►
From YouTube: Quilkin Monthly Sync - January 2022
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).
B
C
A
Oh,
I
don't
know
how
which
is
ridiculous
in
like
covered
times,
I'm
usually
up
at
5
30
or
something
stupid
anyway.
Whatever
reason
my
brains,
don't
work
today,
cool
welcome
again
to
another
episode
of
what's
happening
inside
cooking
for
this
month,
good,
to
see
you
all
again
in
the
new
year.
A
Okay,
stuff
to
talk
about
today,
I
threw
some
stuff
on
the
agenda.
If
you
don't
have
the
bitly
link,
candy.
B
A
B
F
A
That
was
that
was
easy.
I
actually
think
aaron
you
did
this
already
anyway,
it
was
there.
E
A
E
B
Cddl,
I
think
that
is
fine.
Let
me
double
check
that,
while
we're.
A
I
think
it's
not
a
problem
and
I
think,
given
it's
a
dev
dependency,
I
don't
think
we
need
to
include
the
source
code
either.
A
B
A
Careful
and
I'd
say.
E
A
Nice
yeah,
so
I'm
building
out
some
benchmarks
that
are
like
as
pure
as
I
can
make
it
just
right
and
just
read
and
having
them
sit
next
to
each
other.
So
we
can
do
comparative
to
see
if,
like
one
is
slower
than
the
other
in
some
way,
shape
or
form,
because
the
metrics
seem
to
be
a
little
off.
I
have
the
feeling
some
of
that's
tied
to
the
the
fact
that
on
read,
everything
goes
through
that
single
channel.
B
We
haven't
got
that
multi-threaded
as
of
yet,
but
I
think
it'd
be
interesting
to
see
as
well
as.
A
A
That's
my
stuff
luna.
You
have
stuff.
D
I
have
stuff
yeah,
so
so
let
me
open
up
on
our
side
of
things.
If
you
only
left
embark,
that
doesn't
mean
we
are
losing
someone
that
contributed
before,
but
I
mean
he's
obviously
not
decided
from
working
on
the
project
anymore,
but
he's
not
doing
so
in
work
time
anymore,
I'll,
be
poking
up
people
from
my
team
to
start
just
to
see
who
has
the
interest
to
take
over
there?
D
Oh
so
much
needed
internal
details
as
far
as
I'm
concerned,
but
one
some
of
the
things
I'm
looking
at
and
I
think
I'll
start
looking
at
that
more
closely
after
our
c4.3
releases
cut
is
what
are
we
missing
to
run
cloaking
as
a
load
balancer,
so
we
have
an
internal
experiment
where
quilting
runs
on
the
load
balancer
it
works.
D
D
E
Think
the
only
thing
left
on
the
quilting
side
is
the
merging
the
match
filter,
because
that's
what
and
then
on
our
side,
I
just
writing
a
configuration
that
uses
that
match
filter.
That
should
be
it.
E
I
think
that
pr,
I
think
I
just
need
to
update
the
documentation
and
put
in
I
forget
the
almost
a
yama
schema
version
of
the
config,
so
that
you
don't
need
to
look
at
the
rust
stuff
once
that's
done,
it
can
be
immersed.
So
that
should
be
done
this
week.
I.
A
Also
figured
what
was
the
word,
but
I
think
we
document
like
that.
We
use
that
schema
thing
anywhere.
I
think
we
just
do
it.
We
should
we
don't
really
talk
about.
D
E
No
that's
already
there
the
token
the
matching
filter
is
just
more
for
if
you
have
changes
down
the
line
like
you
want
to
have
that
be
forwards
compatible,
so
that,
like,
for
example,
if
the
token
that
is
used
is
prefix
instead
of
suffixed,
that
you
can
properly
handle
that
or
yeah.
That's
a
token.
Is
it
like
a
different
length
and
stuff
like
that.
D
And
then,
on
the
g4
plug-in
side
of
things
me
and
mark
discussed
separately,
like
they,
they
already
published
the
ue4
plugin
for
agonize,
which
is
just
part
of
the
agonist
repto,
was
my
understanding.
