►
From YouTube: Argo Contributor Experience Office Hour 22nd Oct 2020
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
All
right,
so
we
had
several
topics
carry
over
from
the
last
meeting.
I
don't
see
isaac
today,
so
he
proposed
to
talk
about
application
controller
tests
and
since
he
is
not
here,
I
think
we
can
move
on
move
to
the
next
topic
and
if
he
joins
late
then
we
can,
you
know,
talk
about
it.
A
Okay
and
next
one
is
so
the
feature
request
for
ndpr
from
peter.
I
think
peter
kind
of
go
happens.
C
Hey
guys,
I'm
peter
from
bonsai
cloud,
and
I
just
wanted
to
promote
this
pr,
because
I
believe
it
would
help
in
certain
cases.
Let
me
share
my
screen.
Let
me
stop
sharing.
C
And
it
turned
turns
out:
it
failed
because
the
cr
in
that
sync
introduced
used
a
new
field
that
wasn't
available
in
the
existing
crd,
but
it
was
available
in
the
crd
in
the
same
sync
just
not
yet
applied
to
the
server.
C
So
I
was
relatively
new
to
argos
cds,
so
I
try
to
figure
out
how
we
should
handle
this
issue.
I
thought
that
there
could
be
some
workaround,
but
I
haven't
found
any
so.
I
tried
pre-sync
and
putting
the
cr
into
posting
or
using
sync
waves,
but
everything
seem
to
happen
just
after
validation,
so
validation
precedes
everything
and-
and
I
couldn't
even
solve
this
reliably
or
elegantly,
using
or
putting
crds
and
crs
in
different
applications.
C
So
we
discussed
this
internally
and
we
had
an
idea
that
we
could
make
this
work
using
the
crd
in
that
same
thing,
if
it's
available
and
trying
to
validate
the
cr
against
that
that
local
crd
locally,
instead
of
trying
to
validate
it
against
the
api
server,
this
pr
is
a
proof
of
a
concept
on
that
idea.
C
So
to
demonstrate
this,
I
have
an
argo
cd
installation
running
locally.
C
I
applied
the
the
default
orgo
cd
resources,
but
I
scaled
down
the
application
controller
so
that
I
can
start
my
own
from
a
repository
that
has
the
github's
engine
set
to
a
local
pass
where
I
applied
these
changes.
C
C
C
Then
I
should
see
a
failure
that
says:
local
output
traps
is
not
available,
so
it's
an
unknown
field
which
is
true
because
it's
not
yet
applied
to
the
api
server
and
the
change
support
to
fix
that
by
adding
an
annotation
to
this
cr
instead
of
disabling
validation,
which
is
possible
right
now.
But
we
don't
want
to
skip
validation.
C
So
the
point
here
is
that,
with
with
changes
in
crds
and
and
even
when,
we
want
to
migrate
resources
over
to
to
new
api
versions,
we
will
see
we
will
face
issues
like
this,
but
yeah.
So
it's
it's
nothing,
but
but
using
this
local
validation
we
can.
We
can
fix
this
or
work
around
this
without
disabling
validations
at
all.
C
I
tried
to
not
just
give
a
patch
but
to
write
some
tests
for
this,
so
the
test
cases
are
checking
for
valid
and
invalid
crs
and
again.
This
is
this
can
be
enabled
using
a
feature
flag,
so
it
shouldn't
hurt
existing
functionality.
C
A
A
I
just.
I
think
we
had
kind
of
very
similar
problem
when
you're
introducing
cid
for
the
first
time
and
then
you
yes
and
that's
right,
like
my
first
immediate
feedback,
is
that
do
you
think
do
we
will
need
that
flag?
A
I
feel
like
we're
trying
to
add
new
annotations
very
carefully,
because
that
number
keep
growing
so
and
then
we
hit
internally
discussion
between
you
know
how
to
what
is
the
best
way
to
solve
a
problem
when,
for
example,
you
want
to
create
id
ncrs
in
the
same
sync
operation
or
if
you
have
a
namespace
and
resources,
that's
supposed
to
be
created
in
that
namespace,
and
then
we
were
discussing
kind
of
if
you
want
to
introduce
any
kind
of
flags,
and
basically
we
decided
that
most
that
all
looks
like
each
case
and
it's
really
easy
to
detect
it
in
code.
