►
From YouTube: [SIG Network] Gateway API Code Jam for 20230602
Description
[SIG Network] Gateway API Code Jam for 20230602
B
Right
welcome
to
the
Gateway
API
code,
Jam
we're
doing
something
a
little
different
than
the
normal
like
pairing
sessions
that
we
do
today.
This
meeting
is
under
the
kubernetes
code
of
conduct,
which
basically
boils
down
to
be
nice
to
one
another.
So
please
be
nice
to
one
another,
we're
just
kind
of
going
to
do
a
I.
Don't
really
know
how
this
is
going
to
go
exactly,
except
that
the
point
was
there's
a
couple.
B
People
that
have
reached
out
to
me
with,
like
I,
really
would
love
to
get
in
into
the
gateway
API,
but
I
don't
really
know
where
to
start.
So,
let's
just
talk
about
it,
so
it's
just
going
to
be
talking
about
it.
Probably
a
little
bit
of
code
diving
going
through
the
website
very
light,
but
meant
to
be
kind
of
you
know,
lighthearted
fun
interrupt
me
anytime.
Questions
are
probably
going
to
be
the
most
important
part
of
this
little
session.
B
You
know
just
talking
through
questions
and
stuff
like
that
yeah.
So
let's
go
I'll
start
just
by
going
to
the
website.
Actually
I.
Think
again,
I
am
doing
this
like
there's
no
script,
so
I
really
do
want
audience
participation
to
kind
of
drive.
So
please
do
you
guys
see
the
website.
C
C
B
Yeah,
so
I
don't
really
need
to
go
like
read
this
word
for
word
or
anything,
but
I
would
say
if
you
haven't
just
gone
to
the
website
which
you
can
find
from
the
repository.
So
if
you've
made
it
to
the
Gateway
API
repository
it's
right
here
under
about
it
gives
a
pretty
good
overview.
I'm
sure
we
could
tweak
it
and
I'm
sure
it's
a
parts
of
it
might
be
a
little
bit
out
of
date,
but
pretty
good
overview
of
what
we're
trying
to
do
with
Gateway
API
and
so
with.
D
B
From
kubecon
Detroit,
that
was
basically
exactly
that,
and
so
we
can
cover
a
little
bit
of
it,
but
I
think
I'll.
Just
obisman
I
would
be
happy
to
also
share
a
YouTube
link
to
a
talk
that,
like
specifically
focuses
in
on
what
you're
asking
like.
How
does
this
compare
to
Ingress,
but
we'll
do
a
high
level.
So
Ingress
was
this
one
resource,
although
technically
there
was
Ingress
class
later
Ingress
class
would
basically
kind
of
give
you
an
idea
of
like
okay.
B
This
is
my
controller
for
my
ingresses,
so
you
could
have
multiple
controllers
in
a
cluster
and
so
forth.
If
you're
not
sure
what
I
just
said,
don't
worry
too
much
just
come
along
for
the
ride
a
little
bit,
but
basically
one
resource
that
could
Define
some
really
simple,
like
HTTP
routing
rules,
so
that
you
can
get
traffic
to
your
pots,
you
could
use
a
service
as
the
back
end
service
being
the
kubernetes
service.
B
It
had
several
problems,
and
this
is
where
I
feel
like
I'm
almost
going
to
recite
the
talk
I
did
with
Rob
at
Detroit,
but
it
had
several
problems
which
ultimately
resulted
in
Gateway
API,
which
is
kind
of
the
the
follow-up.
If
you
will
to
Ingress,
it
is
role
oriented
in
that
we,
instead
of
having
like
the
the
the
connection
to
in
the
directest
connection
to
Ingress,
is
actually
HTTP
route,
which
is
an
API
which
we'll
go
over
in
a
second
we'll
talk
about
like
the
individual
apis.
B
But
basically
the
idea
is,
you
have
routes
and
you
have
gateways
and
you
attach
your
routes
to
a
Gateway,
and
this
gives
you
different
roles.
So,
like
you,
have
an
infrastructure
provider
making
the
class
which
says
here's
the
operator
or
controller
here's
the
software
that's
responsible
for
handling
the
gateways.
The
Gateway
which
the
cluster
operators
creates
to
say,
I
need
a
Gateway
I'm
going
to
attach
routes
to
it.
Please
and
then
application
developers
can
ship
their
routes
with
their
applications
and
say
this
is
how
you
do
things
for
my
application.
Does
that
all
make
sense?
A
A
The
other
one
that's
probably
worth
pointing
out
is
that
in
especially
in
smaller
organizations,
people
may
not
fall
neatly
into
just
one
of
the
roles
right,
so
you
may
have
people
spanning
roles.
You
may
have
people
only
in
one
role,
but
it
all
works
out.
The
rules
are
still
useful
to
think
about
anyway.
B
Yeah
absolutely
well
said:
Ingress
I
mean
failed,
miserably
is
a
little
harsh,
but
it
wasn't
aware
certainly
about
the
roles
and
now
we
are,
and
we've
had
some
pretty
good
success
with
things
like
what
you
see
here
so
yeah.
A
B
So
just
give
a
little.
A
B
A
A
It
is
an
excellent,
excellent
question:
I
I'm
really
really
hoping
we'll
have
some
constructive
ways.
We
can
talk
about
that
soon
now,
but
but
different
session
I
think
it's.
That
would
be
a
great
session
to
have,
though
you
know
that.
B
New,
which
there's
at
least
a
few
in
here
that
would
this
would
confuse
everything
so
much
so
please
we
do
have
a
thing
called
policy
attachment.
That's
experimental!
Please
don't
worry
about
it.
Yet.
A
Please,
please
don't
worry
about
it
yet.
Yeah.
B
Please
don't
worry
about
it
yet
so
anyway,
I
figured,
it
might
be
a
little
bit
helpful
to
talk
about
like
projects
that
I
work
on
I
won't
go
into
depth
about
them,
but
like,
for
instance,
I
work
on
something
called
the
con
Gateway
operator,
which
is
not
open
source
at
the
moment,
but
we're
intending
for
it
to
be
so
I
could,
at
some
point,
I'll
be
able
to
show
it
to
people,
but
it
basically
was
built
with
only
Gateway
API
in
mind,
so
it
was
kind
of
a
post,
Gateway
API
operator,
whereas
before
we
had
our
Kong
kubernetes
Ingress
controller,
both
of
them
have
support
for
Gateway
API.
B
Sorry
I
have
two
implementations:
there's
a
few
implementations
of
Gateway
API
where
I
work.
There
are
two
that
kind
of
help
me
to
express
what
I'm
trying
to
express
in
terms
of
like
what
the
Gateway
API
can
can
do
for
implementers,
and
that's
that's
the
really
important
thing,
especially
if
you're
coming
into
this
wondering
how
I
can
contribute
we'll
get
to
like
implementations,
are
really
important
to
that.
B
So
I
have
two
implementations
that
I've
worked
on
very
directly
and
one
of
them
is
the
con
Gateway
operator
built
after
Gateway
API,
meaning
it
only
does
Gateway
API
and
it's
it
literally
Provisions
gateways
as
in
proxies.
