►
From YouTube: Maintainer Q&A
Description
Join us for Kubernetes Forums Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
A
Thank
you,
everyone
that
was
totally
awesome.
All
the
talks
were
fantastic
I
want
to
thank
all
of
the
speakers
and
all
of
you
that
came,
it
was
a
great
day.
I
had
tons
of
fun.
I
hope
you
did
I'm
just
a
few
quick
things.
So
I
already
said.
Thank
you
so
I'll
say
thank
you
again,
big
thanks
to
our
sponsors,
I
apologize
to
those
sponsors
that
I
harassed
quite
a
bit
to
get
get
money.
But
thank
you.
Thank
you
very
much.
A
A
We
get
a
round
of
applause
for
the
sponsors
at
this
point,
just
just
real
quick.
This
is
the
same
slide
that
I
put
up
last
year
and
I.
Think
I
talked
about
this
during
my
opening,
but
we
we
need
lots
of
help
from
a
project
perspective.
We
need
help
beyond
just
code
changes.
We
need
help
with
documentation.
I
have
many
many
many
conversations
with
people.
You
know
where
they'll
say:
oh
I
read
the
documentation
and
it
was
a
little
confusing
and
you
know
thank
you
for
answering
my
question.
A
Super
appreciate,
pr's
that
just
fix
the
documentation
or
even,
if
there's
something
simple
like
if
you
wish
the
documentation
had
a
ref
link
to
some
other
part
of
the
documentation,
it's
a
great
way
to
get
involved
in
the
project.
It's
really
really
appreciated.
So,
if
you're
looking
for
ways
of
helping
documentation
is
great
one,
we're
always
looking
for
blogs,
we
can
host
those
on
the
Envoy
proxy
blog
or
happy
to
link
to
your
company's
blog.
A
So
blogging
again
is
a
great
way,
and
these
can
be
user
stories
that
can
be
about
features
they
can
be
just
about
anything.
Also,
always
love
help
with
examples.
We
have
an
increasing
number
of
sandboxes
in
the
project
and
those
are
great
ways
of
helping
people
learn.
So
thank
you
for
those-
and
you
know
again
always
opening
issues
with
questions
is
always
super
helpful.
You
know
for
us
to
understand.
A
What's
going
on
and
as
I
was
saying
before,
we
are
looking
for
maintainer
and
that
does
not
necessarily
require
you
know
a
deep
knowledge
of
C++
networking,
as
I
was
saying
before
I
think
over
the
next
year,
we're
going
to
be
increasingly
looking
at
having
different
types
of
maintainer
roles
and
probably
having
maintained,
errs
just
either
work
on
different
subsystems
or
you
know,
handle
different
parts
of
the
project.
So
if
that
is
appealing
to
you
or
if
your
employer
is
looking
to
help,
you
have
some
time
you
know
for
doing
open-source
work.
A
So,
at
the
current
time
you
know,
I
I
know
that
there's
just
us
in
between
you
and
some
drinks
back
there
or
something
else,
but
I
was
just
gonna,
bring
up
the
maintainer
z--
who
are
who
are
here,
and
we
can
do
a
brief
Q&A.
If
there's
questions
that
people
had
so,
if
you're
here
in
the
audience,
maintainer
z--
would
love
to
have
you
come
up
on
stage,
so
we
have
myself,
we
have
Elissa
Harvey
Stefan
he's
in
snow,
and
then
we
have
D,
Jose,
Josh
and
I.
Think
Dan
is
not
present,
so
so
yeah.
A
So
thank
you
so
much
for
coming
and
like
I
said
we
can
answer
any
questions
that
people
had.
Also
sorry,
if
there
are
questions
for
some
of
the
presenters
who
presented
during
the
conference,
you
did
not
have
time
for
Q&A.
You
can
also
ask
questions
and
assuming
that
people
are
still
here,
we'll
grab
them
and
have
them
come
up.
So
sorry,
so
thank
you.
Everyone
and
yeah
so
can
I.
A
A
The
question
is:
what
is
the
deal
with
the
Envoy
filter
repo
that
anyone
want
to
take
that
one.
A
Okay,
this
is
a
great
example
of
a
case
in
which
we
could
use
some
help.