So
we
can
do
the
exact
same
thing
for
quilking
and
that
shouldn't
be
an
issue.
A
A
That
is
true.
Thank
you.
Yeah.
We
have
a.
We
have
an
sd
case,
folder
in
which
we
have
all
our
different
language
sdks.
We
could
do
something
similar.
I
think
that
would
be
fine.
A
Well,
like,
I
think,
the
only
the
only
complicating
factor
we
ran
into
with
the
corners
wasn't
necessarily
open
source.
It
was
more.
How
do
we
like
do
it?
We,
so
we
don't
do
integrated
tests
for
our
unity
or
unreal
plugins.
It's
pretty
much
managed
by
the
by
the
by
the
end
users.
Some
of
that's
complexity.
Some
of
that's
also
just
managing
licenses.
I
still
haven't
unity-
is
actually
an
easy
one.
You
can
get
a
like
a
build
license,
which
I
actually
do
that
at
some
point
but
unreal.
A
D
D
I
I
think
that
should
be
fine
for
both
embark
and
google.
It's
fine
for
embark,
because
yeah.
A
Already
so
usually
so,
when
the
so,
the
the
the
the
issue
was,
if
we
had
some
sort
of
continuous
integration
system
that
actually
used
the
engine,
that
was
where
I
was
like.
I
don't
know
what
the
license
agreement
requires
you
to
do,
to
use
it
in
a
ci
environment
for
another
just
requires.
F
F
A
I
think
it's
fine
without
ci,
I
mean
you're
using
it,
so
it
works.
I
don't
even
know
how
to
do
ci
for
unreal
stuff.
A
D
To
be
clear,
we
solved
this
problem
internally
because
we
have
right,
but
it's
it's
the
type
of
problem
that
has
an
entire
team
working
on
it.
Just
stop.
Oh.
A
God,
cool
cool
yeah
that
may
end
up
in
the
two-hour
basket.
It
sounds
like,
then
we
people
just
use
the
plug-in
and
if
it
breaks
they
tell
us.
D
A
C
F
A
F
D
D
You
might
want
to
have
chains
of
quill
kids
like
if
you
run
cook
in
an
edge
location.
In
all
likelihood,
what
you're
going
to
want
is
a
setup
where
you
have
quilting
on
the
edge.
You
have
another
quilt
in
as
a
load
balancer
in
the
same
internal
network
as
your
game
servers,
and
then
you
use
the
internal
network
to
bounce
to
the
actual
game
server.
Rather
than
having
your
game
servers
exposed
to
the
internet
only
known
to
the
quilkins,
because
that's
silly,
what
do
we
need
to
do
that.
D
Can
we
pass
traffic
from
one
cloak
into
another
right
now
and
have
them
like
this
implies
like
having
a
quilting
on
the
edge
that
knows
about
pools
of
quill
kings
at
other
locations
and
for
each
routing
token
knows
which
pool
to
go
to,
and
then
the
load
balancer
one
knowing
the
exactly
of
oh
this
is.
This?
Is
a
game
server
within
my
pool?
I'm
gonna
send
it
there
it's
like.
Does
that
work?
We
also
need
to
avoid
loopbacks,
where
a
quilting
that
runs
as
load
balancers
thinks
a
particular
game.
C
D
C
A
A
E
B
E
D
Yeah
we
have
to
modify
a
bit
to
make
it
aware
of
how
we're
setting
up
our
pools
and
stuff,
but
that
that
that's
not
too
hard.
One
thing
I'm
worried
about
is
network
loop
prevention.
That
does
feel
like
something
cool
kim
should
do.
If
I
configure
two
quiltings
to
just
send
traffic
to
each
other
infinitely
can
I
I
I
want
to
one
be
able
to
notice
it
and
to
be
able
to
stop
it
without
taking
down
the
whole
goddamn
pool
of
both
of
them.
How
do
I
do
that.
D
That
or
bigger
loops,
where,
let's
say
you
exactly
you
do
a
round.
