►
From YouTube: SIG Network Gateway API meeting for 20230424
Description
SIG Network Gateway API meeting for 20230424
A
Hello
and
welcome
everybody
to
the
April
24th
edition
of
The
Gateway
API
meeting,
just
a
reminder
as
usual
that
this
is
the
this
is
under
the
kubernetes
code
of
conduct,
which
basically
boils
down
to
be
nice
to
one
another.
So
please
be
nice
to
one
another.
A
We
continue
to
like
to
get
a
list
of
attendees,
even
if
it's
a
little
tedious,
if
you
wouldn't
mind
putting
down
your
attendance,
we're
going
to
start
now
that
we're
past
kubecon
experimenting
with
some
more
EU
friendly
time
zones
and
stuff
and
knowing
when
people
show
up
is
helpful
to
us.
So
if
we
would
do
that,
we'd
appreciate
it.
A
We
do
have
quite
a
few
things
on
the
agenda,
but
I
think
the
vast
majority
of
it
is
triage.
So
just
this
is
an
open
Agenda.
If
you
want
to
put
something
above
triage
that
you'd
like
to
talk
about
today,
please
do
feel
free
to
do
that
anytime.
A
B
Yeah
I
think
I
monopolized
today's
agenda,
just
because
I
may
have
been
the
only
one
to
put
things
on
here,
but,
as
always,
anyone
is
welcome
to
I'll
start
off
with
the
070
release,
since
that
is
absolutely
top
of
mind,
we're
trying
to
get
that
one
out.
I
did
another
check
in
with
Tim
and
Cal
today,
and
they
don't
have
any
major
blockers.
So
there's
one
last
little
PR
that
has
a
few
I
think
are
really
tiny
fixes
once
that
merges
I
think
we're
clear
to
cut
rc1
and
I.
B
Don't
I'm
not
aware
of
any
major
changes
between
rc1
and
a
final
release,
so
this
may
be
a
release.
We
don't
even
need
a
second
release
candidate,
we'll
see
but
yeah
just
tiny
little
changes
on
this
one.
Hopefully
this
is
our
last
last
big
thing
between
now
and
an
RC.
B
And
then
yeah
I
just
wanted
to
do
a
kubecon
recap,
because
yeah
there
were
a
lot
of
really
great
people
there
for
those
who
aren't
familiar
with
kubecon.
They
also
have
a
photographer
that
goes
around
at
most
of
their
events
and
takes
pictures,
and
there
are
a
few
pictures
that
are
evidence
that
we
were
all
well.
Many
of
us
were
in
the
same
room
together.
So
just
a
heads
up
that
if
you
feel
like
looking
through
that
there
are
pictures
of
us
that
may
or
may
not
load
eventually.
B
To
do
that,
oh
yeah,
no,
no
worries,
no
worries
all
right
and
that
that's
it
for
that.
One
next
up
I
wanted
to
highlight
the
slides
that
Nick
and
I
shared
for
Gateway,
update
and
I.
You
know
this
is
there
were
a
lot
of
Gateway
talks,
Shane
and
Nick
presented
Gateway
update
to
contrib
Summit
as
well.
I,
don't
have
those
slides
in
there,
but
there's
some
overlap,
but
lots
and
lots
of
interest
in
Gateway,
Concepts,
I'd
say
most
of
the
Gateway
focused
talks.
B
I
went
to
were
largely
full
and
I
think
there
were
some
like
eight
of
them,
so
I
didn't
I,
didn't
make
it
to
quite
all
of
them,
but
yeah
lots
of
interest
there
I
just
I,
just
wanted
to
run
through
some
of
my
highlights.
There's
this
reference
Grant
scratch
dock.
That
yeah
represents
what
I
thought
was
a
really
important
opportunity
for
us.
Some
one
of
my
main
goals
for
kubecon
was
to
try
and
move
this
forward.
B
We
were
able
to
get
a
lot
of
key
people
all
in
the
same
room
together
and
have
some
discussions,
so
that
meant
all
Gateway
maintainers.
In
addition
to
all,
but
one
cigar
lead
and
TL
were
there,
so
it
was
I,
think
everyone,
but
Mike
was
there
from
the
Sega
off
side.
So
it
was
a
really
great
group
and
we
got
to
you
know,
talk
through
what
the
future
of
reference
Grant
looks
like
when
it
gets
into
cigar.
B
There's
some
really
I
think
really
good
ideas
here,
but
the
high
level
idea
from
ref
from
Sig
auth
is
that
they
want
reference
Grant
to
not
just
be
something
like
that
is
so
reliant
on
controller
implementation,
but
to
actually
also
Grant
access
to
a
controller,
or
something
like
that.
Of
course,
cigarth
would
be
the
ones
to
want
to
do
that,
so
they
have
a
an
idea
of
what
exactly
that
would
look
like
and
how
it
will
work.
