►
From YouTube: Meshery Community Meeting - May 31st 2019
Description
We discuss @smi_spec with @sheriffjackson and @erikveld of @hashicorp in a post #KubeCon #Meshery community meeting with shoutout to @pbouwer, @nservicemesh and @premsankar.
B
A
C
C
C
C
C
C
C
C
E
C
D
Sounds
good
so
essentially,
we
have
one
person
from
frames
team
supreme
as
part
of
the
NSM
adapter
network,
so
a
Spanish
project
and
the
discussion
we
had
earlier
today
was
actually
to
discuss
on
what
is
the
right
approach
for
proceeding
with
implementing
an
adapter
furnace,
because
Edison
really
is
not
a
lol
five
to
seven
kind
of
a
sub
smash.
Like
you
know,
it's
a
different
kind
of
the
mesh
which
actually
works
at
a
much
lower
level,
so
we
were
actually
talking
about
some
of
the
options.
D
So
so
the
call
went
on
for
about
like
40
minutes
where
he
actually
first
introduced
me
to
NSF,
and
then
you
know,
we
then
took
a
total
of
that
measure,
ease
or
touch
or
diagram
and
we're
kind
of,
like
you
know,
going
over
all
the
options,
our
potentials
so
towards
the
end.
You
know
I
kind
of
like
showcased
him,
one
of
the
current
adapters
and
what
the
how
it
is
written.
D
So
he
actually
has
taken
on
some
points
as
the
foot
cathodic
like
know,
so
he
actually
said
that,
like
he,
he
will
work
with
the
person
from
his
team.
Her
name
is
Sahar
and
they'll,
probably
like
you,
know,
kind
of
have
some
follow-up
meetings
with
us
actually
kind
of
ensure
that,
but
are
they
are
working
in
the
right
direction
and
also
it's
it's
also
a
dual
thing
which
or
that
you
know
now,
that
is
a
very
different
kind
of
fun
and
actor.
D
You
know
my
option
or
my
offering
to
him
was
like
we
are
willing
to
support
you
guys.
The
way
you
would
want
this
to
having
no,
what
I
told
him
so
so,
essentially,
that's
kind
of
looking
where
we
ended,
so
he
has
taken
some
points
on
he's.
Gonna
give
a
give
it
a
thought
with
the
team,
and
then
we
come
back
to
us
discussions,
nice,
okay,.
F
Serve
a
little
lake
where
we
were
trying
to
find
some
way
to
the
state
where
him
were
out
at
all,
like
company
retreat
this
week
out
in
Austin
Angie
in
Texas,
they're
tearing
everything
down
so
like
all
of
the
community
rooms
that
we
did
have,
which
is
like
those
people,
shifting
boxes
and
stuff.
So
sorry
I'm,
so
apologies
for
the
delay,
but
really
great
as
see
you
all.
Oh
yeah,.
C
E
Sure,
hey
I'm,
Sacco
I'm
doing
the
mobs
in
open
go
here
in
deadwood
based,
so
that's
before
that
I
was
in
engineers
working
in
John
mesh
project
yourself
as
meshes.
So
it's
really
nice
to
join
to
mesially
community,
which
is
actively
he's
discovering
all
the
service
mesh.
That's
I!
Guess
it's
the
next
thing
for
append:
oh
yeah,
let's
pretty
much
it!
Oh.
E
A
F
F
So
what
we
do
is
we
work
in
the
the
developer
advocacy
team
so
that
developer
advocates
where
developer
advocacy
differs
from
from
some
businesses.
We
are
actually
working
with
engineering,
so
I'll
we're
both
engineers
from
a
background
perspective
in
our
role
was
quite
wide,
so
it
it's
some
education,
it's
also
giving
product
feedback
from
like
a
user
perspective
and
also
kind
of
helping
to
to
contribute
to
the
products
as
well.
So
we
we
have
a
really
wide
gamut
of
stuff.
C
F
C
F
That's
right
so
I
I
worked
on
the
spec
for
the
service
method,
swebus
mesh
interfaith
service,
present
effects.
We've
got
a
terrible,
misappropriation
of
the
and
but
also
Arabic
and
I
delivered
the
the
first
sort
of
me
out
for
release
the
console
connected
integration.
We
you
only
delivered
the
the
soul
of
the
security
policy
side
of
things
right
now,
because
console
Connect
is
is
growing.
By
the
time
we
released
at
one
point,
six
release
which
is
coming
in
about
six
weeks.
C
F
Right,
yes,
we
we
so
console.
We
really.
We
were
literally
discussing
on
this
week.
We
have
kind
of
confused
marketing
around
the
naming.
Console
connect
was
was
the
feature
name,
but
but
actually
the
reality
is
that
the
console
is
is
just
the
thing.
So
a
feature
of
console
is
that
it
can
act
as
a
as
a
service
match.
