►
From YouTube: Community Meeting, November 8, 2022
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
Hello
today
is
November
8th.
This
is
the
kcp
community
meeting.
Welcome
everybody.
If
you
have
a
topic
you'd
like
to
discuss
today,
we
do
have
our
visual
GitHub
issue
for
the
agenda,
so
please
feel
free
to
add
comments
to
this
issue.
If
there's
anything
you'd
like
to
talk
about
we're,
gonna
wait
on
the
first
topic
until
hopefully,
Stefan
has
a
chance
to
join
today,
he's
running
a
bit
late,
so
Paul
I'm
gonna
go
to
your
topic.
First.
B
Okay
sounds
good,
so
I
got
here
a
really
quick
proposal
about
something
about
refinement
processes,
basically
we're
a
pretty
small
group
now
and
we
all
work
pretty
well
together,
but
I
hope
we
grow
a
lot
in
2023
and
I
hope
that
we
scale
to
meet
those
needs
and
I,
don't
think
hope
is
probably
a
good
strategy
for
us.
So
what
I?
B
What
I
just
tried
to
put
on
paper
is
some
things
that
we've
talked
about
in
the
past,
essentially
that
we
probably
need
to
get
a
little
more
formal
in
how
we
write
up
our
enhancements
so
that
people
can
discover
them
understand
the
motivation
and
history
behind
them.
B
We've
got
right
now
a
couple
different
areas
within
the
repository,
and
that
is
growing
just
off
the
top
of
my
head
I
could
think
our
control
plane
the
Tennessee
items,
workload
items
we've
now
got
Edge
workloads.
If
you
haven't
seen
that
there's
a
new
repository
for
Edge
we've
got
controller
development
and
we've
also
got
a
catalog
repository
on
there.
B
So
it
probably
makes
sense
to
split
repositories
where
we
can,
in
order
to
scope
the
responsibilities
of
what's
in
each
and
allow
contributors
to,
you
know,
contribute
on
the
area
they're
most
interested
in
as
part
of
our
enhancement
proposal
process.
We
probably
need
to
have
a
good
group
agreement
on
what
is
our
definition
of
done,
how
much
documentation
Does
it
include
for
features.
Where
does
that
happen?
B
What
are
the
testing
requirements,
and
then
I
also
propose
in
here
that
we
could
probably
backport
a
little
bit
of
this
process
and
get
some
benefit
from
that,
and
by
that
I
mean
we've
got
a
bunch
of
apis
out
there.
Now
what
are
their
current
levels?
What
is
their
graduation
criteria?
Look
like
to
get
to
Beta
or
or
V1,
so
that's
what
I
would
like
feedback
on
the
pr
is
in
the
agenda.
A
We
go
so
yeah
if
anybody's
got
any
comments
happy
to
entertain
them
now.
We
will
obviously
also
accept
them.
Async
Paul.
Would
you
be
willing
to
send
a
quick
note
to
the
kcp
dev
Google
group,
just
referencing
this
and
I
would
encourage
you
to
say
that
we're
going
to
do
one
week,
lazy
consensus
so
by
next
community
meeting.
A
If
there
are
no
significant
outstanding
issues
or
knacks,
we
will
adopt
it
as
going
forward,
if
not
sooner
but
one
week,
lazy
consensus
has
tended
to
work
pretty
well
in
other
projects.
I've
been.
B
Involved
in
yeah
it
sounds
great
and
as
people
read
through
you'll
you'll
notice
that
number
five
on
there
is
that
I
want
us
to
own
this.
As
a
group,
I
didn't
try
to
put
strict
definitions
in
there,
I
think
we
can
probably
come
up
with
a
tailored
process
that
borrows
from
other
communities.
We've
been
active
in
so.
A
Okay,
so
I
had
the
next
topic
here
on
2022
releases.
