►
From YouTube: Istio User Experience Working Group June 11, 2019
Description
Meeting of the Istio User Experience Working Group held June 11, 2019
A
So
hello,
every
woman
I
went
through
the
sto
control
changes
today
and
last
week,
for
the
one
you
know,
I
signed
up
for
is
t
of
control
testing.
The
way
that
I
normally
test
is
to
run
every
single
command,
sto
control
offers,
but
I
noticed
that,
although
they
mostly
work,
I
felt
some
defects
and
fix
them,
we
didn't
have
any
Docs
or
some
blog
articles
for
some
of
the
new
features
of
is
T
of
control.
Rum
was
surprised
to
see
experimental
dashboard
this
way
to
bring
up
key
ally
and
the
various
boards.
A
So
we
should
probably
try
to
target
some
kind
of
some
kind
of
documents
to
highlight
or
monitor
all
right,
I
mean
if
it
works,
then
I
don't
see
any
reason
why
it
can't
be
added
to
the
toxify.
We
definitely
want
to
add
to
the
exit
most
of
the
SEO
controls.
Commands
are
only
shown
in
terms
of
how
to
use
them
to
do
the
tasks
of
the
journeys
we
don't.
A
Sight
inside
the
habit
like
in
the
key
ally
tasks,
for
example.
Yes,
so,
instead
of
instead
of
all
that
stuff
about
you
should
port
forward,
you
can
just
say
is
Joe
control,
experimental
dashboard,
key
ally
and
it
will
figure
out
all
those
ports
for
you
and
just
do
it
so
should
we
should
I
make
a
work
item
to
to
show
that
way
or
do
we
like
not
using
this?
Do
control
just
in
the
old-fashioned
way,
I
mean
our
getting
started.
Guide
goes
through
the
effort
of
setting
up
the
path
for
the
users
for
a
co.
A
B
B
A
A
Probably
the
big
thing
that
we
should
talk
about,
but
don't
know
how
to
talk
about
is
SMI
and
we
should
talk
about
what
what,
if
you
think
is
TOS
should
be
sort
of
with
it
and
I
what
end.
So,
let's,
let's
talk
about
that
in
the
second
half
of
the
meeting
ROM,
do
you
want
to
share
your
screen
and
share
your
slides,
or
do
we
have
any
other
news
items
we
should
talk
about
for
next
week?.
A
You
guys
see
my
screen:
okay,
the
power
plane
perfectly
cool.
So
for
those
who
don't
know
me,
my
name
is
Ron
venom
I'm
one
of
the
product
managers
at
IBM,
work
on
the
IBM
cloud,
kubernetes
service
and
also
Sto,
and
what
I
did
today
is
a
I
wanted
to
run
through
a
list
of
kind
of
pain,
points
that
I've
seen
from
from
users
who
have
used
this
tio
as
part
of
their
evaluation,
POC
or
production
usage.
But
I
experience
has
been
mostly
around.
A
A
Who
evaluated
is
you
heard
adopted?
A
and
I've
also
worked
with
a
lot
of
teams
that
go
through
this,
not
this
journey.
So
again,
my
goal
is
to
kind
of
present
these.
These
pain
points
to
you
guys
and
everybody
else
and
then
see
if
we
can
use
those
to
leverage,
you
know
better
work
items
and
tackle
them
moving
forward.
I've
broken
part.
A
B
A
That
you're
not
adding
unnecessary
complexity.
You
know,
for
example,
things
like
a
production
usage
knob
into
the
Installer
instructions
for
getting
started
or
or
their
learn
phases.
It
helps
you
kind
of
delineate
your
content
and
also
understand
the
flow
of
update
of
the
user
and
making
sure
that
they're
transitioning
from
one
phase
to.
B
A
Where
is
various
logging
or
monitoring
sources,
and
once
that's
set
up,
they
kind
of
hand
it
over
to
their
to
their
dev
teams.
It's
it
kind
of
basically
somewhere
else
that
gets
permission
to
access
this
cluster
and
they're
responsible
for
for
all
tasks
related
to
their
application
lifecycle,
and
that
could
involve
you
know
any
type
of
descriptor
for
for
what
they
expect
out
of
a
service
mesh
and
that,
and
even
if
that
means
just
describing
their
applications
or
even
just
applying
the
SPO
policies
themselves.
So
in
that
case,
I
need
to.
We
need
to
understand.
A
Like
know
where
does
this
deal
fit
into
here
with?
What
is
our
target
persona
and
there's
going
to
be
different
journeys?
