►
From YouTube: 1/29/2020 meeting recording
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Now,
I'll
just
make
two
quick
comments
on
that:
Steve
I,
don't
wanna
do
my
just
do
lots
and
lots
to
talk
about
on
the
one.
Is
that
the
cloud
native
spec
is
a
you
know,
a
modicum
of
a
step
towards
a
standardizing
of
a
like
event:
definitions
right,
which
is
always
a
thing
in
IOT,
that's
been
a
struggle
and
then
the
other
is
that
ya,
the
server
lists
and,
like
you
say,
it's
and
part
of
where
I've
kind
of
disappeared
off
to
is
I've,
been
spending
a
lot
of
time.
A
Working
around
cloud
run
product
which
is
not
an
edge
thing
at
all,
but
it's
serverless
and
scales
to
zero.
So
the
K
native
side
of
it
does
somewhat
solve
that
that
latency
start
because
you
do
run
essentially
the
event
or
request
routing
parts
in
kubernetes
all
the
time
and
then
are
only
starting
up.
You
know
serving
containers,
so
key
natives
a
little
bit
better
than
scaling
up
from
zero
from
four
other
kubernetes
stuff
because
part
of
its
running
all
the
time,
so
it
can
buffer
the
requests
anyway.
A
Well,
I
I'm,
just
I'm,
just
helping
look
for
things
that
are
because
cloud
events
went
one
point:
oh
there's
a
timing,
relevance
of
saying:
hey!
Here's!
Here's
here's
an
introduction
to
this
notion,
but
my
kind
of
issue
with
it
is:
it's
still
very
notional.
It's
not
like.
We
can
go
until
a
super,
concrete,
real
story
around
it.
C
B
C
B
B
So
I
think
I
like
what
we
talked
so
just
to
share
how
I
was
thinking.
You
know
like
I've
reviewed
the
we
use
kubernetes
to
deploy
on
run,
applications
sound
to
edge,
and
then
we
know
like
a
cobra
need
is
to
manage.
It
is
hard
so
for
this
talk,
I
was
also
hoping
to
understand
the
pain
points
for
DevOps
developers
and
operate
operators.
B
So
what
other
like?
What
can
they
do
to
help
them
like
become
a
you
see,
RDR
like
home,
are?
Are
we
having
the
right
tools
for
developers
to
to
write
their
applications
and
then
be
able
to
deploy
it
very
easily?
So
that's
another
real
thought
for
the
for
the
talk
and
then,
as
you
all
know,
we
are
gonna.
Do
a
survey
against
all
the
attendees
so
I'm,
hoping
that
we
can
get
some
broad
understanding
of
coffee
power
like
how
people
are
doing
and
how
people
are
thinking.
B
B
C
C
C
B
I
agree:
yeah
I
was
hoping
like
it,
we
have,
we
can
get
a
better
understanding
and
even
for
service
you
know
or
deploy
a
service
function
to
a
cloud
is
much
easier
than
how
they
do
to
the
edge
I.
D
Think
there's
there's
a
couple
of
broad
topics.
I
would
like
to
see
in
the
survey.
So
one
thing
is
the
deployment
for
those
people
years.
So
what
kind
of
like
a
get
up
source
here
see
ICD
trains?
People
are
planning
to
use
or
what
is
see
missing.
Others
is
all
this
communication
issue
right,
so
cloud
events
or
MPD,
T
or
ERP,
see
you
know.
Where
are
people
now
I
would
like
to
you
know
fish
out
the
problems
for
the
people.
B
And
also
like
for
the
programming
model.
Yes,
where
people
are
writing
applications,
but
they
run.
You
know
this
much
distribute
distributed
fashion
and
then
the
network
topology
is
complicated.
So
I'm
not
sure
if
you
guys
know
a
project
called
dedapper,
that's
what
recently
Microsoft
contributed
to
the
community
I've
seen
a
lot
of
people
are,
are
interested
in
neighbor
and
contributing.
So
the
idea
is
to
ease
people's
pain
when
they
write
this
or
write
or
distributed
application.
D
B
B
B
A
Service
mesh,
sidecar
idea
of
something
like
is,
do
and
then
instead
kind
of
building
it
as
an
SDK
client
library
that
you
deployed
you
out
so,
in
other
words,
just
sort
of
talk
to
other
services
or
do
things
with
other
services
in
this
grouping.
You
you,
instead
of
making
it
like
a
request
or
generating
an
event.