There's
a
there's
a
there's,
a
couple
of
known
problems:
there
we
have
some
sequencing
issues
of
changes
between
the
Envoy
repo
and
the
Envoy
filter
repo.
So
without
getting
into
a
huge
argument
about
mano
repos
there,
there
are
some
issues
there,
but
this
is
so.
We
would
like
there
to
be
more
examples
in
yeah.
C
A
Yes,
sorry
I
had
forgotten
that
that
is
the
plan,
but
yeah,
but
that's
a
that's
a
great
example
of
something
where,
if
you're
looking
to
come
in
and
help
that
would
be
extremely
valuable
for
the
larger
community
and
it
doesn't
require
any
c++
knowledge.
I
mean
it's
it's
a
bunch
of
github
and
CI
and
subhash
scripts,
but
that
would
be
super
helpful.
D
Yeah
daresay
like
another
repo
and
called
my
wisdom
and
webassembly
future
is
living
there.
Currently,
the
target
is
upstream
that
in
next
couple
of
Berwyn
or
to
release
in
the
mainstream,
my
ripple,
it
still
proxies
built
based
on
that
and
already
have
some
wisdom
filter
available.
So
you
can
try
out
there
if
you
want
that
the
current
status.
C
Yes,
so
incremental
X,
yes,
really
actually
has
multiple
components
that
is
Delta
XD
s,
which
is
basically
just
reducing
the
amount
you
to
ship
in
the
while,
which
is
good
for
both
network
and
compute
reasons,
as
we
saw
earlier
today,
and
that
more
or
less
works
right
now
there
is
that
the
only
thing
missing
there
I
believe
and
there's
a
few
roll
backs
and
roll
forwards
going
on
and
around,
and
it
is
around
80s
but
generally
speaking,
that's
that's
available
and
it's
supported
on
demand.
X,
yes,
which
way
you
lazy
load
configuration
really.
C
We
were
approaching
this
XD
s
by
XD
s
and
the
V
actively
worked
on
one
right.
Now
is
the
system
with
VHDs
and
that's
actually
also
coming
pretty
close
to
landing,
but
it's
not
yet
in
a
stable
state.
So
that's
pretty
much
where
we
are
there.
I
would
recommend
anyone
interested
in
incremental
XD
s
to
start
onboarding
onto
Delta
XD
s,
and
then
you
can
think
about
the
other
forms
of
incremental
XE
s
as
they
are
land
further
down.
A
Yeah
I
was
just
gonna
say
one
other
thing
again.
Just
you'll
see
me
keep
coming
back
to
this
theme
of
things
that
you
can
do
to
help
that
don't
require
C++
knowledge.
Is
it's
become
clear,
particularly
around
the
XDS
side
of
things
that
if
we
have
better
integration
tests,
it
would
be
very
helpful,
so
we
recently
merged
a
large
patch.
A
So
that's
just
just
another
area
where,
if
you
are
interested
in
coming
in
helping
out
with
testing
infrastructure,
I
think
there's
probably
some
pretty
low
hanging
fruit,
bringing
up
a
Envoy
binary
and
a
fake
control
plane
and
doing
various
permutations
and
making
sure
that
envoy
works
correctly.
If,
if
you're
into
that
kind
of
thing
very,
very
helpful.
A
C
C
They
work,
they
meet
all
the
needs
of
envoy
and
we're
gonna
slowly
D
envoy
fire
them,
but
we're
not
going
to
you
know
with
that
criteria.
They're
always
going
to
have
to
support
the
entire
set
of
functionality.
That
envoy
does
so
I'm,
not
too
concerned
about
that,
but
it
may
end
up
being
the
case
that
large
parts
of
the
API
need
to
be
proxy.
Specific,
we'll
see
where
we're
spending
a
lot
of
time.
Thinking
about
things
so
far,
the
transport
protocol
seems
pretty
generic
and
I
feel
that
that
is
that's
that's.
C
A
You
know
I
I
think
my
feeling
is,
there
is
risk,
and
unfortunately,
there's
no
I
mean
yeah
there.
There's
no
black
and
white
situation
here.
I
think
there
are
a
lot
of
politics
here.
There
are
a
lot
of
benefits
of
doing
this
work,
so
I
I
think
the
politics
are
probably
worth
it
from
an
industry
perspective,
but
per
what
Harvey
was
seeing.
