►
From YouTube: CNCF Serverless WG Meeting - 2018-10-25
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
Thank
You
sy
to
get
your
attendance
in
there
all
right.
Excuse
me
all
right,
so,
let's
jump
right
into
it.
I,
don't
think,
is
anything
necessary
to
talk
about.
The
AI
is
Austin
sent
on
the
car
call
I'm
gonna
ping,
yet
postured
offline
to
see
what
he
wants
to
go
out.
The
next
SDK
call
because
I
think
it's
been
a
while,
since
we
had
one
and
I
think
we
need
another
one,
all
right
community
time.
A
So
this
is
an
opportunity
for
people
who
may
not
be
normal
people
on
the
call
or
no
visitors
to
the
call
to
bring
up
any
topics
for
discussion.
Is
there
anybody
on
the
call
who
would
like
to
bring
up
a
topic
for
discussion
who
is
in
normal
here
all
right
cool
moving
forward?
Then
okay,
Austin
I,
hear,
but
we
haven't
had
the
SDK
workgroup
meetings
I'm
there
to
talk
about
Cathy.
Is
there
anything
from
the
workflow
subgroup
that
you'd
like
to
mention?
A
No
thank
you.
Okay,
koukin,
Shanghai,
clements,
Cathy
and
I
are
still
working
on
the
charts.
There
have
been
some
updates
since
last
week,
but
I
know
based
upon
what
I
saw
about
an
hour
ago,
there's
still
more
work.
That
needs
to
be
done,
but
obviously
anybody's
free
to
take
a
look
at
the
current
version
of
the
slides.
You
like
the
link,
is
right
here
in
the
agenda.
Any
suggestions
or
feedback
is
welcome.
We
also
a
phone
call
immediately
after
this
one
to
discuss
it
if
they
might
wants
to
join
you're.
A
A
It
was
a
very
exciting
conversation,
though
I
just
let
you
know,
but
so
I'm
gonna
probably
set
up
another
phone
call
to
discuss
my
Interop
at
some
point,
but
in
the
meantime,
if
you're
interested
in
the
interrupt
demo,
please
go
look
at
the
cloud
events
demo
slack
channel.
We
have
at
the
private
channel
so
you're
not
on
that.
Just
ping
me
and
I'll
invite
you,
but
I
would
like
to
try
to
get
some
offline
discussions
going
there
about
the
Interop
demo
in
particular
about
two
hours
ago.
A
I
think
I
pasted
a
rough
outline
for
if
the
flow
that
I
was
thinking
we
may
want
to
do
it
comes.
It
came
after
a
discussion
with
the
class
and
I
had
earlier
today.
So
if
you,
like
it
credit
class,
if
you
hate
it,
that's
put
the
parts
that
I
added,
but
please
review
it.
If
you
have
any
feedback,
let
us
know
I
do
want
to
get
that
going
relatively
soon.
A
The
scenario
that
were
kind
of
looking
at
is
not
terribly
hard,
so
it
should
be
easy
for
anybody
who
wants
to
participate
to
join
in,
but
I'm,
not
half
percent
sure
it
highlights
all
the
right
points
of
the
specs
themselves,
so
looking
for
some
feedback
there
or
ideas
for
how
to
to
beef
it
up
from
that
perspective,
okay,
any
questions
on
any
of
these
four
topics.
Before,
we
jump
into
PR
discussions.
A
All
right
looking
forward,
then
PRS
all
right.
Let
me
just
do
a
quick
double
check
and
see
if
there
any
new
votes
since
the
call
started.
I
think
Fujitsu
was
last
one
invoked:
okay,
okay,
so
the
votes
that
closed
last
time,
I,
checked,
I.
Think
it's
about
10
to
1
in
favor
of
321,
for
the
property
casing
vote
that
we
shared
last
week.
So
321
is
the
one
we're
going
to
go
with.
B
B
A
B
A
Okay,
okay
cool,
so
once
those
once
the
rebase
is
done
I'll
then
merge
that
one
in
so
thank
you
very
much
for
everybody
for
voting.
Let's
see,
move
forward,
the
pro
tip
of
format.
Pr
hasn't
been
updated
since
last
week
and
I
don't
think
there
have
been
too
many
new
comments
since
last
week,
either.
So
two
questions
here
or
two
things:
one,
please
everybody.
When
you
get
a
chance,
look
at
that
PR
I
know
it.