You
do
a
roundabout
like
mark
just
made
where
you
get
game
traffic
from
a
game,
client
pool
a
sends
it
to
pool
b,
pull
b,
sends
it
back
to
pull
a
which
sends
it
back
to
pool
b,
which
sends
it
back
to
pool
a
and
those
packets
will
bounce
back
and
forth
until
those
clusters
are
shut
down.
A
F
A
D
D
C
E
I
I
think,
at
the
very
least
it
doesn't
prevent
like
if
the
source
is
the
same
as
any
of
the
endpoints-
that's
not
prevented
at
the
moment,
but
that
should
be
prevented
I'll
make
it
that's.
E
Yeah
there's
also
happening
another
one
like
I
forget
what
it's
called,
essentially,
where
you
send
you
like,
where
a
malicious
actor
actually
sends
like
a
bad
ip
address
that
essentially
causes
it
to
loop
back
just
within
itself
like
not
even
talking
to
a
different
quilt,
necessarily
but
just
having
it
talk
to
the
same
qualcomm
and
having
it
send
it.
On
the
same.
In
the
same
instance,
yeah.
E
Yeah
exactly
there's
any
similar
over
the
network,
but
the
packet
says
that
came
from
local
hosts
and
stuff
like
that.
D
A
G
So
new
question
here:
do
you
have
the
equivalent
of
traceroute
for
quilking,
because
that
kind
of
reminds
me
of
a
bug
I've
actually
seen
and
traced
at
the
kind
of
tcp
routing
level
where
this
can
happen,
like
you
get
latency
spikes
between
two
locations?
And
you
don't
know
why,
and
that
could
be
a
routing
table
that
changed
somewhere
and
now
it's
rooted
somewhere
else
and
you
get
like
tons
of
loops
before
it
actually
comes
back
and
well.
The
first
thing
is:
it's
cool
to
be
able
to
detect
it,
but
then
what?
G
If
you
don't
have
the
tools
to
actually
see
your
topology
dump?
It
quickly
act
on
it
so
and
the
first
probably
tool
you
would
need
within
quilting
to
detect.
That
is
traceroute
like
give
me
all
the
quilting
service.
I
I
went
through
and
get
me
kind
of
from
there,
the
topology
of
what's
happening
when
I
contact
x
because
it's
gonna
be
like
if
your
mesh
is
really
big,
then
dumping
the
whole
topology
isn't
what
you
want.
G
You
want
to
detect
from
e
to
x
and
then
from
there
you're
gonna
get
metrics
you're
gonna
get
what's
going
on.
You
might
get
loops
and
so
on,
so
traceroute
could
be
an
interesting
tool
within
that,
where
each
working
could
just
add
its
span
information
kind
of
how
much
time.
When
does
it
see
that
packet?
Why
it's
rooting,
how
much
time
it
spent?
D
F
A
It's
just
like
it's
kind
of
like
it's
kind
of
like
okay.
You
could
set
up
some
sort
of
distributed
tracing
because
I
think
that's
useful
anyway,
so
that
you
have,
I
mean
you
might
as
well
use
those
tools
for
visualization
because
they
exist.
But
if
you
wanted
to
be
like,
how
do
I
simulate
it
like?
Is
there
a
simulation
packet?
I
can
send
that
just
not.
A
D
Sound
hard,
but
it's
actually
pretty
complicated.
It
implies
some
sort
of
meta
protocol
for
clothings
to
talk
to
each
other
at
a
bare
minimum
because
otherwise,
well
they
can't
do
that.
A
Well,
this
is
where
I'm
like.
If
we
did
again,
if
we
had
some
sort
of
distributed
like
we
had
a
distributed,
tracing
system
like
open
tracing
is
probably
the
right
bat
which
there
is
a
plug-in
for
the
tracing
library,
which
is
awesome.
We
and
and
luna
you-
and
I
were
talking
about
this.
The
other
day
like
you
would
have.
You
would
have
some
sort
of
trace
for
probably
the
routing
token,
all
the
way
through,
because
you
know
using
ips
and
stuff
is
pii
and
that's
a
whole
other
issue.
