?
You
know
tarble
place
right
versus
like
do.
I
see
it
in
the
registry
kind
of
thing
right
because
in
those
cases
like
there
isn't
even
a
registry
to
pull
from
so
there's
no
point
in
contacting
anything
because
there's
nothing
to
contact
so
yeah.
So
that
would
be
the
the
one
case.
E
C
E
C
E
C
E
D
D
E
Yeah,
so
that's
that's
kind
of
similar
to
that
right.
Where,
when
we,
I
guess
today,
we
have
a
copy
right
and
our
format
is
proprietary.
If
you
will
for
the
for
the
tarball
right,
it's
just
something
that
image
package
knows:
it's
not
standardized
right
intentionally,
and
so
there
could
be
use
for
being
able
to
say
well
image
package
copy
with
the
destination
in
the
oci
directory
format,
or
something
like
that
right
that,
let's
say
container
d
can
understand
right
and
that
again
kind
of
goes
back
to
well.
E
It
would
be
possible
then,
to
let's
say,
build
a
aws
ami
that
contains
the
you
know
the
bits
if
you
will,
but
I
think,
there's
potentially
other
tools
that
may
be
able
to
help
you
with
that.
I
don't
know.
I
don't
know
how,
like
how
you
know,
there's,
I
guess
a
balance
between
us
doing
it
for
convenience
sake,
because
it's
so
easy
versus
delegating
it
to
you
know
some
other
tool,
that's
better
doing
that.
E
C
Yes,
the
other
thing
you
had
mentioned
eli
was
the
like.
How
do
you
describe
it,
but
it
in
my
head?
It
landed
as
like
recursive
bundle
references,
so
that
was
an
interesting
use
case.
You
mentioned
it's
like
a
catalog
that,
like
two
layers,
where
I
have
a
thing
that
references
that,
like
is
a
set
sorry,
I
have
a
overall
bundle
that
represents
a
set
of
bundles
each
one
of
those
are
a
kind
of
product,
and
this
higher
level
thing
is
sort
of
a
suite
that
I
want
to
move
around
together.
C
D
C
C
D
A
C
F
Up,
I
was
just
curious
as
part
of
the
docs
that
are
being
released
or
anything
along.
Those
lines
is
there
any
like,
and
I
know
that
we
have
like
the
basic
sort
of
getting
started
scenario
with,
like
all
the
various
tools
that
come
in
karbal.
Is
there
any
plans
to
like
add
image
package
like
as
part
of
that,
like
hello
world
scenario
or
just
like
part
of
the
basic
docs
there,
so
that
people
can
kind
of
like
have
that
guided
workflow
of
getting
started
with
it?.
D
In
the
actual
repo,
we
are
planning
on
adding
a
couple
of
use
cases
to
the
docs,
which
are
hopefully
going
to
be
like
a
set
of
commands
with
descriptions
about
what
they
do,
that
capture
a
full
workflow.
So
I
think
one
that
we
are
definitely
planning
to
do
is
like
the
offline
relocation
so
showing
the
whole
workflow
from
creating
the
bundle
to
pulling
it
to
a
tarball
to
then
unpacking
the
tarball
and
then
pulling
it
on
the
other
end
yeah,
and
maybe
some
other
ones
like.
D
F
No,
I
think
the
yeah
for
now
it'll,
maybe
it'll,
become
more
apparent
over
time.
As
more
use
cases
are
revealed,
and
we
have
like
a
better
sense
of
how
everyone
is
going
to
use
the
tool,
then
it
would
be
more
reasonable
to
like
add
it
to
other
places,
but
I
think
for
now,
just
starting
in
the
repo
for
it
itself
is
sounds
good
to
me.
G
So
touching
on
what
daniel
asked-
and
I
I
think
that
you
probably
are
familiar
with
the
tensor
developer
portal
and
how
we
have
surface
now-
the
ability
to
do
workshop
with
the
educates
platform.
I
think
that
it
it
could
be
beneficial
to
try
to
work
out
some
of
this
getting
started
or
or
documentation
as
workshops
and
then
link
from
your
site
into
the
tansu
developer
portals
workshops.
