►
From YouTube: CNCF Serverless WG Meeting - 2019-02-07
Description
Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
A
A
A
A
C
A
A
A
A
D
D
D
It's
like
an
e-commerce
Factory,
see
so
there's
the
factory
services
that
send
over
cloud
events
to
the
warehouse
inventory
service
and
those
get
like
added
to
the
inventory
and
then
the
there's
a
store
you
I
hopefully,
and
then
people
can
buy
stuff
through
that
UI.
It
gets
reserved
the
inventory
and
then
sent
to
a
delivery
service
where,
like
maybe
some
fake
updates,
get
sent
back
to
a
delivery
UI.
C
D
And
so
potentially
like
one
of
the
factories
could
be
like
a
cookie
clicker
UI
like
thing
where
you
actually
tell
the
audience:
hey
go
to
this
URL
and,
like
click
on
my
cookie
and
then
every
click
turns
into
a
cloud
event
that
goes
to
the
warehouse
that
adds
to
the
UI
so
that
people
can
like
consume
a
cookie.
On
the
other
side,
like.
D
No,
not
a
real
cookie
like
there
was
a
game
or
like
the
cow,
clicker
thing
right.
It's
just
like
a
busy
task
that
you've
made.
They
do
right,
maybe
sit
on
their
phone,
they
click
the
thing
and
like
maybe
we
have
a
display
that
see
showing
the
active
inventory
counts
anyway.
It
it's
I'm
trying
to
make
it
interactive,
it
may
or
may
not
work
conference.
Wi-Fi
is
kind
of
spotty,
but.
A
A
Always
good
to
refresh
so
thank
you.
Alright,
any
questions
for
Scott
before
you
move
on
okay
yep.
So
please
join
the
call
at
1:00
p.m.
Eastern
on
Monday.
If
you
can
be
you
planning,
so
we
have
the
two
sessions.
We
talked
about
the
intro
and
deep
dive.
We
have
a
call
right
after
this
one
to
discuss.
You
know
what
we're
going
to
talk
about,
but
we
got
a
presenter
who's
going
to
be
presenting
stuff
like
that.
A
A
F
A
couple
of
weeks
or
months
ago,
I
made
a
change
where
I
renamed
your
eye
to
your
eye
reference,
because
this
is
only
the
placeholder
that
we
use
in
this
spec.
It
already
said
this
is
a
URI
reference,
but
the
shorthand
we
used
was
your
eye
and
I
was
pretty
confusing,
so
I
renamed
that
so
one
of
the
mistakes
I
did,
if
you
scroll
down
to
the
spec
markdown,
is
that
I
missed
the
schema
URL,
so
I
forgot
to
rename
that
as
well.
F
Whichever
way
you
want
to
look
at
it,
because
your
eye
reference
is
a
subset
of
the
RFC
and
it's
not
so
clear,
so
I
just
removed
that
part,
and
if
you
want
to
know
what
exactly
the
string
should
look
like,
you
can
go
to
the
time
definition
of
the
URI
reference
I
changed
that
in
the
spec
Jason
as
well.
Hobbie
introduced
this
last
time
and
then
in
their
proton
off
I.
Also
changed
the
shorthand
because
I
think
they're
protobuf
PR
did
kind
of
go
in
parallel
with
mine,
yeah.
A
G
H
F
H
F
H
F
G
C
C
That's
true
like
it.
Try
it
try
it
using
it,
for
example,
so
I'm
tempted
to
use
an
honor
system
here
and
just
assume
that
people
are
being
good-faith
actors.
If
people
are
more
cynical
than
I
am
which
I'm
very
open
to,
then
we
can.
We
can
try
to
like
brainstorm
on
what,
like
what
balances
we
want,
people
to
be
able
to
like
embrace
suspect,
no
matter
what
technology
they're,
using
with
also
being
cynical
about
what
people
will
say
about
their
specs.
So
I'm
really
open
to
comment.
C
If
people
like
my
basis,
if
no
one
else
leaves
comments,
I
might
say
like
I
might
try
to
pull
in
some
of
this
wording
here
into
my
spec,
but
I
not
inclined
to
like
to
raise
the
bar
too
much
more,
because
I
want
people
to
just
like
say
we
support
this
spec
I
went
I
want
people
to
like
be
able
to
like
assent
to
it
pretty
easily.
I
want
that
bar
to
be
low.
A
I
Will
be
concrete
because
it's
right
in
here
right,
so
rockin
mq
is
being
submitted
as
an
incubator
project
in
just
Apache.