It
means
it's
an
important
PR
to
a
lot
of
people
out
there,
especially
obviously
the
protobuf
people.
A
So
please
look
at
that
when
you
get
a
chance,
the
more
eyes,
the
better
it's
actually
a
relatively
small
PR,
but
also
I'm
trying
to
see
who's
on
the
caller
so
Scott,
you
might
be
the
only
Googler
on
the
call.
Could
you
poke
Spencer?
For
me,
there
are
a
couple
of
comments
in
there
that
I'd
like
to
get
his
feedback
on
either
to
say:
I,
don't
like
your
suggestion
or
to
update
based
upon
the
suggestion
he
hasn't
had
a
chance
to
or
haven't
commented
on
the
on
the
comments
in
the
PR.
A
Thank
you
very
much
all
right
any
questions
or
comments
on
the
protobuf
PR
before
we
move
on
all
right,
keep
them
going,
then
the
burgeoning
scheme
we
briefly
talked
about
this
last
week.
I
was
wondering
if
anybody
had
any
new
hope
the
changes
here.
Any
new
comments,
I
know,
Eric.
You
had
some
comments
in
there
and
you're
suggesting
possibly
looking
ads.
What
was
its
scheme
of
work?
That's
another
alternative.
A
A
What
I
would
like
to
do
is
put
a
time
box
on
this,
so
what
I'd
like
to
do
is
push
for
either
a
decision
or
at
least
a
discussion
on
next
week's
call.
No
discussion
implies
I'm
going
to
ask
that
there's
any
objection
to
approving
it,
but
I
do
want
to
give
you
guys
least
one
more
week
to
review
it.
Does
that
sound,
fair
yeah?
That's
good.
E
E
E
A
E
E
A
Yeah
all
right
final
item
on
the
agenda:
it's
gonna
be
a
really
short
call.
I,
think
I
think
it
was
Klaus.
I
can't
apologize
if
again
the
name
wrong,
but
somebody
asked
me
earlier
about
when
we're
gonna
cut
a
zero
2.2
and
I
started
looking
at
our
mouse,
our
road
map
document.
Looking
at
the
zero
point
tool
items
we
had
here
and
I
actually
think
we're
pretty
much
set
for
actually
most
of
these
things,
we
did
the
learning.
A
We
have
some
draft
documentation
out
there,
users
guide,
I'm,
assuming
that's
sort
of
the
equivalent
of
a
primer.
You
guys
can't
really
think
about
that
one.
These
two
are
interesting,
so
I
don't
know
if
I
don't
know
about
any
any
large
design
decisions
that
are
looming,
at
least
that
ones
that
people
have.
You
know
actively
brought
up
or
any
required
attributes.
However,
what
I
was
gonna
do
was
I
was
gonna,
look
through
the
list
of
open
issues
to
see
if
any
of
those
at
the
categories
of
those
two
and
then
bring
those
up
for
discussion.
A
I
would
appreciate
another
set
of
eyes.
Looking
over
the
open
issues
to
see,
if
there's
some
out
there,
that
I
missed
during
that
review,
but
if
I'm
correct
and
we
don't
have
any
really
biggest
any
ones,
then
we
may
be
able
to
assume
those
two
points
are
behind
us
and
then
really
we
have
the
interrupt
on
what
you're
working
on
right.
Now
it
may
be
just
a
matter
of
saying:
do
we
think
we
have
the
initial
set
of
protocol
serialization
mappings
we
want,
but
anyway,
but
many
purpose
here.
A
Yet
what
are
the
outstanding
items
that
we
want
to
tackle
and
then
start,
you
know,
make
sure
those
are
the
on
the
agendas
for
upcoming
calls
and
discuss
those
persistence,
Eric,
yeah,
you're,
right
I
think
there
is
actually
an
issue
out
there.
I
don't
have
the
issue
number
offhand,
but
please
take
a
look
at
that
one
that
might
have
my
personal
point
of
view.
That
gets
it
a
little
bit
of
implementation
detail,
but
I
think
we
should
have
that
discussion
in
the
issue.
A
So
please
people
take
a
look
at
that
persistence
issue
since
Eric's,
specifically
mentioning
that
one
I
do
I'm
pretty
sure
there
is
an
issue
out
there
for
it
right.
Does
that
sound
for
everybody
we'll
try
to
have
a
discussion
next
week
on
where
we
think
we
are
relative
to
zero
point
to
not
hear
any
objection.