There
will
be
more
discussion
on
this
at
the
sigoth
meeting
on
Wednesday.
B
So
if
anybody
feels
like
showing
up
to
that,
one
I'd
recommend
it
there's
a
lot
in
here,
but
just
the
the
high
level
idea
that
reference
grants
should
Grant
access
to
a
controller,
and
the
idea
of
how
that
would
work
is
an
intermediate
resource
called
again.
Who
knows
what
the
name
will
be,
but
we
used
a
placeholder
of
reference
consumer
that
connects
a
set
of
reference
grants
to
a
controller
and
then
all
reference
grants
of
that
type,
Grant
access
to
that
service
account
or
whatever.
However,
that
controller
is
getting
accessed.
B
Yeah
I
think
this
is
really
powerful,
but
it's
going
to
take
some
work
so
just
kind
of
a
heads
up
that
this
is
moving
forward.
Cigarth
is
very
interested,
but
I
realize
that
it's
going
to
be
a
slow
rule
to
to
get
anything
done
here
and
if
you
know,
if
there's
anyone
out
there
that
has
cycles
and
would
like
to
you
know,
is
invested
in
reference
Grant
and
also
interested
in
seeing
this
go
Upstream.
We
have
no
shortage
of
work
here
and
could
appreciate
help.
B
Yeah
I,
don't
know
Shane.
You
were
also
on
most
of
those
well
in
all
of
the
same
conversations.
I
don't
know
if
you
had
a
a
different.
A
Day,
I
think
just
no
same.
Take
I
think
it's
the
the
direction
that
they're
trying
to
head
of
like
actually
granting
our
back
permissions
is
very
interesting,
I
think
for
the
uninitiated.
It
might
be
good
to
know
that
Sig
storage
is
also
involved
in
this
you'll
see
them
around.
In
this
conversation
they
are
currently
using
reference.
Grant
for
I
forgot
exactly
what
it
was,
but
they
they're
bullish
about
it
and
they
actually
showed
a
couple
of
them
showed
up
at
kubecon
and
we
talked
about
it
to
some
extent.
So,
just.
C
A
Heads
up
that
you'll
see
seek
storage
in
the
mix
of
this
as
well.
So
there's
a
lot
of
eyes
on
it.
A
Nothing
else
really
to
add
there,
but
definitely,
if
you're,
interested
Now's
the
Time
to
jump
in.
If
this
is
something
that
like
interests,
you
and
you
want
to
be
involved
in
the
implementation
stuff.
Now
is
the
best
time
to
yeah.
B
For
for
context,
this
feels
like
it's
going
to
be
one
of
the
most
foundational
shifts.
That's
occurred
in
kubernetes
authorization
since
our
back
since
this
is
really
just
a
you
know,
an
extension
of
that
a
way
to
Grant.
D
B
Access,
so
you
know
if,
if
you
have
opinions-
and
you
want
to
be
involved
in
something
like
as
foundational
in
kubernetes,
this
is
one
of
those
rare
opportunities
that
we
get
so
yeah.
E
B
F
B
All
right,
yeah
I'll,
keep
on
moving
and
the
other
thing
that
I
found
really
interesting
about
kubecon
was
a
TV2
Denmark.
B
They
presented
I
thought
one
of
my
favorite
kubecon
talks
about
Gateway
implementation
had
a
really
great
demo,
I
think
a
really
cool
concept
is
as
a
controller
of
Gateway
API,
and
that
concept
is
really
just
allowing
templating
enabling
templating.
So
the
idea
that,
given
these
Gateway
API
values,
how
do
you
translate
them
into
other
apis
which
basically,
what
all
our
controllers
are
doing,
but
they
they
made
that
in
such
a
way
that
it
is
easy
to
extend.
So
in
their
case
they
had
a
demo
where
they
were
using.
B
You
know
they
were
running
two
gateways
in
parallel,
one
they
were
using
cross
planes,
provider
API
to
talk
to
AWS,
and
then
they
were
using
gateway
to
talk
to
istio.
But
from
their
perspective
that
you
know
it
was
the
same
config.
It
was
the
same
everything
they
were
using
Gateway
API
and
it
just
worked.
B
I
thought
it
was
really
neat.
I
think
that
the
patterns
they
have
in
here
are
really
really
cool
they're
using
the
policy
attachment
API
and
get
very
well
I
think
so
it's
a
good
use
case
if
you're
looking
for
examples
of
policy
attachment
and
they've
seem
to
use
it
very
well
and
then
the
very
last
thing
kind
of
the
next
link
I
I
had
here
was
I.
Think
all
of
us
were
really
Amazed
by
this.
B
They
have
a
really
neat
neat
tool
that
simply
converts
Gateway
API
resources
and
policies
of
any
kind
as
long
as
they
have
the
appropriate
labels
into
SVG.
That
shows
the
layout
and
and
graph
hierarchy
of
everything.