Purpose
ona,
the
administrator
just
might
be
the
the
first
half
they
install
and
in
configure
section
and
then
the
app
developer
persona
can
can
start
off
with
a
steal
on
top
of
an
existing
cluster.
A
Some
feedbacks
got
from
the
community,
mostly
Twitter
as
and
we're
seeing
that
a
lot
recently
is
that
this
teo
is
hard
to
use.
The
api
is
difficult
to
understand
the
configure
to
install
and
the
configuration
is
hard
and
that
there's
no
kind
of
like
a
happy
path,
easy
install
route,
but
you
know
for
whatever
the
reasons
that
this
is
what
we're
seeing
a
lot
recently.
So
did
my
best
kind
of
break
all
these
pain
points
down
into
those
phases
and
then
we'll
kind
of
walk
through
them
for
the
SPO
installation.
A
There
are
multiple,
install
methods,
there's
multiple
profiles,
there's
multiple
providers
and
I
think
our
doc
tries
to
satisfy
all
of
them
in
every
page
right,
for
example,
this
getting
started
page
before
even
downloading
this
do
or
the
command
to
download
the
HDL
files.
We
introduced
the
user
to
two
see
cluster.
We
introduced
them
to
mesh
expansion.
We
talked
about
these
multiple
profiles
and
there's
always
like
like
these.
A
We
don't
take
you
straight
to
other
to
a
value.
Add
like
here
launch
this
dashboard
and
you
can
see
this
happening.
You
know
things
like
that.
We
just
kind
of
drop
you
off
with
either
a
verify
command
or
counting
the
number
of
see
RDS
the
configuration
after
you've
installed
sto
there,
the
docs
for
you
to
accomplish
some
of
the
tasks
you
need
to
go
back
and
modify
basically
the
your
install,
and
to
do
that,
you
know
some
of
them
are
in
apps,
and
some
of
them
are
environment.
A
A
There
should
be
an
easier
way
for
you
to
enable
and
disable,
and
you
know
what
features
you
want
without
having
to
go
through
the
helm,
template
install
process
or
whatever
journey
you
took
to
get
to
your
current
SDO
cluster
might
agree.
One
percent
and
also
it
is
possible
that
the
user
who
installed
this
do
is
not
the
administrator
who
is
valid.
B
A
Send
it
there
has
to
be
a
good
way
to
let
more
than
one
person
share
in
the
task
of
administering
and
using
this
do
right,
and
some
of
these
we
might
be
solving
in
in
the
upcoming
releases,
whether
it's
with
it's
do
operator,
work
or
different
method,
but
I
think
from
a
user
experience
point
of
view.
We
still
need
to
define
this
as
a
user
requirement
and
whether
that
gets
implemented
by
another
team
that
that
doesn't
matter,
but
we
need
to
track
this
use
cases.
A
If
statements
where
we're
having
the
user
constantly
like
make
decisions
or
know,
details
of
there
is
still
installation
to
proceed,
there's
a
lot
of
examples
of
this.
But
if
you
just
you
know,
search
for
if
you're,
for
example,
or
look
for
the
number
of
times
we
use
tab
or
we
have
those.
Those
info
or
warning
boxes.
You'll
see
that
some
of
those
problems,
tasks
vs.
examples
that
definitely
causes
confusion
to
some
some
users.
A
I
think
this
is
something
we
all
know
it
for
whatever
the
reason
you
know
we
have
our
API
objects
named
this
way
and
it
causes
a
lot
of
confusion
with
our
users
and
at
every
conference
at
every
presentation,
I
see
the
presenter
describe
these
these
in
a
different
way
and
in
even
in
advanced
presentations.
They
feel
the
need
to
kind
of
reteach
the
user
some
of
these,
because
they
know
that
people
forget
exactly
you
know
what
they
do
and
where
they
fit
in,
but
but
I
think
they're
from
a
UX
point.
We
need
to.
A
We
need
to
drive
some
simplification
layer
on
top
of
this
and
I
know,
there's
multiple
ways
of
doing
it,
whether
we
add
another
layer
on
top
or
whether
we're
renaming
it
or
or
it
might
just
be
a
matter
of
better
documentation
but
regardless
I
think
this
is
a
pain
point
and
it's
not
user
friendly
from
an
application
developer.
Point
of
view
when,
when
they
get
through
the
task
right
with
book
info
and
they
think
they
they
learned
SEO
concepts
and
the
general
flow.
The
next
step
is
to
adopt
this
gear
for
their
own
application
right.
A
Do
we
have
a
ID
that
that
walks
them
through
the
best
practice
for
adopting,
is
here
for
their
own
applications?
What
this
should
they
do
for?
