►
From YouTube: Kubernetes WG IoT Edge 20191106
Description
November 6 2019 meeting of the Kubernetes IoT Edge Working Group
A
And
we're
recorded
and
started.
I
don't
have
the
agenda
doc,
open
dion,
do
you
or
I
don't
know
if
anybody
added
anything
to
that
or
if
we're
going
to
go
free
for
them
today,.
B
A
Okay,
so
we
were
just
talking
about
content
for
the
deck
we're
kind
of
con.
We
already
gave
the
cncf
our
proposal,
so
we've
got,
you
know
the
abstract,
which
has
an
agenda
and
we're.
We
were
just
talking
about
how
we'd
address
that
agenda
in
a
slide
deck.
We
also
talked
about
from
my
perspective.
It
went
well
that
when
we
did
this
in
europe
we
had
an
intermission
because
we
have
a
back
to
back
intro
and
deep
dive.
A
So
if
the
session
is
going
to
go
close
to
two
hours
and
last
time
we
asked
the
audience,
we
told
them,
we
were
going
to
take
a
five-minute
break
and
if
they
didn't
have
to
leave
the
room
to
go,
use
the
restroom
or
something
that
we
earmarked
the
corner
of
the
room
for
different
use
cases
for
iot
edge.
So
people
could
meet
others
who
have
the
same
interest
they
did
and
they
seemed
to
get
heavily
into
that.
A
I
mean
we
had
to
force
them
to
go
back
to
their
seats,
to
restart
the
meeting,
and
a
few
of
them
told
me
afterward
that
they
really
liked
meeting
the
birds
of
a
feather
with
their
use
cases.
So
I
think
I'd
like
to
do
that
again
and
the
other
thing
you
inherently
get
out
of
it
is
you
sort
of
have
an
inherent
poll
where
we
can
see
how
many
people
associated
with
the
group
are
interested
in
particular
niches
and
that's
good
to
know.
If
we're
trying
to
get
content
for
future
meetings
and
future
cube
cons.
B
A
Okay
and
what
we
might
want
to
do,
if
we
have
enough,
you
know,
people
who
are
kind
of
consistent
attendees
is
even
after
we
divide,
have
people
sit
in
the
corners
and
kind
of
facilitate
intros
and
things,
but
because
I
I
would
have
liked
to
get
a
report
back
on
what
the
conversations
were.
I
know
we
had
a
big
group
of
telco
people,
small
group
of
gaming
people,
I
think
yeah,
and
we
have
the
loyalty.
B
B
That
made
it
difficult
to
have.
There
were
a
lot
of
after
after
session
conversations.
I
think
one
of
the
things
that
made
it
difficult
to
sustain
them
was
that
our
our
place
in
the
lineup
was
toward
the
close
of
the
conference.
If
I
remember
correctly,
and
so
a
lot
of
people
were
kind
of
starting
to
head
out.
B
Oh
okay,
all
right!
Well
maybe
we
should
let
people
know
ahead
of
time
like
at
the
start
of
the
session
that
that
you
know
we
will
be
available
afterward
for,
for
you
know,
vertically
oriented
breakouts
so
that
we
could
continue
the
birds
of
a
feather
that
emerged
during
the
session
yeah.
A
A
Oh
yeah,
we've
got
the
last.
We've
got
the
last
day
in
the
afternoon,
okay,
again,
all
right.
So
if
anybody
is
tempted
to
leave
early,
they
might
leave
before
we
get
to
ours,
and
I
haven't
looked
for
over
a
week,
but
right
now
the
last
I
looked.
If
you're
a
speaker,
you
can
actually
get
the
number
of
people
who
indicated
that
they're
interested
by
making
their
schedule
it's
a
little
early
for
that,
because
a
lot
of
people
don't
put
their
desired
session
list
together
till
the
week
of
the
event.
A
A
B
Well,
we're
just
gonna
have
to
make
some
noise
about
the
the
session,
then,
in
order
to
get
people
to
to
stick
around
and
come
in
for
this
one.
A
C
A
A
There
was
actually
one
I
thought
of
somebody
talking.
It
sounded
really
interesting
a
session
on
how
to
put
together
your
own
private,
lte5
network,
and
I
gathered
that
this
is
some
idea
where
you
just
get
the
commercial
off
the
self
chips
and
things
and
you
can
manage
to
become
your
own
mobile
telco,
which
sounded
really
fast.
A
Yeah
I'll
I'll
do
that
by
the
end
of
the
week.
Right
now,
I'm
still
editing
and
I've
had
kind
of
intermittent
bandwidth,
but
because
I'm
moving
around
at
this
conference-
and
you
know
what
conference
wi-fi
is
yeah
yeah.
