►
From YouTube: SIG Network Gateway API Bi-Weekly Meeting for 20220815
Description
SIG Network Gateway API Bi-Weekly Meeting for 20220815
A
All
right
welcome
to
the
gateway
api
meeting
for
august
15.
2022
we've
got
lots
to
get
through
today
before
we
get
started.
I
want
to
remind
everyone.
This
is
a
kubernetes
community
meeting.
That
means
a
few
things,
of
course
follow
the
code
of
conduct,
but
also
this
agenda
is
wide
open.
If
there's
something
you
want
to
discuss,
definitely
feel
free
to
add
it
to
the
agenda
at
any
point
or
add
it
for
a
future
meeting
as
well.
A
This
doc
is
something
that
I
hope
anyone
can
edit
and,
if
not
let
us
know,
but
the
goal
is
for
anybody
to
be
able
to
take
notes.
We
always
appreciate
note
takers
or
to
add
new
things
to
the
agenda
as
they
come
up.
I
also
had
a
note
here.
I
think
there
may
be
one
or
two
new
people
that
haven't
had
a
chance
to
introduce
themselves
yet
so,
if
you're
new
to
this
call-
and
I
want
to
introduce
yourself-
no
pressure-
we'd
love
to
hear
what
brings
you
here.
B
Awesome
yeah,
so
it
looks
like
the
initial
doc
that
we
wrote
that's
about
the
sort
of
the
what
and
why
seems
like
this
reasonable
agreement
on
there.
I
don't
haven't
seen
many.
I
haven't
seen
many
yeah.
You
need
to
change
this
kind
of
topics,
it's
more
like
hey.
We
need
more
detail
here,
which
I
agree.
Obviously
a
couple
of
other
comments,
but.
B
Yeah,
but
I
don't
I
think
in
general,
it
seems
like
this
pretty
good
consensus
there
so
I'll
open
a
pr
to
like
with
the
initial
gap
the
same
as
keith's
already
done
for
the
gamer.
Once
that's
worked
by
the
way.
While
I
was
away.
D
Thank
you
just
a
quick
question
about
egress,
something
that
came
up
in
one
of
my
team
meetings
today
and
I
took
a
look
around
just
to
see
if
it
had
ever
been
discussed
or
there
were
any
plans
coming
up
for
egress.
D
But
it
looks
like
everything's
stale
on
that.
So
I
wanted
to
see-
and
I
see
that
keith
has
added
a
note-
that
it's
a
deferred
goal
for
gamma.
So
just
checking
in
to
see
what
the
community
goals
are
for
egress.
A
I
think
this
anyone
can
feel
free
to
correct
me
on
this,
but
my
take
is
that
it's
something
that
there's
wide
interest
in,
but
no
one
has
volunteered
to
lead
that
exploration,
so
it's
just
kind
of
as,
as
you
mentioned,
it's
just
kind
of
gotten
stale
pasta.
I
see
you've
got
your
hand.
A
E
Subject
and
we
discussed
those
are
discussing
gamma
as
well.
I
think
the
current
spec
can
be
used
perfectly,
as
is
for
egress.
I
mean
ingress.
E
The
same
thing,
if
it's
just
a
different
direction,
so
so,
if
workload
is
a
mesh,
have
a
way
to
configure
the
gateway
as
an
egress
gateway,
you
have
the
same
routing
rules,
the
same
the
same
principles
applied
there
as
well,
it's
nothing
special,
you
have
hosts
and
you
can
redirect
them
to
a
different
video.
You
can
apply
policies
and
you
can
do
things
differently,
so
I
don't
think
we
are
very
far
from
from
from
from
api
perspective.
It's
just
that.
D
You
know
whatever
you
might
call
a
back
end
and
an
egress.
It
would
be
an
external
you
know
outside
of
the
cluster,
even
endpoint
for
egress,
which
makes
things
a
little
different.
E
You
know
creating
a
brand
new
specification
for
egress,
especially
since
so
confusing
an
issue.
At
least
you
know
we
have
egress
directly
from
pods
and
policies
applied
there.
We
have
the
egress
gateway
itself,
which
is
another
thing,
and
and
then
we
have
the
two
types
of
ingress.
I
mean
terminating,
not
terminating
and
we
have
integration
with.
You
know
existing
egress
infrastructures,
which
you
know
sweden
or
the
enterprise
egress
gateways.
So
it's
a
pretty
large
subject.
If
you
want
to
to
go
especially.
D
F
Yeah,
so
I've
got
a
lot
of
thought
here,
not
on
ego
specifically,
but
around
handling
it.
There
are
some
good
comments
in
the
chat
around.
I
think
cuj
stays
for
customer
user
journey
and
collecting
those
requirements.
F
I
think
that's
going
to
be
that's
going
to
be
key
here,
because
even
in
gamma,
when
you
know
egress
is
often
a
service
mesh
use
case
and
when
thinking
went
from
a
gamma
gamer
perspective,
you
know
it's
not
always
clear.
F
F
There
is
that
the
security
problem
of
how
some
folks
implement
egress
gateway
for
extraordinary
services
and
there's
just
there's
a
lot
of
concerns
there
of
folks
trying
to
go
and
use
that.