A
But
if
you
had
your
magic
thing,
you
got
your
tracing
token
and
you
could
see
it
going
through
your
topology
for
all
the
packets
that
go
through
your
system,
one.
You
could
use
that
as
your
warning
system
for
like
hey.
If
you
have
something
that's
a
loopback,
because
you
could
see
it
in
your
trace,
you'd
be
able
to
visualize
an
alert
on
that
which
would
be
sweet,
but
then
it's
it's
then
quickens
don't
need
to
know
about
each
other.
It's
the
open!
A
D
D
A
A
You'd,
send
it
wait
30
seconds
and
then
you'd
see
the
whole
thing
or
whatever
the
batch
time
is
like.
I
would.
I
would
batch
on
time
so
I'd
be
like
okay,
let's
look
at
all
the
packets,
we
got
in,
like
maybe
five
seconds,
aggregate
that
within
that
instance,
and
then
that
send
that
off
to
tracing.
Maybe
maybe
it's
an
idea
because
then
you're
guaranteed
to
always
get
the
path.
A
D
No,
if
you
have
a
sample
rate
of
0.1
you're
sampling,
one
in
every
thousand
packets.
C
D
D
How
traceroute
works,
and
it's
straightforward
is
that
so
packets
have
a
a
loud
number
of
bounces
right
and
if
you
say,
and
how,
how
and
if,
if
you
send
a
packet
and
it
reaches
its
maximum
number
of
bounces,
you
get
a
response
back.
That
says
I
discarded
this
packet
because
it
reached
its
maximum
can't
we
do
the
exact.
C
B
D
Need
any
you
neither
does
ip,
you
don't
need
any
so
wha
how
a
traceroute
works
behind
the
scenes
is
that
you
send
a
packet
with
a
maximum
number
of
hops
with
one
you
see
where
it
lands.
You
see
what
responds
to
you
with
a
hey
I
discarded
this,
then
you
send
one
with
a
maximum
hops
of
two.
You
see
what
responds
to
you
with
hey,
I
discarded
this.
You
send
one
with
a
maximum
number
of
hops
of
three
you
see
who
replies
to
you
with
saying
they
discarded
this
etc.
G
I
I
would
be
a
bit
worried
of
exploding.
All
of
that,
given
at
least
for
me,
one
of
the
interesting
aspects
of
culking
is
kind
of
creating
that
connection.
I
don't
know
after
that
hearing
you
all
talk
about
this.
I
think
I
agree
that
the
mechanism
behind
the
traceroute
you
could
reuse,
what's
probably
currently
acting
as
your
as
your
analytics
system,
that
that
kind
of
measures
how
times
between
ops
operate.
G
Because
that's
what
you
want
at
the
end
of
the
day
you
want
to
span
throughout
each
thing
if
you
don't
want
to
to.
Actually
I
don't
know
if
there's
a
mechanism
right
now
in
kolkin
to
say
well,
this
packet
is
kind
of
a
special
packet
like
activate
the
the
analytics
for
it,
and
I
think
the
answer
is
no.
To
that
question.
E
D
I
have
a
concrete
suggestion
here,
so
technical
suggestion
implement
a
new
type
of
packet
called
trace.
Let's
say
that
the
packet
is
literally
in
ud
in
utfa.
The
word
trace,
followed
by
a
routing
token,
followed
by
a
maximum
number
of
hops,
and
if
trace
is
on
on
that
particular
quilting.
When
the
like,
let
me
let
me
start
that
packet
is
sent
each
hop,
that
cool
kind
sees
the
packet.
It
reduces
the
maximum
hops
by
one
when
the
maximum
hops
reaches
zero.
D
The
let's
say
we
have
an
ip
address
in
that
packet
as
well
that
ip
address
receives
a
udp
data
ground.
Back
that
says,
hi
trace
ended
here.
If
the
trace
feature
is
on
on
that
particular
quilting.
If
it's
off
it
is
silently
discarded
when
it
reaches
zero
and
no
response
is
given
back
this
lets.
You
essentially
define
traceable
parts
of
your
infrastructure,
all
your
edges
and
your
entry
load
balancers,
while
keeping
internal
network
details,
private
and
opaque
to
end
users,
and
that
gives
you
the
visibility.
