►
From YouTube: Cartographer Office Hours No.1 - Nov 22nd, 2021
Description
The purpose of this meeting is to discuss architecture-changing ideas (in the form of RFCs) and provide in-depth support to the community of Cartographer contributors.
You can continue the conversation by adding comments to the RFCs PR: https://github.com/vmware-tanzu/cartographer/labels/rfc
Agenda: https://docs.google.com/document/d/1ImIh7qBrOLOvGMCzY6AURhE-a68IE9_EbCf0g5s18vc/edit?usp=sharing
A
Well,
welcome
everyone
to
the
first
session.
This
is
a
really
happy
day,
for
me,
at
least,
for
the
team
to
have
our
first
office
hours
session
today
is
monday
november
22,
and
remember
that
the
office
hours
is
space,
that
it's
pretty
much
pretty
much
established
in
the
open
source
in
the
open
source
project,
and
the
goal
is
mainly
to
have
an
again
an
open
communication
path
with
the
community
to
answer
tough
questions,
let's
say
and
also
to
discuss
ideas
in
both
ways.
A
A
A
B
A
Okay,
thank
you
yeah,
so
we
will
have
kind
of
15
minutes
per
topic.
I'm
not
using
we're
not
using
any
tool
to
track
the
the
actual
time
for
each
one
of
these
questions.
But
let's
say
that
we
are
kind
of
time
boxing
it
so
everyone
can
chime
in
in
each
one
of
the
four
discussion
topics
for
today's
session
right,
so
that's
it!
The
role
is
to
rotate
again
the
note-taking
role.
For
today,
the
lucky
winner
is
josh.
A
C
C
But
I've
been
it's
not
clear
to
me
what
we
are
to
do
now,
because
the
person
who
opened
this
rc
is
not
on
our
team
and
hasn't
taken
any
action
on
it
and
being
on
the
team.
I
would
like
to
move
this
forward,
so
I
don't
want
to
be
blocked
on
somebody
who
has
like
other
other
concerns,
so
I
don't
know
what
to
what
we
want
to
do.
C
I'd
like
to
know,
you
know
I'd
like
to
take
the
the
updated
suggestion
here
and
make
it
part
of
the
rfc,
and
then
we
could
have
a.
I
think.
Then
we
could
have
an
up
or
down
roman
vote,
and
I
suspect
that
we
would
say
yes,
let's
do
this,
and
my
concern
is
that
yeah
like
if,
if
the
author
of
this
rc
is
like,
I
raised
an
issue
with
you
and
I
just
want
you
guys
to
figure
it
out.
C
I
think
that
that's
totally,
I
think
it's
totally
fine
to
summon
for
someone
to
come
with
us
with
a
problem,
and
we
should
be
able
to
to
say,
like
oh
yeah,
thanks
for
kicking
us
down
this
road
and
we'll
take
it
from
here.
I
don't
know
if
that
means
just
opening
up
a
new
rfc
that,
like
covers
the
same
the
same
ground,
and
then
we
ask
them
to
withdraw
the
rfc
in
rc14
in
favor
of
rfc
17
or
whatever
it
would
be.
C
E
F
G
G
Using
the
like,
the
state
diagram
that
you
shared
last
time
with
schumer
from
I
think
python
that
this
would
be
the
case
of
the
superseded,
we're
like
yeah.
This
one
was
abandoned.
We
picked
up
now
that
one
is
superseded
by
this
if
we
want
to
generate
a
new
number
otherwise,
like
rush
that
we
can
just
use
the
same
number
close
this
one,
the
other
one.
C
F
C
And
then
the
the
only
thing
that
I
would
the
thing
that
I
would
add
to
this
was
what
we
were
discussing
today,
having
augmenting
this
with
one
more
field.
That
is
the
actual
resource
that
the
actual
object
that
could
be
pulled
on
kubernetes.
That
would
say
like
this
is
the
the
object
that
this
artifact
came
from.
C
F
F
B
G
B
C
Can
one
actually
that
might
maybe
even
better,
I
was
about
to
say,
can
one
observe
generation
represent
more
than
one
resource
version?