Are
there
any
other
comments
or
questions
relative
to
that
discussion?
You'll
want
to
bring
up
now.
A
Okay,
it's
gonna
be
a
really
short
call
guys.
Are
there?
Oh,
thank
you
for
the
link.
Eric.
Are
there
any
other
topics
or
PRS
people
want
to
bring
up
I.
Think
all
the
other
PRS
have
our
block
for
some
reason
in
particular
Fabio's,
even
though
it's
relatively
easy
PR,
it
came
in
I
think
just
yesterday,
so
it's
too
new
for
us
to
review.
H
About
the
version
zero
that
to
milestone
are
we
gonna
be
able
to
climb
that
with
the
with
a
very
early
battle,
VSP
key
word,
because
a
new
version
of
club
events
and
actual
very
early
at
the
game,
you're
miles
would
be
a
great
thing
like
people
could
actually
start
implementing
this
version,
0.1
had
actual
industry
interest
and
zero
that
he
would
come
with
some
again.
Even
very
rough
is
the
case.
It
would
be
amazing,
I
think.
A
That
is
a
wonderful
idea
and
that's
a
great
reason
to
have
a
discussion
or
have
an
SDK
workgroup
call
if
nothing
else
that
would
also
force
people
or
I
have
a
good
forcing
function,
because
I
would
love
it
if
we
could
do
0.2
in
time
for
coop
con
Seattle,
which
is
not
that
far
off,
but
it
should
be
I.
Think
there's
enough
time
for
people
to
at
least
get
some
beta
level
or
alpha
alpha
or
beta
level
of
their
SDKs.
I
know,
for
example,
go
he's
already
out
there,
but
I
think
it's
a
wonderful
idea.
A
Okay,
tell
you
what
I
as
a
reminder
to
myself
if
I
keep
I
said
to
be
forgetful
these
days?
Where
is
it
there?
It
is
okay,.
A
C
B
If
you
split
in
you
have
multiple
events
like
the
a
each
of
them
needs
to
have,
it
needs
to
have
a
separate
event
ID,
because
you're
going
to
route
them
differently
and
you
might,
you
might
go
and
use.
We
have
these
these
tracing
extensions.
Where
you,
you
might
go
and
say,
there's
a
causality,
that's
causing
you
know
when
it
all
comes
from
and
you
can
go
and
follow
through
for
for
tracing.
A
E
E
B
A
E
A
B
A
E
B
B
H
J
E
A
J
J
J
J
B
Now
now,
since
I
haven't
been
part
of
the
work,
so
use
work
group
and
that's
basically
just
a
function
of
time
so
far,
but
I
will
eventually
I
will
have
to
the
what's
interesting
here
in
that
slide.
Is
that
you're
effectively
driving
the
state-of-the-state
machine
kind
of
independent
of
the
work
that
you're
doing
like
you're,
advancing
the
state
machine?
Then
you're
you're
reacting
to
the
state
changes
by
doing
the
work,
which
is
a
bit
there?
B
J
B
J
So,
like
you
know
so,
for
this
one
right
when
the
users
say:
oh
I
need
event
wise
I
need
a
stock
trade
right,
so
that's
even
state
waiting
for
that
event.
So
it
starts.
The
starting
state
of
the
workflow
is
an
event
state,
and
that
mean
that's
looking
way.
Well
that
even
happen,
I
don't
function,
one
to
execute,
which
he
doing
some
authentication
and
then
it's
going
to
transition
to
switch
state
and
possible
to
ones
results
over
in
that
switch.
They
come
to
see
check
with
a
stick
about
the
function.
B
I
think
my
objection
here
is
one
that
is
a
bit
more
fundamental,
which
is
something
that
we
shouldn't
just
got
like.
If
this
is
the
discussion
of
of
where
we
are
in
the
works
or
discussion
group,
then
that's
great
and
I
don't
want
to
go
in
and
get
in
the
way
of
this
but
I'll.
The
observation
that
I
have
is
that
effectively
the
state
change
so
I'm,
just
look
what
I'm
looking
at
is
this
operation
stayed
by
and
then,
when
it
hands-off
effectively
control
to
FN
to
write
that
arrow?
What
that
is?
B
B
So
so
effectively,
you
are
then
wiring
up
the
he
event.
Path
of
I,
just
I,
just
I
just
performed
the
transaction.
You're
wiring
this
back
into
this
into
the
state
machine.
