►
From YouTube: CNCF Serverless WG Meeting - 2019-08-29
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
Talk
about
that
probably
shouldn't
be
applicable
anymore,
based
upon
other
change
to
the
spec,
so
yeah
I'm
sure
there's
plenty
of
stuff
yep,
all
right,
all
right!
Three!
After
let's
go
ahead
and
get
started,
so
you
know
13
all
right,
but
I'm,
community
time,
okay,
anything
from
the
community.
If
you
want
to
bring
up
nada,
okay,
STK,
Scott
or
Clemmons
I
know
we
have
a
call
scheduled
for
after
this
one.
But
is
there
anything
you
want
to
mention
on
this
call
or
anybody
else
from
the
SDK
work.
B
A
Cool,
if
you
guys
have
any
questions
about
that
feel
free
to
join
the
SDK
call
right
after
this
one
incubator
status,
okay,
we're
still
scheduled
for
September
17th
I
have
the
proposal.
Powerpoint
deck
is
still
here.
I
have
not
uploaded
into
the
agenda
yet
I.
Do
that
later
this
week,
I
believe
we're
technically
ready
to
go.
We
do
have
three
end
users,
but
we
can
always
use
more.
The
chart
looks
kind
of
small
with
just
the
three.
So
if
you
do
have
any
want
to
mention,
please
let
me
know
other
than
that.
I
do
believe.
A
We
are
ready
to
go
so
fingers
crossed.
I
did
create
an
outline
doc
for
the
two
sessions
at
Keuka,
North
America.
It's
pretty
much.
What
we
agreed
to
before
so
please
feel
free
to
look
at
that
edited
as
you
see
fit
in
terms
of
adding
topics,
and,
of
course,
if
you
do
want
to
talk
to
one
particular
section,
stick
your
name.
There
don't
hesitate,
putting
your
name
next
to
something
they
already
has
someone
else's
name
there.
We
could
work
out
who's
gonna.
You
know
how
we're
gonna
arm
Rusk
figure
out
who's
gonna.
A
Do
it
later,
but
just
having
a
list
of
people
to
choose
from
would
be
nice,
so
feel
free
to
add
to
name
that,
as
you
start
thinking
out
these
things,
you
still
have
time
to
work
that
out.
It's
not
total
November
I,
think
that's
it
in
terms
of
administrative
administrivia,
stuff
anything
else
before
we
start
talking
about
PRS
that
you
want
to
bring
up.
A
C
I
would
so
there's
two
that
are
related
and
I.
Think
the
after
the
discussion
we
had
of
PR
I
was
convinced,
which
occasionally
happens,
but
I
think
that
that
we
should
pull
this
into.
We
should
pull
an
entire
concern
into
the
JSON
encoding
and
to
make
it
not
so
not
as
near
as
it
would
be
if
we
were
introducing
a
extra
attributes.
Just
for
this
I
chose
to
change
the
attribute,
so
I
would
like
to
talk
for
92,
okay,.
C
C
So
in
mapping,
so
this
is
an
adjacent
format.
Well,
that
seems
there
is
basically
saying
if,
if
it's
a
string,
then
it
must
be.
It's
mostly
called
beta
diversity.
The
extra
rules
for
clean
Jason
mapping,
which
also
applied
to
that
field
for
binary.
If
it's
binary,
then
the
the
the
member
name
in
the
Jason
object
must
be
data
underscore.
Basically,
of
course,
that's
our
that's
the
way
how
we
distinguish
between
the
two.
C
That
is
if
we
make
something
that
specific
or
Jason,
which
in
effect
in
effect,
is
then
it
should
be
somewhat
similar
to
the
function
of
an
attribute
in
XML
rather
than
having
introducing
a
whole
new
attribute
that
is
inside
the
inside,
the
the
the
the
events
that
kind
of
sits
in
parallel
but
describes
data
just
in
the
Jason
case.
That
seems
a
little
strange
to
me
if
it's
not
a
general
thing,
but
really
only
applies
to
Jason,
the
Jason
Kony
shouldn't
be
adding
extra
things
in
it.
C
That's
how
this
would
look
then.
Basically,
yes,
like
that,
so,
instead
of
data,
if
it's
basically
encoded,
it's
simply
called
data
underscore
basics
before
and
by
implication
can
be
one
or
the
other
and
then
in
our
main
spec,
then
entire
concert
just
goes
away,
so
the
intercompany
coding
is
gone
and
relating
comments.
D
D
D
C
You
well,
if
you
get
a
data
underscore
basics
before
that,
implies
that
you
that
you
have
your
data
is
binary
and
then
it's
basically
for
coding
and
then
you
have
to
yeah.