A
Imagine
if
you
have
a
cr
and
cid
you
could
simply
in
the
code.
We
can
say:
oh
you
know,
cld
must
be
created
first,
it
seems
to
me
the
same
kind
of
use
case,
so
we
could
say.
Yes,
we
have
cr
and
cld,
so
we,
you
know
we're
going
to
change
the
idea
and
we
should
just
go
back
to
local
validation
without
any
additional.
C
Yeah
for
me,
I
also
think
that
this
should
be
the
default,
but
it's
it's
well,
it's
a
bit
experimental
since
it's
not
the
thing
that
that
qcpr
does,
for
example,
it
also
validates
against
the
api
server,
because
it
doesn't
have
that
crd
available.
C
Also,
it's
different
because
clients,
so
validation
against
the
api
server
means
navigating
against
an
open
api
b2
document.
C
C
So
so,
if
it
may
introduce
some
issues
or
it
may
introduce
different
behavior
than
than
or
how
it
worked
before
so
this
is
why
I
would
be
more
cautious,
otherwise
I
would
be
happy
to
have
this
by
default.
It's
just.
I
didn't.
A
A
So
if
we
only
have
that
behavior
with
that
flag,
then
I
suspect
that
no
one
is
going
to
use
it
until
they
hit
this
issue
and
then
they're
going
to
put
that
flank
and
they
will
keep
it
forever.
It's
like
because
it's
a
it's
kind
of
a
one,
more
thing,
but
you
won't
yeah,
so
maybe
another
option
could
be.
We
could
try
to
run.
You
know
during
the
same
process
we
can
run
the
validation
and,
in
case
of
error,
double
check,
the
local
validation
and
the
focal
validation
passing.
A
Then,
basically,
you
are
upgrading
because
it's
like
a
clear
indication
that
you
know
if
you
have
a
cr
ncr.
So
if
you
have
a
sync
operation
and
then
you
have
cr
and
cld
and
then
cr
validation
passed
in
the
servers
but
sorry
failed
in
the
server
but
passed
locally
and
you're
going
to
change
their
crg
seems
to
be
good
enough.
C
Where
we
do,
the
validation
is
is
way
before
the
actual
validation
is
is
done
right
now,
so
we
do
this
at
the
very
beginning
where
we
collect
resources,
and
it's
done
at
that
stage,
not
the
actual
validation
happens.
C
A
Think
yeah,
maybe
this
but
like
in
general.
I
think
I
want
to
kind
of
at
least
I
will
try
to
now
look
one
more
time
and
then
I
want
to
take
a
look
at
implementation
and
you
know
continue
the
discussion
this
year.
But
basically
I
feel
like
at
least
for
me
now
contribution
would
be.
You
know
much
easier
and
I
will
be
just.
A
I
will
try
to
find
a
way
to
make
this
possible
and
suggest
the
code
changes
inside
because
at
least
I'm
kind
of.
A
This
is
like
oh
hundred
percent.
This
feature
is
needed.
I
just
want
to
find
a
way
to
avoid
asking
people
to
put
that
flag
at
least
okay,
and
there
is
one
workaround.
Basically,
argo
cd
supports
sync,
and
you
during
this
thing
you
can
disable
validation.
You
can
just
say.
C
Yeah
yeah
anything
yeah,
yeah
yeah.
We
realize
that
we
we
just
don't
want
to
disable
validation
because
it
would
be
the
same.
C
We
would
put
that
validation
force
annotation
there
and
we
we
would
forget
it.
So
this
is
why
we
don't
don't
want
to
sleep
or
we
would
have
to
skip
validation
for
the
entire
application,
but
we
didn't
fund
it
either.
So
and
and
as
you
can
see,
this
is
right,
where
you
do
the
automatic
validation
skipping
right
now.