C
A
Always
put
my
draft
there
just
to
have
a
shell,
so
you
tech
you,
you
met
the
requirements
of
meeting
the
deadline,
but
I
mean
I
know
in
europe.
I
think
we,
our
last
update,
was
posted
an
hour
before
the
stage
so
yeah.
I
expect
the
same
here
right
yeah
so
that
you
know
we'll
we'll
definitely
get
enough
that
people
will
have
some
idea
of
and
the
cncf.
A
I
think
the
cncf's
worry
is
that
some
speaker
completely
forgot
about
it
and
maybe
attending
so
they
get.
They
get
some
confidence
that
sometimes
they're
investing
in
something
on
the
air
and
they
won't
have
like
a
no-show
speaker.
B
So,
on
the
topic
of
the
kubecon
north
america
activities,
we
had
in
the
slack
for
this
working
group,
a
post
from
jeff
young
who
was
organizing
an
an
out
of
conference
event
to
get
together,
and
I
just
put
in
the
slack
or
sorry
in
the
the
zoom
chat,
a
link
to
the
the
sign
up
spreadsheet,
which
presently
has
three
signups.
That
would
be
me
and
jeff
himself
and
deon.
B
So
I
think
that
if
there
is
interest
in
in
getting
together,
because
we
often
try
to
do
a
face
to
face
right-
and
we
don't
currently
have
any
face
to
face
organized
other
than
this
one,
so
since
those
usually
go
well,
then
we
like
to
do
them
I'd
like
to
encourage
anybody
who
has
not
yet
put
their
name
on
the
list
to
to
put
your
name
on
the
list.
So
we
can
get
some
gravity
around
this
to
use
the
steve's
word
of
the
day
whose
gravity.
A
Okay,
I
just
added
mine
to
the
list.
My
only
caveat
is
that
I'm
already
obligated
for
some
other
face-to-faces,
so
I'm
I
am
putting
together
a
face-to-face
for
vmware
users,
people
who
run
kubernetes
on
on
top
of
vmware
infrastructure
and
I'm
planning
on
joining
the
storage
cigs,
whatever
meeting
they
might
have
after
hours,
just
because
I've
I
worked
in
that
group
for
years
and
it
almost
evolved
to
like
a
you
feel
like
it's
a
friendship
so
just
want
to
see
the
people
again.
Even
though
I
haven't
touched
the
storage
area
for.
B
A
year
or
more
now,
you're
a
popular
guy.
Well
we'll
try,
maybe
we'll
try
to
organize
this
face-to-face
around.
You
know
a
time
that
you
can
attend.
I
think
there's
a
lot.
A
A
I
think
they
have
a
pretty
long
lunch
period
too
and
I'm
not
sure
even
if
they
do
feed
us,
sometimes
food
provided
by
linux
foundation.
Events
could
be
improved
upon
and
maybe
we
can
just
go
meet
over
launch
because
nobody
should.
A
Yeah,
in
fact,
in
barcelona,
I
think
a
lot
of
the
group
went
to
that
huawei
sponsored
cube
edge
lunch,
which
was
actually
a
really
good
event.
B
It
was,
it
was,
and
I
think
that
that
was
kind
of
the
place
took
the
place
of
the
face-to-face.
Well,
we
went
from
the
face-to-face
to
the
the
sponsored
lunch
right,
so
yeah.
So,
let's,
let's
I
like
dan.
I
like
that
idea.
Let's,
let's
pick
a
place
and
say
we're
going
to
go
and
then
that'll
be
the
way
that
we
do
this
face-to-face
and
that
way
people
already
were
going
to
break
for
lunch
anyway.
A
Yeah,
it
sounds
good
and
I'll
I'll
have
to
check,
but
if
we
only
have
four
in
the
spreadsheet
now,
if
it's
low,
I
can
probably
get
my
manager
to
approve
picking
up
a
bill
for
something
if
it
isn't
doesn't
go
overboard.
I
don't
know
if
we
end
up
having
the
whole
200
people
show
up,
I
know.
Well,
I
don't
think
we'll
find
a
venue
for
that.
They
will,
after.
B
Not
yet
committed
not
its
perspective.
No,
I
will.
I
will
just
post
in
the
slack
that
we
would
like
people
to
sign
up
now
so
that
we
can
start
to
get
our
heads
around
where
to
have
lunch
and
so
on.
A
And
the
other
thing
is
I've
heard
this
event
is
likely
to
hit
12
000
the
full
capacity,
so
lunch
might
be
crowded.
I
mean
that
this
isn't
like
going
to
lunch
in
las
vegas,
where
you
could
perceive
it
as
kind
of
an
infinite
supply
nearby.
A
Wednesday,
I
think,
is
okay
with
me,
I'm
sure,
whatever
day
we
choose,
somebody
will
have
a
conflict
but
yeah.