You
still
have
to.
You
still
have
to
reconcile
it
with
whatever
the
content.
Type
is
right.
So
if
the
content
type
is
text,
slash,
XML,
semicolon
charts,
Nets
exudate,
then
you
have
to
go
and
take
that
basic
before
turn
it
into
a
binary,
rather
through
the
exit,
absolutely
corner
and
then
decode
the
XML.
D
A
Have
a
question:
if
I
raise
my
hand,
I
I
think
you
may
need
to
add
a
little
bit
of
text
in
here
to
make
it
more
explicitly
clear
that
you
can't
have
both
data
and
data
underscore
basic
c4,
because
I
don't
think
it
says
it
yet.
You
know,
and
so
I
may
put
both
in
there
but
I
want
make
it
clear
that
having
both
in
there
is
an
invalid
crowd
event:
okay,
right,
okay,
yeah
I
would
but
the
other
thing
is
that
yes,
but
my
very
concern
is
the
wording
change
it.
A
My
very
concern
is
I.
Guess
it's
related
to
what
Tim
was
saying
is
since
obviously
I
I
haven't
looked
at
this
until
just
now,
and
obviously
even
have
you
encoded
it
up
from
a
coding
perspective,
I
I
am
actually
concerned
about
the
exact
opposite.
What
Tim
said?
I
I'm
wondering
whether
it's
actually
harder
for
people
to
not
know
where
to
look,
especially
if
you're
writing
a
generic
cloud
event.
Processor,
because
now
you
have
to
look
for
both
and
I
understand.
A
Some
of
the
concerns
of
having
a
sort
of
a
modifying
property
doesn't
rise
alongside
things
but
I.
Don't
think,
that's
that
we
of
a
thing,
because
you
kind
of
have
that
already
with
things
like
HTTP
headers,
like
you
know,
your
your
content
type
tells
you
how
to
interpret
you
know
they
should
be
body
right
and
that
can
vary
based
upon
on
the
value.
We
don't
have
two
different
types
of
HTTP
bodies,
one
for
binary.
A
C
Be
the
well,
the
alternative
is
491
right,
which
is
the
rename
we
talked
about,
and
that
is
be
that
basic
renaming
the
data
content
encoding
is
the
data
encoding
simplified?
Not
so
that's
that
change
right
and
that
basically
removes
all
the
references
to
the
prior
to
I
couldn't
get
up
even
more
than
we
discussed
where
I
removed
all
the
references
that
confuse
people
to
the
contents.
C
Was
they're
called
content,
not
what
the
transfer
encoding
of
SMTP,
which
I
was
referencing
because
of
basic
before,
and
there
was
like
continent
coding
and
transfer
encoding.
People
had
all
kinds
of
idea,
so
I
had
I
just
remove
all
that,
and
just
just
point
through
the
basic
before
encoding,
standalone,
RC
and
then
and
then
busy
clarify
that
in
the
reference
that
in
the
in
the
Jason
spec.
So
this
is
this
is
where
this
is
the.
A
B
C
A
B
C
Yes,
you
can
be
SDK,
I
would
make
to
go
away.
This
is
just
the
wire.
This
is
aware
of
the--,
so
if
they
have
main
underscore
basics,
the
course
shows
up.
I
decode
I
decode
this
and
stuff
that
into
a
binary
and
the
binary
becomes
the
type
of
the
data
property
and
then,
if
the,
if
data
shows
up,
then
it
becomes
an
object.
Graph
in
the
simplest
case
is
the
spring
and
then
otherwise,
if
it's
could
take
jason
data,
it's
a
jason
graph.
So
it's
it's.
C
A
When
I
try
understand
theat
the
flip
side,
so
the
user
has
a
cloud
event
object
that
there's
nothing
it's
a
metadata
into,
and,
and
I
stick
something
inside
of
data,
you
would
dynamically
check
to
see
whether
that
data
object
has
binary
stuff
in
it
or
just
straight
text
and
make
the
decision
as
to
whether
use
data
versus
data
underscore
basic
support.
Based
upon
that,
if
you
give,
if.
A
C
You
said
you
you
walk
up
to
it.
You
walk
up
to
the
cloud
event,
object,
you
go
to
data
property
and
you
assign
a
byte
array
to
it.
Oh
right,
then,
under
the
covers,
as
that
gets
flushed
out
to
the
wire
that
data
gets
basically
who
are
encoded
and
gets
put
into
the
data
on
a
score
basically
for
field.
If
I
read
that
back.
