►
From YouTube: SIG Gateway API GAMMA Meeting for 20221101
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Hello,
everybody
Welcome
to
the
November
1st
occurrence
of
the
Gateway
API
gamma
meeting,
just
like
every
other
kubernetes
organization
meeting.
This
is
governed
by
the
kubernetes
code
of
conduct,
so
B2B
nice.
A
So
let
me
just
remember
what
I
need
to
go
ahead
and
share
my
screen.
We've
got
a.
What
do
we
have
an
agenda
today?
If
you
haven't
already,
please
go
ahead
and
add
your
name,
so
we
can
have
a
kind
of
a
record
of
who's
showing
up
to
these
things.
This
is
an
open
Agenda.
So
if
you've
got
an
agenda,
audio
you'd
like
to
add
everybody
should
be
able
to
edit
this
list.
So
no,
oh,
that's
out
of
the
way.
A
minute
ago.
B
So
it
looks
like
we
have
a
couple
of
new
faces
here.
If
anyone
who
show
up
to
this
meeting
for
the
first
time
wants
to
take
a
moment
to
introduce
yourself
we'd
love
to
be
able
to
get
a
chance
to
meet
y'all.
C
I
can
go
first,
I'm,
Tyler,
French
I
work
at
Le,
node
and
just
kind
of
hoping
that
we
can
migrate
some
stuff
to
Gateway
API.
We
use
a
lot
of
project
Contour
right
now,
so.
A
Awesome
great
to
have
you
have
you
anybody
else.
D
I'll
go
next
hi
I'm
Marco
I'm,
an
engineer
at
tetrate
I'm,
just
curious
to
see
how
the
Gateway
API
will
intersect
with
the
gamma
apis
and
yeah
would
love
to
help
out
in
any
which
way.
I
can.
A
Awesome
yeah.
The
short
answer
to
that
question
is
hopefully
they're
the
same,
but
we'll
we'll
see
anybody
else
like
to
introduce
themselves.
A
Right
great,
so
let
me
do
the
screen
sharing
thing.
Can
you
let
me
know
when
you
can
see
my
screen,
so
I
can
confirm
that
this
works.
A
It
always
forget
when
I'm
on
this
is
that
better
yeah.
E
A
You
all
right,
no
problem
all
right,
so
our
first
agenda
item
I,
don't
know
if
anybody's
owning
it.
It's
just
a
a
good
thing
to
start
on
because
was
last
week
there
was
a
couple
of
there
I
think
six
total
Gateway
API
related
presentations
during
food
con,
which
is
fantastic,
John
and
I,
did
one
specifically
on
gamma,
and
you
know
if
anybody
has
any
other
conversations
that
they
had
around
API
gamma
during
kubecon
to
share.
We
can
do
that
or
I
can
share
some
things.
A
B
Yeah
not
directly
demo
related,
but
I
was
definitely
interested
and
would
recommend
anyone
who's
interested
in
thinking
about
auth,
stuff
or
the
future
of
gamma
to
once
the
recordings
are
available
catch
the
admin
Network
policy
talk.
There
was
two
Folks
at
Red
Hat
I
liked
it
on
the
names
of
it.
But
it
was
a
really
interesting
talk
about
this.
Your
project,
that's
being
run
similar
to
Gateway
API
and
like
an
off
tree
project,
that's
focused
initially
on
East-West,
Network
policy
and
I.
B
Think
between
that
and
it
was
it
Alex
from
buoyant,
gave
a
talk
on
how
they're,
using
a
like
off-spec,
experimental
implementation
of
a
Gateway
API
for
Office,
Depot
policy
and
Linker
d
for
HTTP
and
L7
rules.
So
I
think
between
the
two
of
those
there's
a
lot
of
interesting
things
to
think
about
about
what
a
gamma
implementation
of
C
might
look
like
in
the
future.
A
Awesome
thanks
for
the
references,
Mike
I,
think
the
recordings
should
be
posted
in
a
couple
of
weeks.
I
believe
it's
the
timeline,
so
definitely
those
if
off
Z
is
something
interesting
to
you.
I
know
we're
going
to
be
tackling
that
and
we're
going
to
start
attacking
a
lot
of
that
here
and
Gamma
then
take
a
look
at
those
talks.
It
should
be
helpful
anything
else.
Anybody
wants
to
add.
A
A
A
Someone
was
asking
about
a
maybe
it
was
the
upgrade
story
of
their
mesh
or
what
have
you?
Oh,
it
was
metrics,
people
I
think
got
a
couple
of
questions
actually
between
contribute,
Summit
on
Monday
and
then
Ron
and
I's
talk
mine
and
John
stock
on
Friday
asking
what
the
possibility
of
a
standardization
of
metrics.
A
This
person,
in
particular
on
Friday
I,
was
talking
to
had
multiple
their
their
engineering
teams
had
multiple
measures
that
they
were
using
throughout
multiple
clusters
and
it
made
it
difficult
for
their
platform
team
and
sres
to
know
okay,
when,
if
I
want
to
look
at
the
total
number
of
requests
which
metric
should
I
carry
again,
there's
a
bit
of
a
desire
for
a
standard
set
of
metrics
I,
I
I,
mentioned
kind
of
some
anecdotal
experience.
A
I
had
with
standardizing
metric
with
SMI
and
some
of
the
kind
of
inherent
difficulty
with
that.
Just
because
of
you
know
the
existing
work
streams
that
are
present
with
open
Telemetry
but
I
I
think
I'll.
Just
repeat
what
I
said
here:
I
think
one
of
the
benefits
of
gamma
and
what
we're
doing
here
is
that
we've
got
a
lot
of
meshes
in
the
same
room
already
working
on
standardization
and
those
kind
of
problems.