So
I
john
probably
in
the
chat,
john's
comment,
there
probably
sums
up
how
I
feel
about
this.
Is
it's
got
a
lot
of
complicated
semantics
and
security
implications?
F
I
think
to
your
point
costing
me:
yes,
the
technology
is
is
not
necessarily
too
different,
especially
when
you
start
thinking
about
like
the
service
message
cases
forget
your
api.
I
think
there's
some
kind
of
overlap
in
the
technical
aspects
there,
but
from
the
security
semantic
point
of
view,
I
think,
there's
absolutely
a
lot
of
discovery
that
needs
to
be
done
on
what
these
actual
concrete
requirements
are
and
how
we
can
do
it
in
a
in
a
secure
way.
B
B
Mike
said
in
the
chat
that
you
know
things
that
maybe
maybe
pulling
a
get
number
and
doing
this
hour
and
y
discussion
about
it
and
then
saying
and
then
saying
we're
explicitly
going
to
defer
this
to
later,
it
might
be
a
bit
more
clear
yeah.
If
anyone
wants
to
pick
that
work
up,
then
yeah
there's
a
gap
waiting
to
be
written
to
do
the
initial
sort
of
hey.
B
Why
you
know
what
would
what
would
we
do
this?
So
what
what
are
we
doing,
and
why
are
we
doing
it,
not
the
house
all
right
and
so
to
say,
you're
like
hey,
let's,
let's
agree
that
when
we
do
egress,
we
will
be
doing
something
what
is
egress
and
why
would
we
do
that
as
part
of
the
gateway
api
and
the
service
primitives,
and
then
let's
leave
it
and
then,
let's
explicitly
say,
this
is
now
deferred
until
you
know,
until
we
come
back
to
it
later
like
it's,
not
it's
not
that
we're.
B
B
We're
not
ready
to
do
it
yet
that
sounds
like
about
the
right
thing
to
say
to
me:
does
anyone
else
disagree
there
or.
B
B
Yeah
yeah,
I
think,
and
I
think
we've
been
cleaning
up.
Candice
we've
been
cleaning
up
issues,
so
I'm
sort
of
trying
to
you
know
clean
up
the
backlog
a
bit
and
make
sure
that
the
stuff
that's
there
is
actionable,
and
that's
why
some
of
this
stuff
has
been
sort
of
styled
is
because
yeah
no
one's
touched
it
for
a
while.
That's
all.
A
Great
thanks
for
bringing
that
up
and
yeah
yet
another
great
opportunity.
If
someone
is
looking
to
contribute,
we
can
use
all
the
help
we
can
get
move
on
to
the
next
one
here,
which
is
a
coup
con
item
I
added
I
wanted
to
highlight.
I
don't
think
this
has
been
sent
out
yet,
but
the
contributor
summit
will
be
on
monday.
A
I
it
will
it'll,
probably
be
something
that
goes
to
the
kubernetes
dev
mailing
list
soon,
but
all
org
members,
including
kubernetes
board
members,
can
join
it
is
I
want
us:
it's
traditionally
been
free,
even
if
you're
not
going
to
kucon
itself,
I'm
not
entirely
sure
what
how
it
works
this
year,
but
there's
generally
no
extra
charge,
at
least
it's
my
favorite
part
of
kubecon
I'd
love
to
see
as
many
of
you
there
as
possible.
A
So
just
keep
that
you
know
if
you're
thinking
about
going
to
kubecon-
and
you
can
be
there
on
monday-
it'd-
be
great
to
see
lots
of
gateway.
Api
people
there
if
you're,
not
a
org
member,
yet
there's
a
documented
process,
generally
need
a
few
pr's
and
you
need
support
from
org
members
from
two
different
companies.
That's
high
level
so
we're.
A
I
know,
we've
gotten
a
number
of
new
members
through
gateway,
api
and
happy
to
get
some
more
going
as
well
yeah
and
then
next
up,
I've
been
chatting
with
bob
killen
and
jason
about
a
gateway,
api
event
or
some.
I
don't
know
what
this
looks
like
yet,
but
there's
time
available
and
space
available
for
us
to
do
something
probably
on
tuesday.
A
So
especially
if
you're
already
here
for
the
contribute
summit
would
be
cool
if
you
just
left
tuesday
open
if
you
can
yeah
and
if
you
have
any
time
preference
or
anything
like
that
again,
just
ping
in
slack
or
somewhere-
and
this
is
still,
I
think
we
should
know
more
information
next
week
from
what
they've
been
saying.
But
yeah
and
the
main
conference
is
wednesday
to
friday,
and
I
think
we
have
gateway
api
talks
every
day.
I
think
probably
a
couple
every
day.
A
So
thanks
to
everyone
that
submitted
something
like
yeah
any
any
thoughts
or
comments
on
kubecon
before
I
move
on.
B
I
think
tuesday
also
there's
a
number
of
sort
of
co
events
that
people
might
need
to
go
to.
I
plus
one
rob's
thing
about
going
to
contribute
to
summer.
If
you
can
manage
it
personally,
I
will
always
prioritize
the
contributor
summit
over
any
other
thing.