In
that
event,
I
read
that
through
the
Jason
coder,
the
Jason
recently
at
the
document
will
see
neither
underscore
basics
before
will
running
the
basically
for
decoder
over.
C
It
gets
binary
and
that's
the
binary
that
I
signed
back
to
the
data
properties.
That's
all
I
do
like
that
detail
difference.
Nobody
sees
that
detail,
different
thing
if
they
go,
if
you
go
into
the
into
the
I,
have
this
raw
collection
inside
of
the
event
object?
That
is
all
the
attributes,
in
effect,
meaning
the
law
that
raw
thing
will
be
a
data
and
a
property
called
data
that
contains
the
actual,
be
sterilized
data
file.
C
C
I
would
well
I
would
first
followed
I
would
follow
the
the
Jason's
game
rules
for
that
string,
okay,
I
mean,
did
you
see
corner?
Does
that
you
can
go
and
put
whatever
whatever
you
want
in
it,
and
then
there's
rules
that
Jason
quarter,
then
Jason
will
do
to
make
sure
that
it
represent
your
string.
Yesterday,
okay
I
forgot
about
that.
Okay,
so
so
it's
really
just
this
is
literally
just
a
flag
on
data
for
transport
purposes.
That
says,
that
basically
indicates
that
becomes
it
of
this
is
basic
or
right
and.
C
Evan
says
right
rightly
says:
if,
if
for
hypothetical
xml
encoding
in
data
for
data,
if
you
would
probably
then
use
in
coding
basics
before-
and
there
are
actually
in
XML
as
tip,
will
tell
you
four
hundred
ways
to
declare
that
the
data
inside
of
their
basic
before,
and
so
that's
that
seems
neither
this
miss
strikes
me.
That's
the
right
way
to
do
this
for
XML,
and
then
young
amo
also
has
a
way
to
to
go
and
declare
this
as
an
annotation.
A
My
hand
opener
I
said
what's
interesting,
is
when
I
was
talking
to
heaven
about
this
yesterday.
My
assumption
was
that
if
we
ever
did
come
up
with
in
XML
or
yeah
mol
encoding
that
we
would
say
that
the
data
content
encoding
flag
would
appear
as
a
property
or
a
as
an
attribute
here
in
XML,
or
you
encode
data.
This
way
and
yeah
Moe,
we
witness
all
you
have
to
forcing
another
top-level
property,
yeah
and.
A
It's
just
interesting
because,
with
with
your
news
with
your
latest
proposal,
what
you're
basically
doing
is
making
it
a
transport
specific
concern,
and
yes,
the
the
nice
thing
about
the
finding
in
our
was
that
you
at
least
have
consistency
at
an
abstract
level.
It
just
may
appear
differently
at
the
serialization
level
and
we're
saying
nope
we're
not
in
him
to
find
out
the
abstract
level.
It's
completely
a
serialization
problem.
I
was.
C
A
Okay
I
believe,
since
the
pier
was
just
open
today,
it'd
be
unfair
for
us
to
actually
prove
it
today.
However,
last
chance
is
there.
Anybody
I
still
think
may
be
a
little
bit
sour
text.
That
says
you
can't
have
both
in
there,
but
it's
a
minor
type
of
graphical
thing.
Are
there
other
things
relative
to
this
PR?
Who
would
like
to
discuss,
or
is
everybody
pretty
much
yab?
This
is
the
way
to
go.
A
B
C
C
A
C
Rest
is
the
reference
to
something
that's
outside
of
the
envelope,
that
is,
that
is
very
different
from
you
know,
a
data.
This
is
still
debate
afield.
They
understood
the
underscore
business,
of
course,
effectively
just
an
attribute
on
the
data
field,
business
inside
this
name,
if
you
lose
basics,
of
course
it's
not
it's.
It
looks
like
a
different
field,
but
but
conceptually
it's
the
same
thing
as
having
an
attribute
on
an
XML
element.
C
B
A
A
A
A
C
A
A
Okay,
I,
don't
lay
it
so
useful
for
us
to
walk
through
it
right
here.
Could
you
comment
on
this
I'll
talk
to
heavens.
Up
to
this
call
saying,
looks
like
we're
headed
towards
your
your
second
PR
today.
So
that's
gonna
impact.
Obviously
the
data
content
encoding,
but
can
you
make
a
comment
on
why
this
section
might
need
to
change.
E
A
F
That's
a
good
way
to
describe
it,
so
we
still
have
we
removed
map
from
our
type
system
and
we
still
have
it
in
all
the
event
formats.
So
I
move
that
and
we
added
to
edit
the
URI
so
I,
also
edit
this
12
formats
and
then
I
did
one
more
thing
in
the
spec
and
that
I
insert
another
header.