If
you
look
at
the,
if
you
look
at
the
actual
contribution,
it
comes
from
one
place
and
then
certainly
and
then
shortly
thereafter,
I,
don't
know
exactly
what
the
what
the
the
order
is
open
messaging
starts
to
exist
in
the
Linux
Foundation
coming
from
the
same
place
and
except
for
a
benchmarking
project,
also
has
no
contributors
outside
of
that
place.
Wait.
I
That's
literally,
the
protocol
is
the
code
right
and
then
and
then
there
is
an
open
alternatives
that
a
broad
number
of
company
had
companies
have
been
investing
in
and
are
currently
investing
in
with
a
common
protocol
stack
that
has
quite
a
bit
of
traction
and
that
stands
against
that,
and
then
we're
doing
an
internal
project
and
and
effectively
the
game
that's
being
played,
is
obviously
trying
to
wedge
that
proprietary
product.
It's
who
something
is
looking
like
open,
even
though
nobody
supporting
that
open
thing
except
themselves.
C
Means
so
like
we
can
say,
like
they're,
playing
a
game
or
they're
playing
politics,
but
like
I,
don't
understand
what
the
end
result
is
so,
like
is
corporate
capture
open
source
totally
a
thing
totally.
It's
absolutely
a
thing.
That's
definitely
a
thing.
It
makes
for
crappy
open
source
I'm
as
guilty
of
it.
As
anybody
I
totally
agree.
That's
a
thing
really
I,
don't
understand
what
like
they're
playing
the
game
like
what
game?
Are
they
playing
they're,
open
sourcing,
something
and
they're?
Putting
the
work
in
and
like.
I
The
part
so
I'm
I
have
said
about
someone
trying
to
play
a
intro
via
API
story.
That
is
exactly
opposed
to
interrupt
on
the
wire
which
we're
trying
to
achieve
here
and
and
then
what
and
then
effectively
using
writing
a
check
to
legitimize
that
by
by
having
a
project
created
for
them.
That
then
says
open
well
effectively
is
a
proprietary
play.
I
C
I
still
totally
understand,
it's
not
necessary
that
I
understand
honestly,
so
I
can
like
I
can
drop
it,
but
like
just
to
let
you
know
why
I'm
still
skeptical
about
this.
It's
that,
like
so,
if
so
say,
I'm
I'm,
putting
myself
in
their
place
right.
I,
like
I,
have
built
this
thing,
and
now
I
want
to
open
source
it,
and
like
is
anyone
else
using
it.
C
No,
but
I'm
still
really
excited
about
it
and
I
want
to
put
it
out
there
like
I,
have
I
have
like
gained
very
little
like
by
by
like
getting
it
like
I,
don't
understand
like
how
they
are
winning
something
right
like
they
have
put
in
the
plate.
They
have
invested
in
this
thing
and
they
put
it
out
there
for
others
to
use.
C
What
do
they
get
out
of
it
because,
like
I
I
would
be
skeptical
if,
like
they
were
like
secretly
getting
tons
of
like
hidden
benefits
but
they're
not
like
they
are
they
just
like
put
something
out
there
in
the
world
that
will
sit
on
github
and
maybe
no
one
will
use,
which
is
like
the
thing
that
can
happen.
I,
don't
feel
that
bad
about
it,
but.
I
C
Question
is
not,
should
there
be
a
product
interoperates,
it's
like
who
is
responsible
for
figuring
out
if
it
interoperates
and
I
want
to
say,
that's
not
our
job.
I
want
to
say
like
if
you
claim
to
interoperate,
then
you're
responsible
for
actually
in
there
operating
and
like
we're
not
like.
We
don't
have
to
validate
that.
So.
J
K
I
Yeah,
however,
they
end
up.
If
the
question
is
the
question
for
me
is:
do
we
need
to
have
the
so?
We
can
have
a
catalog
of
Optronics
and
support
cloud
events,
and
we
can
point
them
in
that
kind
of
mode.
We
can
point
to
the
respective
front
of
documentation.
I
think
that's.
That
would
be
totally
fair,
but
that's
a
proper
spec.
You
really
the
only
need
if,
if
the
spec
can
really
be
implemented
by
another
party
and
for
the
four
different
protocols
that
are
in
queue,
proprietary.
I
For
those
four,
there
is
no
product,
there's
a
there,
so
there's
no
spec,
and
we
have
to
know
that
because
that's
also,
we
have
a
license
to
one
of
their
phone
calls
and
there
isn't
it's
not
the
only
way
you
can
go
and
do
this
is
by
reverse
engineering
and
listening
for
the
stuff
on
the
wire,
the
documentation
at
the
end
of
the
code
for
it.
So
the
situation
for
them
is
exactly
the
same
as
when
working
in
queue
you
can
try.
A
My
hands
up
next
so
question
for
you,
lemons,
I'm,
trying
to
wrap
my
head
around
this
and
then
see
try
to
see
both
sides
or
understand
both
sides
I'm
wondering
Clemens.