So
if,
if
you
have
the
the
crd
in
the
scene,
you
check
for
it
and
you
have
the
the
annotation.
Oh
okay,
thank
you.
C
C
Same
so
in
in
the
not
found
case,
this
is
automatically
done,
because
if
we
have
the
crd
in
the
sink,
then
we
disable
dry
run.
But
on
the
on
the
other
branch
of
this,
where
we
have
the
resource
already
and
we
have
the
crv
here,
then
we
can
do
validation
and
we
I
I
try
to
do
the
same,
so
I
try
to
skip
dry
run
the
same
way.
A
B
B
I
think
my
my
microphone
was
muted,
when
I
was
talking
just
just
coming
back
to
the
to
the
future,
so
I
I
just
wanted
to
say:
actually
I
think
it's
really
useful,
and
probably
it
also
we
we
could
get
rid
of
the
one
annotation
right
of
the
skip
dry
run
on
missing
resources,
because,
basically,
we
we
wouldn't
have
to
skip
the
dry
run
right,
because
we
have
a
crd
locally
to
value
over.
B
And
I
think
the
what
what
is
what
you
suggested
alex
maybe
to
first
validate
against
the
cluster
and
if,
if
that,
fails
validate
against
locally
it's
it's
actually.
A
A
good
idea,
yeah,
I
think
peter,
was
pointing
out
that
it's
kind
of
hard
to
do
technically.
Just
because
of
how
you
know
all
this
written
and
I
was
trying,
I
was
hoping
to
find
a
way
to
do
it
in
in
the
you
know,
in
during
peer
review,
and
but
I
also
one
more
kind
of
motivation
to
avoid
having
annotations
is
because
there
are
a
lot
of
off-the-shelf
home,
charts
and
people.
You
know
don't
have
control
on
what's
in
the
chart,
so
they
won't
be
able
to
add
validation,
but
same
thing
might
happen.
F
G
H
And
thanks
alex
yeah
first
time,
argo
cd
contributor,
so
I
wanted
to
get
some
context
and
the
design
approach
on
this.
So
the
the
task
says
that
automated
test
that
mimics
user
behavior
for
ui
and
cli,
so
keith
long
he's
taking
care
of
the
the
uips
and
I'm
taking
on
the
clips,
and
I
I
wrote
some
assumptions
in
the
as
a
comment
and
I
wanted
to
validate
with
the
team.
H
If
those
assumptions
are
true,
so
when
I'm
doing
the
cli
testing
am
I
assuming
that
there
is
already
connected
to
a
kubernetes
cluster?
Argo
city
is
installed
all
the
required
dependencies
are
there
and
what
I'm
I
added
some
pseudo
codes
as
well.
So
basically,
I'm
doing
an
app
create
and
based
on
that
create
the
the
next
list
would
show
me
that
was
actually
created.
H
That's
a
very
basic
task,
but
then
I
could
also
see
if
there's
a
certain
number
of
pods
that
would
be
created
based
on
that
sarcastic
gap
create,
and
I
could
do
a
check
on
that
as
well,
and
I
want
it
to
so.
The
other
thing
is,
I
found
this
particular
test
I'll
paste.
The
link
in
the
chat,
so
that's
something
I
also
wanted
to
check
if,
if
I'm
expected
to
do
something
like
that
with
the
go
code
where
the
commands
are
run
within
like
a
goku
rather
than
having
a
shell
script.
A
Yeah-
and
I
think
I
can
give
some
context
about
why
we
created
a
ticket,
it
was
in
response
to
all
the
bugs
we
found
in
1.7
release
so
right
now
we
kind
of
rely
a
lot
on
early
adopters
and
so
that
we
run
our
ocd
into
it,
and
we
just
have
several
clusters
and
the
dog
food
clusters
allow
us
to
test
controller
really
well,
even
before
we
release
anything
and
what
was
missing
is
we
don't
have
much
of
we
don't
have
an
easy
way
to
test
user
interface,
because
even
if
we
release
a
kind
of
pre-release
or
viral
cd
to
internal
dog
food
cluster,
we
don't
have
enough
of
you
know
people
clicking
buttons
so
yeah
and
that
kind
of
answers.