That's
on
that
monday
yeah,
because
I
find
it's
one
of
the
most
used.
It's
the
two
most
useful
days
of
the
year
for
me,
but
yeah
and.
B
A
B
You've
got
a
lot
of
people
who
know
a
lot
about
our
cube
in
the
room,
so
yeah
and
but
yeah,
I
would
say,
yeah
make
sure
you
speak
up.
If,
if
you
want
to
come
to,
I
get
api
event
of
some
kind,
then
you
can't
make
it
because
on
the
tuesday
I
think
maybe
is
service.
A
Yeah,
it's
it's
hard
to
find
something
that
yeah
yeah.
So
I
think
there's
gonna
be
a
lot
of
conflicts
with
everything
we
choose,
so
I
it
would
be
good
to
if
there
are
people
who
are
interested
like
like
nick
said
just
let
us
know
when
you
can't
make
it
or
days
that
are
already
completely
spoken
for
and
we'll
try
and
find
some
like
the
works
around
that.
A
Cool
all
right
next
up
conformance
as
I
I
think
I
discussed
this
last
week
or
weeks
before
I
don't
know,
but
we've
had
this
general
idea
for
a
long
time
that
if
you
have
conformance
tests,
you
need
somewhere
to
describe
that
your
implementation
is
actually
conformant
or
not.
A
I
spent
some
time
last
week,
actually
just
on
friday
talking
with
the
hippie
hacker
one
of
the
cncf
people,
who's
been
really
focused
on
conformance
and
he
had
some
good
tips
and
discussion
and
he
wants
to
help
which
all
great
I
he's
been
more
involved
in
upstream
kubernetes,
conformance,
obviously,
but
had
some
tips
for
what
we
can
do
and
even
some
thoughts
about
how
this
could
eventually
move
from
gateway,
the
other
project
to
cncf
as
a
broader
thing,
but
for
now
starting
here
just
looking
at
the
upstream
pattern,
it
seems
simple
enough
and
largely
doable
for
us.
A
Each
upstream
conformance
report
has
a
product
yaml
that
just
describes
hey.
This
is
I
just
use
cops
as
an
example,
but
just
describes
what
the
project
is.
The
version
the
test
was
run
against
and
that's
about
it,
a
readme
that
describes
every
step.
They
took
to
run
the
conformance
tests
and
then
the
test
logs,
showing
their
test
passing.
A
I
do
and
that's
up
to
each
implementation
to
provide
for
each
version
that
has
been
released.
That
seems
remarkably,
simple
and
something
that
will
get
us
a
long
ways
there.
We
have
as
we're
well
aware
some
unique
bits
that
aren't
really
part
of
conform
upstream
conformance
and
that's
our
conformance
levels
are
core
extended,
et
cetera,
and
so
I
called
out
some
things
that
we
may
want
to
do
on
top
of
what
upstream
does,
but
I
think
really
just
the
top
level
is.
A
If
someone
wants
to
take
this
and
kind
of
document,
how
we
could
accept
conformance
reports
that
that's
kind
of
I
think
step
one.
Just
here's
a
place
where
we
can
accept
conformance
report
reports
and
any
implementation
is
welcome
to
submit
them
here.
I
think,
eventually,
we
want
to
get
to
a
goal
where
our
implementations
page
itself
highlights.
A
This
implementation
was
100
conformant
at
this
version
of
api,
something
like
that.
But
to
even
get
to
that
point
we
need
to
accept
conformance
reports
somewhere.
So
I
added
this
issue
open
to.
Is
anyone
interested?
I
know,
there's
been
a
lot
of
interest
in
working
on
performance.
Is
anyone
interested
in
kind
of
taking
this
forward.
A
Cool
well,
it
sounds
like
nick
is,
has
it
for
now,
but
definitely
reach
out
we're
always
open
to
new
contributors
here
right
and
just
on
that
note,
I
just
as
I
was
going
through
trying
to
find
issues
for
triage.
A
Most
of
them
are
currently
unassigned.
So,
if
you're
looking
for
an
area
to
get
involved
in
the
api,
this
is
this
is
a
great
one.
You
know,
I
think
one
of
our
big
areas
of
investment
in
the
upcoming
releases
has
to
be
conformance.
So
anything
you
can
contribute.
There
are
massive
ones,
and
there
are
tiny
ones
here
anywhere.
You
can
contribute
is
very
welcome.
B
I
wanted
to
call
out,
I'm
not
sure
how
you
say
name,
but
maybe
for
doing
some
great
work,
just
finding
little
things
and
logging
issues
about
them.
Like
that's,
really
awesome,
yeah,
that's
exactly
the
sort
of
stuff
we
need
to
just
to
flesh
out
those
conformance
tests.
If
you
find
stuff
like
that,
please
just
log
an
issue
like
issues
are
free
ish.
You
know
and
there's
not
there's.
No.
I
would
much
rather
have
five
issues
describing
a
thing
than
none.
A
Yeah
completely
agree
he's
been
creating
some
great
issues
here
so
yeah
more
of
that,
and
you
know
as
you
as
you
have
an
opportunity.