If,
if
your
concern
is
more
focused
on
whether
someone
can
then
run
around
and
claim
that
we
are
endorsing
them
in
some
way,
well,
okay,
in
the
sense
that
it's
not
so
much
Oh
such
a
such
protocols,
saying
hey,
we're
clatter
than
compliant
or
worried.
You
know
we're.
We
support
cloud
events,
I,
don't
I,
don't
get
since
that's
what
you're
worried
about!
A
I
I
Interoperability
down
the
stack,
all
right,
TCP
and
then
overlay,
the
the
framing
protocol
made
that
be
HTTP
or
in
deteriorate
or
computed
an
overlaid
on
top
of
that
since
proud
events
with
the
respective
encoding
set
cetera
but
but
I
believe
interoperability
from
the
ground
up
and
not
having
you
know
in
in
drop,
non-interoperable
thing
read:
that's
it's!
The
middleman
prevents
that
you
can
even
benefit
from
bother.
That's
so.
A
It's
not
like
you're
saying
if
we
don't
nest
their
documents
here,
but
we
have
maybe
a
list
of
pointers
to
documents
host
at
other
places
and
say:
oh
by
the
way.
Here
is
some
other
format,
or
here
are
some
transport
bindings
or
whatever
that
that
do
support
cloud
events
and
we
make
no
claim
as
Dora
they're
good
bad
or
anything,
and
it's
just
a
list
you'd
be
more
okay
with
that
it
sounds
like
yeah.
F
I
think
I'm
a
bit
less
religious
than
Clements
I'm
more
from
a
practical
perspective,
so
the
thing
like
I'd
I,
don't
know
about
rocket
MQ
or
open
messaging
or
whatever,
like.
If
a
customer
comes
to
me
and
says,
I
want
to
integrate
your
the
platform
that
we're
building
with
rocket
I'm
cute
I'm
generally
in
support
of
that
and
my
sort
of
dream
is
that
I
have
an
implementation
for
cloud
events.
F
There
is
maybe
something
in
between
that
I
have
to
set
up
like
an
SDK
or
whatever,
but
that
basically
it's
plug-and-play,
maybe
I
have
to
plug
something
in
the
middle,
but
then
they
can
just
implement
well,
they
can
take
our
platform.
We
have
we
amid
cloud
events
and
somehow
they
end
up
in
rocket
and
cute.
L
L
L
Interoperability
is
obviously
a
critical
piece
for
a
lot
of
people
right,
so
you
know
where?
Where
is
that
going
to
fall?
So
where
do
we?
Where
do
we
draw
those
kinds
of
lines?
And
what's
the
seal
of
approval
right
or
Interop
cross
usage
versus
okay?
We
have
an
adapter
that
allows
us
to
you
know
hook
these
things
together
in
this
more
proprietary
hodgepodge
way,
and
maybe
that
solves
problems
for
certain
users,
but
it
solves
it
in
a
way
that
is
not
generic
enough
it'd
be
to
claim
true
Interop
that
my
son
I.
A
He'll
hand
up
to,
let
me
ask
this
question
of
Rachel
the
goals
that
you're
trying
to
achieve
Rachel.
How
important
is
it
that
you
that
they
actually
hosts
their
documents
in
our
repo
as
a
person?
Is
it
okay.
C
C
The
difference
is
not
that,
like
we're,
not
charging
for
our
products
right
so
I'd
like
I,
don't
like
I,
don't
see
this
as
a
stamp
of
approval.
I
see
this
as
a
way
to
gain
adoption
and
and
the
important
the
reason
it's
important
that
it
go
in
the
spec.
Is
that
I
want
people
it's
a
place
to
standardize
it
it's
a
place
to
say
like?
Are
you
interested
in
cloud
events?
Are
you
running
any
of
these
technologies?
Here's
where
you
go
to
look
to
see
like
what
the
like
what
the
standard
like
serialization
is
like.
C
A
You
put
out
a
pseudo
proposal
in
there
of.
Is
there
an
appropriate
wording?
We
can
come
up
with
to
to
say
you
know
your
spec
team
presents
and
disrepair
or
someplace
in
our
org
does
not
give
you
the
right
to
claim
compliance,
use
our
logo
or
anything
like
that.
What
do
people
think
about
hitting
that
direction?.
I
I
Yes,
so
this
so
for
me,
the
higher
and
the
highest
order
of
business,
driving
interoperability
and
and
for
me,
interoperability,
is
the
thing
that
that
ought
to
be
driving
the
adoption,
like
the
desire
to
have
a
common
stack
that
has
a
minimal
amount
of
options,
but
the
options
that
are
appropriate
for
use
cases
and
then
go
and
use
those
rather
than
you
know.