I
guess
we
can
do
wednesday.
We
could
thursday,
I
think
if
we
did
it,
then
it
would
be
before
our
presentation,
yeah
and
the
way
this
has
gone.
Historically.
You
know
we're
all
usually
editing
the
deck
right
then
so
maybe
I'd
avoid
doing
it
right
before
we
go
on
stage
plus
they
get
tired
of
seeing
us.
I
mean
if
we.
A
Right,
we
could
do
it
in
an
evening,
but
the
trouble
there
is
that
there
actually
is
a
kubecon
if
you
google,
for
it
there's
a
kubecon
party
list
of
various
events
sponsored
by
vendors
and
there's
something
going
on
every
night,
so
we'd
be
in
conflict
with
that.
What
we
could
do
is
maybe
have
the
group
crash
one
and,
as
a
group
just
chat
while
we're
attending
some
vendors.
A
A
Okay,
well
anyway,
yeah
well,
I
guess,
as
far
as
I
know,
wednesday
is
as
good
a
time
as
any
and
if
we
want
to
cancel
it
in
for
wednesday
at
lunch,
I
don't
have
any
that's
cool.
B
A
C
James
reached
me
on
slack
in
private.
He
will
be
here
on
silent
because
he's
on
customary
side,
so
I
don't.
I
don't
expect
him
to
talk.
Oh
okay,
well,
this
time
as
well,
yeah
but
yeah.
Let's,
let's
try
to
get
some
things
going
on
after
the
kubecon,
so
I
think
yeah.
You
know
james
had
some
interesting
things.
D
A
Last
time
what
was
it
it
should
still
be
in
there.
If
you've
got
it
open,
you
can
scroll
down
it's
a
it's.
A
provisioning.
A
D
Dealing
with
multiple
use
cases
where
customers
want
to
pre-install
units
with
kubernetes
and
then
have
them,
you
know
essentially
either
install
a
node
in
a
failed
cluster
or-
and
you
know,
we're
probably
a
three-node
physical
environment,
whether
they're,
intel,
nox
or
embedded
boards,
or
even
you
know,
1u
servers.
E
G
D
Based
on
you
know
something
like
a
voting
system
or
whatever
so,
and
I
know
this
is
somewhat
beyond
the
scope,
but
I
don't
know
if
you
know
everybody's
read
the
chick-fil-a
blogs
that
are
out
there
yeah.
So
so
let
me
get
essentially.
A
D
Where
the,
where
these
customers
want
to
get
is
that
they
have
a
have,
a
cluster,
come
up
and
detect
its
identity
with
pre-installed
systems
that
are
vanilla
and
you
know,
essentially
use
something
I'll
d8
like
a
kubernetes
dhcp
type
service
where
it
gets
its
identity
and
then
you
know,
decides
who's
going
to
be
the
masters
and
all
that
sort
of
stuff.
So,
but
the
first
problem
they're
have
they
want
to
solve,
is
okay.
We
go
in
and
configure
this
and
lose
a
node.
A
Having
a
worker
node
joining
existing
cluster
is
a
little
different
from
starting
from
nothing
yeah
and
especially
starting
from
nothing
with
three
potential
masters
or
something.
I
don't
know
that
that
that
possibly,
but
I
can't
say
I've
ever
heard
of
anybody
who
has
solved
this
so
yeah
well,
it.
D
Would
need
to
be
something
that
that
is
flexible
enough,
that
it's
rules-based
so
that
you
could
have
a
rule
set.
That
is
flexible,
that
you
could
define
how
it
figures
out
who
the
master
is
right.
I
mean
you
can
never
write
one
thing
that
does
it.
You
need
some
heuristic
that
that
you're
able
to
populate
yeah
chick-fil-a
was
able
to
do
it
because
chick-fil-a
has
everything
to
find
and
they
know
what
the
heuristics
are
yeah
and
they
just
define
it.
So.
A
The
interesting
thing
is
whether
you're
anticipating
a
use
case,
where
you've
at
least
got
some
sort
of
network
already
in
place
before
you
try
this
and
it
has
dhcp
and
maybe
some
dns,
that
you
could
look
up
well-known
addresses
but
yeah
and
I
think
you'll
literally
start
with
nothing
and
are
doing
nothing
but
peer-to-peer.
I
suppose
that
yeah,
I
assume
that.
D
The
requirement
would
be
that
you
have
at
least
dhcp
or
or
some
way
that
these
pick
up
a
network
identity,
that's
assigned
to
each
of
them
yeah,
because
I
mean
they've
got
to
pick
up
an
identity.
So
once
they've
got
that
identity,
then
it's
a
question
of
okay.
Do
we
pick
the
you
know?
D
You
know
how
how
it?