A
So
if,
if
metrics
turned
out
to
be
something
interesting
to
people
down
the
line,
this
is
a
good
starting
point
for
folks
to
to
start
having
those
conversations
so
yeah,
that's
what
yeah,
open
Telemetry
is
kind
of
where
I
imagine
some
of
that
work
starting,
but
it
doesn't
feel
like
the
lowest
hanging
fruit
at
all
right
now,
given
some
of
the
other
things
that
we
have
had
come
up
so
yeah,
that's
one
conversation
I
had
anybody
else.
F
A
Yeah
when
it
comes
to
Gateway
API,
specifically
I
mentioned
this
during
the
Gateway
API
call
yesterday,
but
I
took
a
video
of
Rob
and
Shane's
talk
on
Thursday
or
is
it
Wednesday?
When
was
your
talk?
Bro
Thursday,
Thursday,
yeah
I
took
a
video,
and
the
entire
room
was
just
absolutely
packed
of
people
wanting
to
learn
more.
It
was.
It
was
pretty
crazy
to
see
so.
A
A
lot
of
people
in
the
community
are
interested
and
and
what
we're
doing
and
the
cool
thing
about
this,
actually
is
that
I
heard
from
I
think
Chris,
a
I
think
around
60
of
kubecon
attendees
this
year
were
newcomers.
It
was
like
it
was
their
first
time
attending
a
kukon,
so
we're
not
the
the
the
hype
and
the
interests
and
the
things
that
we're
doing
isn't
just
people
in
the
community
for
a
long
time
excited
for
things
to
be
easier
than
it
has
been.
These
are
new.
A
These
are
first-time
coupon
attendees,
who
are
feeling
the
pain
points.
I
had
several
end
user
company
Representatives
that
come
up
to
me
and
after
the
game
of
talking
and
Rob's
up
and
say,
Hey,
you
know
these
are
things
that
we're
dealing
with
actively
from
a
user
perspective.
Anything
you
can
do
to
make
this
simpler
would
be
fantastic.
They're
really
excited
to
see
this
so
I
think
the
feedback
for
what
we're
doing
you
know
I
think
initially
gamma.
A
We
started
it
as
like
a
three
month
experiment,
but
we've
gotten
so
much
feedback
at
this
point
it's
kind
of
like
a
resounding
yes,
this
is
needed,
so
that
was
kind
of
my
takeaway
from
groupcom.
A
E
This
is
kind
of
a
sad
well
first,
it
was
great
to
see
everyone
so
Nick.
We
missed
you
and
everyone
else
that
couldn't
make
it.
We
we
missed
all
of
you,
but
on
that
note
the
cfp
for
kubecon
EU
is
way
too
quickly
approaching
so
I'd
love
to
coordinate
with
everyone
on
getting
some
talks
in
you
know,
if
you
have
a
talk
idea
that
you
might
want
to
share,
I
would
love
to
get.
A
Yeah,
absolutely
I
personally,
probably
won't
be
able
to
make
it
to
cook
honey
you
this
year,
just
because
I
actually
don't
have
a
passport
yet,
and
it's
a
pretty
long
turnaround
time,
so
I'm
probably
going
to
make
that
stretch
but
I'd
be
happy,
and
this
goes
through
I.
Think
for
any
of
the
leads
or
maintainers
of
Gateway
API.
If
anybody's
interested
in
giving
a
talk,
please
consider
me
a
resource
to
you
know:
throw
ideas
off
of
or
help
put
a
cfp
together,
a
lot
of
helping
folks
do
that
kind
of
thing.
A
G
It's
Louis
but
I've
hijacked.
His
screen
I've
also
slipped
into
his
DMS
two
things:
Keith.
You
can
get
your
passport
in
under
four
weeks
if
you
book
your
flight,
just
FYI,
because
I've
done
that
so
many
times
before
that
it's
not
even
funny
kind
of
related
to
that
Telemetry
question
that
it
came
up.
I!
H
And
it's
my
screen:
I
I
mentioned
open
Telemetry
is
there
I
mean
open?
Telemet
is
a
cnca
project
that
also
is
trying
to
standardize.
Telemetry
I
mean
the
names
of
metrics
how
to
collect
them
and
everything
how
to
process
them
and
everything
else
and
they
Define.
H
You
know
a
way
to
instrument
applications
which
is
pretty
powerful
and
and
probably
more
powerful
than
what
we
can
do
with
the
mesh,
because
we
as
a
mesh,
you
can
watch
the
traffic
if
it's
not
encrypted,
but
if
instruments
application
directly,
you
get
kind
of
direct
Telemetry
and
more
efficient
and
far
more
data.
So
there
is
an
intersection.
H
F
I'll,
just
I'll
just
hold
up
my
hand,
so
that
you
can
all
imagine
that
XKCD
about
standards
appearing
there,
Tick
Tock
stuff,
so
yeah
I
think
Lewis,
Point's,
really
fair,
that
adding
metrics
is
a
big
chunk
and
and
makes
a
big
statement
about
what
the
API
is.
For.
F
Personally,
my
feeling
is
that
we
should
we
got
more
than
enough
work
to
do
just
to
sort
out
the
networking
stuff
and
we
should
keep
that
scope
tightly
defined
and
and
be
like.
Those
are
very
interesting
discussions,
but
this
is
not
you
know.
This
is
not
the
place.
F
You
know
and
I
think
that
there's
absolutely
discussions
worthy
worthy
of
being
had
about
like
how
you
might
want
to
write
like
policy
style
things
to
do
that
sort
of
stuff.
But
again
this
is
not
the
place
to
do
that.
I
think
yeah,
like
I
I've,
got
a
document
that
I
am
working
on
with
some
updates
to
the
policy
attachment
documentation.
I
really
want
to
watch.
F
I
hadn't
seen
talk
on
the
schedule,
so
I'm
gonna
go
watch
that
because
yeah
I
haven't
seen
a
lot
of
people
use
it
yet,
but
that's
the
document
where
I'm
talking
about
an
updated
policy
attachment
please
everyone
review
that
and
comment
on
it.