Adoption
of
this
specification
at
all
costs
because
I
don't
know,
I,
don't
know
how
much
that
helps
us.
I
C
C
The
like
the
first
way
I'm
gonna,
get
started
using.
This
is
I'm
going
to
like
write
a
shim
that,
like
converts
like
at
the
like
at
the
edge
I
just
convert
the
thing
like
I,
don't
change
any
of
my
internals
and
I
and
I
like
right
assume
at
the
edge,
so
it
like
converts
whatever
I'm
doing
to
a
cloud
of
it
right,
and
this
is
like
in
this
proprietary
like
protocol
and
encoding
like
folder,
is
the
place
you.
I
C
And
I,
just
wanna,
like
I,
agree
with
all
that,
and
the
thing
that
I
just
want
to
add
is
like
the
only
like
the
only
specs
that
should
start
using
this
should
not
be
ones
they're
owned
by
the
Linux
Foundation
right,
like
any
like
any
protocol
that
people
are
using
like
should
have
a
way
to
like
start
using
cloud.
Events
like
you
shouldn't
have
to
like
you
shouldn't
have
to
change
what
you're
doing
you
should
just
be
able
to
like
start
layering
on
a
cloud
of
it.
That's
like
that's
like
what
I'm
coaching
for
yeah.
I
And
I
think
I
think
where
I'm,
where
I'm
at
and
that's
this
disagreement,
I
think
to
make
to
to
make
the
complexity
matrix
of
integration,
and
we
have
some
people
with
it
with
enterprise
integration
background
here.
Is
you
can't
make
the
complexity
matrix
like
you
can't
have
50
protocols
to
choose
from
and
then
everybody
is
is.
Is
you
know
one
is
seeding
this
protocol
and
the
other
one
is
pulling
from
this
protocol
because
then
then,
you're
really
not
compliant
to
anybody.
I
Who
said
then
you
know
you
need
to
have
the
matching
products
which
should
happen
to
speak.
Those
can
protocol
and
that's
why
you
narrow
the
charts
down-
and
you
say,
here's
three
protocols
pick
one
of
the
three
and
and
then,
and
that
makes
integration
more
real.
But
if
you
support
50
different
costs,
the
choice,
the
likelihood
if
you
find
two
products
which
actually
speak
both
because
become
slimmer-
and
that's
my
that's
my
the
core.
My
concern.
C
Okay,
I
think
we're
probably
at
our
time
box,
but
I
would
really
I'm
interested
in
getting
this
like
I'm
happy
to
compromise
on
this
PR,
but
I'm
like
the
thing
that
I
would
most
like,
is
to
get
a
version
of
this
like
with
whatever
caveats
we
need
like
like.
If
I
have
a
proprietary
protocol,
I
want
to
be
able
to
say
here
is
how
I'm
supporting
cloud
events,
and
if
you
want
to
do
it
to
here's,
how,
like
that's
what
I
want
yeah.
A
C
A
I
Think
a
catalogue
of
implementations,
which
you
know,
support,
support,
cloudbeds
I,
think
that's,
okay,
I'm
just
I
mean
I
am
I,
just
I
would
like
to
spec
set
and
what
looks
like
technical
documents
to
be
constrained
to
those
which
which
drive
interoperability
across
products,
and
so
let
me
yeah
I'll.
Let
me
comment
on.
Let
me
comment
on
this
and
then
basically
summarize
summarize
my
my
stance
here.
I,
don't
care
about
us
going
together
to
Plouffe?
A
Over
time,
so
let's
do
this
Clemmons
if
you
could
put
a
comment
or
proposal
or
something
in
there
to
because
right
now
it
appears
Rachel's
proposal
and
then
we
have
the
other.
The
other
side
is
basically
saying:
do
nothing
and
I
actually
think
there
may
be
something
in
the
middle
there.
So
maybe,
if
you
could
write
way,
your
view
of
a
middle
ground
position
would
be
able
to
get
some
discussion
coin.
Maybe
circle
around
the
real
answer
at
some
point
can.
C
I
can
I
also
ask
like
and
like
Clemens
and
I
are
dominating
the
conversation,
but
I
would
also
like,
if
you
just
like,
have
an
opinion
on
this
and
like
I,
would
appreciate
knowing
like
what
the
temperature
of
the
room
is.
So
if
you
feel
so
moved
by
the
spirit,
I
would
love
for
you
to
like
add
a
comment
on
the
pie
or
just
like
I,
don't
know.
Let
me
know
somehow
cuz
like
if
I'm,
what
like
out
on
the
deep
end,
let
me
know
feel
like
that's
what
I'm.
That's
what
I
think?