A
Basically,
we
needed
some
tests
that
we
can
run
against
existing
clusters,
yeah
so
and
then,
basically,
it's
fair
assumption
to
say
that
argo
cd
is
up
and
running
and
the
script
already
it
gets
all
the
information
about
rbcd
from
outside.
Basically
it
would
it
should
be
given
url
and
then
access
key
and
all
of
that.
A
It's
like,
basically,
you
you're
trying
to
choose
between
bosh
and
cola
and
great
yeah
yeah.
That
would
be
another
question
I
I
feel
like.
Okay,
I
did
not.
I
have
not
tried
to
you
know.
Do
it
myself
yet,
but
I
just
adjust
the
suggestion.
So
wait,
isn't
it
saw
a
team,
another
team
and
they
tried
to
automate
they
needed
to
automate
cli
and
team
also
uses,
so
they
needed
to
write
an
automatic.
You
know.
A
A
So,
if
I'm
not
sure
if
you've
heard
about
people
there
but
idea
that
you
have
a
framework
and
that
framework
can
pass
basically
text
files-
and
you
can
write
some
google
code-
and
you
can
say
you
know
if
that
text
happens
to
be
in
the
text
file.
That
means
this
go
function.
So
basically
it's
like
a
combination
of
two
things,
so
the
user
still
can
write.
A
You
know
tests
just
in
the
text
and
we
can
keep
using
golang.
So
I
think
that
seems
to
me
the
best
possible
outcome.
If
we
get
a
framework
which
still
levels
this
calling
and
we
can
reuse
a
lot
of
functions
from
our
test
infrastructure
yeah.
So
that's
but
okay.
I
don't
really
answer
your
question,
but
the
idea
that
it's
like
opening
this
whatever.
H
A
It's
our
internal
e2e
test
and
it
would
be
it
would.
It
will
be
impossible
to
just
reuse
it
yeah
because
it
tightly
coupled
with
where
is
that
guy?
So,
basically
just
to
start
the
test.
You
have
to
use
a
fixture
and
the
fixture
so
methods
from
this
package.
A
H
H
Around
cli
so
for
the
cucumber
path
you
mentioned,
is
there
an
example
or
something
that
I
could
look
at
to
understand
how
to
how
to
do
that
work?
It's
yeah
I
can
okay.
Well,
can
I
get
it?
You
know
yeah
yeah,
you
can.
You
can
comment
later,
whichever
works
for
you
and
the
last
question
I
had
was
rather
than
just
the
basic
validation
of
argo
city
app
list.
If
the
app
was
created.
Can
I
do
some
in-depth
testing,
like
maybe
create
some
pods
or
or
someone?
H
A
A
Imagine
if
you
expect
ports,
you
can
just
use
argo
cd
cli
to
get
all
the
resources,
notification
and
check
if
what
is
inside
so
yeah.
That's.
I
think
I
try
to.
I
I
So
alex
I
have
a
question
here.
Yeah
suppose
I
want
to
yeah
like
the
one
said
that
okay,
so
little
bit
of
in-depth
validation
in
it
and
suppose
we
need
some
in-depth
validation.
Indeed.
So
for
that,
can
we
import
some
packages
like
where
you
can
use
that
to
do
the
like
in-depth
validation
of
like
port
status
and
whatever
we
have
so.
I
Yeah,
okay,
so
so
also
can
we
like
import
any
third
party
packages
here
in
argo
city
so
that
we
can
do
little
bit
of
in-depth
validation.
D
I
For
status
and
maybe
which
will
be
applicable
here.
A
Feel
like
it's
totally
fine
to
import
third
party
packages,
but
I
was
saying
that,
just
because
of
you
know
how
we're
going
to
use
that
test,
it
will
be
difficult
to
give
the
cluster
access
to
the
test.