How
do
you
pick
the
master
and
then
once
you
pick
the
first
master,
you
know,
but
that
sort
of
stuff.
So
I
it
was
just
food
for
thought
and
I
don't
know
if
this
is
something
we
want
to
talk
about
more,
but
I
I
I've
seen
more
more
customers
asking
about
this,
including
technology,
vendors
that
are
saying
that
they
want
to
do
this.
D
G
I'm
I'm
looking
for.
Do
you
know
a
way
that
I
can
pre-assign
labels
to
to
worker
nodes
which
can
basically
switch
clusters.
D
So
what
I
was
thinking
was
dna
second
day
and
you
can
do
a
lot
of
stuff
yeah
and
I
wonder
whether
you
could
use
that
and
then
pick
the
identity
off
of
that
and
it's
secure
and
then
based
on
that
have
some
service
that
that
is
queryable
to
then
pick
up
the
kubernetes
identity.
And
then,
if
it
reboots,
you
know,
if
it
reboots
or
you
know,
re-initializes
or
gets
wiped,
then
it
could.
You
could
have
a
different
identity
in
that
registry
or
whatever
you
call
it.
G
D
That
was
that
was
sort
of
what
I
was
was
thinking,
because
I've
had
customers
that
have
used
dns.
Second
dane
and
it's
it's
real
trustworthy
and
you
can
piggyback
other
things
off
of
the
top
of
it.
A
D
D
No,
no
no
you'd
have
you'd
have
to
have
a
way
for
these
things,
to
pick
up
no
a
at
least
a
temporary
network
identity,
a
dhcp
and
then
be
able
to
get
to
cloud
or
data
center
services.
D
A
A
And
I
could
dream
of
a
scenario
where
central
I.t
ships
out
usb
keys
that
can
be
plugged
into
these
systems.
That
would
imprint
the
ducklings
with
an
identity
as
well,
and
that's
awesome.
E
A
Time
and
be
done
with
it
chuck
the
things
in
the
trash
or
not
yeah.
If
you
had
to
install
certificates
and
things
that
might
be
a
good
way
to
get
those
copied
over
too,
and
perhaps
the
process
of
this
boot
would
scrape
the
thing
out
of
that
key
and
then
erase
it.
B
D
We
could
begin
to
gather
requirements
and
throw
around
potential
design
solutions,
and
then
you
know
see
whether
it's
viable
and
if
it's
viable.
Maybe
we
turn
this
into
a
project
or
something
yeah.
I.
E
D
Out
to
that
to
the
chick-fil-a
people
just
to
see
if
they
were
interested
in
talking-
and
I
haven't
heard
back
they're
in
atlanta
where
I
live
so
I
was
hoping
I
could
tuck
them.
You
know
I,
I
don't
know
what
the
value
of
them
open,
sourcing
their
stuff
would
be,
and
I
don't
know
that
they
would
be
willing
to.
But
given
my
thoughts
on
it,
I
don't
know
how
valuable
valuable
it
would
be
either,
but
we
you
know
just
reading
they're
using
their.
D
I
think,
picking
a
few
use
cases
and
and
beginning
to
sort
of
paper
design
some
stuff
around
it.
You
know
maybe
the
disconnected
use
case
that
you're
talking
about
the
use
case
of
bootstrapping,
one
that's
connected
to
the
internet,
and
then
you
know,
maybe
maybe
one
or
two
others
and
then
just
sort
of
talk
about.
G
G
Something
like
like
device
analysis,
so
you,
you
start
basically
your
your
edge
node
in
your
cluster
and
it
kind
of
scans
through
which
devices
are
detected.
For
example,
you
have
usb
cameras
or
whatnot,
and
these.
E
D
G
D
How
do
we,
how
do
we
represent
that,
and
how
do
we
share
that
connection
among
the
among
the
cluster?
I
don't
know
that
this
is
you
know.
First,
we
need
to
solve
the
first
problem,
but
I
think
that's
also
an
issue
of
you
know
how
do
we,
how
do
we
say
I'm
node
one
and
I
have
direct
physical
access
to
this
thing.
You
know
you
may
have
an
application
that
needs
access
to
it.
You
know,
how
do
we
define
all
that
stuff?
Is
it
by
labels
or
or
what?
G
Going
down,
I'm
currently
working
with
labels,
basically,
but
still
by
hand.
So
so
I
basically.
E
D
Yeah,
because
I
think
this,
if
it's
like
a
registry
and
identity,
you
could
also
begin
to
say
you
know,
someone
could
create
an
application
that
that
would
begin
to
label
these
things
with
with
you
know,
if
it's
a
warehouse
and
one
one
has
proximity
to
a
bluetooth
sensor
or
something
like
that,
then
it
could
register
that
in
some
way
and
then,
if
I
needed
to
interact
with
it,
I
could
go
and
talk
to
the
application.