Yes,.
A
F
Okay,
I
won't
go
through
the
whole
background
of
it.
Basically,
last
time
we
ended
up
saying-
or
we
were
discussing-
if
we
should
have
a
hard
must
or
basically
ended
up
in
two
places,
what
should
the
actual
size
be
and
then
the
other
one
should
we
have
a
must
or
should,
because
we
we
just
cover
some
places
where
maybe
a
must
is
too
harsh.
Let's
say
you
have
an
IOT
or
an
edge
device
that
doesn't
have
enough
bandwidth
or
a
memory
or
is
constrained
in
some
other
way.
F
I
then
proposed
a
third
option,
which
is
say
well
basically
for
me
the
point
is
well:
some
people
will
ride
a
middleware
and
some
people
will
write
the
end
consumers.
What
I
want
is
that
all
middle-earth
sort
of
work
on
a
consistent
way
and
and
if
someone
as
an
end
consumer,
goes
and
drop,
something
it's
maybe
their
own
fault.
F
F
So
that's
why
I
proposed
the
third
option
that
says:
okay,
you
must
support
it,
except
when
you
are
a
consumer
that
is
heavily
restrained.
So
basically
everyone
who
writes
a
middleware
that
runs
in
the
cloud
has
to
accept
these
limits.
But
if
you're
a
consumer-
and
you
happen
to
have
these
constraints,
then
you
don't,
it's
very
ended
up.
Yeah,
ok,.
A
And
there
were
some
comments
put
into
the
PR.
Since
then
we
had
one
person
vote
for
your
new
option,
a
couple
people
I,
think
member
four
or
so
voted
for
option
two,
which
is
changed.
The
must
do
it
should
and
Kathy
his
team
like
she
can
go
the
other
of
the
first
or
the
second.
Anybody
wants
to
chime
in
on
the
call
now
to
add
their
opinion.
A
A
A
Okay,
I'll
speak
up
here
personally,
I
would
prefer
option
two
as
well.
Only
because
I
think
the
third
option
is
it's
fine
other
than
I
think
if
these
same
results
with
the
should
or
strongly
recommend
it
asking
if
I
preferred,
but
either
way,
I
I,
don't
like
any
of
a
must
with
it
out,
because
that's
me
means
well,
why
not
just
make
it
a
should
or
strongly
recommended,
but
that
was
just
my
take
on
it
and
we
also
have
an
opinion.
A
P
May
change
it
because
you
know
once
we
process
ice
there
I
think
you
know
if
we
put
a
size
that
can
be
applied
to
more
yours
cases
or
more,
our
component
I'm
een,
more
even
consumers,
I
think
they'll
be
good.
People
are
trying
to.
He
made
sure
people
would
like
to
be
compliant
with
that
size,
but.
A
P
Yes,
I
think
you
know
it's
yes,
I
think
we
can
give
it
a
smaller
size
and
then
put
sure
I.
Think
more.
F
I'm,
okay,
to
compromise
on
a
shirt
or
strongly
recommend
the
only
thing
I
really
want
to
have
is
that
sort
of
everyone
in
this
group
then
agrees
to
honor
that
limit.
So,
basically,
then
I
would
also
want
to
pick
the
64
kilobytes
if
Clemens
or
event
grid
basically
wants
to
keep
that
limit
on
their
side
that
we
should
have
that
limit.
Sort
of
that.
Everyone
in
the
group
agrees
to
support
with
the
products
that
we
build
here.
A
A
A
Q
A
A
A
Okay,
so
you'll
get
that
on
there.
Okay
revisit,
oh,
maybe
vote
I
should
say
who
don't
know
for
sure
tell
what
those
reaction
to
it
is
all
right:
cool,
okay,
let's
go
back
to
the
claim
check.
One
did
you
did
you.
E
F
F
I
F
Well,
that
is
not
if
I
may
respond.
That
is
not
the
intention.
The
attention
is,
you
have
a
data
object
which
then
in
turn
may
contain
other
URI
references
or
your
eyes,
but
here
is
like
the
main
outer
date.
Data
object
does
not
fit
within
your
message
and
then
instead
putting
it
into
the
data
rack.
So
then,
if
you
catch
that,
so
you
can
kind
of
replace
both
either
you
have
the
data.
I
Yeah
I
think
I
think
my
my
point
is
it
is.
It
is
just
as
easy
to
go
and
put
a
your
I
into
data
and
your
I
update
into
data
that
in
having
a
first-class
construct
for
this,
like
it,
doesn't
necessarily
make
a
lot
of
a
lot
of
sub-areas
easier
to
force
everybody
to
implement
that
evil
net.
That
pattern
so.