I
think
the
answer
is
yes,
yeah
that
even
that
speaks
more
strongly
than
it
should
be
the
resource
version
yeah,
because,
for
example,
kpac.
If
the
build
store
changes,
the
observed
generation
will
not
change,
but
the
resource
version
presumably
would
so
that
would
but.
G
B
B
F
F
C
F
C
Yeah,
the
right
now
we
when
we
read
a
value
off
of
a
so
yeah,
you
use
a
template,
a
source
template
to
wrap
some
object,
and
if
you
look
at
our
tests,
you'll
see
that,
like
we
just
accept
anything
when
you
say
like
url
field
revision
field,
we're
like
you
can
put
anything
in
there,
it
could
be.
It
could
be
a
whole
json
object
that
you
that
you
use
as
your
url,
and
I
think
that
we
should
you
know
if
somebody
misconfigured
and
did
that
in
their
actual
supply
chain.
C
I
couldn't
stamp
this
and
they'd
have
to
like
trace
back
and
figure
out
like
what's
what's
the
problem,
and
I
think
it
would
make
a
lot
more
sense
for
us
to
just
actually
do
do
some
validation
of
when
we,
when
we
read
a
value
off
of
a
templated
object-
and
we
say,
oh
I'm,
going
to
put
this
into
an
image,
I'm
going
to
expose
this
to
later
resources.
As
as
an
image
value.
C
B
F
B
H
B
E
A
B
Yeah,
let's
just
make
sure
we
get
it
right.
That's
that's
the
thing
right,
I
think
yeah,
I
think
at
the
moment,
configs
are
considered
to
be
any
old
string
that
can
be
serialized
in
many
many
ways,
including
I
have
heard
people
bandy
around
the
idea
that
it
could
just
be
a
string
like
it
could
just
be
a
config
template
could
be
a
thing
that
gives
you
like
a
secret.
Not
a
secret
object,
just
a
the
value
of
a
secret,
for
example.
H
C
B
B
G
C
C
I
think
it's,
I
I'm
totally
fine
having
an
object
that
you're
having
a
template
type,
that's
just
sort
of
like
yeah
there's
an
arbitrary
template
type.
I
just
wonder:
do
we
want
to
have
config
b
this
thing's
a
you
know
what
you
stamp
out
in
this
is
meant
to
be
a
valid
kate's
object
and
if
you,
if
something's
gone
wrong-
and
it's
not
a
valid
case-
object
we'll,
let
you
know.
B
C
B
I
think
we
reach
out
to
the
users
we
know
who
are
currently
using
the
system
for
configs
in
in
complex
ways
for
delivery
and
just
see
what
the
shape
of
that's
looking
like,
and
if
I'm
wrong,
then
assume
that
it
can
just
be
a
valid
case,
object
and
validate.
On
that
worst
case
scenario.
It
means
a
bunch
of
people
are
going.
Oh,
my
god.
This
is
not
going
to
work
for
me
when
I
move
up
to
the
next
version
and
we'll
we'll
either
give
them
a
new
solution
or
backtrack.
C
Yeah,
I
think
I
think,
even
if
even
if
we
found
folks
that
are
using
it
the
way
that
you're
you're
saying
I
would
be
with
this
story,
I
would
be
saying
like
okay,
we're
gonna
have
a
breaking
change.
You're
gonna
have
to
use
this
other
arbitrary
template
type,
but
yeah
it
is.
It
is
worth
first
finding
out.
Would
we
need
to
have
those
conversations
with
anyone
anyway,.
B
C
C
B
But
I
do
have
one
other
question
for
this
one,
so
your
rsc
does
not
read
like
what
you
propose.
It
re
reads
like
there
could
be
a
multitude
of
things
that
we
need
to
check
all
right.
Could
we
at
least
have
the
concrete
examples
very
clearly
called
out
dot
image
should
be
an
image
url
unless
I'm
not
looking
at
the
right
document,
but
have
something
concrete
about
what
you're
asking
for
and
then
is
there
further
scope
that
you
care
about
other
than
inputs?