If
y'all
were
at
the
last
community
meeting
a
week
ago,
I
brought
up
the
idea
that,
rather
than
trying
to
do
two
more
releases
this
calendar
year
before
we
get
to
2023
that
it
might
make
more
sense
to
try
and
just
do
one
and
we
can,
if
we
all
agree
to
doing
that,
we
can
decide
if
we
want
to
try
and
do
that
sooner,
like
in
the
next
week
or
so
or
sometime
towards
the
beginning
to
middle
of
December.
A
Before
we
get
into
the
later
holidays,
so
I
sent
an
email
to
kcp
Dev
last
week,
I
think
and
I
didn't
get
any
feedback,
so
I
think
that
counts
as
lazy
consensus
to
a
degree,
so
I'm
probably
going
to
suggest
that
we
just
do
a
release
in
about
month.
A
There's
unless
there's
something
pressing
where
we
need
to
get
0.10
out
the
door
now
so
again.
Call
for
feedback
and
I
will
follow
up
to
my
post
to
kcp
Dev
with
a
more
concrete
proposal
and
see
if,
if
folks,
have
any
comments
but
we're
we're
not
rushing
to
get
a
release
out
at
this
time
and
there
there's
bug,
fixes
and
features
in
flight.
So
so
yeah,
that's
pretty
much.
It.
A
All
right,
I'm
not
going
to
make
you
wait,
Frederick,
given
that
Stefan
is
still
in
another
meeting,
so
I'll
share
this
or
if
you
want
to
share
and
drive,
we
can
go
ahead
and
start
the
discussion.
What
would
you
prefer?
Oh
yeah.
D
All
right,
so,
you
should
see
my
screen
right,
so
I've
created
a
Academy
notes,
which
is
a
summary
or
process
as
a
result
of
the
discussion.
We
we
added
in
the
Google
Doc
and
basically
add
two
to
flow
as
a
pattern
from
previous
documentation
with
motivation
called
and
goal,
and
the
proposals
I'm
quickly
over
that
because
we
I'll
discussed
start
last
week
and
I
would
like
just
to
present
the
proposal.
D
So
this
proposal
is
to
introduce
three
new
resource
type,
the
primary
one
being
API
export
endpoints
and
this
resource
type
would
be
a
sync
you
specify
in
the
spec
an
API
export
that
you
locate
next
for
API
bending
by
providing
a
pass
and
a
name,
you
can
also
optionally
specify
partition
and
when
you
have
done
that
as
a
research,
you
get
a
list
of
that's
our
the
end
point
for
your
API
export
service
in
in
the
shards.
D
So
one
question
is
why
to
introduce
with
a
new
resource
instead
of
using
it
API
export,
as
it
was
in
the
past.
One
reason
is:
you
can
create
this
resource
locally
in
in
a
workspace
and
gets
this
information
populated
by
The
System
Controller.
Instead
of
trying
to
access
the
global
resource
like
API
exports.
D
And
another
reason
is
the
possibility
to
Partition,
so
we
can
look
now
as
his
second
resource.
D
So
a
partition
is
basically
the
select
or
list
of
selectors
in
the
spec
and
based
on
this
selector,
you
will
select
a
set,
a
group
of
shards,
for
instance.
Here
you
can
see,
we
select
the
cloud,
we
select
a
region
and
you
can
also
exclude
things
if
you
want.
So
in
this
case
we
exclude
one
country,
we
we
don't
want
any
shot
from
North
Korea.
In
this
example,.
D
So
this
next
step
is
that
we
want
to
provide
the
mechanism
to
slice
and
dice
automatically
and
for
that
we
have
defined
a
resource
partition
set
that
you
may
create.
So
this
is
a
service
provider.
May
create
his
resource
in
the
workspace.
Where
say,
API
exports
is
created
and
you
can
Define
Dimensions.
So,
for
instance,
here
Dimension
is
origin
and
cloud
and
it
will
automatically
create
a
partition
matching
this
Dimension.
So
you
will
have
a
partition
with
Cloud
gcp
Europe.