That's
how
you
do
this
right.
B
J
B
B
You
know
detail
definition
of
what
an
event
really
is
and
and
differentiate
that
for
messages
and
then
basically
forward
reference
to
your
section
and
say:
Kathy
is
going
to
have
a
example
that
shows
that
some
things
are
jobs
and
their
messages
and
other
things
are
events
and
we're
showing
how
both
of
those
elements
are
useful.
Within
the
scope
of
a
of
a
single
system,
the.
A
B
B
B
That
no
they're,
both
they
both
look
suspiciously
identical,
but
they
are
different.
There
are
somewhat
different
in
terms
of
how
they
what
what
they
carry
so,
for
instance,
the
message
that
you
send
has
intense
and
a
destination
like
you
are
sending
the
message
to
someone
for
them
to
go
and
execute
that
trade
and
you're
interested
in
that
in
the
result
of
that
in
the
result
of
that
trade.
B
J
B
J
But
I
just
want
to
clarify
firm
workflows,
point
of
view.
Actually,
you
know
only
care
about
the
event
that
you
not
only
have
event
event
informational
event.
If
you
need
event
identifier,
that's
it
the
videos.
You
know
the
workflow
doesn't
differ
any
event
details
it
just
defined
what
kind
of
event
it
is
right.
J
A
A
B
You,
if
you
click
in
the
middle
yeah
that
works
too.
It's
like
yeah,
so
here
that
that's
the
section
here,
basically
I,
so
that
there's
there's
a
word
cloud
in
the
prior
slide
and
basically
categorizing
them
and
then
I
split
this
up
into
the
next
slide.
Is
that
and
you
know,
there's
intense
and
there's
facts
intense
then
work
out
the
messages.
Facts
are
effectively
events,
and
so
the
and
messaging
deals
with
events
and
inventing
the
deals
with
facts
and
I
can
go.
B
I
can
summarize
that
more
successfully
in
one
slide
and
I
also
have
ensure
that
also
in
this
chat,
something
that's
a
little
bit
more
product
product
focus
and
that's
a
blog
article
that
basically
summarizes
the
summarizes
that,
as
the
differentiation
between
messages
and
events,
so
I
can
go
in
through
the
one
slide.
So
Clements.
A
I'm
a
little
confused,
though,
if
you
were
to
remove
the
I
want
to
say
semantics:
that's
probably
not
the
right
word,
though,
but
if
you're
removed
sore
the
purpose
behind
why
some
message
exchange
happened.
What's
the
difference
between
messaging
and
eventing,.
B
You
can't
remove
the
semantics
and
semantics
actually
matter,
because,
because
this
meant
so
so,
if
you
are
just
telling
a
fact,
you're
effectively
just
reporting
or
reporting
out
it
exhaust
and
you
are
not
making
anybody
do
work
if
anybody
needs
to
do
work.
That
is
you
need
to
go
and
assign
that
message
to
a
receiver,
and
you
also
need
to
make
sure
that
that
that
that
there's
only
one
receiver
that
gets
that
data
and
handles
that
handle.
B
It's
that
job
right
or
you
have
a
message
to
transfer
value
in
itself
in
the
extreme
case,
Bitcoin,
where
you
only
have
one
processor
which
can
go
and
and
put
that
into
a
ledger.
These
are
all
very
different
things
from
effectively
just
telling
you
that
something
has
happened,
which
is
something
that
we
can
affect
the
broadcast
to
the
world
as
much
as
you
like.
So
these
the
from
a
it's
all
the
same.
It's
effectively
the
same
mechanism
like
it's
the
same
protocols.
A
B
For
instance,
let's,
let's
briefly
talk
about
while
we're
at
it,
let's
talk
about
the
addressing
discussion
that
we
had
right
with
with
events
what
we
currently
do.
We
have
these
this.
You
know
we
have
the
source
of
this
notion
of
sorts,
which
really
is.
Is
it
says
this
event
is
about
this
general
scope
or
this
general
context
right?
It's
not
really
truly
the
source,
it.
B
You
know
the
the
overall
categorization
it's
topic,
even
though
that
people
have
problems
with
that,
but
that's
what
what
source?
What
source
really
means?
There
is
no
there's,
no,
nothing
we
have
in
in
the
in
the
specification
that
says
this
is
where
this
goes
to
in
messaging
is
different,
because
there
it's
there's
actually
a
stress
on
where
things
go
to,
because
you
care
about
a
particular
destination
where
things
are
heading.