Should
they
enable
automatic
injection?
Should
they
should
do
it
per
app?
And
what
is
the
flow
and
I?
Don't
think
that
exists
or
if
it
exists,
people
are
not
finding
it
because
I
see
issues
all
the
time
where
they're
like
they
just
enable
automatic,
sidecar
and
then
restart
stuff
and
things
are
broken
and
they're.
A
Like
oh
crap
like
install
this
deal
now
I'm
a
broke
in
and
it's
just
that's
not
the
experience
we
kind
of
want
them
to
go
through.
It's
got
to
be
incremental,
and
if
it
is,
you
know
what
is
that
experience
so
that
guide
I
think
is
missing
so
for
users
that
have
adopted
it.
Some
of
the
pain
points
they
hit.
Is
you
know
they?
This
cryptic
error
messages
like
503
cyber
force
or
no
healthy
upstream
messages.
What
can
we
do
to
to
help
a
user
debug?
A
There's
just
got
to
be
something
we
can
do
to
help
with
that.
Most
of
those
gotchas
are
documented
in
this,
like
port
name
and
service
requirements
dock,
but
that
again
people
don't
people,
don't
read
that
and
people,
and
if
you
don't,
then
you
don't
know
that
you've
made
a
mistake.
That's
documented
in
this
documentation.
A
What
can
we
do
to
identify
that
a
user
might
be
making
this
mistake
and
point
them
to
the
dock?
Maybe,
and
maybe
have
that
and
the
logs,
or
have
done
as
part
of
this
do
TPL
to
point
them
to
these
requirements
so
that
they'll
go
read
and
implement
it.
A
lot
of
questions.
A
lot
of
people
ask
us
about.
A
You
know
if
they're
the
problems
in
other
application,
but
it's
the
configure
around
their
applications
that
they
now
have
to
migrate
to
SEO
and
things
like
you
know
if
they
already
have
like
engine
X
in
quests
and
they're,
using
in
genetics,
rules
and
annotations
etc.
Like
now,
they
know
that
they
have
to
move
over
to
sto.
Is
virtual
services
and
destination
volts?
What
is
the
migration
guide
for
that
and
there's
a
lot
of
activity
on
in
this
kit
repo
and
this
github
issue
that
talks
about
these
requirements
and
a
lot
of
responses.
B
A
The
community
enabling
security
in
my
experience
is
the
number
one
reason
the
number
one
use
case
for
for
customers
to
adopt
is
Tio's.
They
want,
you
know
pod
to
pod
encryption,
so
this
should
be
dirt
simple
to
implement,
and
currently
it's
not
I
think
it
is
error-prone,
and
it's
also
one
of
those
that
causes
those
common
503
issues
and
and
and
for
the
user.
There's
no
easy
way
to
deal
of
what
they
did
wrong
and
when
they
go
start
debugging,
and
this
is
where
I
think
the
the
complexity
is
exponentially
grows
right.
A
If
they
start
reading
envoy
logs,
it's
they're
very
noisy
when
you
apply
and
hiss
do,
configure
steal
rules,
there's
no
way
for
you
to
know
that
the
rules
were
successfully
applied
because
there's
no
like
describe
like,
like
you,
would
describe
a
kubernetes
resource
and
get
a
status
message
back
that
that
this
rule
is
is
incomplete.
Oh
because
it's
waiting
on
something
else,
for
example
a
subset
or
or
that
this
rule
is
invalid,
or
this
rule
it
was
broke
the
service
mesh
and
now
the
mesh
is
frozen.
A
So
things
like
if
my
pot,
a
can't
talk
to
pod
B,
is
there
just
a
co,
CTL
check
connection
a
B
and
then
have
it
do
some
of
these
checks
for
me,
and
you
know,
give
me
back
a
list
of
things
that
had
verified
and
found
a
potential
problem.
That's
just
a
solution,
but
think
of
things
like
that
or
I
think
what
people
would
kind
of
expect
when
they
get
these
type
of
issues.
A
So
those
are
all
the
pain
points
and
another
things
that
I
wanted
to
capture
is
a
desired
experience.
You
know
what
is
our
end
goal
for
someone
who's
adopting,
sto
and
kind
of
tried
to
put
based
on
you
know
talking
to
a
few
of
you
I
put
together
this
rough
draft
of
you
know
what
I
think
we
should
aim
for.
So,
as
a
user
who
wants
to
learn,
if
do
I
want
to
be
able
to
easily
install
sto
on
top
of
an
existing
cluster,
that
I
already
have
that
cluster
will
most
likely
have
applications.