I
feels
like
a
few
people
have
reviewed
it
and
it
feels
close
like
I'm,
ready
to
start
yeah,
basically
to
tldr.
F
This
I
want
to
start
I
want
to
make
it
I
want
to
change
that
documentation
so
that
we're
clear
that
you
meta
resources
are
resources
that,
like
attach
to
some
other
resource
and
and
augment
costan
I
picked
a
different
word:
I'll,
augmented
function,
and
you
know,
and
policy
attachments
are
a
meta
resource,
but
not
all
meta
resources
are
policy
attachments.
Reference
Grant
is
also
a
meta
resource
and
stuff
like
that.
F
I
just
want
to
put
some
language
in
there
so
that,
as
we
talk
about
things,
it's
more
clear
that
policy
attachment
is
more
about
things
that
have
a
hierarchy
and
having
settings
so
up
and
down
so
anyway.
I
don't
want
to
derail
this
too
much.
I
just
wanted
to
put
that
in
there,
so
that
I
can
get
some
more
eyes.
D
F
It
I'd
very
much
appreciate
some
more
people.
Some
awesome
time
spent
having
a
look
at
that
thanks.
A
Yeah
I
want
to
also
plus
one
Louise
comment
and
mix
as
well
so
coming
from
SMI
land,
and
you
know
where
SMI
was
a
is
a
full
specification.
A
You
know,
metrics
is
something
that
we
kind
of
try
to
do
and
it
is
best
to
be
more
complicated
than
people
may
may
expect
it
to
be,
because
there
are
also
two
layers,
the
one
that
Kaufman
said
is
okay.
How
do
we
actually
instrument
the
application
to
get
these
metrics
and
then
two,
which
is
the
slightly
easier?
A
It's
probably
easier
technical
question,
but
the
harder
human
question
is:
how
do
we
get
all
these
message
meshes?
Who
use
the
same
metric
names
across
Prometheus,
Hotel,
Etc
and
I
I,
agree:
I,
don't
think!
Gamma
is
the
place
for
that
because
we
have.
You
know
our
place
full
with
existing
networking
things
I
I.
Would
it
would
be
great
to
see
somebody
kind
of
champion
that
work
kind
of
parallel
to
this,
but
here
does
not
block
the
place
to
to
do
that.
A
Okay,
awesome!
Let's
go
ahead
and
move
to
the
next
topic:
Nick
I'm!
Guessing
you
don't
want
to
talk
about
the
policy
attachments,
Resources
stock.
Again,
it's
just
nine
dollars
to
me
taking
notes.
B
A
Guess
appreciate
it
always
as
always:
Mike
John,
if
you're
there
performance,
testing
agenda
item
I,
believe
this
is
a
holdover
from
yesterday's
Gateway
API
meeting
John.
If
you're
you're,
talking
I
cannot
hear
you
I,
don't
know
if
it's
just
me
or
not,.
A
And
now
I
don't
see
anything
at
all
I'll.
Let
John
work
on
his
AV
stuff,
guys
here
for
this
topic
yesterday,
so
I
can
kind
of
give
a
a
summary
of
what
we
talked
about
yesterday.
As
he's
getting
Navy
stuff
worked
out,
basically
John
I
believe
yeah
John
has
a
issue
open
in
the
Gateway
API
repo
detailing
the
need
for
mesh
conformance
tests.
A
A
Performance
tests
really
do
help
make
sure
that
we're
describing
the
kind
of
behavior
that
we're
expecting
and
having
that
early
is
is
really
useful.
However,
there
are
some
incompatibilities
is
the
best
way.
I
can
think
to
put
it
between
how
the
Gateway
API
currently
things
with
performance
tests
and
how
we
would
need
components
to
test
for
the
mesh
use
case.
Gateway
API,
currently,
you
know,
stands
up
a
great
cluster.
A
You
use
a
service
type
load
balance
for
usually
to
expose
a
Gateway
of
sorts
and
then
requests
are
made
using
the
external
IP
of
that
service
type
of
load
balancer
replicating
the
Gateway
in
order
to
look
for
particular
Behavior
with
mesh.
That's
not
good
enough
because
you
know
in
a
mesh
configuration
you've
got
services
and
pods
talking
directly
to
one
another.
So
you
know
you
need
some
way
to
orchestrate
those
requests
use
of
different.
A
You
know
different
protocols,
things
of
that
nature,
so
that's
kind
of
the
background.
The
context
of
the
agenda
item.
John.
Are
you
able
to
speak
now.
I
I
think
so,
can
you
hear
me?
Yes,
I
can
okay
cool
yeah,
so
if
I
kind
of
the
one,
if
we
briefly
talked
about
this
but
like
the
thing
that
we're
trying
to
figure
out
is
like
okay,
what
do
we?
What
do
we
actually
use
to
do
this?
So
you
know,
obviously
we
are
testing
HBO.
So
we've
made
a
solution.
I'm
sure
all
the
other
meshes
have
a
solution.
I
So
I
was
looking
to
see
if
there
was
opinions
on.
If
we
should
build
our
own
solution,
we
should
copy
one
from
a
mesh.
The
only
one
I'm
familiar
with
is
person
who's
used
to
you.
So
I
can't
you
know,
compare
it
to
others,
but
if
others
have
implementations
that
they
like
or
don't
like
or
want
to
use
or
don't
want
to
use.
That
would
be
interesting.
F
You
know
yes,
I,
don't
think
it's
worth
to
use
to
use
another
construct
that
I've
used
many
times
before.
I,
don't
think
it's
worth
spinning
Innovation
tokens
on
writing,
CI
to
link
no.
I
I
definitely
don't
think
so.
I
just
didn't
want
to
like
be.
F
Yeah
I
appreciate
you
doing
the
right
thing
here
and
not
being
like.
Well,
we
got
this
one
does
everything
we
should
just
use
that
yeah
yeah
yeah
but
I
think
like
it
doesn't
seem
worthwhile
Reinventing
the
wheel
for
me.