G
C
We
want
to
provide
a
a
lot
of
sort
of
front
door
options
when
folks
show
up
some
people
want
to
just
jump
in
kick
the
tires
and
and
redox
other
people
like
to
have
a
curated
tutorial,
a
lot
of
folks
do
and
having
that
support
around
it
and
the
infrastructure
around.
It
would
be
great
for
us
to
make
use
of
appreciate
the
offer
we'll
be
talking
about
that.
C
G
G
The
experience
why
the
cli
has
the
alpha
flag
or
the
alpha
word
in
it,
because
then
the
problem
is
that
whenever
it
promotes
to
non-alpha
you'll
need
to
change
the
cli,
and
then
I
was
doing
a
little
bit
of
a
review
of
all
the
tools
and
one
thing
that
I've
I've
constantly
been
asking
is
that
there
is
no
consistency
from
from
tool
to
tool
on
how
they
works.
So,
for
example,
if
you
run
some
of
the
tools
without
any
flags,
they
will
give
not
give
you
no
options.
Some
other
ones
will
print
the
help.
G
Some
other
ones
will
print
the
help,
plus
an
error.
Some
of
the
ones
will
print
their
error
or
directly
and
then
the
flag.
So
so
my
ask
is
that
you,
if
the?
If
there
can
be
some
consistency,
moving
forward
like
or
in
the
future,
and
some
like
developer
experience
thought
into
what
then
the
ui,
or
this
the
let's
say
the
developer
experience
aspect
of
the
cli
can
be
because
this
this
can
be
a
problem
for
people
if
they
automate
things
using
this
flag
and
then
the
flag
changes
the
name,
it
can
break
people.
G
C
Yeah,
so
I'm
hearing
from
from
what
you're
saying
sort
of
the
general
theme
of
there's
a
couple
different
things
that
you
see
that
that
are
that
are
inconsistent
across
across
the
tools.
So
yeah,
that's
something
definitely
definitely
look
into.
I
know
dimitri.
You
did
a
lot
of
work
on
that.
That
particular
fact
did
you
want
to
talk
about
that.
One.
E
Backwards,
comfortability
is
something
that
is
obviously
very,
very
important.
I
really
hate
when
tools
break,
so
there
needs
to
be
a
way
to
indicate
right
that
something
is
not
ready
to
be
solidified
and
made
to
be
relied
upon
right.
So
the
flag
that
I
think
is
in
question
is
diff.
Ui
alpha
was
intentionally
added
as
an
alpha
feature
and
it
was
annotated.
E
You
know
in
that
as
a
suffix
alpha
with
the
intention
that,
if
we
do
expect
it
to,
if
you
will
graduate
eventually
in
its
maturity,
then
at
some
point
the
alpha
thing
will
be
removed
now,
because
it's
also
a
ui
component
right,
none
of
our
clies
offer
a
live
ui.
On
top
this
made
it
even
more
kind
of
a
reinforce
that
hey.
We
should
put
it
as
alpha,
because
we
don't
quite
know
how
I
want
to
actually
keep
this
set
of
functionality
long
term
right
today,
it's
a
web
ui.
E
Maybe
that
changes-
maybe
not
maybe
the
web
ui
changes
right,
and
so
that's
why
that
specific
flag
is
marked
as
alpha.
I
do
expect
that
if
we
do
have
any
other
flags
that
have
a
significant
suspicion
that
they
may
change,
they
will
also
be
annotated
as
alpha
and
then
the
expectation
would
be
is
you
know
when
they
do
graduate
to
something
more
stable
than
the
alpha
will
be
you
know,
removed?
E
G
No,
I
I
understand
the
thinking
and-
and
I
think
that
you
you
answered
in
the
in
github,
dimitri
about
the
promise
to
me.
Maybe
it's
my
perception
of
how
people
should
feel
from
the
the
user
interface
of
this
maybe
you're
right,
it's
an
api.
But
if
you
look
into
some
other
tools
there
is,
there
is
different
ways
to
mark
things
as
alpha,
so
the
the
the
experience,
the
cli,
the
integrations
and
everything
doesn't
break
if
there
is
promote
too
stable.