If
you
see
one
of
these,
they
say.
Oh,
I
can
help
with
that.
That's
great,
too
just
comment
on
the
bug
that
you
want
to
take
that
one
on
cool
all
right.
I
wanted
to
spend
a
bit
of
time
talking
about
gaps.
I
know
we
already
mentioned
nick's
gap
earlier
on,
but
we
have
a
few
gaps
that
are
in
the
proposed
state.
F
Yeah
yeah,
I
yeah,
I
forgot
to
assign
myself
to
them
cool.
A
B
Yeah
yeah,
if,
if
again,
if
no
one
else
wants
it
I'll
take
it,
because
I
think
between
you
and
I
rob-
I
probably
did
the
most
damage
here
with,
of
course,
the
attachment.
I
think
we
definitely
need
to
swing
back
around
and
talk
about
it
again.
I
have
noticed
a
lot
of
people
talking
about
it
and
wanting
to
use
policy
attachment
for
service
much
stuff
as
well.
B
So
I
think,
there's
probably
some
space
for
us
to
talk
again
about
for
us
to
go
through
like
why
we
build
it
the
way
we
did
what
sort
of
stuff
we
were
thinking
about
when
we
built
it
and
like
if
we
should
consider
extending
because
I
yeah,
I
know
I've
sort
of
been
surprised
a
couple
times
and
people
like
oh
just
use
policy
attachment
for
that.
I'm,
like
I
hadn't
thought
about
that
so
yeah.
I
think.
G
Any
like
separation
of
concerns
between
that
and
the
proposed
back-end
properties
gap
between
progress
right
now.
B
Yeah,
I
think
the
discussion
on
back
end
properties
should
help
us
figure
out
like
what's
the
what's
the
dividing
line
between
using
a
policy
and
doing
the
back-end
properties.
I
think
I
have
an
idea.
I
don't
have.
It
very
clearly
explained
yet
so
I
won't
say
it
now,
but
yeah.
So
I
think
that's
between.
B
We
should
yeah,
so
it
probably
makes
sense
for
me
to
be
honest.
A
C
A
I
have
a
pr
that's
working
on
on
conformance
tests
anyway,
but
I'll
chat
with
you
afterwards.
I
can.
I
can
try
and
help
out,
but
working
on
on
better
dots
here,
gotcha
sounds.
F
Yeah,
just
if
anybody
has
had
a
chance
to
come
to
some
of
the
gamma
meetings
or
if
they
have
other
thoughts
in
general
about
what
gateway
api
should
be
trying
to
accomplish
when
it
comes
to
service
mesh.
This
is
this
is
the
pr
to
take
a
look
at.
It
really
focuses
on
high
level
what
and
why
around
service
mesh,
and
this
kind
of
is
going
to
be
the
springboard
for
the
other
two
gaps
around
mesh
service,
our
mesh
representation
and
service
binding.
F
I
was
actually
working
on
those
two
gaps
and
just
found
a
a
lack
of
concrete
description
and
guidance,
and
that's
kind
of
in
with
talking
to
john.
This
gap
came
out
of
it.
So
if
you
have
thoughts
are
interested
like
this
is
a
really
good,
like
checkpoint
gap.
If
you're
wanting
to
track
progress
for
what
we're
doing
over
on
the
gaming
side,.
A
Great
thanks,
it's
cool
to
see
this
moving
along.
A
I
I
was
impressed
by
this
html
bit
here
too.
F
I
spent
some
time
as
a
front-end
web
designer
for
a
little
bit
and
it's
it's
messy,
but
it
gets
the
job.
B
Keith
I'll
have
a
look
at
that
today
because
I
can't
make
the
the
gamma
meeting
tomorrow,
because
it's
two
o'clock
in
the
morning
for
me
so
yeah.
I
love
you
all,
but
yeah,
not
that
much
so
yeah
I'll
see.
If
I
can
put
some
comments
on
that
today,
or
at
least.
F
That's
good
yeah
sounds
good.
Oh
I
forgot
to
mention.
I
had
a.
I
was
kind
of
trying
to
to
time
box
the
other
two
submissions
by
like
pre-august
22nd.
That's
probably
not
gonna
hold
quite
as
well
now
with
this
with
this
new
gap,
but
I
I
do
want
to
try
to
get
that
gap
merged-ish
around
that
same
time,
so
just
kind
of
shifting
the
the
timeline
so
yeah.
That
sounds
good.
A
Great
there
was
a
new
gap
that
came
up
just
in
the
past
a
few
days.
I
think
sanskar,
I
think,
is
his
name.
I
proposed
this
and
it
makes
a
lot
of
sense
to
me.
This
is
one
of
the
few
guests
that
just
feels
like
a
very
obvious,
straightforward
edition
of
the
api.
We
have
a
request-
header
modifier
in
the
api
already,
but
we
have
no
response,
header
modifier,
I,
which
could
look
near
identical,
that's
the
proposal
anyway,
I
don't
think
sanskar
is
on
this
call
correct
me.
A
Yeah
doesn't
look
like
he's
on
this
call,
but
yeah
reviews
welcome
on
this.
I
think
this
is
pretty
straightforward.
