►
From YouTube: Network Service Mesh WG - 2018-06-22
Description
Join us for Kubernetes Forums Seoul, Sydney, 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
B
Sort
of
stealthily
added
to
the
agenda
under
events,
we
wanted
to
ask
if
the
case
and
their
advice
for
the
meetings,
it's
a
networking
I
know
that
there
was
some
comment
last
week
where
there
was
some
confusion
from
the
travel
pink
eyes
where
they
had
sort
of
you
know
said
hey.
You
know
we
wish
you
guys
would
have
sent
this
to
me.
It's
a
networking
list
like
everyone
else
does,
then
we
know
that
the
meeting
was
here
and
then
I
know
Sergey
wasn't
aware
of
the
meeting.
So
do
we
want
to
consider
politely
asking.
B
B
A
C
E
A
E
A
D
B
A
Yeah
I
completely
agree
and
the
easier
we
are
to
to
understand
the
the
architecture
for
newcomers
to
understand.
Not
only
means
people
get
to
work
faster
on
them,
but
also
means
that
our
architecture
remains
simple,
which
means
less
bugs
and
ideally
we're
more
potentially
more
flexibility
with
the
type
of
projects
can
they
can
and
also
yeah
completely
agree,
so
the
you
know
and
any
any
help
that
can
be
given
in
that
area
as
well.
It's
definitely
highly
highly
appreciated.
A
A
B
F
F
We
had
a
thing
at
what
is
on
Sudan,
also
I,
see
here
both
of
his
here:
okay,
hey
man.
He
may
want
to
speak
to
how
cumulative
voting
works.
In
general,
there
was
a
few
issues
on
the
how
the
votes
came
in,
but
I
will
share
them
and
then
won't
go
through
that.
G
F
B
F
A
B
F
Potentially
the
original
instructions
didn't
have
enough
on
that,
so
end
up
with
some
votes,
where
eight
was
placed
as
yeses
and
zeros.
This
knows
so.
Some
of
the
numbers
look
a
little
weird
it
ends
at
should
be
okay
with
the
time
because
it's
basically
the
preferred
times
everyone
else.
So
it's
okay,
thanks,
like
this
ones,.
H
F
And
I
am
I
was
actually
trying
to
go
in
and
see
if
I
could,
like
turn
them
all
into
proportional,
which
is
why
some
of
these
are
about,
and
so
this
was
a
proportion
of
the
total.
It
just
didn't
work,
because
some
of
the
others,
instead
of
being
ranked
like
that,
were
literally
like
a
proportionate
amount.
B
I
mean
it's
almost
feeling
to
me:
I
mean
we.
We
had
an
interesting
suggestion
for
this
portable
voting
then
always
like.
We
should
just
drop
back
to
doing
a
freaking
doodle
poll,
because
that
sure
works
and
the
real
question
is
is:
can
we
maximize
the
ability
of
the
community
to
attend
the
meeting,
and
the
doodle
poll
is
literally
just
a
I
can't
make
it
I
can't
make
it
in
the
generic?
Does
that
sound
reasonable
to
folks
I.
F
Think
we
end
up
with
the
same
amount
anyways
from
what
we
were
seeing
here.
It's
still
the
it's.
We
ended
up
with
proportionate
plus
the
point,
so
it
seems
to
be
okay
Watson.
Do
you
want
to
take
a
look
and
say
he's
he's
done
a
lot
of
the
voting
and
calculations
and
could
communicate
how
the
the
dot
voting
would
work
with
ranking.
G
B
G
G
Basically,
we
almost
had
it
it's
just
that
we
needed
to
have
some
validation.
It
just
says:
hey.
The
maximum
amount
of
points
is.
If
you
have
eight
points,
you
have
eight
choices.
You
have
eight
points,
you
have
to
distribute
them,
put
them
all
on
one.
You
can
spread
them
across,
but
that
way
the
people
who
you
know
when
you
have
a
second
choice.