A
Yeah
I'm
I'm,
stealing
that
construct
nick
the
the.
If
not
oh
yeah,
it's
not
not
a
great
use
of
time
right
now
and
I.
You
know
oscme
to
the
deck
right
now,
because
we're
not
you
know
we're
not
at
the
scale
we
need
to
have.
You
know
to
write
them
separate,
so
kind
of
anticipate
you
need
to
scale
or
something
like
it
with
get
with
gamma
I'm.
Finally
go
on
like
Ron
described
yesterday.
It
seems
like
it
does.
H
I
Yes,
yeah.
We
definitely
have
to
do
that.
Yes,
please,
okay,
so
I
don't
know
if
that
becomes
like
just
in
dedicated
Easter,
repo
or
one
under
Gamma
or
something
whatever
people
prefer,
but.
H
I
That
some
of
the
stuff
we
have
in
in
it
currently
is
like
super
tied
to
istio,
so
that,
like
maybe
we
just
take
that
strip
it
and
then
put
it
in
kubernetes
in
the
issue
itself,
may
not
even
use
it.
To
be
honest,
because
we
have
to
remove
some
of
these
two
stuff
and
it's
too
painful
to
make
it
extensible
and
whatnot.
But
then
then
we're
not
Reinventing,
but
we
are
like
maintaining
our
own
thing
as
kubernetes
Gateway
API.
H
A
Yeah
I
think
I'm,
probably
fine,
except
in
the
burden
of
maintaining
just
not
Reinventing
yeah
right,
where
I
am.
F
Is
a
thing
where,
where
once,
if
we
can
extract
it
out
of
so
that
it's
you
know
as
constant
says,
there's
not
many.
Doesn't
it
we
minimize
the
dependencies
on
istio.
This
feels
like
a
thing
where
we
could
end
up
with
something
like
the
Ingress
to
Gateway
thing.
That
Rob
has
started
just
like
the
standard
way
to
test
this
sort
of
stuff
that
could
live
under
kubernetes
six.
That
would
make
sense.
F
A
Okay
seriously,
at
some
point,
I
want
to
get
Mike's
dog
like
a
medal
as
an
honor
80
honorary
gamma
member.
A
Okay,
so
it's
it
sounds
like
the
this
I
haven't
heard
any
dissent
at
this
point,
so
I'm
tentatively,
going
to
just
kind
of
call
that
result
and
and
say
that
we're
just
going
to
use
the
SEO
conformance
it's
about
the
SEO
testing
tool
that
you
know
will
be
struck
by
your
Co
team
of
specific
things
and
then
kind
of
essentially
donate
it
to
the
kubernetes
repo
costume
Luigi
on
that
sounds
that
kind
of
represent
the
action
item
there.
F
So
I
I
agree
that
that
sounds
resolved
to
me,
but
it
would
be
good
to
have
an
issue
or
something
recording
somewhere
recording
the
decision
that
way
we
can
all
plus
one
and
whatever,
and
it
can
be
all
above
wood.
A
Sounds
good
yeah
I
will
report
it
here
in
the
notes
and
then,
if
somebody
wants
to
create
a
issue,
please
do
if
I
don't
see
one
next
couple
of
days.
I'll
just
go
ahead
and
create
it.
Yeah.
A
Okay,
yeah.
A
Oh
yes,
there
is
one
already
a
good
call
out,
okay
kind
of
as
a
sub
Point
under
conformance
testing,
which
could
have
been
its
own
thing.
But
the
contact
thing
was
important
when
talking
about
this
yesterday,
I,
don't
remember
who
it
was,
who
brought
it
up,
but
when,
when
the
current
Gateway
API
performance
test
currently
use
the
same
cluster
and
have
multiple
gateways
in
the
same
cluster
as
the
kind
of
expedite
the
performance
testing,
so
you're
not
spitting
up
a
new
cluster
each
time
with
the
mesh.
A
You
know
we've
kind
of
punted
in
gamma
the
idea
of
namespacing
romance
multiple
meshes.
It's
in
a
cluster
deployment
models
and
and
topology
in
general,
we've
kind
of
punted.
All
of
that
yeah
I
think
that's
right
decision,
because
these
are
all
kind
of
complex
problems
as
we
start
getting
into
okay.
How
do
we
have
these
conformance
tests
running
a
reasonable
amount
of
time?
A
It
might
be
worth
revisiting
those
I'm
not
trying
to
set
any
kind
of
policy
or
Direction
here
I'm
just
trying
to
open
it
back
up
to
the
group
to
see
if
we
feel
announced
the
correct
time.
We've
got
Direction
on
HTTP
route,
skipping
ahead
a
little
bit
but
get
1426
has
been
merged
as
provisional
yay.
That's
awesome,
I
think
it
represents
a
big.
A
You
know
set
of
work,
a
milestone
for
us
and
so
now
I
think
it's
an
appropriate
time
to
at
least
ask
the
question.
If
we
want
to
do
that,
Nick
I
saw
you
on
oh
you're,
not
unmuted,
but
Customs
hand,
person
and
John.
H
I,
probably
don't
want
to
say
the
same
thing.
I
suspect,
I
think
is
a
locker,
for
this
is
some
agreement
on.
How
do
we
deal
with
with
certificates
and
and
Trust?
If,
if
we
want
to
test
with
with
security
enabled
if
we
want
to
do
testing
or
whatever
with
security
turned
off,
it
may
be
easier,
but
not
sure
how
valuable,
because
in
practice
most
people
will
probably
use
measures
with
security
on.
H
B
It
also
might
be
something
I
I'm,
just
thinking
ahead.
This
is
entirely
hypothetical
of
just
like
a
add,
a
callback
for,
like
a
pre-running,
hawk
and
post
run
hook
for
tests
such
that,
like
concurrent
implications,
could
do
something
like
add,
obviously,
rules
if
needed,
or
something
like
that.
