►
From YouTube: Quilkin Monthly Sync August 2021
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
Now
I've
also
completely
forgotten.
Does
we
have
some
new
faces?
I
think
that
we
haven't
seen
before
I've
seen
nodding
heads
if
you
feel
comfortable.
Did
you
want
to
introduce
yourself?
We
can
do
a
random
intros.
B
Yeah
candy,
I'm
giles
I
work
for
cisco.
We
came
across
quilkin
a
few
weeks
ago
and
projects
I'm
working
at
the
moment,
looking
at
kind
of
real
time,
almost
like
service
mesh
type
stuff
in
kubernetes,
but
for
real-time
apps
and
I
was
like
cooking,
looks
like
it
could
be
helpful.
B
So
we've
we're
sort
of
thinking
of
my
code
nikolas
might
join
my
we're
sort
of
thinking
of
maybe
creating
a
new
filter
for
quillkin
to
our
nefarious
ends,
and
that's
that's
like
I
guess
we'll
be
interested
to
find
out
is
what's
the
kind
of
policy
on
on
up
streaming
new
filters
and
that
kind
of
stuff,
because
I.
D
D
E
F
Hi
everyone
guys
nice
to
meet
you.
I
am
luna.
I
work
at
embark.
Studios
crooked,
originally
started
in
a
mark's
twitch
chat
channel
with
me
ranting
at
him.
How
how
we
did
this
thing,
how
it
wasn't
cool
that
only
steam
and
a
few
other
major
studios
had
something
like
this.
F
So
that's
how
we
got
started
since
then.
I
haven't
done
much
development
on
wilkin.
Rather
I
I'm
the
tech
lead
for
the
team
at
embark
that
is
working
on
cloaking
in
which,
if
you
have
me
and
erin
are
part,
and
I
try
and
more
like
give
product
level
stuff
of.
I
want
to
be
able
to
do
this
and
that-
and
I
I
like
iffy
only
and
aaron
we're
both
amazing
engineers
just
fill
that
in
because.
A
F
Oh
god,
no!
I
I
still
do
a
lot
of
coding
and
I
spent
most
of
my
time
working
on
stuff,
but
I
most
of
my
time
I
spent
doing
either
surplus
plus
rust
or
go,
but
I've
just
had
other
stuff
to
work
on
for
the
past
year
and
amusing
anecdote
mark.
I'm
in
fact
getting
someone
to
help
me.
F
Do
the
actual
pming
bit
of
my
job,
who
got
hired
two
weeks
ago
and
who's
slowly,
ramping
up
so
yeah
I'll
be
doing
less
pm
work
going
forward,
hopefully
better
than
more
me
mark.
You
can
go
next.
A
A
Yes
lunar
and
I
had
this
wacky
idea
over
a
year
ago
for
this
in
a
twitch
stream
that
I
was
doing
so
that
was
been
fun
yes,
but
I
one
of
the
founders
on
agones
and
started
a
lot
of
the
initial
code
on
cool
kim,
but
I
think
opinion
and
aaron
have
written
a
lot
of
it
rewritten
a
lot
of
it.
These
days
I
seem
to
be
focusing
on
our
build
system
and
ci
tools,
which
is
fine
and
it's
lots
of
fun.
A
So
that's
all
that's
all
good
fun,
so
yeah
just
doing
stuff.
That's
me
I'll,
head.
G
Sure
yeah,
so
if
I
I
I
work
as
a
developer
on
the
same
team
as
luna
and
aaron,
we
work
at
embark
and
mostly
I've
been
helping
out
with
yeah
just
code
pr
reviews
things
like
that.
H
Sure
my
name
is
aaron.
I
work
as
an
open
source
engineer
at
embark
studios.
The
project
I
currently
work
on
is
quilting,
but
I've
worked
on
a
bunch
of
open
source
products
like
during
rush
gpu
and
blackboard
itself.
So
yeah.
That's
me.
I.
I
Yeah,
so
I'm
marcus,
so
I
work
as
an
infrastructure,
architect
embark,
I'm
probably
like
the
other
side
of
luna
in
the
requirements
and
defining
requirements
and
how
we
go,
how
we
should
use
clicking
at
the
embark,
but
rather
from
a
a
network,
an
infrastructure
site
rather
than
game
site.
So
that's
me.