H
E
G
One
update
is
that
I've
put
a
like
a
reference
implementation
of
the
first.
I
remember
that
this
rc
has
like
true
suggestions,
or
he
talks
about
true
use
cases.
They
remember
exactly,
but
one
was
literally
like.
Oh
what,
if
we
had
our
selectors
allowing
for
multiple
fields
in
the
map,
and
then
our
workloads
performed
this
best
math
best
label
matching
thing:
there's
an
open
pr
about
it.
Sorry,
I'm
really
like
hazy
on
the
deep.
B
No
you're
right,
so
there
was
two
two
suggestions:
one
was
and
then
they
could
be
either
or,
and
it
was
one
was
to
actually
have
labels
at
select
meaningfully
and
then
inferred
selection
based
on
fields
being
present.
That
meant
a
thing,
and
I
think
we
had
decided
at
the
time
that
the
only
thing
that
had
to
happen
anytime
soon
was
the
first.
B
C
C
It's
a
it
points
out
that
we
could
have
resources
where
you
define
not
one
template,
but
multiple
templates
and
we
just
say
like.
Oh:
if
you
try
to
fulfill
template
one
and
it
doesn't
work,
then
you
can
just
fall
back
to
template
two,
and
if
template
two
doesn't
work,
you
could
fall
back
to
template
three.
We've
kind
of
obviated
the
who
sidestepped
the
need
for
that
sort
of
strategy
with
our
use
of
ytt
templating
but
yeah.
I
thought
it
was
interesting.
They're.
B
C
B
C
B
C
B
C
B
Totally
get
us
over
the
line
yeah
and
then
and-
and
it
still
solves
the
same
problem
it's
just
that
there
is
another
way
that
a
user
might
want
to
be
able
to
not
have
to
be
specific
and
just
go.
Look.
Why
do
I
have
to
put
that
up
there
when
clearly,
whenever
that's
there?
One
of
these
is
here
right.
B
C
Yeah,
I
I
think
also
I
would
be
if
we're
going
to
do
features
one
of
the
things
that
I
think
about
on
the
supply
chain
about
falling
back.
Is
you
know
at
the
top
of
the
at
the
top
of
the
dag?
It's
like
really
easy
to
see
like
oh
yeah,
like
the
workload
either
does
or
doesn't
include
these
fields,
but
then
there's
as
you
continue
traversing
through
the
dag.
There
are
like
later
nodes
that
depend
on
an
earlier
node
to
like
output,
something
it's
not
totally.
B
C
I
think
I
was
concerned
that
you
could
end
up
with
node
a
create
some
output
and
then
node
b,
depending
on
whether
node,
a
like
the
shape
of
its
output
like
either
can
or
can't
fulfill
but
node
a's
output
is
deterministic
right.
It's.
It
is
if
it's
a
source
template
it's
some
url
in
some
revision
and
you're
kind
of
like
guaranteed
that
it's
going
to
do
that.
There's!
No!
F
C
Yeah,
so
if
you
are,
if
you're
like
misusing
cartographer
and
you're
saying
oh
later
on
in
supply
chain,
I'm
going
to
assume
that
some
object
exists
on
the
cluster
and
I'm
going
to
access
it
directly.
Then
yeah.
You
could
totally
shoot
yourself
in
the
foot,
but
I
think
our
our
expectation
is
all
information
that,
like
node
b,
needs
to
know
from
node
a
it
should
be,
should
be
passed
by
node.
A
as
the
output
like
is
what
the
template
is.
Exposing.
B
And
I
think
this
rfc
is
supposed
to
make
it.
This
is
the
we
don't
do
any
clever
selection
approach
to
it.
This
is
the
look
it's
either
it
is
or
it
isn't.
This
is
the
one
we
will
select.
We
still
got
to
select
one
you
will
know
which
one
you
selected,
and
it
has
a
very
clear
graph
at
that
point
and
then
it's
whether
your
supply
chain
made
sense
in
the
first.