We
see
that
as
kind
of
like
it's
like
a
higher
level.
F
Abstraction
features
on
top
of
the
the
primitives
which
were
always
built
being
built
into
the
console
system,
such
as
the
system,
catalog
health
checking,
distributed
configuration
and
stuff
like
that,
so
it
where
I
apologize
on
the
the
naming,
but
yet
console
console
one
point:
six
and
that's
a
control
plane.
We
use
envoy
as
a
data
plane,
and
we
also
have
a
built
in
layer
for
proxy
to
support
Windows,
but
predominantly
our
lawyers
is.
These
are
preferred
deadly,
but
okay.
F
So
yeah
and
then
that's
something
that
we're
quite
quite
keen
on
well,
but
one
of
the
advancements
will
be
making
this
year
is.
We
are
introducing
a
decoupling
of
the
data
plane
from
the
control
plane,
so
console
console
service
mesh
will
be
able
to
support
a
wide
variety
of
control
planes
as
long
as
that
length.
Yes,
deflector
planes
I'm.
Sorry,
it's
like
July,
so.
F
Which
I
think
is
gonna,
be
really
exciting
from
a
community
perspective
is
alone
yeah
a
new
office
will
be
who
build
their
own
data
planes
and
I.
Think
envoy,
has
the
market
mind
share
on
that
right
now,
but
you
know
everybody
might
be
using
own
voice
right
and
you
know
how
this
stuff
works,
we're
what
they
call
in
the
tech
industry,
so
everything's
subject.
Well,.
C
F
And
actually,
you
know
some
of
the
work
that
the
reasoning
behind
the
work
of
this
sort
of
the
decoupling
and
presenting
a
pluggable
interfaces
is
to
enable
organizations
like
nginx
to
to
be
able
to
leverage
the
the
console,
control
plane
and
we've
kind
of
spoken
with
with
nginx
and
H.
A
proxy
NHA
have
done
some
premium
to
sort
of
firm
exploratory
stuff
around
burden.
Yet
so
all
those
things
ongoing,
but
I'd
be
really
nice
to
to
be
able
to
have
that
flexibility
and
freedom.
I
think
you
know,
ultimately,
you
might
use
nginx
as
your
control.
F
As
far
as
your
data
plane
for
your
room
for
your
edge
in
Brett's,
be
you
might
this?
You
know
just
for
service
the
service.
You
want
something
just
super
lightweight
and
being
able
to
have
a
service
mesh
which
supports
that
as
a
pluggable
pluggable
trend
of
architectural
model,
certainly
for
large,
larger
enterprises.
I
think
is
gonna,
be
a
desirable
feature.
It.
C
Really
is
you
know,
I
wouldn't
well.
We're
gonna
have
some
fun
this
afternoon,
a
lot
to
talk
about
yeah
we
went
out,
and
so
Sacco
certainly
is.
You
know
intimately
familiar
with
engine
mesh.
That
was
a
contributor
there
and
then
just
having
gone
round
and
explaining
engine
mesh
as
as
nice
do
deployment.
Debt
was
leveraging
anyway,
is
basically
having
east
EO,
initially
being
described
as
sort
of
envoy
as
the
battery
included,
but
display
scible
with
you
know,
with
a
couple
of
other
choices,
not
not
that
many,
but
a
few
other
yeah
there's
a
cost.
C
This
is
so.
The
nice
thing
about
SMI
are
I,
expect
that
this
is
I'm,
putting
which
words
in
your
mouth
that
I
expect
maybe
you're
in
agreeance
with,
but
they
are
the
nice
thing
about
something
like
SMI.
Is
that
there's
a
cost
in
terms
of
the
overhead
of
trying
to
be
data
trying
to
be
agnostic
to
whatever?
That
thing
is,
and
in
this
case
is
the
data
plane
and
there's
caught?
C
You
measure
the
cost
of
various
ways,
but
hey
if
there's
a
significant
community
effort
around
eliminating
that
cost
to
rather
than
which
there
it
is
and
what
you
guys
are
participating
and
then
yeah.
You
would
expect
data
plane
vendors
if
you
will,
if
that's
a
thing
to
come
over,
and
you
know
readily
help
at
the
answer,
particularly
the
part
of
SMI,
that
I
think
just
understanding
very
young
initiative
and
effort,
and
that
of
the
limited
time
that
we've
spent
there
to
make
measuring
compatible.
I'm
I
understand
that
the
extensibility
bit
of
the
spec
is
ill-defined.
F
You
get
it
so
the
the
intention
behind
the
extensibility,
the
the
spank
laws,
to
to
kind
of
give
some
leeway
to
to
people
who
are
integrating
with
the
spec,
where
there
isn't.
Maybe
a
formal
specification
element
right
now.
The
intention
behind
the
the
kind
of
third
that
the
running
process
is
that
what
you
want
to
see
is
those
extendable
elements
actually
become
formal
specifications,
but
as
the
spec
rose
and
complexity
is
much
more
input
into
the
spec
like
with
all
specs.
