►
From YouTube: SIG Network Gateway API Bi-Weekly Meeting for 20220919
Description
SIG Network Gateway API Bi-Weekly Meeting for 20220919
A
A
B
Okay
yeah:
this
is
a
bit
of
a
joint
effort
here,
Nick
added
the
first
thing
about
the
Milestone
and
I
just
kind
of
expanded
on
that
I
wanted
to
just
run
through
the
060
Milestone,
we're
getting
awfully
close
and
make
sure
we
were
all
on
board
with
the
major
themes
here.
One
Nick
I
think
the
most
relevant
thing
that
we're
really
trying
to
get
in
is
your
Gap
1364.
yeah.
B
The
big
conditions
overhaul
I'm
excited
to
get
that
one
going
then
reference
Grant,
the
beta
I
know
we
have
a
lot
of
implementations
of
this
I'm
curious
anyone
who's
implemented.
This.
Have
you
gotten
feedback
on
it
has?
Can
you
tell
if
anyone's
using
this
yet.
D
B
Okay,
yeah
I
think
that's.
That
seems
to
be
the
overall
theme
Here
for
those
on
the
call.
If
we
were
to
continue
with
the
plan
of
graduating
this
to
Beta,
are
there
any?
Would
anyone
push
back
on
that.
B
Okay,
it
see
it
seems
non-controversial
enough
that
we
should
just
run
for
run
with
it.
We've
had
this
in
the
API
for
a
while,
it's
broadly
implemented,
so
I
think
we
need
an
issue
to
track
this.
If
we
don't
already
have
it
but
yeah,
we
should
just
get
this
one
in.
B
And
then
next
step
on
here
was
a
grpc
route.
Alpha
I,
don't
know.
If
Richard
is
on
this
call.
Oh
you
are
so
I
think
we
still
need
some
docs
and,
as
a
stretch,
goal
also
conformance
tests,
yeah,
so
I'm.
C
Working
on
that
with
priority
I
think
the
docs
I
should
be
able
to
have
in
by
the
end
of
the
week
and
I'm,
aiming
for
the
end
of
September
for
the
conformance
tests.
B
Awesome
thanks
the
response.
Header
modifier,
that's
a
PR
that
we'll
get
to
in
triage
later
today.
It
just
needs
review,
I
I
feel
bad
I.
It
it
I
think
it's
been
ready
to
go
for
probably
a
week
plus
and
I
just
haven't
had
the
bandwidth
too,
prove
it,
but
if
anyone
else
has
time
to
take
a
look,
that'd
be
great,
otherwise
I'll,
try
and
get
to
it
tomorrow,
but
I
think
this
one
is
just
about
good
to
go
as
well
and
bonus
points.
Sense.
B
Car
I
think
that's
their
name
added
test
to
begin
with
a
conformance
test,
which
is
great,
but
we
still
need
docs
on
this
one,
and
you
know
adding
conformance
tests
before
there's
an
implementation
is
always
a
little
scary,
but
hopefully
this
practice
works.
These
are
opt-in
tests
and
at
least
it
provides
a
framework
for
what
we
think
is
right,
but
maybe
in
the
future
we
need
an
experimental
channel
for
conformance
tests
as
well.
Instead
of
just
opt-in,
yeah
I.
D
B
B
Yeah,
that's
my
take
yeah
the
next
one,
gamma
I,
don't
know
if
anyone's
here
I
can
talk
about
what
we
might
want
to
get
into
the
release
for
gamma.
C
I'm
here,
I
think
so.
We've
spent
the
past
couple
weeks
are
focusing
on
kind
of
like
shifting
targets
and
we've
been
working
on,
defining
the
how
HTTP
route
should
be
used
for
East-West
traffic
events
that
is
kind
of
like
what
we
determined
after
a
few
false
starts
on
different
potential
implementations,
we're
working
on
a
proposal
to
kind
of
move.
The
existing
work
in
progress
stock
towards
a
gap
that
will
I
think
try
to
share
during
the
gamma
meeting
tomorrow.
C
I
think
the
intent
would
be
to
like
get
that
Gap
approved
for
o60
I.
Don't
think
it's
going
to
require
any
code
changes.
It's
really
going
to
focus
on
just
documentation
around
like
use
cases
and
expected
behavior
and
like
maybe
what
status
should
be
set
when.
B
Cool,
that's
great
I
know
it's
got
to
be
difficult
to
get
through
and
build
consensus
on
that,
but
excited
to
see
where
we
land
and.
E
B
Like
you
said,
hopefully,
we
can
get
some
kind
of
like
a
standard
recommendation
pattern
whatever
into
the
API.
C
For
this
release,
that's
that's
the
hope,
but
having
all
the
flexibility
in
it
to
be
able
to
iterate
as
we
move
forward
with
minimal
or
without
breaking
changes,
makes
sense.
B
C
Added
it
I
know
I'm
supposed
to
be
doing
something
for
this
I'm
supposed
to
be
creating
a
proposal
and
I
haven't
done
it
yet,
but
I
would
like
to
make
sure
it
gets
still
gets
in
the
060
if
even
if
I'm
a
little
bit
late
to
the
school.
B
Yeah,
it's
gonna
we'll
have
to
move
quickly
on
it,
but
I'd
love
to
see
it
get
into
right.