You
will
have
a
partition
with
Cloud
gcp
us.
D
You
will
have
a
partition
with
Azure
Europe,
a
partition
with
issue
us
and
you
can
use
this
partition
so,
as
we
mentioned
in
the
API
export
endpoint,
but
you
can
also
use
this
information
for
the
scheduling
of
the
controller
So.
Currently
we
have
in
cluster
workspace
a
label
selector,
and
we
will
have
that
also
in
the
workspace
here
in
the.
D
That
it
will
match.
So
if
you
specify
this
selector,
when
you
create
a
new
workspace,
you
you
will
have
a
replacement
that
much
the
partition.
D
D
Yes,
is
that
basically
it
so
the
the
main
points
do
you
have
any
questions.
D
A
Some
time
to
read
through
this,
but
what
you
presented
seems
to
make
sense.
I,
unfortunately,
haven't
had
a
chance
to
look
at
the
Google
Doc
that
precedes
this.
So
I'm,
assuming
that
this
came
out
of
the
Google
Doc
and
is
largely
agreed
upon,
is.
D
That
accurate
yeah-
yes,
that's
correct,
so
I've
been
if
you
couple
of
iteration,
it's
also
the
last
bit
of
the
google.com
on
this
one.
So.
A
D
I
recall
today,
earlier
with
Steven
and
Stefan,
to
like.
D
The
last
iteration
on
it
and
and
that's
the
result
for
for
now
so
obviously
we
can
amend
it
further
with
this
additional
inputs
as
a
die.
Cuts.
A
Yeah,
so
the
the
only
comment
that
I
have
for
anything
that
stands
out
is
that
we
don't
want
to
name
the
resource
ending
in
an
S
like
API
export
endpoints.
We
don't
want
it
to
be
plural,
so
we'll
need
to
come
up
with
another
name.
That
is
not
plural
and
I.
Don't
have
a
good
one
off
the
top
of
my
head,
maybe
something
like
export
locator
or
something,
but
it
needs
to
be
singular.
D
Okay,
all
right
I
think
so
so
we're
supposed
so.
D
All
right
so
there
are
similarities,
but
it's
not
one
one.
So
the
user
will
be
responsible
for
creating
the
resource
what
it
gets.
It
gets
the
status
automatically
populated.
So
that's
a
full
match.
Endpoint
in
services.
F
D
F
One
yeah
and
that's
the
reason:
API
exports
will
be
distributed
across
charts
eventually,
so
there's
a
caching
layer
and
we
distribute
those
objects
if
they
are
big
and
changing
frequently,
it's
not
good
right.
So
exports
should
be
mostly
constant,
not
immutable,
but
they
don't
change
much,
that's
why
we
want
to
move
it
out
and.
D
All
right
so
as
there
would
be
two
steps,
so
the
first
one
would
be
to
create
a
partition
set
a
resource.
A
D
D
That
that's
that
Buzzy
Pleasers
it
was
a
spec.
He
is
a
selector.
So
that's
the
selector
that
you
would
apply
when
you
create
a
step
work
space,
for
instance,.
F
A
A
F
Probably
there
will
be
assume
Target
endpoints,
as
Felix
said,
and
we
can
make
it
so
it's
owned
by
the
API
Sim
Target
right,
so
the
owners,
the
same
Target,
also
owns
the
endpoints
option.
So
you
can
fill
in,
for
example,
a
sparse
list
of
endpoints
so
like
not
every
chart
but
I.
Think
I
should
connect
to
those
three
shards
because
yeah
so
think
about
this
between.
We.
F
F
Saying
no
because
we
said
the
same
thing
and
we
found
reasons
why
we
don't
want
it
okay,
so
one
of
these
is
permissions
like
the
owner
of
the
API
string,
Target
or
the
owner
of
the
API
API
export
has
to
fill
that
right.