E
F
I'm
just
thinking
about
I'm
just
trying
to
think
of
a
concrete
example
to
illustrate
the
the
use
case,
you're
bringing
abushima
and
is
it
like.
Can
I
frame
it
as
like
in
your
workload?
You
can
either
define
where
your
source
comes
from
or
where
an
image
comes
from
and
like
you
could
potentially
get
into
the
case
where
the
first
item
in
your
supply
chain
would
either
have
to
be
a
source
template
or
an
image
template.
C
Yeah,
so
so
there's
the
the
features
is
exactly
that,
like
maybe
you
define,
maybe
you
have
your
workload
and
you
find
source
for
you
to
find
image
so
exciting
over
here.
Let
me
tell
you
guys,
thanksgiving
man,
so
you
define
yeah
you.
You
have
your
workload.
It
has
some
fields
on
it
and
based
on
those
fields.
It
should
be
very
easy
to
say.
C
Like
okay,
you
need
supply
chain
a
or
supply
chain
b,
and
I
was
concerned
that
I
later
in
your
supply
chain,
you
have
some
node,
that's
saying
like
oh,
I
expect
some
output
from
an
from
an
earlier
node
in
the
graph
and
what,
if
it
doesn't
give
you
that,
and
I
think
that
we
don't
need
to
worry
about
that,
but
they
will
always
output
what
you
know.
They
have
a
contract
they're
going
to
output,
and
if
they
don't
output,
the
supply
chain
will
stop.
C
C
Also,
I
I'll
just
put
a
quick
note
rash.
I
I
totally
heard
you
when
you
say
like
we're
not
using
that
with
the
handoff
between
supply
chain
and
delivery,
and
my
hope
is
that,
as
kate's,
like
you
know,
kind
of
like
gets
its
multi-cluster
story
better,
that
we
will
be
able
to
leverage
some
of
that
so
that
we
can
have
some
direct
communication
between
our
supply
chains
and
our
deliveries
so
that
we
don't
have
that
implicit
communication
we'll
see
so.
B
F
C
C
The
other,
the
other
is
to
just
use
a
flag
to
say
if
something's
overrideable,
you
know
we
have
a
when
you
specify
a
value,
you
just
explicitly
write
a
flag.
True
or
false.
Is
this
overrideable
and
if
you
do
then
you
would.
The
supply
chain
is
delegating
to
the
workload
josh
had
written
a
story
for
the
flag
version,
I'm
pretty
I'm
pretty
agnostic.
At
this
point
as
to
which
one
I
am
totally
open
to
hearing
one
of
it,
people
think
is
clearer,
but
yeah.
C
Are
you
saying
like
what
is
the
case
right
now
in
the
code
base
right
right
now
in
the
code
base,
we're
in
this
really
weird
situation
where
what
it
really
is
is
supply
chains
can
overwrite
templates
and
then
supply
chains
can
use
some
json
path
specification
to
say:
oh,
I'm
going
to
grab
a
value
from
the
workload
and
like
write
it
in
here.
So
one
oh
yeah,
one
of
the
things
I
was
thinking
over
the
weekend
was
oh
with
this
change.
C
C
H
F
So
I
think
part
of
the
problem
is:
if,
if
you
want
to
be
totally
flexible
right,
if
you
think
about
the
hierarchy
of
objects,
then
it's
like
you
have
one
supply
chain
operating
on
multiple
workloads.
So
if
you
want
full
flexibility,
you
need
to
be
able
to
override
a
supply
chain's
decision
at
each
individual
workload
right.
F
H
H
H
C
Be
that
hasn't
happened
but
and
at
the
same
time
the
request
for
this
whole
feature
has
been
asked
for
once.
So
I
don't
think
we
have
so
yeah
to
your
point.
No,
we
don't
have
a
request
for
it.
We
have
very
limited
information
and
it
doesn't
seem
unreasonable
to
say
if
the
template,
author
and
supply
chain
out
there
are
different
personas
there
and,
of
course
the
workload
author
is
as
well.
Then
like
there
may
come
times
when
the
template
author
is
like
sure
this
can
be
any
value
and
the
supply
chain
offers.