You
know,
because
we're
not
the
final
line
of
reviewers
for
this
API.
We
really
need
to
have
everything
ready
early
October
if
we
want
to
have
this
ready
to
get,
and
so
that's
not
just
the
Gap,
but
it's
also
implementation.
B
You
know
a
variety,
a
series
of
things
merged
in
so
I'm,
very
supportive,
just
paying
any
one
of
us
for
help.
If,
if
we
can
help
move
this
forward,
I
think
we
all
want
something
like
back-end
properties
in
API,
so
yeah.
If
we
can
help,
let
us
know.
A
Cool
so
there's
a
couple
of
action
items
in
here.
We
will
be
available
to
you
canvas
as
you
work
on
your
proposal.
I'm,
going
to
take
a
look
at
that
PR
and
then
Rob.
We
just
need
to
double
check
that
we
have
an
issue.
That's
blocking
the
b060
Milestone
anything
else
that
anybody
wants
to
talk
about.
As
far
as
v060
goes
before
we
move
on.
B
Do
have
a
couple
other
items
below
this,
but
I
I
want
to
at
least
my
my
main
goal
here
was
to
make
sure
that
the
major
items
in
the
in
the
o6o
release
are
accounted
for.
So
if
you
see,
if
you
have
really
an
API
change
of
any
size-
and
it's
not
represented
here-
definitely
speak
up,
I
know
we
have
more
than
this
in
the
Milestone,
but
I
want
to
make
sure
that
we
all
understand
and
are
working
towards
the
same
goal
of
this
release.
D
Yeah
I'd
argue
that,
probably
that's
probably
that's,
probably
all
we're
going
to
be
able
to
get
done.
I,
don't
think
that
we'll
be
able
to
get
done
too
much
more
than
what
what
we've
listed
just
there
yeah.
Those
are
large
names,
so
I.
D
B
D
Just
a
this
is
a
reminder
that
there's
a
there's
a
bunch
of
help
wanted
issues.
Some
of
them
are
pretty
straightforward.
I
have
noticed
a
few
more
people,
sort
of
wanting
to
pick
up
issues
and
so
yeah.
This.
This
search
is
shows
the
Milestone
ones
and
yeah
pick
something
that's
not
assigned
now.
So
the
admission
notebook
one
this
doesn't
have
to
be
done
on
o60.
D
It
could
easily
be
done
later,
but
it'd
just
be
nice.
If
someone
had
a
look
and
saw
what
Upstream
Cube
has
available
at
the
moment
for
security
scanning
be
nice
to
have
that
done,
but
I
think
the
publishing
the
the
doc
the
docker
image
to
the
production
registry
would
be
really
good
egos.
Engine
X
has
a
this
I
put
some
links
in
there
to
Ingress
engine
X's
how-to
on
how
they
do
it.
We
probably
just
need
to
update
our
release
processors,
some
extra
stuff,
so
there's
a
mm-hmm.
D
D
This
is
the
Securities.
This
is
the
security
one.
Sorry,
it's
the
second
one
that
one
yeah
yeah
so
right
right
now
you
have
to
pull
this
from
staging
from
the
staging
kubernetes
repo.
D
D
D
A
B
All
right
and
my
last
thing:
yeah:
if
we
can
open
up
the
Milestone,
which
is
linked
all
the
way
at
the
top
I
wanna
should
have
put
it
down
anyway,
but
it
is
a
work
yeah
that
too
I
wanted
to
run
through
some
of
these
and
see.
If
it
were
some,
we
could
take
out
for
sure
I
think
there's
one
for
so.
Okay,
let's
start
at
the
top
I
guess
the
API
specification
I
think
this
is
one
that
we
all
want
to
get
in,
but
I
don't
think
it
needs
to
be
release
blocking.
D
I
think
that
we've
actually
done
most
of
the
work
here.
This
was
yours,
but
I
think
we've
I've
done
a
bit
already
to
yeah
we've.
C
D
Yep
so
there's
two
sort
of
things
that
adding
the
adding
the
adding
the
anchor
links
and
then
moving
the
the
rules
about
how
a
field
how
a
structures
used
to
inside
the
struct
rather
than
to
the
field
containing
a
struct,
yeah
and
so
I
think
we've
done
I
did
that
already
for
backendriff,
yeah,
yeah
and
so
and
I
think
there
might
I
think
the
other
one
was
allowed
it
outside.
C
A
B
D
Too
much
other
stuff
going
on
with
you
know:
status,
back-end
properties,
you
get
my
stuff
yeah.
B
I'm
fine
with
clearing
it
I
I
do
want
it
to
be
next
but
I.
You
know
there's
a
lot
going
on
too.
So
is.
B
So
reference
policy
seems
good
grpc
route.
That's
good,
yeah,
I
think
the
web
hook.
2
prod,
makes
sense
if
we
can
yep
the
other
than
AMD
64
one
someone
assigned
himself
to
that
right.
D
B
D
B
C
B
A
D
B
And
somebody
already
volunteered
for
the
shell
script,
yeah
route
matching
precedence
language.
Can
we
yeah
I.
D
Is
ringing
a
bell,
so
this
is
Sanjay
and
Steve
found
this
when
they
were
doing
Contra
implementations
that
it's
just
that
yeah.
