►
From YouTube: CNCF Serverless WG Meeting - 2018-05-17
Description
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.
Join us for KubeCon + CloudNativeCon in San Diego November 18 - 21. Learn more at https://bit.ly/2XTN3ho. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.
B
C
D
A
Of
varun
am
headed
morning
good
morning,
vladdy
there
twice
yep
I'm
here:
okay,
let's
see
Roberto,
okay,
I,
don't
know
where
Bertos
last
name
I,
don't
think
we
have
a
Roberto
yet
on
a
John.
The
attendee
list,
but
I
could
be
wrong.
E
A
G
A
A
I
I
I
just
stared
at
it
the
the
PR
as
is
and
I,
could
not
remember
what
the
objection
was
and
okay
went
back.
I
actually
went
back
into
the
notes
and
looked
and
yes,
it
was
not
notice,
and
that
was
in
the
middle
as
we
were
preparing
for
the
funny
demo.
So
in
that
kind
of
okay,
I
apologize
for
fondly
no.
A
I,
probably
we
messed
up
and
taking
a
meeting
minutes
l
there
you
go
all
right,
let's
see,
Colin
are
you
there.
A
A
But
Laurie
are
you
there.
K
A
L
A
William
you.
A
A
Who's
that
last
one
Erik.
K
A
K
A
M
A
O
So
well,
I
represents
AWS
on
the
CNC
F
board,
so
I
do
watch
all
the
activities
closely
in
case
in
future.
If
there
are
any
issues,
anything
around
CNC
have
any
activity
related
to
that
I'm
always
available
on
slack
or
you
can
shoot
me
a
mail
at
a
RG
you
and
amazon.com.
So
that's
my
email
id
for
answer.
You
should
be
ml
now
for
this
activity,
the
version
that
is
been
at
this
point
of
time
in
0.1.
So
we
are
following
the
specification
we
have
looked
at
the
specification.
O
A
And
to
that
point,
I
did
at
that
item
to
the
agenda
profile
discussion,
so
I
wanted
sort
of.
Has
the
question
of
the
group
itself?
Does
anybody
have
any
customers
asking
for
this
I
think
we're
going
to
share
with
the
broader
group
to
help
people
answer?
That
kind
of
question
of
you
know
is
someone
asking
for
this
I.
H
D
F
O
It's
not
really
about
the
money
at
this
point
of
time.
You
know,
that's
not
how
we
look
at
any
product
or
any
project
we
get
involved
with.
The
key
thing
is
yes,
I
understand
the
influencer
part,
but
what
I'm
still
lacking
at
this
point
of
time
is
what
which
customers
any
service
I
mean,
as
I
said,
90
to
95%
of
our
roadmap
on
what
we
do
is
driven
by
customers.
O
So
the
reason
we
launched
lambda
is
because
customers
were
asking
for
the
longest
time
that
give
us
some
compute
capability
for
storage
and
then
now,
at
a
time
we
have
added
a
bunch
of
triggers
as
well.
The
reason
we
launched
eks
is
because
customers
are
asking
run
kubernetes
for
me,
so
if
I
were
to
justify
for
cloud
events
within
Amazon,
the
first
question
that
will
be
asked
by
the
management
is
which
customers
are
going
to
benefit
from
this.
What
are
they
benefit?
O
I
Have
we
have
you
know
in
our
in
a
repo
here
we
have
a
presentation
by
rocky
Rome
which
is
just
showing
how
much
dispatch
code
he
needs
to
write
inside
of
AWS
to
distinguish
between
all
the
events.
So
just
just
to
harmonize
all.
The
events
that
you
guys
are
yourself
are
throwing
around
in
the
AWS
platform,
which
have
absolutely
no
standardization
whatsoever,
will
be
a
giant
internal
benefit
for
you
guys.
G
Can
you
send
that
into
the
chat
I'd
like
to
look
it
as
well,
I
think
it's
important
to
the
what
we're
doing
is
important,
because
when
you
want
to
get
an
event
of
AWS,
do
you
use
SQS
SNS?