So
I
think
if
you
scroll
down
a
bit
yeah,
that's
that's
the
kind
of
output
you
get,
and
you
know
that
they
use
this
for
the
demo.
It
was
live.
Updating
it
just
seemed
like
magic,
but
if
you
ever
are
ever
looking
for
a
tool
like
this,
it
exists
and
I
thought
it
was
really
cool.
B
B
Cool
yeah,
it
was
one
of
the
the
best
examples
of
you
know,
really
getting
and
understanding
Gateway
API
and
just
yeah.
B
I
I
recommend
checking
it
out,
it'll
be
on
YouTube
sometime
soon,
but
yeah
next
couple
weeks.
B
All
right-
and
the
very
last
thing
I
wanted
to
highlight
from
kubecon-
was
Envoy
Gateway
and
Arco
you're
on
this
call,
so
you
can
probably
give
more
context
than
I
could
but
I
just.
This
was
my
first
time
realizing
that
you
were
using
custom
route
filters
and
maybe
you're
not
the
only
one
but
I
just
thought.
It
was
really
a
good
use
case
of
them.
E
Yeah,
so
we
released
this
in
February
of
this
year
and
at
that
point
we
were
at
with
linked
to
zero
six
zero.
At
that
point,
direct
policy
attachment
didn't
exist,
so
the
natural
extension
way
to
extend
authentication
was
to
add
a
filter,
and
we
also
added
one
for
rate
limiting,
because
I
think
we
discussed
and
for
rate
limiting
I
think
we.
We
decided
that
the
Persona
ties
more
to
an
application
developer,
and
so
we
decided
to
keep
the
config
closer
to
the
route
and
so
yeah
we
added
two
customizations.
E
We
plan
on
adding
some
more
in
the
next
release
for
authorization
and
yeah.
More
sub
authentication
features.
E
With
direct
policy
attachment
being
created
and
experimental
I
think
we
need
to
rediscus
whether
some
of
these
need
to
be
moved
into
a
policy
attachment
or
not.
B
Cool
yeah,
I
I
would
love
to
you
know,
also
see
the
the
pros
and
cons
of
using
either
one
I
know
that
they're
they're
definitely
getting
a
little
bit
closer
together
in
terms
of
pros
and
cons.
B
I
really
like
the
idea
of
you
know,
route
for
custom
route
filters
just
because
they
you
know
it
feels
like
the
right
place
for
this
kind
of
configuration,
but
maybe
yeah
I,
guess
that's
kind
of
hard
to
quantify
so
I'm
interested
in
how
you're
making
decisions
as
well
of
whether
it
becomes
policy
attachment
or
this
kind
of
custom
route,
filter.
B
Yeah,
just
out
of
curiosity,
anyone
else
on
the
call
using
custom
route
filters
or
planning
to.
E
B
That
in
gk's
implementation
for
something
but
yeah
interested
as
always,
and
how
these
extension
mechanisms
are,
are
being
used.
Foreign.
B
With
that
maybe
I'll
just
keep
on
running
through
this.
Anyone
feel
free
to
interrupt
me
at
any
point,
but
yeah
the
this
one
came
out
of
a
a
discussion
actually
with
a
customer
that
really
wanted
to
have
redirects
drop,
a
trailing
slash,
and
that
seemed
like
entirely
reasonable
request
and
something
like
that
is
as
far
as
I
can
tell
either
impossible
or
unnecessarily
difficult
with
Gateway
API.
Today,
the
first
part
of
the
question
is
what
people
are
doing
with
exact
matching.
B
Today
we
have
for
prefix
matching
we've
been
very
clear
that
a
trailing
slash
is
just
ignored,
whether
or
not
it
exists
it
doesn't
matter.
It's
it's
a
prefix
match
the
path
match.
On
the
other
hand,
it
the
exact
match,
on
the
other
hand,
is
just
we
match
exactly
does
exactly
include
a
traveling
slash.
You
would
think
because
we
don't
say
that's
what
it
means,
but
I
don't
know
anyone
with
visibility
into
your
implementation,
that
that
knows
what
how
this
is
being
implemented
today,
it's
kind
of
a
open
question
here.
A
But
so
you'll
be
curious
if
somebody's
already
trying
it
yeah
but
I
I'm
thinking,
there's
some
rfcs
out.
There
that'll
probably
give
us
the
the
legend
by
which
to
go
off
yeah.
G
Yeah
yeah,
this
is
in
the
way
you
know.
E
G
And
Apache
dealt
with
this
in
the
past
with
how
they
did
prefix
matching.
Maybe
look
at
those.
G
B
Squid,
okay,
yeah
yeah,
good
idea
and
then
I
guess
oh
yeah,
and
thanks
for
our
RFC
reference
and
then
I
guess
kind
of
the
the
next
natural
follow-up
to
that
is
the
other
issue
which
is
you
know,
even
if
the
trailing
slash
an
exact
match
is
Meaningful
redirecting
to
drop
a
trailing.