E
And
I'm
april
I
work
in
the
google
open
source
programs
office
and
I
partner
in
crime
with
mark
on
all
of
our
gaming
open
source
and
also
work
on
our
cloud
native,
open
source
so
jealous.
I'm
glad
that
you're
here,
because
I
did
tell
mark
that
quillkin
was
going
to
be
quite
interesting
to
a
lot
of
folks
in
the
cloud
native
space
beyond
just
gaming.
So.
B
Yeah
volunteer
away.
I
should
I
should
probably
be
really
annoying
and
do
the
the
how
to
say
my
name
bit.
So
it's
giles
being
an
english
persuasion.
Anyone
who's
got
any
french
in
them,
would
know
the
names
kind
of
the
same,
but
for
our
version
thereof
and
my
colleague
may
be
joining
us
and
if
he
was
trying
to
join
nikos
he's
interesting
to
listen
to
you
guys,
I
guess
I'm
the
I'm
the
infrastructure
network,
part
of
our
side
of
things.
Yes
giles
and
buffy.
That's
the
one.
B
Yeah
that
the
guy
who
who's
charles
and
buffy
was
was
one
half
of
this
really
famous.
I'm
diverting
a
bit
here,
really
famous
series
of
instant
coffee
adverts
for
nestle
gold
blend
here
decades
ago.
It
was
one
of
those
will
they
won't
they
with
a
couple.
B
A
Awesome
yeah
and
if
anyone
else
has
anything
they
want
to
add
to
the
agenda,
please
for
free.
I
just
threw
some
stuff
down
that
I
thought
might
be
interesting
and
give
it
a
starting
point
and
move
them
from
there.
A
Take
him,
oh,
you
start
that's
starting
to
do
things.
So,
according
to
our
release
schedule,
we
said
that
we
would
get
together
every
month
and
say
hey:
should
we
cut
a
release
so
should
we
cut
a
release
initial
review
from
when
I
was
going
through
it?
I
think,
since
a
month
ago
we've
had
some
api
changes
that
are
quite
nice
and
some
updated
documentation,
I'm
trying
to
work
out
if
we
had
any
specific
things
that
are
either
blocking
or
stopping
us
from
doing
that.
I
think
the
only.
H
H
A
A
Documentation
had
a
couple
of
things
that
were
old
in
it.
I
don't
know
whether
we
want
to
we
want
to
stop
on
that.
There
was
a
I
found
something
when
I
was
going
through
it.
I
A
Well
then
we'll
be,
you
know
we
could
just
we
could,
should
we
write
a
ticket
and
then
make
it
make
a.
H
C
People
think
I
actually.
A
Don't
mind
on
timing,
I'll,
tell
you
what
my
theory
was.
Do
we
want
to
just
create
a
milestone
track,
anything
that
we
think
needs
to
be
fixed
before
then?
I
think
that's,
I
think,
that's
the
only
thing
I
could
come
up
with
and
then
once
those
are
complete
I'll,
just
be
like
hey
I'll,
set
a
release
ticket
and
we'll
we'll
just
do
it.
A
C
A
We'll
probably
find
other
stuff
after
the
release,
but
I
think
that's
yeah.
That's
fine!
All
right!
Let
me
here
actually
I'm
going
to
make
an
imagination
after
this
moment,
because.
A
C
C
F
Regarding
upstreaming
custom
filters,
my
take
on
the
giles
is
pretty
simple
if,
unless
it's
incredibly
obscure
and
has
like
a
super
weird
dependency
chain
or
is
hard
to
compile
or
something
long,
as
I
see
no
issues
that
just
upstream
it
if
it
is
super
obscure
or
or
has
some
weird
dependency,
I
don't
know
what
you're
making
yet
so
it's
hard
to
say
but
like
in
that
case
I'd
say:
let's
evaluate
if
it's
worth
it
and
work
the
maintenance
burden.
But
but
that's
about
really
my
only
caveat.
I
have
to
extremely
think.
B
Yeah,
I
can
probably
explain
what
we're
planning
to
do.
In
fact,
initially
we
were
thinking
of
modifying
the
existing
load
balancer
filter.
What
it
is
is
we
want
to
be
able
to
effectively
route
udp
flows
to
specific
endpoints
based
on
the
source
address.