If
you,
if
you
have
a
look,
it
says
each
ties,
it
still
exists
to
cost
multiple
routes.
Matching
presidents
must
be
determined
in
order
of
the
following
criteria.
Continuing
on
on
Fires
is
the
key
part.
D
If
you
just
scroll
down
a
little
bit
more,
it's
like
it's
like
the
number
of
it's
there's
some
confusion
about
whether
one
of
the
typewracking
things
is
how
many
header
matches
there
are
or
like
the
exact
definition
there,
and
so
that's,
Sanjay
and
Steve
were
just
asking
us
to
clarify
that
a
bit
to
make
sure
that
to
make
sure
that
this
is
a
super
clear.
A
D
C
D
Yeah,
it's
just
this
is
just
to
record
yeah
like
you
know,
it's
1370
the
the
right
way
to
do
it
or
do
we
want
to
flip
to
what
Steve
has
suggested
in
the
comment
and
what
Sunday
is
suggesting
here.
B
A
That
puts
me
there
all
right,
so
we
have
a
four
that
are
blank.
A
couple
of
them
are
like
web
hooky
things
that
we
talked
about
kind
of
being
like
nice
to
haves
yeah
that
one's
one
of
those
okay
and
reference
policy
yeah
that
one
has
to
happen.
Okay,
yeah
I,
don't
see
any
more
any
more
grass
to
mow
here.
B
Yeah
I
think
we're
in
a
better
place
and
we
can
keep
on
coming
back
to
this
in
the
next
couple
weeks.
But
this
is
good
evidence
that
there
are
lots
of
places
to
jump
in
and
contribute
if
you've
got
time-
and
you
know
a
reminder
that
we
are
hoping
to
get
this
release
out
pre-cubecon.
So
not
a
lot
of
time.
A
B
Oh
I
forgot
to
attach
my
name
to
that
one,
the
Milestone
one
that
you're
on
okay,
sorry,
but
that
that
is
me,
I
filed
one
more
cherry
pick
and
then
I
think
we're
good
to
go
for
this
release.
I
included
three
things
in
here:
I,
don't
think
any
of
them
are
particularly
controversial.
B
One
is
in
addition
to
a
conformance
test
that
covers
a
gap
and
two
or
just
doc,
updates
that
either
fix
or
clarify
something,
and
then
you
know
o5.
The
release
o5
Branch
already
includes
some
big
bug,
fixes
that
I
would
love
to
get
a
release
out
for
so
my
goal
is
to
get
you
know
once
this
merges
I'll
put
together
a
changelog
branch
and
we
can
get
a
release
out.
Hopefully
this
week.
A
B
Cool
thanks
for
making
an
exception,
yeah.
A
D
Yeah,
so
look
I,
think
we're
getting
closer
here.
I
think
there
was
just
a
few
things
here
that
I
wanted
to
ask
people
about
so
yeah.
If
you
could
pull
this
up
and
then
head
to
the
in
the
change
file
part
way
down,
there
is
a
section:
that's
like
the
that
is
the
set
of
rules.
D
The
proposed
site
changes
summary
one
other
thing
that
I
thought
that
I
did
want
to
ask
about,
though,
is
chain
you
asked
for
for
me
to
clarify
where
I
was
getting
some
of
this
some
of
the
stuff
a
bit.
In
the
background.
From
that
whole
background
section,
the
numbered
list
is
actually
pretty
much
a
straight
copy
summary
of
what's
in
the
API
conventions.
D
Yeah
that
numbered
list
that's
in
the
down
a
bit
further.
A
Yeah
I
think
somebody
mentioned
something
and
I
just
like
I,
just
kind
of
piled
on
that.
We
just
could
put
a
link
in
there
so
that,
if
anybody's
ever
looking,
they
know
where
it's
coming
from.
Well,
that's
where
the.
A
D
Okay,
cool
yeah,
so
yeah.
If
you
could
check
that
one
for
me
and
just
tell
me
what
you
think
I
should
do
there
I
don't
mind,
but
yeah
like
I
kind
of
the
problem.
Is
that
to
really
give
you
the
background
of
of
exactly
what
the
API
convention
say.
You
just
need
to
like.
Well
like
really.
We
could
just
put
the
API
convention
that
section
the
API
conventions
of
a
bad
gym
in
there.
It's
very
pithy,
it's
very
difficult
to
summarize
yeah.
A
No
I
think
I
think
I,
don't
know
exactly
what
I
was
thinking
at
the
moment,
but
right
now,
I
think
all
we
really
need
is
just
a
I
would
just
like
to
see
a
reference
somewhere
so
that,
like
we
can
kind
of
people
can
see
this
and
then
say
we're
doing
this
because
of
the
conventions
that
are
already
set
for
us.
So
if
that's
not
already,
there.
D
A
D
So
if
we,
the
yeah
cause
I,
say
Constance
sort
of
asked
for
a
couple
more
references,
the
this
one's
hard,
because
it's
the
all
of
these
references
are
in
Old,
PRS
and
issues
and
stuff
I
definitely
was
involved.
I
was
part
of
a
proposal
to
API
machine,
read
that
that
I
helped
with
Evan
from
K
native
work
on
a.