Slash
is
unnecessarily
complicated,
there's
not
kind
of
a
way
to
configure
that
in
any
simpler
path,
rather
than
just
you
know,
per
path
for
exact
match.
B
It
seems
like
it'd,
be
nice
to
provide
a
simpler,
a
simpler
method.
To
do
this.
Don't
have
any
ideas
right
now,
but
just
want
to
track
this
and
and
see
what
kind
of
what
kind
of
interest
there
is
for
this.
B
This
is
just
again,
you
know
showing
what
drop
trailing
slash
could
look
like
with
if
trailing
slash
was
meaningful
on
exact
match,
but
if
it's
you
know
either
way,
we'll
need
to
spend
some
time.
Thinking
about
how
we
want
to
handle
this
drop,
redirect
drop,
trailing
slash,
redirect.
A
A
Are
at
the
end
of
everything
but
triage
with
quite
a
bit
of
time,
left
to
go
if
anybody
has
any
pop-up
items
or
just
things
that
they
want
to
discuss
before
we
move
into
triage,
that
would
be
totally
fair.
We
can
give
a
couple
seconds,
otherwise
we're
going
to
go
into
trash.
F
B
Mean
yeah,
so
please,
please,
do
I
tried
to
just
precede
this,
this
triage
list
with
items
that
I
haven't
seen
or
haven't
taken
enough
of
a
look
at,
but
this
is
an
entirely
meant
for
anybody
to
add
items
to
so
this
is
an
incomplete
list.
I'll
say
that
over
kubecon,
the
list
of
things
to
review
grew,
and
you
know
this
is
there's.
E
B
Lot
more
I
want
to
look
at
than
is
here
right
now,
but
you
know
at
a
high
level
thing
just
to
again
and
differentiate,
trying
to
make
sure
that
we're
triaging
things
that
are
between
us
and
GA
first
and
then
as
there's
time
going
through
anything
outside
of
that.
C
C
H
Problem
so
I
went
ahead
and
unsuccessfully
created
him
yep,
but
so
I
appreciate
the
direction.
H
And
I
did
it
as
a
draft
mode,
because
I
just
thought:
well,
why
not
just
be
careful
so
I
think
the
first
thing:
what
would
you
just
recommend
going
for
just
not
a
draft
mode
and
just
straight
into
a
pull
request
when.
A
You're
actually
ready
for
people
to
start
looking
at
it.
Market
is
non-draft,
so
even
if
it's
still
kind
of
whip,
but
if
you
want
review,
it's
I
mean
you
can
specifically
ask
for
review
too.
If
you
want
to
leave
it
in
draft,
but
like
ask
specific
people,
it's
kind
of
up
to
you,
I,
don't
I,
don't
think
there
needs
to
be
a
hard
and
a
fast
rule
about
when
something
is
a
draft
generally
I'll
put
something
in
draft
in
when
I
don't
want
it
to
be
reviewed,
and
then
it
gets
reviewed
anyway.
So
I'm
not.
A
A
Yeah,
if
you're
ready
for
review
I
can
personally
start
giving
like
I,
can
jump
on
this
for
you
and
start
doing
I'm
already
a
review
around
there.
So
I'll
get
to
this.
H
H
B
Usually,
usually
the
way
this
works
is
different
for
every
employer.
I
know,
in
my
case,
I
just
have
to
have
my
work,
email
associated
with
my
with
my
GitHub
and
with
my
commits
and
then
the
rest
just
is
automatic,
but
I
I
know
it's
different
depending
on
your
employer,
but
there
is
probably
some
internal
guidance
for
for
working
with
this
right
and.
H
A
Yeah
yeah
I'm,
sorry
because
I
signed
this
thing
like
nine
million
years
ago,
so
I
don't
even
but
the
I'm
pretty
sure
on
the
cncf
slack.
There
are
places
where
you
can
get
help
with
this
as
well
with
questions
and
stuff,
so
not
on
the
kubernetes
slack
but
on
the
cncf
slack
I
think
there's
a
few
channels
related
to
to
this
and
we
can
help
you
find
them
too.
A
A
H
H
Okay,
I'll
just
need
to
dig
a
little
bit
digging.
H
A
E
A
Let's
just
get
the
groundwork
going,
you
know,
don't
don't
make
it
too
big
I.
Think
that's
the
way
to
to.
B
D
H
A
Me
feels
like,
oh
no,
like
there's
a
lot
of
implications
of
adding
an
egress
route.
We've
held
off
on
doing
things
like
even
adding
like
IP
routes
in
the
past,
because,
like
it
breaks
some
of
the
we
actually
talked
about
this
at
kubecon.
Everything
thus
far
when
it
comes
to
like
routes,
is
effectively
just
programming.
Rules
like
instructions
for
programming,
your
data
type
plane,
your
gateway,
egress
route,.