If
you
want
to
call
them
that
in
response
to
a
Gateway
resource
being
created,
does
that
make
sense.
B
But
these
eight,
in
addition
to
just
this,
is
to
try
to
help
just
make
it
more
clear
why
we
have
these
rule
orientation
like
we
literally
have
something
that
just
handles
gateways
in
our
code.
Right
just
works
on
like
let
me
deploy
my
Gateway
with
these
settings
and
stuff
like
that
and
another
thing.
Another
controller,
if
you
will,
which
is
just
a
fancy
word
for
saying
another
loop,
that's
sitting
there
working
on
other
things,
it's
focusing
on
the
route
part
and
attaching
them
to
the
gateways
so
programming,
the
data
plane.
B
B
But
it
works
really
well
for
us,
because
it
basically
the
model,
fell
right
over
what
we
were
already
doing
with
our
customers,
so
that
is
kind
of
part
of
what
you
get
at
a
very
high
level
compared
to
Ingress,
whereas
with
Ingress
you
would
have
in
our
in
our
stuff.
We
would
have
originally
like
to
helm
deployment
of
the
proxy.
We
still
do,
there's
the
helm
deployment
of
the
proxy
you
manage
everything
by
hand.
The
life
cycle
of
the
Gateway
is
managed
entirely
by
the
person.
B
The
human
operator,
whereas
having
these
apis
as
apis
like
having
these
distinctions
in
these
roles,
makes
it
easier
to
model
kubernetes
operators
that
is
autopilots
for
handling
these
components
and
their
life
cycle.
That
is
from
begin
from
the
point
where
they're
created
to
all
the
things
that
they're
configured
with
to
when
they
have
bad
days
and
good
days,
all
managed,
at
least
in
part
by
some
autopilot
software.
Does
that.
B
Bad
days
like
some
when
I
said
bad
yeah,
that's
funny.
No,
when
I
said
bad
days,
I
was
thinking
more
like
it's
there
for
the
ride.
When
things
like
you
know,
the
traffic
gets
so
heavy
for
the
data
plane
that
it
goes
down.
Your
autopilot
system
using
these
apis
may
be
maybe
capable
of
healing
that
and
stuff
like
that.
A
B
A
It's
it's
a
little
different
in
terms
of
some
of
the
focus
of
things,
and
it
can
understanding
that
that's
there
can
help
understand
some
of
the
complexities.
When
you
look
at
some
of
the
code,
Gateway
API
is
for
north-south
traffic.
East-West
traffic
might
be
handled
by
kubeproxy
or
it
might
be
handled
by
a
service
mesh
or
it
might
be
handled
by.
A
You
know
a
variety
of
other
things,
but
the
distinction
that
the
Gateway
API
was
originally
for
north
south,
not
East,
West,
is
important,
and
then
we
have
to
talk
about
Gamma,
where
people
are
trying
to
figure
out
how
to
make
the
Gateway
API
also
be
able
to
configure
East-West
traffic.
But
that
also
is
probably.
A
The
two
of
us,
for
example,
yeah,
yeah
and.
B
B
Actually
it's
in
here,
if
you're,
if
you
are
interested
more
in
like
mesh
traffic
East-West
traffic,
there
is
a
page
for
gamma
I,
don't
think
we're
going
to
dive
too
deep
into
that,
but
basically
it's
an
experimental
sub-project
of
the
the
Gateway
API
project
to
use
things
like
HTTP
route,
East
and
West
in
inside
the
cluster
between
applications,
as
opposed
to
Ingress
into
the
application
from
outside
the
cluster.
E
Yeah
I
had
one
question
so
is
so
is
the
Gateway
API
built
so
so
from?
What
I
can
understand
is
that
the
Gateway
API
Works
separately
from
the
Ingress
not
built
on
top
of
the
Ingress
correct.
B
It's
it's
built
on
the
I.
It's
built
like
in
spirit
on
Ingress,
like
a
lot
of
some
of
the
same
people
were
involved
in
stuff,
like
that,
it's
it's
those
of
us
that
did
Ingress
for
many
years,
built
Gateway
API.
It
is
you'll,
find
this
and
I'll
send
you
a
link
to
rob
in
my
talk
from
Detroit.
But
basically
we
are
at
a
point
where
Ingress
is
not
going
to
get
any
new
features.
B
All
the
new
features
will
go
to
Gateway
API,
and
that
means
Gateway
HTTP
route
to
the
apis
within
the
Gateway
API
project.
So
I
think
that's
just
it's
it's
in
that
way.
It's
kind
of
its
successor,
but
yeah
anyway,
good
question.
F
I
I
also
have
a
question
like
what's
the
like,
once
we
come
to
1.0
so,
do
we
encourage
people
to
start
using
it
for
production
and
stuff,
because
yeah.
F
B
Yeah
I
mean
we
have
plenty
of
people
using
it
in
production,
I
mean
I,
know
for
a
fact,
Google
does
and
so
what?
What
V1,
which
we're
hoping
for
to
do
for
coupon
Chicago,
so
we're
hoping
to
do
it
in
October
before
November
should
be,
is
just
well
back
to
the
overview.
Actually
just
the
parts
you
see
here,
Gateway,
class,
Gateway
and
HTTP
route
as
GA,
meaning
backwards
compatibility
will
be
preserved.
We
are
ready.
B
We
are
saying
this
is
usable
in
production
now,
but
as
demand
was
pointing
out,
there
is
more
than
just
HTTP
route,
so
the
Gateway
is
what
you
attach
routes
to,
but
we
have
routes
of
type
HTTP
route,
TLS
route,
TCP
and
UDP
route.
We
have
a
grpc
route.
Now,
that's
actually
moving
along
pretty
quickly
and
the
sky's
the
limit
we
could
add
more
in
the
future
if,
if
the
need
shows
up-
but
this
is
pretty
good
coverage
of
a
lot
of
what
people
are
doing
in
a
kubernetes
network
today-
so
that's
all
there.
B
The
things
that
are
not
HTTP
route
on
this
table
are
mostly
in
Alpha,
I,
think
they're,
all
in
Alpha
State
and
like
on
their
way
to
Beta,
whereas
HTTP
route
is
in
beta
and
on
its
way
to
GA
it's
the
main
use
case.
It's
it's
the
most
used
route,
so
it's
kind
of
driven
forward.
Does
that
make
sense,
and
does
that
answer
your
question.
C
A
Yes,
okay,
you
Shane,
does
http
route
already
cover
HTTP
3
or
not?
Do
you
remember
I.
D
B
B
Let's
see
probably
good
to
just
look
at
some
stuff,
real,
quick,
let's
like
look
at
these
actual
apis.
So
like
here's
some
examples,
a
Gateway
class
is
a
very
simple
resource
and
yeah.
Let's
just
let's
go
into
breaking
it
down,
but
do
stop
me
if
there's
like
things,
you
want
to
focus
on
so
we
talked
about
all
the
rules
and
we
talked
about
the
apis
by
name.
Let's
just
look
at
them,
real,
quick
and
then
we'll
kind
of
we'll
kind
of
move
on
Gateway
class
is
pretty
much
just
a
controller.
