►
From YouTube: Istio Networking (Data Plane) WG meeting - 2019-10-17
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
B
B
Yeah,
so
we
have
both
my
implementation
down
the
merge
to
the
master
that
we
opium
112
my
beliefs,
and
we
also
get
that
updated.
You
II
steel
dependency,
proxy,
wasn't
so
and
so
forth
and
going
helpless
to
get
the
pilot
to
change
down
and
so
right
now
it's
merged
for
you
master
branch,
and
we
have
a
cherry-pick
PRM
1.4
branch
get
that
into
that
release.
Branch
as
well
and
diem
actually
just
play
the
visa
feature
a
couple
minutes
ago.
C
I
mean
the
good
name
is
work.
The
bad
news
is
I
think
we
need
to
rise
a
clear
user
guide,
how
user
to
migrate
form
the
current
state
into
this.
How
to
empty
OS.
Look
like
you
have
to
restart
everything
or
every
deploy
everything
I
try
to
find
out,
which
component
is
necessary
or
not,
but
the
first
trial
at
speeds
of
flag
doesn't
work.
I
have
to
restart
almost
everything,
including
this
work
lost,
is
no
control
plane,
so
it
could
be
a
bad
experience.
C
C
I
want
to
find
out
I
try
to
find
out,
I
think
that
some
dependency
on
policy
as
well
mixer
I,
don't
know
why
it's
like
exactly
chains
here
so
breaking
meow,
but
look
like
there's
some
condition
on
the
mixer.
That's
why
the
connection
actually
might
get
between
the
client
and
server
whatever
server
lose
a
policy.
B
Make
that
ham
flag
to
add
the
memcache
labor
ready
label
to
mixer
workload
as
well.
That's
a
prerequisite,
and
we
include
you,
sell
you
mine
for
I.
Hopefully
many
customers
upgrade
to
one
for
they
have
mixer
yeah
and
added
and
I
thought.
Okay.
I
can
confirm
that
even
the
flag
is
you
say
what
we
always
added
label
two
were
close,
you
might
for
yeah.
E
C
A
F
F
C
G
H
F
I
think
the
other
thing
is
that
if
one
point-three
proxies
will
continue
to
do
what
they
are
doing
today,
which
is
if
they
have
destination
rule
base,
MK
does
then
it
will
just
continue
to
work,
and
one
point
for
proxies
will
be
the
one
that
will
have
these
labels
and
the
whole
auto
MPLS
will
kick
in
only
when
we
fully
remove
the
destination
rules,
and
so
it's
like
you
know,
we
can
keep
the
destination
rule
around
for
one
more
release
and
then
remove
it
by
name
or
in
a
point
release
for
that
matter.
So.
G
F
F
I
F
B
F
D
D
G
G
B
C
B
F
So
I
think
what
we
have
is.
We
have
a
lot
of
a
serious
issue
with
the
whole
LPN
stuff
because
for
protocol
sniffing
to
work
on
the
inbound
path,
we
need
to
encode
LPNs
click
on
the
client
side.
They
still
HTTP
one
is
the
HTTP,
2
and
so
on
and
so
forth.
But
if
we
do
that,
then
the
problem
is
that
during
upgrades
one
point
for
proxy
talking
to
one
point
three
proxy:
that
communication
would
break
because
the
one
point
three
proxy
would
not
have
those
alts
encoded.
F
A
Yes,
so
two
things
to
comment:
one
is,
we
can
definitely,
you
know,
add
them
on
the
client
side.
So
we
are
future
proof
in
you
know
future
releases
isn't
of
requiring
all
the
time
list
of
appearance
in
the
released
right,
because
then
we
have
this
log
said
that
all
the
order
proxies
won't
work
with
with
this
feature,
so
we
should
definitely
at
least
configure
a
peon.
F
J
F
The
proxy
that
one
point
for
proxy
will
get
both
the
old
and
the
new,
a
LPNs
in
the
same
connection
right
and
when
it
sees
both
old
and
new.
The
proxy
needs
to
basically
still
pick
the
the
newer
LPN
and
activate
the
HTTP
filter
chain
and
not
fall
back
to
like
oh
I,
see
the
old
al
can
call
this
to
you.
Let
me
just
fall
back
to
the
plain
PCB
proxy
for
the
chain,
but.
F
A
F
A
F
Is
that
if
traffic
is
coming
from
a
outbound
listener
that
is
doing,
protocol
sniffing
and
and
it
detects
that
this
is
a
h2
protocol
and
that
information
has
to
be
passed
through
on
the
outbound
path?
That
is
where
we
wrote
this
ALP
and
filter
that
basically
looked
at
the
sniped
protocol
and
attach
the
appropriate
LPN
on
the
outbound
part
where
it
said
that
this
is
a
is
T
1.
This
is
H
2
connection,
ICU
and
HTTP,
1
connection
or
SEO,
and
a
HTTP
1.0
connection
or
whatever
it
is.
So.
F
This
was
a
LPN
that
the
client
basically
is
adding.
This
was
a
new
addition
in
1.4
and
but
on
the
server
side
when
you're
doing
protocol
sniffing
right,
because
this
is
for
the
port's
that
are
not
named
so
on
the
server
side.
When
you
are
doing
political
sniffing
today
by
default,
you
would
only
have
a
TCP
proxy
or
you
will.
F
Connect
you
to
either
have
a
TCP,
proxy
or
HTTP
connection
manager,
depending
on
the
sniffing
thing
right.