I
did
ask
for
a
non-envoy-based
example
of
this
functionality,
but
otherwise
I
it
feels
straightforward
to
me.
I
think
shane
and
I
are
the
only
ones
who've
looked
at
it
so
far,
so
if
you
have
any
additional
feedback,
definitely
get
it
in,
but
otherwise
I
think
we
can
probably
move
it
straight
to
implementable.
B
No
I'll
have
a
look
at
it
today,
but
yeah
like
it
looks.
It
looks
pretty
straightforward
to
me.
A
Cool
thanks
and
the
the
massive
one
round
delegation
this
one
I
I
painted
jeff
and
I
think
he
he's
not
on
this
meeting
so
I'll
I'll
follow
up
with
him
offline,
but
I
think
we,
we
know
the
next
steps
here,
we've
kind
of
started
to
split
things
out
into
different
categories
and
different
pr's.
A
A
I
I
think
that
it's
probably
not
worth
adding
more
dot
more
comments
to
this
specific
pr,
and
instead
we
can
wait
till
we
get
a
provisional
gap
in
and
we
can
take
on
individual
questions
as
they
come
from
there
yeah,
okay,
the
any
any
other
guess
that
we
should
cover
before
I
move
on
to
next
section.
A
Cool
I'll
keep
on
going
reviewing
conformance
test
vrs.
I
know
I've
reviewed
a
bunch
of
conformance
test
pr's,
but
I
know
I'm
not
the
only
person,
I'm
curious,
what
other
people
are
doing
for
our
workflow
here,
because
all
I've
been
doing
is
just
you
know,
checking
out
the
pr
and
running
conformance
tests
against
implementation
of
choice
and
it's
random
depending
on
the
day,
but
that
that
has
been
my
entire
workflow
and
it
feels
very
random
and
and
manual
and
not
not
ideal,
and
if
every
reviewer
is
doing
the
same
thing.
A
That's
that's
not
great,
so
my
question
is:
does
it
make
sense
to
have
some
kind
of
non-blocking
informational,
only
pre-submit,
that
runs
against,
say
one
to
three
implementations
of
the
api?
I
don't
know
I
know
I
know
this
is
getting.
We
don't
have
a
standard
implementation
of
gateway
and
we
never
want
to
give
that
appearance
of
this
is
the
standard
implementation
of
the
api,
but
if
that's,
what
we're
effectively
doing
when
we're
reviewing
conformance
tests
automating
some
portion
of
that
could
maybe
be
helpful.
A
Any
any
thoughts
on
that.
One.
C
A
This
could
never
be
a
anything,
but
if
it
were
at
least
informational
in
a
sense
of
you
added
this
conformance
test
and
it
broke
these
two
implementations
that
were
previously
passing-
and
this
is
why
or
the
test
completely
failed
to
run
like
we
don't
even
have
anything
that
catches
the
test
completely
failed
to
run
other
than
manual
testing
right
now
I
don't
have
any
problem.
C
C
A
I
feel
like
we
need
to
be
respectful
of
test
infra
costs
and
you
know
like
there's
a
non-zero
cost
for
every
pre-submit.
We
run
so
you
know
if
we,
I
think
we
have
14
implementations
today,
and
I
expect
that
will
only
go
up
so
if
we
had
14
different
pre-submits.
I
think
that
would
be
unfortunate,
but
I
I
do
want
to
avoid
favor
favoring
favoritism
of
any
one
or
three
or
whatever
implementations
out
there.
So.
A
G
So
I
don't
know
if
this
is
going
to
be
the
case
for
what
percentage
of
conformance
tests,
because
sometimes
we're
testing
stuff
that's
already
implemented
widely,
but
oftentimes
like
what
you're
asking
about
the
performance
for
grpc
route.
I
think
it's
unlikely
that
there's
maybe
more
than
one
implementation
of
that
out
there
currently.
A
I
guess
what
I
was
meaning
for
this
is
that
there
are
a
number
of
conformance
test
changes
and
tweaks
that
we
see
come
in
that
are
really
just
bug
fixes
right.
You
know,
like
I,
I
completely
agree
net
new
conformance
tests
are
likely
going
to
fail
on
the
majority
of
implementations,
but
there
is
still
something
to
be
said
for
not
having
to
you
know
like
right
now,
anytime,
I
see
a
conformance
test
pr.
E
Yeah,
I
I
think
it
will
f,
should
never
be
the
case
where
a
conformance
test
doesn't
pass,
you
know
or
implementations,
or
you
know
at
least
a
significant
number.
You
know
these
four
core
features
for
optional
features
any
features.
Many
any
implementations
that
claims
to
to
support
that
particular
feature
obviously
should
pass.
E
I
mean
it
said:
you
first
implement
the
behavior
and
then
you
test
it
with
you
know,
servlets
and-
and
you
know,
with
java
in
general,
it's
a
compatibility
test,
suite
reflected,
represent
agreement
and
implementation,
not
you
know,
dreams
or
what
we
hope
people
implement
or
as
a
way
to
force
people
to
implement
something
in
a
particular
way.