Do
you
have
to
create
a
lambda
and
get
it
triggered
directly?
There's
all
these
options,
I
think
does
mean
that
the
events
can
become
portable.
That's
going
to
be
really
useful.
The
customers
why
the
platform
they're
working
on
so.
O
N
O
C
O
C
O
C
B
Hi,
this
is
a
worry
from
pure
SEC.
I
wanted
to
also
clarify
on
the
fact
that
even
within
AWS,
the
the
the
difference
between
each
type
of
event
structure
makes
it
really
hard
for
third
parties
to
analyze
the
events.
Sometimes
you
have
to
use
heuristics
to
understand
where
the
event
came
from,
so
it's
definitely
missing
some
sort
of
standardization
that
allows
you
to
understand
what
kind
of
event
it
is,
what
what's
the
schema
of
of
this
specific
event,
type
and
so
forth,
so
that
that
would
make
it
I
think
much
easier
for
a
third
party.
J
O
All
right,
this
has
been
great
okay
input
to
me,
so
Nordstrom,
an
AIC
Arriba
and
early
from
pure
sec.
I
would
really
like
to
understand
more
on
use
cases
and
how
is
impacting
you,
and
once
I
have
a
better
understanding.
This
will
allow
me
to
kind
of
socialize
the
proposal
a
bit
more.
The
lambda
team
is
aware
of
the
cloud
events
proposal,
I
didn't
bring
it
up
to
them
and,
as
a
matter
of
fact,
I'm
giving
a
keynote
tomorrow
at
swamp,
Park
and
the
title
is
the
circle.
O
S
tidal,
wave
and
I
will
be
mentioning
about
cloud
events
over
there
and
the
problem
domain
that
is
trying
to
solve,
and
so
I'm
personally
I'm
for
supporting
cloud
events.
But
you
know
on
what
the
lambdas
service
team
is
going
to
do
and
how
it's
going
to
impact
our
customers
is
the
biggest
thing
that
I
need
to
understand
that,
just
by
the
time,
so
I'm
willing
to
cannot
explain
that
to
the
lambda
team.
O
What
I
could
do
is
I
could
fire
up
each
okay,
I'm
thinking
out
now
here?
That's
fine
I'm
thinking
that
each
customer
use
case
may
be
different
and
I'm,
not
sure
how
much
they
are
willing
to
talk
about
it
in
public,
so
I'm,
okay,
if
they
want
to
do
it
and
totally
I,
would
rather
preserve
their
own
use
cases.
So
we
are
very
vigilant
about
customers,
security.
So
that's
the
reason
I.
A
Okay:
okay,
what
if
you've
sent
a
public
email
asking
for
the
data
you're
looking
for
and
if
people
don't
feel
comfortable
sharing
it
publicly,
they
can
email
back
to
you
privately,
but
at
least
for
those
people
who
are
comfortable
sharing
it
publicly.
At
least
then
other
people
can
see
it
and
use
that
data
for
their
own
purposes
as
well
to
help
convince
others
of
the
value
yeah.
A
There
is
14c
if
in
general,
I,
don't
think
we
have
one
just
for
our
group.
Our
Mentalist
is
pretty
pretty
non
active,
so
I,
don't
think,
there's
a
problem
with
flooding.
It
should
I
send
an
email
or
today's
in
a
slack
message.
I
would
do
an
email,
because
not
everybody
watches
the
stock
very
closely.
Gotcha.
Okay,.
O
E
J
O
A
Thank
you
very
much
and
thank
you
guys
for
speaking
up
there.
Things
are
really
useful,
all
right.
So
moving
on
the
agenda,
I
believe.
Last
time
we
talked
about
trying
to
come
up
with
modifications
for
our
milestones,
I'm,
not
quite
sure
of
the
best
way
to
move
forward
other
than
to
say,
I
did
put
forward
a
proposal
earlier
today
and
I
know
it
was
just
put
out
there
today.
A
So
I'm
not
gonna,
ask
any
rights
unless
I
vote
on
it,
but
I
did
put
out
some
ideas
for
how
to
categorize
what
go
on
and
do
going
forward
on
a
sec.