So
it
won't
have.
You
know,
keep
config
to
talk
to
the
cluster
yeah,
because
it's
just
given
giving
more
so
yeah,
it's
like.
A
Basically,
we
will
try
to
use
that
around
the
test
against
kind
of
production
cluster
view,
so
which
should
be
that
we
need
a
guard
which
will,
which
would
be
basically
I'll,
go
see
the
api.
So
it
should
be
really
not
really.
It
should
be
test,
but
let's
make
changes
in
rvcg,
but
most
likely
we
will
have
to
so
configure
rgocd
and
then
create
a
test
project
with
limited
permissions
and
the
test
project
test.
A
A
Thank
you
and
then
yeah,
so
I
I
was,
I
kind
of
I
almost
want
to
take
my
system
about
cucumber
it's.
It
seems
to
me.
Maybe
that
you
know
cucumber
is
like
a
next
step,
but
maybe
it
is
the
best
possible
way
is
to
just
reuse
golden
framework
which
we
have,
but
to
kind
of
to
to
reduce
it
we
would
have
to.
We
cannot
use
codes.
A
We
cannot
run
tests
in
the
in
that
folder
if
we,
because
it's
just
all
the
tests
are
implemented
in
a
way
that
code
inside
of
the
folder
assume
that
we
have
access
to
test
cluster
and
they
try
to
do
too
much.
They
clean
up
very
aggressively
yeah,
but
maybe
the
best
possible
way
would
be
to
create
a
smoke
test
in
inside
of
test
yeah
and
then
just
reuse,
whatever
needed
from
from
existing
functions.
I
H
Yeah
so
smoke,
let's
say:
smoke
test
folder
under
test,
but
not
e
to
e
and
within
that,
whichever
third
party
package
we
need
similar
to
testify
or
anything
that
runs
a
shell
command
and
that
or
cli
command.
That
would
be
argo,
cd
command,
and
so,
whatever
validation
we
do,
we
will
target
the
api
and
validate
it.
A
H
A
And
I
think
we
can
assume
kind
of
safely
that
this
test
can
do
whatever
it
needs
to
do
with
argo
cd
using
calculated
api,
because
we
will
need
to
configure
it
so
it
has
limited
access
to
electricity.
So,
for
example,
you
can
delete
applications
exactly
you
know.
If
you
need
to
prepare
a
safe
state,
then
you
know
it's
okay
to
delete
applications
and
you
create
them
because
arbor
city
lets
us
configure
access
properly.
I
I
was
thinking
I
feel
like.
A
We
could
we
can
run
headless
home,
and
I
know
that
I
did
it
myself
before
it's
like
it
doesn't.
Basically,
as
long
as
you
can
run
a
container,
you
could
so
we
can
run
a
selenium
grid,
so
it
must
be
basically
this
syringe
server
which
and
it
has
access
to
a
headless
chrome.
And
then
you
can
just
use
any
api
to
talk
to
the
syrian
server
and
it
will
interact
with
first.
A
I
A
And
like
to
be
honest
right
now,
we
kind
of
rely
completely
on
github
accounts,
so
whatever
github
actions
provide,
we
try
to
stick
to
it.
I
think
in
worst
case
we
can.
We
also
have
a
kubernetes
cluster,
which
can
run
a
little
more
heavier
things.
It
will
be
harder
to
manage,
but
it's
also
a
possibility.
A
B
Yeah,
so
basically,
this
this
emerged
from
from
several
questions
on
slack
or
conversations
I
I
recently
had
on
slack
and
with
all
the
new
contributors
now
coming
up,
and
everyone
faces
basically
similar
questions,
how
to
do
stuff
and
and
where
to
find
stuff
and
like
how
do
I
change
api?
And
what
do
I
have
to
change
for
that,
and
I
feel
we
have.
B
We
have
a
place
currently
for
contributed
documentation
that
lives
in
git
right,
but
I
think
that
git
might
not
be
the
right
place
for
for
such
kind
of
fast,
changing
short,
maybe
short
living
documentation,
and
I
just
wanted
to
put
in
the
discussion
because
the
the
github
repository
it
provides
like
a
wiki.