B
It's
the
most
of
them.
Are
this
simple?
Although
there
are
a
couple
other
fields
you
can
fill
in,
this
is
just
a
resource.
That's
cluster
scoped.
That
kind
of
says
this
is
my
controller
by
name
so
like
example.net,
my
Gateway
controller,
that's
responsible
for
gateways
attached
to
this
and
then
I
don't
care
about
all
this
stuff
right
now.
Oh
sorry,
Gateway!
B
So
then
a
Gateway!
Why
don't?
We
have
examples
in
here?
Where's
all
the
examples,
simple
Gateway
here
we
go.
Okay,
so
gateway
then
attaches
to
that
Gateway
class
that
we
looked
at
and
says.
Okay
for
the
controller,
that's
responsible
for
me
here
is
a
gateway.
I
need
you
to
provision
for
me
if
you
will
and
I
need
it
to
have
these
listeners,
which
in
this
case
is
just
a
basic
HTTP,
Port
80
listener
that
needs
to
be
listening
on
that
Port.
Does
that
make
sense?
B
So
now
we're
at
the
point
where
Ingress
used
to
kind
of
Step,
In,
Like
Ingress
couldn't
do
any
of
this
and
we
can
go
down
to
the
the
route
level,
so
an
HTTP
route
then
can
attach,
as
the
Gateway
attached
to
the
Gateway
class,
to
say
this
is
the
controller
responsible
for
me,
the
ACP
route
attaches
to
the
Gateway
to
say
this
is
where
the
data
plane,
where
my
configuration
needs
to
be
pushed
to.
If
you
will-
and
this
one
says-
okay
for
that
Gateway
I'll-
take
this
hostname.
B
Anything
from
this
host
name
needs
to
go
to
Port
80
and
to
this
service.
B
That's
a
kubernetes
service
and
here's
where
some
things
kind
of
get
interesting,
so
that
there's
a
lot
more
I,
don't
think
we
necessarily
need
to
dive
into
it
right
here,
but
we
can,
if
there's
questions
but
there's
a
lot
more
rules
than
Ingress
had
a
lot
of
different
ways
in
which
you
can
kind
of
shape
your
traffic
and
get
things
to
like
certain
back-ends,
which
you
should
dive
into
in
the
documentation
in
the
API
and
stuff
like
that,
if
you're
interested
there's
also,
obviously
this
attachment
to
the
actual
Gateway,
which
is
new,
so
that's
that's
very
different
from
HTTP
route,
but
in
that
in
that
video
I'll
send
you
there's
basically
a
one-to-one
comparison
that
comes
up
in
terms
of
Ingress
versus
HTTP.
D
D
Yeah,
so
just
a
quick
one,
really
I
suppose
we
have
parent,
refs
and
back
end
refs.
Are
they
ever
going
to
be
anything
other
than
a
Gateway
reference
and
a
service
reference.
B
Maybe
I
mean
the
idea
is
that
we
learned
some.
We
learned
lessons
about
like,
for
instance,
with
Ingress.
You
pretty
much
just
point
that
as
a
service
right,
there
are
potentially
use
cases
out
there
for
pointing
it
at
something
else,
and
it's
really
better
for
us
to
leave
that
open-ended,
so
that
people
can
have
implementations
that
aren't
like
somehow
locked
into
service.
If
there's
some
other
thing
that
they
want
to
point
out.
One
thing
that
concretely
for
a
back
end
draft
for
HTTP
route
is
done.
Today
is
service
Imports.
B
So
if
you're
doing
multi-cluster
services-
and
we
actually
have-
we
have
documentation
on
this
in
this
website-
I
won't
get
into
this
too
much
because
again,
kind
of
like
net
policy
attachment,
multi-cluster
Services
is
like
way
too
much
for
what
we're
trying
to
do
here.
But
if
you
have
multiple
clusters,
service
Imports
basically
collect
all
the
back
ends
across
those
clusters,
and
you
can
point
an
HTTP
route
at
that
and
we
have
documentation
which
says
how
to
do
that
and
it's
pretty
much
the
same
thing
as
service,
but
with
a
service
import.
So,
yes,.
A
And
one
concrete
example:
we
also
have
the
concrete
example
that,
although
this
is
not
yet
part
of
the
standard,
the
gamma
initiative
is
allowing
a
back-end
or
sorry
a
parent
drift
to
be
a
service
as
well,
so
right,
yeah.
These
are
both
things
that
can
evolve
over
time
and
that
we
expect
to
evolve
over
time.
B
D
B
B
We
do
one
that
does
open
resty,
which
is,
if
you
don't
know
it,
it's
basically
a
wrap
around
nginx,
and
then
we
have
another
one
that
does
envoy,
which
is
a
very
popular
proxy
as
well,
and
for
any
of
these,
like
h8
proxy
Google,
does
I
think
their
implementation
is
over
their
their
Network
load,
balancers
in
the
cloud
provider
and
so
forth.
So
it
can
really
be
it's
abstract
so
that
you
can
really
have
anything.
C
B
Yeah,
you
can
have
anything
you
want
underneath
it
and
we
have
to
help
make
this
work.
We
have
a
model
of
conformance
support
levels,
I'm
I'm,
so
glad
I
was
able
to
find
the
page
right
when
I
was
thinking
of
this,
but
basically
for
those
20
plus
implementations.
It
might
feel
kind
of
hard
like
how
do
all
of
these
implement
the
same
API
we
created
leveled
tiered
support,
if
you
will
tiered,
is
probably
not
the
right
actual
word,
but
it's
it's
kind
of
like
that.
In
that
core
is
every
implementation
can
do
this.
A
B
B
Okay,
that's
okay,
like
I,
said:
I
want
audience
participation
for
sure
there
are
tests
that
test
this,
so
that
you
can
say:
I
am
conformant
and
we're
working
on
more
stuff
related
to
this
testing,
which
is
probably
not
worth
talking
about
too
much
here.
Unless
you
have
questions
about
it
basically
core
is
everything
can
do
it
extended?
Is
some
can
do
it
an
implementation
specific?
Is
we
are
aware
that
people
have
something
they
want
to
do
here?
We
don't
really
have
any
tests
or
anything
for
it.
B
We
just
have
it
here
for
awareness
and
maybe
one
day
it
graduates
to
extend
it.
If
a
few
people
are
able
to
do
it
and
we
can
start
coming
up
with
some
tests
for
it
and
along
this
path,
basically
from
bottom
to
top
things
can
graduate.
We
can
get
to
a
point
where
somebody
implements
something
implementation.
Specific
theoretically
can
go
to
extended,
like
I
said
and
in
theory,
if
we
get
to
a
point
where
every
implementation,
Checks
In
and
says
okay,
this
one
extended
feature.
B
We
can
all
do
it
now
that
could
become
core
might
take
time,
but
these
things
we
left
it
open
so
that
that
could
happen,
and
we
actually
like
I,
said
I
won't
dive
into
it
too
deep.