K
That
Clemens
had
touched
on
last
time
was
the
fact
that
you
know
just
having
a
reference
to
something
could
be
potentially
dangerous.
If
you
don't
know
the
size
of
the
thing
you're
pointing
out
or
maybe
what
it
is
yeah
so
I
I
understood
that
concern
I
again,
you
know
this
seems
to
be
double
whammy
for
me,
because
I
think
I
was
one
of
the
original
instigators
for
this
as
well
again.
K
K
You
know,
work
around
that
limit
if
they
have
to,
and
so
that
was
really
again.
My
only
drive
was
to
say,
let's
just
show
people
a
way
to
do
it,
rather
than
e
everybody
coming
up
with
their
every
implementation
coming
up
with
their
own
way
to
do
it,
so
it
was
more
of
a
design
pattern
which
was
then
codified
right.
A
A
A
I
One
of
the
so
a
reference
to
an
external
reference
to
a
lot
to
a
large
object.
You
would
you
would
point
to
could
be,
it
could
be
manifold
right,
it
could
be.
It
could
be
that
you're
pointing
to
a
large
graph
that
sits
in
some
in
some
document
database.
If
you
wanted
one
reference
and
where
you
speak
a
particular
protocol
to
get
at
it.
Let's
say
you
point
into
a
record
in
mongrel
database
and
it's
impossible
to
go
and
send
an
entire
record
of
the
bungle
database
with
all
the
dependent
dependencies
with
it.
I
So
you
may
see
point
to
it
or
you
have
a
large
file
or
you
have
a
video
or
you
have
I
mean
there's
all
kinds
of
duck
because
you
can
have
and
you
going
to
further
qualify
them
and
how
you
qualify
them
and
probably
how
we
get
further
ends
like
the
size
of
the
document,
etc
might
differ.
Based
of
the
kind
of
document
that
you
have
so
I'm
I'm
I.
I
Have
a
fit
I
also
have
like
you
said
it.
Fidelity
concern
around
that
lake
in
terms
of
how
much
metadata
you
further
need
to
make
sense
out
of
that.
Out
of
that
reference,
and
and
that
this
might
not
be
the
only
this
might
not
be,
the
only
link
that's
efficient
could
go
and
express
that
basically
giving
a
client
that
receiver
that
received
an
event
a
choice
of
of
what
they
might
of
what
they
might
want.
I
Like
say,
let's
say
you
have
an
image
database
in
a
store
or
a
new
picture
right
and
you
send
them
an
event.
The
event
is
about
the
picture
and
ultimately
want
to
have
the
picture,
but
maybe
in
the
data
you
want
to
give
them
the
choice
of
ending.
You
cut
me
off
or
the
original
resolution,
or
a
choice
of
thumbnails,
and
so
all
of
a
sudden,
you
point
in
to
four
versions
of
the
same
or
ten
versions
of
the
same
picture.
I
J
I
That's
that's
kind
of
where
I'm
coming
from,
because
that's
the
principle
that
we
have
in
general
in
our
product
is
to
always
point
to
stuff
and
never
and
rarely
carry
the
stuff.
So
we'll
always
have
these
references
in
line
in
the
in
the
data
and
then
leave
it
to
the
application
will
interpret,
interpret
them
according
and
here
this
seems
like
good,
the
most
rudimentary
case
specialized.
I
F
Let
me,
as
a
I,
think
that
this
Funday
example
doesn't
really
well
will
be
a
misuse
of
this.
Your
I
reference.
That's
not
what
I
would
like
to
use
it,
for
that
is
my
point.
So
I
think
there
are.
There
are
Ella
use
cases
so,
for
example,
a
doe
yes
for
SKS
they
have
a
limit
or
they
have
built
in
there,
at
least
in
our
Java
SDK
away.
For
that
you
in
the
Java,
SDK
can
say
I'm
sending
this
message
and
then
someone
else
with
the
same
SDK
can
consume
this
message.
F
So
if
you're
sending
a
message
below
that
limit,
it
will
be
put
directly
into
ask
us
if
you're
sending
a
message
above
that
limit
it
will
actually
put
into
as
free
and
then
the
pointer
to
that
as
free
object
is
put
into
the
message
and
then
the
SDK
on
the
other
side
will
pull
out
that
data
from
me
as
as
free
blob
storage.
So
the
good
thing
is
as
a
wolf
as
a
producer
and
as
a
consumer.
F
You
really
don't
know
if
that
happens
in
the
background
or
not,
so
that
is
I
think
something
that
that
is
worth
considering
having
I'm,
not
sure
how
many
people
run
into
this
particular
issue.
For
me,
that
is
in
my
daily
work.
This
is
often
a
case
where
I
really
don't
or
let's
put
it
this
way,
I
already
point
to
the
object
itself
by
having
that
source.