So
in
terms
of
eventing,
you
generally
have
a
pub
sub-model
well
in
messaging,
you
generally
have
a
routing
model.
B
B
A
B
A
Okay,
maybe
I
miss
phrase.
The
question
so
so
I
understand
what
you
said
there
and
I
understand
that
when
someone
is
setting
up
their
infrastructure,
they
need
to
understand
there's
between
the
two,
but
once
you've
decided.
Okay,
I
am
going
to
send
an
event
and
cloud
events
looks
like
a
good
way
to
format
or
serialize.
My
events
does
the
does
the
the
split
between
messaging
versus
eventing
come
into
play
at
all?
Yes,.
B
Because
you're
sending
your
are
sending
to
a
particular
executing
party,
the
who
ought
to
go
and
buy
like
you
have
one
designated
receiver,
whether
you're
choosing
whether
you
choosing
to
decouple
that
in
the
infrastructure,
that's
a
different
story
by
you
know
having
a
designated
receiver.
That
is
a
queue
and
then
you
have
the
al-qaeda
Sivir
cannot
in
quotes
anonymously,
picking
it
up
from
that
queue,
but
you
have
one
place
where
the
control
transfer
happens.
That
is
not
an
event
and
we
have.
B
We
don't
have
a
mechanism
to
even
express
that
in
in
our
spec
right
we
can't
say
and
intentionally
we
can't
say
this
goes
here.
We
simply
say
this
is
this.
So
this
is
this
event.
We
fully
describe
the
event
where
it
came
from
and
what
the
type
it
is.
Well,
we
don't
have
any
notion
of
addressing
in
the
cloud
event
spec
and
for
that
transfer
you
actually
need
to
not
address
it.
J
So
here's
a
chroma
key
I
thought
I
think
you
know,
okay
from
workflows
point
of
view.
I
do
not
see,
there's
a
need.
You
know
to
print
the
messaging
into
the
picture
because
that
event,
so
let
me
clarify
this,
and
so
when
that
event
one
happens,
that's
a
stock
trade
request,
that's
an
event
and
then
you
know
the
function.
J
One
is
going
to
go
to
the
event,
the
payload
or
the
event,
how
to
say
the
major
data
event
payload
and
see
whether
it's
a
it's
a
buy
or
sale
and
has
that
information
which
is
part
of
that
event?
You
know
the
event
could
have
many
information
right,
even
payload,
but
is
it
chocolate
out
just
past
that
that
you
know
Phi
or
sale
information
down
to
the
next
state
and
then
down
to
the
operation
state?
So
there's
no
messaging
involved
here
see.
B
B
In
addition
to
telling
you
that
you
now
are
about
to
own
that
stock,
but
the
event,
one
really
is
someone
saying:
I
want
to
go
and
buy
something,
and
that's
not
a
pubsub
thing.
What
that
is,
is
you
are
issuing
a
purchase,
a
purchase
order
to
a
system
that
now
needs
to
go
in
and
and
execute
so
that
is,
that
is
technically
seen,
not
an
event,
well,
that
it
might
be
expressible
in
the
cloud
events
formats,
but
it's
not
an
event,
because
it's
not
a
fact
that
has
happens.
J
Ok,
so
ok
I,
don't
quite
get
what
you
mean,
but
I
think
you
know
these
are
the
event.
So,
no
matter
whether
the
stock
trade
request
of
the
managers
of
all
the
customers
say
of
confirm
it's
all
come
from
the
web.
So
for
this
stock
trade
system
right,
the
user
entered
on
the
website,
enter
some
information
and
then
say:
oh
I
want
to
trade.
This
stock.
B
B
F
A
B
B
If
you
differentiate
out,
there
is
actually
a
job
that's
being
executed
by
by
the
operation
by
that
then
calls
fn2,
but
then
there
might
be
also
an
event,
that's
being
raised
by
that
fact
that
you
can,
though,
then
have
insights
into
the
workflow
state
so
effectively.
Monitoring
of
the
workflow
state
is
something
that
you
would
do
with
events,
but
the
control
transfer
into
the
execution.
You
would
do
effectively
with
messaging.
So
I'll
write
this.
B
J
Okay,
so
I
think
probably
what
not
I
want
to
first
make
sure
I
think
we're
on
the
same
page.