B
That's
totally
hypothetical
I,
don't
know
if
that
is
practical,
but
it
may
be
one
way
to
do
something
like
that.
I
don't
know
foreign.
I
Yeah
I
was
just
going
to
say
it
like
on
multiple
meshes.
There's,
there's
not
really
one
definition
of
that
I
think
like
even
in
the
spec
today,
you
can
have
multiple
meshes
in
the
same
cluster.
There's
just
no
way
for
a
given
HTTP
right
to
say
which
one
it
applies
to
right.
It
implicitly
applies
to
all
of
them,
but
it's
totally
fine
to
have
multiple
of
them.
I
So
from
a
conformance
testing
perspective
we
just
need
to
you
know,
have
one
namespace
and
buy
one
mesh,
presumably,
and
that's
that's
totally
fine
I
think
a
lot
of
people
will
have
multiple
messages
actually,
because
we'll
have
silly
and
mesh
my
news,
because
they
don't
like
the
mattress,
because
that's
now
the
Baseline
on
almost
all
Cloud
providers
and
then
they
may
add
a
different
mesh
on
top
right.
E
I
E
I
Set
of
namespaces
yeah,
okay,
like
I,
just
mean
like
you,
shouldn't
you
probably
don't
want
to
like
Linker
D
and
E
Studio,
both
in
the
like
conformance
namespace
and
like
have
two
sidecars
or
something
I,
don't
know
what
happens,
but
it
doesn't
work
so
sure
right.
I
So,
like
it's
just
like
on
the
test,
you
say:
I'm
testing,
this
Gateway
class,
we'll
instead
we'll
need
some
way
for
them
to
enable
the
mesh
on
the
test
name,
spaces
I'm,
not
sure
what
that
looks
like,
but
probably
just
create
the
namespace
and
have
the
label
for
in
our
case,
if
others
have
other
enablement
methods,
I'm
sure
they
can
figure
something
out.
F
So
I
mean
it
sounds
like
we
are
going
to
need
to
saw
to
have
something
for
mesh
enablement
in
order
to
be
able
to
do
CI
properly
in
order
to
be
able
to
conforms
properly.
But
there's
going
to
be
some
way
that
you
can
signal.
There
needs
to
be
some
way
that
you
can
see
your
what's
enrolled
in
the
mesh
so
to
be
able
to
make
the
complexes
work
badly.
I
F
F
Yeah
yeah
I
should
also
make
clear
that
celium
service
mesh
does
require,
like
you
know,
twiddle
some
knobs
for
it
to
be
enabled
if
psyllium
is
installed.
So
the
service
mesh
is
not
there
unless
you
ask
for
it
to
be.
H
But
in
general,
in
in
most
cases,
when
you
talk
internal
property
and
the
performance
testing,
it's
important
to
know
if
you
know
Gateway
from
vendor
a
can
work
with
a
measurement
vendor
b
or
if
a
mesh
from
vendor
a
can.
You
know
be
in
the
same
organization
as
a
version
of
the
interview,
and
so
that's
why
I
think
it's
important
to
address
authentication
and
communication
issues
first,
because
you
may
have
a
very
large
measure.
You
have
you
know
your
acquire
a
company.
They
use
that
one
mesh,
you
use
a
different
mesh.
F
Yeah
I
think
all
of
those
things
sound
like
extremely
worthwhile
problems
to
solve
that
are
very,
very,
very
hard
and
require
like
enormous
Matrix
tests,
that
will
be
very
expensive,
CI
and
so
I
would
like
to
kick
those
down
the
road
and
make
them
future
us.
This
problem.
H
A
This
is
theme
like
side
note,
but
I
was
looking
through
the
Harry
opponent
notice.
There's
studies
in
the
G
of
the
gcr.io
image
registry
and
Nick
said
expect
to
see.
How
make
me
think
about
this
is,
is
does
kubernetes
six?
Is
there?
Are
they
taking
part
in
the
kind
of
registry.k
to
I
o
move,
or
is
that
just
the
K
slash
K
repo.
E
I
think
we
should
we
it's
just
uninvested
right
now,
I
mean
so
any
anyone
who
has
time
to
make
the
transition
is
more
than
welcome
to
it's.
Just
no
one
currently
has
taken
the
time.
I
think
we
have
an
issue
tracking
it
yeah.
F
And
we
don't
even
we
don't
even
publish
to
the
production
cage
registry.
At
the
moment,
it's
only
all
the
Gateway
API
stuff
is
only
available
in
the
staging
codes
for
Just,
because
no
one's
done
the
work
to
actually
like
push
the
buttons
to
make
that
happen.
So
I
think
that
both
of
those
things
like
both
of
us,
these
publishing
to
the
coach
for
history
and
then
using
registered
or
cage.io
instead
are
available.
It
just
needs
resourcing
for
someone
to
actually
sit
down
and
do
the
work
to
make
it
happen.
A
Wow,
that
is
a
that
is
a
a
popular
phrase
in
open
source,
yeah
Nickerson,
so
it
just
it
just
popped
in
my
head.
One
just
double
check:
okay
cool,
so
it
sounds
like
there
is
no
real,
like
action
for
the
spec
to
take
here
as
far
as
enabling
performance
tests
as
long
as
and
I
think,
we've
already
done
this
as
long
as
there
is
nothing
prohibiting
meshes
from
having
their
own
movement
mechanisms,
then
we're
essentially
good
to
go
or
doing
good
moments,
which
is
a
good,
a
very
good
thing.
Now.
A
What
we
need
is
resourcing
for
people
to
start
writing
components,
tests
which
is
another
conversation.
Do
we
want
to
yeah,
let's
go
ahead
and
do
this:
let's
have
a
conversation.
So
at
what
point
do
we
feel
comfortable
starting
that
Journey
I
know?
A
We've
got
that
accident
Number
issue
to
kind
of
make
that
the
the
echo
server
and
the
various
utilities
around
that
available
in
a
third-party,
neutral,
Koreans
repo
once
that's
done,
is
that
kind
of
the
green
light
to
start
utilizing
it
and
writing
tests
for
it
are
there
any
other
things
that
need
to
be
done?