A
Installing
is
T
on
that
cluster
should
break
nothing.
I
want
to
deploy
a
sample
that
comes
from
this.
Do
o
it
should
be
a
modern
good-looking
sample
that
represents
my
application.
Representin
represents
my
architecture
or,
as
close
to
it
as
possible,
once
I
deploy
the
sample
without
having
to
do
too
much.
A
I
want
some
confirmation
that,
if
t
o
is
working
and
I
can
get
some
value
at,
and
that
could
be
something
as
simple
as
launching
a
dashboard
like
key
ally
to
see
to
defeat
traffic
flow,
for
example,
so
deploy
a
sample,
hit
the
run
sample
and
then
see
the
various
services
and
get
that
service
graph.
That's
a
nice
confirmation.
A
A
Working
in
a
blessed
video
that
we
put
directly
on
this
deal
so
I
think
that's
kind
of
at
the
end
of
my
my
ramble,
so
I'll
stop
there
and
kind
of
gain
feedback
from
from
you
guys
and
see
what
you
think
about
about
converting
some
of
these
pain
points
into
epics.
What
you
think
about
this,
like
experience,
etc.
I,
don't
think
it
was
a
ramble
at
all.
I
think
you
organized
it
very
well
and
I
think
these
are
exactly
the
problems
that
we
need
to
tackle.
D
As
one
of
the
new
pre
people
in
the
sto
project,
I
can
identify
with
almost
every
single
one
of
these
yeah.
It
took
me
a
while
to
even
understand
exactly
what
what
is
this
deal.
Would
you
know
it
took
me
a
little
while
to
kind
of
wrap
my
mind
around
it
and,
like
you
said
some
of
the
some
of
the
examples
you
know
they
take
a
while
to
kind
of
to
work
in
and
I
like
this
idea
of
getting
doing
the
getting
started
tasks
in
less
than
two
minutes.
D
B
D
There's
also
one
thing
that
I
noticed
on
on
the
docker
website:
they
actually
have
a
why
docker
page,
and
so
that
might
be
something
that
we
could
look
why?
Why
do?
Why
do
we
need
something
like
yeah
I
mean?
Maybe
that
could
be.
You
know
something
where
you
can
kind
of
explain.
The
point
of
having
is
anything
like
is
dia,
because
again
at
first
I
had
to
just
understand
what
what
it
was
before.
D
A
A
Nekron,
you
and
I
have
spoken
separately,
I
think
there's
another
persona
of
the
sre
sort
of
persona
who
is
using
sto
to
keep
this
mesh
up
and
to
troubleshoot
wrong
when
his
work
fails.
So
that
is
a
section
that
you
or
someone
else
should
do
more
on
to
come
up
with
what
sort
of
rules
that
persona
has
and
to
make
sure
the
documentation
is
tailored
towards
towards
that
user.
Right,
cuz,
that's
up
and
that's
re
persona
would
would
kind
of
come
in
after
they've
adopted,
sto
right,
like
once.
A
They
have
it
running
in
production,
for
example,
and
then
I
would
think
that
that's
where
there's
re
persona
and
their
task
would
would
would
really
shine.
I
agree.
Okay,
see
we
talked
a
lot
about
diagrams
and
we
have
many
and
but
they're,
far
away
from
the
Yambol
one.
My
biggest
frustration
is
that
the
yeah
MO
and
the
diagram
that
goes
together
are
kept
at
the
opposite
ends
of
the
concept
page.
A
But
I
was
wondering
about
those
diagrams
if
people
are
going
to
be
being
pointed
to
key
ally
or
some
of
these
other
tools
to
visualize
their
mesh.
Should
we
consider
dumping
are
pretty
handcrafted
diagrams
and
using
something
like
key
ally
having
the
user
see
one
set
of
diagrams,
and
then
we
told
that
they
should
use
key
ally
means,
like
a
translation
of
visual
metaphor,
do
we?
What
do
we
think
about
trying
to
recreate
the
diagrams
that
we
use
to
explain
our
sto
rules
with
key
opping
or
with
whatever
tool,
recommending
people
to
use
in
production?
A
C
Kind
of
interesting
it,
but
by
the
way
this
is
Dan,
I
had
a
dial
and
I
couldn't
get
audio
to
work
in
my
computer,
you
use
so
our
diagram,
some
of
which
are
conceptual
until
which
illustrate
the
hgo
components.
Edie.
Are
you
basically
saying
redo
those
with
a
styling
that
matches
that
which
they
will
see
in
key
Ollie
I.