B
So,
for
example,
you
can
imagine
we're
here
on
on
meet.google.com,
and
there
was
some
problem
with
google's
peering
with
british
telecom
in
london,
and
so
google
had
an
edge
proxy
and
they
wanted
to
say
route
that
traffic
to
another
pop.
Then
the
edge
proxy
could
could
divert
it
based
on
you
know,
look
up
my
source
address
and
say:
oh,
that
person's
on
you
know,
name
of
whichever
isp
it
was
british
telecon
virgin
whoever
and
would
say.
B
C
B
That's
a
bit
tricky
in
the
in
the
packets
already
arrived
at
the
at
the
edge
proxy.
At
that
point
so
yeah,
I
guess,
there's
only
a
subset
of
cases
you
can
solve
with
that,
but
you
could
you
know,
I
guess
imagine
other
cases
where
being
able
to
pick
an
end
pick
your
end
point
based
on
the
source.
B
Ip
could
be
quite
handy,
and
so
what
we
were
thinking
is
probably
do
kind
of
a
five
to
four
lookup
based
on
the
you
know
the
flow
and
then
use
that
to
route
packets
rather
than
doing
a
round
robin
or
a
random,
and
so,
as
I
say
initially,
you
looked
at
changing
the
existing
load,
balancer
filter,
but
there
is,
there
is
stuff
there
that
would
have
to
change.
I
mean,
for
example,
the
filter
only
takes
the
it
doesn't.
B
I
was
looking
at
the
code,
the
sort
of
context
for
the
packet,
the
that
five
to
put
info
doesn't
make
it
into
that
filter,
for
example,
in
terms
of
the
trait
that's
defined.
So
another
reason
I
guess
to
have
a
new
filter
for
it.
B
A
B
A
I'm
just
I'm
just
so
when
you
say
there's
something:
that's
not
there.
What's
not
there.
B
Oh
sorry,
just
so
when
I
looked
at
the
code,
the
existing
load,
balancer
filter,
you
don't
you
know,
there's
all
that
information.
You
have
remembering
it
from
my
shocking
memory.
You
have
kind
of
like
a
context
for
the
for
the
flow
and
one
of
the
things
in
the
context
is
the
effectively
the
client,
ip
and
port
or
whatever
and
yeah
that
does
not
get
passed
into
that
filter.
B
So
if
you
look
at
the
trait
that's
defined
for
the
load,
balancer
filter,
that's
not
one
of
the
parameters
that
gets
passed
in
like
kind
of
the
main
bit.
I
guess
gets
it,
but
when
it
goes
into
the
next
bit
of
thing
it
doesn't
it's
not
there.
I'd
say
I
still
remember
the
code
structure,
memory
shocking,
but
but
when
I
looked
at
where
the
traits
were
defined
for
the
the
random
and
the
and
then
implemented
by
the
random
and
the
the
around
robin
and
that
that
parameter
wasn't
there.
B
H
B
B
H
H
Yeah
this
is
this
is
something
I
wish
we
could
be
better
off
because,
like
for
example,
I
don't
know
how
many
people
have
used
tower,
which
is
this
rust
framework,
but
that
has
like
a
lot
of
load,
balancing
implementations
that
are
completely
generic
and
work
for
any
protocol.
I
would
love
if
we
could,
you
know
from
those
and
be
able
to
use
those
within
our
stuff,
be
nice
if
we
could
have
more
imagine
yeah
a
nice
way
to
do
that.
F
I
I
guess
for
a
filter
like
that,
the
xds
api
could
eat
in
a
routing
table
or
of
sorts
based
on
before
these
source.
Ips
do
this
or
that,
and
if
you
nice
way
to
configure
it,
if
you
want
to
make
it
nice
and
generic
there's
plenty
of
other
ways
to
do
it
as
well,
but
that's
probably
the
most
quill
kimmy.
That's
a
word
now.
One.
B
Yeah
initially,
I
was
wondering
funny
enough:
I'm
getting
bad
network
conditions
on
this,
which
is
just
the
thing
I
started
talking
about
the
initially
that
was
kind
of
I
was
thinking
of
maybe
pushing
pushing
the
routing
rules
into
coolkin
itself,
but
I
guess
in
a
sense
I
was
thinking
in
the
end.
Well,
you
probably
want
to
keep
cooking
pretty
lightweight
and
since
you
only
do
this
for
the
first
packet
in
the
flow,
you
know,
I
don't
think,
there's
a
huge
performance
hit
and
taking
it
out
and
just
having
an
api
call.