It
takes
time
to
formalize
those
those
elements.
F
I'm
very
much
looking
forward
to
kinda,
seeing
where
the
community
sort
of
wanted
to
drive
s,
mind,
drive,
drive
it
to
because
I
think
the
reality
and
we're
starting
to
talk
about
it.
Now.
Anybody
who
is
around
Cooper
was
hearing
bits
and
pieces
around
a
sort
of
Federation.
It's
certainly
a
topic
which
is
becoming
more
and
more
prevalent.
I
think
is
larger.
F
F
Think
I
think
this
is
kind
of
this
or
this
to
to
predominant
use
cases
where,
where
a
situation
like
that
so
originates,
the
the
first
one
is
is
acquisition.
So
you
know
an
organization,
one
airs
another
organization,
you
don't
make
acquisition
decisions
based
on
the
cloud
or
the
technology
by
which
they're
operating
so,
but
ultimately
you
have
to
face
the
technological
challenge
that
there
needs
to
be
integration
at
some
point
between
the
the
two
systems
and
and
I
think
as
service
mesh.
Matures
we're
going
to
see
the.
F
Having
that
sort
of
giant
virtual
overlay
is
is
really
beneficial
to
everyone.
The
the
kind
of
the
the
second
pattern,
which
I
think
you
see
in
larger
organizations,
is
that
departments
potentially
have
freedom
of
choice
choose
their
cloud.
F
I
just
thought
of
the
third,
a
third
element
as
well,
which
I
think
is,
is
probably
less
prevalent,
but
that
there
are
certainly
people
who
are
looking
at
redundancy
across
cloud
and
wanting
to
leverage
multiple
clouds
to
give
them
selves
sort
of
extreme
high
availability.
I
mean
that's
a
total,
like
you
know
it's
a
real
edge,
but
it
does
exist
and,
as
I
said,
that
was
probably
a
fourth
point
as
well.
So
it's
two
to
four
narrow
zone.
F
If
you
kind
of
consider
operating
outside
of
a
type
of
miracle
or
outside
of
waxed-
and
you
want
a
new
business
in
somewhere
like
China,
where
cloud
is-
is
a
lot
more
controlled,
then
you
can't
just
create
a
federated
VPC
across
to
AWS
regions.
Like
you
know,
you're
gonna
be
operating
some
of
your
stack
in
probably
a
LaBarbera
or
one
of
the
other
sort
of
government-sanctioned
clouds
and
you're
going
to
be
operating.
F
C
And
taking
notes,
taking
that's
fantastic,
I
agree.
I
would
thank
you
for
that
Nick
to
get
us
back
to
the
agenda
quickly.
Actually,
a
couple
of
things
I
know,
I
said:
I
do
desire
and
I
think
the
in
Pavan
and
Sacco
and
gireesham
Santosh
speak
up
if
I'm
sending
something
you
guys
disagree
with,
but
that
you
desire
for
the
initial
interoperability
with
SMI
to
be
to
take
further
steps
with
respect
to
mastery.
C
There
been
a
couple
of
folks
that
have
suggested,
and
maybe
we
talked
about
this
I-
don't
recall,
but
then
and
actually
was
Prem
of
the
NSM
team.
It
was
one
and
then
Paul
Bauer
of
Microsoft
is
another
one,
but
was
suggesting
some
mastery
is
something
of
a
reference.
Implementation
of
SMI
and
and
I
may
be
from
an
adapter
perspectives
that
you
might
might
make
sense
your
respective
of
what
that
is,
but
just
ping
Jason
Hanson
to
ask
me:
when
are
the
what,
with
you
know
the
post
Kubek
on
community?
C
F
Not
them
not
as
yet
I've
been
meaning
to
chase
that
up
this
week,
and
so
local
and
a
person
from
Microsoft
is
gonna,
be
taking
over,
be
at
the
mic
from
a
sort
of
an
engineering
perspective.
Lucky
also
and
said,
like
Sina
things
like
that,
that
there
needs
to
be
some
on
one
of
the
last
calls
we
talked
about
the
need
to
fall,
lies
a
sort
of
governance
board
and
to
formulize,
like
you
know,
process
around
ensuring
that
people
can
make
make
suggestions
and
make
make
contributions
to
the
to
the
specification.
F
E
F
That's
gonna
we're
not
gonna
kick
off
in
in
in
a
more
sort
of
formal
state,
I
believe
sort
of
next
week,
I
just
needed
once
lucky,
it
sure
big
early.
When.
C
F
C
No,
but
you
know
no,
no
I
wasn't
no.
No!
What
all
the
other
things
he
said,
I
didn't
I
was
mostly
trying
to
make
a
joke
about
the
very
good.