You
use
a
very
rich
SDK
and
then
it
handles
all
the
service
to
service
or
event
type
stuff.
Yeah.
B
B
B
B
A
A
B
C
C
We
started
this
meeting
with
nothing
in
the
agenda.
I
can
throw
in
a
few
community
notices
I,
since
some
of
us
are
on
the
west
coast.
This
maybe
isn't
super
attractive
to
the
rest
of
you.
Could
there
are
definitely
IOT
related
sessions
at
the
upcoming
scale
conference
and
then
in
April.
There
is
this
Linux
Foundation,
oh
and
me
yes,
conference
with
CR
CFP
is
open
for
a
little
less
than
another
week,
and
they
have.
One
of
the
categories
is
kubernetes
at
edge
for
the
conference.
C
B
B
B
C
Experience
because
people
have
to
go
to
lunch
anyway
and
we
can
either
meet
at
a
restaurant
nearby
or
the
other
thing
that
works
is
to
simply
get
some
signs
made
in
the
conference
itself.
I
presume
is
going
to
have
a
lunch
area.
We
can
cut
out
some
signs
on
tables
so
that
we
can
find
each
other
and
just
meet
during
the
standard
conference
lunch
whatever
people
would
prefer
yeah
familiar
with
the
area
in
Amsterdam.
Sometimes
the
Linux
Foundation
ends
up.
Locating
these
in
kind
of
isolated
areas.
There's
a
lot
of
options.
B
B
C
We
did
we
got
approached
at
the
event
by
a
couple.
People
one
was
the
new
stack
and
the
other
was
I,
don't
remember
the
name,
but
it
was
a
telco
focused
podcast
based
in
Argentina,
then
I
wasn't
even
familiar
with
and
it's
a
CN
CF
people
get
approached
by
press.
So
this
wasn't
something
that
we
saw
coming
in
advance.
It
was
just
during
the
event
somebody
from
the
CN
CF
tracked
us
down
and
asked
if
I.
C
You
know
I
think
it's
just
take
advantage
of
the
opportunities
to
reach
out
and
talk
about
subjects.
People
are
interested
in
and
I
think
that
those
podcasts
amounted
to
something
where
you
got
the
initial
visibility
so
that
people
knew
you
were
out
there
and
then
word
spread
and
they
found
us.
It
wasn't
the
week
when
seeking
them,
yeah.
C
C
A
C
A
A
A
Is
the
the
chip
alignment
between
Google,
Amazon
and
Apple
around
weave
and
thread
which
has
inside
it
deep
in
weave
is,
is
an
event
spec
and
some
other
API
stuff,
but
I
think
in
general,
when
it
comes
to
all
things
in
in
IOT
and
edge
that
people
would
rather
build
to
open
standards
that
define
things
like
event,
payload
shapes
or
envelopes,
at
least
right
and
I.
Think
cloud
events
will
be
interesting
for
two
things.
A
One
is
if,
if
there's
enough
traction
among
event
producers,
you
know
like
cloud
platforms
to
produce
events
so
that
they
can
be
picked
on
and
then
the
other
is
to
reuse.
The
spec
in
custom
bespoke
deployments
that
that
aspire
to
create
their
own
ecosystems.
At
some
point
right,
that's
would
be
the
other
way
that
that
cloud
events
would
be
relevant.
D
I'm
thinking
from
the
perspective
of,
for
example,
it
Clips
Conover,
where
we
try
to
not
do
the
same,
but
but
a
little
bit
similar
thing
just
try
to
define.
Basically
the
the
messaging
API
is
for
for
telemetry
Zanden
commands,
and
things
like
that.
So
so,
and
that
also
then
Maps
so
different
protocols
to
HTTP
to
impute
et
I'm
trying
to.
But
what
I'm
trying
to
figure
out
is
there
something
we
could
do
there
to
to
align
with
the
cloud
events
as
well
and-
and
maybe
maybe.
A
C
I
think
the
adapter
does
make
sense
when
something
already
got
established
to
bring
it
into
a
different
thing.
That's
become
popular.
That
makes
sense,
I
mean
the
whole
gateway
thing
is,
could
be
viewed
as
that,
either
from
a
software
or
hardware
perspective
right
to
bridge
between
something
with
an
existing
installed
base,
rather
than
try
to
revolutionize
the
world.
Well,.
A
And
I
mean
I
think
the
pragmatic,
brutal
reality
is
that
in
any
given
project
you
just
write
whichever
adapter.