If
TMC
is
not
privileged
anymore,
it
cannot
see
all
the
endpoints
it's
responsible
for
that's
what
that's
why
it
should
be
a
different
object
for
TMC
swing,
Target
inputs,
and
the
second
reason
is
the
semantics
behind
that
is
pretty
different.
I
mean
the
filtering
in
swing.
E
F
It's
intentional
for
permission
reasons:
I
see,
I
mean
we
came
from
generic
and
moved
away
for
reasons.
It's
it's
a
balance.
We
have
to
decide
I.
G
Yeah,
because
these
are
different
types
but
but
the
you
know
there
are,
they
are
common
in
the
at
least
parts
of
their
structure
and
and
their
logic,
so
maybe
having
the
ability
to
you
know,
reconcile
them
through
a
a
common.
F
A
Okay,
so
I
know,
Paul
had
talked
about
a
proposal
for
the
process
on
doing
these
sorts
of
things,
but
we
haven't
ratified
that
yet
so
next
steps
on
this
I
know
you.
You
have
a
to
do
and
a
question
here
about
self
and
the
partition.
F
F
A
So
it
seems
reasonable
to
me
other
than
coming
up
with
a
new
name
for
this.
That's
not
plural.
F
Is
this?
Is
this
technical,
or
is
this
just
taste,
because
endpoints
in
Cube
up,
plural,
yeah,.
A
Endpoints,
in
Cube
being
plural,
was
a
source
of
problems
for
various
things,
so,
like
the
new,
the
new
V2
so
to
speak,
is
called
endpoint
slice.
F
A
I
mean
there's
nothing
stopping
implementation
at
this
point
like
it's
fairly
easy
to
rename
something
with
search
and
replace
and
code
gen,
so
yeah
I
would
start
with
slice
and
if
we
can
come
up
with
a
better
name,
we'll
try
and
find
one.
A
All
right
so
just
to
recap,
stuff,
on
real,
quick
on
this.
One
I
was
following
up
from
last
week
and
proposed
that
we
do
one
more
release.
This
calendar
year
in
approximately
a
month
and
I'm
gonna,
send
a
follow-up
post
to
the
Google
group.
A
Anybody
have
any
other
topics
before
we
go
into
new
issue,
triage.
A
All
right,
if
you
think
of
something,
please
feel
free
to
speak
up,
we
got
20
things
in
here:
lots
of
bugs
wow,
mostly
bugs
all
right,
we'll
we'll
go
from
top
to
bottom
here.
A
So
we
have
this
one
which
I
saw
three
weeks
ago
and
haven't
gotten
to
I,
also
saw
one
that
just
came
in
that
I
think
is
most
likely
a
duplicate.
So
if
anybody's
interested
in
looking
in
to
the
quota
sub
system,
I'm
happy
to
provide
some
mentoring
there,
I'm
actually
gonna
say
it
might
be
a
dupe
of
2220.
B
A
All
right
intermittent
error
trying
to
create
a
workspace.
A
Interesting,
so
I
think
this
one
just
needs
somebody
to
look
into
it.
Anybody
interested
and
have
time
to
look
into
some
of
the
CLI
and
workspace
creation
stuff,
if
not.
A
As
they
can
find
any
export
through
a
VW
of
an
API
expert
that
claims
bindings,
did
we
I
think
this
one
probably
needs
meteors
or
actually
surges?
Are
you.
D
A
And
you
said
you
had
to
not
be
here:
okay,
Stefan,
have
you
looked
at
this
one.
A
For
API
export
virtual
workspaces
doesn't
work
probably
should
add
that
in
there,
mainly
because
they're
they're,
like
Helm,
is
trying
to
use
it.
A
A
E
E
E
A
E
A
E
And
that's
that
was
there
like
the
original
types,
existing
API
returns
properly,
but
it's
more
like
you
see
it.
The
error
type
is
different,
it
doesn't
return,
then
you
list
it.
It
lists.
It
gives
you
an
empty
list
or
like
little,
but
this
is
says
404.