G
I
understand
that
that
in
kubernetes
you
can
you
whenever
you
look
into
into
them
into
the
api
and
the
resources
they
are
mark
as
alpha,
but
then
the
experience
that
the
that
manages
all
those
interfaces
like
all
these
cube,
ctl
commands
and
everything
like
that.
They
don't
change
or
if
they
change
it's,
because
they
have
been
marked
as
they
pre
as
as
alpha
in
the
documentation,
but
the
the
command
itself
doesn't
change.
So.
E
E
E
That's
not
a
question
because
let's
say
we
feel
strongly
that
it's
stable
enough
to
not
be
marked
as
alpha,
but
there
still
needs
to
be
a
non-documentation
differentiator
to
indicate
that
hey
this
is
watch
out.
This
is
alpha.
So,
for
example,
if
you
did
one
rely
on
defui
flag
in
your
code,
you
should
be
very
very
aware
that
it
is
alpha
feature
and
it
is
annotated
that
way
intentionally
right.
So
it
it
is.
I
guess
the
the
the
experience
of
this
flag
is
intentionally
done.
That
way,
and
it
is
an
experience
right.
E
I
guess
it
sounds
like
you.
Don't
necessarily
agree
with
the
way
that
you
know
alpha.
Things
should
be
annotated,
but
I
would
say
that
for
for
questionable
functionality
that
may
even
go
away
tomorrow
right,
we
don't
want
people
to
rely
on
it
as
if
it's
stable
right,
because
documentation
stating
something
in
documentation
that
it's,
stable
or
not
stable,
typically
isn't
a
large
enough
indicator
to
somebody
that
it
may
just
go
away.
F
So
just
a
note
there
there
are,
for
instance,
like
tricky
ways
with
cobra
that
you
can
sort
of
support
like
renaming
a
flag,
but
also
having
these,
like
the
the
flag
of
a
previous
name,
still
work
and
there's
also
like
ways
to
even
print
like
warning
messages.
If
someone
uses
a
particular
like
name
of
that
flag
to
like
warn
people
more
explicitly
so
there,
there
are
kind
of
work
around
ways
that
we
could
potentially
explore.
F
G
Okay,
yeah,
I
mean
I,
I
understand
the
rationality
I
I
just
didn't
like
I
didn't
feel
really
consistent,
but
yeah.
I
do
agree
that
there
might
be
like
use
cases
and
if
you
want
to
be
explicit,
I
understand
that
then
related
to
consistency.
I
guess
that
that's
a
different
story,
consistency
between
the
different
tools.
E
So
yeah,
the
other
thing
that
you
mentioned
is
consistently
about
various
tools,
and
I
think
you
know
I
don't
know
if
you
have
other
examples.
I
think
the
example
that
you
filed
on
github
was
regarding
the
root
command
being
specified.
Maybe
daniel
actually
filed
a
root
command
being
specified
without
any
flags
and
behaving
differently.
E
Yeah,
I
think
I
think
you
know
we
definitely
want
to
make
the
experience
that
is
between
these
tools.
Consistent,
the
the
one
for
the
root.
One
is
an
interesting
one
right,
because
it's
kind
of
a
it's
like
different
tools,
have
slightly
different
functionality
right
and
so
in
some
cases,
it's
perfectly
valid
to
have
a
scenario
where,
like
for
ytt
right
today,
if
you
give
ytt
zero
flags,
you'll,
just
exit
with
no
output
and
exit
start
of
zero,
and
so
that's
done
on
purpose
to
allow
people
to
potentially
dynamically.
E
Have
you
know
like
when
they're
scripting
with
ytt
right,
if
you,
if
you
are
doing
some
kind
of
a
dynamic
variable,
substitution
or
whatever,
there
are
cases
where
you
actually
don't
end
up
with
any
kind
of
input,
and
so
you
just
want
an
empty
output
right
versus,
for
example,
for
cap
there
is
never
a
case
of
where
you
want
to
run
cap
without
you
know
like
in
ytt.
The
root
command
is
the
templating
command
right
versus
in
cap.
E
There
is
no
root
command
right,
and
so
that's
why
cap,
for
example,
shows
you
an
error
message.
Saying:
hey.
You
didn't
actually
give
us
the
command
to
run
versus
ytt,
happily
says
hey,
there
was
nothing
to
do
for
me.
