►
From YouTube: CNCF Serverless WG Meeting - 2019-06-06
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
All
right,
what
can't
smoke
up
with
those
guys
later
everything
just
slip
away
for
a
second,
it's
three.
After
so
when
are
we
gonna
get
started?
Oh
okay
got
to
Tim
those
pesky
mute
buttons,
all
right.
Let's
see!
Where
are
we
all
right
community
time
anything
from
the
community
that
you'll
want
to
bring
up
I,
don't
see
anybody
new,
so
I
guess
I'm,
no!
A
Okay!
Your
word
no
escy
update,
since
we
don't
call
this
week,
just
a
slight
update
for
kook
on
China
Cathy
is
not
gonna
be
able
to
make
it
so
I
will
be
presenting
both
sessions.
Melissa
somebody
else
happens
to
be
there
who
wants
to
join
in,
but
I
haven't
heard
it
for
anybody.
So
I'm
assuming
that's
not
true
and
the
slides
are
not
quite
yet
available,
but
they
are
due
June
10th,
so
I'm
hoping
to
make
that
or
do
that
this
weekend.
A
So
then,
you
guys
can
take
a
look
at
that
and
make
sure
I
didn't
go
off
the
rails,
all
right,
so
TOC
call
this
week.
The
topic
of
what
three
independent
and
users
meant
was
brought
up
and
the
agreement
I
got
from
TOC
was
that
it
is
users
of
the
product
that
implements
our
spec.
It
is
not
just
three
implementations
of
the
spec
itself,
which
is
kind
of
I.
A
Think
what
a
lot
of
us
expected
anyway,
but
it's
good
to
have
a
confirmation
and
I
asked
if
they
actually
are
going
to
use
for
the
honor
system
and
say
yes,
you
know
it's
good
after
to
say:
I
have
customers
or
do
they
actually
need
names,
and
they
said.
Ideally
they
want
names
as
proof.
If,
for
some
reason
the
names
are
confidential,
then
we
can
work
out
some
mechanism
to
get
them
that
information
offline.
So
it's
not
public,
but
they
do
want
confirmation
that
it
is
three
end
users
of
products
they
don't
put
the
spec.
A
So
oh
and
the
other
question
was
somebody
at
chemical
has
asked:
if
there's
a
version
requirement
that
we
have
to
reach
before
we
go
to
incubator
and
the
answer
was
no,
the
documentation
or
the
governance.
Documentation
doesn't
say
anything
about
version
numbers
for
any
of
the
levels,
including
being
a
graduated
project,
even
though
I
think
there's
an
implication
that
you
should
at
least
see
one
point
though,
but
the
government
of
stock
doesn't
actually
say
it,
but
for
incubator,
we're,
okay
being
beta.
What
were
you
going
to
call
us?
Okay
so
anyway,
any
questions.
B
A
C
C
A
All
right
cool
so,
like
I,
said
I
think
we
can
revisit
this
going
to
incubator
status
after
we
get
0.3
out
the
door.
I
think
it
makes
sense
to
wait
and
with
that
Before
we
jump
into
PRS.
Are
there
any
other
high-level
topics?
People
would
like
to
bring
up
all
right.
In
that
case
Clemmie
and
this
I
believe
you
are
up.
I
was
kinder
ever
what
do
you
make
any
significant
changes?
I.
A
D
Well,
senator,
why
we're
mean
with
settable
is
that
you
should
be
able
to
to
so
effectively
effective
of
this
type
system.
Definition
that
I
have
now
implies
that
that
there's
always
a
conversion
or
there's
a
canonical
way
to
express
to
express
a
value
of
that
type,
and
we
have
a
canonical
string
representation
for
every
single
one
of
them,
and
but
you
should
use
the
most
the
best
native
representation
for
sdks,
but
also
for
protocol
mappings.
So
a
time
the
time
values
should
be
at
a
timestamp
in
ATP,
so
setup
setting
means
means
broadly.