You
need
specific
for
your
use
case
in
payload
and
anything
any
project
that
kind
of
tries
to
come
up
with
the
universal
translation
framework
of
all
like
an
all-encompassing
adapter
thing.
Isn't
something
you'd
probably
want
to
use,
because
it
becomes
overly
complicated
for
what
you
need
to
do
when
you
can
just
write
your
own
adapter
for
what
you
need
right
so.
C
D
Yeah
but
I
think
that
there's
a
complementary
features
between
between
the
cloud
events
and
what
Hall
is
trying
to
do
because
yeah
we
don't
consider
payload
at
all
so
so,
for
example,
for
n
qdt,
it's
just
it's
it's
just
a
you
know.
Binary
blob
did
get
past.
So
what
we're
focused
on
is
like
a
topic
name
ever.
D
Yeah,
oh
no,
just
just
more
like
a
communication
patterns
like
like
dumping,
name
namespaces,
and
then
how
do
you
address
things?
And
how
do
you
things
and
the
idea
is
to
have
exactly
a
stream
of
telemetry
data
on
a
single
address
in
the
backend,
no
matter
how
it
came,
but
but
the
the
method
metadata
we
put,
there
is
really
very
minimal.
A
D
B
Okay,
so
I
suggest,
like
we
said
I,
can
do
that
as
well.
Just
ask
people
in
the
channel
to
see
what
kind
of
survey
questions
they
are
interested
in
will
consolidate.
I
mean,
while
I
set
up
a
meeting
with
Kim
from
CN
CF.
We
can
follow
the
right
process
and
create
them,
like
my
thought,
is
to
to
use
the
CAF
SurveyMonkey
and
then
we
can
start
from
there.
So.
C
B
C
C
Then
you'll,
hopefully
come
up
with
more
than
we
can
afford.
An
interesting
aspect
of
these
surveys
is,
if
you
put
one
out
there
with
a
hundred
questions,
people
give
up
on
it
and
don't
hit
submit,
and
then
you
hardly
get
any
entries.
So
what
you
really
need
in
one
of
these
surveys
is
something
where
you
say:
ncf
will
send
out
an
email
and
the
email
has
to
say
something
like.
C
Could
you
give
us
five
or
ten
minute
if
you
suggest
this
survey
is
gonna,
take
more
than
about
ten
to
complete
the
only
people
with
enough
skin
in
the
game
to
put
that
kind
of
time
into
it
are
gonna,
be
vendors
trying
to
promote
something
and
the
results
will
show
that
kind
of
influence.
So
if
you
really
want
the
users
to
participate,
you're
going
to
have
to
have
that
letter,
give
them
a
time
box
saying
that
look.
We're
only
asking
for
you.
C
People
suggest
five
minutes
some
ten
and
to
get
it
down
with
it.
It
means
you
can't
have
more
than
about
twenty
questions.
In
my
opinion,
maybe
if
they
were
really
quick,
yes/no
things,
but
if
you've
gotta
think
about
that,
you
can't
many
and
we're
gonna
have
to
collect
respective
questions
them.
Maybe
have
a
internal
survey
that
votes.
B
B
D
B
C
Thing
maybe
we'd
want
to
do
in
the
survey,
but
some
is
get
a
question.
Some
of
them
are
to
gather
information,
but
some
of
them
might
be
to
classify
the
audience
or
the
answer.
Aires
and
one
of
the
issues
with
edge
is
that
we've
got
some
fundamentally
different
use
cases
where
you
know
the
one
I
described
is
kind
of
retail
remote
office.
C
Our
have
enough
resource
budget
money-wise
and
space
wise
to
run
multi
node
kubernetes
clusters,
yet
device
edge,
they're,
probably
hard-pressed
to
come
up
with
the
resource
for
one
ein,
expect
quite
different
answers
on
the
survey.
Questions
from
those
but
it'd
be
nice
to
know
what
they're
coming
from
so
that
when
we
present
the
results,
we
maybe
can
characterize
it
as
saying
look.
People
with
this
use
case
think
these
items
are
critically
important,
yet
ones
with
a
different
one
lean
a
different
way.
Yeah.
B
D
C
B
E
Just
gonna
add
that'll,
be
helpful
to
have
a
docking
drive
so
that
we
can
take
a
look
at
what
we're
asking
there
and
I'm
definitely
interested
in
getting
information
out
of
it,
but
Steve
I
think
you're
right,
which
is
you
you?
You
have
a
lot
to
handle
here.