That
could
perhaps
be
worked
on
in
parallel.
A
A
E
I
mostly
agree
with
what
you're
saying
I
think
it
would
be
helpful
because
these
conformance
tests
are
going
to
be
slightly
different
than
what
we
already
have.
My
I
would
suggest
writing
some
kind
of
either
addition
to
the
existing
conformance,
Gap
or
a
net
new
Gap
that
basically
proposes
a
structure
for
what
these
new
conformance
tests.
You
know
how
they
will
function,
so
we're
not
just
going
straight
to
here's.
My
first
attempt
at
how
this
will
work,
but
we
have
some
kind
of
agreement
on
how
they'll
be
structured
first,
but
otherwise.
A
Yeah
I
completely
miss
that
there
is
a
conformance
get,
but
that
probably
makes
sense
that
there
is
yeah
I
agree
that
should
be
done
prior
before
we
start
writing
writing
code.
So
yeah
I'll,
plus
one
there.
B
Finding
some
more
functionality
like
the
ability
to
opt
into
specific
stuff
and
part
of
this
is
probably
kind
of
figured
out
conformance
levels
and
how
those
map
over
to
gamma,
because
with
you
may
have
Divergent
implementations
between
a
mesh.
That
also
has
a
north-south
component,
or
only
once
one
or
the
other,
so
definitely
some
stuff
to
figure
out
there.
B
But
I
am
happy
to
start
writing
tests
in
parallel
with
a
build
out
for
us
and
also
happy
to
help
onboard
anybody,
who's
interested
in
contributing
tests
and
help
review
and
facilitate
some
of
that
work
so
definitely
excited
to
see
us
like
get
a
framework
in
place
and
then
start
writing.
A
A
Personally,
we
can
chat
offline,
you
can
initiate
me,
but
I
completely
agree.
I,
think
we
need
to
figure
out
conformance
levels
here
and
then
have
that
you
know
we've
hinted
at
it
a
couple
of
times
and
basically
that's
been.
Oh,
this
is
different.
We
should
figure
this
out
at
some
point
and
we
need
to
actually
like
set
our
time.
Yeah.
A
D
A
A
Okay,
moving
right
on
along
so
yeah.
This
is
I
mentioned
this
earlier,
but
get
1426
was
merged,
which
is
super
exciting.
A
A
In
the
specification,
there
are
seven
different
items
here:
two
have
been
assigned
with
PR
as
a
link
to
the
one
to
go.
John
I'll,
probably
add
myself
to
a
couple
of
these
and
look
there,
you
go
Define
the
performance
levels
and
how
that
should
be
applied
to
gamlets
right
there
in
the
Milestone.
So
this
is
something
that
you
know
that's
rapidly
approaching
anymore
Nick.
It
is
here
it's
in
the
Milestone,
so
we're
gonna
have
that
conversation.
A
You
know
here
soon,
as
in
like
now
not
now
about
next
week
and
yeah,
just
investigation
of
limits
and
digging
into
some
of
the
semantics
I.
Think
of
the
existing
of
the
gaps.
So
if
any.
A
Let
me
get
the
zoom
bar
out
of
my
way.
That's.
F
B
Yeah
cool
yep
also
just
wanted
to
like
thanks
so
much
everyone
who's
taking
the
time
to
read
through
review,
comment
on
on
the
Gap
like
that
was
so
incredibly
helpful
in
like
driving
it
towards
being
something
that
we
can
merge
and
breaking
out
some
of
the
stuff
into
open
questions
for
anybody
who
has
not
had
a
chance
to
read
it.
B
Yet,
like
would
love
you
to
take
the
time
and
if
there's
areas
that
are
unexplored
or
not
clear,
like
please
open
issues
open
PRS
and
we
can
add
them
to
the
Milestone
to
help
get
a
pretty
solid
Clarity
before
we
before.
This
is
implementable.
So
just
thanks
everyone,
everyone
who's,
contributed
so
far
and
looking
forward
to
working
with
more
of
y'all
soon.
F
I
think
one
thing
that
I
just
wanted
to
say
that
it
is
important
to
note
is
that
a
gap
is
not
set
in
stone.
You
know
especially
a
gap
that
is
still
flagged
as
experimental
or
implementable.
It
can
be
changed
until
it
changes
to
like
implemented
finished.
I
can't
remember
what
the
actual
state
is.
Well,
maybe
you
can
help
me,
but
yeah
like
that
until
the
Gap
is
actually
like
marked
as
done
then
then
it's
it's
up
for
modification
and
check.
E
Yeah,
so
we
recently
emerged
a
change
to
the
Gap
documentation
and
process
that
goes
from
implemented
to
experimental
to
standard.
So
there's
a
distinct
difference
between
you
know
like
this
will
go
once
it's
input
implementable.
It
will
go
to
experimental
once
it's
released
and
hopefully
eventually
get
the
standard.
A
Phenomenal
so
yeah,
like
I,
said:
if
anybody
else
has
you
know
one
of
those
issues
in
the
mouse
when
they'll
actually
start
tackling,
then
please
assign
yourself
or
one
of
the
maintain
some
will
assign
you
excited
to
get
this
stuff
closed
out.
A
All
right,
Shane,
finally
wondering
about
attached
to
mesh
semantics.
We've
covered
a
little
bit
of
that,
but
you
wanna
tackle
that
again.
J
We
covered
kind
of
what
I
wanted
to
talk
about
pretty
good
earlier,
with
revisiting
multiple
meshes
in
the
same
cluster.
I
just
wanted
to
make
sure
we
were
still
thinking
about
this.
I
have
a
pretty
strong
inclination
that
in
time
we
will
users.
We
will
want
users
to
have
the
ability
to
say
that
they
can
attach
to
a
mesh.