B
But
we
have
upcoming
work
right
before
GA
that
I'm
working
on
with
a
couple
other
people
in
the
community
called
conformance
profiles
which
will
help
to
identify
these
things
better
and
Report
these
There's
an
actual
Reporting
System,
where
you
send
reports
back
to
us,
and
then
we
have
an
index
if
you
will
of
who's
actually
doing
what
that
we
can
search.
Whereas
currently
we
query
slack
for
that
today
and
it
sucks
very.
A
Has
a
question
on
slack
that
ties
in
pretty
directly
to
the
conformance
profile.
Word
sorry
question
on
chat.
Zoom
chat,
not
slack.
That
ties
pretty
directly
into
the
conformance
profiles
about
The
L4
project,
trying
to
pick
up
Gateway
API
at
some
point.
If
we
see
a
lot
of
interest
in
that,
I
would
expect
that
we
would
end
up
with
a
layer,
4
conformance
profile
for
exactly
the
reason
you
just
said:
it's
not
possible
for
an
L4
gateway
to
support
HTTP
route.
A
So
if
enough
people
want
to
do
that,
we
would
need
to
do
a
profile
that
says
you're
conformant
with
The
L4
profile.
If
you
do
TCP
route
and
UDP
route
and
nothing
else
meshes
have
a
similar
problem.
Where
asking
a
mesh
to
support
the
full
Ingress
role
doesn't
really
work
out,
and
so
there's
a
mesh
conformance
profile.
That's
under
active
developments
to
make
that
sort
of
Distinction
work.
B
B
We
can
talk
about
that.
If
we
have
time
for
that,
because
I
L4
has
been
a
big
important
thing
for
me,
and
I'm
I
currently
have
a
branch
where
I'm
working
on
the
conformance,
the
initial
conformance
tests
for
TCP
route
UDP
route.
B
B
Well,
the
in
the
face
of
what's
going
on
with
Gateway
API
today
with
layer,
7
being
the
only
thing
going
to
GA
layer,
7
stuff
gets
the
higher
priority
generally,
but
we'll
we'll
the
work
will
be
done.
It's
it's
underway
and
we'll
get
to
that
point.
Hopefully
it
didn't
make
that
too
complicated.
Does
that
make
sense
that
we
have
many
many
implementations.
We
have
different
levels
of
support
for
like
the
fields.
If
I
wasn't
clear
about
that
like
in
HTTP
throughout
there
are
fields
that
are
core.
Every
implementation
can
do
it.
B
There
are
fields
that
are
extended.
You
may
deploy
an
implementation
like,
let's
say
Kong,
for
instance,
where,
like
it,
may
not
support
that,
and
so
it's
marked
as
such,
like
just
so,
you
know
this
may
not
be
supported
enough
implementation.
That
makes
sense.
Yes,.
C
B
Cool
moving
right
along
we've
talked
for
a
half
hour
and
kind
of
just
gone
over
like
topical
kind
of
important,
topical
stuff
for
Gateway
API.
Are
there
any?
Quite
let's
stop
for
questions.
You
got
any
questions
like
you,
have
any
directions.
You
kind
of
want
to
go
from
here.
F
B
That's
a
good,
that's
a
good
segue,
so
we
just
talked
about
all
these
implementations
and
that
they
all
use
these
apis.
So
let's
talk
about
the
apis.
Our
repository
is
mostly
what
lior
just
said
like
it's,
mostly
just
the
apis
like
it's.
We
don't
have
an
implementation
in
this
repository.
If
that
wasn't
clear
before
every
we,
we
only
have
what
we'll
we'll
call
third-party
implementations,
but
there's
one
caveat,
which
is
the
thing
I
potentially
want
to
get
to.
If
we
have
time
so,
let's
look
at
HTTP
route
types,
for
instance.
B
This
is
pretty
indicative
of
what
you'll
find
in
the
code
base
the
API
for
HTTP
route,
the
specification
we
just
looked
at
one
that
had
what
was
it
hostname
right
so,
like
here's,
the
host
names
field
that
we
were
just
looking
at,
where
it
says,
example.com
the
rules.
So
it's
pretty
much
that
an
HTTP
route
is
pretty
much
the
common
route.
Spec
parent
ref
like
where
you?
What
Gateway
are
you
putting
this
route
on
host
names
and
rules?
B
So
this
is
a
lot
of
what
the
code
base
is
and
that's
what
you'll
find
under
the
and
I
guess.
We
can
just
go
right
into
code,
diving
because
I
think
that's
what
you
want
it
originally
to
mount.
That's
what
you'll
find
in
apis
is
the
actual
apis
that
eventually
become
those
yaml
Declarations
of
HTTP
route
and
Gateway
and
TCP
route,
and
all
that
that
you
saw.
C
B
We
use
Cube
Builder
and
we
use
Cube
Builders
I
love
their
their
their
annotations
for,
like
validation
and
stuff,
like
that,
it's
really
helpful
admission
is
just
our
admission
web
hook
that
one's
kind
of
squirrely,
if
you
want
to
get
in
there
and
dig
into
that
we'd
appreciate
it.
Config
is
our
actual,
like
crds
and
yaml,
basically
that
you
can
actually
deploy.
B
Release
channels,
we
also
have
release
channels.
So
all
these
support
levels
are
things
that,
like
our
support,
these
are
supported
things
at
these
levels.
Release
channels
is
how
we
get
a
standard
channel
of
like
these
are
the
standard
crds
versus
how
we
are
able
to
kind
of
like
host
some
experimental
things.
B
You
can
have
things
in
an
experimental
release,
Channel
at
any
of
these
levels
or
a
standard
release
channel
to
any
of
these
levels
technically,
but
experimental
is
basically
where
the
newer
things
we're
not
sure
are
going
to
continue
on
go
and
in
the
code
base.
You'll
see
that
here,
because
this
is
where
we
emit
our
experimental
crds
with
the
extra
fields
that
are
experimental
versus
standard
and,
generally
speaking,
in
production,
people
are
using
these
ones.
B
The
standard
ones
and
you'll
see
there's
a
lot
less
of
them
here,
because
things
like
TCP
routing,
UDP
route,
because
they're
Alpha
are
considered
experimental
today.
That
will
probably
be
changing
in
the
next
few
months.
Sorry,
next
few
months
is
probably
pushing
it
a
little
bit,
but
it
should
be
changing
relatively
soon
after
ga,
okay.
B
F
Shane
yeah
one
thing
I
want
to
ask
is
like
theory:
if
something
goes
from
experimental
to
standard,
so
wouldn't
it
be
reasonable
to
say
like,
wouldn't
be
reasonable
to
say
that
it
must
be
like
it
should
be
extended
or
core
like.
Why
would
something
be
implementation
specific
and
be
standard.
B
Yeah,
so
implementation
specific
is
kind
of
special
in
that
it's
it's
just
a
level
that
exists
to
point
out
that
something
may
exist,
and
we
do
occasion
I
think
we
have
a
couple
of
fields
for
that,
but
it's
kind
of
like
No,
Man's,
Land,
more
or
less,
but
if
it's
in
standard
and
implementation
specific,
that
means
it's
something
we're
aware
of
people
can
do
and
we're
we're
comfortable
enough
with
it
that
we
know
that
it's
just
part
of
the
package
like
it
should
be
part
of
the
package.