B
D
Machinery
reviewers
were
not
in
favor
of
that;
they
didn't
want
it
to
be
too,
like
they.
D
Fact
that
you
need
to
look,
people
need
to
look
at
these
or,
if
you're
handling
them
in
code.
They
need
to
be
like
special
case
handled
right.
You
can't
have
a
generic
implementation
that
handles
these
yeah
I
haven't
looked
at
Constance
comments,
yet
so
I'm
not
sure
about
those
all
right,
so
I
think
yeah.
So
the
the
key
part
here
is
that
that
just
go
back
up
a
bit
is
where
Customs
comment
is
changes.
D
Well,
so
I
think
the
problem
is
the
problem
is
that
for
us
to
do
this
properly,
we
have
to
break
backward
compatibility
because
we're
going
to
be
flipping
the
meanings
of
things,
we're
going
to
be
saying:
hey,
you
know,
do
you
think
so
it's
either
we
break
back
with
ability
to
compatibility
now
or
we
do
it
in
a
future
release.
Okay,.
E
C
D
E
Less
people,
okay,
can
you
explain
what
exactly
is
breaking
because
from
from
what
they've
seen
is
you
know,
there's
no
point
to
remove
some
of
the
field
of
views.
The
world
Fields
don't
need
to
be
removed,
so
that
should
not
be.
We.
D
D
They
are
included
in
the
conformance
tests
right
now,
so
we
will
either
need
to
change
the
update,
the
conformance
test
to
use
the
new
process
or
we
don't.
If
we
don't
update
the
conformance
test
to
use
the
news,
the
new
process,
then
there's
no
point
making
the
change,
because
there,
because
there's
because
we're
not
checking
that
that
people
are
doing
what
the
specs
says
you
should
do
so.
D
The
critical
thing
here
is
updating
the
conformance
test,
because
the
conditions
are
part
of
the
conformance
tests,
and
so
when
we
do
that,
implementations
will
no
longer
be
conformant
until
they
update
their
conformance
tests,
and
so
that's
the
braking
change.
It's
not
that
the
behavior
of
the
API
is
going
to
change.
As
you
say,
it's
status,
but
the
conformance
will
break,
and
so
we.
D
Two
options
that
we
test
in
conformance
you
can't
be
like,
oh
maybe,
is
it
attached
or
accepted.
So
like
that's
what
I
thought
it
was
better
to
get
this
out
of
the
way
and
get
it
done
and
get
the
you
know
and
get
any
breaking
changes
out
over
now
and
be
like
this
is
the
way
forward
and
then
and
then
we're
done.
I.
E
Right
so
I'm,
not
talking
about
you,
know,
changing
the
conferences
and
everything
else.
What
I'm
saying
it
from
from
your
proposal.
What
I've
seen
is
that
you
are
adding
the
new
standard.
I
mean
one
field
that
is
going
to
be
used
and
that's
very
reasonable
and
performance
is
going
to
be
changed
to
verify
that
this
field
is
added,
but
at
the
same
time
there
is
no
need.
E
D
We
can
do
that,
but,
like
that,
like
it
feels
to
me
like
oh
we're,
we
don't
need
to
break
them,
yeah
sure,
but
like
it
would
just
mean
that
those
things
are
checking
a
thing
that
implementations
may
or
may
not
set
right
like
you
know,
they
will
be
doing
a
bunch
of
custom
Behavior,
because
the
spec
says
that
you
upload
once
I've
updated
this.
The
spec
will
say
attached
is
the
way
not
okay,
okay,.
B
I
I
would
just
I
would
just
go
one
step
further,
that
I
I
think
it
is
kind
to
have
some
kind
of
transition
period
where
both
are
present.
I,
don't
think
it's
a
requirement,
but
it
is
certainly
a
a
nice
thing
to
do
and
if
it
just
means
leaving
cons,
you
know
cons
around
and
calling
them
deprecated
and
saying
they'll
be
removed
in
the
next
one.
I
I
feel
like.
That
is
probably
a
good
practice
here.
A
Yeah
is
that
maybe
going
against
the
because,
because
of
how
long
it
has
historically
at
least
taken
for
us
to
release,
is
that
maybe
falling
into
the
problem
that
made
Nick
want
to
do
this
so
soon
like
there
will
be
a
release
where
you
can
do
the
thing.
We
don't
want
you
to
do
and
that
will
last
for
a
while.
D
No
is
it
you're
there
I
think
what
what
Rob
is
saying
is.
Forgive
me
if
I'm
wrong,
but
that
the
key
part
here
is
that
we
don't
remove
the
consts
that
have
the
types
you
know
right
now.
If
you
remove
the
cons,
then
then
things
will
stop
compiling.
If
we
leave
the
console
present
and
mark
them
as
deprecated,
then
you
can
still
use
them
right.
D
Your
code
will
still
compile
you
know,
but
you,
your
conformance
test,
will
fail
because
we
should
be
flipping
the
conformance
test
to
use
the
new
condition
types
all.
B
D
B
I
I'd
actually
go
I'd,
go
two
steps
further
than
that.
Well,
so
I
agree
with
everything
you
said.