B
It
was
used
some
ages
ago.
I
seen
the
couple
of
documents
in
it
and
yeah.
I
just
wanted
to
put
up
for
discussion
whether
we
we
want
to
push
all
this
non-user-facing
documentation
into
the
wiki,
maybe
so
it
it
can
be
easily
edited
and
updated
without
having
to
create
a
new
branch
in
your
local
copy,
create
a
pr,
have
someone
review
and
prove
it
and
stuff
like
that,
and
I
think
the
wiki
yeah
it
would
at
least
provide
us
with
with
a
revision
history
right
who
changed
what
so.
B
Yeah
this
is
this
is
basically
a
discussion
topic
like
what
do
others
think
of
it.
B
You
know
dogs
they
have
so
I
think
it's
it's
good
to
have
the
user
facing
docks
somewhere
in
the
git,
because
we
we
generate
the
site
from
it
and
stuff
like
that
yeah,
but
for
the
so
things
like
developer
how-tos.
Let's
call
it
like
that,
so
maybe
someone
submitted
documentation
that
contains
a
typo
in
a
path
or
something
like
that
and
to
fix
this.
It's
it's
quite
exhausting
right.
If
you
have
to
create
a
pr
and
and
and.
B
I
don't
know
if
it's
possible
to
further
restrict
it
down
or
if
it's
even
required
to
further
restrict
it
down.
A
B
So
yeah,
I
think
that
there
will
be
lots
of
mistakes
in
the
documentation,
it's
in
the
nature
of
things,
yeah
right,
but
it's
actually.
J
G
I
mean
that's
just
like
a
good
idea:
it's
mature
from
red,
I
it
might
be
simpler
right
and
then,
if,
if
at
some
point
the
dark
start
stabilizing,
we
could
always
bring
that
into
the
to
be
part
of
the
repo,
but
it
allows
that
kind
of
a
iterative
inputs.
I
mean
I
like
that
idea.
Yeah.
J
J
I
can
also
look
at
look
to
see
if
cncf
provides
any
kind
of
this
kind
of
service
that
we
might
be
able
to
use,
but
lacking
that
we
could
set
up
a
document
repo
and
just
let
anyone
who's
a
maybe
an
org
level
member
to
update
it
or
even
anyone
else.
I
mean
if
there's
no
problem,
there's
no
harm
in
letting
other
people,
but
there
could
be
some
noise.
I
suppose.
A
A
A
Yeah
github,
I
think
you
talked
about
github
rights.
Yes,.
B
E
D
D
I
didn't
know
you
could
do
this,
but
dejuan
just
posted
in
the
chat
that
you
can
make.
You
can
remove
that
check,
so
anyone
can
edit
it,
but
then
anyone
can
edit
it.
So
it's
either
contributing
collaborators
only
or
everyone.
J
Yeah
so
then
that
means
that
any
say
org
member
could
commit
the
change,
but
someone
who's,
not
an
org
member,
could
send
apr.
A
And
I
I
might
be
wrong,
but
I
feel
like
so
we
had
we
had
to
go
for
the
procedure
of
you
know
we
had
to
figure
out
who
I
remember,
you
know
which
expensive,
but
it's
done
already,
and
I
guess
the
next
step
is
to
pretty
much
add
everyone,
at
least
from
red
hat
team
who
actually
contributes
as
members.
Is
it
correct?
I
think.
J
Yeah,
so
the
org
membership
appears,
I
think,
jesse
just
recently
updated
the
member
rules,
but
the
intent
is
that
if
you
are
an
active
community
member,
it
should
be
very
easy
to
become
an
org
level
member.
It
doesn't
give.
A
J
J
A
J
F
B
All
right,
yeah
sounds
fine,
okay
and
yeah.
If
it's,
if,
if
we're
gonna,
try
it
I'd
step
forward
and
and
set
up
a
general
structure,
maybe
for
the
for
the
new
wiki
and
okay
yeah.