F
That's
the
thing,
though,
right
network,
sorry,
sorry,
I'm
interested
to
do
ipod,
chats,
please
people
notice
all
right.
So
that's
what
I
was
thinking
about
like
if
you're
having
a
video
call
for
three
hours,
the
network
conditions
might
change
right.
I
think
you
actually
get
more
value
out
of
pushing
it
in
yeah
it's
yes,
then
you
would
from
necessarily
only
checking
it
with
the
first
packet
in
the
flow
right
because
then
you're
stuck
with
it.
The
first
packet
was
fine
and
the
others
are
not
necessarily.
B
Yeah
yeah,
that's
potentially
true,
I
mean,
I
guess
yeah.
We
should
have
think
about
other
ways.
I'm
thinking
there
are
other
ways
you
can
sort
of.
Potentially
you
know
push
a
change,
but
you're
right,
yeah
you'd
have
to
know
which
flows
you're
moving,
so
maybe
yeah,
maybe
having
almost
a
routing
table
you
could
move
around
on
the
fly.
B
I
think
the
challenge
is
that
for
most
for
a
lot
of
things
you
might
be
doing,
the
problem
is
the
flow
itself
effectively
will
create
state
and
whatever
your
end
point
is
at
the
back
end.
So.
C
I
B
A
I
guess
that
was
more.
I
think
yeah.
The
only
other
thought
I
also
had
is
standardizing
around
xds
and
making
sure
that,
like
the
xds
is
our
our
mutation
integration.
Api
means
that
we
can
integrate
with
a
whole
bunch
of
other
stuff
without
sort
of
building
out
our
own
standards.
For
that,
if
we,
if
we
don't
have
to
it,
makes
things
a
lot
better
for
us
in
the
long
run
yep.
B
A
C
F
It's
worth
mentioning
we're
working
on
solving
a
very
different
problem,
a
very
similar
problem,
but
with
a
very
different
solution
that
might
be
worth
touching
on.
So
something
that
we
at
embark
want
to
do
is
also
have
quilting
at
various
edge
locations
and
we
want
our
actual
client.
So
our
game
client
in
our
case,
but
it
could
be
anything
really
pick
several
quiltings
from
a
list,
whichever
one
gets
the
best.
The
actual
real
world
conditions
from
and
the
way
we're
planning
to
solve.
F
That
problem
is
we're
routing
tokens
so,
in
our
case
we're
dealing
with
matchmakers,
but
the
solution
does
translate
when
our
game
client
matchmakes.
Currently
it
receives
a
server
ip
and
a
server
port
to
connect
to
right,
which
establishes
a
direct
connection.
If
you
add
cooking
in
between
what
we
could
do
and
what
we
plan
to
do
is
to
to
match.
F
Have
the
matchmaker
return
some
sort
of
routing
token
instead
and
then
use
the
xds
api
to
give
other
edge
quiltings
an
update
on
which
routing
token
matches
which
server
ends,
point
game
servers
in
our
case,
but
the
abstraction
holds
and
then
the
quilting,
the
the
quilting
knows.
F
All
the
quiltings
know
the
routing
tokens,
the
client
knows
its
own
running
token,
and
then
it
can
talk
to
any
quilting
that
it
wants
to
get
to
its
server
and
they
will
find
the
way
there
because
they
all
know
the
routing
token
and
how
to
get
there.
That
would
be
another
way
to
solve
the
same
problem.
F
If
you
want
to
add
something
similar
in
your
stack,
that
could
work
as
well,
but
the
advantage
there
is
that
we
can
let
the
client
actually
measure
the
latency
to
the
server
and
back
through
the
qualcans
and
have
it
decide
for
itself,
which
is
better
or
faster,
and
it
can
redo
that
every
now
and
then
reach
the
same
server
and
still
be
able
to
route
around
many
outages,
even
if
they
happen
in
the
game.
In
our
case,
so
that's
a,
I
think,
very
similar
problem
with
a
slightly
different
solution.
B
F
B
F
A
And
give
the
thumbs
up
cool,
we
happy
moving
on
just
checking
excellent
all
right.
I
really
wanted
to
get
some
there's
been
some
great
conversation
between
aaron
and
finney
and
myself
on
scriptable
filters.