D
How
do
you
map
from
a
field
over
here
to
a
field
over
there,
and
it's
not
necessarily
being
prescriptive
about?
You
know
how
your
API
should
look
or
how
your
your
stack
should
look
but
effectively.
You
should
always
be
ready
to
set
whatever
that
value
is
in
your
native
stack
with
a
string
type
and
then
you're
ready
to
do
that
conversion
and
do
the
check
at
that
point,
and
you
should
always
have
it.
D
You
should
always
have
a
way
to
go
and
convert
that
value
that
you
have
there
in
your
native
Texas
student
back
into
that.
Canonical
string,
okay,
I,
think
that
helps
me.
So
that's
that's
that's
what
what
I
mean
with
it?
What
I'm
not
restricting
is
in
that
text
event
is
and
I
think
I
have
to
have
a
comment
about
this.
It's
perfectly
okay,
for
you,
you
go
from.
D
You
know
a
go
SDK
which
has
a
timestamp
to
m
QP,
which
has
a
timestamp
and
never
used
the
strength
time
at
all.
Right.
That's
that's
fine!
It's
just
that
that
when
you
have
a
mismatch,
when
you
can't
not,
then
you
should
be
able
to
go
to
a
string
from
that
string
to
the
other
native
type.
So,
for
instance
the
example
that
I
brought
up
in
the
comment
was
UNIX.
Unix
epoch
is
used
widely
used
to
represent
timestamps
and,
of
course
the
c-sharp
types
then
has
an
epoch.
A
D
Just
for
for
people
having
who
haven't
read
visit
or
on
the
call,
the
objection
was
that
there
are
codes
around
strings
in
headers,
specific
HTTP,
and
what
this?
What
this
here
does,
is
it
does
in
a
very
wordy
way,
does
away
with
just
simply
using
Jason
as
the
type
system,
but
introduces
a
type
system
that
has
the
effect
of
the
strings
being
exactly
the
same
as
they
would
be
in
Jason.
D
If
you,
if
we
will
use
the
date/time
and
a
string
and
a
number
I
sent
around
so
and
actually
using
the
I'm
deeply
into
the
Jason
number
definition
and
I
think
the
integer
part
of
that
number
definition.
For
that
to
be
a
number,
and
the
overall
effect
is
that,
while
all
Jason
will
stay
exact
in
terms
of
how
you
formulate
a
message,
you
can
literally
use
a
jason
encoder
to
go
and
and
turn
your
strainings
into
the
right
format
and
and
back
the
Tyrian
here
in
types
of
the
right
format
and
back.
A
D
Yeah,
so
extensions
are
so
you
would
define
extensions
as
with
their
proper
types.
The
types
of
expect
that
you
have
an
integer
for
your
sequence,
the
the
sequence
number
thing,
and
so
you'll
define
acceptance
with
a
proper
type,
how
extension
or
generally
how,
how
attributes
travel
of
the
wire
is
not
so
important
if
you
have
a
way
for
your
typing
abstraction,
to
turn
it
back
in
to
the
right
side.
D
So
what
you
would
do
is
in
your
I
expect
that
if
you
care
about
an
extension-
and
if
you
care
about
the
semantics
of
that
extension,
you
will
have
an
expectation
about
that
type,
which
means
but
data
arrives
to
you
either
in
the
right
in
the
right
type.
You
know
natively
mapped
from,
let's
say
a
message:
it
has
some
type
system
like
the
injured,
her
maps
to
an
integer
from
the
MVPs
go,
the
more
it
arrives
its
string
and
then
at
that
point
you
pick
up
the
straining.
D
But
since
your
extension
expects
integer,
it
will
go
ahead
and
do
then
a
conversion
from
that
string
into
an
integer
and
the
rules
that
we
set
here
that
we
say
here.
This
is
the
this
is
the
the
wire.
This
is
the
wire
type
description.
We
basically
mandate
that
that's
that
that
string
is
convertible.
It's
an
integer
so
effectively
the
conversion
into
the
native
type
system
of
the
programming
language
happens
at
the
edges,
but
we're
not
caring
so
much
about
it
being
the
right
type
on
the
wire
all.