Everyone
agrees
on
their
second
choice:
it
rises
to
the
top,
whereas
if
you
just
do
yes/no,
you
have
situations
where
you
have.
B
The
thing
that
you
understand
and
I
literally
don't
haven't,
thought
it
all.
The
way
through
is
what
I'll
have
is
if
I
minority
really
really
likes
the
particular
today,
the
majority
can't
make
it
on
that
day,
and
then
you
have
the
smearing
of
points
across
other
days
by
the
majority,
where
there
are
dates
that
more
people
could
make.
It's
just
that
minority
really
really
like
a
particular
day,
I'm
a
little
bit
that
you
would
wind
up
with
a
situation
where
we
have
a
day.
G
And,
and
so
the
way
it
happens
in
real
life
is
we
are
we?
Actually,
it
doesn't
happen
in
a
vacuum
right?
We
actually
can
so
you
communicate
before
you
put
the
before
you
put
the
options
out
you
communicate
after
they
put
slops.
Is
that
you're
like
I
like
Friday
and
I
like
Saturday?
It's
not
like
you're
doing
it
in
a
in
a
secretive
way,
where
no
one
can
use
your
true,
true
right.
B
So
I
mean
it
does
look
like
the
the
leading
candidates
of
the
current
time,
I
dunno.
This
is
kind
of
a
brutal
time
both
for
Israel
and
also
for
Europe,
because
it's
basically
you
know
evening
Friday
night
for
Europe,
it
looks
like
the
sort
of
second
most
favorite
time
is
on
Tuesday
7:00
to
8:00
a.m.
there
are
two
the
clues
that
there
are
four
people
who
couldn't
make
that
versus
six
people
who
can't
make
this
time
right.
B
B
So
it
does
look
like
this
time.
Has
this
time
does
have
oh
wait.
Now
it's
kind
of
the
wrong
number
0,
so
I
was
cutting
along
the
line,
so
Friday
8:00
to
9:00
a.m.
is
1
2
zeros
right.
So
we
have
2
people
today,
so
it
does
look
like
they
actually
also
maximizes
the
number
of
people
who
could
make
it
because
I
don't
see
any
as
few
reserves.
The
right.
B
H
E
B
B
A
H
B
I
would
like
to
do
and
probably
look
I
can
improve
the
degree
to
which
is
documented.
What
I
essentially
did
was
I
documented,
how
you
figure
out
from
inside
upon
information
you
can
send
outside
the
pod.
That
will
allow
you
to
know
what
the
namespace
of
the
hog
was
and
there's
a
link
to
it.
There
and
I'll
stick
a
link
into.
B
So
that
it's
in
the
meeting
minutes
as
well,
but
the
net
that
is,
you
can
actually
find
out.
You
can
always
find
out
inside
the
pod
in
Brock
the
inode
of
your
network
namespace,
and
if
you
are
in
a
position
to
be
mucking
with
network
namespaces
anyway,
which
the
NSO
has
to
be
then
you'll
have
access
to
something
like
far
run
that
NS,
where
you
can
look
for
that
I
note,
and
that
gives
you
the
name
of
the
network
name
space
to
work
with
and
a
file.
You
can
grab
ahold
of.
B
A
And
I
posted
the
link
to
the
to
the
wiki
as
well,
so
there's
a
lot
of
to-do
items.
If
anyone
wants
to
pick
any
to-do
item
up
and
work
on
it,
it
would
be
highly
appreciated,
otherwise,
I'm
going
to
be
working
down
them
as
well,
and
that
that
can
be
seen
on
the
bottom
of
the
wiki
getting
a
namespace
from
inside
a
pod
and.
C
D
New
button
yeah,
it's
nothing,
really
fancy
it's
just
I
ran
it
and
there's
a
couple
of
I
am
things
we
had
to
get
to
change
so
how
to
set
apart
from
that,
it's
really
just
taking
a
snapshot
of
the
current
code
and
getting
it
to
run
just
like
it
runs
today
on
mini
cube.