H
So,
and
are
we
talking
about
two
separate
ripples?
Where
would
be
one
would
be
the
developer
facing
documentation
and
one
would
be
the
the
user
docs?
One
of
the
link
I
pasted
it
says,
like
the
documentation
is
out
of
date
and
will
be
updated
later.
Is
that
the
developer
doc
or
the
user
doc
you're.
A
B
But
darwin
the
the
link,
you
posted,
read
the
docs
I
think
alex
and
me
we
have
discussed
in
the
past
that
this
could
be
an
option
for
user
facing
documentation
actually
right
so
because
it
it
features,
versions
and
and
stuff
like
that
and
but
yeah
probably
we
can
discuss
it
the
next,
the
next
meeting,
maybe
I
can
put
it
on
the
agenda
because
we
only
have
10
minutes
left.
So
I
think
this
this
might
be
a
longer
discussion.
Maybe.
A
K
K
What's
important
is
that
you
know
the
quality
of
the
content
and
its
accessibility,
discoverability
search
ability,
the
big,
the
big
issue
we
had
in
our
go
seating
and
the
big
issue
we
have
on
especially
on
workflows
today
is
peop.
There
are
a
lot
of
questions
that
have
been
answered
out
there,
but
people
can't
find
their
way
to
the
to
those
answers
and
in
the
documentation.
So
so
more
of
our
effort
is
really
focused
on
making
sure
the
documentation
is
up
to
date.
Incorrect-
and
I
I
I
feel
like
my
question-
was
about
pretty.
A
Much
just
the
dev
documentation,
you
know
like
internal.
I
like
how
it's
like
kind
of
yes
and
what
we're
thinking
is
that,
because
it's
so
difficult
to
change
it
now
in
the
main
import
you
have
to
you
know,
at
least
you
have
to
wait
for
like
20
minutes
long
pull
request.
The
proposal
is
to
move
it
into
a
separate
repository
where
you
can
make
changes
much
easier
easier,
even
without
even
pr,
basically
just
have
a
knowledge
base
in
the
keyboard.
A
And
then
I
ask
yeah
what
about
using
a
github,
wiki
page
yeah,
we're
going
to
consider
that
it's
like
either
wiki
or
and
if
we
ended
up
using
the
git
repository.
We
have
to
give
it
a
name
and
we
have
to
either
name
it
there
or
argo
dev.
What
do
you
think?
So?
If.
K
A
K
J
All
right,
okay,
so
alex
you're,
saying
that
you
would
continue
to
use
argo
cd
for
developer
or
official
project
doc.
But
you
would
use
that,
like
the
community
contribution,
docs
repo
for
yeah.
K
K
I
mean
people
have
trouble
with
the
code
generation
because
you
know
golan
code
generations
is
pretty
flaky
instead
of
script
at
the
end
of
the
day,
isn't
it
you
know
painful
to
manage
dependencies,
but
if
people
want
to
run
the
the
system
up
and
install
it,
they
just
run
check
the
code
out
and
make
starting.
You
know
they
can
start
making
their
code
changes
about
that
point.
I
don't.
I
don't
really
know
why
we
don't
have
those
issues,
but
we
don't
really
seem
to
have
them.
H
Was
I
was
there
a
question?
Sorry,
I
was
having
a
quick
question
about
faqs.
Would
that
be
the
regular
dog
or
developer
dog,
because
on
the
slack
I
see
a
lot
of
people
asking
like
the
same
question
and
it
gets
lost
and
maybe
in
order
to
be
more
efficient
if
there
was
a
way
for
us
to
include
an
faq
of
commonly
asked
questions
and
every
time
someone
asks
a
question
we
could
keep
referring
to
that
until
a
point
where
most
common
questions-
people
just
start.
Referring
to
that
and
what
the
ideal
place
for
that.
J
A
Yeah,
it's
an
open
kind
of
issue
and
I
I
think
that
we
were
supposed
to
do
something
for
that.
We
just
we
didn't
have
time.