A
Actually,
I
should
put
no.
I
did
link
to.
I
did
link
to
that
there
and
I
just
wanted
to
sort
of
highlight
it
and
possibly
get
some
external
feedback
outside
of
the
three
of
us.
The
big
thing,
that's
in
my
mind,
is
around
the
performance
of
it
and
like
what
our
use
cases
are
around
that
and
whether
that
performance
hit,
that
we
would
get
with
scriptable
filters
is
actually
acceptable
and
worth
the
effort.
A
H
F
I
guess
the
main
upside
they
get
is
they
need
to
do
it
once
per
tcp
connection,
usually
since
envoy,
almost
strictly
deals
with
tcp
right,
it's
http
connections,
grpc
connections,
etc
and,
like
oh,
no.
I
took
20
ms
to
initiate
my
connection
instead
of
14
whatever
exactly
for
udp
packets,
that's
a
bigger
problem.
So
that's
indeed
I
don't
have
any
thoughts,
but.
B
Yeah,
but
is
it
a
bigger
problem
if
you're
only
doing
it
on
the
first
packet
in
the
flow
again?
That
kind
of
you
know
the
gaming
sessions
tend
to
be
quite
long-lived,
don't
they,
whereas
unless
you
go
to
the
world
of
http
3
and
the
rest,
you
know
with
the
web.
Historically,
there's
been
a
lot
of
short
shorter
lived
flows.
A
H
H
Current
yeah,
we
don't,
we
don't
have
control
flow
in
the
filters
at
the
moment,
so
you
couldn't
say
like
if
you
couldn't
write
this
in.
You
could
write
this
as
part
of
the.
If
you
as
a
library,
you
could
probably
write
this,
but
you
couldn't
right
now
in
the
configuration
have
like
a
if
expression
is
this?
Do
this.
H
That's
something
that,
if
that's
something
that's
worth
having
like,
I
would
love
to
know,
because
I
would
also
love
this
for
testing,
because
I'm
working
on
testing
prometheus
metrics
right
now
and
they're
all
untyped.
So
if
we
needed
something
like
a
way
to
check
things,
expressionally
in
the
conflicts.
A
A
B
F
B
Yeah,
like
maybe
that
filter
could
then
be
the
one
that
you
would
script,
so
it
could
be
the
framework
to
do
something
on
the
first
packet
in
the
flow
and
then
you
would
allow
whatever
you
chose.
You
know
lure
scripting
or
whatever
for
the
for
that.
One
decision
that
you
make,
but
it's
hard
to
see,
yeah
I
expose
what
decisions
other
than
load
back.
What's
the
set
of
decisions
that
you'd
only
want
to
make
once
per
flow
versus
the
ones
you'd
want
to
make
once
per
packet,
I
mean
load.
B
Is
emphatically
once
the
flow
round?
That's
why
I
was
so
confused,
because
around
robbing
on
every
packet
causes
problems.
If
there's
any
state
in
the
back
ends,
plus
the
absence
of
an
ecmp
hash
option
in
the
load.
Balancer
was
what
threw
me
off,
because
that's
traditionally
what
you
do
if
you're
doing
on
every
packet
but
yeah,
I
don't
know
what
what
other
things
you
might
only
want
to
do
once
per
session.
F
Other
stateful
things
might
include
setting
state
for
encryption
if
you're
doing
encryption
between
two
quiltings,
for
example-
it's
kind
of
fuzzy.
There,
though,
because
like
let's
say
one
of
your
quiltings
picks
a
different
path
towards
the
same
end
server
you
you
might
need
to
re-hand
shake
there,
so
it's
not
quite
once
per
session,
but
once
per
I
don't
know
what
to
even
call
it
but
like,
if
you're,
using
a
stateful
protocol
for
two
quiltings
to
talk
to
each
other,
the
typical
aaron,
you
call
them
udp,
plus
plus
type
protocols.
F
I
think
that's
a
great
name
like
if
using
one
of
those
you
might
want
to
redo
some
things
if
you're
breaking
one
of
those
and
reinitializing
it
potentially,
if
you're,
adding
an
authentication
system
you
might,
you
would
also
do
that
at
the
start.
Have
an
extra
authentication
handshake,
I'm
thinking
of
dtls
like
things
here
where
dtls
has
a
handshake
at
the
start
and
then
a
bunch
of
extra
work
for
each
packet.