However,
that
ends
up
looking
I
have
a
preference
for
the
mesh
resource,
but
I
know
that
right
now,
I
think
I'm.
The
only
one
who's
really.
J
A
A
Particularly
one
thing
that
comes
to
my
mind
is
the
upgrade
case.
This
was
brought
up
like
I
mentioned
yesterday,
where,
okay,
you
do
technically
have
multiple
messages.
When
you
go
into
an
upgrade,
that's
something
that
I
think
probably
needs
to
be
solved.
A
I
I
know
osm
and
I
spent
some
time
in
istio
world
before
I
know
that
istio's
got
the
whole
revision
and
tax
semantic
not
just
to
mentioned,
but
it's
this,
you
know
the
construct
to
you
know
represent
different
versions
of
the
mesh
within
a
a
cluster
I.
Don't
know
that
you'd
ever
want
a
route
to
this
in
one
about
one
version,
not
the
other
version,
but
you
know
that's
something
a
user
might
reach
for
for
like
how
do
I
represent
this.
A
So
even
if
you
know,
I
I
think
I'm
still
fine
punting
it
for
now.
If
it
was
like
the
other,
more
important
things,
but
I
feel
like,
as
we
start
like
moving
HTTP
route
to
from
implementable
to
maybe
to
experimental
and
then
for
sure
experimental
standard,
I
think
we
are
going
to
have
to
answer
that
question.
A
Even
if
we
don't
have
a
mechanism,
you
should
be
explaining
be
explaining
how
to
represent
certain
things
or
explicitly
say
these
things
aren't
supported
by
the
spec
for,
for
whenever
users
are
looking
for
that
functionality,
it's
matching
sense.
Anybody
else
have
thoughts
on
it.
J
H
Yeah
I
just
want
to
say
my
only
requirement
really
would
be
that
if
I
have
an
HTTP
route
attached
to
a
service,
any
mesh
implementation
should
obey
it.
I
mean
it
doesn't
matter
what
implementation
I
use
or
implementation
should
be
conformant
and
should
respect
the
the
food
magnetic
specs
that
is
defined.
I
agree
that
it
would
be
very
useful
to
have
a
standard
way
to
specify
this
workload
is
using
a
particular
admission,
implementation
or
version
I.
H
Don't
know
how
critical
it
is,
because
each
vendor
probably
has
its
own
ways
and
it's
specific,
but
as
long
as
we
can
guarantee
that
all
meshes
are
conformant
and
hopefully
can
interoperate.
So
we
can
have
coexistence
yeah.
Probably
it's
not
a
big
deal.
We
can
delay
it
a
bit.
B
I
think
some
of
that
may
be
difficult,
particularly
as
you
get
into
like
the
mess
like
implementation.
Specific
extension
points
like
if
you
have
a
filter
or
something
like
that
on
your
HTTP
route,
then
istio
by
respects
and
understand
that
and
console
might
not
accept
the
route
because
it
doesn't.
It
has
a
filter
defined
that
it
doesn't
understand
so
I
I
guess
I
would
lean
more
towards.
B
I
external
engagement
points,
whether
that
be
like
a
Gateway
implementation
of
some
sort,
whether
that
be
some
of
the
multi-cluster
service
Imports,
something
that
we
haven't
quite
gotten
to
yet
I.
Think
there's
a
lot
of
like
Edge
Casey
type
problems
of
expecting
every
mess
to
understand
resources
intended
for
a
different
mesh.
F
I
agree:
I
think
that
there
are
mechanisms
already
in
the
HTTP
route
status,
particularly,
namely
the
controller
name
stuff
that
will
allow
you
to
slice
the
status
to
say
you
know,
controller
name,
istio,
Sports
this
and
control
and
console
doesn't
yeah
I,
think
you
know
it's
probably
it
could
be.
F
It
certainly
could
be
approved
to
handle
the
message
case
a
little
bit
better,
but
I
think
that
today
there
are
ways
to
make
sure
that
that
is
clear,
that
that
this
doesn't
does
and
does
not
work
for
this
route
doesn't
does
not
work
for
certain
parents,
because
those
parents
will
then
get
roll
up
into
something
that
has
a
controller
name.
Presumably
well,
I
think
that
that's
probably
the
one
thing
that
we
will
they
just
keep
no
matter
what,
because
that's
how
the
status
works.
Sorry.
E
Hey
so
I
just
kind
of
want
to
follow
up
on
on
Shane's.
You
know,
like
the
the
LA
I,
don't
know,
I
have
some
concerns
too
about
just
I.
Think
right
now
we're
we're
implicitly
saying
that
if
you
attach
to
a
service
as
a
parent
ref,
that
is,
that
means
you
mean
mesh
and
that's
kind
of
like
a
leap
in
terms
of
understanding
that
does
not
feel
like
and
and
to
be
clear,
I'm
not
trying
to
slow,
slow
down
this
process.
I
think
this
is
still
a
reasonable
starting
point.
E
I
just
want
to
have
this
conversation
available
that
maybe
some
kind
of
decoration
to
say
I
want
to
attach
this
mesh
or
just
mesh
as
a
whole
to
this
service,
as
opposed
to
like
it,
you
can
imagine
that
it
would
start
to
get
very
confusing
to
users
if,
if
it's
a
service
type
load,
balancer
and
we're
saying,
hey,
Gateway
replaces
service
type
load
balancer
and
usually
have
an
L4
Gateway
that
you
want
to
attach
to.
Why
is
that
different
than
attaching
to
an
L4
load?
E
Balancer
server,
you
know
I
I,
think
just
my
main
concern
is
in
terms
of.
Does
the
does
the
kind
of
parent
then
influence
what
can
implement
it
again?
I
I,
don't
think
this
should
slow
down
I
I
don't
want
to
get
stuck
going
going
around
in
circle,
but
I
just
would
like
to
leave
some
room
open
to
have
some
kind
of
addition
to
this.