A
If
you
think
the
diagrams
are
difficult
to
understand,
I
I
know
that
I
thought
the
diagrams
that
we
created
were
fine.
I,
just
think
we
need
more
of
them.
They're
beautiful
I
think
they
need
to
be
near
the
yellow,
so
I'm
agreeing
I'll,
look
at
them,
I'll
scroll
down
to
the
mo,
but
don't
be
like
which
diagram
was
this
I'll
have
to
go
back
so
yeah.
We
can
reorganize
them,
I.
Think
they're,
good,
pretty
diagrams
I
have
no
trouble
understanding
our
diagrams.
B
Hi,
this
is
Len
I.
Think
next,
step
probably
would
be
categorized
some
of
these
issues,
because
I
think
a
lot
of
them
would
go
to
the
documentation,
workgroup
and
then
some
of
these
we
should
looking
into
which
will
be
able
to
help
fix
some
of
these.
Oh,
maybe
some
of
the
old
work
item
already
being
addressed
in
different
workgroup
I,
think
that
will
be
a
good
exercise
to
actually
turn
these.
C
B
A
C
Right,
that's
what
I
wanted
to
get
at
thanks
for
bringing
that
up,
Lynn
and
and
run
by
the
way.
I
said
it's
rum
earlier
we
spoke
last
week.
I
think
this
is
fantastic
work
that
very
much
needed
to
be
done
for
a
long
time.
So
thank
you,
yeah,
so
the
environment.
So
a
lot
of
stuff
here
is
installation,
I,
think
that's
the
environments.
Working
group
I
think
we
should
get
ROM
on
the
agenda
there
to
do
this
presentation
and
then
yeah
for
documentation
and
that's
that's.
B
A
A
This
is
part
of
the
the
table.
Fear
cube,
comes
started,
IBM,
hosted
an
event
to
teach
people
how
to
use
this
deal.
When
three
people
showed
up,
you
can
see
this
giant
room.
These
people
spent
over
three
hours
going
to
the
guest
book
demo
trying
to
do
it.
You
can
see
ROM
way
back
here
and
Linda
way
back
here
to
the
idea
for
how
big
this
room
was
when
rum
is
telling
you
about
the
problems
he
sees
people
having
at
these
events.
He's
he's
doesn't
present
for
real
on
this.
A
That
evening
we
had
a
meet-up
organized
by
dennis
truly
something's
call
here
is
me
talking
about
the
performance
of
envoy
and
the
most
interesting
talk
from
me
was
not
any
of
the
deep
technical
talks
about
how
a
steel
works,
but
this
time
by
a
local
telecom
user,
Pablo,
Mankato
and
I
liked
I
thought
previous
slides
were
really
good.
Everyone
should
look
at
his
whole
presentation.
Let
me
pull
up
his
his
slide
deck,
which
I
thought
was
really
interesting.
A
So
this
is
a
lessons
running
SQL
in
production
in
Barcelona.
This
is
a
local
Barcelona
guy
Pablo
Montoya.
He
had
some
very
nice
diagrams
that
teach
sort
of
hollow
sto
works,
and
while
you
should
use
a
mesh
that
we
should,
we
should
be
using
to
teach
people
about
the
mesh,
but
then
at
the
end
he
talked
about
his
sort
of
pain,
points
and
I.
Think
that
would
be
worth
addressing
those
in
detail.
The
first
one
is:
do
is
a
break
by
bad
config.
A
You
see
what
happens.
You'll
see
proxies
going
stale
because
two
rules
conflict.
This
is
what
means
when
he
says
that
the
mesh
seems
to
freeze
and
Pablo
solution.
Is
that
he's
actually
using
helm,
two
templates
that
he
writes
themselves
to
right?
Sto
rules,
so
you'll
generate
a
steel
rules
automatically
to
prevent
all
of
the
errors
and
I
actually
had
been
working
on
something
similar
and
have
a
PR
that
gives
a
helm
like
function
for
this
with
error,
checking
that's
not
in
helm,
helm,
let
you
do
strings
and
it's
probably
enough
for
most
people.
A
My
peer
takes
it
a
little
further
giving
validation
to
the
strings
like
it's
really
a
service
game
and
stuff
I'm.
Looking
for
some
people
to
collaborate
on
taking
that
PR
further
and
getting
emerged,
but
he's
telling
people
that
they
should
be
running
proxy
status,
which
we
nose.
Although
level
debugging
command,
he
has
a
temple
to
increase
the
proxy
verbosity
and
all
this
debugging
stuff,
and
that's
just
because
he
had
so
many
problems
with
conflicting
rules.
Making
his
pots
go.