H
So,
just
on
the
question
read
back:
do
we
want
black
frost?
Should
people
go
tell
which
would
tell
people
to
write
roster?
Should
they
write
a
script?
I
think
the
main
hesitancy
that
I
would
have
right
now
to
recommend
someone
writing
is
that
they
would
have
to
deploy
qualcomm
binary
entirely
themselves.
F
H
End
yeah
yeah,
but
you
wouldn't
have
to
build
the
entire
cooking
thing.
You
could
have
install
quilt
and
then
install
this
filter
to
this
location
and
go
and
pick
it
up.
So
like
good
yeah,
it's
complicated
either
way.
That's
kind
of
my
point
is
like
either
like
do
you
have
a
a
good
way
for
people
to
provide
custom
filters?
F
H
Good
one
of
the
big
things
where
this
came
off
was
because
we've
been
doing
testing
and
like
one
of
the
big.
The
advantages
with
having
this
crypto
filter
is
that
we
would
have
to
add
essentially
less
cases
in
our
configuration,
because
you
could
have
a
filter.
That
is
the
test
that
runs.
You
know
it's
actually
assert
that
the
packet
is
equal
to
this
packet
at
the
end
of
the
entire
chain.
H
So
this
is
like
one
of
the
areas,
but
like
one
of
my
problems,
right
now,
I
was
like
was
missed
the
best
at
this.
The,
but
also
is
the
least
convenient,
because
there's
no
ahead
of
time
like
this
is
also
the
reason.
Probably
that
employee
is
not
using
this
there's.
No.
I'm
then
like
a
really
good
ahead
of
time
compiled
language
like
someone,
someone
should
make
this
a
lower
to
western
ahead
of
time.
Compiler.
That's
like
a.
H
But
right.
C
A
The
yeah,
the
idea
I
had
on
the
ticket
was
at
least
maybe
for
now
given
given
the
use
case
error
which
for
testing
100
like
do
we
do.
We
do
something
like
like
a
lure
or
pick
out
pick
a
scripting
lang
and
say
like
okay.
This
is
pretty
much
just
for
running
tests
and
maybe
integrate
it
with
that
cool
contest.
Sweet
we
were
talking
about,
maybe
in
some
way
shape
or
form
too
and
then
see.
If,
if
we
get,
we
find
more
things
down
the
line
and
expand
from
there.
Does
that
sound.
H
F
I
I
think
that
makes
sense.
I
I
almost
are.
I
argue
I
want
to
push
her
back
a
little.
Do
we
actually
need
this
for
testing
like
for
unit
tests?
We
can
just
do
everything
directly
and
implement
the
trade
in
our
test
and
have
a
custom
filter
there
directly
in
rust
code
right,
since
our
unit
tests
are
rust
and
for
real
world
testing.
I
guess
is
where
you're
talking
about
it.
That's
the
print
debug
info
and
stuff
right.
H
Well,
it's
more
like
integration
testing
and
also
to
sort
of
reduce
a
lot
of
the
units
that
we
currently
have,
because
we
have
initial
work.
We
have
a
lot
of
unit
tests
that
sort
of
are
that
require
that
are
interdependent
and
so
that
a
lot
of
changes
in
one
place
causes
a
lot
of
changes
to
happen
in
another
place.
That's
what
and
like
a
lot
of
the
tests
are
the
the
stuff
that
they're
testing
are
stuff.
H
That's
only
available
in
the
public
api,
so
it
is
instead
have
the
main
goal
right
now
is
to
have
okay,
just
basic
input,
output
checks
and
then
to
check
hey.
Did
this
prometheus
metric
tick
up?
That's
like
the
main
things
we're
going
for
now,
so
I
I
also
don't
think
the
scripting
thing
is
not
needed
immediately,
though
we
can
do
those
three
with
just
configuration.
We
don't
need
scripting
for
that.
H
F
I
I
think,
I'd
sooner
jump
on
like
a
d-block
filter,
which
you
can
feed.
Please
print
out
this
thing,
every
x
packet
or
something
along
those
lines
that
probably
serve
most
of
my
deep
debugging
needs.
If
I
need
to
start
dumping
packet
data,
I
probably
start
there
as
opposed
to
actual
dual
scripting
personally.
