►
From YouTube: SIG Network Gateway API Bi-Weekly Meeting 20220627
Description
SIG Network Gateway API Bi-Weekly Meeting 20220627
A
Record
all
right
welcome
to
the
gateway
api
meeting
for
june
27
2022
we've
got
a
few
new
people
so
before
we
go
any
further,
let's
see
anyone
who's
interested
again
completely
optional,
but
anyone
who's
interested
in
introducing
yourself
that
hasn't
had
a
chance
before
we'd
love
to
hear
who
you
are
and
why
you're
here.
B
Yeah
I'll
start,
my
name
is
keith
maddox,
I'm
a
my
maintainer
for
the
smi
project.
I
also
work
on
the
open
service
mesh
project
looking
forward
to
chatting
about
gaba
api
and
some
of
the
service
mesh
use
cases
and
ways
to
go
about
doing
work
on
that.
C
All
right
and
I'm
bridget
cromhold
and
I'm
a
pm
on
a
team
supporting
keith's
dev
team
and
I'm
really
excited
to,
I
think,
there's
there's
got
to
be
a
pokemon
collect
them
all
right.
If
you
get
a
chance
to
go
to
every
sig
and
every
working
group
there's
some
sort
of
badge
right,
I
I'm
I
go
to
cig
network
all
the
time
and
I'm
excited
to
come
to
the
gateway
api
meeting.
A
Well,
welcome
both
and,
yes,
I'm
sure,
familiar
faces
to
many
here,
but
I
should
actually
before
I
go
any
further
anyone
else
who
might
want
to
do
an
intro
here.
A
Cool,
okay,
well,
welcome
everyone
to
gateway
api
meeting.
Shane
is
actually
going
to
be
running
the
vast
majority
of
today's
meeting,
which
is
why
he's
sharing
his
screen
today,
but
I
wanted
to
do
a
brief
intro.
We
are
welcoming
a
new
maintainer
to
gateway
pi
today
and
that
is
shane
fittingly
enough
so
congrats
to
shane
for
making
it
he's
been
a
great
contributor
for
years
now
and
well
deserved.
A
We're
excited
to
have
him
officially
on
board
as
a
maintainer.
I
know
he's
been
active
for
some
time,
there's
a
pr
that
I
filed
just
before
this
meeting.
I
talked
with
all
the
different
maintainers.
I
think
it's
very
well
earned
and
thank
you
so
much
for
all
your
contributions
already
and
yeah
happy
to
formalize
it.
With
that
I'll
hand
it
over
to
you
and
yeah
yeah
you
can,
you
can
run
the
rest.
D
Yeah,
actually
nick
yeah,
I'm
super.
I'm
super
glad
to
be
yeah
to
have
shane
on
board
as
a
maintainer
as
well.
Yeah,
more
maintainers
is
always
better
right
and
yeah
shane.
You
do
a
great
work.
Man
thanks
appreciate
it.
Thank
you
very
much.
E
A
Yeah,
oh
man,
there's
a
okay,
so
there
there's
both
a
lot
and
a
little.
I
I
I
can't
you
know
part
of
me
says:
oh
man,
we
could
have
this
done
and
out
the
door
by
thursday,
but
that
is
my
you
know.
Just
everything
is
perfect
and
everything
works.
You
know
perfectly.
Between
now
and
thursday
realistically
says:
yeah
yeah.
A
You
see
a
couple
of
these
issues
already
have
prs
filed
against
them,
we're
getting
close,
but
I
think,
given
the
holidays
coming
up,
at
least
in
the
u.s
well
in
canada,
canada
day
on
friday,
u.s
we've
got
the
fourth
coming
up,
probably
other
holidays,
I'm
missing,
but
I
imagine
this
is
a
time
that
a
lot
of
people
are
going
to
be
in
and
out,
and
so
the
next
week
or
so
may
may
not
be
as
productive
as
other
weeks
of
the
year.
A
So
that
said,
I
think
that
just
to
give
general
timeline
I've
gotten
this
question
a
lot.
I'm
not
I'm
sure.
I'm
not
the
only
person
who
who
keeps
on
getting
asked
like
when
is
this
thing
releasing
you're.
So
close,
you
know,
like
you've
been
so
close
for
so
long.
I
would
like
to
to
set
a
rough
goal
of
july,
11
or
12
just
kind
of
after
the
holiday
bump.
Does
that
seem
reasonable?
E
The
blog
post,
one
is
just
about
done,
although
it
depends
on
whether
or
not
we
actually
want
to
add
anything
significant
for
like
talking
about
service
mesh,
which
I
don't
really
know
when
I
wrote
the
blog
post,
I
wasn't
like
so
so.
If
we
want
to
talk
about
that,
there
might
be
a
little
bit
more
to
do
there.
Otherwise,
I
think
it's
just
about
just
about
ready,
yeah
conflict
guidance.
What
needs
to
be
reversed
to
an
invalid
back
end.
A
This
is,
this
is
a
good
one.
I
apologize
for
starting
this
issue.
Yeah
we've
had
some
great
discussion
on
this
one.
I
think
nick
you're,
the
one
who
no
okay,
sorry,
I
I've.
A
A
Conversational,
yes,
it
has.
I
missed
the
last
two
comments,
but
I
think
the
yeah.
What
what
I
remember
reading
the
last
comment
I
read
was
nyx
and
that
was
that
500s
were
preferable.
I
think
everyone
was
caught
was
falling
behind
the
option.
Two.
I
gave
to
start
this
threat
nick
you
you've
been
commenting
more
than
I
have
on
this.
Does
that
sound,
accurate.
D
Yeah
yeah,
I
think
so
the
this
this
comes
down
to
sort
of
that
general
philosophy
thing
I
mean.
Generally,
we
have
the
philosophy
hey.
D
We
should
try
not
to
break
existing
configs,
but
I
think
the
we
need
to
be
really
really
careful
with
that,
because
trying
not
to
break
existing
conflicts
can
very
easily
turn
into
the
api
must
only
be
used
in
this
very
specific
way,
and
so
that's
why
I'm
generally
in
favor
of
take
whatever
the
config
is
and
if
that's
a
broken
config,
then
you
get
a
broken
config,
because-
and
in
this
case
that
means
that,
if
the
that
that
means
that's
option
two
and
that's
because
you
know
if
you've
asked
for
a
back
end
to
have
50
of
the
traffic
and
that
back
end
is
invalid
for
some
reason,
because
no
reference
grant
because
there's
no
end
points
because
there's
a
bunch
of
different
reasons
it
could
be.