But
what
I'm
sending
in
a
message
is
often
a
change
set
and
it
can
be
very
small
or
it
can
be
very
big
even
for
the
same
event
type.
F
P
I
think
this
SPI
is
talking
about
out
of
time,
I'm
beta
right
now
in
band,
so
I
I
think
you
know
it's
a
good
thing
to
have
that,
because
if
you
want
listener
either
large
large
set
of
data
you
want
to
not
to
send
it
in
band
or
restriction,
does
not
allow
you
to
send
it
in
BAM,
so
this
could
have
a
data
reference,
but
whether
we
have
you
know
one
URI,
URI
or
zero.
That's
another
thing
we
can
discuss.
I,
think
you
know
there
are
use
cases.
A
A
A
All
right,
Sonia
he's
still
there
no
Sonia
to
leave
it.
What
about
Alex?
Okay?
Is
anybody
I
missed
for
roll
call?
Okay,
thank
you
guys
very
much.
Please
do
take
time
to
comments
on
those
how
sanity
ours,
the
way
that
you
guys
are
free
to
go,
except
for
the
people
who
would
like
to
potentially
present
at
the
next
koukin.
So
those
guys
please
stay
on
the
call
I'll
just
give
people
a
second.
H
A
A
Let's
see
whoops
okay,
so
who
should
be
here?
Let's
see
what
you
know,
Clements
is
here
Scott's
here
Christophe
still
here.
Well,
that's
still
here
Chad
his
mother,
neither
called
Dan
it's
later,
okay,
hear
what
he
was
supposed
to
be
here
is
Harry
good,
so
we
can
get
started
all
right.
Mr.
Scott,
if
you
don't
mind,
I'm
gonna
lean
on
you
to
drive
more
of
this
discussion
since
this
baby's.
Your
idea
in
terms
of
how
you
want
to
proceed
this.
A
Sorry,
what
are
we
talking?
Oh
gosh
darn.
It
I'm
mixing
up
things.
Okay,
you're
right
turn
it
okay.
So
in
that
case,
presentations,
okay,
so
donÃt
you
go.
A
J
A
Okay,
so
let's
focus
on
the
intro
first
I
think,
based
upon
the
description
of
the
way
of
the
way
intro
calls
are
supposed
to
bar
the
intro
sessions
are
supposed
to
be.
We
have
to
at
least
give
the
generic
one
that
we
give
almost
every
time,
which
is
what
is
caught
events.
Why
are
we
here?
Why
we
matter
that
kind
of
stuff?
A
It
does
not
necessary
long
session,
but
you
think
we
have
to
at
least
give
that
brief
intro
for
people
who
don't
know
what
the
heck
it
is,
but
after
that's
were
then
free
to
fill
the
time
whatever
we
want.
So
as
a
follow-on
for
intro
session,
do
you
get?
What
would
you
guys
like
to
see
the
next
topic
or
a
set
of
topics,
be.
D
Maybe
like
the
hello
world
of
how
to
use
it
like
how
to
send
an
event.
Okay,
so.
A
I
I
I
R
Might
have
something
present
on
this,
but
I
don't
know
if
we'll
be
ready
before
then
like
we
becoming
currently
working
with
AI,
implement
the
whole
pipeline
for
all
20
deployments
and
the
whole
internal
dev
tools
will
be
using
all
events.
Well,
the
main
driver
for
this
is
the
fact
that
we
have
to
send
trace
IDs
around
and
we
have
to
use
SNS,
sqs
and
stuff
like
that,
and
there
is
no
easy
way
to
use.
R
I
D
R
D
I
There's
there's
a
good
news
on
that
front
by
the
way
that
in
the
w3c
Aubry
traces
back
I
said
what
it's
called
yeah
I
think
so
that's
being
that's
being
referenced
from
there.
That
is
not
also
getting
them
QT
and
a
maybe
people,
I
think
so
we
can
go
online,
but
only
on
HTTP
what
we
can
also
align
on
the
that
they
QP
behind
it.
Nietzsche
yeah
I
know
that,
because
I
had
to
review
it.
A
D
I
There's
there's
the
don't
use
this.
Don't
use
this
to
do
a
general
messaging
aspect
there
as
well
right,
there's
like
patterns
and
to
enter
patterns
and
I.
Think
I
can
the
anti
pattern
section
is
pretty
long,
because
you
can
bet
that
people
will
come
and
make
their
reply
to
extension,
and
really
it
really
really
shouldn't
like
things
you
should
not
do.
Please
yeah.
F
I'm,
not
sure
god
I'm.
We
assume
that
the
people
who
are
there,
probably
don't
use
it
yet
and
may
have
only
heard
the
intro.