That's
on
something
like
that.
D
D
A
What
you
can
do
is
there
there
is
the
membership
alias.
So
what
is
it?
Okay,
burnetti's
iot
edge,
there's
the
full
common
thing
and
all
these
things
is
to
actually
just
shift
the
group,
and
then
there
is
a
sad
history
that
if
these
things
get
published,
that
trolls
go
and
edit
this.
So
my
advice
is
that
you
actually
restrict
it
not
to
anybody
with
the
link
but
to
members
of
this
group
like
implemented
authority
and
that's
generally
safe.
A
D
B
That's
exactly
what
I
was
gonna
say
is,
I
think
if
we
solve
this
for
a
use
case
or
two,
then
it
would
lead,
it
would
be
like
a
guiding
light
toward
figuring
out.
How
does
this
become
an
abstracted
universal
layer
right
or
approach?
Yes,
so
I
think
participating
and
solving
it
for
a
couple
of
stand-up
instances
would
probably
inform
the
entire
group
really
well.
A
A
What
you
could
do
is
have
the
it
admins
ship,
like
a
burner
grade
cell
phone,
one
of
these
20
20
bucks
cell
phones,
that
the
advantage
is
it
it
presumably
even
has
gps
in
it,
and
you
know
kind
of
authenticates
that
this
has
ended
up
at
the
desired
location
and
didn't
get
misdirected
to
or
stolen
by
somebody
who's
not
entitled.
So
it
only
works
from
the
right
location
and
these
nodes.
When
you
say
how
do
they
know
how
to
connect?
A
Even
though,
in
the
final
operational
context,
they
never
intend
to
be
using
bluetooth,
it
actually
establishes
a
proximity
because
it
isn't
going
to
go
very
far,
because
I
would
presume
that
it's
safe
to
assume
that
if
I've
got
three
clusters
in
a
kubernetes,
node
they're
installed
very
close
to
one
another,
and
you
just
hold
the
phone
near
them
and
they
wake
up
expecting
to
find
somebody
on
the
air
that
is
authoritatively,
imprinting
them
with
their
configuration.
D
B
B
It's
it's
not
the
first
time
it's
been
exposed,
that's
actually
some
of
the
earlier
demonstrations
of
the
bluetooth,
low
energy
abstraction
layer
in
in
the
eclipse,
io
fog,
we're
demonstrating
that
a
node
could
identify
where
it
was
without
any
gps
or
anything
by,
like
you
know,
kind
of
proximity
to
a
beacon.
So
this
is
starting
to
shape
up
pretty
interesting.
B
I
had
probably
I
don't
think
it
would
cost
much
more
than
usb
keys.
So
no,
in
fact
the
le
beacons
are
can
be.
You
know
dirt
cheap
and
some
of
them
are
even
flashable
so
that
you
can
avoid
having
anyone
understand
your
identities
beforehand.
B
I
had
put
for
the
talk
down
in
a
couple
weeks
at
kubecon,
north
america.
I
had
put
some
info
about.
You
know
I
could
do
a
demonstration
of
bluetooth,
low
energy
hardware,
abstraction
on
edge
nodes.
I
don't
know
if
it
fits.
I
know
what
we
haven't
yet
figured
out
our
thing,
but
well.
A
B
Don't
we,
what
do
we
do?
Why
don't
we
do
a
bluetooth,
low
energy
location
detection
on
on
wake
up
that
can
pass
that
information
along
that's
a
great
demo,
because
we
can
also
demonstrate
all
of
the
bluetooth
low
energy
communication
layer
right,
which
is
which
is
a
non-tcp
low
power
protocol.
You
know
layer
or
communication
layer
so.
A
Because
we
said
we'd
have
protocol
demos
so
sure
it
sounds
that
qualifies
and
then
on
top
of
that
we
can
throw
out
james
ideas
that,
as
far
as
we
know,
nobody's
done
this
before
but
we'd
like
to
have
other
people
contribute
use
cases
and
call
it
feature,
requests
and
see.
If
we
can
get
volunteers
from
around
the
world
to
help
build
it.
C
I
think
they
will
be
happy
to
see
some
actual.
You
know
deliverables
yeah.
I
don't
think
even
if
it's
be
like,
like
a
you,
know,
side
project
that
can
actually
come
to
cncf
or
whatever.
D
A
D
Starting
to
see
is
some
some
standardization
around
cluster
api,
but
it's
still
we're
still
several
years
from.
You
know
how
it
is
yeah.
A
But
yeah
anyway,
okay,
but
I'm
glad
you
brought
up
this
subject
because
it's
it's
interesting
and
yeah.
I
think
it's
unique,
but
I
think
it's
solvable
yeah
and
yeah
I
mean
it
looks.