D
Then
you
get
that
percentage
of
you
get
whatever
the
weight
of
that
thing
is
as
five
hundreds
in
my
mind.
That's
that
matches
up
well
because,
like
that's
what
you've
actually
asked
for
like?
Maybe
it's
not
what
you
want,
but
it's
what
you've
asked
for,
and
you
know
like,
and
we
can't.
I
think
that
trying
to
impute
you
probably
don't
want
to
do
this
in
the
actual
api
is
a
mistake.
I
think
the
api
should
say
this
is
what
you've
asked
for.
D
This
is
what
you
get
and
then
it's
up
to
implementers
to
be
like
to
do
stuff
to
be
like
you
know,
if
you're
supplying
an
invalid
backend
to
try
and
call
that
out-
and
you
know
and
stop
that
from
happening
rather
than
making
it
so
that
when
you
have
it
in,
like
some
coffee,
that's
not
great
that
that
we
try
and
rescue
you
from
yourself.
A
Yeah
and
so
that's
why
I
think
that
makes
sense
the
what
I
was
already
seeing
on
that
made
me
think
that
we
had
largely
reached
consensus,
and
the
last
two
comments
agree
with
your
comment,
which
makes
me
think
we
basically
have.
So
I
think
this
is
what
this
needs
need,
someone
to
follow
up
and
actually
update
the
spec
based
on
these
comments,
but
it
seems
like
we
may
be
at
a
point
where
we
understand
what's
needed.
D
That
makes
sense
to
me,
I
guess,
as
the
person
who
has
been
the
most
strident,
let's
be
terrible
about
this.
I
probably
should.
It
should
probably
should
be
on
me
to
to
walk
the
talk
here
and
actually
update
the
spec.
E
All
right
so
that
covers
that
one,
this
one's
one
or
two
things,
maybe
clarified
behavior
behind
multiple
query
parameters.
I
think
this
is
the
one
I
commented
on
yeah,
so
this
one
is
just
a
documentation
issue
more
or
less.
I
think,
and
unless
somebody
has
something
else
to
say
on
this
one.
The
idea
is
that
with
http
header
matches
we
say
that
anything
past
a
first
match
of
a
name
like
basically
gets
ignored.
E
So
if,
like
you
have
a
header-
and
you
have
like
three
of
them
with
the
same
name-
the
next
everything
after
the
first
two
or
the
first
one
gets
ignored
it's
kind
of
weird
but
the
the
parameters
ref
works
kind
of
similarly,
but
isn't
documented.
That
way,
so
chris
wants
to
go
ahead
and
get
that
updated.
So
the
documentation
is
consistent.
So
unless
anybody
has
something
to
say
about
that
one,
I
would
say
that
one's
pretty
close.
I
think
he
already
put
in
a
pr
for
it.
A
Yeah
steve
put
in
pr
I
spent
a
little
bit
of
time.
Looking
at
I
got
distracted,
I
shameless
plug
I'd
say:
if
anyone
has
time
to
do
reviews
here
there
are,
we
have
a
huge
review
backlog
so
any
help
you
can
like
don't
feel
like
you're
under
qualified
to
review
code.
We
just
you
know
as
a
maintainer
looking
at
this.
If
I
see
someone
else,
that's
reviewed
it
and
says
hey.
This
looks
good
to
me
that
that
really
helps
me
as
a
sanity
check.
A
I
hate
to
be
the
only
person
or
nick
or
you
know
it's
or
shane
or
anyone
just
you
know
when
you're
the
one
person
reviewing
the
pr,
it's
always
better
when
there's
just
someone
else
another
set
of
eyes
confirming
that
yeah
this
does
make
sense,
so
the
the
very
high
level
you
know
five
minute
approach.
A
E
Yup
some
documentation,
some
validation,
all
right
cool,
so
not
too
far
along
on
that
one.
So
so
far
so
good,
I
think
I
saw
you
put
this
one
in
this
morning,
so
this
one
is
we
just
gotta
this
one's
just
a
little
bit
of
lifting
work
where
we
gotta
move
things
around
and
copy
things
over.
A
Right,
I
I
tried
to
make
this
something
that
was
approachable
for
as
many
people
as
possible,
so
if
you're
looking
for
a
way
to
get
started
in
gateway
pi,
a
lot
of
these
things
are
really
hey.
We
need
to
move
these
things
over
here.
We
need
to
rename
this
stuff.
I
I
I'm
hoping
it
it's
a
an
all
right
place
to
get
started.
A
The
downside
is
that
it's
relatively
time
sensitive
in
that
it's
blocking
blocking
this
release,
and
the
other
downside
is
that
some
of
these
things
in
this
list
will
conflict
with
each
other.
I've
tried
to
make
that
as
minimal
as
possible,
but
there's
a
chance.
You'll
have
to
do
some
rebasing
if
you
choose
to
take
one
of
these
on
and
someone
else
takes
a
different
one
on
and
yeah
whatever
so
heads
up,
but
again,
if,
if
anyone
has
some
time
to
to
take
this
on
and
really
wants
to
help
get
this
release
out.
A
These
are
good
opportunities
here
and
I
you
know,
I
tried
I
you
know,
I'm
not
I'm
not
trying
to
say.
I
have
a
complete
list
here
of
everything
that
needs
to
be
updated
in
the
docs,
but
I
think
this
is
a
a
good
starting
point
and
would
make
it
an
acceptable
release,
ready
documentation,
at
least.
D
D
Absolutely
a
large
issue
that
the
way
that
we
have
done
the
doc
structure
at
the
moment
does
not
work
very
well
with
having
we
kind
of
assumed
when
we
built
the
dock
structure,
originally
that
all
the
resources
would
be
at
the
same
api
version
at
the
same
time,
which
is
now
no
longer
the
case
so
like
yeah
and
so
there's
definitely
a
there's,
a
big
chunk
of
work
to
be
done.
That
maha
was
looking
into
us
for
us.