I
think
we're
gonna
have
to
be
nimble
and
stay
flexible
and
see
what
happens.
E
Yeah
so
right
now
the
the
project
is
hosted
under
under
lift,
and
that
is
deliberate
in
the
sense
that
we
wanted
to
start
with
the
flexibility
to
to
really
integrate
with
a
lift,
app
and
work
for
for
lift
concerns,
but
with
the
vision
that
we
want.
We
want
a
strong
partnership
with
a
with
the
Android
community
and
as
hopefully
they,
the
project
gets
more
traction
and
and
more
more
use
cases.
Then
we
can.
We
can
move
back
into
into
the
Android
ecosystem.
I,
don't
know
if
you
know.
A
A
I
I
think
that
we're
very
close
to
having
it
be
the
production
networking
path
in
the
lyft
app,
and
there
are
quite
a
few
other
companies
I
think
that
are
pretty
interested,
but
even
if
it
were
just
in
use
in
the
lyft
app
I
think
the
chance
of
it
being
abandoned
is
quite
low.
So
would
I
say
that
there's
guaranteed
upkeep
right
now?
No,
if
we
get
it
rolled
out
to
production
in
the
first
half
of
next
year,
I
think
the
chance
of
it
going
away
is
almost
zero.
A
B
A
A
C
Yes,
I
mean,
as
I
mentioned,
you
might
talk,
you
know,
v3
is
going
to
not
be
too
exciting
right.
It's
not
really
gonna
have
anything
new.
It's
really
just
gonna,
be
a
deprecation
of
the
various
fields
in
v2
and
a
chance
for
us
to
get
our
process
in
place.
I
feel
probably
I.
Would
you
know
we're
intend
to
end
at
the
end
of
year?
I
would
give
it
a
few
weeks,
maybe
just
make
sure
there
are
no
roll
backs.
C
You
know
everyone's
pretty
happy
with
it
and
wait
for
folks
to
pick
it
up,
but
certainly
by
end
of
year
that
you
know
next
year,
you're
gonna
have
to
move
over
I.
Would
you
know
I'm,
not
one
of
these
people
who's,
like
necessarily,
you
know,
super
early
adopter
on
anything,
so
I
I
would
recommend
that
you,
you
know
probably
little,
but
if
it
was
me
I
would
let
someone
else
for
me.
C
Be
that
that
first
adopters
to
be
honest
but
I,
feel
we
certainly
are
we're
strongly
committed
to
having
v3
be
a
stable
or
a
bust,
API
and
I
would
say,
probably
within
let's
say
by
the
midway
through
the
first
quarter.
We
should
have
a
pretty
good
signal
of
whether
or
not
everything
landed
as
it
was
expected
to
land.
A
Once
we
snap
the
three
that
we
are
going
to
allow
small
features
and
again
I'm
really
waving
my
hands
here
to
be
added
to
the
v2
API,
so
it
won't
be
completely
frozen,
probably
again
like
for
I,
don't
know
if
it's
one
quarter
or
some
number
of
weeks
or
something
like
that,
mainly
because
it's
pretty
terrible
experience,
if
we
tell
consumers
that
you
know
to
add
it
to
have
access
to
one
field,
you
have
to
migrate
your
entire
API.
With
that
said,
we
have
to
eventually
encourage
people
to
migrate.
A
So
there'll
be
some
grace
period
by
which,
at
the
disk
we
will
not
back
port
every
feature,
but
at
the
discretion
of
the
person
who's
developing
it
they'll
be
given
some
period
of
time,
which
will
let
them
put
that
field
non-breaking,
obviously
into
v2,
but
then
after
some
date
we
just
block
it,
but
that
policy
has
to
be
worked
on.
So
if
you
have
thoughts
on
that,
that
would
be
helpful.
A
Will
you
be
my
ghostwriter?
The
question
is:
are
we
gonna
write
a
book
on
envoy,
I,
III
I
think
I
talked
to
Harvey
at
one
point
about
writing
a
book
or
something
like
that.
I
I
think
on
envoys
five-year
anniversary.
I
was
thinking
about
writing
a
medium
post
on
this
topic,
but
I
don't
know,
maybe
I'm
not
did
the
morning
yeah
sorry,
no
plans.