So
there's
nothing
that
you
know
fancy.
So
please
try
it
if
you,
if
you
need
to
use
Jiki,
try
it,
and
if
you
have
any
problems,
let
me
know
and
I'll
update
documentation,
try
to
help
you.
A
A
A
A
So
that
code
landed
a
couple
of
days
ago,
and
we
also
did
a
couple
improvements
to
the
CR
d2.
We
have
them
use
a
single
shared
Informer,
so
that
should
that,
should
that
should
reduce
memory
footprint
and
reduce
the
complexity
of
the
of
chef
of
shuffling
around
different
different
informers
at
this
particular
time.
A
A
So
these
are
part
of
the
kubernetes
client
client
api
sand.
Initially
we
were
creating
multiple
informers
turns
out
that
we
only
need
to
create
one
and
it's
and
everything
still
works
and
ideally
should
should
work
more
more
efficiently
and
the
last
part
that
we've
been
working
on-
and
this
is
still
active-
is
we're
working
on.
A
We,
we
had
a
version,
1.0
or
and
I
won't
point
one
of
the
network
service
of
like
getting
a
pod
to
the
network's
service,
mesh
daemon,
and
so
now
that
we
have
the
the
UNIX
socket
landing
where
we
can
use
it.
We
were
revisiting
the
client,
connectivity,
API
and
iterating
over
it
to
try
to
work
out.
What
kind
of
messages
can
we
can
we
send
to
like?
How
do
we
request
a
connection?
A
How
do
we,
how
do
we
expose
a
service
and
and
interact
with
the
network
service
manager,
and
how
do
we
set
the
labels
for
for
a
given
connection
to
requesting
and
specify
the
type
and
so
on?
So
this
is
probably
one
of
the
more
I
would
argue
this
one
of
the
more
important
parts
that
we're
going
to
be
working
on
over
over
the
next
several
over
the
next
several
weeks.
A
So
if
anyone
wants
to
get
involved
with
this
conversation
and
share
share,
your
experiences
is
to
like,
based
upon
what
you
know
of
of
the
domain
like
why
this
would
or
would
not
work.
That
would
be
incredibly
useful,
so
feel
free
to
join
in
on
the
conversations
there
best
way
to
join
in
in
these
particular
conversations
would
be
in
two
areas.
One
is
to
look
at
the
and
work
on
the
pull
request.
A
B
B
Know
one
of
the
things
that's
probably
worth
noting
is
this
is
something
that's
not
going
to
set
in
concrete
for
a
little
while.
So
you
know
it's
very
much,
how
can
we
get
working
so
well?
Well,
more
feedback
progressively
as
we
go
is
important,
don't
feel
like
the
train
is
leaving
the
station
just
now.
That's
a
terrible
feeling.
Yeah.
A
Completely
agree
with
that
as
well
like
the
do,
this
is
going
to
evolve
as
we
get
more
use
cases
and
involved
the
time
when
it
becomes
setting
concrete
is,
and
it's
going
to
be
malleable
concrete
to
a
degree
is
gonna,
be
when
we
get
to
the
to
the
1.0
0.0
release,
because
their
intention
is
to
use
semantic
versioning
and
so
a
that's.
That's
what
we
want
to
be
very
careful
with
with
changes
and
that's
not
to
say
that
we
can't
change
it
at
that
point.
A
But
it's
with
the
realization
that
if
we
do
change
it
and
people
are
people
who
are
depending
on
this,
are
going
to
have
work
to
do
in
order
to
ensure
that
they
migrate
because
of
a
backwards,
incompatible
change.
So
so
before
will
point
out,
we
know
like
if
you
see
something-
and
you
feel
like
this-
is
like
the
wrong
direction
or
we're
missing
a
use
case
or
some
important
detail
like
bring
it
up.