D
She
I
haven't
talked
to
her
for
a
while,
so
she
may
still
be,
but
basically
we
want
to
version
the
whole
site
same
way.
Cube
does,
via
the
bundle
version,
the
release
version
and
then
that
should
make
some
of
this
stuff
a
little
easier
to
handle.
But
right
now
it's
gonna
be
a
bit
funky.
How
we
do
this
until
we
fix
that,
but
I
think
having
something
is
better
than
them
being
like.
Oh,
we
can't
do
this
yet
until
we
conversion
the
whole
site,
because
that's
gonna
take
a
while.
E
A
D
Yeah
and
yeah,
please
feel
free
to
if
you
see
me
online
on
slack.
That
means
that
I
am
ready
to.
You
know,
answer
questions
about
this
sort
of
stuff
as
well
all
right
that
goes
for
any
of
these
yeah.
I
I
remember
that
also
that
I
am
in
australia,
so
I
am
online
in
most
of
yours
like
afternoon,
slash
evening.
A
I
feel
like
between
nick
shane
and
myself,
we
have
near
24
7
maintainer
coverage
coverage,
just
the
different
time
zones.
We've
got
going
on
more
maintainers
welcome,
but
we
can
try
and
spread
that
that
time
zone
coverage
around.
E
Last
one
here
was
there
a
pr
for
this
one.
A
Yeah
spencer
has
a
pr
for
this.
This
one
is
just
needing
review.
I
think
someone
already
reviewed
it
and
he's
responded.
I
can't
remember
yeah.
A
So
I
said
july
12
as
a
target
that
that
gives
us
you
know,
I
don't
I
hate
to
skip
ahead
in
the
the
agenda,
but
I
think
this
is
a
really
short
thing.
The
next
meeting
would
fall.
I
think
on
the
fourth
in
the
us
it
it
it
is
a
holiday
for
what
I
would
imagine
is
a
lot
of
the
attendees
on
this
call.
A
D
D
Lts
working
group,
I
ran
a
series
of
like
eight-pack
meetings
and
I
think
it
was
about
three
months
where
I
would
dial
in
for
10
minutes
and
sit
there.
Looking
at
my
screen,
sadly,
I
record
the
meeting.
E
D
I
think
we
should
front
up
and
if
we
all
agree
on
that,
we'll
take
a
lazy
consensus
moment.
But
then,
if
we
do
agree
on
that,
we
I
think
we
should
just
update
the
milestone
and
put
the
date
on
there
get
home
milestones.
Let
you
put
dates.
D
Yep,
exactly
I'm
going
to
give
people
you
a
count
of
five
to
to
object.
Please
feel
free
to
eject
in
the
chat
or
by
you
know,
speaking
five,
four
three,
two
one:
okay,
good
good.
E
Cool,
so
next
is
keith.
Some
of
you
may
have
seen
his
email
and
sig
network
and
chat.
I
think
he
posted
in
the
chat,
maybe
as
well,
but
go
ahead.
Keith
tell
us
about
your
draft
proposal
for
the
work
group
for
mesh.
B
Sure
so,
as
as
many
of
y'all
know,
smi
has
been
a
cncf
project,
looking
at
kind
of
a
common
api
for
service
meshes
and
traffic
general
traffic
thingies
in
kubernetes
kind
of
around
the
same
time
as
gateway
api
actually,
but
looking
at
the
api
services
of
both
projects
and
a
lot
of
the
fantastic
work
that
this
group
has
done,
get
your
api
talking
to
people
across
the
community.
B
There's
been
some
interest,
a
good
bit
of
interest
actually
in
coming
together
and
really
firming
up
some
of
the
use
cases
and
the
mesh
use
cases
for
gateway
api.
There
are
little
references
here
and
there
within
the
specification
itself
and
so
looking
to
put
together
a
group
a
a
place
for
people
to
to
have
conversations
and
such
specifically
for
mesh
use
cases.
The
reason
why
a
separate
working
group
has
kind
of
been
suggested
or
been
pursued
is
because
you
don't
want
to
distract
from
a
lot
of
the
great
work.
B
That's
that's
happening
here
for
the
east,
for
the
north,
south
ingress
egress
use
cases,
and
so
this
would,
you
know,
naturally
kind
of
report
into
kind
of
gateway,
api
gateway.
Api
would
still
you
know
own
all.
The
resources
and
such
we'd
just
be
exploring
use
cases
and
proposing
patterns
or
modifications
for
mesh
use,
use
cases
upstream.
I
wanted
to
get
a
pulse
on
how
folks
here
feel
about
that
before
moving
any
further,
so
yeah
kind
of
opening
the
floor
to
hear
what
folks
think.
D
D
What
you'll
come
up
with?
I
think
it's
a
great
idea
to.
I
know
that
we've
definitely
had
conference
lots
of
conversations
about
how
this
or
that
or
this
thing
would
be
served
by
in
the
mesh
use
case
and
yeah
it'd
be
good
to
really
get
some
good
use
cases
and
for
you
all
to
say,
you
know,
hey.
We
think
it
should
work
like
this.
We've
tried
to
make
sure
that
there's
spots
for
it
in
lots
of
places.
That's
why
there's
references
you
know
we've
sort
of
tried
to
be
like.
D
Oh,
that
sounds
like
something
mesh
might
want,
but
yeah
like
I
would
love
to
see
it
yeah.
I
think
I
I'm
only
supposed
to
plus
one,
but
I
would
plus
ten
so.
A
Yeah
I'll
I'll
follow
up
and
say
I
agree
with
what
nick's
saying
I
mean
I
I'm
really
excited
for
this.
You
know
I
looked
at
the
the
proposed
chairs
of
this
group.
You've
got
an
awesome
list
of
people
working
on
this.
I
I
am
pumped
for
for
where
this
is
gonna
go.
You
know,
like
nick
said,
I
know.
A
As
we
were
building
this,
we
we
tried
to
leave
holes
or
like
maybe
this
could
work
for
mesh
like
you
know,
but
it
so
far
gateway
api
has
been
led
with
a
lot
of
ingress
focused
implementation
implementers.
So
I
am
not
a
mesh
expert,
I
I
look
at
it
and
say
yeah.