E
Api,
like
as
a
rough
example
like
what,
if
parent
refs
had
some
kind
of
field
like
mesh
that
said,
hey
I
want
to
attach
to
this
service
using
this
mesh
or
I
want
to
attach
to
this
service
via
this
mesh
or
I
I,
don't
know,
I,
don't
know
what
that
looks
like,
but
it
feels
like.
We
need
just
like
one
additional
thing
to
indicate.
This
is
a
mesh
specific
concept.
J
E
H
Even
for
Gateway,
it's
not
something
that
the
user
can
easily
do.
I
mean
I,
say:
I
want
to
use
nginx
as
a
as
a
data,
implementation
or
I
want
to
use
history
as
I
get
implementation.
It's
not
as
easy
as
I
go.
I
put
this
you
know,
Gateway
classes
to
an
Easter
will
suddenly
show
up,
usually
both
gate
to
implementation
and
meshes
need
to
be
provisioned.
You
may
need
to
install
whatever
demon
sets
permissions.
You
know
sign
up.
H
We
do
all
kind
of
stuff
that
is
necessary,
so
it's
not
something
that
a
regular
user
will
be
able
to
do,
and
then
how
do
you
know
the
list
of
messages
when
you
deploy
the
cluster?
How
do
you
find
out
what
meshes
are
available
in
that
cluster?
If
you
have
a
CI
CD?
So
it's
a
far
more
complicated
problem
and
given
that
mesh
is
not
interoperate
and
we
don't
have
interoperability
testing
between
them,
then
the
reason
is
that
usually
select
my
shape
user
B
select
with
mesh
B
and
since
they
cannot
talk
with
each
other.
H
E
Yeah
I
think
I
think
my
that's
a
good
that
those
are
all
good
points.
I
I
think
my
main
concern
is
adding
something
inside
inside
parent
refs.
That
is
saying,
I
want
to
attach
this
Resource
as
mesh
specific
like
this.
This
is
a
mesh
specific
attachment.
I
I,
don't
I'm,
not
too
I,
don't
care
too
much
about
how
we
do
that.
E
I
just
would
like
to
differentiate
it
from
because
you
can
imagine
that
at
some
point,
if
Gateway
supports
cluster
IP
routing
you're
going
to
want
to
attach
to
that
as
a
parent,
ref
and
so
you're
going
to
want
to
be
able
to
distinguish
how
you're
attaching
to
that,
whether
it's
for
mesh
or
Ingress,
routing.
A
What
comes
to
my
mind
as
far
as
like
how
I'm
interpreting
what
you're
saying,
rob
and
Shane
I
guess
is,
it
seems
like
that
the
current
get
implementation
kind
of
puts
HTTP
routes
in
this
one
kind
of
global
name
space,
if
you
will
not
literally
namespace,
but
this
one
Global
Group
and
any
all
messages
in
that
particular
cluster
are
reading
from
that
same
group
and
what
users
and
I
can
I
can
see
this
pretty
easily
as
well.
A
Users
would
look
for
a
way
to
segment
this
route
from
another
set
of
routes
and
put
them
in
different
groups.
Essentially
because,
like
to
the
point
you
just
made,
you
know
when
Gateway
supports
Tech
load
balance.
You
know
sorry
type
cluster
IP,
you
know
there's
not
much
distinguishing
it.
It
feels
like
the
underlying
the
underlying
question
that
we've
had
that
were
that
we're
circling
around
is
is
kind
enough.
It's
the
kind
of
the
parent
rep
enough
to
delineate
what
this
route
is.
Attaching
to
and.
A
I've
gone
back
and
forth
like
thought
about
this
several
so
many
times,
but
you
know:
I
went
to
there's
a
a
trip,
Summit
talk
about
API
design
and
it
feels
like
using
kind
in
order
to
segment
this
namespace.
A
This
virtual
interface
is
overloaded
too,
but
this
figurative
namespace
for
all
these
routes
lives
doesn't
feel
like
the
strongest
abstraction
and
yeah
I
I.
Don't
know
it's
not
something
that
we
have
to
solve
immediately.
I!
Think
we've!
You
know
open
the
door
to
allow
us
to
continue
to
iterate
and
make
progress.
A
Now
you
don't
think
for
what
it's
worth,
that
they
exist
and
get
close
at
the
door
on
adding
anything
later,
but
The,
more
I've
heard
you
know,
Rob,
Shane
and
Coston
said
specifically
with
Costa
and
the
idea
of
being
able
to
see
what
meshes
are
in
my
cluster.
A
It,
it
does
feel,
like
kind,
might
not
be
the
correct
abstraction
or
this
on
its
own.
Rather
so
that
was
LinkedIn.
I'll
I'll
stop
talking
there.
F
I
propose
I
propose.
We
call
the
namespace
that
all
of
the
routes
live
in
the
mesh
space.
E
Yeah
and
I
want
to
be
super
clear
and
I
know
this
kind
of
came
out
of
nowhere
here,
I
just
this
is
not
a
problem
today.
It's
not
a
problem
tomorrow.
This
is
more
just
future
thinking
of
the
direction
of
API
and,
as
these
things
start
to
converge,
a
little
bit
more
I
want
to
make
sure
that
we
have.
We
have
left
space
for
these
things
to
happen
and,
right
now
it
seems
like
we're
not
leaving
space.
It's
fine,
it's
experimental!
A
So
yeah,
let
me
know
if
you
create
a
slash
when
you
create
that
issue
and
I
can
add
some
some
mesh
contacts
there
cool
all
right.
We
are
right
at
time.
We
have
one
minute
left.
We
have
any
last
any
last
comments
or
questions
before
we
wrap
it
up.
A
All
right
awesome!
Well,
thank
you.
Everybody
for
your
attendance
look
forward
to
the
next
gamma
meeting
next
week
at
8,
A.M
Pacific
time
at
the
end
of
it
there
everybody
have
a
good
rest
of
your
day
week,
but
everything
talk
to
y'all
later
cool
thanks.