C
Okay,
so
there
is
actually
yeah
there's
a
dedicated
envoys
security
team
and
open
source.
It's
actually
mostly
the
maintainer
now
is
really
essentially
a
subset
of
them.
There's
also
a
team
at
Google,
which
is
also
a
heavily
contributing
to
this
process,
and
we
discover
most
of
there's
probably
three
different
ways:
we're
discovering
most
of
these
security
vulnerabilities.
Some
of
these
are
a
pretty
far
actually
there's
something's
reported
to
us
for
our
private
disclosure
email
lists,
and
then
we
reach
and
that's
great
they're
under
embargo
someone
discovered
by
fuzzing.
B
I
mean
basic,
basically,
every
time
we
we
get
a
report
or
if
I
don't
find
something
in
fuzzing.
We
end
up
doing
analysis,
kind
of
that
surrounding
area
of
code
and
there's
best
practices
in
terms
of
like
you
know,
smuggling
and
line
breaks
and
data.
You
know
extra
data,
injection
and
whatnot
that
we
tend
to
think
about
when
we
find
an
area
that
could
use
some
firming
up
and.
C
B
Say
one
more
point
to
that
again,
you
know
when
we
get
emails
to
the
envoys
security
list,
I'd
say
you
know,
maybe
half
of
them.
We
decide,
don't
merit
CDE,
but
we'd
much
much
rather
have
you
err
on
the
side
of
caution
and
tell
us
hey,
there's
something
crashing
I'm,
not
sure
if
it's
my
filter
or
if
it's
actually
exploitable
and
Envoy,
we
will
look
into
it
for
you.
We
will
help
you
debug
it
it's
much
much
better
for
everyone
involved
in
than
having
a
zero-day
posted
on
github,
where
you
basically
can't
delete
it.
A
Yeah,
the
the
only
thing
that
I
would
add
is
we
are
starting
to
think
again
about
our
general
process
around
how
many
stable
releases,
how
many
releases
should
we
support
the
the
thing
that
I
would
add
here
is
that
yes,
we're
all
paid
by
companies,
but
from
a
security
perspective,
you
know
we
have
different
focus
areas.
We
work
on
different
products
and
we
use
envoy
in
different
ways
and
we
care
about
different
things
and
the
reason
that
I
bring.
A
That
up
is
that
we
simply
do
not
have
the
resources
to
treat
every
line
of
code
in
the
Envoy
code
base
with
the
same
rigor
and
the
same
response
as
as
certain
certain
other
areas.
So
we've
been
increasingly
working
on
better
definitions
of
threat
models
and
processes
of
like
which
areas
of
the
code
base
you
know
received
here:
zero
support
and
Tier
one
and
so
on
and
I
think
increasingly
moving
forward.
We're
going
to
be
getting
more
and
more
rigorous
about
you
know
this
piece
of
code
or
this
extension
is.
A
It
is
not
supported
for
full-on
CBE
handling
and
if
there
are
companies
that
care
about
that,
we
are
going
to
need
volunteers
to
join
the
security
team
and
only
Security
Response
for
those
for
those
extensions,
and
that's
also
why
increasingly
you're
going
to
see
us
try
very
hard
to
push
more
and
more
things
into
extensions
because
for
people
that
have
you
know
more
stringent
security
postures
where
they
want
to
compile
out
code
that
they're,
not
depending
on
it,
allows
us
to
have
a
trusted.
Computing
basis
is
much
easier
to
understand.
A
E
F
A
Yep,
so
the
question
is:
are
we
going
to,
you
know,
are
going
to
support
more
stable
releases,
I?
Think
and
and
that's
actually,
what
I
was
saying
before
is
that
there
is
a
proposal
right
now,
that's
floating
around
it's
driven
by
the
sto
community,
but
we
think
that
there's
other
people
that
will
be
interested.
A
This
is
the
kind
of
thing
where
we
do
not
have
enough
time
to
do
stable
release
management
for
four
different
releases,
so
we're
looking
for
stable
release,
maintained,
errs
and
a
bunch
of
other
stuff,
so
these
are
again
things
where
they
don't
require
deep,
C++
knowledge,
but
if
you
want
to
really
help
the
community
and
help
with
cherry-picking
and
stuff
like
that
and
doing
release
management,
these
are
things
that
we,
you
know
the
community
would
really
appreciate
it.