Stale
I
think
we
should
really
address
that
I
I've
thought
about
that
myself.
A
I
have
some
what
Rob
said
about
having
a
respect
status
field
and
the
rules
and
leg
cube
control
describe
work
in
this.
You
know,
I
think,
would
be
a
good
start
and
letting
rules
sort
of
go
in
provisionally
and
until
the
envoys
have
all
accepted
them.
I
think
would
be
another
helpful
thing.
Anybody
wants
to
talk
about
that
meet
with
me
on
proposal.
For
that
talk
to
me
health
line.
A
He
is
heavy
use
of
this
you're
fond
of
feature
that
I
had
not
used
before.
If
envoys
reject
the
rules,
because
they're
inconsistent
you'll
see
these
these
things
happen
here.
These
things
can
happen
for
all
kinds
of
reasons.
Some,
like
it's
discovery
service,
not
having
a
secret
or
whose
conflicting
rules
say
he's
right
away.
A
One
in
capability
should
all
be
using
this
feature,
which
suggests
were
not
sort
of
doing
a
good
job
and
having
pilot
percent
good
data
to
envoy,
or
maybe
we
need
to
put
pairs
in
the
Envoy
to
relax
things
and
in
some
cases
his
top
experiences
were
about
gateway,
port
name
uniqueness.
This
naming
convention
that
ROM
also
mentioned
service
entries
on
TCP
ports
and.
A
A
Yes,
so
these
races
that
he
already
opened
right,
some
of
them
are
opened
by
him.
Some
of
them
are
just
his
complaints
and
we,
but
he
didn't
open
his
issues.
Those
are
some.
Those
are
4-digit
issues
that
are
long-standing,
okay,
okay,
so
for
me
the
biggest
takeaway
from
cube
con
was
this
SMI
announcement.
Microsoft
has
their
own
service
mesh
I'm,
going
to
call
it
an
API
without
an
implementation.
I
recommend
that
everyone
sort
of
look
at
the
announcement
and
watch
this
eight-minute
video
I
gave
Monroy
to
sort
of
see
it
as
a
screenshot.
A
Cord
I'll
show
you
in
a
moment,
so
they,
instead
of
using
you,
know,
resources
specific
to
your
mesh.
They
want
to
have
a
common
one
and
they
have
all
these
partners.
I.
Think
that
it's
a
good
idea
in
terms
of
things
that
build
on
top
of
the
mesh
like
analytics
pieces
I
think
that's
going
to
be
great
for
customers,
I'm
nervous
about
sort
of
how
this
is
going
to
play
for
places
where
is
geo,
does
something
that
other
things
don't
lik,
say:
fault,
injection
or
different
retrying
stuff.
A
It's
going
to
be
very
tricky
to
convince
people
to
use
our
stuff
when
there's
this
subset.
That's
in
this
different
language
that
they're
used
to
so
I'm,
hoping
that
the
TIA,
so
you
can
come
up
with
a
better
response
on
what
I
just
said
about
what
we
should,
how
the
hell
we
should
deal
with
this,
so
the
documentation
for
it
is
at
this
site.
Smi
sky.
Oh,
when
I
look
two
weeks
ago,
it
didn't
seem
fleshed
out.
I
listened
to
the
criminais
podcast
where
eat
it.
A
Levine
talked
about
super
glue
and
how
they're
going
to
switch
over
and
use
these
api's
and
that's
worth
listening
to
for
everyone
and
the
whole
benefit.
The
Kerberos
podcast
is
great.
This
is
especially
good
episode,
so
here's
a
screenshot
from
the
video
that
Microsoft
showed
a
Q
Khan
I
have
a
zoom
in
on
this
later
on,
but
they
showed
three
screens
like
this
one
for
sto,
showing
a
Canarian
waited
traffic
balancing
one
with
linker,
D
and
telemetry,
and
one
with
console.
A
They
showed
her
the
same
system
working
so
so
that's
nice
here
is
their
actual
sort
of
JSON
that
they
used
to
control
the
system.
I
just
typed
the
sentence
from
the
screenshots.
It
was
too
hard
to
see
so
they're
sort
of
saying
they're
proposing
this
new
object,
called
SMI,
spec
traffic
split
canary
with
a
weight
and
services
and
they're
using
Flagler
to
implement
it,
and
this
ran
using
Sto
as
well,
so
slider
was
sort
of
manipulating
it.
Sto
is
sort
of
making
happen
so
anyway,
it's
worth
checking
out.