D
You
don't
say:
yeah,
it's
a
meteor
intermediary
doesn't
know
right,
so
you,
the
message
shows
up
in
an
intermediary
against
an
event
by
HTTP
and
since
the
event
on
words
using
ad,
and
then
you
have
a
receiver
and
the
receiver,
and
now
once
it
go
and
evaluate
the
extension.
The
way
that
works
is
the
event
comes.
If
you
have
a
field
that
comes
in
is
supposed
to
be.
A
date
comes
in
and
astraying
over
HTTP
you
copy
that,
because
you
don't
know
what
that.
D
Is
you
copy
that
in
as
a
string
into
the
MVP
message,
the
Inc,
even
though
it
ought
to
be
a
day,
but
it's
travels
as
a
string.
You
send
that
to
the
consumer
and
the
consumer
now
goes
and
walks
up
to
the
AVP
field,
expect
it
to
be
in
timestamp,
but
now
finds
it
to
be
a
spring,
and
it
should
not
be
able
to
blindly
apply
the
conversion
rule
or
day
time
that
a
from
spring
to
daytime,
and
that
should
work,
because
we
have
specified
effectively
a
for
that.
D
D
A
A
D
A
Those
quotes
were
annoying,
so
thank
you,
alright,
so
I
believe
just
for
clarity's
sake
or
referred
out,
there's
a
completeness,
Scott
I
believe
you
think
we
can
close
out.
This
pull
request.
Is
that
true,
that's
right,
yeah
and
then
I
know
I,
don't
think
this
was
from
Alan
right,
no
he's
a
it's
Adam,
Adam,
sorry,
Adam.
A
A
A
F
D
Strong
but
constraints-
maybe
he
doesn't
doesn't
say
that
so
I
called
limits,
because
I
also
speak
about
limits
in
the
second
line
here
in
the
first
paragraph
and
then
I
personal
phones,
the
constraint
when
the
publisher
redundant
did
I
agree.
The
fat
so
I
took
that
out.
So
that
was
like
four
four
seven
down
there,
so
they
used
to
also
be
a
rule.
D
The
publisher,
though,
that
I
we
moved
there
was
a
objection
from
discussion
from
Eric
saying
that
the
at
least
rule
here
for
intermediaries
were
inclusive
and
we're
saying
you
have
to
go
and
and
do
forward
events
of
every
size,
I
think
I.
Think
with
you.
So
in
terms
of
what
I
think
this
means
normatively
is
that
if
you
support
64k,
you
are
compliant.
If
you
are
confirmed
and.
D
D
G
D
I'm
effectively,
producers
should
be
able
to
publish
events
up
to
4k
safely,
which
means
they
will
get
into
and
through
intermediaries,
and
then
it's
still
up
to
each
particular
consumer,
whether
they
want
to
go
and
take
those
events,
but
the
middleware
will
not.
It
will
not
go
on
strike
if
you
publish
events
of
CT
4k
and
then
up
to
a
consumer
whether
it
wants
to
go
and
deal
with
events
that
happen
to
be
larger
and
that's
the
that's
the
the
effect
of
all
this.
E
Well,
I
guess,
most
simply,
I
think
we
agree
that
the
the
first
statement
there
above
the
comments
in
includes
amounts
of
64
kilobytes.
What
I
my
reading
is
that
it
also
includes
amounts
that
are
greater
and
and
there's
a
must
on
the
forwarding
of
those
events
so
I.
It
was
just
a
wiggle
on
wording
and
I
think
that
it
was
easy
to
correct
for.
D
E
Yeah
yeah
rather
right
yeah.
My
understanding
is
that
to
get
the
intent
that
I
think
you
mean
do
you
would
say
it
should
be
of
size
at
most
64
kilobytes.
So
the
requirement
would
be
for
all
events
up
to
64
kilobytes
and
then
there
would
be
an
addition
area.
Dition
addition
of
kind
of
you
can
do
what
you
want
with
anything
over
that.