Does
anyone
want
anything.
G
A
The
carrot
of
of
which
we're
also
thinking
about
about
doing
some
of
these
things
is
that
the
way
that
you
get
on
the
early
security
release
list
right
now
is
we
have
a
set
of
requirements
which
basically
say
that
you
cannot
be
an
end-user.
You
must
be
a
distributor
of
envoy,
so
you
have
to
be
a
package
maintainer
or
you
have
to
be
running
envoy
as
a
service.
It
would
be
a
public
named.
A
This
is
done
because
it's
mostly
in
line
with
what
other
similar
open-source
projects
do,
because
if
we
let
all
end-users
on
the
list
well,
then
we
start
having
arguments
over
how
big
do
you
have
to
be
in
blah
blah
blah
blah
blah?
It's
not
a
very
reasonable
thing
with
that
said,
we
are
thinking
about
a
change
in
policy
or
at
the
discretion
of
the
security
team.
If
you
are
a
large
enough
end-user
and
you
are
willing
to
help
us
test
risky
changes
and
test
security
fixes,
we
will
let
you
on
the
list.
A
So
what
it
means
to
be
large
is
TBD
and
likely
at
the
discretion
of
us.
But
if
that's
a
thing
that
interests
you
please
let
us
know,
but
that
would
mean
that
you
would
get
access
to
security
patches
two
weeks
in
advance
of
everyone
else
knowing
and
be
able
to
pre
patch
without
it
being
a
giant
fire
drill.
B
G
A
B
Things
I
know
off
hand
that
were
working
on
there's
a
bunch
of
work
again
on
security,
largely
front
about
Harvey's
team
in
terms
of
fuzzing,
but
also
looking
into
kind
of
refactoring.
Some
of
the
core
can
put
it's
an
envoy
to
pull
as
much
code
out
of
the
core
libraries
as
possible
to
kind
of
decrease
that
attack
and
bug
surface.
Obviously,
there's
a
lot
of
work
being
done
in
quick
integration.
B
G
Now
that
I've
had
a
second
think
so
actually
somebody
asked
me
about
this
earlier
and
I
think
it's
a
really
good
idea.
We
don't
have
any
specific
plans,
but
it
would
be
great
to
get
some
in
integrated
into
the
continuous
integration
testing
some
performance
testing
for
the
proxy,
so
that
we
can
understand
the
trend
over
time
of
how
it's
performing
and
under
various
scenarios
and
again
I'll
call
out
if
anybody
wants
to
help
with
that,
we
would
love
to
talk
to
you.
C
F
I'm
working
on
a
couple
small
things
right
now:
the
JSON
access
log
output
actually
not
being
all
strings
all
the
time
like
numbers
as
numbers
and
and
then
think,
I
need
to
look
at
using
URL
query
parameters
for
consistent
hashing,
as
opposed
to
just
headers
and
I,
want
to
further
around
with
the
memory
allocator
too,
but
we'll
see
if
I
get
to
that.
I.
B
Think
over
the
next
year
that
there's
also
gonna
be
a
lot
of
work
put
into
on
weight
scale,
so
both
handling
massively
scaled
up
numbers
of
configurations
handling
multi-tenancy.
Better,
probably
you
know
throughput
improvements.
Just
a
lot.
You
know
load
testing,
as
Josh
already
mentioned
a
lot
more
focus
on
it
being
kind
of
more
robust,
both
from
a
security
and
a
traffic
standpoint.
I
think
that
they're
also
probably
over
then
hopefully
over
the
next
year,
but
if
not
maybe
a
little
out
more
latency
improvements
right,
prefetch
and
connections
was
mentioned
at
one
of
the
talks.
B
H
Yes,
what
I
see
happening
is
more
flexible
load,
balancing
we're
seeing
that
right
now
with
the
aggregate
cluster.
That's
about
LAN,
which
kind
of
allows
you
to
do
load
balancing
across
multiple
cluster
at
the
same
time
and
I.
Imagine
that
will
just
kind
of
keep
increasing
as
people
have
more
in
different
requirements
and
so
yeah.
Hopefully,
load-balanced
will
be
more
configurable
extensible
in
general
and
kind
of
making
that
all
more
making
it
easier
for
people
to
do
exactly
what
they
want
to
do
in
terms
of
load,
balancing
and
failover,
and
all
that.