Okay,
so
for
this
example
right
so
from
the
user
is
going
to
be
fun.
This
is
like
stock
trading
platform,
okay
or
stock
trading
application.
That's
supports
the
stock
trade.
It
can
receive
the
stock
trade
requests
from
different
users
from
making
hundreds
of
thousands
of
users
right.
So
whenever
there's
a
user
that
says,
okay
I
want
to
do
a
stopped
red,
stop
trade,
no
matter
is
100
shares
or
1000
shares
right,
that's
event.
H
B
The
reason
why
I'd
wear
disagree,
if
you
get
so,
if
you
get
the
order
to
go
and
execute
to
execute
a
trade,
you
only
can't
do
that
execution
once
and
if
that
execution
fails.
For
some
reason
you
should
go
and
go
basically
put
that
job
back
into
the
queue
and
then
retry
it's
a
hot.
That's
not
that's
an
operation
that
requires
high
reliability,
that's
not
an
event,
because
an
event
is
something
that
effect.
That's
a
that's!
That's
like
a
photo!
B
You
send
it
and
then
you
receive
it,
but
once
you
have
it,
you
have
received
it
and
you
fail
to
execute
that
action.
There's
no
way
to
go
back.
So
that's
why
you
need
to
have
you
need
to
have
a
messaging
system
and
need
to
have
a
queue
to
go
and
and
make
sure
that
you
really
really
execute
that
job,
because
it's
control.
J
So
that's
the
platform's
job,
okay,
not
the
event,
job,
okay,
so
the
platform
here
like
it
for
each
state
for
each
function.
If
it
fails,
no,
my
tea
is
function
while
or
function
to
even
the
authentication
that
could
fail
right.
Any
function
that
could
fail.
Those
are
really
try
mechanism
or
there's
a
you
know
in
the
work
flow.
J
If
you
go
to
the
work
flow,
the
user
can
specify
the
retry
interval
that
retry
wait
time
and
how
many
times
you
know
it
maximum
retry,
and
then
this
platform
will
do
that
retry
based
on
the
user's
you
know
whether
the
user
can
specify.
Oh
this
requirement.
So
it's
not
the
events
systems
job
to
do
the
retry,
because
it's
a
service
platform
because
that's
a
function
execution
it
has
nothing
to
do
with
the
event
event
is
just
event:
that's
it!
You
know
how
you
are
going
to
handle
the
event
the
serve.
J
B
J
B
But
how
do
you
prevent
how
you
prevent
that
you
have
two
subscribers
which
both
are
able
to
execute
to
execute
trades,
which
are
then
both
observing
that
event?
How
you
prevent
that,
because
you're
gonna
have
a
scalable
system,
where
you
have
many
many
many
different
nodes
which
all
need
to
go
and
execute
those
trades.
How
do
you
realize
that,
with
an
event
system,
you
can't.
J
It's
not
the
event
job,
okay,
it's
a
somebody's
problem.
So
why
need
skills
are
right?
It's
going
to
make
sure
you
know
one.
How
does
it
work
instance?
Basically,
one
instance
handle
that
one
workflow
instance
handles
that
trade
to
specific
trade.
If
there's
another
trade
from
another
user
is
another
instance.
B
J
Okay,
so
that's
a
performance.
Job
sohow
platform
is
on
the
implement.
I
mean
implement
this
in
a
workflow,
the
backend
problem:
how
to
export
this
workflow?
How
the
primer
implements
I
think
that's
a
preference
job.
That's
the
implementation
easier,
either
use
a
few.
What
kind
of
chill
right
there's
so
many
different
types
of
Karaka?
Okay
here
so
I,
think
that's
the
implementation,
easier,
I
think
here
we
do
not
need
to
get
into
the
detail
or
into
the
implementation,
but
it
could
be
different
types
of
implementation.
Different
company
will
have
different
implementations
here.
A
L
Think
I'm
just
saying
saying:
I
think
we
do
need
to
kind
of
think
about
how
a
platform
with
implement
some
of
these
things,
defining
it
in
a
way
which,
like
this
is
kind
of
one
of
the
reasons
I
actually
honestly
didn't
backing
out
of
flight.
Some
of
the
workflow
meetings
is
because
we
kept
on
saying
hey.
We
don't
want
to
think
about
how
a
platform
will
implement
it.
L
J
So
how
the
platform
implement
I
think
that
if
people
have
interests,
we
can
have
another
whatever
is
we
call
spec
or
whatever,
or
we
can
call
it
implementation?