D
Like
just
baseline
ideas,
I'd
love
to
see
what
people
think
of
this.
I
I'm
not
saying
it's
a
good
idea
of
the
top
of
my
head.
This
is
what
I
can
think
of.
This
is
basically
almost
replicating
what
ip
does
ip
has
the
same
concepts
like
you
have
a
sender
to
your
packet
and
when
max
hop,
reaches
zero.
That
sender,
not
the
previous
hop,
but
that
actual
sender
gets
a
reply
back
hey.
D
E
Yeah,
just
off
the
off
the
top
of
my
head.
What
comes
from
how
to
implement
that?
You
know
it
within
the
current
system
is:
do
you
have
to
have
a
filter
at
the
front
that
says:
hey
attach
this
this
piece
of
data
onto
the
packet
and
then
you
also
have
a
filter.
That's
like
that
checks
for
that
piece
of
metadata
on
every
single
one.
That
says:
hey
is
this
one
here,
okay
decremented
it.
E
I
think
I
think
it's
relatively
straightforward
with
like
with
the
dynamic
metadata
and
just
having
a
cooking
that
picks
it
up
and
applies
a
packet
that
then
flies
back.
That
has
that.
Has
it
like
the
identifier
that
you
can
then
other
questions
could
pick
up
and
be
hey?
Okay.
This
is
the
part
that
we
use
for
the
counter,
and
then
they
just
parse
the
counter,
and
then
you
know
decrement
it
or
increment
it.
So
I
think
it'd
be
relatively
straightforward,
but
obviously
have
to
try
it
out.
A
For
whatever
reason
yeah,
I
think
you
could
do
it.
I
think
the
only
other
concern
that
I'll
concern.
The
thing
is
sweats
right
now:
filters
can't
readily
send
packets
too
easily
or
like
return
packets
to
cinder,
but.
A
A
E
It
could
it
could
just
change
the
address,
that's
sending
too
so
like
if
it
runs
out
at
the
counter.
Couldn't
it
just
say.
B
A
A
G
A
Handle
something
that
has
a
handshake.
We
don't
really
have
a
way
of
doing
that.
Right
now,
I'd
say:
there's
a
there's,
a
probably
a
more
interesting
discussion
or.
A
Interesting,
but
a
discussion
around
that
kind
of
topic
that
would
probably
tie
into
this
yeah.
The
only
thought
I
have
on
this
as
well
is
this
is
a
this
would
definitely
be
a
handy,
debugging
tool.
I
definitely
want
to
see
us
lean
more
like
how
do
we?
How
do
we
do
observability
on
the
platform
as
a
whole,
and
that's
where
I
look
more
distributed
tracing
we
have.
A
D
D
A
A
D
It
solves
that
the
trace
packet
proposal
will
be
just
discussed
so.
C
C
A
Yep,
all
right
cool
cool
cool
that
actually,
that
has
that
actually
has
multiple
use
cases.
Then,
in
that
instance,
because
as
you,
because
you
you
and
I
were
discussing
this
the
other
day,
you
could
have
it
set
up
so
that,
like
every
30
seconds,
it
just
sends
it
through
the
system
and
just
tracks
that
data
and
takes
that
part
of
the
observability
thing.
So
you
have
that.
D
A
D
A
A
Is
especially
if
it's
a
filter
in
the
way
the
filters
are
set
up
like
what
makes
a
tracing
packet
could
be
unique,
it
could
be
cryptographically
generated.
So
if,
if
you
were
to
like
expose
that
to
the
outside
world,
if
no
one
else
knows
what
that
that's
a
bit
of
security
through
security
but
like
if,
if
no
one
else
knows
what
it
is,
and
maybe
it
rotates
every
so
often
it
might
give
you
some
security
in
that,
like
I.
D
Mean
yeah
that
makes
sense.
You
could
do
something
that
way.
If
you
wanted
to
yeah
right,
it
doesn't.