H
A
A
F
H
Also,
just
while
that's
on
that
screen
there,
I
just
wanna
point
this
out,
but
there's
I
don't
know
why.
But
on
254
bytes
on
mac,
I
can
very
consistently
get
very
long
delays.
You
can
see
in
those
plots
that.
A
Wonderful,
the
only
other
thing
I
had
was
for
these
monthly
sinks.
I'm
gonna.
I
need
to
still
set
up
the
calendar
april.
I
just
need
to
coordinate
some
stuff
with
you.
We
should
also
check
the
bitly
link
in
the
calendar
description,
which
I've
got
yet
to
ask
you
about
yesterday,
because
you
own
the
calendar
entry
ever.
A
You
did
all
that
oh
you're
on
it
all
right.
Well,
then,
I
will
review
that
and
I
will
send
a
pr
to
add
it
to
the
github
page
and
so
that
people
can
see
it
that
way
as
well.
Click.
A
A
Oh
okay,
oh
I'm,
sorry
I'll
do
that.
I've
got
a
pr
sitting
in
the
wings,
all
the
links
on
it
cool
and
then
once
the
recording's
live.
We
can
check
that
on
the
youtube
account
as
well
sweet
and
also
on
that
note
there
is
now
an
announcement
channel
and
discord.
A
I
promise
not
to
spam.
You
in
the
user's
meeting
anymore.
A
H
To
just
talk
about
it,
but
not
necessarily
like
that
thing,
which
is
sort
of
related
to
the
you
know
the
refactor
into
distinct
types
here
now
it
also
comes
up
because
we're
doing
the
test
subcommand.
I
have
to
add
this
ready
q
to
have
it
so
that
you
could
then,
when
the
test
fire
is
it's
the
starts
of
the
server
and
then
the
test
runner
knows
when
the
server
is
ready
and
it
can
start
sending,
it
doesn't
have
to
do.
You
know,
duration,
sleep
but,
like
I
finally
pointed
it
like.
G
You
mean
the
shutdown
channel
that
every
pretty
much
everything.
H
Listens
to
but
yeah,
but
also
like
yeah,
but
also
in
the
set,
like
I
said,
sending
the
stuff
like
the
cues
that
I
have,
for
example,
in
the
distinct
types,
is
another
example
of
where
there's
that
separation
of,
like
okay,
the
upstream
just
wants,
or
the
downstream
just
wants
to
send
the
packet
and
the
message.
But
right
now
it
has
to
construct
the
whole
thing,
whereas
that
could
just
be
a
message.
That's
sent
to
a
another
actor.
G
Yeah,
I
think,
like
overall,
like
my
thought,
was
mostly,
I
think,
they're
they're
like
pros
and
cons
to
both
sites
like
if
we
have
it
the
way
we
sort
of
the
way
we
have
now
versus
moving
to
a
cube-based
model.
Currently,
the
current
implementation
has
like
a
lot
of
boilerplates
that
you
mentioned,
but
I
think
that's
like
the
that's
the
evil,
then
tracking,
like
how
things
are
being
passed
between
channels
how
packets
are
being
passed
between
channels.
G
Like
I
mentioned,
there's
like
some
perf
considerations,
there
there's
also
the
fact
that
we
can't
we
can
no
longer
like
track
in
the
code.
How
a
packet
has
been
is
being
processed
without
like
understanding
the
big
picture
yeah,
it's
mostly
those
kind
of
things
that
that's
currently
the
my
thoughts
on
it.
I
would
say
I
don't
know
if
anyone
else
has
thoughts
on
that.
Well,.
H
Just
more
the
idea
of
hey,
we
have
it's
also.
We
have
like
essentially
a
lot
of
dynamic
components
that
need
to
change
a
lot
at
runtime.
Like
that's
also,
the
part
where
this
comes
in
is
we
aren't
able
to
take
advantage
of
a
lot
of
compile-time
stuff
to
reduce
a
lot
of
our
boilerplate,
because
we
have
to
write
v
tables
ourselves
and
stuff
in
order
to
be
able
to
do
everything
so
like
a
lot
nicer
to
use
an
actor
model
that
is
sort
of
designed
to
be
a
bit
more
dynamic,
but
can
also
be
typed.
H
F
A
There's
a
you,
you
did
it,
I
mean
aaron.