It
sounds
to
me
like
something
that
half
the
edge
deployments
could
use
too.
I
mean
oh
yeah
when
you
think
about
it.
A
lot
of
these
edge
locations
are
not
attended
to
by
skilled
I.t
people.
Correct.
A
A
The
interesting
thing
with
the
idea
of
it
being
a
cell
phone
is
that
it
maybe
could
get
some
internet
connectivity
to
draw
in
kind
of
latest
versions
of
containers
and
things
as
you
come
up.
If
you
have
this
idea
that
devices
are
fedexed
out
from
some
factory,
the
vendors
are
going
to
are
going
to
want
to
produce
at
volume.
A
Probably
don't
want
to
do
like
last
minute,
squirt,
the
last
updated
version
of
kubernetes
out
there
I
mean
granted.
The
releases
are
quarterly,
but
there's
there's
been
a
history
of
a
lot
of
security
patches
coming
up
at
intervals
more
like
every
couple
of
weeks,
and
if
you're
going
to
bring
one
up,
you
don't
exactly
want
to
bring
one
up
with
known
vulnerabilities.
B
I
think
that
the
any
any
edge
node
that
comes
online
from
you
know
it
goes
through.
The
bootstrap
process
should
probably
include
steps
to
verify
what
I
mean,
what
what
agent
software
right
is
running
on
the
edge
node
before
it
begins
running
application
layer,
software.
B
D
A
A
B
Out
whatever
process
you
go
through
to
get
ready
to
join
the
the
you
know,
the
the
network
or
the
cluster
once
you
finish
it,
I
think,
then
you
should
receive
your
your
identity
for
for
being
part
of
the
application
pool,
and
it
should
probably
only
last
as
long
as
you're
valid
meaning
that
if
you
fall
out
of
order
in
any
sort
of
way
that
that
you
know
we
should
be
able
to
detect
that
and
and
make
you
go
through
a
process
of
recertifying
yeah
a
lot
a
lot
of
thoughts
about
it.
D
A
B
That
that's
great
and
it
should
result,
then,
in
a
lot
of
kubernetes,
ready
pieces
of
hardware
right
so
correct.
E
B
Off-The-Shelf
stuff
that
you
can
just
drop
in
and
I
think
that
would
that
would
spur
a
little
bit
more
industry
adoption
too,
if
it's
that
easy
yeah,
that's
that's.
D
I
we
can
get
to
the
point
where
you
know:
company
x
builds
an
appliance
three,
you
know
appliance,
you
drop
three
of
them
in
in
a
remote
location
and
they
they
come
up
with
the
software
and
federate
and
could
just
come
up
as
nodes.
Then
that's
you
know,
I
think,
that's
that's.
That's
products
being
born
there.
That's.
B
D
Yeah
they're
more
they're
more
mechanical
than
than
I
t
yep
awesome.
Well,
I've
got
a
drop.
I
just
wanted
to
listen
in
and
then
bring
this
up,
but
okay.
Well,
thanks.
D
Excellent
well
I'll
see
you
guys
in
san
diego.
A
C
A
A
You
know,
I
don't
even
know
if
bluetooth,
low
energy,
it
would
be
viable
if
it
looks
enough
like
what's
out
there
to
have
it.
Look
like
a
pixie
boot
or
something.
B
I
don't
don't
think
anything.
I
I
think
that
stack
doesn't
even
come
up
until
after
we're
well
yeah
well
into
the
above,
the
the
the
bootloader
yeah
but
interesting
what
the
future
could
hold,
because
it
is
the
the
way
that
you
do.
Pairing
of
bluetooth,
low
energy
devices
typically
is
or
sorry
the
way
that
you
give
iot
devices
their
internet
connection.
Usually,
is
you
use
bluetooth,
low
energy,
which
is
you
just
basically
need
to
stand
near
it
to
get
it
to
get
access
to
it
via
your
cell
phone?
B
G
B
Fire
tv
stick
doing
that
it
might.
I
don't.
G
G
But
what
is
really
cool
when
you,
when
you
add
a
new
device
from
them?
It
already
has
wi-fi
installed
and
just
boots
up,
and
it's
just
like
way
too
easy.
B
Oh
boy
yeah,
that's
that
crosses
over
into
the
the
lines
of.
If
you
make
it
so
easy,
then
you
know
yeah
in
the
industrial
space.
I
think
we're
dealing
more
with
abstracted
identities
that
have
only
meaning
in
the
back
end,
not
not
actual
amazon
accounts
but
yeah.
You
know
who
knows
where
amazon's
going
to
play
next.