But
since
the
traffic
is
coming
over
TLS,
we
have
the
ability
to
attach
the
LPN
on
the
client
side
itself,
with
the
port
Yan's
filter,
while
doing,
and
it
basically
specified
that
you
know
this
is
a
HTTP
1
over
m
TLS,
the
HTTP
2
or
M,
TLS
or
whatnot,
and
then
on
the
receiving
side.
This
would
actually,
like
you
know,
look
at
the
AL
and
then
activate
either
the
HTTP
connection
manager
or
the
TCP
proxy.
A
F
F
F
Why
is
your
truck
I
think
the
column
was
this
right,
because
if
even
that's
coming
in
as
just
plain
history
of
then
the
TCP
proxy
would
always
get
selected.
We
wanted
the
maximal,
LT
and
set
match,
which
meant
that
you
know
either
that
they
were
two
options
in
this
case,
one
as
we
we
collapse.
This
whole
thing
into
like
this
ginormously
large
al
can,
like
we
steal
x,
blah
such
that
on
the
receiving
side.
F
What
I
meant
so
that's
that
if
we
do
that
change
then
then
we
would
not
need
to
invent
a
new
LPN.
So,
like
now
I
remember
yes,
I!
Don't
think
we
need
to
invent
new
lps.
We
can
retain
the
old
al
pians,
but
have
the
option
of
like
matching
everything
all
the
LP
ins
in
a
set
in
envoy
no
you're
touching,
and
that
way
we
would
just
think
so.
Just
work
seamlessly
and
but.
F
A
G
A
F
F
F
F
F
F
G
F
Yeah,
that's
true.
Now
everybody
at
even
went
in
the
comment
that
I
made
like
like
two
months
ago
for
the
things
on
the
summary
of
the
issue.
We
have
adjusted
this
and
semantics
and
but
then
like
know
when
I
saw
this
like
is
eh:
okay,
okay,
fine,
as
long
as
it
works
it
works,
and
then,
when
yesterday,
when
we
were
like
drawing
the
entire
matrix,
we
realized.
This
is
gonna
break
in
ten
different
ways.
So
we
still
need
the
envoys
I'd
change.
The
filter
chain.
G
G
F
F
A
One
mentioned
this
I
didn't
open
properly
issue
for
tracking
this,
yet,
but
in
the
end
to
beach
to
upstream
on
worry
releases,
an
upcoming
history
of
proxy
and
keep
it
going
forward
because
we
keep
too
many
like
we
spend
too
much
time
backwards.
Part
14
fixes
to
our
randomly
selected,
comets
and
stuff,
relying
on
upstream
providing
that.
A
A
A
Where
there
is
shown
like
three
to
four
months
scheduled
right
for
which
we
would
need
to
kind
of
adjust,
or
at
least
like
we
would
have
to
resit,
based
on
st.
Roch's
key
with
some
cherry
picks.
But
the
point
is-
and
one
thing
that
we
need
to
discuss
with
them,
is
whether
they
can
work
on
stability,
because
right
now,
on
way
of
seeing
only
the
lightness,
they
will
release
some
starting
next
week.
Then
1/12
right.
So
we
will
need
to
some
resources
for
backporting
to
stable
releases.
A
G
F
G
A
So
on
so
only
is
gonna
ship
112
next
week
and
then
we're
gonna
ship
one
for
based
on
that
and
one
five
either
we
stick
to
this
release
if
they
don't
release
anything
anywhere.
Let
me
switch
to
you
know
one
on
thirteen,
at
which
point
I
would
like
all
of
our
custom
extensions
to
use
the
web
assembly
API,
whether
it's
no
sandbox
or
now
like
they
would
be
still
compiled
as
small
sandbox
right,
because
what
was
somebody
isn't
GI
but
then
like
to
lease
from
now,
we
would
I
would
like
us
to
sweet
move.
A
G
F
A
J
Now,
right
now,
I
don't
think
there
are
any
committed
extensions
from
within.
We
have.
We
don't
have
plan
of
testing
out
one,
but
since
our
back
a
back
is
already
up
streamed.
It's
probably
unlikely
that
we
are
going
to
use
the
exactly
implemented
in
in
was
oh,
maybe
we'll
hit
for
a
performance
study.
A
F
F
F
F
A
The
first
step
is
to
port
them
to
no
sandbox
right
correct,
because
this
is
the
same
API.
Basically
we
we,
you
know,
we
just
compile
the
the
same
extension.
That's
different
target,
don't
think
that
didn't
mean
code
and
we
just
have
option
to
compile
it.
I
see
their
web
assembly
module
or,
as
the
native
extension.
F
H
G
A
A
I
If
there's
nothing
on
the
agent
I
just
want
to
introduce
myself,
my
name
is
Anil.
I
worked
for
a
redhead,
and
here
I'm
kind
of
attending
will
last
one
or
two
meetings
just
so
this
time.
I
want
to
introduce
so
people,
don't
wonder
who
is
this
guy
listening
it
so
I'm
planning
to
contribute
to
his
tea?
Oh
and
basically,
looking
at
you
know,
Federation
and
the
networking
aspect
of
it.
I
My
experiences,
mostly
in
the
networking,
mostly
in
Sdn
from
last
ten
years,
working
through
multiple
there
from
device
table
to
the
control
plane
and
a
wide
variety
of
that
so
looking
forward
to
working
with
you
closely,
hopefully
once
I
settle
in
and
warm
up
toward
the
code.
So,
hopefully
you
know
hoping
to
have
a
more
interaction.
Nice.