I
think
this
could
work,
but
I
am
really
excited
to
get
some
more
mesh
experts
really
showing
how
this
api
could
fit.
I
I
think
this
working
group
is
a
great
idea.
A
Let
us
know
how
we
can
support
you
and
yeah.
I'm
excited.
B
You
have
a
channel
yet
that
we
can
watch
on
the
grenady
slack
not
at
the
moment,
but
assuming
this
goes
through.
That
would
be
something
to
to
discuss
and
figure
out
what
the
best
form
of
communication
would
would
be
for
that.
But
I
imagine
that
would
be
kind
of
customary
for
working
groups.
E
We
have,
we
have,
I
have
ingress
and
mesh
kind
of
in
my
purview,
so
I'm
excited
to
join.
H
Yeah,
I
I
I'm
kind
of
in
a
similar
spot
to
shane.
My
primary
focus
is
ingress,
but
you
know,
being
you
know,
my
project
doesn't
work
without
the
console
service
mesh,
so
so
that's
obviously
near
and
dear
to
my
heart
and,
of
course,
mike
morris
who's.
One
of
the
proposed
chairs
of
the
group
is
actually
one
of
my
engineers
as
well.
I
think
the
only
only
caution
I
would
say
is.
H
I
would
like
to
see
it
be
such
that
the
working
group
would
bring
gaps
to
the
group
to
this
group
and
then
the
the
overall
gateway
api
community
can
work
on
reviewing
the
gaps,
and
you
know
first
thing
about
them
and-
and
you
know,
come
up
with
a
final
a
final.
You
know
proposal
that
we
can
get
consensus
on
and
move
forward
with.
I
think
just
be
cautious
about
we're
having
a
with
a
group
coming
in
that
hasn't
been
involved
with
the
api
this
api.
H
Before
going
about
doing
things,
you
know
in
ways
that
end
up
being
different,
or
you
know
inconsistent
with
how
we've
wanted
to
do
things
and
tried
to
do
things
historically
here
you
know
so
get
end
up
with
like
weird
inconsistencies
or
something
so
as
long
as
we,
I
think,
as
long
as
we
kind
of
follow
that
process
of
having
you
know
bring
gaps
here
or
you
know
or
pre-gap,
but
I
think
that's
that
would
work
well.
B
Yeah,
I
think
bridget
put
there
in
the
chat
a
good
point.
There
is
that
working
groups
don't
own
any
code
or
api
surface
by
definition
of
what
they
are
so
the
existing
get
process.
I
I
what
kind
of
spurred
from
this
conversation
actually
was
the
I
think
it's
kept
713
policy
attachment,
and
so
I
I
imagine
that
a
lot
of
the
modifications
or
proposed
additions
to
the
api
spec
would
take
place
in
a
similar
fashion
to
that
so
yeah.
I'm
I'm
on
board.
With
with
that.
D
Yeah,
I
definitely
think
side
note,
since
you
mentioned
policy
attachment.
That
is
definitely
an
area
that
needs
further
exploration
from
anyone
who
is
interested.
I
think
we
could
really
do
with
like
some
standard,
some
sample,
implementations
and
stuff
like
that
talked
about.
I
think,
there's
a
that's
a
that's
a
framework
for
a
framework
rather
than
an
actual
functional
thing.
At
the
moment,
it's
on
my
it's
on
my
to-do.
B
Yeah
I'll
off
z,
I've
been
looking
at
policy
attachment
for
all
z
use
cases
for
a
while.
So
so
should
I
get
a
chance
to
I'm
gonna?
Be
I'm
gonna
be
talking
about
that
with
anybody?
I
can
because
that's
a
personal,
personal
pursuit
of
mine.
E
B
Yeah,
based
on
the
documentation
and
sig
network
or
in
commerce,
general
kubernetes,
community
lifecycle
documents.
The
next
step
would
just
be
to
create
a
pr
into
the
kubernetes
community
repo
with
this
information,
and
then
we
be
good
to
go.
E
Yeah,
let
us
know
if
we
can
be
helpful
and
we'll
join
the
channel
as
soon
as
it's
up.
E
Candice
you're
next
has
anyone
ever
discussed
plans
for
a
gen?
Sorry?
Has
anyone
ever
discussed
plans
for
a
general
purpose
gateway
api
operator?
I've
heard
this
talked
about
before?
I
Yes,
yes,
I'm
here
I
had
wanted
to
mention
to
keith,
though
that
I
saw
this
proposal
earlier
today.
I
work
for
red
hat
and
somebody
brought
this
proposal
to
our
meeting
today,
which
is
the
meeting,
is
there's
a
lot
of
mesh
guys
in
the
in
the
meeting.
I'm
not
a
mesh
person
myself.
I
work
on
the
network
edge
at
red
hat,
but
just
wanted
to
mention
that
there
was
some
interest
from
red
hat
today,
but
this
topic
is.
I
I
So
one
of
the
things
that
we
want
to
do
is
so
we
in
in
openshift.
We
have
an
ingress
implementation,
an
ingress
operator
implementation,
and
we
want
to
move
from
it
being
strictly
h.a,
proxy
and
kubernetes
ingres
to
something
more
with
gateway,
api
and
another
implementation.
I
What
we've?
So
we
were
going
to
be
starting
with
a
specific
implementation
looking
into
a
specific
implementation
like
istio,
probably,
but
what
we?
Our
project
managers,
have
recently
sort
of
seen
how
many
different
implementations
there
are
of
gateway
api
recently
and
have
asked
us
to
look
into.
You
know
comparing
some
of
the
implementations,
and
maybe
we
would,
in
the
end,
offer
more
than
one
sort
of
in
a
plugable
way.
I
And
that's
the
extent
that
I've
gotten
to
so
far
in
terms
of
our
our
discussion,
and
we
know
that
there's
a
contour
operator.
That
was
something
that
we're
kind
of
looking
at
that
dane
hansen
worked
on
for
some
time
and
that's
something
that
we
could
use
as
a
as
a
base.
Although
we
decided
in
the
end
that
we,
we
probably
wouldn't
choose
contour.
I
But
I
guess
what
else
did
I
forget
to
mention
yeah,
a
vanilla
gateway,
api
operator
that
would
support
more
than
one
gateway
api
implementation,
so
that
our
customers
could
pick
and
choose.