C
A
C
A
You
some
thoughtful
comments
and
we'll
kind
of
try
to
make
sure
we're
also
not
blocking
it
like
we'll
we'll
come
up
with.
Whatever
we
feel
like
you
know,
is
a
good
way
to
move
forward,
but
also
where
the
boundaries
are
because
e-rest
route
is
definitely
going
to
be.
That
would
be
very,
very
interesting,
yeah
and.
B
And
interesting,
to
say
the
least,
and
just
to
set
expectations
that
you
know
we're
I
think
almost
all
of
us
agree
that
we
want
egress
to
be
in
in
scope
of
this
API
at
some
point,
but
also
so
we
want
to
see
something
like
this.
At
the
same
time,
we
are
very
hyper
focused
on
getting
GA
out
the
door,
and
so
you
know
we're
doing
what
we
can
to
try
and
prioritize
things
that
fall
between
us
and
GA,
and
this
is
unlikely
to
be
one
of
those.
B
A
More
more
consider
like
quick
things
that
can
be
consumed
quicker,
definitely
helps
to
get
a
little
bit
of
attention
during
this
time.
So.
H
A
A
Okay,
so
add
validation
for
Gateway
spec
addresses
I.
Think
I
just
looked
at
this
for
a
hot
second
and
just
put
it
in
triage.
So
we
didn't
miss
it.
Current
validation
server
does
not
have
validations
for
Gateway
spec
address,
addresses
and
that
is
needed.
A
B
B
B
Said
yes,
yeah:
it's
like
a
quality
of
life
Improvement!
It's
it's
not
like
a
strict
requirement,
but
it
sure
it
sure
seems
reasonable.
Might
as
well
I
mean
it's.
Gonna
have
to
be
something:
that's
in
web
hook,
validation,
anyways
but
doesn't
seem
to
have
any
harm.
A
F
B
A
You
know
it,
it
occurs
to
me
now
in
retrospect
that
there
would
be
such
there
would
definitely
be
implementations
of
Gateway,
where
I
would
think
you
would
not
want
immutable
addresses
like
absolutely
right.
I.
A
The
other
issue,
just
so
I'm
I,
feel,
like
my
brain's,
a
little
broken
but
I'm
also
jet
lagged
like
you,
wouldn't
believe.
What's
the
what's
the
what's
the
other
issue
you
were
just
referring
to
that
relates
to
that.
F
In
cluster
local,
there
was
a
discussion
of
reachability
of
a
Gateway,
it
should
be
mutable,
and
maybe
the
answer
is
no
either
or
I
guess:
I'm
I'm
not
coming
in
with
opinions.
For
me,
it
might
make
sense
that
it
is
immutable
and
then
I
was
looking
at
address
as
sort
of
a
hey.
What's
the
discussion
there.
F
Yeah
so
I
guess
the.
E
F
B
A
A
If
there's
something
to
be
to
be
discussed
there,
but
I'm
having
a
little
bit
of
trouble
with
it
personally,
because
I
just
can't
I
can't
see
not
being
in
a
position
where
you
couldn't
remove
an
address
that
you
were
no
longer
going
to
use
for
the
Gateway
like
that
being
a
thing,
but
definitely
want
to
hear
you
out
on
it.
If
there's
some
ways
in
which
this
is
making
sense,.
F
No
I
don't
really
have
opinions.
It
was
more
just
like
if
there's
a
precursor
to
the
reachability
immutability,
which
to
me
I
would
say
like,
doesn't
make
sense
to
change
the
gateways
reachability.
But
then
I
was
wondering
if
there
was
a
any
relation
to
like
changing
addresses.
D
A
All
right
that
one's
ready
to
go,
let's
go
on
to
the
next
one
documenting
HTTP,
header,
modifiers,
I,
didn't
see
this
one
I
think.
B
No,
this
is
great.
This
is
a
Nico
filed.
This
I
took
a
quick
glance
and
it
looks
it
looks
great.
This
is
a
pretty
big
gap
in
our
documentation
right
now,
I
I've
added
a
you
know
this
to
my
list
of
things
to
review,
but
you
know
it
would
be
great
if
others
have
some
time
to
help
review
it.
This
is
yeah
huge,
Improvement
and
definitely
something
like
we
need
pre-ga.
A
B
Yeah
no
I'm
excited
to
see
Nico
writing
docs
for
us,
because
he's
he's
great.
A
A
B
Yeah
so
I
think
any
of
us
can
answer
this.
It's
it's
it's
something
that
we
need
to
clarify
in
our
documentation
that
it
is
only
required
for
for
cross
namespace
references
and
is
at
this
point,
Irrelevant
for
same
namespace
references,
but
the
singoff
changes
that
are
coming
may
change
that,
but
for
the
purposes
of
Gateway
API
yeah.
So
this
this
needs
someone
to
respond
and
say
something
like
that
and
then
somebody
else
to
actually
make
the
change
or
or
potentially
the
same
person
to
make
the
change
in
API
spec.