Hopefully
you
guys
did
see.
My
email
from
yesterday
basically
moved
everything
into
point,
one
that
we
did
from
before
and
then
point
two
is
this
stuff
here?
Basically,
what
I
tried
to
do
is
to
group
things
in
terms
of
the
items
that
I
thought
impacted
the
spec,
the
most
meaning,
the
most
breaking
most
serious
breaking
changes.
A
I
tried
to
move
up
to
the
top
and
the
things
that
were
either
sort
of
on
the
optional
line,
or
clarifications
I
move
those
little
bit
down,
but
I
was
wondering
if
people
wanted
to
take
time
just
to
review
that
off
line
because
I
don't
know
if
I
want
to
do
it
right
now,
but
other
than
that
are
there
any
suggestions.
People
have
in
terms
of
ways
to
move
forward,
to
lay
out
our
roadmap,
going
forward,
I'm
open
to
any
ideas.
Basically,
at
this
point.
J
D
J
H
J
A
Yeah,
like
I,
said
I,
don't
I
have
no
intention
of
asking
for
a
vote
during
like
that
right
now,
I
just
wanted
to
know
if
there
was
an
alternative
proposal
or
path
forward
to
developing
the
list.
That
people
would
like
us
to
consider
or
does
wordsmithing
AMA
we
have
here
is
a
sufficient
way
to
go
forward.
So.
J
How
about
give
us
two
weeks
and
if
we
have
like
some
some
new
items,
I
did
we
can
pour
PR
or
we
just
sorry,
okay,
how
we
should
do
this.
A
Yeah,
if
you
want
to
see
changes
in
here,
I'm
very
open
to
putting
a
comment
on
the
on
the
PR
and
I'll
I'll
make
the
changes.
That's
not
a
big
deal.
Okay
sounds
good!
Thank
you!
Okay,
so
you
had
mentioned
two
weeks.
There
is
it
okay.
If
we
try
to
shoot
for
next
week,
but
people
need
more
time,
then
we
can
go
another
week.
I'd
rather
be
aggressive
and
then
slip
a
little
rather
than
assume.
A
full
two
weeks
is
that
okay.
B
A
I
I
don't
think
it's
critical
for
us
to
move
forward
because
we're
gonna
address
PRS
as
they
come
in
anyway,
but
we
do.
We
should
probably
lay
down
on
paper
or
someplace
what
our
plans
are.
So
let's
do
this
then,
for
Kathy's
request.
Well
well,
may
be
briefly
discussed
next
week,
but
look
for
a
vote
in
two
weeks
is
that
okay.
A
A
Okay
with
this
is
ask
people
to
spend
just
a
couple
minutes
and
I
really
do
mean
a
couple
like
two
or
three
per
item
had
it
here
just
to
briefly
summarize
it,
and
then
we
could
talk
about
next
possible
steps
here
in
terms
of
determining
what,
if
anything
you
want
to
do
next
is
that's
not
okay,
yeah,
okay,
mattering
objections,
Kathy
I
think
this
first
one
right
be
yours.
Is
that
true?
Yes,
okay,
so
maybe
there's
a
two
or
three
minute
summary!
Oh.
J
Okay,
oh
you
asked
me
to
give
a
summary
okay,
so
there
have
been
several
presentations
in
the
in
the
recent
CSF,
the
keuken
conference
about
the
you
know:
function,
workflow
or
function
composition,
people
give
it
different
names.
What
it
really
means
is
you
know
the
user,
you
know
if
your
it
needs
a
way
to
specify
their
service
use
case
workflow,
for
example.
You
know
some
use
case
could
be
like
do
a
image.
J
You
know
on
face
recognition
on
a
photo
and
the
photo
is
uploaded
onto
the
scroll
storage
and
then
another
use
case
could
be
in
could
involve
you
know.
Multiple
steps
of
the
function
functions
either
executed
in
sequence
or
in
parallel,
and
also
you
know,
could
also
you
know
evolve.