A
As
a
downside
of
that,
I've
read
an
article
recently
out
of
I
think
I
found
it
on
hacker
news,
but
somebody
observing
that
they
had
rented
a
car
that
was
internet
connected
with
a
mobile
phone
app
and
they
went
to
the
trouble
to
configure
it,
and
this
app,
then,
is
anticipated
you're
the
car
owner.
B
There
is-
and
the
first
thing
it
makes
me
think
of,
is
if
I'm
a
hired
technician
to
go,
do
an
install
at
a
factory.
That's
that's,
you
know
kubernetes
at
the
edge
and
then
I
decide
to
come
back
by
on
a
saturday.
Can
I
just
go
ahead
and
reboot
the
whole
system,
with
my
handheld
device
that
I
use
to
bootstrap
them?
This
is
definitely
well.
B
A
Lingering
credentials.
B
Yeah
it
all
points
to
you
know,
which
is
something
that
we
we
already
have
exposed
well
in.
The
security
white
paper
from
this
group,
but
fixed
identities
at
the
edge
are,
are
probably
not
a
good
idea,
just
as
in
general
that
concept,
because
we
put
so
much
trust
in
things
that
we
normally
can
protect.
But
you
can't
protect
once
the
car
is
no
longer
your
rental
right.
It
might
be
somebody
else's
problem,
but
the
identity
remains.
G
G
Preston
talked
about
this.
I
I
think,
four
weeks
ago,
with
this
keycard
idea,
where
you
kind
of
kind
of
oh,
yes,
different
processes,
which
I
found
really
interesting,.
A
Well,
your
idea
of
the
bluetooth
beacon,
if
you
envision,
that
is
something
you'd
epoxy
down
to
a
location
so
that,
even
if,
if
the
equipment
moves,
you
know
it
loses
its
identity,
that'd
be
interesting.
I
had
another.
A
B
Yeah
on
the
note
of
of
where
you,
where
you
mentioned,
you
know,
being
able
to
detect
the
the
location
and
then
make
a
change
one
of
the
very
first
demonstrations
of
of
eclipse
io
fog
before
it
was
that
project,
but
but
it
didn't
even
have
a
name
at
the
time.
B
B
When
you
moved
it
physically
over
to
another
location,
with
the
with
the
pharmaceutical
cart,
it
could
detect
based
on
bluetooth,
low
energy
beacons
that
you
had
entered
into
a
space
where
the
same
ability
to
issue
the
medication
was
not
valid.
You
were
no
longer
in
a
protected
space
and
it
and
it
locked
out
the
drawers
on
the
cart.
So
you
could
do
some
work
in
one
room,
but
not
in
another,
and
it
knew
where
you
were,
and
so
it
shut
down
some
of
the
interfaces.
It's
pretty
interesting
potential
here.
A
How
solid
are
those
with
regard
to
spoofing
like?
I
know,
people
kicked
around
the
idea
that
you
put
gps,
but
unfortunately,
hackers
have
definitely
learned
how
to
how
to
spoof
gps
locations
and
these
bluetooth
beacons.
Are
they
reasonably
secure
with
regard
to
imposters
and
situations
like
that.
B
Nope
but
but
work
could
be
done
to
provide
a
layer
that
that
allows
you
to
understand.
If
it's,
you
know
to
verify
the
identity
of
one
right
now
I
mean
they
broadcast
on
a
pulse
and
what
they
send
out
is
by
kind
of
the
definition,
is,
is
accessible
for
everyone
to
read,
and
they
really
are
just
a
beacon.
Everybody
can
see
it
meant
to
point
away,
but.
B
No
there's
not
a
way,
there's
not
a
way
of
just
using
like
the
the
standard.
You
know,
beacon,
definitions
that
are
there.
You
know
ibeacon
and
eddie,
stone
and
whatever,
but
that
doesn't
mean
that
using
something
as
a
beacon
can't
be
developed
into
a
a
much
more
secure
layer.
B
If
we
could
do
a
demonstration
with
with
the
ble
beacon,
I
would
say
that
the
first
thing
we
should
we
should
point
out
is
that
that
serves
as
an
easy,
easy
demonstration
to
get
started,
but
that's
not
the
layer
that
we
would
trust
in
enterprise,
but
very
good
points
and
sounds
like
additional
material
for
the
security
white
paper.
A
On
a
related
note
to
that,
I
read
another
article
about
how
somebody
started
coming
up.
I
think
the
patents
expired
on
those
hid,
chips
that
are
used
pretty
commonly
for
building
access,
and
things
like
that,
and
it
has
put
out
kiosks
to
duplicate
these
things
for
less
than
ten
dollars.
Yep.
B
A
Yeah,
I
the
article
said
they
sell
both
the
cards
and
even
chips
that
you
could
potentially
glue
to
your
fingernail
or
your
watch
and
the
sales
point
wouldn't
be
to
have
a
backup,
but
just
that
it's
more
convenient.