A
Yeah
I
mean
it's
fairly
straightforward
to
run
conformance
tests,
at
least
for
the
in
cluster
implementations
that
have
a
you
know.
You
can
run
them
in
kind
for
that,
for
that
matter
I
at
least
I
would
think
it
should
be
straightforward,
but
I'm
interested
in
perspective
of
others.
A
E
B
B
It
is
basically
just
a
sanity
check
that
we'll
check
you
know,
check
those
small,
pr's
and
stuff
like
that.
I
think,
there's
a
whole
bunch
of
problems
that
we
have
there
about
yeah.
When
you
add
a
net
new
one
like
that
recently
will
obviously
fail,
because,
because
probably
implementations
won't
have
a
new
spot
for
it,
yet
right
like
and
so,
and
even
if
you're
not
doing
that.
Like
then,
you've
got
the
problem
of
like
you,
I
know
for
contour
like
we
were.
B
We
were
doing
the
performance
at
like
a
tagged
gateway
api
version,
so
you
know
how
do
you
then?
You
need
to
build
out
like
each
implementation
will
need
to
build
out,
make
sure
they
build
out
like
a
runner
that
can
run
the
tagline
or
remain
and
you'll
need
to
make
sure
you're
running
main
rather
than
the
yeah
or
the
pr
branch
or.
B
Fiddly
detailed
stuff
there
that
would
need
to
happen.
Is
it
a
worthy
goal?
Yeah
absolutely
do
I
think
it's
like
a
ton
of
work,
yes
very
much
so
yeah,
and
I
think
we
need
to
be
careful.
We're
not
trying
to
boil
the
ocean
here
and
you
know,
sort
of
build
an
arbitrary
framework,
for
you
know,
tested
service
which
would
be
cool
but
like
it
feels
like
that's
probably
larger
than
what
we
need.
B
I
think
if
this
feels
to
me
like
yeah
this,
this
would
be
really
nice
to
have,
and
if
someone
wants
to
spend
some
time,
like
figuring
out
how
to
make
it
work
like
more
power
to
you,
maybe
we
need
an
issue
to
sort
of
park
the
discussion
and
that
we
freeze
raw
but
yeah.
I
think
for
now
we're
stuck
unless
someone
can
put
a
reasonable
amount
of
work
into
making
this
work.
B
Yeah-
and
I
mean
maybe
maybe
maybe
a
better
way
to
do
this-
would
be
for
the
implementations
for
us
to
have
a
maybe
it
would
be
a
better
way
to
do.
This
would
be
for
us
to
have
like
a
place
where
the
implementations
can
put
here's.
How
you
run
our
performance
tests
in
this
repo.
C
We're
not
there
yet
we
should
at
least
make
it
so
that
all
the
implementations
that
are
listed,
it's
just
very
trivial,
for
each
implementation.
To
tell
the
someone
authoring
a
conformance
test
to
be
like
here's
a
cut
and
paste
you
do
this,
you
can
run
the
conformance
test
against
it.
So
you
know
just
if
you
have
a
change,
then
just
here's
how
to
check
that
it
worked.
B
Yeah
so
I
mean,
I
think
I
think
that
in
my
mind
I
100
agree,
but
I
think,
like
a
set
of
runbooks,
maybe
like
that
that
you
know
an
implementation
puts
in
puts
into
this
repo
that
says:
hey
here's,
how
you
test
with
our
thing
and
then
part
part
of
adding
a
new
conformance
test
is
that
we
say
hey
we
want
to
see.
Part
of
this
is
that
we
want
to
see
the
you
know
the
new
conformance
tests,
or
we
want
to
see
a
couple
of
outputs
from
from
the
conformance
testing
for
various
letters.
B
Assuming
that
you
run
the
run
book
and
we
want
to
see
them
passing
like
and
just
you
know
just
do
this
manually
by
people
to
start
with.
We
can
just
make
it
easier
so
that
you
don't
have
to
figure
out
figure
out
every
time
how
to
get
your
infrastructure
set
up
and
figure
out
how
to
how
to
run
each
each
implementations,
things
there's
just
a
standard
like
template
that
we
say:
hey.
B
Here's
a
template,
direct
here's,
a
directory
with
all
the
templates
for
your
implementation,
put
it
in
here
and
then
we'll
expect
people
to
tests,
expect
people,
and
then
people
can
test
against
your
implementation.
Something
like
that.
Maybe.
A
Yeah
I
want
to
be-
I
I
think
it's
I
think
I
I
agree
with
what
you're
saying
of
making
it
very
clear
how
anyone
can
run
conformance
tests,
but
I
also
want
to
be
clear
that
it
should
not
fall
on.
You
know
say
me
as
a
reviewer
to
run
conformance
tests
for
every
every
employee,
and
I
don't.
B
B
B
A
So
one
of
my
concerns
there
is
that
we
could
run
into
slower
velocity
just
in
terms
of
conformance
prs.
You
know
we
have
the
1415
conformance
issues
that
are
open
right
now
and
if
everyone
needs
to
figure
out,
you
know
not,
everyone
has
access
to
even
a
kubernetes
cluster.
They
can
spin
up
and
and
run
these
tests
on,
for
example,.