H
C
All
right,
so,
if
we,
if
we
take
that
as
a
reasonable
scenario,
the
other
scenario
we
have
come
across
where
a
template
says,
here's
a
parameter
that
can
be
set
and
the
supply
chain
author
didn't
have
control
over
that
template
and
did
want
to
be
able
to
say
if
the
workload
doesn't
provide
this
value.
I
want
to
provide
my
own
default
value,
so
do.
B
F
G
B
So
we're
changing
what
the
field
means,
but
doing
so
by
changing
a
boolean
somewhere
else.
Just
feels
weird
right.
I
mean
ultimately
I'd
like
to
have
a
designer
chime
in
on
this
sort
of
thing,
but
still
my
designer
on
this
stuff,
I
guess
but
yeah
that
would
be
yeah
that
that's
how
I
thought
about
it
this
morning
and
that's
how
I
responded
to
washuma
and
I
was
like:
why
did
this
end
up
being
the
one
that
was
chosen.
C
D
C
A
third
case
where
you,
okay
and
where
it
doesn't
it
where
the
template
has
set
a
default,
the
oh
yeah,
well,
okay,
yeah!
So
in
the
third
case
you
said,
the
supply
chain
does
not
specify
a
value
and
the
workload
quote.
Unquote
must
supply
a
value,
but
I
don't
know
that
that
fits
conceptually
with
the
design
of
with
the
template,
because
the
template
does
supply
a
default.
B
B
Now
I
mean
it
is
a
third
scenario,
but
it
was
just
a
scenario.
I
don't
see
why
it
shouldn't
exist,
which
is
I'm
waiting
for
you
to
give
me
this
thing.
There's
no
way
I
can
default
it
or
I
can
default
it,
but
you
will
never
like.
Let's
put
it
another
way
like
if
your
default
is
an
empty
string
for
a
url
that
must
be
provided
right,
then
it's
not
going
to
finish
so
you
still
must
provide
it.
It's
just
that
you're
not
going
to
find
that.
E
B
I
still
think
that's
true,
even
though
you
can
provide
a
default
often
the
defaults
are
ones
that
don't
make
sense.
We've
seen
that
in
some
of
our
example
chains,
we've
gone.
Oh
sorry,
in
our
example
templates
we
go.
Oh,
we
have
to
provide
a
default,
but
it's
going
to
be
empty,
especially,
I
think
for
the
git
push
stuff
that
you
were
doing,
you
don't
provide
a
username
and
email,
it's
not
healthy.
B
It's
best
to
leave
it
right
and
then
have
the
tool
complain
that
you
didn't,
provide
the
right
information
so
and-
and
that's
true
all
the
way
up,
I
think,
to
the
workload
author
in
scenarios.
I
think
there
would
be
plenty
of
scenarios
where
you
say
the
workload
author
is
the
only
person
who's
going
to
know
this
so
have
them,
provide
it
or
the
cli
tool
that
they're
using
is
the
only
thing
that
knows
it.
Maybe
it's
pushing
from
a
local
github
and
it
will
just
use
their
author
name
or
something.
C
B
C
So
to
that
point,
do
we
just
say
like
hey,
we've
had
the
community
meeting
we've
had
you
know
this
rc
open
for
a
long
time.
We
said
we
want
to
you
know:
josh
your
commenters
like
11
days
ago.
We
want
to
find
an
implementation.
Do
we
just
say,
like
it's
been
open,
we're
discussing
and
now
we're
going
to
roam
and
vote
on
these
two
options
and
implement
one
this
week.
F
No,
we
need
to
get
buy-in
for
the
change
we
want
to
make
right
because
for
the
record,
I
I
like.
I
totally
agree.
I
much
prefer
this
pattern.
It
makes
a
lot
more
sense
to
me
as
well,
but
we
need
more
buy-in
on
from
the
vmware
side
to
pursue
this
path
and
I've
started
getting
I've
tried
I'm
trying
to
get
that
buy-in,
but
I
don't
know
if
that's
going
to
happen
before
we
need
to
deliver
this
feature
so.