So
you,
oh
boy,
glue
it
to
your
fingernail,
so
you
don't
have
to
carry
around
the
id
card.
B
A
G
Just
one
question
about
the
use
case:
what
is
exactly
now
the
plan?
We
kind
of
just
write
some
some
ideas
down
what
we
want
to
tackle
first
or.
A
You
mean
the
use
case
of
booting
up
a
kubernetes
cluster
at
edge.
Is
that
exactly,
I
think
james
said
he
was
going
to
post
a
document
and
us
I
assume,
he'll,
put
a
little
structure
in
it.
Okay
and,
I
think
just
add
whatever
you
like,
I
mean
after
he
left
the
call.
We
came
up
with
a
few
more
ideas.
I
think
at
this
page.
A
The
best
thing
to
do
is
free
format
form
at
brainstorm
mode,
where,
if
you've
got
anything
potentially
valuable
to
add
if
it
relates
to
something
already
there
add
it
in
that
part,
but
feel
free
to
have
a
complete
new
section.
If
you
like.
G
Yeah,
because
what
I
was
just
thinking,
what
would
be
really
nice
is
actually
changing
clusters.
So
basically,
let's
say
you
have
you
have
your
your
edge
device
with
your
camera
system
and
you
just
move
the
camera
to
a
different
location
where
another
kubernetes
cluster
is
running
and
you
just
plug
it
into
the
network.
There.
It
kind
of
boots
up
into
the
system
as
well
and
then
grabs
the
the
camera
software
that
is
actually
run
there
in
this
specific
cluster.
So
you
can
basically
transfer
devices
in
between
clusters
to
to
to
and
manage
the
software.
B
This
is
moritz
I'd
say
this
sounds
like
a
a
great
topic
for
the
next
working
next
one
of
these
working
group
meetings.
So
to
keep
this
rolling
and
and
if
we
get
the
the
documents
set
up
from
james
there
to
kick
it
off
and
you
had
some,
I
bet
you.
We
could
spend
a
whole
session
just
looking
at
the
document
and
starting
to
get
more
structure
into
it.
A
Yeah
yeah
and
that
idea
of
moving
from
one
cluster
to
another
I
mean
ordinarily
you'd
expect
that
any
worker
node
is
knowledgeable
as
to
the
identity
of
its
masters,
and
now
those
have
changed.
So
there
is
a
security
aspect
to
that
yeah,
especially
if
it's
a
security
camera
you're
talking
about
you,
don't
want
it
to
become
too
easy
to
go.
Seize
control
of
the
camera.
G
Yeah,
let's:
let's
talk
about
the
use
case
or
the
vision
of
that,
so
I'm
basically
working
in
a
laboratory
environment
for
different
manufacturing
systems.
So
what
we
have
here
is
different
different
machining
parts
or
different
sensor
systems
lying
around
and
we
do
experiments
with
that.
So
you
might
have
your
your
production
machine
in
the
middle
and
then
you
have
your
your
camera
system
and
you
have
your
temperature
sensor
and
so
on
and
our
stuff
actually
gets
moved
around
a
lot.
G
So
you
might
have
your
your
production
system
a
and
you
you
try
to
do
your
experiments
on
there
and
then
you
go
to
another
machine
and
you
take
the
camera
with
you
because
you
just
have
one,
but
you
just
need
it
for
this
one
experiment.
So
what
you
do
is
you
move
your
your
camera
to
this
location
and
what
would
be
nice
is
in
case
or
in
in
the
case
where
I
move
the
camera
to
another
cluster
or
into
another
machine.
G
It
kind
of
boots
up
itself,
and
you
get
your
your
your
database
with
your
with
your
camera
system.
You
get
your
your
access
interface,
your
web
interface,
everything
with
this
movement
of
the
of
the
note,
so
this
is
like
heavily
experimental
bass.
I.
A
Can
think
of
another
video
camera
related
thing
with
more
use
case,
and
that
is
drones.
So
if
you
have
cleaned
a
drone,
that's
moving
from
one
region
of
a
city
to
another,
maybe
handing
off
between
clusters
or
autonomous
vehicles
other
than
drones
might
be
a
similar
thing.
Yeah
interesting
idea.
A
B
Ones,
there's
some
some
some
work
that
I
I
can't.
I
can't
talk
about
the
details
of
right
now
due
to
customer
confidentiality,
but
I
can
say
that
in
the
automotive
space,
even
without
automation
for
the
driving,
so
is
it
just
a
regular
person.
Driving
handoff
is
a
is
going
to
be
a
very
important
way
to
address
the
the
you
know,
the
v2x
scenarios
that
we're
encountering.
A
C
C
A
Okay:
okay,
bye
everybody
thanks
for
joining,
see
ya.