A
Okay,
I'll
leave
this
tab.
Open
I
took
it
I'll,
get
it
moved
forward
and
get
some
those
comments
out.
There
invest.
A
Conformance,
oh
I
worked
on
this
one.
Okay,
so
I
think
I
was
expecting
to
put
out
a
prototype
of
conformance
profiles
before
kubecon,
but
Matia
had
a
requested.
He
basically
we
had
talked
and
he
was
interested
in
helping
out
with
it
and
helping
me
particularly
to
plug
it
into
the
implementation
he
works
on.
So
during
kubecon
he
was
kind
of
doing.
C
A
And
here's
the
pr
so
this
is
I'm
not
going
to
really
review
this
since
three
of
the
commands
are
mine,
but
for
those
who
are
interested
in
the
conformance
profiles,
this
prototype
does
ultimately
at
this
point,
produce
a
conformance
report.
A
Things
are
still
a
little
rough
there's
still
a
couple
things
like
this
one
which
we
got
to
figure
out,
but
the
the
basics
are
kind
of
there.
The
basic
Machinery
is
working,
it's
all
underneath
an
experimental
build
tag,
so
it's
literally,
you
know
not
part
of
the
conformance
suite.
Yet
you
have
to
opt
in
with
a
go,
build
tag.
A
We
would
like
reviews.
Obviously,
but
then
the
idea
is.
This
is
just
a
small
chunk
to
move
us
a
little
bit
closer.
We
have
like
the
basic
Machinery
working
like
other
people
to
jump
in
and
try
their
implementation
see
if
they
also
can
get
a
report
going
and
then
make
the
little
changes
ultimately
to
just
iterate
on
this
profile
prototype
as
part
of
the
get
process.
Until
you
get
to
the
point
where
we're
like.
Okay,
we
have
a
few
implementations
kind
of
liking.
What
we're
seeing?
Let's
talk
about?
B
Yeah
just
so,
this
is
awesome,
but
one
high
level
thing.
How
are
we
running
this
like
I
I,
see
there's
a
lot
of
metadata
in
here
like
organization,
name
and
description.
Whatever
is:
are
those
just
command
line?
Flags.
A
Figured
out
how
to
how
to
there's
things
like
even
getting
the
Gateway
API
version,
because
we
just
basically
assume
that
everything
is
going
to
be
pre-deployed
before
you
run
the
test,
Suite,
it's
kind
of
LOL,
so
you
can't
like
dynamically,
really
get
that
right
now,
so
we're
thinking
about
that,
and
it
might
end
up
being
that
there
needs
to
be
Flags
or
maybe
a
metadata
config
file
that
could
also
be
Flags
or
whatever
that
people
need
to
provide.
B
At
right
now,
yeah
I
think
the
way
it
works
in
upstream-
and
someone
can
correct
me
on
this,
but
I
think
the
way
it
works
is
there.
You
know
every
ever
release.
Every
minor
release
has
a
directory
within
that
every
implementation
that
claims
conformance
has
a
directory
and
then,
within
that
there's
a
like
a
metadata
I
think
yaml
file,
a
markdown
file
that
provides
you
know
this
is
these
are
the
steps
we
took
to
run
the
conformance
test
and
then
just
a
another
file.
C
A
We
weren't
really
focused
on
that
stuff
just
yet,
because
the
Gap
was
where'd.
We
got
all
these
fields
exist
because
they're
things
that
we
decided
we'd
like
in
the
gap,
but
at
the
point
where
prototyping
we
literally
just
wanted
to
get
to
the
point
where
we
were
actually
able
to
produce
a
report.
So
the
details
of
how
we
get
some
of
these
fields
is
definitely
open,
and
that
sounds
fine
and
I
think
it's
kind
of
a
because
this
is
a
prototype
because
it
is
behind
the
build
tag.
A
I
think
the
idea
at
this
point
is
Judge
it
in
in
the
pr
review
based
on
what
it
can
do
so
far,
but
we'll
iterate
on
these
other
bits,
rather
than
trying
to
solve
them
all
in
one
place,
and
that
also
opens
up
the
door
so
that
more
people
can
contribute
and
like
this
is
kind
of
the
fun
part.
So
we
didn't
want
to
monopolize
it
like
there's
a
lot
people
writing
the
code
is
actually
the
part
that
a
lot
of
us
enjoy
so
go
ahead.
David.
F
Just
a
quick
question
about
this:
is
it
easy
to
create,
for
example,
a
conformance
profile
Downstream,
because
one
use
case,
I
kind
of
realized
is
while
like
Downstream
key
native,
would
probably
be
wanting
our
own
performance
profile,
which
might,
for
example,
include
core
extend
it
and
we
might
want
to
turn
on
some
additional
features
that
way
we
can
sort
of
say
hey.