Well,
so
one
of
the
individuals
that
wasn't
able
to
join
us
today,
CTO
sorry
CTO
of
octarine,
which
is
a
security
centric
surface
mesh.
They
were
had
abused.
A
looser
kou-kun,
calm,
I
had
them
on
stage
at
docker
Kong
for
a
panel
as
well.
I'm
gonna
give
his.
He
sends
his
regrets,
but
did
send
an
update.
C
C
That
is
the
notion
that
they
need
to
set
up
a
public
use
cluster,
that
just
those
that
want
to
use
Mestre
can
go
and
not
use
a
free
truck
I,
don't
know
what
what
they're
gonna
come
up
with,
but
just,
and
so
he
so
the
update
is
that
there
this
is
forcing
their
hand
on
setting
up
a
public
cluster
they've
got
a
functionally
complete.
Adapter
Girish,
you
don't
recall,
I,
don't
recall
I'm.
Having
pushed
the
first
draft
of
the
octarine
adapter.
Yet,
though
I.
C
C
So
maybe
I
think
you
probably
saw
this,
but
so
you
helped
create
and
in
Santosh
I
don't
know
if
you
guys
had
yeah
you
had
you
and
Santosh
gonna
tie
it
off
on
Mishra
CTL.
E
C
E
C
That's
great
I
know
you
did
some
documentation
updates
as
well,
but
so
those
are
looking
good.
It
did
any
I'm
sensitive
to
the
time
that
we
have
any
anything
else.
You
want
to
speak
to
for
the
week
up
coming
up.
C
E
C
E
E
E
E
B
Like
I
have
okay,
here's
the
thing
in
VirtualBox.
If
we
install
the
SEO
like
load
balancers
like
if
we
install
a
CEO
in
GCP
or
in
AWS
like
we
get
the
load
balancer
because
has
IP
addresses
were
there
but
better
than
VirtualBox.
We
don't
have
I
peered
as
an
automatical
sign,
so
it's
in
old
automatically
allocates
the
IP
address
in
watch
was
box
all
right,
that's
the
only
the
thing
is
see
the
difference.
B
C
E
G
E
D
Poe
on
I
think
the
issue
is
actually
what
to
get
started
with,
like
yeah
you're
right,
like
a
nose.
If
there
is
no
external
load
balancer
you
not
probably
high,
or
do
not
probably
see
an
external
IP
assigned
to
the
clusters,
but
as
long
as
the
ports
on
the
cluster
or
accessible
from
your
host,
you
should
be
able
to
use
the
IP
address
of
any
of
the
any
of
the
VirtualBox
nodes,
and
you
know
you
should
be
able
to
access
it
directly.
D
C
Very
good,
oh
wait!
We
don't
have
much
time
left
a
couple,
quick
things
it
since
the
Eric
and
Nick
are
here,
it
would
grease
you
wanted
to
give
a
run.
I,
don't
know
that
we've
in
fact
run
these
guys
through
the
constable,
adapter
and
kind
of
what
it
currently
does
and
maybe
an
really
kind
of
a
question
to
them,
and
so
what
they
think.
Maybe
it
should
do
next.
Oh.
C
D
You
know
we
one
thing:
I'm
missing
their
property
is
actually
a
link
to
the
console,
not
for
repo,
which
I
can
definitely
let
out
up
there,
so
that
you
know
Nick
and
every
can
actually
take
a
quick
look,
quick,
look
at
it.
D
So
what
we
have
is
pretty
functional
so
or
what,
like
you
know,
what
we
get
is
deployment
off
console
using
mesh
tree
on
a
communities
cluster
with
the
cycler
injector
enabled
or
deployed
is
what
we
have
like
you
know,
so
to
start
would
like
no,
it
doesn't
do
any
fancy
configurations,
but
what
we
have
is
we
have
us
a
llamo
that
is
generated,
I,
think
from
helm,
and
we
have
it.
You
know
kind
of.
We
have
measure
II
you
through
the
adapter.
Let
me
know
deploy
that
on
our
coordinators
cluster.
D
It
works,
I
mean
equals
cluster,
and
the
next
thing
we
did
was
that
one
of
the
things
we
were
trying
or
working
towards
yeah
you
know
for
coupon-
was
actually
as
close
to
possible
or
as
close
as
possible.
An
apples-to-apples
comparison,
so
we've
been
pretty
much
using
the
HTTP
been,
and
this
year
was
canonical
booking
for
app
quite
a
bit.
So
what
we
did
was
you
know:
I
took
that
you
know,
took
their
yeah
Milles
kind
of
looking
added
the
console
annotation.
D
So
those
were
needed-
and
you
know
I
kind
of
like
you
know,
have
made
it.
You
know
pre-configured
in
the
adapter,
so
that
if
you
don't
load
the
console
adapter,
you
will
be
seeing
all
these
three
options.