A
H
H
D
For
well,
what
we
have
is
a
sir
on
Metacritic,
which
is
all
the
events
that
are
being
sent
when
you
create
a
new
blob
and
there's
a
you've
yet
has
deployed
and
there's
a
queue
available,
etc.
Most
of
those
events
are
like
1,
K,
2
and
K
3
K
boys
fight
all
right,
and
those
sizes
are
mostly
due
to
very
long
URLs
that
are
in
them.
H
G
D
A
Okay,
Chris
nothing
at
pick
on,
even
though
your
hand
went
down.
Is
this
thing
you
want
to
say.
F
Post
a
link
in
the
chat,
that's
the
original
issue
that
has
like
a
couple
of
different
Oracle's
or
products
and
their
size
limits
all
over
the
place,
but
and
I
think
for
64
K
is
basically
the
lowest
so
we're
safe
on
that
we
don't
exclude
any
technology
and
that's
the
kind
of
the
point
where
I,
don't
why
would
fueled
size
guarantee
would
be
a
better
name
because
it
more
tells
us
that
this
okay
is
not
a
lower
limit.
Well,
it's
just
the
lowest
guarantee
that
we
can
give.
A
H
Think
so
what
I
was
basically
trying
to
get
to
see?
Do
we
really
need
to
set
up
a
maximum
limit
or
heed,
but
I
see
PV?
You
guys
are
saying
that
you
know
if
somebody
wants
to
have
a
larger
event
size,
they
can
still
do
that,
and
and
also
the
practical
numbers
are
much
lower
than
64
kilobytes,
so
I'm,
hungry
low-key,
okay,
cool.
A
Okay,
okay,
then,
circling
back
around
to
Eric's
original
question
or
concern
our
people.
Okay,
with
this
slight
wording,
change
of
this
sent
out
of
line
four
four
seven
two
ended
with
of
a
size
of
64
kilobytes
over
less
just
for
clarity,
sake,
Thank,
You,
Christoph.
You
know
anybody
else
want
to
speak
up.
Clemens
I,
assume
you're.
You
said
you're,
okay,
with
that
right,
I
already.
A
A
A
All
right
cool,
thank
you
guys
very
much.
Another
tough
one
behind
us.
Thank
you
guys
very
much
for
your
patience
on
that
and
then
push
it
already
excellent.
Thank
you.
Oh
I
mean
you
come
up
there
approved
with
wording
change
cool
all
right
next,
one
on
the
list.
This
one
was
mine.
This
is
just
a
reminder.
This
is
just
for
the
primer,
so
it's
completely
non
normative.
A
In
particular,
it
focuses
a
lot
on
producers
that
are
not
part
of
the
event
source
directly,
so
they're
acting
on
behalf
of
the
event
source,
and
you
know
how
they
should
populate
those
fields
and
stuff
like
that.
I
think
it
also
does
touch
a
little
bit
on
intermediaries
on
what
they
should
or
should
not
do
as
the
message
goes
through
them
and
at
a
high
level,
I
believe
I
basically
say
for
the
most
parts
in
tomato
really
shouldn't
have
touched
the
message
it
but
much
like
HTTP
proxies
it.
A
It's
okay
for
them
to
add
additional
properties,
but
generally
they
don't
touch
things
unless
there
was
a
very
good
reason
for
them
to
touch
a
certain
property,
because
generally
these
things
are
supposed
to
be
passed
through
is
because
the
end
receiver
of
the
message
or
they've
is
here
with
the
event.
In
this
case,
it
really
shouldn't
know
or
care
that
it's
going
through
an
agent
mediary
for
the
most
part.
That's
why
you
know
they
may
extra
stuff.
A
A
A
A
Okay,
what
I'd
like
to
do,
then,
is
ask
for
everybody
to
look
this
over
during
the
next
week
or
so,
and
hopefully
we
might
be
able
to
I
get
this
one
approved
next
week,
assuming
there
aren't
any
major
concerns
with
it,
god
this
one's
been
out
there
for
a
very
long
time,
so
I
know
the
guys,
walk
with
this
and
be
very
happy
to
get
this
one
in
there.