If
you
run
this
conformance
profile,
then
you
know
it's
been
all
work
well
with
you
native
rather
so
I'm
wondering
is.
A
Easier
yeah,
if
I'm
understanding
your
question
correctly.
Well,
the
the
direct
answer
to
whether
or
not
it's
easy
to
do
entire
profiles
by
yourself.
You
can
in
fact
create
a
profile
by
your
like
Downstream,
but
I
think
in
practice
that
they
won't.
Those
named
profiles
will
not
be
recognized
as
part
of
like
the
certification
process
right,
so
you
could
do
it.
You
get
a
report
that
includes
your
custom
one,
but
the
idea
and
we
need
to.
A
Here
is
that
we
would
have
name
profiles
like
in
this
case
http,
which
should
be
pretty
Universal
to
everybody,
and
then
you
can
do
what
you
were
saying
where
you
can
say
like
we
only
support
specific
when
it
gets
out
of
core.
When
you
get
into
extended,
we
only
have
like
specific
features
that
is
represented
by
the
extended
section
where
you
can
actually
it'll,
actually
identify
the
features
that
you
do
or
do
not
support
from
extended.
Does
that
make
sense
yeah?
Does
that
capture
kind
of
what
you're
getting
at.
F
Yeah
look
I'm
not
looking
to
push
a
conformance
profile,
Upstream,
just
essentially
for
us
to
have
our
own
Downstream.
That
makes
it
easier
to
run
a
specific
set
of
conformance
sets
and
features.
Yeah.
C
A
Is
all
prototypical
so
definitely
want
you
involved
like
because
this
is
part
of
the
fun
of
it,
but
in
its
current
state,
if
you
run
a
profile,
you
you
get
a
conformance
report.
A
You
know
you
can
not
use
profiles
technically
like
during
a
development
period
where
you
just
pick
pick
the
specific
tests
or
specific
features
that
cover
a
certain
number
of
tests
that
you
want
to
run,
and
we
recently
made
the
change
that
all
tests
have
to
be
associated
with
features
so
that
helps
with
that
as
well.
So.
C
A
F
Yeah
because
I
could
imagine
there
might
be
some
experimental
features
that
some
implementations
support
and
it
might
be
a
test.
That's
Upstream,
but
that's
not
broadly
enabled
so
be
good
to
have
like
a
profile
that
I
can
use
myself
to
configure
these
tests
and
I
can
run
all
these
tests
Downstream
against
different
implementations,
yeah
cool,
great
thanks.
A
So
yeah
looks
like
arco's
already
taking
a
look
at
it.
Thank
you,
Arco
and
we'll
pull
in
a
few
more
and
kind
of
keep
giving
it
a
go.
We
don't
have
to
spend
too
much
time
on
this.
One
though
we're
getting
close
to
to.
E
A
End
here
and
we
still
have
a
few
items
we
want
to
look
at
so
please
do
check
that
out.
That's
1967
that
kind
of
affects
everyone,
I,
think
anybody
who's
implementing
yeah
all
right.
Let's
start
looking
at
some
of
the
others.
A
F
I've
done
that
I
left
some
of
them
open
because
I
when
there's
a
question
and
then
I
reply,
then
I
also
update
the
Gap.
So
ideally
I'd
want
the
author.
F
Two
also
I
also
cannot
leave
it
open,
yeah.
Sorry,
it's
a
fine
art
because
it
isn't
what
happens.
Is
people
will
come?
They
will
read
the
comments,
but
they
don't
read
the
gap
or
they
read
the
Gap
and
they
don't
read
the
comments.
So
that's
why
I've
been
kind
of
very
diligently,
throw
as
I
resolve
some
things
put
them
into
the
Gap
based
on
the
discussion
in
the
comments,
but
I
still
like
to
think
it's
worth,
leaving
some
comments
open
to
reading
so
that
always
you
kind
of
gain
some
context.
F
You
can
bit
mixed
on
that,
but
I
have
resolved
a
bunch
and
I
think
for
most
of
it
like
some
are
just
really
old
and
don't
apply
anymore,
so
they
really
should
be
resolved.
A
Yeah
you're
right
that
it's
a
fine
art,
my
my
default
personal
and
and
very
personally,
is
because
99
out
of
100
times
the
original
commenter
will
not
resolve
their
own
comment.
A
It
is
to
once
you
are
like
really
sure
that
you've
resolved
what
they're
saying
I
I
will
often
put
a
comment
that
says:
I
believe
I
have
captured
exactly
what
is
needed
to
resolve
your
comment.
Here's
how
I
did
that
links
to
the
commit
or
whatever,
please
unresolve
this.
If
that
isn't
resolved,
that's
kind
of
the
the
default,
but
again
it's
a
fine
art.
You
don't
always
do
that
and
sometimes
comments
just
are
too
nebulous.
But
okay.