C
Although
that
begs
the
question,
then
that
they're
submitting
a
pr
that
they
haven't
tested
so
like
yeah.
Well,
what
do
you?
I
don't
know
about
that?.
B
I
mean-
maybe
maybe
just
saying
okay,
you
for
this
to
be
for
conformance
powers
to
be
reviewable.
You
need
to
have
run
it
against
one
implementation
using
the
you
know,
and
you
need
to
put
some
logs
in
here,
sort
of
showing
that
showing
the
performance
test
run
that
you've
done
and
you're
in
the
absence
of
having
standard
steps,
putting
your
steps
in
the
in
a
couple
of
spr
that
way
like
you
need
to
have
done
it
again,
at
least
against
at
least
one
right
like
so.
B
Yeah,
I
I
think
at
least
one
is
like
a
reviewable
thing,
but
like
for
for
everything
and
then
for
two
for
adding
new
performance
tests.
It
should
be
more
than
one
right
like
the
yeah,
so
I
think
for
for,
like
bug
fixes
like
it
should
be
fine
to
just
be
like
hey.
This
works
on
one
of
them.
It
should
work
on
all
of
them.
I
don't
know
maybe,
but,
but
I
think
at
least
for
the
it
feels
to
me
like
that's
the
way
to
do
it
to
say
your
conformance
test.
B
B
Like
the
steps
to
do
that
and
then
more
that
we
should
be
able
to
make
this
easier
by
giving
people
the
steps
to
do
that
so
that,
then
we
can
say:
hey
pr,
author,
who's,
writing
conformance
test
stuff.
This
is
on
you
to
do
it
and
they
should.
If
there's
a
standard,
run
book
or
something
like
that,
then
that
makes
it
easier
and
it
should
be
easier
for
people
to
do
it
and
then
that's
then
those
set
of
runbooks
is
then
the
set
of
things
that
you
would
base
any
eventual
automation
on
right.
A
Okay,
yeah,
that
makes
sense.
We
have
a
new
performance
docs
page
coming.
It
seems
like
we
can
link
out
to
some
run
books
from
that
page,
and
so,
if
anyone
already
has
a
run
book
for
conformance
tests
or
is
able
to
add
one
soon,
that
would
mean
that
there's
a
chance,
your
conformance
test
will
be
run
automatically
against
aprs
well,
not
automatically,
but
as
a
requirement
for
them
being
reviewable.
B
Yeah,
I
think
I
think
I
did
see
your
conformance
stocks.
They
are.
I
think
we
should
get
that
one
in
and
then
add
this
stuff
as
a
you
know,
as
an
additional
issue
to
then
put
on
that
page
to
figure.
C
A
Cool
thanks
I'll
add
an
issue
to
track
that
and
yeah
and
go
from
there
great
okay.
Next
up.
I
just
added
this
as
a
note,
because
it's
the
next
one
of
these
meetings
anyway
keith
and
gamma
team,
are
going
to
be
presenting
something
next
week.
I
just
does
a
informational
status
update,
so,
if
you're
interested
in
how
that's
going
and
you
haven't
been
able
to
attend
their
meetings,
that'll
be
next
week.
A
With
that,
let's
get
into
pr
and
issue
triage,
I
wanted
to
highlight
this
issue
that
was
basically
asking
for
a
back
port,
a
dot
something
release
for
conformance
tests
only.
A
I
think
that
makes
sense
and
shane
agrees
with
me,
but
I'm
open
to
broader
discussion
here
right
that
it
will
be
the
first
time
we've
had
a
conformance,
only
release
that
feels
acceptable
to
me,
but
interest
in
what
others
think
here.
B
Patcha,
a
patch
version,
conformance
only
release,
makes
perfect
sense
to
me.
That's
one
of
the
best
candidates.
I've
ever
heard
for
a
patchwork.
G
B
Yeah,
I
think
the
important
the
important
thing
is
just
to
make
sure
that
we're
not
bringing
in
new
conformance
tests
for
like
performance
tests
for
new
behavior
we're
only
bringing
in
like
bug,
features
to
behavior
that
existed.
You
know,
5-0,
you
know
what
I
mean
rob.
We
need
to
be
pretty
careful
about
that.
Sorry.
A
That
that's
a
very
good
and
and
close
distinction
here
and
depends
on
okay.
So
there
there's
two
interpretations
of
that.
Does
that
mean
we
should
avoid
adding
in
all
net
new
conformance
tests
that
have
been
added
after
the
release,
or
does
that
mean
we
should
avoid
backboarding
just
conformance
test
for
net
new
behavior.
B
You
know
we
tightened
up
some
of
the
language
about
the
conditions
that
actually
changed
the
behavior
a
little
bit
like
you
know
in
this
case
the
condition
should
be
this
other
one
and
not
this,
and
so
doing
that
will
actually
break
people's
implementations
if
they
were
conforming
to
the
old
one.
That
will
actually
be
a
breaking
change.
So
should
that
be
a
patch
release,
probably
not
so
yeah
like,
so
it
is
important
for
us
to
be
like
okay,
okay,
you
know
we
need
to
think
carefully
about
what
that
api
guarantee
thing.