A
B
A
E
A
H
Well,
I,
don't
know
what
I
get
it
all
done
in
a
week,
but
I
should
we
should
be
able
to
make
some
progress
by
next
week,
but
I
just
want
to
start
I
I
want
to
I
stopped
working
on
it
on
anything
here
for
a
few
weeks
and
I
want
to
start
again
and
start
from
scratch
and
document
every
step,
I'm
I'm
doing
so
the
others
that.
J
H
A
A
F
F
The
actual
CNF
project
repo
itself
is
right
here.
I
can
draw
that
in
where's.
The
chat
go
when
you
are
sharing
here,
so
we
have
two
main
repositories
on
this
and
the
first
one
doesn't
have
a
lot
of
info
I'm
going
to
give
it
because
that's
River
moving
to
work
some
of
the
similar
stuff
with
what
y'all
were
talking
about
with
the
wiki's
and
everything
and
this
project
is
type
in
life
ona,
and
let
me
do
one
more,
this
BCP
a
use
case
which
is
bring
this
up
on
the
right
screen.
F
Okay,
so
and
we
are
and
Watson's
on
this
team
as
well,
so
we
are
part
of
the
cross
cloud,
CI
team
and
that's
actually
tied
on
to
an
action
item
for
later
as
well.
So
that's
this
project
here
and
we
helped
build
out
the
dashboard
for
the
CNC
F
and
we're
helping
now
on
this
project
for
trying
to
convert
or
port
or
recreate,
depending
on
how
you
want
to
look
at
it.
The
Venus
of
virtual
network
functions
to
work
on
kubernetes
and
as
containers.
F
So
that's
the
end
goal
and
what
we're
looking
at
is
these
containers
here,
which
that's
on
that
the
link
I
gave
to
this
wiki
and
aptoide
org.
So
looking
at
the
different
functions
like
the
broad,
the
Brookman
network
gateway,
this
vnf
and
the
this
gateway
here.
All
of
those
including
stuff
like
VD,
HCP
and
DNS,
and
all
of
these
pieces
that
tie
together-
and
this
particular
use
case-
is
of
interest
to
folks
like
AT&T,
wrong
way.
Orange
telecom
and.
F
This
implementation
that
we
see
here
in
this
diagram
is
the
all
these
network
functions
that
are
built
and
running
on
VMs
and
OpenStack.
So,
and
this
is
actually
a
pretty
old
diagram
and
the
set
up
was
I
think
originally
a
little
every
year
goes
and
what
we're
coming
in
to
do
on
this
particular
project
and
there's
not
a
lot
of
info
on
the
front
page.
Yet
it's
actually
more
of
it.
F
The
last
few
months
that
we've
been
working
on
us
trying
to
get
up
and
utilize
what
the
ona
group
and
Intel
and
all
the
different
folks
have
been
working
on
I
apologize,
I
think
he
said,
I
explained
his
CN
CF
is
as
well
so
that's
cloud
native
foundation,
so
the
foundation
is
interested
in
trying
to
get
any
software
working
as
cloud
native
and
containerized
and
portable
across
the
different
clouds.
So
that's
kind
of
been
our
focus
with
the
CI
staff.
F
So
right
now,
our
we've
this
originally.
What
we
were
focused
on
was
understanding
what
all
they
had
and
and
most
of
the
stuff
again
is
kind
of
in
the
projects,
and
issues
will
be
building
out
the
Dickey
and
readme,
based
on
what
we've
been
doing
the
last
couple
of
months,
which
is
really
digging
into
the
demo
code.
That
was
available
again.
F
F
F
Well,
once
you
start
dealing
with
all
of
these
other
network
functions,
they
will
need
something
like
network
services
mesh,
so
I
can
see
tying.
That
in
is
one
area
to
show
off
here's.
What,
as
a
a
I,
don't
want
to
say
proof
of
concept,
but
similar
very
network
service
mesh
could
really
help
and
then
I
see
potentially
we're
kind
of
in
the
early
stages
of
this
build-out,
but
potentially
having
a
non
own
app.