C
Depending
on
I
didn't
want
to
interject
one
thing:
I,
don't
know
exactly
what
it
means
when
it
says
typically
was
going
to
rebase
on
these
I
have
heard
them
communicating
privately,
that
they
basically
wrote
the
SMI
spec
and
gave
it
to
Microsoft
I.
Don't
think
that
github
backs
that
up,
but
she
made
it
sound
like
these
were
our
API.
It's
like
they
are
the
Superboy
guy
I,
don't
know
if
that's
true
or
not
/.
If
there
are
something
changes
that
were
made.
This.
A
C
A
B
By
the
way
lotto
the
last
weekend-
I
don't
know.
If
you
guys
see,
there
was
a
treatise
storm
that
talk
about
SMI
between
Tim
Hawkins
and
Brandon.
I
forgot
his
license
for
Michael.
Sorry,
he
was
one
of
the
Cuban
etiqa
founders
and
also
our
Christian
poster
Zac.
Petra
was
also
chiming
in
Christian
post,
actually
summarized
the
discussion
as
as
on
his
blog.
So
there
was
people
talking
about
the
best
comment:
demand
demand
LCD,
the
least
common
Demonica
API
for
the
podium
for
the
new
API
and
how
it
is
going
to
be
helpful.
B
C
Well,
yeah
I
am
not
convinced
that
we
are
dealing
with
honest
actors
here.
I
think
that
the
criticisms
you
know
Google
does
have
more
PR
than
anyone
in
in
is
do
but
Google
had
this
had
a
seminal
similar
number
it's
about.
Two-Thirds
at
this
stage
in
kubernetes
lifetime
as
well
and
and
when
Brendan
says
things
like
SMI
is
much
more
like
kubernetes
than
is.
Do
he's
really
saying,
because
this
geo
is,
is
not
a
multi
cloud
day.
He's
you
know
he's
saying
SEO
is
is
for
a
particular
he's
essentially
saying
it's
particular
about
a
cloud.
C
It
is
a
almost
as
if
it
is
not
open
source
and
it's
it's
completely
disingenuous
and
I.
Don't
know
that
it's
worth
having
a
debate,
our
marketing
team,
when
I
did
talk
to
them
their
feeling
about
SMI
was
don't.
There's
no
reason
to
talk
about
it.
I'm
not
convinced,
that's,
correct,
yeah,
but
I,
don't
know
what
the
ideal
strategy
is.
C
C
So
so
they
say:
I'm
not
actually
sure
that's,
correct,
I'm,
not
actually
sure,
that's,
correct
and
and
I
think
that
had
they
had,
they
really
wanted
to
get
a
an
industry-wide
effort
at
this.
They
would
have
involved
us
from
the
beginning
right,
I,
don't
think
you
were
involved
in
the
API
design
where
you
are
offered.
Were
you
late
well,.
B
C
That's
a
very
same
day,
I
think
they
did
the
same
for
us
for
the
announcement
they
notified.
Someone
here
that
they
announced
was
happening.
That's
very
different
than
saying
hey,
as
an
industry
we'd
like
to
get
together
and
so
what
they
did
was
they
get
everybody
who's,
not
IDM
and
Google
and
say:
hey,
let's
get
together
and
do
this.
That
in
itself
is,
is
pretty
interesting.
It's
an
interesting
sure.
Yes,.
C
C
I,
don't
know
that
I
think
the
SMI
kind
of
takes
for
granted
that
people
will
want
to
run
different
control
planes,
different
places
and
and
therefore
they,
a
common
API
among
them.
I
think
that's
pretty
suspect
and,
and
that's
that
the
the
biggest
per
the
biggest
issue
I
have
is
that
it
seems
to
assume
that
everyone
will
want
to
run
a
different
control
plane
everywhere.
C
B
E
B
C
B
C
Because
it's
because
right
now,
there's
only
one
that
could
be
right
and
I
think
that
they
could
live,
that
the
other
vendors
who
are
in
there
probably
aren't
thinking
we're
going
to
be
the
one
and
so
they're
just
trying
to
prevent
there
from
being
a
one
and,
in
reality,
I
wish
that
these
I
wish
the
companies
would
find
better
ways.
It
is
if
people
are
trying
to
differentiate
on
the
wrong
things.
I
think
yes,.
C
B
They
kind
of
believe
is
still
is
the
winner.
It
just
says
they
want
to
I
mean
they're,
not
solely,
is
still
camp
and
they
want
to
support
everything
out
there.
So
then
that's
interesting
to
build
something
on
top
of
all
that,
but
they
did
say
you
know.
If
the
winner
is
defined,
you
know
they
would
just
move
some
of
their
stuff
to
be
based
on
is
still.