My
was
one
idea
was.
I
just
recently
found
a
report
where
I
tried
to
create
an
issue
and
then
the
issue
had
some
pre-filled
text
which
basically
encouraged
me.
It
gave
me
a
link
to
stack
overflow
and
I
was
thinking
to
propose
to
do
the
same.
You
know
just
just
provide
the
link
to
stack
overflow,
which
already
includes
argos,
dpeg
and
then.
A
Yeah,
so
this
is
answer
to
your
question
like
this
is
one
idea
to
kind
of
just
discourage
people
from
creating
issues.
That
is
a
question
and
tell
them
please
go
to
stackoverflow
and
apply
this
stack.
A
J
A
J
K
I
can
give
you
guys
a
bit
of
feedback
on
our
experience
of
this.
It's
quite
it's
early
days.
We've
only
been
pushing
what
encouraging
users
to
ask
their
questions
on
stack
overflow
by
installing
a
config.yaml
file
which
changes
the
options
that
people
are
given
when
they
go
to
create
an
issue.
K
So
when
you
go
to
create
an
issue,
we
previously
had
make
a
enhancement
issue,
make
a
bug
issue
or
make
a
question
issue
and
the
question
one
now
is
actually.
Instead,
it's
changed
to
be
a
link
to
stack
overflow.
That's
that
technical
change.
What
what
I
mean?
That's
meant
to
drive,
behavioral
trains
and
we
do
see.
I
think
I
think
it's
pretty
reasonable
to
say
we
see
fewer
questions
asked
via
github,
but
we
haven't
seen
the
same
volume
of
questions
being
asked
on
stack
overflow.
I
don't
think
I
think.
Actually
people
are
asking
them
there.
K
You
are
asking
more
of
them,
but
it's
not.
You
know
not
the
two
or
three
questions
we
used
to
get
every
single
day.
It
seems
like
people
are
now
asking
those
questions
in
slack
to
me,
that's
how
that's
what
their
behavior
change
has
become.
So
we've
kind
of
driven
people
out
of
github
into
into
slack,
maybe
not
driven,
use
different
words
to
driven
encouraged.
B
Yeah,
probably
it's
it's
a
little
bit.
It's
it's
a
it's
a
small
hurdle,
because
I
think
you
need
to
sign
up
for
stack
overflow
as
well
right
and
I
don't
know,
do
they
have
a
single
sign-on
with
github.
B
Maybe
because
I
think
it's
yeah,
it
might
be
a
small
hurdle,
but
talking
for
myself,
I
I,
for
I
don't
have
a
stack
overflow
account
personally,
so.
K
Maybe
enough
those
single
sign-ons
don't
work
very
well.
We
had
a
problem
in
our
community
meeting
yesterday,
where
somebody
one
of
the
presenters
couldn't
get
in
because
he
couldn't
sign
on
with
single
sign-on,
because
he
incorrectly
entered
his
date
of
birth
as
being
under
16
years
old,
and
the
error
message
about
the
fact
that
he
entered
his
date
of
birth
incorrectly
by
accident,
and
you
know,
spent
20
minutes
bashing
his
head
against
zoom's
single
sign-on,
unsuccessfully.
So
it
means
it
can
completely
prevent
a
small
proportion
of
people
engaging
in
the
software
application.
K
I
I
think,
like
might
be
better
option
than
stack
overflow
for
many
people,
but
it's
certainly
a
better
option
than
than
having
lots
of
issues
opening
github.
That
kind
of
consume
developer
bandwidth.
But
you
kind
of
the
downside
is
you
lose
the
documentation
for
their
solution?.
L
L
K
A
A
To
make
it
successful,
I
feel
like,
as
you
know,
we
needed
several
people
who
can
subscribe
to
questions
and
then
answer
them
just
to
start
it
kind
of
rolling.
So
people
get
feel
that
yes,
technical
form
is
working
and
then
keep
asking
questions
instead
of
because
if
we
just
redirect
people
to
stick
overflow
and
then
no
one
answers
yes,.