It's
not
an
experiment.
B
F
A
Pretty
much
yeah
if
you
can
write
a
conformance
test
for
it,
it
implies
that
is
capable
of
doing
it
in
the
same
way,
so
it
can
be
extended.
But
a
good
example
of
this
one
is
timeouts
our
perennial
example
of
something
that
should
be
simple
and
is
not
where.
If
you
go
look
over
the
existing
implementations
of
the
Gateway
API,
you
will
find
that
there
isn't
a
single
timeout
that
all
of
them
actually
agree
on
in
terms
of
where
it
begins
and
where
it
ends.
So
writing
a
conformance
test
for
that.
A
You
know,
there's
a
timeouts
enhancement
proposal
now
that
I
personally
think
is
pretty
good,
but
it
acknowledges
the
fact
that
yeah,
you
know,
there's
going
to
be
some
there's
going
to
be
some
variation
between
implementations,
which
is
okay.
In
this
particular
case,
you
can
still
use
timeouts
in
a
meaningful
fashion,
even
if
trying
to
go
through
and
write
a
conformant
test
that
requires
multiple
implementations
to
be
the
same
down
to
the
milliseconds
is
really
hard
right.
B
F
You
could
ask
for
a
timeout,
you
know
input
like
you
can
get
a
demo
but
yeah
like
I
I,
recall,
I,
I
think
the
HP
pass
as
well
was
implementation
specific,
but
there
are
some
conformance
tests
that
I
I
written
and
they
are
excellent
patients
specific
so
I
didn't
I
understand
this.
Like.
D
F
B
A
Mean
direct
seems
like
something
that
people
ought
to
generally
agree
on,
so
that.
B
F
B
F
B
Can
do
that
yes,
I
can
I
can
do
that
on
your
behalf.
I
can
I
can
kind
of
bring
it
to
the
to
the
table
for
you
if
you'd
like
so,
please
do
put
it
on
the
agenda
and
put
our
names
on
it
and
I'll
know
to
take
that
on.
So
what
Leora
is
talking
about
goes
directly
into
conformance,
where
this
is
all
the
tests,
basically,
that
somebody
can
run
this
test
Suite.
A
Me
these
conformance
tests
and
I'm
being
I,
am
being
slightly
facetious
here,
but
I'm
only
being
slightly
facetious,
because
I
just
tried
to
I
tried
this
recently
and
literally
could
not
get
this
to
work,
and
that
seems
odd,
because
I'm
pretty
confident
this
runs
in
CI
all
the
time
right.
Yeah.
B
Well,
so
I
have
a
little
bit
of
a
bias.
I
don't
use
the
CLI
in
the
Pro
in
the
two
projects
that
I
help
that
I
maintain.
We
just
use
it
as
go
library
and
I
can
show
you
one
of
those.
B
It
yep
so
and
I'm
sorry
I'm,
sorry
for
that
bias,
because
it's
not
really
fair
to
the
question
or
anything,
but
that
is
I
don't
mess
with
the
CLI,
but
under
in
the
kubernetes
Ingress
controller
for
Kong
we
have
test
conformance
tests
and
basically
you
can
find
I
think
it's
in
here.
It's
been
a
minute
since
I've
been
in
here,
but.
F
Until
you
find
it,
I
can
elaborate
on
the
original
question
if
it's
okay
but
yeah
recently,
yeah
recently
just
explained
how
to
run
it
with
the
CLI
as
well.
So
basically
you
just
need
to
do
that,
like
you
need
to
have
a
cluster.
What
do
you
want
to
like
you
wanted
to
test
it
against
your
your
implementation?
F
So
you
do
it
like
you
better,
do
it
with
with
the
go
library
and
integrated
in
CI,
but
if
you
just
want
to
like
start
experimenting
so
you're
essentially
gonna
run
the
tests
as
go
tests
with
some
parameters,
while
you
are
in
the
Gateway
API
Fork
so
like
I
can
elaborate
it
later.
If
we
have
time
or
like
even
offline
but
yeah,
essentially.
F
C
F
A
Actually,
okay
with
using
Envoy
Gateway,
as
you
know,
as
a
vehicle,
because
I'm
involved
with
that
project
as
well
and
like
I,
said
I,
have
to
go
and
talk
to
talk
to
some
of
those
folks
about
places
where
it
looked
like.
I
was
seeing
tests
fail,
but
there
are
two
things
that
I
think
are
relevant
here.
One
is
well
yeah.
A
We
we
probably
should
have
a
readme
in
the
conformance
directory
and
I
would
volunteer
to
write
that,
except
for
the
fact
that
I
didn't
actually
get
it
working,
so
maybe
I'll
get
it
working
and
then
volunteer
to
write
it.
And
the
other
thing
that
comes
to
mind
is
I
kind
of
wonder.
If
we
could.
B
We
do
that
that
that's
that's
I'll,
be
getting
to
that.
That's,
okay,
inbound
yeah!
That's
inbound!
Because
of
this
kind
of
problem.
Okay,.
A
B
Means
so.
A
B
I
appreciate
that
so
just
to
give
you
an
idea
of
what
ours
is
doing,
it's
there's
a
lot
of
stuff
in
here,
but
basically
long
and
short,
we
have
a
test
harness
that
allows
you
to
take
add-ons
and
apply
them
to
a
kubernetes
cluster.
So
we
just
say
we
want
metal,
B
and
Kong.
We
build
the
environment
and
then
I
think
it's
I
think
it's
just
build
yeah.
B
Just
we
build
the
environment,
which
then
creates
a
kind
cluster
drops
these
add-ons
into
it
and
then,
ultimately,
back
over
here,
we
feed
that
to
sweets
which
in
this
case
is
actually
the
Gateway
conformance
utils
suite
and
we
say:
here's
our
client,
which
points
to
that
cluster.
We
just
created
here's
our
Gateway
class
name,
so
you
know
what
controller
we
belong
to
and
then
all
the
other
features
this
one's
kind
of
interesting.
This
one's
actually
going
to
change
a
lot
or
has
already
changed
in
a
branch
somewhere.
B
But
basically
you
can
see
here
the
extended.
We
skip
some
tests
right
now.
This
will
be
different
when
conformance
profiles
is
done,
but
we
we
skip
like
extended
tests
that
we
don't
actually
do
and
then
experimental
conformance
that
we
don't
care
about.
Yet
we
actually
technically
have
one
thing,
that's
standard
and
core
that
is
in
here,
but
this
is
actually
fixed,
but
just
to
give
you
an
idea
of
what
the
kind
of
things
you
can
do
with
this
test,
Suite
like
you,
can
configure
everything
run.
B
It
run
the
setup
actually
which,
like
deploys
gateways
and
stuff
I,
wish
this
there's
there's
some
rules
under
which
github's
code
reviewer
can
like
send
you
to
the
thing,
but
I
can't
I,
don't
think
I
can
do
it
with
libraries
and
stuff
I,
don't
know,
I,
don't
spend
money
time
in
this
thing,
I'm,
usually
in
them,
and
then
ultimately
run
the
test
suite
and
get
the
result
in
the
future.