But
you
know
ask
you
first,
you
can
actually
deploy
or
remove
console
along
with
the
sidecar
injector
and
the
next
one
is
a
capability
to
actually
deploy
the
stos
canonical
booking
for
app
on
console
and
the
same
thing
for
HTTP
then.
So
we
support
both
deployment
and
the
removal.
D
So
so
this
is
kind
of
like
how
far
we
got
now.
This
did
help
our
team
to
actually
kind
of
run
performance
tests
and
not
sure
if
you
guys
were
actually
there
at
the
presentation
at
coupon.
So
some
of
the
results
we
presented
on
the
stage
were
actually
you
know
generated
while
using
the
console
adapter,
you
know,
I
mean
so
our
team
actually
used
this
adapter
to
deploy
the
sample
apps
and
with
intestine.
So
this
is
kind
of
where
we
are
like
I
said
it
was
like
probably
a
few
days
effort.
D
D
That's
SMI
compatible
and
you
know
buy
it
on
the
cluster,
so
in
a
similar
fashion,
I'm
not
super
up-to-date
on
SMI
knowledge.
I
did
know
that
as
somewhere
that
prefers
D
existed,
but
in
the
last
week
I've
been
like
kind
of
keeping
busy
with
some
other
stuff.
So
I'm
not
sure
if
there
is
a
MSO,
my
adapter
for
console
today,
but
if
there
is
one
I
think
it
would
be
nice
if
we
can
actually
add
support
for
that
in
my
free
to
you
know
to
do
deployment.
D
F
F
One
of
the
things
that
I
think
we're
happy
to
to
kind
of
help
them
and
sort
of
chip
in
as
well
around
on
is
the
the
Canada
1.5
release
of
console
introduces
the
easy
configurability
of
layer,
7
observability
and
in
reality
anybody
who's
using.
Let
us
mesh
in
in
in
production
is,
is
going
to
be
pretty
much
one
be
using
layer.
So,
if
you're
running
any
problem
of
reliability,
factor.
F
Getting
getting
that
sort
of
minor
configuration
change
as
well
just
to
make
the
more
sort
of
realistic
before
Rome
from
from
console,
because
it's
actually
configure
to
use
layer
for
not
just
pure
tcp,
be
the
other
thing
I
think,
which
is
interesting
and
I
apologize
all
I
need
to
dig
in,
but
is
is
security.
So
in
terms
of
thinking
about
the
request,
flow
and
and
the
sort
of
the
performance
for
the
request
flow.
F
Is
that
the
they
obviously
need
the
authentication
stamp,
but
the
service
measure
traffic
is,
is
just
m
TLS,
that's
gonna,
add
yeah
and
overhead,
but
m
TLS
is
a
pretty
pretty
known
so
in
pretty
quick
process
a
lot
of
the
sector.
Well,
a
lot
of
the
meshes
are
going
to
use
sort
of
secondary
authentication
authorization
as
well
for
the
service
mesh
traffic,
and
that's
often
done
as
a
as
an
out
of
bounds
so
step.
One
is
obviously
do
the
the
m
TLS
process.
F
The
second
step
is
that
the
property
is
going
to
say
well
here
receive
this
traffic
from
this
service
mesh
identity.
Is
it
actually
allowed
to
make
a
communication
and
then
implement
it'll,
make
an
outbound
call
to
the
control
plane
to
be
able
to
validate
that
and
the
different
implementations
on
that
validation
can
have
a
huge
bearing
on
the
performance
of
the
service
match.
F
D
Awesome,
thank
you.
That's
a
great
thing
apart
from
that
Oh,
so
that's
a
fantastic
start
and
you
actually
let
me
give
out
another
point
which
is
fantastic.
One.
Other
point
like
you
know
we
were
thinking
about
during
another
site.
Discussion,
like
you
know,
was
actually
maybe
a
capability
to
support
deployment
of
you
know
the
console
side.
Cars
on
mediums-
or
you
know
something
along
those
lines-
would
actually
be
kind
of
a
you
know.
D
F
And
that's
so
that's
possible
right
now.
As
long
as
the
traffic
is
outbound
from
kubernetes
I
mean
technically,
you
can
do
it
for
a
virtual
machine
inbound
in
to
kubernetes.
Realistically,
you
probably
wouldn't
because
of
the
complexities
with
network
communication
right
absolutely
figure
issue
yeah,
we
only
own
pke.
F
The
Gateway
will
forward
the
request,
where
that's
interesting
for
measure
as
well
is
the
Gateway,
because
it's
adding
a
hop
and
ok
we're.
You
know
we're
not
we're
doing
we're
going
to
be
doing
TCP
area
TLS,
transparent,
proxying,
it's
still
gonna
matter,
hot
and
there's
you
know
potential
performance
loss
there
as
well.
So
there's
something
interesting
there
or
I.