You
know
the
second
function,
wait
for
another
event
and
to
happen
and
then
execute
that
function.
J
So
so
I
think
you
know
we
need
a
standard
way.
It
will
be
used.
It
would
be
good
for
the
user
if
we
can
have
a
standard
way
or
consistent
way
for
the
user
to
specify,
and
they
are
workflow,
for
example,
to
specify
what
combination
events
should
go.
What
functions,
how
those
functions
are
executed
and
what
information
is
passed
from?
One
function
to
the
next
function
and
also
like
you
know
whether
the
next
step
function
execution
needs
to
wait
for
another
event
to
happen.
J
For
example,
we
already,
like
you
I,
mean
I
like
issue.
Is
we
discussed
a
correlation
ID
right,
I?
Think
that
correlation
I,
the
Association,
is
associated
with
a
use
case
workflow
and
do
need
to
be
specified
in
the
workflow
specification,
like
you
know,
if
that
workflow
in
multiple
events
and
if
there
are
multiple
instances
of
each
event
and
so
how
we
correlate.
You
know
those
in
my
instance
together
to
the
right
and
workflow
execution
instance.
I.
Think
I
think
that's
a
motivation
for
for
this
work.
I.
This
work.
A
J
G
J
Yeah
yeah
I
mean
many
start
working
on
this
yeah
we're
going
to
yeah
and
we're
aware
of
about
you
know
is
that
function
or
other.
You
know
as
Microsoft
is
your
function
and
also
always
you
know,
laughs,
yeah
I'm,
aware
of
that
we
can,
we
can
be
settings
will
be
really
start
to
work
on
this.
Those
will
be
our
reference
from.
G
It
from
the
perspective
of
the
workgroup
do
you
do
people
think
that
function
workflows
are
more
important
than
the
signatures,
seeing
as
nordstrom
are
delaying
the
adoption
of
lambda,
because
it's
so
hard
to
move
from
one
cloud
to
another.
Would
it
make
sense
to
to
see
which
of
those
two
items
is
more
important,
whether.
A
A
Alex
I
think
that's
going
to
be
a
question
we
have
to
ask
once
we
get
done
well
quickly,
evaluate
it.
We're
talking
about
each
of
the
items
on
the
list
is
to
figure
out
which
ones
should
come
first
versus
second
and
prioritize
them.
So
I
think
that's
that
that's
so
the
next
question
after
this
review,
yeah.
Okay,
all
right
any
other
questions,
comments
for
Kathy
before
we
move
on
to
the
next
one:
okay
event,
orchestration
and
changing
Kathy
was
this
year's,
or
is
this
somebody
else?
I
can't
remember
I!
Think.
Q
I
wrote
part
of
this
when
it
was
still
combined
with
what
Kay
Kay
talked
about,
but
this
is
I
can
talk
about
it,
but
I
still
am
new
to
the
working
group,
so
I'm
myself
other
wrong
stuff.
But
there
are
a
lot
of
questions
about
how
events
and
functions
are
transmitted
and
chain.
There
is
bad
cue
come
face
to
face.
There
was
the
discussion
about
the
light
bulb
or
an
event
you
you
have
an
event,
but
an
actual
device
in
a
room.
Q
That's
the
generating
an
event
now
well,
a
light
bulb
might
not
know
what
room
it's
in
if
it's
in
a
warehouse
or
a
flat,
and
you
might
have
even
gateways
forwarding
that
event.
How
is
that
bank
are
they
allowed
to
modify
the
event?
Do
they
generate
a
new
event?
What
is
happening,
and
there
are
a
lot
of
questions
on
this
and
people
wanting
to
just
before
the
meeting
Doug.
Q
You
responded
to
a
PR,
for
somebody
wanted
to
add
the
source
attribute
for
IOT
events
to
help
with
adoption,
because
they
did
mean
that
it's
the
whole
thing
we
have
an
event
has
been
generated
and
want
to
forward
it.