E
Am
I
am,
I
understand,
I'm
sorry,
I
may
be
a
misunderstanding.
It
sounds
like
maybe
you're
talking
about
like
controller
machinery
that
could
then
back
into
any
proxy,
like
any
proxy
could
kind
of
be
plugged
into
controller
machinery.
That
would
allow
them
to.
Is
that?
What
is
that?
What
you're
getting
after?
Does
that
not
sound
right.
I
No,
that
doesn't
doesn't
exactly
sound
like
it.
In
openshift
we
have
an
ingress
operator,
so
it
would
be.
You
know,
to
some
extent,
like
the
ingress
operator.
Increase
operator
works
with
an
ingress
class,
and
you
know
works
between
an
ingress
class
and
an
proxy
router.
That's
how
we,
how
we
implement
increase,
but
we
want
to
instead
of
using
ingress
ingress
operator
and
h.a
proxy
we'd,
be
sort
of
in
the
ingress
operator.
H
I
think
one
thing
to
consider
is:
you
know:
you've
got
here
listed
that
you
know,
contours
has
got
an
operator
and
istio
has
one
console
for
api
gateway,
we're
looking
pretty
seriously
at
implementing
one.
I
don't.
I
can't
obviously
speak
for
for
others,
but.
H
If
you're,
if
you're,
trying
to
if
you
implement
a
generic
one,
then
if
an
implementation
has
its
own
operator
that
handles
gateway
api,
you
know
how
would
those
interact
or
would
it
cause
a
problem
for
you
know
that
implementation
from
the
third
party
implementation,
you
know,
like
bars
or
istios
or
contours.
D
Okay,
cool
so
yeah,
I
think
for
contour
operator,
contour
operator's,.
E
D
Is
to
set
up
the
individual
sort
of
contour
instances
that
fulfill
individual
gateways?
I
think
that
one
of
the
one
of
the
smaller
problems
with
the
idea
of
having
something
more
generic
is
that,
because
we
kind
of
left
it
a
little
bit
up
to
the
implementation,
exactly
sort
of
how
they
build
their
bits
that
say:
reconcile
like
gateway,
class
and
gateways
and
exactly
you
know,
do
the
gateways
get
merged.
E
D
They,
you
know,
spin
up
different
data
planes
and
a
few
other
things
like
that.
It's
going
to
be
tricky
to
build
something
that
would
that
would
do
lots
of
different
ones.
I
think
that
it
would
be
reason
that
it's
definitely
possible.
If
you
have
like
a
limited
set
of
options
that
are
that
work
in
reasonably
similar
ways,
you
could
build
an
operator
that
would
sort
of
let
you
pick
between.
C
D
I
think
that
the
way
that
I
would
think
about
this
is
like,
what's
the
so,
if
this
is
an
operator,
what
resources
are
reconciling
and
what
you
and
what
happens
when
you
get
one
of
those,
so
it
sounds
like.
Maybe
you
want
to
reconcile
gateway
classes
and
when
you
create
a
gateway
class,
like
you
pick
up
that
this
is
a
gateway
class
of
like
a
your
contour
type
or
an
istio
type,
and
then
you,
you
know
provision
the
the
the
gateway
controller
deployment
to
sort
of
satisfy
that
gateway
class.
D
If
that's
the
case,
then
then
that
seems
like
a
reasonably
straightforward
thing,
but
and
then,
and
then
it's
the
implementation's
job
to
handle
like
the
gateways
inside
the
gateway
class.
But
that
would
be
that
that
you,
if
I
was
looking
to
build
something
like
this,
that
would
that
would
be
how
I
would
think
of
it
like
which
which
pieces
of
the
puzzle
reconcile
which
kubernetes
resources
and
like
which
bits
do
you
want
to
be
able
to
switch
out
is,
is
how
I
would
start.
D
I
think
I
know
that,
for
I
mean
obviously
I'm
a
maternal
contour,
but
I'm
also
going
to
turn
around
for
gateway
and
for
contour,
because
we
decided
that
a
gateway
equals
an
envoy
deployment
that
meant
that
the
contour
operator
sort
of
will
deploy
will
manage
satisfying
your
gateway
request
by
setting
up
onboard
infrastructure
for
you,
inova
gateway.
We're
actually
doing
it
slightly
differently
that
the
onward
data
plane
matches
to
the
gateway
class.
D
So
we
are
only
going
to
be
we're
going
to
be
leaving
the
creation
of
the
on
gateway
deployment
to
the
to
the
to
the
human
for
now
and
then
so,
there
is
a
place
where
you
could
put
like
a
layer
above
the
top
that
says:
hey,
there's
a
you
created
a
gateway
class
that
you've
marked
as
being
an
onboard
gateway,
one
in
some
fashion,
tvd,
and
then
that
that
one,
that
would
that
operator,
would
then
take
that
gateway
class
and
stand
up
the
onboard
gateway
installation.
D
That
would
then
stand
up
for
the
envoys
to
satisfy
the
gateway
thing.
If
the
key
part
is
that
in
all
of
that,
like
I
said
the
the
so,
the
key
distinction
is
that
you've
got.
You
know
that
you've
got
to
figure
out
like
which
programs,
which
containers,
which
controllers
are
going
to
reconcile,
which
resources
and
that's
how
you
would
build
that.
I
think,
and
I
think
right
now.
I
don't
think
a
lot
of
people
have
thought
too
much.
D
Yet
to
answer
the
exact
question,
I
don't
think
we've
talked
too
much
about
a
generic
go
api
operator,
because
people
have
been
doing
things
in
such
different
ways
and
not
really
what
the
commonalities
are
between
the
ways.
People
are
doing
things
yet
because,
although
we've
got
12,
implementations
like
we
haven't,
spent
a
lot
of
time
sort
of
comparing
and
contrasting
how
everyone's
doing
everything.
Yet
I
think
spending
some
time
doing
that
exercise.
D
I
would
dearly
love
to
read
like
a
report
of
like
all
12
implementations
and
how
they
all
how
they
all
manage
the
the
different
bits,
reconciling
which
bits
you
I
yeah.
I
would
like
to
know
that
information,
but
I
do
not
have
the
benefits
to
do
the
analysis
myself
so
yeah.