F
Think
and
ultimately,
as
I've
seen
earlier,
I
think
the
end
goal
is
that
most
service
matches
are
going
to
support
this
sort
of
gateway
model
and
realistically
I
might
be
connecting
console
connect
on
a
virtual
machine
to
sto
some.
You
know
some
really
interesting
sort
of
use
cases
for
future
imagery
as
well
it
down
that
route.
Gotcha.
C
Fantastic
yeah,
it
is
I,
think
I've
captured
a
number
of
those.
Those
ones
are
really
good,
didn't
actually
Nick
edification
for
the
community.
For
the
folks.
Here
too,
you
were
talking
about.
You
know,
console
in
the
near
term,
supporting
gateway
to
Gateway
or
sort
of
cluster
to
cluster
or
service
mesh
instance
to
service
mesh
instance,
connect
gateway,
gateway,
connectivity
between
them
and
NAT.
F
Right,
so
what
we
don't
want
to
do
I
mean
from
a
supply
so
from
a
security
perspective,
if
you
just
had
like
a
gateway,
so
you
you
need
to
understand
what
the
destination
is
is
going
to
be
or
what
the
intended
destination
is.
There's
there's
a
number
of
ways
of
doing
that.
The
most
common
way,
if
you
just
dealing
the
standard
non
encrypted
traffic,
is
look
at
the
host
header.
Be
the
problem
with
the
host
header
in
with
with
SSL,
is
that
my
header
is,
is
obviously
encrypted,
so
to
get
it
the
the
host
header.
F
F
So
what
we're
going
to
do
is
we're
going
to
use
the
the
headers
inside
of
the
TLS
protocol
and
we're
going
to
encode
the
destination
into
that,
so
the
gateways
just
gonna
peek
at
that,
and
then
it
can
just
literally
forward
on
the
roll
packets
to
the
destination
and
that's
bad.
That's
gonna
be
kind
of
kinda
neat
because
the
as
long
as
you've
got
like
a
federated
trust
across
multiple
service
matches,
the
the
gateways
aren't
a
weak
point.
They
you
know
Ito,
they
don't
need
to
decrypt
and
recon
trip
traffic
to
a
destination.
F
You'll
have
that
pure
pure,
transparent
forwarding
services
again,
wouldn't
even
need
to
be
aware
that
they're
getting
that
traffic
forwarded
to
to
another
cluster
or
another
entity.
As
far
as
they're
concerned,
they're
still
talking
localhost
the
the
difference
is
that
the
the
endpoints
which
are
configured
inside
of
the
data
plane
would
be
what
would
be
a
gateway
as
opposed
to
physical
service
address,
and
then
the
Gateway
would
say:
okay,
you
know
it
will.
It
will
then
do
the
the
final,
the
final
sort
of
routing
and
balancing
two
to
the
destination.
F
F
C
C
F
I
mean
we're
gonna
use
the
S
and
I
had
it.
So
the
part
of
this
sort
of
the
TLS,
the
TLS
spec,
but
been
able
to
do
that
because,
obviously,
that
that
sent
well,
we
won't
be
encrypting
those
that
might
be
something
that
we
look
at
little
later
date.
I
know
like,
but
the
yeah
the
destination
I
the
destination
service
name
would
be
encoded
into.
This
I
had
a
announce
block
against
the
Gateway,
the
ability
to
to
be
able
to
successfully
forward
it
without
having
to
decrypt
the
traffic.
C
D
I
can
sure
try
to
show
whatever
I
have
here.
Of
course,
I
can
I've
broken
some
pieces
of
that
because
you
know
I
was
doing
some
development,
but
I
do
have
something
opening
you
see.
13
quickly,
show
me
sprint
now,
while
while
I'm
doing
that
now
I,
you
know
yeah
I
had
this
section
within
which
we
created,
like
you,
know,
a
discussion,
social
items,
so
we
can
I
mean
people
can
actually
be
true.
It
and
I
can
feel
free
to
actually,
like
you
know,
kind
of
nominate
their
names
if
they're
interested
in
participating.
D
So
these
are
some
of
the
things
that
can
be
found.
Clan
or
mystery
I
mean
these
are
like
not
long-term
stuff,
but
these
are
sort
of
short-term
enhancements
to
my
tree.
So
if
the
folks
on
the
call
are
interested,
people
actually
put
your
name
right
next
to
it
and
again,
if
you
have
any
questions,
we
can
definitely
talk
about
it.
There.
D
D
So
so
this
is
actually
one
of
the
pieces
that
are
more
on
now,
which
is
moving
away
from
using
poor
planner
charts
Logano
your
mesh.
We
had
a
good
integration
with,
like
you
know,
so.
Here's
a
section,
for
example,
some
connected
to
Cortana,
using
this
interface
and
right
after
the
connection
it'll,
actually
pull
all
the
boards
and
for
everybody
you
know
it'll
it'll.