I
am
running
the
templating
command
and
you
know,
because
there's
nothing
to
do
for
me.
I
exit
successfully.
E
So
the
way
I
look
at
it
is
there
is
a
consistency
on
a
kind
of
a
metal
level
of
what
operation
we're
doing
and
if
that
operation
succeeds,
then
it's
actually
start
with
zero,
and
so
in
ytt
we
do
the
operation
of
templating
and
with
zero
inputs
it's
successful
versus
in
cap.
There
is
no
operation
to
do.
Hence
it's
exit
status,
one
because
you've
given
us
a
wrong
thing
right,
so
that's
at
least
the
way
I
rationalize
that
that
that
level
of
consistency-
but
you
know,
there's
obviously
other
ways
to
to
look
at
it.
G
Yeah
and
that's
I,
I
understand-
that's
consistent
in
the
in
the
restaurant,
in
the
rationale
that
you
are
mentioning.
But
if
you
look
into
cap,
you
print
out
the
help
and
if
you're
looking
to
vendor,
for
example,
you
don't
so
there
is
no
consistency
in
whether
you
are
entering
without
because
you
are
entering
and
you
are
saying
hey,
you
need
to
specify
one
command
in
one
of
them.
You
are
showing
the
help
to
so.
The
user
knows
what
command
to
run
in
the
other
one.
E
Yep
that
that
that
that
that
case
I
would
agree.
I
think
that
actually
is
an
artifact
of
us
not
updating
to
the
particular
cobra
miscellaneous
package.
We
have
that,
does
some,
you
know
kind
of
a
genetic
configuration,
so
the
cap,
I
think,
might
be
using
a
newer
version
versus
vendors
using
a
older
version,
but
yeah
the
intent
would
be
is
that
both
of
those
cases
would
actually
be
printing
out.
Let's
say
help
message
and
the
error
since
in
vendor
case,
there
is
no
to
run.
C
Cool
well
jorge,
really
appreciate
that
feedback
and
that
you've
been
contributing
in
various
ways
and
again
you
know
you
showed
up
offering
to
help
out
with
the
the
the
onboarding
experience
really
appreciate
your
support
in
in
the
project.
Sure
thank
you
and
yeah.
This
is
great.
Let's
keep
talking
about
these
kinds
of
things.
You
can
see
that
like
where
we
miss
out
on
on
one
of
our
own
goals.
We
love
it
when
we
get
feedback
to
help
get
better
and
then
to
be
able
to
like
unpack
rationale.
C
E
C
E
Yeah
yeah,
because
because
the
example
that
was
just
brought
up
right
off
vendor,
for
example,
not
doing
something,
but
cap
is
doing
something
right
like
I
was
not
aware
of
that
part
right,
and
so
the
question
would
be.
Is
I
mean
yeah?
We
would
want
to
file
an
issue
against
vendor
right,
but
is
there
some
place
where
we
can
say
here's
the
general
patterns
that
we
use
across
multiple
clis
that
you
know
comply
and
then
we
can
always
have
individual
issues
to
fix
the
discrepancy.
A
Yeah
a
few
items,
so
we
have
this
agenda
item
on
hackmd
and
you'll
have
the
link
it's
in
slack,
you
can
find
it
and
for
the
future,
always
feel
free
to
add
your
topic
there.
It
will
be
really
great
if
we
know
various
topics
ahead
of
time,
so
that
we
can
give
more
time
for
that,
because
we
do
want
to
hear
from
community
members
and
yeah.
So
that's
one
and
then
also
there's
a
feedback
form
if
there's
anything
that
we
can
improve
on.
Let
us
know
also
we'll
be
uploading.
A
The
recordings
to
youtube
channel
so
I'll
add
a
link
into
this
agenda
doc.
So
you
can
see
we
started
recording
the
previous
one
and
this
one
and
then
the
future
will
record
them
as
well.
So
recordings
of
these
meetings
on
a
youtube
channel
and
you'll
find
the
link
embedded
in
this
agenda
at
the
top,
so
stay
tuned
thanks.
Everyone
for
joining
today
have
a
good
rest
of
your
day.