C
B
C
B
I
personally
own
a
lot
of
comments
that
aren't
resolved
right
now,
so
you
know,
I've
I've
got
a
few
things
ahead
of
this
one
in
my
you
know
in
my
prioritization
right
now,
but
I
will
try
and
get
back
to
it
this
week.
If
you,
if
I,
missed
that,
don't
hesitate
to
ping
me
on
it.
A
A
Roger
that
yeah
and
thanks
for
re-raising
it
please
don't
ever
hesitate
to
raise
a
thing
when
it's
been
sitting
all
right,
1950
the
tester
interface
I
think
Matia
was
taking
a
look
at
this
one.
First
and
I
hadn't
looked
at
this.
Yet
this
allows
us
this
will
allow
users
to
run
conformance
tests
with
custom
logging
headers
and
enable
emitting
important
info,
such
as
timestamps.
B
Yeah
this
is
this
is
coming
from
the
gke
implementation,
just
it
takes
a
while
for
our
test
to
run
and
it'd
be
really
helpful
when
we're
debugging
to
actually
get
some
time
stamps,
along
with
the
the
logs
and
Brian
looked
at
a
few
different
Alternatives
I
think
it
was
John
suggested
this
specific
one,
just
a
custom
interface.
It
seems
reasonable
to
me,
but
I
wanted
to
get
some
other
reviewers
to
make
sure
before
just
going
ahead
and
merging
it.
A
A
B
Yeah,
this
is
a
bug
it
probably
I
might
have.
Maybe
should
have
put
this
as
a
ga
block,
not
not
blocker,
but
we
have
you
know
again:
one
running
conformance
tests
you're
getting
a
lot
of
extra
noise
in
there,
because
we're
printing
out
everything
and
so
garov
fixed
that
with
this
just
basically
removing
the
certs
are
truncating
them,
I,
guess
yeah,
so
yeah
anyway.
I
think
this
is
non-controversial.
Maybe
I
should
have
just
approved
it
myself,
but
I
guess
raising
it
here.
B
S:
cool
yeah:
if
I
don't
hear
anything,
I'll
just
merge
it
tomorrow,
I
guess,
but
it
seems
very
fine.
A
Okay,
we'll
start
there
I'll
come
back
to
this
one
myself
take
a
look
all
right:
spammy
logs.
A
I,
don't
know
why
I
put
them
both
in
my
bad
nope
I
did
that
because
I
opened
the
relevant
issue,
yeah
I
might
have
gotten
lost,
because
we
just
look
at.
C
B
Think
I
think
we're
I
think
you
just
didn't
open
all
of
them
up,
maybe
or
I
don't
know,
looks
like
Candace
has
one
at
the
bottom.
Maybe
we
should
make
sure
we
get
that
yeah.
Let's
do
that.
One
real
quick.
B
D
I,
don't
think
so.
I
think
Nick
went
back
in
and
addressed
a
couple
of
comments
that
more
open,
I
didn't
get
any
feedback
based
on
some
changes
that
I
posted
so
far
as
far
as
I
can
tell
everyone.
A
Involved
in
this,
by
the
way
everyone
who's
been
involved
through
silver,
was
at
kubecon.
So
there
was
some
business,
but
we
can
definitely
start
pinging
and
getting
everybody
to
take.
Another
look
I
think
there's
at
least
one
comment
here
which
I'm
just
going
to
resolve
with
Howard
or
John
Rather,
but
we
can
start
going
like
this
and
be
like
hey
guys,
we're
all
back
from
kubecon
wake
up.
A
D
C
D
I,
don't
think
yeah,
so
so
far,
I,
don't
think,
there's
anything
really
controversial
because
we
just
kept
coming
back
and
saying
this.
Is
you
know
we're
not
talking
about
implementations
at
this
point
and
then
you
know
once
we
are
talking
about
implementations,
there's
there's
going
to
be
some
heat,
but
would
be
good
to
to
get
this
put
away
and.
B
A
B
All
right,
yeah
I,
just
thanks
again
for
for
adding
this
I
I,
should
have
added
it
myself.
This
is
a
pretty
high
priority.
Gap
and
yeah.
E
B
Also,
try
and
take
another
look
at
it:
I'm,
supportive
of
it
and
we'll
try
and
make
sure
any
last
comment.
Threads
here
that
aren't
resolved
do
get
resolved
ASAP
and
we
can
get
this
through.
D
1959,
that
was
another
conformance
test.
One
I
think
we
skipped
just
let's.
A
B
E
A
All
right
and
I
think
for
we're
at
59
minutes
after
the
hour,
so
we
should
probably
just
go
ahead
and
wrap
up.
We
got
through
a
lot
of
triage
again
for
the
second
meeting
in
a
row.
That's
this
is
unusual,
which
is
good,
maybe
a
good
problem
to
have.
Thank
you,
everybody
for
coming.
We
have
not
decided
on
the
EU
friendly
times
that
we're
going
to
do
yet.
So
just
those
aren't
decided.