What
I
would
also
say
is
we
should
allow
conformance
tests
to
continue
to
pass,
even
if
you're
setting
the
old
conditions
like
performance
tests
can
never
say
you
can
only
set
these
conditions
just
that
you
must
set
these
specific
conditions
that
we're
aware
of,
and
we
can't
have
an
opinion.
D
D
It
attached
right
like,
but
the
spec
says
that
you
have
to
set.
You
know
if
you
said
accepted,
but
the
spec
says
you
have
to
set
attach
right.
You
can
set
attached
and
accepted
fine,
no
problem
but
like
that,
but
the
spec,
the
conformance
tests,
only
care
about
attached
right,
like
yeah
and
so
we'll
leave
the
things
in
there.
People
can
do
that
if
that's
what
they
want,
I
think
it's
a
bit
confusing
for
users
personally,
but,
like
you
know,
as
a
migration
thing
shot
like
you
know,
yeah
yeah.
That
seems
fine
to
me.
D
C
C
C
Effect
that
makes
sense
that's
similar
to
what
we
did
for
reference
policy
too.
I
think
we
were
one
of
the
few
implementations
that
supported
both
reference
policy
and
reference
Grant,
and
that's
exactly
how
we
did
it.
D
Yeah
so
I'll
update
the
Gap
I'll
put
that
in
the
the
Top
Line
changes
we
run
back
to
the
software
changes
I
think
there
was
one
more
question
I
had
about
those
yeah,
so
the
other
question.
The
other
question
is
that
originally
in
this
Gap
I
had
that
so
there's
actually
a
couple
so
originally
in
this
Gap
I
had
it
that
all
the
conditions
that
we
had
would
be
positive.
Polarity
I
think
Rob
convinced
me
that
that's
not
a
good
idea.
D
D
D
D
And
that
double
negative
is
very
confusing
and
difficult
to
understand,
but
having
everything
positive.
B
D
D
I'm
suggesting
in
this
now
is
that
we
have
a
number
of
summary
positive
conditions.
Attached
is
definitely
one
ready,
I'll
get
to
in
a
sec
and
then
I'm
thinking
that
we
might
we
might
in
in
that
discussion
about
ready,
we're
gonna
talk
about
adding
a
new
condition
called
programmed,
but
the
the
one
there's
one
that
that
is
present
already,
that
is
currently
positive.
Polarity,
which
is
resolved
reps
resolved
refs,
is
currently
positive.
Polarity
because
it's
like
all
the
references
are
resolved.
That's
that's.
D
What
resolve
refs
means
so
resolve
ref's,
true
means
everything
is
good.
It
doesn't
matter
if
we
flip
that
or
not
so
I've
left
that
as
positive
for
now.
It's
a
little
bit
confusing
to
explain
in
this,
but
I
think
it
should
be
fine,
but
the
intent
is
that
for
any
other
type
thing
that
would,
that
is
a
problem
with
your
config.
That
does
not
make
your
resource
unattached
from
the
thing
that's
attached
to.
That
should
be
surfaced
with
a
more
specific
error
condition.
D
So
there's
a
lot
of
sort
of
nuance
here,
like
HTTP
route,
is
very
special
here
because
and
the
rule
that
I'm
trying
to
use
here
is
that
if
it
doesn't
produce
any
config
in
the
underlying
data
plane,
then
it
shouldn't.
Then
it
it
shouldn't
be
attached
right.
So.
D
Going
to
be
produced
in
the
underlying
data
plane,
then
the
thing
you
shouldn't
attach
there's
some
business
questions
there
about
like
there
should
be
no
valid
config
and
I'm
like
well.
Yes,
I've
tried
to
update
most
of
the
references
there
to
say
that,
but
like
the
idea
here
is
that
a
couple
of
people
had
sort
of
asked
for
for
us
to
have
like
a
separate
condition.
That's
like
this
thing
is
semantically
valid.
D
B
D
Things
that
are
semantically
invalid,
so
there's
no
there's
no
point
having
like
an
extra
condition
that
just
represents
that
the
thing
about
attachment
is
and
the
the
thing
that
I'm
really
thinking
about
with
attachment
here
is
when
you
have
a
HTTP
route,
if
you
make
a
mistake,
like
you
reference
a
service
that
doesn't
exist
or
you
know,
reference,
a
filter
that
doesn't
exist
or
filter
is
incompatible.
Like
there's
rules
in
the
thing
that
say,
you
need
to
program
500s
in
to
replace
the
traffic
that
would
match
those
things.
D
That
then
means
that
all
of
those
problems
that
would
otherwise
cause
the
thing
to
become
attached
and
produce
config
in
the
data
plane,
so
they
don't
cause
the
HTTP
to
become
attached
in
the
case
of
TCP
route.
If
you
don't
have
a
back-end
service,
you
there's
no
logical
way
that
you
can
program
that
rule
into
the
data
plane,
so
that
will
that
will
cause
the
TCP
route
to
become
unattached,
and
so
that's
the
that's.
Why
I'm
sort
of
using
that
rubric
as
the
as
the
way
to
decide
what
becomes
attached
and
what
doesn't
so.
C
D
E
Any
chance,
instead
of
polarity,
positive
negative,
knows
that
we
introduced
an
explicit
sort
of
condition,
error
or
error
condition
or
dedicated
types.
