►
From YouTube: Kubernetes SIG API Machinery 20180926
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
A
I'll,
take
that
doesn't
nope
now
I
have
to
remember
what
I
was
gonna
say
about
this
I.
Oh
yeah
I
thought
it
might
be
worth
giving
a
heads
up.
I
noticed,
while
somebody
was
fixing
something
unrelated
that
the
JSON
patch
library
that
we
use
permits
negative
indices.
You
can
like
count
back
from
the
end
of
an
array.
This
is
used
when
we
apply
a
patch
and
I
also
happen
to
know
from
having
read
the
spec
for
unrelated
reasons
that
the
defect
does
not
allow
this.
A
C
B
So
I
think
that
for
turning
off
a
feature
like
this,
that
we
don't
want
to
support,
I
definitely
I
can
definitely
think
of
this
and
other
things
that
we
would
want
to
do.
But
when
we
start
going
about
that,
we
normally
go
through
a
process
where
you
indicate
that
it's
deprecated
and
can
opt
into
the
new
behavior
and
then
the
next
release.
We
have
it
so
that
the
new
behavior
is
on
by
default
and
you
can
opt
out
if
you
wish
and
then
the
next
release.
After
that,
you
just
no
longer
get
a
choice.
A
B
A
B
I
guess
what
I
would
I
would
like
to
see
the
transition
happen
where
it's,
where
we
don't
accept
the
thing,
but
you
can
turn
on
that
feature
again
if
you
need
it
for
one
release
and
then
the
release
after
that
is
just
gone.
I
think
that
would
be
behavior
I
could
get
behind
in
terms
of
whether
it
is
supported
by
the
spec
in
a
bug.
B
If
it's
been
there
long
enough
it,
whether
we
think
it's
a
bug
or
not,
if
we
don't
have
a
security
reason
for
just
breaking
it,
I
would
rather
go
through
like
a
one
release.
You've
been
warned:
here's
a
flag
turn
it
back
on.
If
you
really
need
it
and
the
release
after
you
just
have
to
deal
with
new
Baker.
A
I
mean
we
can
also
document
that
if
a
large
number
of
people
find
the
need
to
use
this,
then
then
we
can
reevaluate
whether
this,
but
we
should,
if
we
have
like
stuff
like
this,
we
shouldn't
also
be
claimed.
We
shouldn't
also
claim
to
follow
this
back
right,
like
we
should
either
claim
too
small
on
the
spec.
B
A
A
D
E
E
B
Mean
we
know
there's
a
patch
command
on
the
CLI.
We
know
that
people
use
it,
I
can
positional.
Positional
checks
are
problematic
in
general,
so
I
wouldn't
expect
much
of
an
issue.
It's
just
using
this
as
a
test
case
for
like
a
general.
What
would
we
do
and
we
tighten
our
API
I?
Think
we
I
think
we
can
tighten
our
API,
but
I
think
we
need
to
be
somewhat
measured
about
how
we
do
it.
Yeah.
B
A
A
A
Some
a
blocker
that
prevented
us
from
going
with
this
design
in
the
first
place
is
that
there's
no
general
way
to
do
a
conversion
on
such
a
field
set
like
you
can't
run
it
through.
Our
existing
conversions
functions
and
figure
out
what
the
corresponding
field
set
is
in
a
different
version,
but
it
turns
out
it's
not
so
simple
to
do
that
with
partially
specified
objects,
as
in
the
original
document
and
the
technique
that
I
came
up
with
the
that
will
work
for
the
original
design,
like
variations
on.
That
will
also
work
for
this
design.
A
A
So
that's
that's
good,
and
we
also
think
that
we
can
limit.
The
cardinality
like,
like
the
previous
design,
had
had
a
problem
where,
if
a
bunch
of
people
applied
things
like
you
could
just
grow
the
list
bigger
and
bigger.
So
we
think
we
can
keep
the
list
down
to
like
a
like
a
constant
factor
of
the
object
size.
The.
A
Think
that,
like
literally
every
test
case
in
my
original
document,
will
give
the
same
behavior
that
that's
that's
the
plan,
so
it's
like
you
shouldn't
notice
that
this
this
should
be
an
implementation
detail.
It's
like
you're
rather
sick.
It
will
have
it.
It
will
have
a
profound
impact
on
the
API,
so
will
will
have
to
specify
that
and
get
a
new
API
in
progress.