That's
old,
all
right
in
that
case
I,
try
to
think
of
anything
else
worthy
of
discussion
here.
Let
me
pick
on
Klaus
for
a
sec.
H
A
A
What's
interesting
is
I
looked
at
the
governance,
stock
and
I,
don't
believe
we
technically
require
a
week
for
approval
of
the
spec.
However,
votes
in
general
do
require
a
week
and
I
feel
like
approving,
of
a
version
of
the
spec,
is
kind
of
a
big
deal,
and
so
I,
don't
want
to
say,
I
feel
like
anybody
like
we
rushed
things
past
them.
So
what
I'd
like
to
do
is
this
is
once
we
go
through
the
process
of
actually
merging
these
approved
PRS
into
master.
A
What
I'd
like
to
do
is
send
out
a
note,
officially
kicking
off
the
1-view.
What
do
you
want
to
call
it
a
review
cycle
or
formal
vote?
It's
up
to
you
guys,
but
what
I'd
like
to
do
is
get
people
a
week
to
look
it
over
and
assuming
no
one
finds
anything
too
egregious
in
the
spec
that
can't
be
fixed
later
on
through
wordsmithing
I'd
like
to
do
a
close.
The
vote
down
next
week's
call
and
get
the
0.3
out
the
door
on
the
beginning.
Next
week's
call
is
that
sound,
ok
to
everybody,
it.
C
Sounds
ok,
doc.
Well,
I
have
a
question.
This
is
Roberto
I,
I'm
kind
of
struggling
a
little
bit
to
understand
how
I'm
supposed
or
everybody's
supposed
to
keep
up
with
it
with
respect
version
changes.
So
when
we
did
the
gdpr
events
for
using
cloud
events,
we
use
point
two
because
I
was
the
one
that's
there,
but
us
they
spank
evolves
over
the
next
months.
I'm,
not
exactly
sure
what
we're
supposed
to
do
to
keep
up.
A
D
D
D
F
Christoph
your
hands
up,
you
know
what
I'll
do
is
that
customers,
basically
we're
just
through
there.
Don't
only
say:
I
want
cloud
events,
but
I
have
to
specify
the
corrosion
of
call
events.
They
want
so
right
now,
I
support,
0.1,
yeah
and
then
I
think
once
because
we
also
integrated
banquet
I
wait
until
the
event
could
also
support
0.3,
and
then
people
can
basically
choose
which
version
they
want
to
get
for
some
time
and
then,
after
some
time
we
will
remove
the
old
versions
when
all
people
have
migrated
off
of
the
old
version
and
I.
D
Don't
you
know,
let
me
just
add
one
thing
because
no
tension,
so
what
we,
what
we
did
effect
this
you
as
you
subscribe
into
a
topic.
You
can
choose
what
caused
events
where
he
once
delivered
and
that's
the
thing
we're
going
to
change
so
effectively
when
you
walk
up
to
the
event
grid
and
you
subscribe,
you
can
say:
I
want
to
have
the
native
the
grand
event
grid
format.
D
A
D
C
I
know
but
I
mean,
like
everybody
says
it
requires
a
product
change
to
actually
make
it
I
mean
unless
you
make
it
a
choice
at
the
event,
a
meter
level
to
see
which
version
of
the
schema
you
want
to
comply
with
it
on.
In
our
case,
it
would
be
like
a
product
change
to
make
the
changes
in
the
spec
level,
so
I
need
to
figure
how
we're
gonna
do
this
at
some
point,
what
we're
planning
to
do
in
a
tied
adobe
is
to
offer
all
our
events
at
the
subscriber
level.
C
C
A
Let
me
pick
on
somebody
else
here
Tim,
you
actually
open
up
an
issue
in
the
serverless
repo,
as
opposed
to
the
connivance
repo
you
know,
but
it
is
a
con
event
issue.