That
is
so
it's
clear
that
it's
an
error
and
that's
negative,
followed
it
in
your
document.
D
I
mean
yeah
so,
like
I
think
that
in
the
API
we
should
refer
to
them
as
error
States
for
sure,
but
they
are
represented
by
our
condition,
with
negative
value
type.
D
You
know,
and
it
has
a
very
it
has
a
name
and
a
set
of
behavior
that
you
get
from
Upstream
and
the
the
polarity
is
about
like
what
happens
when
the
thing
is.
That
is
true,
so
it's
like,
like
I'm
using
it
in
a
very
technical
sense
to
to
but
I
100
agree
that
those
should
be
no.
We
should
consider.
E
D
I
think
that
this
is
this
is
to
use
the
same
phrase
that
Chinese
do
it
like.
As
far
as
Upstream
kubernetes
is
concerned,
these
are
the
way
this
is
the
way
that
you
do.
This
is
the
way
that
you
put
status
information
that
you
don't
want.
That's
not
like
extremely
specific
to
the
to
your
API,
it's
okay,
yeah
yeah.
D
So
people
should
so
conditions
are
relatively
new
to
the
Upstream
API,
but
they
are
the
way
that
that
you
should
be
expecting
to
check
if
your
resource
is
working
or
not
they're
not
widely
used
across
crds.
Yet,
but
that's
because
they're,
real
relatively,
they
are
heavily
used
on
node
on
pod
and
on
deployment
like
online
sources.
So
I
think
that
people
should
be
pretty
familiar
with
them
and
if
they're
not.
D
D
Okay,
so
I
guess
the
things
that
I
want
to
erase
here
are
like
this
is
why
I'm
trying
to
do
this
and
the
you,
the
negative
polarity
conditions
thing
is
important:
I
haven't
specified
many
negative
polarity
conditions
here,
because
I
kind
of
don't
want
to
have
to
specify
them
all,
and
so
the,
but
the
intent
is
definitely
that
you
know
if
there's
something
that
would
that
does
make.
That
is
a
problem
but
doesn't
make
things
unattached.
Then
there
should
be
another
condition
that
appears
on
there
when
that,
when
that
is
true,.
C
D
So
yeah
I've
got
I
absolutely
not
costing.
You
absolutely
can
add
your
own
conditions.
It
is,
however,
expected
that
if
you're
adding
your
own
conditions,
you
use
the
domain
name
spacing
to
be
able
to
add
your
own
conditions.
So
you
have
like
that.
That
is
actually
again
in
the
in
the
Upstream
gold
lens
of
the
the
condition
type
is
allows
a
namespaced
like
a
domain
namespace
type,
so
you
can
put
yo
project
contour.io,
custom
condition
right,
that's
an
acceptable
type
of
condition
and
it's
kind
of
expected
yeah.
D
So,
okay,
yeah,
I'll,
I'll,
add
something
in
there
as
well.
That
implementations
may
add
their.
D
They
they
need
to
be.
You
know,
domain
prefix,
strings,
yeah,.
D
They're
supposed
to
be
like
that,
they
don't
have
to
be.
Obviously
our
spec
will
is
only
going
to
mandate
certain
names
and
we'll,
and
the
conformance
test
will
check
for
those
names,
so
yeah
like
and
practically
yes
and
the
intent
is
that
if
you,
if
lots
of
people
are
using
similar
domain,
namespace.
B
B
Yeah
this,
this
is
great
all
around
I,
just
want
to
you
one
thank
you
for
pushing
this
through,
but
two
I'm
trying
to
see
like
we
have
a
tight
deadline
here,
and
this
is
huge
change.
Would
it
be
reasonable
to
put
some
kind
of
deadline
on
feedback
discussion
around
this
people
would
just
want
to
get
this
in.
D
Yeah
I
think
that's
very
reasonable.
I'd
certainly
like
to
see
us
sort
of
maybe
like
any
comments,
need
to
be
in
sort
of
the
next
24
48
Hours
and
then
so
that
I
have
a
chance
to
fix
everything
up
and
then
get
this
in,
because
then
yeah
it's
going
to
take
me
a
while
to
actually
do
the
implementation,
because
there's
a
lot
of
changes
here,
you
know
lots
of
conformance
tests
are
going
to
be
needed
to
be
updated.
There's
like
lots
of
spec
stuff
is
going
to
need
to
be
updated.
B
Yeah,
okay,
so
let's
let's
aim
on
comments
in
in
48
hours.
We
can
mention
that
in
slack
on
the
pr
wherever
and
I'll
I
know,
I
am
due
to
do
another
round
of
review
on
this.
D
The
rest
of
the
my
day
of
your
evening
tomorrow,
tomorrow,
your
time
I,
should
have
a
change
done
on
this
to
sort
of
posts.
The
updates
that
we've
just
talked
about
to
to
this,
so
there
should
be,
it
should
be
reviewable
tomorrow
morning.
Your
time.
C
One
of
the
bigger
questions
I
had
on
this.
It
seems
that
we
would
lose
the
ability
to
represent
something
where
the
syntax
is
valid,
when
it's
unable
to
configure
the
underlying
data
plane.