That
will
have
potential
reporting
added
to
it.
B
So
yeah,
sorry
that
it's
not
the
greatest
experience
I
will
be
happy
to
catch
up
with
you
afterwards
as
well
to
try
to
help
and
to
shorten
the
time
from
now
until
when
you
get
something
working.
So
please,
let's
do
that,
though,
but
because
we
have
a
little
over
10
minutes
and
I
do
want
to
get
to
a
couple
of
the
the
floating
topics
that
I've
been
pushing
back
on.
Let's
keep
moving,
there's
some
stuff
in
here
for,
like
container
bills,
there's
documentation
examples
of
resources
geps.
B
B
A
A
B
Package
is
full
of
some
like
go
client
Library,
stuff
site,
Source
I
think
is
just
like
where
the
actual
like
HTML
for
the
site
is
emitted
and
tools.
I,
don't
remember
what
we
put
in
here:
oh
okay,
well,
okay,
don't
I,
don't
know
so
that.
B
That
one
exists,
but
anyway
so
yeah,
that's
that's
kind
of
long
and
short
of
what's
going
on
in
the
code
base,
but
let's
take
a
let's
round
it
up:
real,
quick,
let's,
let's,
let's,
let's
collect
ourselves,
it's
an
ape.
This
is,
for
the
most
part,
just
the
API
and
the
tests.
The
implementations
have
to
plug
into
this,
or
this
is
nothing
it
doesn't.
B
Do
anything
I
want
to
make
that
abundantly
clear,
because
that's
important
one
of
the
main
reasons
why
we're
even
having
this
agenda
topic
today,
which
is
I,
would
imagine
Damon
and
abysman.
You
both
have
like,
potentially
an
interest
in
contributing
which
we
appreciate
it's
kind
of
hard
to
contribute.
Okay,
it's
nearly
impossible
to
contribute
without
an
implementation
that
you're
also
kind
of
working
with.
B
So
this
has
been
a
problem
for
a
while
I
guess:
I
should
I
should
say:
let's
go
back
to
contributing
here.
Where's
the
I
always
lose
this.
If
there's
an
implementation.
B
A
B
B
It's
it's
very
on
the
edge
anyway.
Suffice
to
say
that
was
just
to
point
out
that,
like
we
took
a
different
approach
intentionally
here
by
not
having
a
reference
implementation
or
a
default
implementation,
because
that
is
a
maintenance
nightmare
to
be
to
be
supporting
a
production,
the
ape
not
only
the
API,
but
the
production
implementation
of
the
API
that
people
are
using
from
the
perspective
of
volunteers,
working
on
an
open
source
community-led
project
is
nightmarish.
B
So
I
am
go.
I
am
always
very
careful
to
not
provide
biases
if
I
can
avoid
it.
I
obviously
have
biases,
but
go
through
this
list
find
the
one
that
speaks
to.
You
is
basically
what
I
tell
people
more
or
less
like
find
like
whatever
you
know,
because
I
I
obviously
have
no
bias
here
like
if
you've
worked
with
aha
proxy
before
go
there,
you
know.
Do
that
that
kind
of
thing
make
a
decision
for
what's
best
for
you
in
terms
of
getting
attached
to
an
implementation.
B
We
do,
on
the
other
hand,
technically
have
an
upcoming.
Because
of
this
problem.
We
have
an
upcoming
project
which
is
currently
at
Kong
disclaimer
but
being
and
there's
a
there's,
a
discussion
for
this
being
donated
to
the
cncf
called
blixt
and
I'd
like
to
talk
about
it
just
for
a
minute,
but
it
is
an
implementation
of
Gateway
API
that
only
does
L4.
B
B
And
the
intention
with
this
originally
so
at
Kong,
this
just
started
as
pure
experiment.
This
used
to
this
thing
had
a
big
logo
on
it.
That
said,
just
it
still
has
something
like
that,
but
it
originally
started
with
just
don't
we're
not
using
it
for
anything
we're
just
playing
with
it
as
we're
donating
it
that
what
we
decided
in
Gateway
API,
that
is
the
maintainers
decided,
is
that
this
would
be
a
good
tool
to
provide
a
reference
implementation
if
we
never
if
it
never
ends
up
in
production.
C
B
B
No
we've
the
question
of
layer.
Seven
remains
open,
but
the
important
part
is
that
this
implementation,
the
rules
that
we
had
for
whether
we
would
have
finally
have
a
reference
implementation
is
it
has
to
it-
can't
be
trying
like
yeah.
It
cannot
step
on
anybody's
toes
would
be
like
we
technically
Envoy
is
cncf,
but
we
made
the
decision
that
we
would
not
use
Envoy
because
it
there's
kind
of
a
I.
Don't
know
if
this
is
the
right
word,
but
kind
of
a
tribalism
to
that
right.
It
kind
of
plays
a
favorite.
B
At
minimum,
it's
a
conflict
of
interest,
so
our
data
plane
is
Linux,
no
conflict
of
interest,
everybody
everybody's
happy
with
Linux
being
on
the
data
plane
and
then
the
control
plane
is
Gateway
API.
So
this
worked
for
us.
We
could
have
made.
We
might
they're,
probably
some
other
things
we
could
have
chosen,
but
ebpf
was
fun
at
the
time
too,
and
this
has
been
a.
This
project
has
been
around
to
be
fun
as
well.
B
There
was
a
there
was
a
certain
amount
of
Spirit
of
like
just
kind
of
building
something
fun
together
that
went
into
this
project.
Let.
B
Out
these
yeah
yeah
you
could
you
could,
but
the
that
that
was
another
option,
but
we
felt
ebpf
is
more
fun,
so
the
intention
is
never
in
production.
We
literally
talked
about
having
a
kill,
switch
that
would
like
shut
it
down
over
time
and
say
nope.
You
cannot
run
this
in
production,
don't
you
dare
yep
and
then
it's
meant
to
also
be
the
first
like
we
don't
current
because
of
this
problem
of
like
implementations,
kind
of
having
to
be
these
third-party
things.
B
We've
never
had
on
PRS,
where,
like
the
CI,
actually
runs
the
conformance
tests,
we
always
have
to
wait
for
feedback
from
people
going
off
and
trying
them
or-
and
this
has
happened
quite
a
bit-
we
produce
some
conformance
tests
because
they
don't
run
actually
when
we
merge
the
pr
have
broken
the
the
conformance
test.
So
this
is
supposed
to
help
a
little
bit
with
that
and
if
we
can
get
a
layer
7
option
that
is
not
a
conflict
of
interest,
then
we
could
potentially
do
that
for
that
too.
B
But
again,
so
this
is
kind
of
our
answer
to
we
want
people
to
be
able
to
jump
in
and
like
help
with
something
we'd
never
want
to
in
production,
and
we
want
to
kind
of
like
be
able
to
provide
a
reference
for
people
and
we
want
to
run
something
in
CI.
This
came
in
together
at
all
those
things,
big
caveat,
I
would
say,
because
it
I
am
basically
saying
please
do
feel
free
to
come
and
join.
This
project
is
feel
free
to
wait
until
this
is
actually
ported
over
to
kubernetes.