Actually
they
show
the
show
the
panels
and
from
here
you
can
actually
pick
and
choose
like
another
ones.
D
We
need-
or
you
can
just
select
everything,
for
example,
here
I'm
working
with
one
panel,
because
you
know
I'm
working
on
a
specific
use
case.
So
here
like
earlier
use,
be,
like
you
know,
all
the
iframes
embedded
here
on
the
page
but
know
we're
only
trying
to
move
away
from
it.
So
so
here,
I'm
I'm,
working
on
a
POC
with
charge,
is
fundamentally
like
extracting
all
the
metadata
from
Ravana
and
trying
to
you
know
all
the
data
and
present
it
as
a
charge.
Yes,
chart
eventually
will
actually
probably
like
not
go
directly
to
Prometheus.
Now.
D
This
has
two
benefits.
One
capability
to
actually,
you
know
kind
of
provide
a
responsive
interface,
because
when
iframes
come
into
the
game
you
know,
there's
always
a
mess.
So
so
this
will
actually
provide
a
responsive
interface.
That's
the
first
thing
or
that's
like
the
shortest
or
the
simplest
advantage.
The
other
one
is
that
we
are
going
to
actually
provide
capabilities
in
Missouri,
I,
guess
part
of
our
performance
section
to
actually
you
know
showcase.
You
know
the
host
or
the
permanently
the
right
word.
There
is
the
infrastructure
metrics
like,
for
example.
D
So
here
is
actually
the
performance
section.
So
right
now
look
we
have
capability
to
actually
a
lot.
You
know,
then
you
circle
runs
from
here.
You
can
see
at
the
bottom,
but
you
know
once
Cortana
is
integrator
like
you
know,
you
will
actually
start
seeing
the
right
charts,
so
this
is
actually
a
book
on
a
charge,
but
not
for
example,
but
that
one
another
page
was
not
so
so.
This
is
just
an
example.
D
So
when
a
person
is
running
a
lot
yes
like
if
they
can
see
the
metrics
from
Griffin,
but
all
these
have
been
looking
pretty
specific
there's
the
other
like
you
can
see
misty
own
gallery
dashboard.
What
I
mean
we
aren't
very
tightly
coupled
with
this.
Do
is
Cortana.
It
can
be
any
or
Connor,
but
it
so
so.
D
The
thought
here
is
to
actually
have
capability
to
also
present
the
server-side
metrics,
irrespective
of
Ravana
existing
or
not,
but
there
will
be
a
dependence
on
Prometheus,
so
that's
actually
the
eventual
direction
or
bringing
in
the
support
for
working
with
charges.
So
this
is
one
of
the
things
you
know
for
the
folks,
oh
I
can
who
haven't
seen
this
interface.
You
know
this
is
the
testing
interface.
D
We
have
like
twenty
seconds
so
I'm
just
going
to
showcase,
like
you
know
the
results
section,
so
the
results
section
is
actually
the
truss
every
time
the
load
test
is
run.
We
versus
the
results,
along
with
the
charts.
So
you
know
people
can
actually
compare
them,
we're
using
different
ways
of
comparison.
So
this
is
so
one
interface
like
again,
you
can
select
as
many
charts
you
want
and
let
you
know
use
it
for
comparison
so
and
its
results
agitated
et
cetera,
you
know
sorting
paging
searching
is
all
on
this
page.
D
The
other
one
is
actually
the
playground
interface
of
a
playground.
You
have
to
configure
it
with
an
adapter
I.
Don't
have
an
adapter
running
at
the
moment
in
the
next
sorry,
it's
already
passed
so
essentially,
like
you
know
this
interface
takes.
The
cube,
can
pick
one
right
after
that.
You
have
to
go
to
the
next
interface
and
connect
the
adapters.
So
this
one
is
already
listening
to
known
adapters.
You
can
connect
to
it
once
they
are
connected.
They
will
show
up
here
on
the
right
and
the
graph
on
I
integration
is
optional.
D
That's
the
reason
why
you
don't
see
a
starter
in
my
case,
I
already
have
an
integration
of
mana.
So
once
once
an
adapter
is
connected,
you
will
actually
see
this
page
populated
with
the
operations
supported
for
that
adapter
and
from
then
you
can
actually
interact
with
the
source
mesh
using
the
adapter
from
this
coin.
I,
given
the
time
I'm
actually
cutting
chart,
if
you
guys
want
to
hang
around
looking
for
a
bit
longer,
I
can
definitely
show
is
correcting.
It
I
mean.
C
I'll,
just
let
me
reiterate,
part
of
the
goal
is
sorry
well,
so
a
number
of
people
have
investments
in
bigger
fauna
or
a
number
of
the
service
meshes.