You
did
a
lot
of
that
work
in
the.
Where
is
it
where's
the
pr
refactor
quickens
data
type
pipeline
into
distinct
types?
Pr.
H
No,
it's
a
uh-huh,
it's
one
thing
I
would
say
to
that.
Is
that
there's
like
it
does
reduce
code
and
like
if
we
actually
a
lot
of
our
boilerplate
could
be
reduced
even
further,
because
we
do
kind
of
do
half
actor
model
type
stuff
for
the
stuff
we
do
need.
So
if
we
move
to
like
actics,
for
example,
you
know
there
is
like,
like
supervising
like
we
have
a
run,
pruned
session,
which
essentially
an
actor
supervisor.
A
Actually,
it's
a
good
question.
I
haven't
done
a
lot
of
stuff
with
actor
models.
Do
you
have
a
favorite
like
this
is
how
the
pattern
works
and
well.
H
I
just
use
activex
to
be
honest,
like
okay,.
H
H
No
okay,
I'm
pretty
bad
at
that.
I
have
to
think
about
that.
A
Yeah,
I
thought
I'd
ask
the
question
just
as
actually
that
way,
assuming
if
we
decide
that
that's
a
path
we
wanted
to
go
down,
that
would
probably
be
handy
to
point
to
in
our
country.
H
Yeah
definitely
well,
this
also
sort
of
comes
into
the
there's
another
issue
I
mean
you
know
creating
an
architecture
to
md,
because
we
are
like
at
like
ten
thousand
lines,
so
it'd
be
nice
to
have
a
very
high
level
overview.
I,
like
actress,
also
makes
actors
can
be
more
high
level
because
you
can
say
hey,
there's
this
person
that
just
sends
a
message.
A
A
Just
because
that
would
you
would
be
able
to
go
okay
as
like
a
first
contributor
coming
to
the
project.
How
hard
like
I
could
do
it.
You
could
almost
do
that
side-by-side
comparison
of.
Is
this
harder
to
understand,
or
is
this
easier
to
understand
which
side
which
some
pros
and
cons
would
be?
I
think
that
would
be
really
interesting.
Data.
F
A
Actually,
just
out
of
general
curiosity
as
well,
because
now
I'm
just
digging
into
this,
you
said
the
act.
Would
you
be?
Would
you
think
is
your?
Is
your
vision?
There
we
go,
there's
a
good
there's,
a
good
phrasing.
Is
your
vision,
kind
of
using
actics
or
do
you
because
I
know
you've
done
some
stuff
where
it's
sort
of
lightweight?
Oh.
H
A
H
E
I'm
playing
routine,
that's
all.
I
play
nice,
I'm
addicted.
H
I
just
finished
12
minutes
yesterday.
Don't
want
to
talk
about
it,
it's
not
a
good
game.
H
E
Once
we
get
cross
world
luna,
we'll
have
to
or
cross
data
center,
we'll
have
to
yeah.
Indeed,
hang
out.
F
E
F
F
F
Me
at
least,
I
think,
they're
still
wanting
to
add
new
servers,
though,
because
yeah
the
blizzard
shenanigans
have
caused
a
way
bigger
player
influx
than
your
average
expansion
release
does
for
them.
So
yeah
lots
of
players.
F
E
Just
saying,
there's
clearly
a
need-
and
you
know
so
yeah
anybody
else
playing
anything
exciting.
F
D
Yeah,
I
really
enjoyed
it
as
well,
like
I
probably
completed
90ish
percent,
except
for
one
quest
where
I
got
a
bug
but
yeah,
it's
a
good
game,
at
least
in
my
point
of
view,
and
I'm
playing
wise
seven
on
stereo,
which
is
pretty
cool.
It's
pretty
nice
rpg
game,
okay,.
E
Yeah
and
then
destiny,
two
just
had
a
new
season
or
release
drop
yesterday,
which
I
have
not
checked
out
yet
because
again
final
fantasy,
but
everyone
seems
very
excited
about
that
too.
So.
H
E
E
Thanks
everybody
for
joining,
especially
the
cisco
folks
for
being
interested
in
other
uses
that
we
had
not
considered.
So
we
will
get
recording
up
on
the
google
for
games,
youtube
channel,
we'll
put
a
link
in
the
doc,
and
then
we
will
see
everyone
again
next
month.