Maybe
add
some
word
details
to
it
again
in
relation
to
the
correlation
ID
issued
of
capital
throughout
and
there's
a
lot
of
questions
here
and
it's
unclear
who's
allowed
to
modify
what
how
events
are
forwarded
and
there's
been
some
discussion
about
putting
everything
in
headers
and
not
necessarily
having
labels
again.
Q
This
is
still
very
tied
up
with
a
correlation
ID,
but
then
again
at
the
face-to-face,
the
issue
of
encrypted
or
signed
the
payloads
was
brought
up
and
if
the
payload
is
encrypted,
obviously
the
even
gateway
could
not
do
that
much
unless
there
are
some
labels
that
are
outside
this.
This
needs
to
be
discussed,
because
there
are
a
lot
of
questions.
I.
I
Think
we
agree
that
we
want
to
go
and
amend
the
proposal
that
Thomas
made
for
that
to
be
properties
that
are
associated
with
the
message
and
then
we're
also
going
to
have
an
annotation
property
back.
So
I
thought
that
was
something
that
we
had
kind
of
semi
resolved
at
least
that
there
will
be
proposal
that
we
can
go
and
then
go
and
discuss.
I
J
I
think
we'll
agree.
This
will
be
part
of
the
cloud
events
it's
back.
First
I
think
this
is
a
very
important
thing,
but
I
think
we
agree
where
this
will
be
part
of
the
event.
Spec
will
not
be
a
separate
item
also
talked
about
I.
Think
Thomas
has
a
sauce
label
proposal.
I
think
he'll,
add
also
has
something
which
put
into
my
PR
some
suggestion.
Putting
my
my
carnation
DPR,
we
can
probably
look
at
you
know.
Do
this
so.
A
It
sounds
like
there
may
be
some
things
that
fall
back
in
scope
of
the
spec
and
but
it's
like
I'm
glad
you,
you
think
there
may
be
some
items
that
we're
falling
outside
of
scope
and
I'm
wondering
whether
what
we
should
do
is
assume
they're
in
scope.
Right
now
get
people
to
open
up
PRS
for
those
features,
and
then
once
you
look
at
the
PR
is
then
they
can
make
the
determination
about
whether
it's
in
or
out
of
scope
and
if
they're
out
of
scope,
then
we
can
push
it
back
to
this
list.
Q
Think
this
is
still.
There
was
the
issue
of
okay.
We
do
set
a
bag
of
events
or
labels
or
whatever,
and
we
do
that
in
cloudy
banks.
That's
in
scope
of
cloud
events,
but
if
we
do
that
we
don't
specify
okay,
there
is
just
a
bag.
You
put
everything
you
want
in
here:
oh
I
do
think
there
there
is
need
to
say
okay,
what
are
we
gonna
put
here?
A
So,
okay
I,
don't
want
to
get
into
a
deep
discussion
about
how
to
solve
this
particular
problem.
But
what
I
think
I'm
hearing,
though,
is
at
least
part
of
this-
may
follow
then
scope
of
the
spec
so
so
like.
Maybe
it
would
be
best
if
we
discussed
this
within
the
scope
of
the
specification
until
we
definitely
decide
it's
not
in
scope,
because
I
don't
want
to
leave
something
out
if
it
does.
If
we
do
think
it
should
be
there
yeah.
J
A
Okay,
okay,
okay,
any
other
questions
or
comments
on
that.
One
for
you
on
the
next
one:
okay,
functional
signatures,
I
came
over,
who
added
this
one
but
you're
on
was
that
you
or
is
that
somebody
else
I.
G
I
think
any
of
us
could
really
speak
to
this
it's
this
is
when
you
like,
like
we
had
from
Lourdes
from
you,
you're
thinking
of
adopting
surplus.
You
start
writing
some
code
for
a
sure,
and
then
you
realize
that
it's
not
going
to
work
out.
You
want
to
move
it
to
AWS
and
you
can't
because
the
signatures
are
all
different
and
the
events
are
different,
so
this
is
about
finding
a
way
to
get
a
common
ground.
A
A
G
Legitimate
I
think
that's
true.
The
other.