Otherwise,
so
you
get
noise
and
then
you
rely
on
it
and
that's
bad
right.
So
we
definitely
don't
want.
D
C
C
F
C
If
you
know
they're,
they,
both
in
common,
have
kind
of
cut
out
unneeded
things
in
kubernetes
to
put
kubernetes
on
a
diet
to
demand
lesser
resources,
but
maybe
they've
make
taken
some
different
elections
on
how
they
about
that.
You
know
in
terms
of
replacing
at
CD
or
or
not,
and
what
you
replace
it
I
think
they
both
taken
out
cloud
providers
and
storage
providers,
but
they
might
have
different
aspects
of
how
should
you
want
to
get
one
of
those
things
back
in
you'd
managed
to
do
it
so.
F
Wait
wait
and
my
crickets
does
not
take
anything
out
from
the
binaries.
It's
the
upstream
binaries
right
now,
all
right
and
it's
we
don't.
We
don't
drop
anything.
The
the
part
that
I
think
is
relevant
to
this
group
and
I
did
say
we're
doing
a
good
job.
There
is
because
we,
my
croquettes,
is
basically
snuff.
F
F
Yeah,
so
so
yes,
this
exactly
what
we,
what
we
do
in
place,
we
basically
release
the
same
day
as
upstream.
We
take
the
binaries
package
them
ship
them.
Well,
if
you're
following
a
stable
release,
we
we
get
that
you
get
a
delay
of
one
week
just
in
case
but
yeah.
This
is.
This
is
one
of
the
issues
that
we
one
that
we
feel
would
have
addressed
very
elegantly
and
the
other
issue
that
the
IOT
devices
often
face
is
because
they
might
be,
and.
F
We've
seen
that
that
often
on
the
IOT
world
in
the
edge
world,
you
need
some
extra
security.
So
there
is
this
idea
that
the
entire
kubernetes
cluster
ants
under
what
we
call
three
confinement,
then
is
essentially
an
akan
security
state
state
where
everything
runs
under
a
farmer
and
we
its
if
that's
another
layer
of
security.
So.
C
F
Yes
yeah,
so
so
these
are
really
good
points.
I
hope
you
will
be
able
to
cover
them
in
the
next
fitting,
but
this
comes
the
this
comes
again
as
when
sherry,
please
as
part
of
the
snobs,
where,
basically,
when
you
get
a
new
revision,
it
tries
to
start
if
it's
not
starting
it
rolled
bugs
exactly
the
way
it
was
so
yes,
you're
right.
This
is
a
this
is
a
and
a
common
ask,
don't
break
us,
and
if
you
do,
please
don't
back
the
entire
infrastructure,
not
just
the
their
work.
Look
that
I'm
running
on
ok.
D
E
If
you
can,
when
you
present,
you
could
talk
about
the
the
scale
out,
Federation
and
opportunities
right,
because
I
think
you've
got
a
a
lot
of
great
thinking
around
what
life
is
going
to
be
like
to
bring
these
down
to
the
edge
in
this
way.
But
then
you
know
bringing
bringing
kubernetes
itself
down,
introduces
a
lot
of
challenges.
F
C
F
C
Call
it
this,
the
other
thing
for
edges
may
be
coverage
of
what
the
hardware
requirements
might
be.
You
know
it's,
it's
pretty
obvious.
You
can
run
it
on
something
like
an
x86
look
or
even
a
pond.
Is
there
experience
base
running
it
on
things
that
look
more
like,
say,
Hardware
made
for
Android
phones
and
things.
C
F
So,
on
that
point,
I
think
k3s
has
gone,
has
done
a
better
job
in
the
sense
that
the
kubernetes
binders
by
themselves,
the
way
they
come
from
upstream
they
they
are
not
light.
Okay,
then
I,
don't
think
in
the
past
were
targeting
IOT
devices.
Okay,
but
maybe
they
they're,
probably
where
were
targeting
it
or
thought
ID
devices.
So
kaitlin's
has
done
some
some
interesting
work,
they're
run
silent
with
kiiiers
I.
Don't
want
you
to
misrepresent
so,
but
from
what
I've
heard
they
they
have.
F
Yes,
they
have
removed
some
of
the
storage
providers,
for
example,
and
they
have
also
put
on
binaries
on
executable
ISM
on
binary.
So
this
has
a
smaller
memory
footprint
anyways.
They
have
looked
a
lot
into
this
subject
and
I
would
be
really
interested
to
see
a
presentation
from
those
guys
as
well.