So
it's
just
something
but
yeah
I
will
take
out
more
details
and
do
some
chasing.
A
Steve
Hardy:
are
you
still
working
on
this
or
pieces
of
this?
The
Sharda
test
server
with
Standalone
virtual
workspaces.
H
Yeah
I
started
going
through
your
branch
and
push.com
a
couple
of
PRS
which
were
kind
of
not
related
to
the
actual
separation
of
the
virtual
workspace
server,
but
I
was
planning
to
try
and
get
the
rest
of
it
up
this
week.
I'm
just
got
a
sidetracked
by
some
other
tasks.
Okay,.
A
All
right
root,
compute,
maximal
permission
policy,
doesn't
Grant
access
to
oh
right
yeah.
This
is
just
something
that
needs
to
be
implemented
anyway.
So
if,
if
folks
are
interested
in
contributing
and
have
some
time,
this
is
just
editing
some
yaml
happy
to
provide
a
pointer
also,
it
can
just
do
it,
but
I
would
love
to
let
folks
participate
as
well.
So
if
anybody
is
interested
in
working
on
this,
let
us
know.
A
I'm
Gonna
Leave
it
unassigned
but
feel
free
like
if
you
know,
if
it's
the
sort
of
thing
where
you're
not
going
to
get
to
it
for
several
days,
let's
leave
it
unassigned
and
then,
if
it's
still
unassigned
and
you're
free
in
a
week
or
so
feel
free
to
pick
it
up.
A
This
flake
has
been
around
for
a
while
David
have
you
ever
Joachim
had
a
chance
to
figure
out?
What's?
Oh,
is
this
the
missing
spec
bit.
A
Got
it
yeah,
so
I
don't
know
if
you've
seen
this
stuff
on,
but
we're
not
really
sure
why.
But
something
is
snipping
out
the
spec
from
various
objects
like
either
as
soon
as
they're
being
created
or
sometime
later
you
end
up
in
this
case
it
happens
to
be
with
deployments,
but
I
think
we've
seen
it
elsewhere,
where
this
Peck
just
disappears.
So
I.
G
Yes,
what
what
led
to
think
this
was
that
you
are
Kim
tested
this
in
the
entrance
test,
which
is
obviously
Synchro
related.
But
finally,
the
end-to-end
test
creates
directly
in
in
kcp,
creates
a
deployment,
and
the
deployment
you
get
just
out
of
the
create
call
is
already
without
the
spec
can.
F
You
can
you
check
audit
locks
and
see
eyes
are
pretty
bubbles.
You
have
all
the
objects
everything
sent
and
received
from
a
request.
Yeah.
C
We
just
started
this
started
happening
when
Steve
bumped
up
the
the
actual.
Well,
the
the
number
of
parallel
tests
so
looks
like
something
that
is
happening
when
trying
to
I
don't
know
hit
something
that
isn't
ready.
G
Yeah
it's
obviously
it
was
happening
with
quite
a
high
number
of
parallel
threads
I
mean
where
I,
don't
remember
what
you
said.
Joking,
but
maybe
15
I
mean
as
soon
as
soon
as
you.
As
long
as
you
wear
an
under
a
certain
number
of
of
a
certain
load.
I'd
say
it
was
okay
and
then,
if.
A
Yeah
I
mean
I
I,
know
that
we've
seen
this
on
a
kcp
instance
that
we're
running
there's
just
random
legitimate
deployments
that
are
missing
spec.
So
you.
A
Sorry,
I
I,
don't
think
that
it's
like
there
is
an
issue
somewhere
and
like
upping
the
test
parallelization
to
exacerbate
the
issue
is
probably
a
good
thing,
so
you
know
it's
not
just
like
a
test
issue.
It's
a
real
issue.
C
I
was
able
to
rip
reviews
that
somehow,
luckily
and
I
just
stopped
the
test
and
I
was
able
to
create
a
full
deployment.