You
actually
asked
a
very
similar
question
about
AWS
as
possible
support
for
this
and
whether
you
guys
should
wait
or
something
like
that.
Did
you
want
to
sort
of
summarize
your
concerns.
G
A
G
Geek
I
get
cold
chills
at
the
prospect
of
promising
to
support
something
that
isn't
finished
yet
so
so
why
honestly
I'm
not
100%
sure
what
the
best
in
AWS
could
do
is
now
I
guess.
One
thing
we
could
do
is
just
stay
with
our
native
event,
format
that
we
have
and
say,
oh
and
we'll
also
give
you
a
cloud
event
as
if
you
specify
what
version
you
want.
G
I'll,
be
honest:
I,
don't
like
that.
The
reason
I
don't
is
that
we
have
like
huge
numbers
of
people,
processing,
millions
of
events
and
they
all
every
one
of
them.
Has
the
down
field
names
hardwired
into
their
code
and
and
they
really
don't
want
to
think
of
a
version
numbers
and
super
version,
numbers
and
I?
Don't
really
want
to
think
about.
You
know:
maintaining
support
for
an
arbitrary
number
of
old
version
numbers
going
forward.
G
A
I
think
that's
a
good
question
and
it
oh
I
know
I
know.
A
I
saw
the
waiting,
let's
go
back
over
here
for
a
sec.
This
is
a
great
topic
since
we
have
20
minutes
to
discuss
it.
Obviously,
if
we
go
through
all
the
issues
and
we
don't
think
any
of
them
are
worthy
41.0,
that's
one
criteria
right,
but,
let's
see
so
technically
I
believe
we
would
have
reached.
We
would
have
done
our
zero
point
three,
because
that's
we're
going
to
be
voting
on
if
you're,
looking
what
we
have
for
zero
point,
four
additional
C
realizations
of
protocols,
I,
don't
believe.
A
A
A
Thomas
had
won
or
non
goals,
routing
I
think
I
haven't
back
in
double
check,
but
I'm
pretty
sure
there's
at
least
one
or
two
issues
out
there
that
might
fall
into
the
category
of
of
of
clarifications
or
things
that
run
into
like
the
PR.
We
just
closed
today
about
size
limits,
right,
actual
usage
of
the
spec
itself
and
whether
we
need
to
make
some
changes
based
upon
the
real-world
usage
I
think
we
have
some
issues
around
that.
A
So
if
we
can
get
those
behind
us,
in
my
opinion,
I,
don't
think
there's
any
reason
why
we
can't
jump
to
1.0
very
very
quickly.
Obviously
it's
up
to
you
guys
and
whether
you
think
some
of
the
real-world
experience
people
have
like
in
K
native
or
Adobe
and
Microsoft,
whether
that's
sufficient
experience
under
our
belt
to
say
yes,
we're
ready
to
consider
1.0.
A
Thank
you
other
thing
to
keep
in
mind
is
we
did
agree
that
at
some
point,
when
we
reach
the
equivalent
of
0.9
that
we
were
gonna,
let
the
spec
bake
for
a
little
bit
of
time
to
allow
people
the
time
to
get
some
real-world
experience
before
we
tag
it.
Officially.
1.0
now
it's
possible
that
K,
Natives
usage
and
the
other
guys
uses
is,
you
know,
counts
as
that.
It's
up
to
you
guys
to
decide
that
or
not
so.
The
net
of
this
is
I.
C
G
C
A
Don't
want
you
to
make
up
a
number
without
actually
looking
at
the
issue
backlog
first.
So
let
me
do
this.
Let
me
take
the
action
item
and
I'd
actually
like
everybody
to
do
this
as
well,
is
to
make
sure
my
analysis
is
it's
not
incorrect.
I'm
gonna
go
back
and
look
at
all
the
open
issues
and
pull
requests
and
make
sure
or
pull
out
the
ones
that
I
think
we
have
to
resolve
before
1.0
and
I
think
once
we
have
that
list,
then
a
next
phase
call.