Maybe
a
cleanroom
version
of
the
this
use
case
built
out.
F
B
A
F
F
B
H
F
F
A
B
Also,
since
we're
talking
about
the
wiki
I
did
want
a
sort
of
crowdsource
one
idea,
which
is,
if
you,
if
any
of
you
guys,
have
seen
the
kubernetes
story
of
Fibby
if
you've
not
seen
it
do,
go,
look
for
PhIP.
Why
Fibby
and
kubernetes
it's
awesome.
It's
sort
of
the
story
explaining
kubernetes
as
a
children's
story
and
I've
been
kind
of
mulling,
doing
something
similar
for
MSM,
but
I
need
a
character
to
it
when
the
character
name
for
the
protagonist.
So
you
know,
if
you
guys,
have
any
interesting
ideas.
Please
resurface
them
at
some
point.
K
K
Books,
so
the
idea
is
when
you
deploy
you
get
qbp
a
server
calls
your
API
and
you.
You
are
running
a
server
inside
the
cluster
and
you
get
all
the
details
about
the
admission
and
then
you
can
mute,
you
can
patch
add
a
container,
add
a
volume
or
add
any
other
annotation.
So
that's
how
you
can
add
us
in
the
sidecar
inside
that
pod
and
that's
what
I
have
pasted
a
link
in
the
meeting
minutes
document
about
that
also.
K
B
It
certainly
sounds
potentially
quite
promising,
you
know,
I,
don't
think
any
of
us
actually
understand
it
well
enough
to
know
what
way
they
are,
but
but
it
certainly
sounds
very
interesting
to
me.
So
maybe
in
a
little
more
detail
might
be
helpful.
I've.
C
B
Cool,
so
would
you
be
thinking
that
we
could
use
that
for
injecting
like
an
in
container
or
something
to
to
set
up
the
connections
to
various
or
or
a
site,
or
a
sort
of
sidecar
in
the
pod?
To
talk
to
the
NSM
to
set
up
and
manage
the
stuff?
Is
that
the
Theban
thingy
pratik
yeah
yeah?
That
sounds
very
promising
to
me?
A
No
worries
thanks
for
thanks
for
coming
in
and
discussing
with
a
little
bit
of
time.
So
so
we
have
four
more
minutes.
I
want
to
make
sure
I
got
all
the
action
items
lined
up
properly,
so
top
of
the
action
items
lined
up
for
the
CI
Working
Group.
If
you
haven't,
if
you
intend
to
go,
we
have
a
distributed
action
item
to
invent
a
character
for
Edie.
A
We
have
Tom
looking
at
some
we're
looking
at
adding
documentation
to
help
people
who
are
entering
and
getting
started
if
Taylor
who's,
gonna,
add
documentation
for
C
and
C
F
and
for
them
for
CNF
and
the
project
that
he's
that
he's
been
working
on.
We
have
pratik
who
is
going
to
document
the
sidecar
admission
process,
and
so
is
there
anyone
that
I
have
or
anything
that
I
have
missed
or
any
action
items
that
we
have
not
discussed
that
you
want
to
add
yourself
to.
A
I'm
going
to
add
one
more
as
well.
Oops.
Sorry
gonna
add
that
if
people
who
want
to
get
involved
with
the
pod
to
NSM
API,
just
as
a
reminder
for
for
people
who
are
looking
at
this,
you
know
feel
free
to
feel
free
to
get
involved.
If
you
have
the,
if
you
have
the
time
at
this
particular
point
as
well
and.
A
I
B
Wait,
there's
a
YouTube
channel
they're
putting
along
and
there's
a
link
at
the
top
of
the
meeting
thing
to
the
YouTube
channel.
So,
okay,
yep
yep
anyway
cool
I
gotta
run
the
MINIX
call
you
later
cool,
take
care.