I
would
definitely
love
for
someone
to
do
that,
and
it
feels
like
that
would
be
something
that
you
would
need
to
do
to
figure
out
like
to
figure
out
what
you're
talking
about
does.
That
is
that.
I
I
Yes,
we
we
have
started
doing
that
to
some
extent,
we
probably
won't
do
like
all
12
implementations,
but
we'll
and
we'll
choose
the
ones
that
that
are
more
likely
to
to
have.
You
know
it
have
acceptance
for
for
a
variety
of
different
criteria
that
will
will
apply
and
which
we
may
not
even
know
yet
so
we
know
some
of
them,
but
not
all
of
them,
but
I
was
wondering
if
anyone
had
had
ever.
You
know
started
doing
something
like
that
and
I
guess
it
looks
like
if
they
have.
I
They
haven't
really
talked
about
it
in
these
in
this
meeting,
so
you
know
it's
worth
seeing
if
someone
else
is
doing
something
like
this.
First
before
we
try
to
embark
on
such
a
project.
So
thank
you.
H
I
just
make
one
quick
comment:
is
you
know
I've
already
got
customers
that
are
using
openshift
and
have
a
console
api
gateway,
my
product
deployed,
you
know
so
and,
like
I
said,
we're
we're
looking
at
doing
an
operator
down
the
road.
So
you
know,
obviously
from
my
my
bias
perspective.
You
know
I'd
be
concerned
about
you
know,
compatibility
issues
with
if
you,
if
red
hat,
implements
a
a
generic
controller
for
gateway
api,
you
know,
would
that
be
locking.
You
know
other
implementations
out,
like
you
know,
console
or
or
istio
or
whoever
you
know.
I
H
I
H
So
you
know
I
I'm
I'm
no
expert
on
on
controllers
and
operators
here,
so
maybe
there
wouldn't
be
an
issue
having
you
know,
both
in
the
environment
or
something
so
yeah.
Often
I'd
have
to
defer
to
others.
That
was
just
the
concern
I'd
raised
so.
I
F
A
It
is
a
single
controller,
but
we
do
have
different
gateway
classes,
for
example
one
for
xlb
one
for
ilb,
and
then
I
think
we
have
a
couple
multi-cluster
gateway
classes,
but
it
is
a
single
controller.
D
E
Yeah,
no
that's
okay!
I
was
just
gonna
suggest
that
it
seems
like
there
could
be
potentially
a
lot
more.
That
could
be
discussed
on
this,
but
at
least
in
my
mind,
there's
a
lot
of
question
marks.
This
might
be
a
good
candidate
candace
for
a
discussions
on
our
discussion
board
where
we
get
a
maybe
like
a
deeper
definition
of
what
generic
gateway
api
operator
means.
E
We
start
digging
into
some
of
these
kind
of
like
each
person
with
their
implementation
can
start
kind
of
talking
about
how
that
relates
to
their
implementation,
and
we
can
kind
of
have
a
a
longer
async
discussion,
because
I
feel
like
there's
a
lot
here.
Are
you
up
for
that?
Maybe
putting
in
a
discussion
for
it.
I
Yes,
let
me
go
back
to
my
team
first
and
you
know
it'd
be
better
if
we
had
some
more
stuff
to
discuss.
First,
I
just.
I
It
off
and
see
if
anyone
else
was
doing
anything
and
you
know
glad
to
know
all
the
comments
that
that
this
group
has
has
mentioned
so
far.
So
so
thanks
for
giving
me
the
chance
to
to
bounce
it
off
you
here
and
I'll
come
back
and
let
you
know
you
know
what
the
plans
are
and
if
I
have
any
good
docs
that
we
come
up
with
I'll
share
them.
E
I
E
Next
one
who's
this
guy,
so
I
just
wanted
to
see:
is
there
somebody
other
than
bobby
tables?
He
doesn't
seem
to
be
responsive
right
now
that
can
help
us
with
this.
A
I
so
I
know
he's
really
busy
he's
there
there's
a
summit
that
he's
running
tomorrow
inside
google,
so
I'm
assuming
that's
why
he
like,
I
know,
he's
busy
to
begin
with,
but
I
know
he's
got
a
big
thing
coming
on
this
week.
I
I
can
try
and
ping
him
as
well.
I'm
sure
there
are
other
people,
but
he's
just
who
I
happen
to
ping
most
often
about
these
things.
Okay,.
E
A
There's
yeah,
I
ca.
I
can't
remember
that
there's
a
there's,
a
slack
group
that
is
supposed
to
be
the
one
you
mentioned
for
these
questions,
but
I
can
never
remember
what
it
is.
I
know
that's
not
helpful,
but
there
are
other
people,
I
just
don't
know
who
they
are.
So
sorry.
E
Yeah,
no,
it's
fair.
Maybe
this
is
one
of
those
things
where
I
can
like
ping
the
world
and
get
like
famous
just
for
like
pinging
everyone
in
slack.
No,
I
don't
want
to
do
that.
All
right
next
thing.
Did
you
try
control
x?
Did
you
try
to
contribute?
E
No,
I
didn't
that's
a
good
point.
So
yeah
I'll
go
I'll
go
check
with
contributes.
G
Oh,
like
I
would
love
to
do
that,
but
I'm
not
sure
how
what
exactly
me
to
do
like
this
is
my
first
meeting,
but
I
was
quite
keen
about
network.
I
was
just
following
the
descriptions
so
welcome.
G
Oh,
oh,
my
gosh,
like
yeah,
I'm
right
now
working
confident,
but
I've
worked
before
like
maybe
as
interesting
for
four
years
and
quite
into
maybe
what
you
can
say
traffic
space
for
in
this,
and
I
just
noticed
that
you're
looking
for
like
you
know
so.
The
work
group
match
discussion
found
quite
interesting
and
I
thought
it
would
be
a
good
start
to
dig
it
because
that's
like,
I
guess,
a
new
thing
starting
off.
But
as
soon
as
you
said,
you're
looking
for
something.
Maybe
I
think
it
will
be
a
good
start.
E
A
I
think
he
was
looking
for
a
way
to
contribute
and
I
think
this
specific
issue
is
mostly
result.
Well,
most
yeah.
It's.