That's
called
type
system.
We
had
this
before
we
removed
it,
but
all
the
event
formats
basically
say:
please
refer
to
our
type
system
and,
for
example,
to
Jason
one
even
links
back.
F
A
F
F
A
A
A
A
Yeah
he's
needs
to
do
some
updates,
so
we
can't
even
look
at
that.
Okay,
so
let's
let's
skip
that
one
all
right.
So
let's
talk
about
timelines.
My
hope
is
whether
this
was
my
hope
that
we
resolved
Paris
today,
obviously
we're
not
there
yet
so
this
may
be
pushed
out
by
a
week,
but
I
was
wondering
if
maybe
on
next
week,
we
could
approve
and
say
we
approve
Clemens,
PR
and
maybe
Evans
PR
about
the
tricky
use
cases
stuff
and
this
clarification
in
the
binder
which
I
don't
really
think
changes
anything
normatively.
A
If
we
can
get
all
those
PRS
in
there
say
by
like
Monday
or
so
so,
he'll
have
a
chance
to
review
it.
What
do
people
think
about
vote
in
next
week's
call
to
approve
that
as
0.9,
which
is
technically
release
candidate
1
for
version
1.0
and
then
give
us
two
to
resolve
any
outstanding
issues
that
people
find
as
the
review?
A
The
docs
start
of
votes
on,
September
19th
do
an
offline
vote,
since
it's
obviously
a
very
big
decision,
close
it
on
September
26
and
call
it
done,
and
then
we
can
make
an
announcement
on
Kuk
on
that.
Obviously
gives
us
all
of
October
if,
for
some
reason,
we
can't
meet
these
these
deadlines
to
push
it
out,
but
I
start
off
with
being
aggressive.
What
do
people
think.
F
A
This
is
too
easy:
ok
cool
in
that
case,
looking
at
other
issues,
clip
Christoph,
I
think
you
said
this
one
obviously
will
go
away.
If,
if
we
approve
Clemens,
PR
number,
2,
right,
yeah,
ok,
webhook,
I,
think
Clemens.
You
know
I
still
need
to
talk
about
that.
I.
Don't
think
it
changes
our
course
back.
I
don't
need
to
worry
too
much
and
the
SDK
that's
gonna
be
talked
about
in
the
SDK
call.
I
think
they
just
double
check
here
in
terms
of
issues
for
version.
One.
A
Clemens,
your
PR
addresses
this
one
completely
right,
I
guess
we
talked
about
that
nevermind
I'm
gonna
assume
it's
true,
so
we
I
think
your
PR
addresses
all
three
of
these
that
one
that
one
okay.
This
is
the
big
one.
Then,
okay,
so
a
long
time
ago,
Thomas
from
Google
asks.
Basically
you
know
what
is
what's
the
criteria
for
going.
One
point,
though,
and
as
of
right
now,
this
list
may
be
old,
but
I
believe
we
have
at
least
these
implementations
out
there
and
I
think
actually
the
list
has
grown
now.
A
If
you
look
at
the
the
incubator
proposal
for
our
for
our
work,
I
actually
list
quite
a
few
more
things
in
there.
So
this
list
is
actually
quite
short
and
compares.
For
example,
I
know,
RIT
has
some
that
has
some
products.
Oracle
has
one
I
think
there's
a
couple
of
others
out
there.
So
my
question
for
the
group
is:
does
the
current
implementation
of
the
spec
satisfy
people's
conceptual
definition
of
exit
criteria
or
do
we
need
to
add
things?
The
list.
D
A
A
D
C
D
C
So,
on
our
side,
we've
been
perfectly
waiting
for
one
photo
to
lock
so
that
we
can
start
updating
our
stuff,
so
some
from
the
product
side
that
will
be
there
will
be.
Actually
you
see
her,
but
I,
don't
think
we
will
rush
it
because
now
or
getting
serious
about
thing
with
issue
stuff,
that's
product,
so.
A
D
A
The
reason
I'm
asking
for
a
little
more
specifics
on,
if
you
can't
that's
fine
reason,
I
was
asking
that
was
because,
if
we
did
want
to
put
together
some
sort
of
demo
we'd
have
to
know
you
know
the
shape
of
the
events
coming
in.
So
people
can
know
what
to
do
with
as
opposed
to
just
yes
I
got
it.
I
can
parse
it
right
mm-hmm.
D
I
mean
we're
all
so
yes,
I
mean
go
look
at
the
things
that
flow
through
the
event
bridge,
which
is
a
product
so
in
existing
product.