Then
we
can
say
how
we
implement
this,
but
I
think
how
you
implement
this
I.
Don't
think
this
should
be
a
walk
flow
standard.
You
know,
I
mean.
B
I
have
literally
built
systems
like
this
right.
Iii
have
a
background
in
financial
systems
and
I
have
been
building
workflow
systems
that
provision
stuff
in
either
on
very
shaky
ground.
In
the
beginning
of
the
platform
like
I've
been
building
I,
don't
know
how
many
of
these
workflow
systems
and
there
are
two
different,
two
different
floors
of
messages
in
those
systems.
B
Infrastructures
exist
and
Hughes
exists
and
Kafka
exists,
because
those
all
have
very
different
different
reasons
for
for
existing,
so
ii,
so
before
I've,
I'm
gonna
write
this
up
specifically
to
the
slide,
but
the
second
link
that
are
sent
into
the
chats
where
I'm
explaining
in
a
blog
post,
why
we
have
so
many
different
messaging
services
and
azure
basic
goals
and
rationalizes
that
what
that
what
that
does?
Is
it
basically
explains
along
the
side
of
you
know?
We
have
real
and
we
event
event
ups
and
we
have
event.
B
We
have
service
bus
if
you
go
and
take
this
as
event,
bird
as
the
event
browser
service
bus
as
your
message
broker
and
event
hubs
as
Kafka.
If
you
read
it
like
this
I'm
going
to
give
you
I'm
giving
you
in
that
article,
the
argument
for
why
we
need
to
have
so
different,
it's
so
many
different
services
why
people
use
those.
So
the
the
distinction
between
control
flow
and
the
end
events
is
a
very
important
one
and
it's
realized
in
today's
distributed
systems.
Architecture,
components
that
we're
all
using.
J
Mean
really
to
face
off
all
this.
You
know
so
so
I'm,
here's
what
I'm
thinking.
So,
if
you,
okay,
I,
think
the
the
message
in
the
control
flow
itself
as
part
of
the
service
platforms,
job
I,
don't
think
he's
event.
Specs
job
I,
think
events
that
we
just
wait.
I
think
you
know
country.
We
have
all
those
information
like.
B
J
B
B
Lately,
you
literally
take
the
the
logical
threat
of
execution
and
delegate
that
out
to
someone
else,
you
need
to
have
something
you
need
to
have
a
system
that
deals
with
that.
You
need
to
have
timeouts,
you
need
to
have
dead-letter
queues,
I
mean
there's
all
kinds
of
stuff
that
you
need
to
make
that
control
transfer
work,
and
we
don't
have
that
romantic,
because
we
don't
need
that.
You
don't
need
that
in
eventing,
but
the
venting
is
too
primitive
to
deal
with
with
execution,
specifically
in
a
financial
scenario
where
you
have
stuck
execution.
J
So
when
you
say
that
timeout
retry
is
different
in
the
workflow
okay,
well,
you
talked
about
automation,
about
correlation,
that's
a
very
good
point.
So
that's
why
I
keep
mentioning
that
you
know
really
to
you
know
just
have
an
event
itself.
You
know.
There's
no
there's
I
mean
there's
no
function,
not
real
functionality,
but
to
connect
all
this
event
and
functions
together.
We
need
a
correlation,
that's
also
defined
in
the
water
off.
So
hobby
must
remain.
You
didn't
join
the
work
flow
group
discussion.
J
You
know
so
you
probably
you
know,
meet
some
information,
but
I
think
that's
part
of
the.
You
know
what
prospects
the
user
need
to
specify
how
you
correlate
these
multiple
units
together,
because
now
they
are
because
now
the
functionality
is,
you
know
we
see
introductional
service.
The
functionality
is,
is
hot
to
say
the
compote
between
the
service
problem
and
the
function
the
function
developer,
so
they
need
to
Carly.
That's
why
they
also
need
a
work
flows
back
got
previously
right.
J
Everything
is
done
in
one
system
by
the
in
by
the
application
developer
by
the
stock
plate
application
developer
in
his
own
system.
Now
some
heart
of
the
work
like
the
wool
flow
control,
the
the
scheduling
of
the
resources,
Lily
rate,
try
and
the
correlation
all
these
are
done
by
the
platform.
The
function,
that's
it.
So
that's.