There's
a
couple
examples
where
we
have
things
like
that,
whether
it's
like
Port
unavailable
or
Gateway,
or
a
few
other
things
given
the
like
Reliance
assume
the
Reliance
from
the
validator,
is
that
okay,
like
in
some
ways,
I
think
it
actually
makes
the
ux
a
little
easier
because
you
have.
C
D
Exactly
so,
if
the
Gateway
can't
be
configured
on
the
underlying
data
planes,
then
it
should
not
attach
successfully
to
the
Gateway
class,
but
that's
that's
the
idea
that
attached
is
the
thing
that
you
look
at
to
be
like
you
is
this:
is
there
some
fundamental
problem?
D
You
know
that
that
you
can't
that
can't
be
resolved
like
you
know,
or
is
it
like?
Is
there
some
small
syntax
error
right
like
if.
D
And
it
says
attached
true,
then
you're
like
okay
things
here
are
mostly
okay,
like
that's
the
that's
the
intent,
yes
and
and
absolutely
it
could
become
attached
unattached
later
absolutely
and
that
that
that's
one
of
the
reasons
why
we've
been
really
careful
with
the
reference,
Grant
and
HTTP
route
is
that
we
didn't
want.
You
know
a
reference
Grant
being
removed
to
like
have
a
huge
blast
Radiance.
D
D
Rather
than
have
the
cause
the
HTTP
route
to
become
attached,
there's
a
bunch
of
rules
there
as
well
about
like
what
happens
if
you've
got
three
three
rules
inside
of
HTTP
route
and
one
of
them
works
and
the
other
two
don't-
and
you
know,
there's
a
bunch
of
other
stuff
like
that
that
I've
needed
to
think
about
a
lot.
So.
B
Thank
you,
okay.
This
is
a
big
change
and
I
wanted
to
highlight
it
and
make
sure
it
is
reasonable
and
it's
something
we
want
to
do
so.
This
change
that
I
made
has
two
parts.
One
of
it
is
setting
the
default
storage
version
to
V1
beta
one,
that's
I
think
uncontroversial.
That's
every
recommendation
ever
the
other
is
making
the
decision
not
to
serve
V1
Alpha
2.,
that's
something
that
is
widely
recommended
by
API.
Machinery
drop
your
Alpha
versions
as
soon
as
reasonable.
B
What
that
means,
though,
is
if
your
implementation,
like
many
or
most,
has
been
reading
or
writing
to
V1
Alpha
2,
it's
not
going
to
work
with
these
crds.
If
we
release
has
the
current
API
is
designed
or
the
current
crds
are
designed,
that's
a
significant
change.
B
We
can
move
forward
as
exists.
This
is
you
know.
For
example,
the
GK
implementation
has
decided
just
to
read
and
write
from
V1
beta
1
and
move
forward
with
that.
But
I
recognize
that,
as
just
one
of
many
implementations
involved
here,
the
rationale
for
dropping
Alpha
2
as
from
a
served
version,
is
that
I
you
don't
want
to
get
a
large
set
of
new
Alpha
users
and
it
is
an
alpha
API
version
and
we
should
drop
it
as
soon
as
reasonable.
B
With
that
said,
I
know
this
particular
transition
could
cause
for
a
painful
update
process,
upgrade
process
for
other
implementations,
so
yeah
just
wanted
to
start
a
discussion.
Nick
I
think
you're
up.
First.
D
D
And
that's
I
think
that's
the
key
right
like
it's
only
for
those
resources
like
yeah
yeah.
We
really
want
to
encourage
you
to
move
to
the
beta
version
of
this,
of
the
ghost
rocks
as
soon
as
possible
for
your
implementation,
the
idea
behind
leaving
the
V1
Alpha
ones
be
served
is
that
it
means
that
people's
tutorials
and
stuff
don't
stop
working,
but
like
once,
you
remove
the
V1
Alpha
2
versions
from
being
served.
Everyone's
gonna
have
to
update
the
yamls
to
apply
it
to
bov
1B
to
one
that's.
D
The
API
version
at
the
top-
not
you
know
not
anything
else,
because
the
spec
is
the
same,
but
it
does
mean
that
we
can
remove
those.
You
know
we
can
do
this
type
of
Alias
thing
to
sort
of
make
it
so
that
those
V
well,
we
could
do
the
type
of
earliest
thing,
but
the
the
idea
here
is
that
those
V
there's
no
V1.
E
D
B
So
yeah
so
on
a
note,
I
linked
out
to
the
new
type
aliasis
thing.
This
is
something
that
John
Howard
did,
which
he
was
thinking
about,
how
they
were
going
to
do
this
for
istio,
and
this
I
thought
was
a
pretty
clever
implementation.
B
If
you
look
at
the
files
changed,
it
really
is
just
type
aliasing
for
every
type
that
is
shared
between
beta
1
and
Alpha
2.,
which
should
make
it
much
much
simpler
if
you
are
trying
to
interoperate
between
using
V1
beta
1
for
some
resources
and
V1
Alpha
2
for
others,
which
I
expect
all
of
us
are
going
to
have
to
deal
with
yeah
that
that's
all
so
I'm
I'm
hoping
this
will
be
less
painful
than
it
once
was.