A
As
we
can
take
it,
and
we
need
some
like,
we
need
higher
level
help
from
oh
outside
of
this
project.
For
the
last
step
or
two
yeah.
E
Set
of
people
with
that
one,
but
if
you'd
like
to
contribute,
we
definitely
can
follow
up
with
you.
Catch
us
in
the
sig
gateway
net,
our
gateway
api,
slack
channel
on
kubernetes
slack
after
this
and
just
say,
hey,
I'm
looking
for
something
to
do
and
we'll
we'll
have
a
thread
with
you
about
it.
Sound
good,
perfect,
yeah,
thanks.
D
I've
tried
to
make
a
few.
There
are
a
few
good
first
issues.
I
think
not
many
we've
been
pretty
bad
at
setting
up
issues
for
success
there,
so
yeah.
That
is
definitely
something
that,
like
I
could
probably
spend
some
time
and
sit
down
with
someone
to
go
through
and
find
some
good
first
issues
and
tidy
them
up
a
bit
yeah
yeah.
Definitely
like
we
haven't.
We
haven't.
D
I
like
those
of
us
who
have
been
maintainers,
we've
been
kind
of
so
busy
with
doing
the
maintainer
stuff
that
we
haven't
had
much
time
to
do
the
proper
community
building
stuff.
So
if
anyone
can
help
with
any
of
that,
then
that
would
be
really
appreciated,
probably
the
fastest
way
to
do
that
is
going
to
be
to
get.
You
know,
one
of
us
on
a
call
and
sort
of
walk
through,
like
you
know,
walk
through
some
issues
and
try
and
find
some
ones
that
are
good.
First
issues.
G
Yeah,
I'm
more
than
happy
to
do
that.
Probably
I
will
drop
a
message
in
a
flag
group
and
if
somebody's
free
today,
tomorrow,
whatever
anyway,
I'm
on
calendar
more
than
happy
to
jump
on
a
call,
I'm
sorry
to
test
of
the
meeting
in
between
no
not
at
all,
not.
E
A
Oh
yeah,
this
is
actually
a
reference
to
your
pr,
but
I
wanted
to
make
sure
I
didn't
get
missed.
I
I
love
this
idea
from
you.
We've
gotten
several
support,
requests
on
gateway,
api
and
gateway.
Api
repo
is
not
really
where
I
want
to
be
supporting
implementations.
A
So
one
idea
I
think
this
actually
came
from
shane-
was
in
your
implementation.
Information
actually
link
out
to.
If
you
want
support
with
this
implementation.
Here's
where
you
go.
I
I
think
this
is
great.
I'm
glad
shane
added
it
for
kong.
I
think
shane.
Your
other
idea
was
to
add
something
in
the
issue:
template
itself.
E
E
Merged
now,
okay,
so
I
think
if
you
go,
try
to
create
a
new
issue,
oh
wait
that
one's
not
done.
I
also
updated
it
in
the
documentation
which
tells
you
to
create
an
issue
but
yeah
I
this
this
was
one
I
forgot.
So
thank
you
for
reminding
me,
but
I
think
we
just
wanted
to
kind
of
ask
everybody
like
hey,
since
we
do
seem
to
get
at
least
a
few
issues
where
people
are
like
gateway
api.
This
is
where
I
put
in
my
issue,
and
it's
not
quite
the
right
place.
E
It'd
probably
be
helpful
to
put
stuff
like
this,
like
everybody
go
to
your
implementation
page
and
just
put
something
that
says:
hey
here's
where
you
can
get
support
for
this
implementation,
but
also
it
would
be
kind
of
cool
if
we
could
have
an
issue,
template
kind
of
no
remind
people
like
hey
this,
isn't
for
specific
implementation
support.
If
you
need
that
and
then
link
to
the
implementations
page
so.
H
I'll
update
the
implementation
page
for
console
and
put
our
support
resources
in
there.
H
E
Thank
you
so
yeah,
just
async
whenever
you
guys
get
time,
please
add
those.
I
think
it's
helpful
and
like
we'll
kind
of
get
people
to
the
right
place
quicker.
Does
anybody
else
have
any
last
minute
agenda
items?
We
still
have
nine
minutes.
So
if
there's
any
other
topics
conversation
we
have
some
time.
Otherwise
we
can
go.
Go
home
early.
J
J
E
D
Okay
sure
so
personally,
my
thought
is
my
thought
has
always
been
that
the
layer,
four
resources
to
superiority
in
udp
route
will
end
up
basically
kind
of
replacing
the
surface
type
load,
balancer
use
cases,
so
in
cloud
providers
yeah
that
I
would
expect
that
those
will.
D
Those
will
be
that
those
will
be
implemented
using
the
sort
of
the
network
load
balancers
that
are
that
are
used
at
the
moment
when
you
do
service
type
load,
balancer,
and
that
the
that
yeah,
similarly
for
other
sort
of
when
you
don't
have
a
cloud
provider
available,
it
can
give
you
one
of
those
that
it
will
be
that
something
that's
something
that
would
give
you
a
service
type
load.
D
Balancer
should
be
able
to
do
your
layout,
for
I
know
that
we've
been
asked
a
lot
in
contour
to
do
layer,
4
support
for
using
sort
of
the
envoy
path,
and
it
is
really
not
trivial.
Once
you
have
your
data
path
running
as
a
container
inside
the
cluster
to
do
layer,
four
stuff,
it's
super
hard
because
of
layer,
four
things
so
yeah,
I
guess
yes,
it
would
be
my
short
answer.
Shane,
do
you
have
anything.
E
J
D
I
think
did
you
log,
you:
did
your
logo
use
facebook,
yeah
yeah
and
that
the
use
case
you
were
talking
about
was
like
using
a
like
a
gateway
and
with
disappearance
to
kind
of
describe
the
please
make
me
a
dmz,
like
use
case
roughly
like
if
we
were
talking.
J
Sometimes
because,
when
I
put
in
the
slight
channels
like
we're
trying
to
figure
out
how
to
do
session,
resiliency
and
therefore
right,
so
the
network
load
balances
today,
I
think
in
ews
and
google.
If
this
session
changes,
if
the
pod
dies,
then
the
session
gets
killed.
We
want
to
be
able
to.