B
You'd
be
super
super
happy
participating
in
the
AMQP
working
group,
because
all
the
things
that
are
that
you
need
for
for
your
scenarios.
What
it
sounds
like
is
what
we
have
a
name
QP,
because
that's
where
you're
you're
describing
you're
really
describing
a
messaging
scenario
and
not
an
emetic
scenario
and
and
that's
and
and
there's
places
for
both.
And
this
is
what
we
have
our
feet
in
in
both
of
those
areas.
But
the
eventing
scenario.
B
Your
overlay,
a
lot
of
complexity
over
the
event,
a
scenario
that
is
already
solved
in
in
in
messaging
infrastructures.
Right
your
Adi
you're,
trying
to
add
stuff
to
something
that
it
ought
to
be
something
that's
fairly
simple
and-
and
you
know,
unify
across
a
ton
of
platforms.
By
just
you
know,
delivering
events
and
in
facts
with
effectively
a
correlated
molten.
J
Okay,
I
think
I'm.
You
know
what
I
think
we
can
pop
it
later.
I
think
we
can.
You
know,
discuss
more
about
this.
You
know
how
that
fitting
to
the
inventing,
but
at
this
stage
right
I
think
you
know,
we
probably
you
know,
I,
don't
know
whether
it's
good
to
bring
this
because
there's
people
is
going
to
be
it's
not
part
of
the
events
spec.
It's
not
a
part
of
the
work
flows
back
I'm,
not
sure
that
that
will
confuse
the
audience
rather
than
clarifying
more.
A
H
A
B
And
I
would
appreciate
if
you
would
read
that
blog
article,
ignoring
ignoring
and
Kathy,
and
and
just
ignore
the
fact
that
there
are
product
names
there,
because
it's
really
an
this
description
of
categories
of
products
like
so
service
must
a
stands
for
message,
brokers
and
event.
Table
stands
for
Canon
jesters
like
Kafka
and
event
grid
is
affecting
what
we're
defining
here.
B
J
L
The
Shanghai
perspective
right
I
think
if
we
can
talk
about
a
KB
working
on
workflow
and
we
have
some
ideas
like
I
personally,
don't
think
we're
a
point
where
we
can
say
hey.
This
is
the
thing
and
you
decided
all
but
kind
of
bringing
it
up.
So
other
people
who
might
not
agree
with
their
ideas
get
more
engaged.
That's
good,
yeah.
B
J
Actually,
I
think
the
purpose
of
presenting
this
is
to
collect
more
people
to
in
to
me
to
be
involved
on
this.
You
know
I,
think
you
know
the
word
definite
is
not
the
final
draft
there,
but
there
are
errors.
They
are
even
just
you
know,
kind
version.
I
can
see
some
arrows
there,
because
different
people
put
in
different
section
right,
yeah.
H
H
J
Bring
people
in
introduced,
you
know
the
work
we
are
doing,
yeah.
A
Okay,
so
I
think
we
might
keep
this
dead
horse
enough.
So
what
if
we
do
this?
What
if
we
all
try
to
say
that
by
next
Thursday's
phone
call,
we
have
our
are
all
of
our
slides
in
there
and
ready
for
for
review,
because
I
thought
I
know.
I
have
more,
do
I
need
to
put
in
as
well
but
I
put
in
some,
but
not
enough.
But
what,
if
we
shoot
for
next.
J
G
A
We
need
to
reschedule
that.
That's
obviously
fine,
because
it's
obviously
jury
Duty's.
That's
then
you
can
just
avoid
what
why
don't
we?
Why
don't
we
shoot
for
Thursday?
And
if
you
don't
make
it,
then
we
can
try
to
reschedule
for
some
other
time,
but
at
least
everybody
try
to
get
their
stuff
done
by
Thursday
and
then
we'll
figure
out
when
to
meet
after
that
how's
that
yeah.
I
J
J
A
J
B
A
B
A
J
H
A
L
A
J
I
J
A
Hold
on
a
minute
wait:
wait.
Wait,
wait,
wait!
Kathy!
That
is
an
excellent
question,
so
hold
on
a
minute
me.
I
will
ask
on
next
week's
call
cancel
coop
con
call
kook
on
calls
no
cancel
with
us
even.
A
Figure
it
out,
but
it's
definitely
a
good
topic
for
next
week's
phone
call
to
bring
up
and
make
sure
everybody's
on
the
same
page.
So
thank
you
for
reminding
me.
Okay.