So
the
assumption
is
that,
assuming
you
know
assuming
cloud
event,
stabilizes
and
assuming
this
is
not
a
promise.
Okay,
I'm,
not
making
the
promise
yeah
yeah
and
assuming
things
go
the
way
I'd
like
we
would
probably
be
able
to
provide
all
of
those
of
which
there
are.
You
know
millions
per
second
in
in
cloud
events,
format
right
and
there's
a
couple
of
hundred
different.
You
know
payloads,
okay,
so.
A
A
B
A
So
obviously,
if
you're
interested
in
that,
please
join
the
SDK
call
right
after
this
one.
So
let's
go
back
around
to
the
first
question.
I
asked
exit
criteria.
Do
people
think
we
need
to
add
more
beyond
the
current
set
of
implementations
that
we
know
about,
and
possibly
the
conformance
test
suite
that
Scott's
mentioning.
A
A
H
A
H
A
A
D
Can
I
make
a
remark
or
two
on
that
yeah?
Please
I've
we've
discovered
that
once
you
actually
start
shipping
these
things
at
scale,
the
inertia
that
develops
inertia
develops
huge
inertia
develops
almost
instantly
and,
aside
from
adding
fields,
new
fields,
it
is
insanely,
difficult
verging
on
impossible
to
change
anything.
So
it
you
know.
Please
do
not.
D
A
Wait
yeah
and
that's
why
I
think
the
the
two
or
three
week
review
period
that
we're
gonna
have
here
is
critical,
yes,
but
I
believe
the
overall
question
is
still
a
good
one
and
I
and
honestly
from
my
perspectives,
I
I,
don't
think
we
can
necessarily
lock
down
too
many
things
until
we
actually
start
seeing
issues,
because
if
it
all
the
issues,
are
you
know
syntactical
in
nature?
Obviously
we
can
do
those
things.
They
don't
change
anything
normatively.
A
If
someone
was
to
find
something
significant
enough
that
we
thought
we
can't
change
this
without
breaking
backwards
compatibility,
then
we're
just
gonna
need
to
decide.
Okay,
do
we
ignore
the
issue,
or
do
we
go
over
version?
Two
I?
Don't
think
we
can
make
that
decision
until
we
see
how
big
of
an
issue
it
really
is,
so
my
I'm
inclined
to
say
well,
we
can't
answer
too
much
until
we
get
there
until
we
see
the
issue
itself
like.
D
A
G
Just
wanted
to
start
the
discussion
around
it,
basically
to
make
sure
that
we
know
what
we're
getting
into
after
we
release
1.0
and
it
sounds
good
to
me
like
we
got
a
version
out
and
we're
gonna
actually
get
some
usage
and
we're
gonna
see
what
this
develops
into
yep,
but
I
know
deep.
Only
Mother
comment
I
can
make
is,
do
you
want
a
longer
I?
You
know
preview
period
where
this
is
added
as
supported
by
multiple
tools,
and
then
we
see.
Is
there
any
point
in
that
I?
Don't
really
see
it
so
I,
don't
know.
A
Sorry
how's
you
I
was
assuming
that
on
September
19th
we
would
vote
to
start
the
vote
and
if
people
have
concerns
large
or
small,
about
starting
the
vote,
because
the
spec
isn't
ready,
I
expect
them
to
raise
those
by
then
and,
like
I,
said,
we
still
have
a
whole
month
before
coop
con
to
work
do
any
of
those
issues,
but
did
this
becomes
a
a
a
very
important
step
for
the
working
for
the
entire
working
group?
A
A
Okay,
in
that
case,
I
think
we're
done.
Oh
I
didn't
mention
one
thing:
I'd,
like
Chris
Hanna
check
ping
me
earlier
in
the
week
asking
about
our
version
1.0
status,
basically
hinting
that
they
would
love
to
make
some
PR
noise
around
this
and
then
not-too-distant
future.
So
there
there
are
obviously
our
people
within
at
least
the
scenes.
Yes,
aside
from
our
with
little
group
here
that
are
very
anxious
for
the
scene
to
go
forward.
So
I
thought
that
was
a
very
nice
little
sign
that
that
it's
not
just
us
on
this
call
that
are
interested.
A
There
are
other
people
who
definitely
want
to
see
this
thing
go
forward,
so
I
just
thought:
I'd
mention
that
and
the
CNC
F
is
all
eager
to
make
noise
around
this,
which
is
great
all
right
in
that
case,
unless
there's
anything
else,
I
believe
we're
done
and
we'll
resume
the
call
again
in
about
ten
minutes
for
the
SDK
call
all
right.
Do
you
think
you
guys?
Okay,
bye,
guys.