B
You
add
a
link
in
this
doc,
pointing
to
yeah
like
point
to
your
new
design
and
particularly
the
API
changes
to
actually
track
ownership.
A
A
A
B
A
Which
leads
me
to
the
other:
half
of
this
is
I
actually
started.
A
repository
to
hold
apply
code
needs
to
be
renamed
at
this
point,
because
we've
shifted
direction
just
a
little
bit.
But
if
you
look
in
that
repository
I
added
a
apparently
the
template,
the
repository
template
has
this:
releasing
got
in
the
file
that
you're
supposed
to
fill
out
how
the
release
process
works.
So
I
wrote
up
a
sketch
of
a
new
release
process
they're.
A
Basically,
my
intention
is
to
make
this
this
library
and
everything
underneath
a
like
100
percent
go
compatible.
So
it's
like
no
bash
scripts
in
it.
I'd
like
to
follow
the
new
go
module
semantics.
So
that
means
once
we
get
to
a
stable
release,
it
will
be
if
we
change
anything
at
all
about
the
interface
we'll
be
adding
a
B
to
prom,
because
I
don't
know,
I've
done,
I've
done
hacking
on
go
projects
just
out
in
the
wild
and
hacking
on
kubernetes
related
go
projects,
and
it
is
a
huge
pain
to
hack
on
kubernetes
related.
B
A
E
A
B
I
think
I'd
really
like
to
see
the
tool
that
automatically
let
you
know
when
you
were
breaking
it
more.
We
expanded
it
I
like
I'm,
not
saying
it
has
to
be
here
for
you
to
do
this
here,
but
before
we
considered
expanding
it
to
other
repos,
I
really
I
mean
like
I'm,
not
gonna,
do
it
on
purpose,
but
chances
are
yeah.
A
A
B
A
Clayton
says
that
that
Tim
would
of
course
make
a
make
file
to
generate
a
make
file
form,
which
is
true.
F
A
A
D
B
B
G
So
two
issues
with
that:
there's
a
button:
a
missionary
had
a
workaround
and
working
on
a
proper
extruder.
Okay,
and
there
is
that
design
issue
on.
Regarding
the
defaulting
we
discussed
it
extensively
with
hallway
and
Eric
and
I
think
we
have
a
solution
now
going
to
send
the
PR
and
also
share
a
dog
that
compared
all
of
the
solutions,
the
the
PR
you're,
sending
us
to
do,
PR
going
to
be
to
the
ground
with
the
solution
that
we
are
proposing,
but
also
have
a
link
to
the
doctor,
control
solutions.
Okay,.
B
Great
I
just
want
to
make
sure
we
weren't
holding
something
up
there
for
that
and
then
separately
for
Chow
I've
seen
some
action
from
him
in
the
storage,
migration
tool
and
I
wanted
to
make
sure.
We
were
on
the
same
page
with
why
the
why
using
a
typed,
client
or
using
protobuf
was
a
dangerous
thing
to
do
for
the
storage
migration
tool.
B
B
F
B
Resource
quota
yeah,
actually
there
is
a
I
guess:
I
can
back
up
and
give
a
brief
history
back
in
111,
a
poll
came
in
very
late
about
trying
to
add
resource
quota
for
custom
resources
and
I
had
to
hold
that
PR,
because
it
was
duplicating
some
very
sticky
code
from
the
garbage
collector
and-
and
we
really
don't
want
a
copy
of
that
code.
B
So
there
is
a
community
member
I,
don't
think
regularly
attends
this
call,
but
a
guy
named
polynomial
is
actually
building
out
a
dynamic
Informer
and
a
dynamic
lister
based
on
the
dynamic
client
that
we
now
have
once
we
have
those
pieces,
I'm
expecting
to
be
able
to
clean
up
how
the
garbage
collector
manages
dynamic
resources
and
then
clean
up
how
and
then
add
a
to
resource
quota
as
well.
Right
that
way,
we'll
be
sharing
the
same.
Informer
will
have
the
same
logic
for
handling
when
these
things
get
added
and
removed.
F
F
A
Any
other
topics
I
thought
of
one
other
thing
that
I
could
say,
which
is
everybody's
favorite,
the
SSH
tunnels
and
update
they've,
been
deprecated
for
a
year
they're
going
away.
This
quarter
I
will
probably
be
circulating
a
design
for
a
or
something
that's
functionally
equivalent,
but
cloud
neutral,
just
a
just
a
heads
up.
This
might
be
coming.