B
But
it
is
a
thing,
and
my
biggest
concern
here
is
that
if
implementations
are
reading
and
writing
from
alpha
two
today,
they're
going
to
have
to
add
some
kind
of
documentation
saying
hold
on
wait
till
you
install
the
latest
version
of
crds,
because
we
haven't
updated
to
support
them.
Yet
something
like
that
or
get
a
release
out
before
Gateway
API
releases.
A
For
a
little
bit
of
feedback
at
Kong,
we
have
two
Gateway
API
implementations
and
we
have
not
released
the
beta.
We
have
feature
gates
for
Gateway
apis.
We
have
not
released
the
beta
version,
we're
about.
A
To
release
the
beta
version
for
just
feedback,
like
we
already
kind
of
said,
yep
and
tore
it
out
and
replaced
all
of
our
examples
and
so
forth
and.
B
B
B
And
Mike
asked
the
question
in
chat,
so
V1
Alpha,
2,
strucks,
so
yeah.
Absolutely
you
can
still
use
V1
Alpha
2
structs.
The
problem
is
when
you're
making
the
actual
API
calls.
If
you're
using
v060
crds
that's
going
to
fail,
you
have
to
use
the
V1
beta,
1,
API
call,
which
would
mean
V1,
beta1,
structs,
yeah.
D
B
I
think
I
think
this
required
I,
don't
know
what
the
best
action
here
is,
whether
I
should
or
we
should
have
a
discussion,
an
issue
something
to
track.
This
I
want
to
make
sure
that
this
does
not
come
as
a
surprise
to
anyone,
but
I'm
also
not
hearing
any
compelling
reason
to
hold
off
on
this
change.
C
So
sorry
yeah,
so
today,
if
I
using
V1
Alpha
2,
does
that
mean
is
not
going
to
work?
Okay,
once
you
make
this
change.
B
So
with
the
crds
that
are
released
in
v0.6.0,
so
in
a
month
or
so,
what
would
need
to
happen
is
instead
of
using
like
if
your
controller
is
using
V1
Alpha
2
apis
you'd
have
to
just
do
a
find
and
replace
and
use
V1
beta,
1.
A
B
I
think
probably
I
want
to
make
sure
I
wanna.
What
I
want
to
do
is
is
start
with
a
smaller
group.
First
make
sure
we,
as
you
know,
most
active
implementers,
are
on
board
with
it
and
then
once
we
are,
then
we
can
send
it
as
more
of
an
announcement
than
a
question,
but
I
just
want
to
make
sure
that
yeah.
D
One
thing
that
I
just
thought
of
robies:
if
we
stop
supplying
the
V1
Alpha
2,
CID
definitions
and
they're
already
installed
in
a
cluster,
won't
they
just
stick
around.
B
So
what
yeah
good
good
question,
so
the
key
Point
here
is
that
V1
Alpha
2
is
still
included
in
the
crd.
It's
just
not
stored
or
served
it's
just
it's
just
sitting
there
for
the
purpose
of
conversion.
For
the
purpose
of
anything
else.
You
can't
actually
write
to
it.
You
can't
use
Alpha
2
for
anything,
but
it
still
exists
there.
So
you
can
do
kind
of
a
seamless
upgrade
from
040
crds
all
the
way
up
to
060
without
things
breaking.
B
Okay,
it
seems
like
we're
okay,
here,
I'm
fine
to
move
on
to
next.
Thank
you,
I
think
it.
Thank
you
yeah
Shane!
Thank
you
for
taking
all
these
notes.
A
Yep
no
problem
all
right:
Rob
contributor,
Summit
cfp
submitted
okay.
B
B
Yeah,
thank
you
yeah
sure.
A
Do
we
we
already
know
that
we're
going
to
come
back
around
to
this
one
yeah
I
know
I
am
at
least,
and
we
talked
about
it
earlier,
so
this
is
kind
of
the
only
one.
We
really
need
to
talk
about
right.
B
I
mean
this
is
self-selected.
There
may
be
other
things
on
there
that
we
should
cover.
It
may
be
easier
just
to
look
at
the
files
change,
because
this
is
a
pretty
tiny
one
with
this
is
kind
of
making
room
formally
making
room
for
next
changes
and
incoming
conformance
tests.
B
I
think
we've
talked
about
this
in
meetings
we
just
hadn't
formally
documented
it
basic
idea
is:
we've
said
in
patch
releases.
We
are
okay,
with
expanding
conformance
test
coverage
to
cover
things
that
are
already
in
the
release.
The
you
know
the
spec
is
already
there.
We
just
don't
have
good
coverage,
so
we're
going
to
expand
that
that
sounds,
like
we've
said,
is
in
scope
for
patch
and
then
for
minor
versions.
B
We're
saying
that
we
can
make
changes
to
the
recommended
recommended
conditions
to
be
set,
and
we
can
change
conformance
tests
to
match
updates
to
the
spec,
which
really
is
just
a
status
at
this.
C
Point
makes
sense.
A
C
C
A
A
Is
it
yeah?
It
might
be
cool
well
for
those
of
you
who
are
have
some
time
this
afternoon,
the
next
couple
days
there's
lots
of
stuff
to
review
and
a
couple
things
to
pick
up
for
v060.
If
you
need
any
help,
we
really
appreciate
you
jumping
in
on
some
of
those
issues
that
have
been
tagged.