They
they
are
doing
a
good
job
but
yeah.
Definitely
if
this
group
wants
to
push
something
for
something
a
smaller
footprint
on
kubernetes
might
be
one
of
these
things.
Yeah.
C
Yeah,
if
certainly
this
group,
we
don't
own
the
kubernetes
codebase,
but
we
can't
contact
the
SIG's
that
so,
if
in
your
development
of
micro
k8s,
you
come
up
with
ideas
of
Whitely
you'd
like
to
see
in
terms
of
reducing
the
size
of
the
kubernetes
binaries
I
think
that
we
can
approach
the
people
who
actually
owned
those
Patrick
packaging
aspects
and
put
the
idea
on
the
table.
Mm-Hmm.
F
C
Yeah,
the
way
that
is
usually
done
is
a
group
would
start
something
called
and
kept
the
kubernetes
enhancement
proposal,
which
is
just
a
docx
entered
under
github,
but
I've
gone
through
the
process
before
for
other
things,
so
that
that's
where
we
could
take
that.
But
with
sounds
like
we
need
a
little
bit
more
definition
of
what
the
ask
would
be
mm-hmm.
F
Another
thing
that
we
are
looking
at
both
us
and
k3
hazards-
probably
a
common
pain
point-
is
the
the
data
store
that
is
more
or
less
kubernetes
is
fixed
on
one
TCB,
which
is
which
is
great
because
it
scares
it's
it's
well
tested
in
everything
for
IOT
devices.
You
would
like
something
lighter,
and
this
is
why
we
are
experimenting
with
a
couple
of
another
couple
ways:
I
can
baby
embedded
databases
so
so
that
we
reduce
the
the
memory
footprint.
F
C
C
B
C
C
C
C
The
other
thing
is:
if
you
look
at
the
actual
docks,
were
running
kubernetes
production
grade,
one
of
the
things
that
this
isn't
unique
to
kubernetes,
but
in
general
you
need
to
start
worrying
about
backups
and
addy
our
plan
and
the
kubernetes
Doc's
tell
you
that
the
way
to
do
a
backup
of
your
kubernetes
state
is
to
do
a
net
CD
backup,
yet
they
don't
document
and
equivalent
process.
Should
you
be
doing
something
else,
so
you
really
to
take
it
into
production.
C
F
F
C
I
think
this
would
be
an
interesting
thing
to
try
to
take
on
if
you
could
get
the
interesting
parties.
You
know
people
doing
kubernetes
distros
for
edge
like
micro,
Kade's,
k3s
in
a
room
and
try
to
whiteboard
and
kick
around
ideas
for
an
NC
d
replacement.
I
that'd
be
a
fascinating
meaning
whether
it
was
sometimes
those
kinds
of
meetings
are
best
on
face
to
face.
In
my.
F
Though
this
meeting
has
never
officially
taken
place
and
in
reality
we
have
canonical,
has
a
distributed
database
that
is
embedded
and
it's
based
on
SQLite
called
TQ
light
and
the
kiiiers
guys
the
answer.
The
answer
guys
created
and
adopted,
for
that
would
basically
seem
etcb
and
it
has
a
back
end
for
DQ
light
and
they
call
it
kind.
So
these
I,
these
ideas
come
actually
actually
evolved
organically
and
the
collaboration
is
there.
It's
just
that.
It's.
C
To
use
one
of
the
contributor
summit
events
and
those
things
are
set
up
as
a
pre
event
to
a
cube
con,
and
you
would
typically
have
representatives
there
from
the
kubernetes
architecture
sig
and
those
are
the
people
ultimately
who
would
have
to
sign
off
on
that
being
accepted
into
the
project
as
a
hall.
But
the
contributor
summit
has
I
think
the
one
for
Europe
is
is
still
not
locked
down
so
that
you
may
still
be
able
to
submit
meetings
there
and
managed
to
put
that
on
the
agenda.
C
If
you
wanted
to
try
to
get
that
as
soon
as
cube
con
Europe,
which
is
last
week
of
March,
they
tend
to
have
enough
of
a
quorum,
they're
I,
think
at
the
contributor
summits
in
Europe
and
North
America
to
actually
get
things
done,
I'm,
not
sure,
historically,
that
they've
had
enough
people
show
up
at
Q,
Khan
China.
Events
to
you
know
to
have
a
reasonable
quorum,
but
you
might
want
to
try
to
put
that
on
the
contributor
summit
agenda.
I,
don't
know
if
you're,
what
are?