E
Also,
just
on
like
having
meta
protocols,
I
think,
as
long
as
we
have
it
so
that
you
know,
essentially,
the
metaphorical
remains
within
the
clock
and
network.
It
should
be
mostly
fine,
because
that's
kind
of
what
we're
trying
to
do
is
like
transparent
transform.
But
you
know
it's
fine.
If
we
have
bits
of
metadata
between
point
a
and
b
as
long.
C
B
D
E
So
I
think
a
good
question
to
think
about
is
because
we're
doing
udp
and
it's
a
velocity
protocol
like
we
could
have
that
protocol,
be
the
prefix
of
the
packet
which
guarantees
that
I'll
basically
always
be
there
or
if
we
have
it
as
the
suffix,
it
could
be
lost
and
like
do
it
like
no.
E
C
A
The
other
thing
I
was
going
to
say
about
meta
protocols.
I
don't
I
don't
think
that
should
be
something
we
show
away
from.
I
think
we're
already
down
that
path.
A
Anyway,
I
mean,
and
even
even
down
to
like
whether
you
prefix
or
suffix
a
routing
token,
like
that's
kind
of
the
cook
and
meta
protocol
already,
I
think
cooking
is
really
about
like
you're
building
up
your
own
protocol
within,
like
your
set
of
filters
that
is
particular
to
how
your
game
works
and
because
udp
is
the
wild
west
and
everyone
does
it
differently
right,
like
we
kind
of
have
to
do
that
anyway.
So
I
don't,
I
don't
think,
that's
a
problem
anyway.
D
A
A
Yeah,
it's
people
will
complain.
You
know,
you
know
what
game
developers
are
like.
I
love
everyone's
smile.
My
only
thought
there
is
actually
probably
more
of
a.
If
there's
no
prefix,
then
we
know
it's
a
game
packet.
If
there
is
a
prefix,
then
it's
a
quilting
thing
which
we
can
check
for
and
I
think
that's
difficult,
but
at.
D
If
we
define
a
game
packet
format
like
off
top
of
my
head,
let's
say
it
is
q,
l,
k,
n
for
quilkin,
without
old
vowels,
followed
by
two
bytes
for
a
packet
type,
that's
six
bytes
and
then
like
you,
could
have
a
game
packet
type
and
then
the
game
packet
type
v2.
If
you
ever
want
to
change
how
your
game,
packets,
are
formatted
and
yeah,
I
mean
as
long
as
we
make
sure
that
can
be
turned
off.
D
G
Are
you
proposing
that
those
bites
be
there
and
be
added
by
the
client
themselves.
G
E
E
A
Which,
admittedly,
we
have
is
a
problem
anywhere
with
routing
tokens
you
have
to
you
have
to
consider
your
writing
text
now.
The
good
thing
about
that
is,
we
make
the
end
user,
define
their
routing
tokens,
so
they
have
to
be
aware
of
them.
So,
ideally,
it's
something
they
should
be
aware
of
right
like
we're
like
what
what
you
create
your
routing
tokens.
So
you
know
what
their
length
is
going
to
be
within
your
system.
A
If
we
start
adding
our
own
extra
stuff,
that's
not
explicit
to
packets,
not
impossible,
but
I
think
it
becomes
a
little
less
transparent.
G
So
again
new
here
I
don't
know
the
details
of
quilting,
but,
and
that
might
be
a
good
thing,
though,
does
the
client
define
well?
Does
quill
quilting
impose
mtu
on
packets
and
if
so,
then
you'd
have
basically
the
the
you
could
define
the
space
for
an
envelope
that,
at
the
quality
level
you're
guaranteed
to
have
this
many
bytes
you
can
use
for
envelopes
and
and
types
and
so
on.
So
that
could
be
a
way
to
keep
the
clients
kind
of
sending
naked
packets.
G
You
would
still
kind
of
remove
those
bytes
from
them,
but
without
defining
something
at
the
client
level.
So
that's
something,
then
you
can
change
at
the
quilting
level.
You
can
update
that,
as
you
see
fit
can
be
reused
for
different
purposes.
You
don't
need
to
update
clients.