Does
it
really
make
sense
to
go
and
talk
to
like
okay?
This
is
the
stuff
that
doesn't
work,
and
you
know
or
focus
on
the
positive
examples
of
what
it
can
be
used
for
and
odd.
You
should
do
this
is
this
well,
then
they
don't
really
have
an
idea
and
they
had
what
they
should
do
with
it.
Yeah
I
think
you,
thanks
to
both
both.
A
D
M
A
Okay,
what
are
the
things
we
do
for
deep
dive?
We
only
have
35
minutes
and
glads
thing
or
whatever
we
come
up
with
whether
it's
ads
or
something
else
could
take
a
quite
a
while,
so
that
could
be
I
may
not
have
as
much
time
as
we
think
it
is.
This.
Are
these
two
high
level
topics
good
enough
for
a
deep
dive?
M
A
H
A
A
I
We've
done
it,
we've
done
the
demo
in
the
intro
in
the
last
time
and
I.
Think
the
demo
in
in
the
intro
makes
more
sense
to
at
least
show
well
the
defense.
What
the
demo
will
be,
but
kind
of
showing
the
demo
off
in
the
in
the
intro
is
a
good
cliffhanger
for
the
deep
dive
session,
because
everybody
wants
to
see
code
and
so
you'll
show
the
demo
over
there
and
say.
Well.
If
you
want
to
know
how
we
put
this
together,
yeah.
D
A
A
M
A
I
thought
you
had
a
really
good
starting
point
for
an
outline.
What
do
you
guys
want
to
do
in
terms
of
next
steps
here?
Go
off.
Think
about
this
with
a
bike.
Put
this
to
a
document
to
ask
you
guys
to
then
comment
offline
to
try
to
expand
upon
the
thoughts
or
we
can
talk
about
it.
Now
we
have
45
minutes
unless
you
guys
want
to
go
eat.
M
M
R
One
more
question
and
then
I
can
go,
get
pizza
or
so
no.
This
is
very
cloudy.
Events
focus
and,
at
the
last
cue
corn
in
Europe,
I,
really
liked
how
there
was
also
China.
How
are
you
using
called
events?
What
issues
do
you
have
and
general
talking
about
this
kind
of
stuff?
Would
it
be
worth
to
have
something
like
this?
She
just
yet
where
the
community
is
with
it
and
see
if
we
need
any
other
working
groups
or
stuff
like
that,.
R
A
Now
I
I
think
you're
planning
on
having
a
service
day.
I
can't
believe
that
term
they
use
for
it,
I'm
wondering
if
it,
if
it's
better,
to
save
that
discussion
for
that
thing
or
take
a
part
of
our
time
in
cloud
of
ensk,
because
it's
a
bit
of
a
shame
that
well
actually
they
back
up.
We
originally
were
given
the
option
of
having
a
an
intro
in
a
deep
dive
for
both
cloud
events
and
service.
Because
technically
it's
two
separate
groups.
R
S
A
A
There's
no
question
I
I
kind
of
view
this
as
sort
of
follow-on
for
the
working
for
the
surrealist
working
group
itself
to
say:
okay,
look
we
put
up
this
white
paper,
we
can
talk
about
where
we
see
the
state
of
service
in
general,
and
you
know
talk
about
all
the
various
service
products
to
set
on
top
of
kubernetes.
You
know
open
faz
kami.
They
probably
do
those
things
and
then
sort
of
turned
into
like
a
birds
of
a
feather
kind
of
thing
that
someone
mentioned
and
just
get
the
audience
engaged
and
say.
A
Okay,
where
do
you
guys
think
services
should
go?
We
do
you
think
the
service
working
group
should
tackle
another
topic
right
because
well
because
we
talked
about
you
know,
what's
the
next
thing
be
and
we
picked
workflow,
but
that's
not
getting
a
whole
lot
of
traction,
so
maybe
there's
just
something
else
we
should
be
focusing
on
and
maybe
we
need
input
from
the
broader
community,
so
we're
not
just
talking
to
ourselves.
I,
don't
know
that
was
my
initial
thought
when,
when
you
guys
started
hating
on
this
path,
well,.
A
D
A
L
M
E
B
A
D
A
A
D
A
A
So
what
I'd
like
to
do
is
see
if
I
can
get
agreement
to
at
least
play
your
name
on
one
thing:
now:
it's
okay,
so
you
can
not
lie
to
your
management
team
and
say:
yes,
you
are,
you
will
be
talking
yeah
right
and
glad
you
sound
like
you
were
going
to
do
something
here
or
you're.
Gonna
really
really
try
what,
if
we
tentatively
Scott
put
you
on
the
hook
for
in
essence,
part
two
of
the
intro
to
cover
these
things.
Okay,.