We
want
to
make
it
we
wanted.
We
want
to
redirect
it
to
another
pod
and
do
some
stitching
ourselves.
So
it
looks
like
it's
right
to
the
end
user.
J
It's
you
know
continued
on
and
gives
you
more
resilience
both
for
just
for
failures,
but
also
for
scale-in
events,
I.e
this
sort
of
normal
cycle
of
day
it
comes
at
five
o'clock.
You
have,
you
know
four
pods
all
10
to
the
load.
I
I
want
to
kill
a
couple
of
them,
but
I
don't
want
to
go
and
have
customers
go?
Oh
all
my
sessions
died.
Why
did
that
happen?.
D
Yeah
it
sounds
it
yeah.
It
sounds
to
me
like
it'd,
be
hard
to
be
honest,
but
but
I
also
think
that
the
thing
that
I
would
say
there
is
the
layer.
Four
resources
are
severely
understood
at
the
moment
we
haven't
spent
much
time
thinking
about
them.
We
haven't
you
know.
They've
got
some
basic
stuff
in
there
that
we
kind
of
just
like
kind
of
seems
about
right,
but
like
there's,
definitely
that's
why
yeah
that's
why
it
was
really
awesome
with
you
to
look
at
use
case.
D
I
think
there's
definitely
a
lot
of
work.
We've
got
to
do
around
doing
layer,
4
stuff,
but
right
now
we
just
haven't
had
enough
implementations,
doing
lamb4
things
to
sort
of
try
it
out
and
be
like
what
do
we
need?
So
if
you
want
to
build,
if
you
want
to
work
on
an
implementation
that
does
some
cool
layout
for
stuff
like
and
tell
us
about
what
you
need
in
order
to
be
able
to
do
that,
then
awesome,
like
you
know,
I
think
the
the
thing
there
is
to
start
building
it
out.
D
D
D
We
think
it
should
work
like
this
and
we've
done
a
controller,
and
this
is
how
it
works,
and
this
is
what
this
is,
what
happens
and
then
and
then
sort
of
take
that
spec
and
be
like
okay,
I'd
like
to
bring
that
back
to
the
broader
project
and
say:
okay,
please
add
these
trucks
like
like
we
did
in
this
one.
D
Please
see
this
is
our
implementation
and
we've
proved
that
it
works
here,
but,
like
that's
kind
of
always
been
expected
that
people
would
run
ahead
of
like
what
the
official
titles
are
and,
and
one
way
to
do,
that
is
to
just
build
out
your
one.
That
has
the
bits
you
want
and
show
us
how
it
works
and
then
sort
of
come
back
and
say
this
is
what
I
want
to
do,
and
this
is
what
I
did
in
my
one,
and
this
is
how
it
worked
yeah.
J
D
D
Then
I
think
it's
a
hundred
percent
fun
for
you
to
suppose
like
what,
if
we
were
to
have
an
ip
route
but
like
and
the
gateway
then
describes
like
you
know.
If
that
was
the
case,
how
would
that
work?
We
have
talked
a
little
bit
about
that.
I
think
right
now
you
kind
of
get
hung
up
because
port
is
required.
D
You
don't
say
I
remember
those
guys,
yes,
so
that
that's
the
that's
the
probably
the
hang
up
to
having
an
ip
right
now
and
but
if
you
were
to
do
that,
like
you're
like
equally,
it's
okay
for
you
to
sort
of
fork,
the
the
gateway
definition
and
say
like
here's,
how
we
change
the
gateway
definition,
here's
what
we
did
and
here's
what
we
want
to
do
and
here's
why
we
want
to
do
it
like
with
that
with
the
with
having
an
ip
route
and
so
like.
D
Personally,
I
would
love
to
see
like
if
you
have
things
you
want
to
change
about
the
api
like
walk
us
through,
like
what
you
would
change
about
the
api
and
why,
like
and
and
that
that
then
helps
us
all
be
on
the
same
page
about
what
we're
changing.
I
think
one
of
the
things
we
found
with
the
port
discussion
is
like,
oh,
but
how
does
you?
How
does
that
interact
with
what
we're
already
doing
and
stuff
like
that?
So
there's
some.
D
I
think
that
one
of
the
reasons
we
made
the
port
stay
required
for
beta
is
that
changing
a
required
field
to
an
optional
one
is
not
an
ap,
a
breaking
api
change,
but
changing
a
optional
field
to
a
required.
One
is
a
break,
so
it's
easier
to
sort
of
change
that
light
up
that,
then
it's
not.
K
E
We're
running
close
to
time
here
so
I
just
wanted
to
it,
sounds
like
the
the
next
steps
here.
John
sound,
like
you
kind
of
be
trailblazing
in
this,
so
this
might
be
a
place
where,
like
you
know,
we
can
provide
some
support
and
guidance
and
stuff
like
that,
might
be
another
good
place
for
a
discussion,
because
you
could
get
pretty
deep
into
this.
Is
there
anything
else
that,
like
does
that
sound
good?
Does
she
like,
following
up
async,
sound
like
reasonable,
like
the
github
discussion,
is.
J
E
Slack
this
is
everybody's
gonna
have
a
different
opinion
on
this.
My
personal
opinion
is
that
things
disappear
into
slack,
but
github
discussions
are
basically
just
issues
that
don't
need
to
be
in
milestones
like
it's
questions.
It's
you
know
that
kind
of
thing.
So
that
would
be
that.
That's
why
I
suggest
it,
but
I
think
you
know
do
whatever
feels
right
to
you.
D
Yeah,
I
think
a
slack
thread
is
probably
going
to
generate
a
discussion
anyway,
like
I'd.
Give
up
discussion
at
some
point
anyway,
and
I
100
agree
with
shane
the
good
part
about
the
github
discussions
is
they're
there
forever.
So
you
know
it
means
we
don't
end
up
losing
the
context
when
your
slack
scroll
back
kind
of
becomes
hard
to
search.
D
E
Yeah,
thank
you.
Well,
I
think
we're
a
minute
pass.
So
that
concludes
today's
meeting.
Just
a
reminder
that
we're
not
trying
to
do
next
week's
meeting.
If
you
want
to
do
it,
hang
in
the
slack
channel
that
you
want
to
and
we
can
get
something
working.