D
I
Since
I
worked
the
same
company
with
Liz,
and
so
mostly
we
really
the
same
yeah,
we
work
mostly
on
how
making
a
steal
better
in
terms
of
integration
with
envoy
and
you're,
bringing
more
probably
more
examples
on
assembly.
So
we
yeah
we
pay
for
information
that
it
currently
the
web
assembly
filter
inside
Steel
repo
is
like
using
VM
know.
So,
basically,
it's
C++,
but
using
the
web
assembly
API
is
not
it's
not
basically
not
web,
simply
assembly,
but
it's
it's
called
statically
linked
internet
proxy
and
that
probably
and
then
the
placing
capabilities.
C
D
E
Guess
my
answer
is
envoy
envoy
everywhere
and
that's
not
only
about
the
work
that
we're
doing
without
going
mobile,
but
actually
related
to.
This
is
some
of
the
work
that
we
want
to
upstream
is
like
we
were
presenting
earlier
the
the
API
listener
that
will
allow,
for
example,
other
libraries
to
inject
directly
into
Envoy
as
an
engine.
A
I
think
my
my
main
interests
over
2020
is
around
making
envoy
simpler
to
use.
So
we
have
a
lot
of
people
at
lyft
we're
working
on
generally
producing
common
configurations,
so
things
like
adaptive,
concurrency
adaptive,
timeouts
things
like
that.
There's
a
lot
of
ideas
that
we've
talked
about
around
how
to
have
people
have
better
configuration
defaults.
A
So
we
talked
about
introducing
configuration
profiles
so,
for
example,
you
would
start
up
on
by
with
perf
edge
servus
mesh,
and
that
would
preset
a
bunch
of
things
like
time
outs
and
you
know,
overload
manager
and
blah
blah
blah
blah
blah
and
then
basically
warn
you
if
you're
using
configuration
that
we
think
is
not
safe,
I
guess
on
a
personal
level,
be
honest,
I'm,
I'm,
tired,
so
my
you
know,
my
personal
interests
is
is
I,
think
it's.
It
think.
A
It's
clear
at
this
point
that
we're
like
I,
said
I
think
we're
in
this
for
the
long
haul.
So
my
interests
is
making
sure
that
we
continue
to
scale
the
number
of
maintained
errs
the
number
of
contributors,
the
people.
You
know
that
are
helping
to
answer
questions
and
all
of
those
things
so
yeah
I
mean
there's
lots
of
other
features,
but
those
are
the
things
that
are
top
of
mine
I.
B
Think
it's
worth
calling
out
is
one
of
the
unique
things
about
the
envoy
project.
Isn't
that
said,
is
that
it's
not
owned
by
specific
companies,
so
the
envoy
read
map
is,
as
you've
heard,
is
a
collection
of
what
various
companies
think
they're
working
on
right
now,
but
any
given
team
at
any
given
company
can
get
preempted
by
their
actual
day
jobs.
At
any
point
and
frequently,
the
envoy
roadmap
gets
driven
by
some
other
company
coming
in
and
being
like.
B
Hey
I
really
want
to
use
envoy
Forex,
and
you
know
we
sit
down
and
have
a
couple
design
conversations
and
help
Shepherd
them
to
that
a
process.
So
you
know
every
couple
times
a
year
we
get
pulled
by
like
scene,
CF
or
whatnot
of
like
what's
the
Envoy
roadmap
and
frankly,
we
don't
really
know
like
we
know
what
we
plan
on
working
on,
but
we
don't
know
what
you
plan
on
working
on
and
there's
more
of
you
than
there
are
of
us.
So
I
guess
we'll
just
see.
Yeah.
A
Well,
on
that
note,
we,
a
bunch
of
us,
will
be
tomorrow
on
the
show
floor
at
coop
con
there's
a
there's,
an
envoy
booth
somewhere
in
the
CN
CF
area.
A
bunch
of
us
will
be
there
from
around
noon,
to
like
2:30
or
to
20
to
12,
22,
20
or
something
anyway,
we'll
we'll
be
there
go
out,
I
think
there's
some
snacks
and
drinks
out
there
and
thank
you
so
much
to
everyone
for
coming.
Thanks.