E
So
just
how
it
works
at
the
moment
is:
it
is
just
the
maximum.
U
u16
max!
Every
time
there
is
a
issue
opened
up
to
be
like
because
it's
not
the
the
max
mtu
is
not
always
fixed.
It
can
actually
be
set
at
like
the
linux
level,
like
you
can
set
the
kernel
to
say
hey.
This
is
the
max
mtu
that
you're
going
to
allow.
So
eventually
we
probably
want
to
have
it
so
that
it
just
matches
that,
but.
D
Even
worse,
every
piece
of
networking
gear
between
your
computer
and
the
answer
can
have
an
opinion
on
what
the
maximum
mtu
is
going
to
be
before
it
forwards
your
packet
or
decides
to
drop
it.
So
your
real
maximum
mtu
is,
in
fact
the
lowest
mtu
between
all
the
hops
between
you
and
the
server
that
that.
A
A
B
D
E
E
Other
interesting
possibilities
there
yeah.
Definitely
you
could
definitely
just,
but
I
think
the
thing
at
the
moment
is
we
don't
have
like
configuration
for
like
generating
those
pieces
on
the
fly
from
like
kulken
or
something.
G
E
A
A
D
A
D
Ue
the
engine
plug-ins
we
built,
could
have
the
same
same
thing
features
yeah.
One
thing
that
you
also
need
is
to
be
able
to
strip
data
off
a
packet
with
a
filter.
If
you
need
that,
so
the
final
hop
can
drop
those
packets
that
will
be
a
match,
filter
or
if
you
have
a
game
packet
type.
That's
let's
say
it's
qlqn,
followed
by
zero
one
to
strip
that
off
before
giving
the
data
to
the
game.
E
Well,
for
like
optionally
stripping
it
like
what
we
could
have
is
that
is
for
that
quilting
recognizes
different
types
and
just
says:
there's
a
piece
of
metadata
that
says
what
kind
of
packet
it
is
that
you
can
look
at
and
say.
Is
this
a
game
or
a
network
or
you
know,
trace
in
the
config
and
then
you
can
say
if
it
is
strip
out
these
nine
bytes
or
whatever
yeah.
D
E
Just
to
go
just
go
back
to
this
mtu
thing,
thinking
more
about
like,
so
how
does
it
work
because,
like
for
example,
like
again
you
udps
mtu
is
like
u16,
but
ethernet
is
only
like
1500
bytes,
which
I
believe
is
like
when
you
have
over
that
that's
when
you're
in
the
risk
area
of
like
packet,
loss
and
stuff.
So
how
do
you
know
how
that
works
at
that
transition?
Yeah.
D
From
the
yeah
I
do
so
a
udp
datagram
is
defined,
so
a
udp
datagram
has
a
length
on
the
ip
level
like
a
udp.
Datagram
corresponds
to
one
ip
packet
and
the
ip
packet
has
a
length.
If
an
ip
pack
is
too
long
to
be
sent
over
the
lower
level,
physical
layer
or
any
of
the
lower
layers
below
the
ip
level.
I
forgot
their
names,
they
don't
matter,
then
your
ip
packet
will
become
fragmented
and
a
fragmented
packet
will
only
be
received
by
the
other
end.
D
D
If
I
do
a
quick
google,
an
mtu
of
1492
is
standard
for
ppoe,
which
is
commonly
used.
D
Which
is
why
1500
is
risk
area,
because
1492
is
the
most.
You
can
send
overlap,
which
is
by
the
way
1500
minus
8
bytes
for
the
ip
level,
things
that
get
added
onto
your
ip
packet.
C
D
Yeah
yeah
for
things
that
are
tagged
with
my
name.
I
think
it's
only
fair,
I'm
really
interested
in
the
meta
protocol
discussion
in
particular,
and
I
think
it
would
make
things
a
lot
easier
to
diagnose
if
we
have
it
so
I'll
make
sure
to
log
all
of
those.
A
E
Oh
well,
marcus
isn't
here
anymore,
but
I'll
answer
his
question
from
the
chat,
which
is
like,
isn't
a
single
bite
enough
for
the
identifier.