B
A
Can
talk
a
little
bit
more
about
that
offline?
It's
I
ended
up
having
to
talk
with
an
open
source,
a
colleague
of
mine,
who
is
also
really
really
rapidly
pro-open
source,
and
so
we
ended
up
talking
about
this
at
one
point
and
it's
kind
of
an
interesting
story
that
is
both
really
confusing
and
also
ends
up
making
a
fair
amount
of
sense.
So
yeah,
okay,
I'd.
A
Don't
know
what's
going
on,
it
does
make
me
wonder
a
lot
about
you
know
so.
Yeah
to
Dan's,
Point
I
know
for
a
fact
that
this
is
impacting
istio
and
psyllium
and
blixt.
You
know
at.
B
A
I
mean
it's
as
I
understand
it.
It
is
the
thing
that
is
in
the
way
of
istio
and
psyllium
being
able
to
graduate,
for
example,
and
yeah
I
mean
don't
get
me
wrong.
I'm
happy
to
have
fewer
graduated
service
meshes
wearing
my
buoyant
hat,
but
actually
yeah.
This
needs
to
be
dealt
with.
It's
a
problem.
Yep.
B
F
So
one
one
thing
I
wanted
to
call
out
as
well
is
like
it
depends
if
you
want
to
contribute
to
like
just
you
know,
more
conformance
tests,
so
you'd
probably
want
to
choose
an
implementation
that
it's
more.
You
know
that
have
more
features.
B
B
Yeah
that's
different,
but
yeah.
If
you
go
pick
a
different
implementation,
if
you
want
to
work
on
L7,
if
you
just
want
to
kind
of
start
messing
with
the
waters,
this
might
be
a
good
place.
I
I
totally
understand
for
people
that
want
to
wait
until
it's
actually
migrated
to
the
kubernetes
six
org,
but
yeah
the
it's
a
good
point
and
I
definitely
encourage
you
to
go
find
like
I
said.
The
main
thing
I
would
say
is
find
the
one
that
speaks
to
you
like
something
you've
used
before
or
you're
using.
F
B
If
you're,
if
you're,
using
Contour
in
production
right
now,
you're
interested
in
Gateway
API,
and
you
want
to
contribute
to
Gateway
API,
why
shouldn't
you
go
into
again
I'm
using
this
as
an
unbiased
example:
I,
don't
work
on
Contour
at
all.
You
know
like
why
shouldn't
you
get
involved
there,
but
that's
how
that's
the
best
way
for
you
to
contribute
to
the
Gateway
API
is
if
you
can
connect
yourself
with
an
implementation
yeah,
if
that
makes
sense,
yeah.
B
And
this
meeting
actually
started
as
the
blitz
meeting
yeah,
which
is
why,
like
I
said
most
of
the
time
like
last
week
when
we
were
on,
there
was
probably
four
or
five
of
us
on
here
last
week.
Oh
yeah
I
think
Victor
was
here
too,
and
he
was
like
what
are
you
guys
doing
because
Victor's
kind
of
new
to
the
community,
but
we
were
literally
like
pair
coding,
Rusty
BPF
code
live
like
on
Blix
to
get
some
stuff
done
in
there.
We
recently
just
got
the
TCP
support
added.
D
B
But
that's
I
need
to
get
back
to
that,
so
that,
hopefully,
that
rounded
off
a
lot
of
the
things
I
was
trying
to
get
to,
but
the
most
important
one
was
where
and
how
can
I
even
contribute,
because
it's
not
like
there's
an
nginx
Ingress
controller.
For
this
thing
there
kind
of
is,
but
we're
being
extremely
guarded
about
that.
If
that
makes
sense,.
A
But
you
know
the
the
point
about
the
existing
implementations
and
such
does
make
a
lot
of
sense.
It's
certainly
handy
for
me
to
know
that
yeah,
the
official
supported,
sanctioned
way
would
be
pick
something
like,
in
my
case,
it'll
probably
be
onboard.
Gateway
use
that
as
the
mechanism
to
test
a
conformance
test
but
put
the
conformance
test
into
the
gateway
API
right.
B
A
Yeah,
yep,
and
just
in
case
anybody
is
watching
this
later.
My
previous
wise
crack
about
it
being
to
my
benefit,
to
have
fewer
graduated
service
meshes
was
a
joke.
Actually,
you
know
having
other
things.
Graduated
is
not
at
all
a
bad
thing.
Let
me
be
clear
about
that.
Yeah.
B
F
B
A
F
B
Have
it
it
needs
help
like
it
needs
more
contributors,
so
it's
actually
right
now
it
is
Rob
myself
and
I.
Think
Sanjay
are
I,
think
are
the
the
technically
considered
the
maintainers
anyway,
but
it
exists
it's
in
kubernetes.
We
did
in
the
talk
that
oh
I
won't.
Forget.
I
need
to
share
a
talk
with
you,
obvious
mom,
but
the
in
the
talk.
B
B
We
need
people
to
who
have
like
real
implementations
with
lots
like
we
need
people
who
have
like
annotations
all
over
the
place
to
come
in
and
help
us
like
map
out
annotations.
Potentially
that's
something
we're
open
to
adding,
but
like
the
basic
case
of
like
a
simple
Ingress
to
Gateway
works
today,
but
we
will
need
it
will
need
more
love
from
the
community
to
get
to
the
point
where
it's
able
to
do
some
of
the
more
complex
things
with
like
I.
Don't
know,
specific
implementations,
annotations
and
stuff
like
that.
A
B
Meeting
is
over
and
meeting
no
I
mean
you
can
try
it
I
I,
don't
know.
D
D
B
I
see
that
but
I
also
I
see
a
future
that
we
don't
do
anything
like
what
we're
doing
today.
So
the
future
that
I
see-
and
you
know,
if
I'm
wrong,
I'm
wrong,
but
is
that
AI
is
kind
of
silly
how
it's
being
used
today,
because
it's
I
this
is
opinion.
Sorry,
my
opinion
is
AI
being
used
in
the
current
way
to
generate
code
is
a
little
silly,
because
it's
basically
saying
we
have
all
this
junk.
B
We
don't
want
to
write,
let's
just
automate,
that
what
I
think
is
going
to
happen
as
a
as
like
an
evolution
of
this
is
an
eventual
future
where
compilers
and
languages
actually
have
more
AI
at
the
lower
levels
baked
into
them,
so
that
the
languages
become
more
expressive.
So
it's
not
so
much
I'm
just
automating
the
junk,
but
it's
that
the
junk
isn't
really
there,
because
the
language
is
more
expressive.
That's
the
future
I'm
hoping
for
with.
A
B
A
There's
a
there
are
a
couple
of
scenes
in
there
where
Mozart
is
working
with
salieri
and
it's
an
interesting,
an
interesting
parallel
to
some
of
the
stuff
you're
talking
about
right,
yeah.
A
B
Yeah
yeah
I
think
we're
I
think
we're
probably
going
to
get
something
more
like
that.
I'm
sure
it
will
never
be
exactly
what's
in
my
head,
but
something
more
like
that,
where
the
languages
just
become
way
more
expressive
because
of
AI
the
syntax
and
everything
is
influenced
by
it.