I
mean
even
at
the
star.
The
deployment
was
getting
the
spec
removed.
Then
I
was
able
to
just
you
know:
Cube
CTL
create
deployment,
and
everything
was
was
perfect.
So
really
weird.
A
Okay,
thank
you
so
time
to
Steve.
G
Maybe
sorry,
let
me
just
just
Demi
thought
that
Summerside
apply
with
things
like
deployments
is
based
and
it's
it's.
A
quite
old
patch
in
one
of
the
first
batch
in
in
kcp,
is
based
on
the
open
API,
which
is
you
know
the
open
API
for
which
is
based
on
the
crd,
which
was
not,
which
is
not
in
standard
Cube.
G
Instead
of
cube,
you
can
apply
server
side
apply
only
to
objects
that
have
you
know
that
that
are
entry
real
deployments,
so
it
there
might
be
some
something
that.
F
F
No,
no,
no,
you
were
talking
about
strategic,
merge
patch.
Oh.
G
Sorry
I,
don't
know
strategic,
yes,
you're
right
next.
B
A
I'm
gonna
put
this
to
next,
because
it's
definitely
not
backlog.
I
know
Steve's
owl.
A
If
anybody
has
API
server
background
and
wants
to
look
at
this,
please
feel
free
I
will
try
to
get
to
it
as
well,
but
this
week
is
kind
of
busy
foreign.
E
A
Oh,
this
is
the
CEO.
Oh
I
found
okay,
this
is
actually
a
duplicate,
so
this
is
a
duplicate
of.
A
Interesting
that
it's
in
the
same
code
path
like
when
you
look
at
the
stack
Trace,
that's
coming
through
I,
can't
make
this
one
much
bigger.
So
it
comes
through.
Sorry
comes
through
CEO
trying
to
track
cost,
but
the
ultimate
seg
fault
comes
from
the
underlying
go
interface
code
and
it's
a
very
strange
one.
F
Because
we
bumped
bumped
CL
or
we
are
oh.
A
No,
this
was
from
you
know
two
three
weeks
ago
and
I
found
this
issue
from
five
years
ago,
where
it
had
like
the
same
sort
of
same
sort
of
Stack
race
where
it
was
going
through
iface.go.
And
if
you
looked
at
the
code
it
they
like
they're,
basically
like
it
shouldn't
happen,
and
it's
not
repeatable
and
it's
weird
so
I
don't
know
what
to
do
with
it
other
than
if
we
keep
seeing
it
like,
maybe
updating
to
a
newer,
CEO
or
different
Go
version
or
something
but
I
don't
know.
A
All
right,
Sinker
health
backlog,
anybody
working
on
this
or
no.
A
To
you,
yeah.
A
F
A
E
I
have
a
question
yeah,
so
I
posted
this
in
the
weekend,
I,
basically
I'm,
not
everybody
had
a
broken
weekend
and
didn't
come
to
at
some
point.
We
discussed
why
the
names
for
the
names
for
the
workspaces
must
start
with
a
letter
instead
of
the
standard,
kubernetes,
validation
pattern
and
remember,
we
discussed
it
like
months
ago
and
we
didn't
get
anywhere.
D
A
I
think
maybe
at
some
point
we
thought
they
needed
to
start
with
a
letter,
and
we
talked
about
this
either
in
slack
or
in
a
PR
somewhere
in
the
past
month
or
two
Stefan.
Do
you
recall.
A
It
was
about
workspace
names
and
why
they
can't
start
with
a
number.
F
E
E
F
A
Yeah
I
can
follow
up
on
that
I'm.
Trying
to
find
the
slack
thread
you
posted.
A
Yeah
I
will
I'll
take
that
and
either
get
it
fixed
or
explain.
Why
how's
that.
E
A
Yeah
yeah
I'll
follow
up
thanks.
A
Okay,
anything
else
or
you
all
wanna
call
it
a
day.