You
know
you
find
some
things
that
are
fairly
common
and
make
you
call
up
a
couple
earlier,
but
some
of
those
are
exposure
of
metrics
through
in
Prometheus,
literally
or
at
least
in
a
Prometheus
exposition
format,
and
so
and
so
to
do
you
find
many
of
the
mesh
is
supporting,
maybe
pre-configured
dashboards
in
Euro
fauna
or
a
number
of
people
that
are
running
these.
C
F
What
that
would
have
enable
you
to
do
is
if
you
had
specific
adapter
it
you,
you
would
interact
with
the
traffic
metrics
API.
The
traffic
metrics
API
is
going
to
be
the
abstractions,
the
underlying
metric
itself,
but
it
would
allow
you,
for
example,
if
somebody
wasn't
using
Prometheus,
they
might
be
using
in
flux,
DB
or
data
doll
or
or
like
some
vendors
cloud
metrics,
you
get
a
nice
clean
abstraction
with
with
massery
there
as
well.
Again,
it
was
just
a
just
an
idea.
It's
ya.
F
C
Well,
Nick
thanks
for
that,
it's
not
like
I
think
we've
gotten
more
suggestions
out
of
this
call
than
that.
Maybe
the
last
three
combined.
So
this
is
great
with
I
understand
we're
at
the
top
of
the
hour
to
Santosh
I,
understand
we
didn't
and
and
Pavan
didn't
get
to
spend
a
lot
of
time
on
things
that
you
guys
be
working
on
forthcoming.
C
If,
let's
do,
though,
find
someone
one
on
one
or
one
or
two
or
what-have-you
time
before
next
community
meeting
the
there
is
a
list
of
chemistry,
to-do
items
that
that's
laid
out
in
in
bullet
points.
One
of
those
is
that
Kimball
Kidman
commissioned
by
a
link
by
buoyant
to
produce
performance
s
and
spent.
You
know
good
good
friends
with
Kim
bulk
and
they
were
getting
some
compelling
insight
as
to
what
coordinated
omission
is
in
terms
of
WR
k2
and
how
it
has
a
load.
Balancer
is
a
car
I'm.
C
Sorry,
a
load
generator
is
accounting
for
some
fanciness
with
respect
to
measuring
latency
and
how
we're
using
for
IO
right
now,
but
maybe
maybe
we
should
partner
up
and
they
did
help
put
in
WR
k.
There's
some
things
to
be.
We
need
to
make
sure
we
have
a
mental
note
and
for
the
community
here
for
us
to
reach,
invite
them
to
the
call
look
at
either
supporting
fort
ILO
and
w
RK
2,
or
maybe
just
disclosing
4l
of
w
RK
to
getting
a
blog
post
up
about
coordinated
emission
and
what
that
concept
is.
C
But
the
the
other
reason
why
I'm
looking
over
these
bullet
points
is
to
say
when
we
met
with
Paul
Bowers
of
Microsoft
this
last
Friday,
his
ass
was
to
see
a
bit
more
of
a
measuring
roadmap
and
I.
Don't
think
he
in
part
wanted
to
influence
it,
which
is
great
because
he's
taking
my
Sheree
to
customers
and
try
and
help
them
evaluate
as
they're
going
through
evaluations
of
meshes
and
he's
helping
them
deploy.
C
The
wonderful
so
hopefully,
they'll
provide
a
great
set
of
feedback,
but
he
also
I
had
mentioned
to
him
that
just
just
how
much
time
I
have
spent
over
the
last
three
years
within
the
same
CF,
how
we're
about
to
go
off
and
establish
a
networking
or
we
reincarnate.
The
word
networking
working
group
into
the
new
cig
in
the
seems
yeah
and
within
there
may
push
measure
e
to
the
CF
sandbox,
which
I
think
made
him
feel
warm
and
fuzzy.
And
but
we
need
to
make
some
of
those
statements
publicly
and
so
on.
C
There
I
set
it
on
the
recording,
so
Paul
said
he
watches
some
of
these.
So
as
long
as
he
pays
attention
for
an
hour
and
six
minutes
till
her
to
begin
with
that,
we
were
slightly
over
guys
I,
just
thanks
thanks
so
much
for
spending
this
time.
Sakho
you
caught
the
where
we
were
I
was
kind
of
doing
this
on
stage
last
week.
You
know,
thank
you
very
much
shirts,
you
guys
so
with
that.
Maybe
is
anyone
have
anything
else?
That's
not!
Let's
wrap
up
and
we'll
go.
Have
some
barbecues
little
bit.
C
Really
do
I
am
it
may
Jewish
may
have
to
take
you
up
on
the
offer
I'm
going
to
I'm
gonna,
well
make
sure
that
that
both
Nick
and
Erica
barbecue
and
their
bellies
and
their
time
here
otherwise
we'll
have
to
get
a
rope
and
hang
them
out
back
so
I,
really
I
really
want
to
buy
I've
got
a
soap
to
riche.
Can
you
can
you
hang
out
with
Santos
for
a
little
bit
sounds
good
gentleman?
It
is.