It's
like.
We
want
to
have
enough
disambiguation
that
it's
obvious
that
it's
quoting
because
like
if
you
did
something
that
was
like
a
two-bit
identifier
or
something
like
that
like
like
it'd,
be
very
easily
mixed
with
real
data.
E
D
It's
why
I
said
four
rather
than
one
because
like
if
it
just
happens
by
chance
to
hit
that
yeah.
C
D
A
It's
an
interesting
question:
would
you
want
yeah?
We
should
talk
about
this
because
there's
like
I'm
like,
would
you
want
your
envelope
to
start
with
something
that's
game
specific
rather
than
like
per
game?
Would
you
want
it
per
client?
Would
you
like
again,
but
then
you
end
up
in
a
routing
token
right.
If
it's
per
client
is
the
is
the
is
the
routing
token
enough
for
an
envelope,
because
then
you
know
what
it
is
like.
I
don't
know,
I'm
just
yeah,
I'm
throwing
ideas
there
and
there's
a
big.
E
Yeah,
I
also
think
configuration
is
oh
yeah,
a
better
value.
Add
because,
just
even
for
okay,
you've,
already
shipped
a
game
and,
like
you
know,
people
who
want
to
just
put
this
in
front
of
the
game
like
they're,
not
even
the
developers
of
this
game,
they
want
to
they're
just
hosting
these
servers
and
they
want
to
configure
it
to
work
like
that,
like
cooking.
If
it's
configurable
quilting
can
work
in
those
scenarios
where,
if
we
have
this
requirement,
it
becomes
more
restricted.
A
Yeah,
I
think
in
and
like
when
we
look
at
like
the
the
topologies
that
we
have
like
the
integration
options
of
you,
can
pick
and
choose
how
how
much
you
want
to
do
with
it
that
that
becomes
a.
I
think.
It's
the
value,
add
then,
like
anything
else,
you
can,
it
does
mean
configuration
becomes
more
complex
because
we're
less
opinionated,
but
I
think
that's
probably,
I
think
the
the
ethos
that
we're
getting
for
with
the
project.
D
But
I
don't
think
we
can
get
away
from
that.
I
don't
think
so
like
if
you
look
at
invoice
config,
it's
absurdly
and
annoyingly
specific,
but
I
I
don't
see
a
way
around
that
if
we
want
to
support
all
these
options
and
having
different,
like
you
end
up
having
different
hops
of
your
quilting
network
configured
differently
right,
like
your
site,
car
is
configured
differently
from
your
load
balance,
which
is
configured.
A
A
D
C
D
A
It
like
you're
like
do.
I
want
the
envelopes.
Do
I
not
want
envelopes?
Do
I
want
a
prefix?
Do
I
want
a
suffix?
Do
I
want
to
like
you're
you're
you're,
choosing
your
bits
as
you
go
through
it,
so
choose
your
own
adventure
like
how
you
how
you
want
to
make
your
udp
protocols
work,
which
I
think
is
super
handy.
D
I
think
it's
super
cool
yeah.
I
I'm
more
time
to
go
towards
your
suggested
approach.
Rather
than
have
things
hard
coded
in
with
modes
and
stuff,
I
think
it
makes
more
sense
and
more
flexible
and
what's
gonna
happen
in
reality.
Is
you're
gonna,
hard
code
that
in
your
management,
server
right,
yeah,
you're,
just
gonna,
say
in
your
management
server.
All
my
sidecars
will
be
configured
like
this,
and
all
my
load
balancers
will
be
configured
like
this.
D
B
E
Out
there
is
like
there's
nothing
stopping
us
from
also
supporting
things
like
dns
and
quick
and
stuff
that
can
be
like
a
whole
filter
onto
itself.
It's
just
like
what
we
are
trying
to
provide
is
for
all
these
options
that
are
like
bespoke
and
you
don't
you
don't
want
to
write
like
a
whole
filter.
You
just
want
to
say:
hey
just
append
these
bytes.
Just
do
this
small
thing
on
top
of
udp,
I
don't
want
to
have
to.