So
but
anyway,
we've
got
digression
to
the
I'm.
Sorry
I
can
I
can
be
I
can
go
off
on
the
digression,
like
you
wouldn't
believe.
Do
we
have
questions
that
are
relevant
to
Gateway
and
stuff
like
that.
E
I
have
one
question
yeah,
so
you
said
that
Ingress
and
AP
Gateway
they
have
been
kept
a
separate
projects
right,
so
this
project
showed
it
right
now
on
the
screen
that
Ingress
to
Ingress
to
Gateway.
So
isn't
that
making
Ingress
more
like
Gateway
working
more?
Are
you
making
Christmas
pick
up
moment.
B
Having
Ingress
be
more
gateway,
conformant
to
me,
the
wording
that
you've
used
sounds
like
we're
trying
to
make
sure
that
the
things
that
like
come
out
of
Ingress
to
Gateway
are
functionally
working
and
I
think
by
they
would
be
like
I
think
that
they
I
I'm
having
a
little
bit
of
trouble
kind
of
getting
at
the
point
of
the
question.
But
if
the
question
is
you
know,
should
Ingress
resources
today
be
conformant
as
they're
converted
to
Gateway
API
Resources,
with
something
like
Ingress
to
Gateway?
Yes,
like
that
everything
should
just
work.
B
Is
there
room
for
some
error
where,
like
something
doesn't
quite
work,
the
same
way
that
is
kind
of
up
to
the
implementations?
The
implementations
could
definitely
mess
that
up
like
there
could
be
subtle
differences
in
their
implementation
of
HTTP
route
versus
Ingress
that
cause
things
that
are
converted
to
actually
do
something
different,
and
these
are
things
that
maybe
we
need
to
be
better
about.
Looking
out
for.
If
that
was
your
point,
is
that
your
point.
A
I
think
part
of
the
point
of
the
Gateway
API
is
excuse
me,
I.
Think
part
of
the
point
of
the
Gateway
API
itself
is
trying
to
deal
with
some
of
the
complexities
in
interpreting.
What
exactly
Ingress
is
supposed
to
mean
giving
all
the
given
all
the
annotations
that
proliferated
on
the
Ingress
resource.
A
So
in
a
lot
of
ways,
the
point
of
the
Gateway
API
was
acknowledging
that
the
Ingress
resource
wasn't
really
working
out
and
trying
to
come
up
with
something
better
So
to
that.
Given
that
background
I
think
it
makes
an
enormous
amount
of
sense
to
give
people
a
path
to
migrate
off
of
Ingress
onto
something
that
is
structured
and
less
flexible
or
not
less
flexible,
structured
and
less
subject
to
random
interpretation.
How
about
that?
D
B
Have
said
that
way
earlier
on,
considering
we're
being
recorded
but
yeah
Ingress.
C
B
A
B
If
you're
previously
using
it,
you
don't
need
to
rush
to
upgrade
or
migrate
or
anything
like
that.
But
if
you
are
not
previously
using
Ingress,
probably
start
with
Gateway
yeah,
don't
use
Ingress
avoid
it.
E
B
B
So
please
do
reach
out
to
me
personally:
I,
don't
I'm
at
shame
in
kubernetes
slack
and
then
just
I
do
I
would
I
generally
think
it's
better
to
post
in
the
Gateway
API
Channel,
but
if
you're
not
comfortable
with
that
PMA,
because
you'll
get
a
lot
more
people
and
I'm,
not
always
online
and
whatnot,
but
the
really.
D
A
D
B
As
a
matter
of
fact,
although
with
the
Advent
of
chat,
gbt
I
may
be
able
to
be
online
constantly
soon
we'll
see
if
I
can,
if
I
can
create
a
shamebot,
no
so
yeah,
so
please
do
reach
out.
You
know
if
you're
getting
stuck,
don't
let
yourself
feel
stuck
we'll.
You
know
reach
out
to
the
community
there's
so
many
people,
it's
it's
such
a
big
group.
B
It's
the
it
is
the
biggest
sub
project
in
kubernetes
under
a
Sig
that
I'm
aware
of
in
terms
of
sheer
number
of
contributors
and
like
active
people
on
syncs
and
stuff,
like
that,
it's
close
it's
one
of
the
top
ones
and
then,
personally
speaking,
it's
the
most
intense
open
source
thing
as
a
maintainer
I
have
I
have
come
to
an
appreciation
of
just
how
busy
we
are
so
I'm.
A
At
Flynn
on
pretty
much
all
the
slacks
in
the
world
as
far
as
I
can
tell,
based
on
how
many
show
up
in
my
slack
app
but
yeah,
yeah
and
pretty
much
everything
Shane
just
said
applies
to
me
as
well,
happy
to
hear
from
people
and
help
out
if
I
can
and
I
am
also
not
always
online.
Although
there
are
people
who
won't
believe
that.
A
Unlikely
are
you,
are
you
lior,
or
is
your
slack
handle
more
complex
than
that?
Yes,.
A
Yeah
I
didn't
realize
that
you
were
working
on
Envoy
Gateway.
Actually,
so
that's
I'm,
not
okay,
yeah.
A
I'll
I'll
tell
Luke,
you
said
that
and.
B
A
Envoy
Gateway
is
you
know,
as
Ingress
controllers
go,
Envoy
Gateway
is
still
very
straightforward
and
lightweight
mostly
because
it
doesn't
do
a
whole
lot,
so
its
build
system
is
straightforward
and
the
people
who
wrote
its
build
system
I
think
the
most
polite
way
of
saying
it
is
learned
a
lot
from
the
horror
that
is
emissary's,
build
system
and
did
things
differently.
A
So
it
would
be
nice
to
steal
some
of
that
stuff
back
for
Emissary,
but
there
are
reasons
why
it's
harder
for
Emissary
as
well,
but
end
result
if
you're
doing
something
vaguely
like
Envoy
Gateway
in
terms
of
trying
to
run
the
conformance
tests.
It
actually
is
a
pretty
straightforward.
Make
file
setup.
A
A
B
Is
talking
about
Gateway
operator
or
I'm?
Sorry.
A
B
Gateway
feel
free
to
make
your
pitches
for
your
own
things
like
I
I,
didn't
really
do
a
pitch,
but
like
open
door,
if
you
want
to
say
hey
I'm
working
on
this
project,
we'd
love
to
have
you
join
us.
That's
fine!
This.
F
F
F
A
B
Yes,
please
please
more
for
kubernetes,
which
everyone
is
using
right,
all
right,
so
I
guess
we're
about
15
minutes
over
I'll,
probably
probably
should
wrap
it
up.
Thank
you
for
all
the
questions.
Please
do
reach
out
in
slack
and
stuff
like
that.
I
hope
this
was
very
helpful
for
you
and
again,
if
it
wasn't,
please
reach
out,
in
slack,
feel
free
to
be
like
man.
That
really
wasn't
that
helpful.
Could
you
help
me
with
these
things
yeah?
So
this
is
a
good
one.
Thank
you
all.
Thank
you.