B
Actually
is
there
yeah,
it
feels
like
we
shouldn't
be
breaking
if
you're
conformant
to
o50,
you
should
and
we
pat,
and
we
do
a
patch
release
to
to
do
conformance
bug
fixes
you
should
still
be
confirmed
without
having
to
change
too
many
things
right.
Like
you
know,
we
shouldn't
break
you
in
a
patch.
We
back
your
conformance
in
a
patch
release.
A
B
B
You
know
it's
kind
of
like
we
just
haven't
written
down
like
what
are
the
conformance
guarantees.
Yeah
we
need
to
just
agree
on.
You
know.
Is
this
okay
and
and
say
that
that's
the
expectation
I
mean
this
is
the
first
time
that
we've
done
05
was
the
first
time
we've
done
this
right.
So
it's
probably
okay
for
us
to
say,
hey
your
conformance
and
actually,
probably
this
discussion
is
kind
of
dependent
on
the
outcome
of
the
discussion
about
how
we
record
conformance
like
what
level
of
components
passed
and
what
version
of
components
testing
you
passed.
B
And
that
way
you
say:
hey,
maybe
you're
conforming
to
the
o5o
conformance
tests,
but
you're
not.
B
B
Releases
as
well,
because
the
performance
has
no
change,
that's
okay,
but
we
just
haven't
set
that
expectation
yet
and
it's
just
about
expectation
management.
So
we
can
do
anything.
We
like.
We
just
have
to
tell
people
what
we're
going
to
do
and
then
do
that.
A
A
The
reason
the
reason
I'm
even
suggesting
that,
even
though
it
seems
to
go
against
december
and
all
these
things
is
that
if
we
wait
between
minor
versions
to
include
these
kinds
of
breaking
changes,
even
if
they're
fixes
it
cements
this
behavior
and
this
expectation
that
when
this
bad
thing
happens,
I'm
gonna
get
this
condition.
A
A
B
C
B
C
C
A
A
Yeah
that
makes
sense
to
me
I'll
work
on
a
pr
to
document
this,
and
then
we
can
work
on
back
porting
as
we
see
fit.
A
A
It
seems
like
broad
agreement
that
we
need
to
do
this,
but
nobody
working
on
it.
Yet
I
would
love
to
get
this
fixed
sometime
soon.
So
this
is
just
a
reminder
that
this
issue
exists.
It's
well
defined
and
if
you're
interested
would
love
to
get
this
one
in
yeah.
A
Next
one
is
the
okay
yeah
this
one,
we
had
an
issue
that
goes
back
yeah
to
the
800s
that
it
would
be
cool
to
have
utils
to
deal
with
all
our
custom
types.
I
we
introduced
all
kinds
of
fun
custom
types
to
simplify
crd
validation
and
in
the
process
made
it
more
annoying
to
actually
implement
the
api.
A
This
is
basically
just,
I
think,
some
libraries
are.
Some
implementations
already
have
utils
like
this,
just
the
to
and
from
pointer
for
basically,
every
custom
string
type
we
have
or
or
anything
like
this.
A
C
C
C
C
D
A
C
I
have
it
on
my
code
base.
I
will
use
it.
A
Great
okay,
that's
good
to
hear
I
that
sounds
like
we
should
move
forward
with
this
I'll.
Make
sure
that
we
have
the
comments
that
were
mentioned
here,
but
anyone
else.
If
you
want
to
review
as
well,
we
would
appreciate
it.
B
Yeah,
I
would
also
say
shane,
I
see
your.
I
see
your
concerns
about
having
it
in
the
api's
package,
but
I
kind
of
feel
like
it's
best
to
have
it
really
close
to
the
actual
crd,
like.
B
B
Yeah,
I
think
I
think
it
the
the
one
thing
that
I
would
add
here
and
I'll
put
a
comment
on
the
pr
to
this
effect
is
that
it
should
have
a
little
globe
at
the
top
being
like
you
know,
what
sort
of
what
sort
of
things
should
be
in
this
file?
You
know,
like
you
know
it
shouldn't,
have
you
know
we
shouldn't
be
adding
custom,
validation,
things
or
anything
like
that
in
here,
that's
the
web
hook.
This
is
for
like
very
simple.
A
A
Cool
okay,
I'll
make
sure
I
add
a
couple
more
follow-ups
on
this
one.
More
feedback
welcome.
This
is
13
10.
A
A
The
last
one,
this
one
is
something
that
I
covered
briefly
at
the
very
beginning
here
and
that's
steve's
pr
that
adds
hp
route
method
matching.
This
is
something
we
know
we
need.
It
looks
very
straightforward.
I
would
just
having
looked
at
it
would
almost
approve
it,
but
of
course
we
actually
need
to
run
it,
and
I
haven't
run
this
yet
so
yeah.
A
That's
all
I've
got
for
this
one.
That's
everything
on
triage.
I
think
we've
got
everything
on
the
agenda
any
last
things
or
can
have
our
minute
back.
A
Okay,
all
right!
Well,
thanks
everyone,
it's
been
great
to
discuss
this
and
talk
to
everyone
next
week.
Thanks
see
ya.