A
Okay,
buddy:
that's
this
dragon!
You
could
move
on
from
it!
Somebody,
although
I,
don't
have
much
much
more.
These
are
sort
of
the
other
sto
talks.
I
saw
at
the
cube
con.
The
other
talks
that
I
thought
were
be
interesting
or
especially
interesting
to
me
was
the
linker
D
talk,
which
was
very
deep
dive.
The
webassembly
talked
because
that
might
be
coming
to
Envoy,
and
also
this
native
experience
of
CR
DS,
which
did
not
have
the
cupids,
will
explain.
A
A
E
B
A
A
A
B
A
A
If
we
can
identify
a
couple
dozen
apps
that
we
think
I
just
want
to
use
that
maybe
1/2
work
with
this,
do
you
know
and
have
don't
we
could
see
if
that
number,
the
things
that
just
work
goes
up
and
down
I
think
that
would
be
sort
of
good.
So
that's
something
that
I'm
doing
sort
of
on
the
side
of
the
UX
work
to
measure
improvements.
C
A
Like
to
see
it
yet
I'd
like
to
see
a
survey
for
demand
and
I'd
like
when
something
doesn't
work,
not
just
saying
don't
work,
but
to
sort
of
say
why?
So
is
it
not
working?
Because
it
expects
networking
to
be
up?
Well,
we
have
initiative
that
is
it
not
working,
because
we
expect
court
naming
well,
we
have
a
shoe
for
that.
Is
it
not
working
because
it's
a
stateful
set
and
it's
not
listening
right
or
it's
doing,
direct
polities?
A
C
So
the
for
the
for
the
finding
out
the
kind
of
demand
I
think
we
could
do
it
survey
and
then
maybe
based
on
the
results
of
the
survey
and
finding
out
what
which
workloads
are
important
to
people.
Then
we
could
see
if
we
could
get
volunteers,
perhaps
to
kind
of
give
provide
some
details
with
each
about
what's
wrong.
We.
A
C
Yeah
I
mean
it's
certainly
part
of
it.
If
we,
if
we're
able
to
improve
the
lead,
getting
started
experience
and
the
ability
of
SEO
to
tell
you
hey,
there's
a
problem
with
your
with
your
llamo,
because
you
don't
have
a
port
name,
that's
right
or
even
better,
just
behave
properly.
Despite
that,
it
might
fix
some
of
those
things
right
right,
so
so
we're
working
on.
A
E
A
B
Yeah
I
think
that's
great
to
have
it
in
its
ETL
as
the
additional
check
just
so
it
was
just
so
you
know
the
network
team
is
working,
a
particle
sleeping
well,
they
would
try
to
detect
a
particle
based
on
the
traffic,
but
then,
on
top
of
that,
it's
going
to
be
always
also
a
support.
What,
if
you
want
performance
optimization?
Well,
you
don't
want
the
particle
sniffing.
What
rather
you'd?
B
A
Totally
true
and
earlier
on,
I
was
thinking.
We
needed
more
done.
How's
it
going
earlier
stuff
I
was
thinking
we
needed
stuff
to
validate
stone
stuff
like
sto
vet
does
to
give
people
suggestions
what
I
realized.
What
was
that?
It's
really
hard
to
rules
that
simulate?
What
the
is
going
to
do
and
pilot
keeps
improving.
Things
are
less
necessary,
so
I.
What
I'm
looking
for
are
ways
that
without
rules
we
can
figure
out
the
problem
like
if,
if
envoy
is
rejecting
configuration
from
pilot,
can
we
improve
that?
A
A
A
Hopefully,
I'll
be
able
to
present
some
preliminary
results
on
what
works
and
what
doesn't
and
I'll
be
able
to
talk
about
things
we
created
around
Rams
slide
deck,
yeah,
I.
Think
by
that
point
we
should
have
a
list
of
epochs
and
there
are
sub
child
tasks
and
have
some
sort
of
prioritization.
So
we
should
be
able
to
be
starting
assigning
them
to
people
I'd
like
to
start
making
epics
for
your
bullets
even
before
you
show
this
at
the
environments
group,
but
I'll
start
with
the
right.
B
A
A
Okay,
well
thanks:
everyone
hey
before
link.
Is
there
a
link
to
this
cube?
Con
EU
highlights,
did
I
miss
that
I,
don't
think
I
gave
it
I
think
it's
on
my
personal
thing
and
I
camera.
If
I
show
you
this
link
or
not
what
I
see
no
reason,
people
can
can't
just
view
that
I'll
put
that
in
the
you
egg
slack
Channel
how's
that
sound
okay.