The
other
thing,
though,
is
that
we
have
this
implicitly
and
people
already
have
committed
to
that.
A
juror
or
the
opens
has
API,
and
that's
a
fact.
It's
just
that.
If
there's
now
a
problem-
and
there
may
be
some
projects
that
can
find
a
common
ground
and
then
moving
between
them
would
be
incredibly
useful
as
an
end
user,
I
am
concerned
as
well
about
breaking
the
interface
so
with
everybody.
That's
got
so
far.
Have
people
actually
using
this
and
making
use
of
it.
G
A
Okay,
but
I
think
everybody
understands
that
what
we're
talking
about
here,
I,
don't
think,
there's
any
confusion
about
that.
So,
let's
move
on
to
the
next
one
API
is
for
accessing
cloud
events
and
urine.
I
think
this
one
might
be
yours
in
conjunction
with
the
next
one.
Is
that
true
API
is
for
accessing
cloud
events
and
then
the
cloud
event,
Client,
SDK
I,
think
those
might
be
related
and
I
think
they're.
Both
yours
aren't.
They.
P
P
Yes,
you
know
each
one
of
the
frameworks
have
or
most
of
the
frameworks
have
a
logging
facility
like
if
you
go
to
Amazonia
bus
I'm
logging
facility,
that
writes
things
to
crowd,
watch
eventually
go
to
nuclear.
We
have
something
that
can
export
to
a
variety
of
logging
services,
so
there's
Asia
and
others
so
when,
if
you're
starting
to
write
the
function,
so
you
focus
on
the
events
and
the
signature.
Everything
is
cool,
but
then,
within
your
functions,
you're
calling
a
logging
facility
in
each
one
of
those
logging
facility
have
its
own
proprietary
API.
P
That
means
that
every
every
place
you
put
debug
print
or
logging,
etc.
You
have
to
modify
your
code
or
you
have
to
build
some
wrappers.
So
if
we
come
with
a
suggestion
for
a
common
API
for
locking
and
that
you
know,
we
can
also
build
wrappers
to
the
different
log
implementations
within
the
different
function.
Providers
and
people
can
update
these
semantics
and
for
the
open
source
community.
P
We
can
also
support
that
in
all
trying
and
converge
on
one
definition:
it's
not
only
for
the
function,
implementation,
but
also
we
can
have
a
facility,
especially
on
the
open
source
community,
where
you
could
have
like
log,
consumer
or
logging
service
that
consumes
the
function
event.
So
it
couldn't
be
kubernetes
okay,
but
it
may
be
a
direct
injection
of
a
log
entry
into
something
like
elasticsearch
or
cloud
lot
or
a
paean
sites,
so
that
that's
the
proposal
there's
also
we're
using
something
from
uber
called
there's
app
and
I've
also
listed
how
it
works.
P
So,
if
you're
interested
in
looking
into
it
but
yeah,
it's
it's
more.
The
point
of
trying
to
standardize
the
way
are
doing
logging.
We
can
expand
it
later
to
also
observing
maybe
no
open
tracing
serve
API
built
into
the
function
context,
or
maybe
some
programmable
counters.
You
know
there
would
be
pushed
to
things
like
prometheus.
I
I'm
just
I'm
wondering
how
that
path,
which
is
like
so
Frances
in
narrative
functions
as
a
first
is
logged
into
the
azure
monitoring
pipeline,
which
is
used
by
all
the
other
200
services
as
well,
and
so
I'm
wondering
how
this
is
specific.
To
what
we're
trying
to
achieve
here
with
in
the
working
group
is
that
special.
P
Alleviates
the
complexity
from
the
programmer
side
of
trying
to
integrate
with
each
one
of
the
logging
facilities.
You
know
II
what
writes
the
code
once
you're
aware
but
locally.
You
can
also
publish
events
to
a
juror
app
insights,
so
it
is
possible
for
for
service
frameworks
to
publish
to
multiple
there's.
G
A
bigger
question
about
logging,
as
well
as
like
what
what
do
you
log,
especially
if
you're
working
with
a
big
payload
I,
think
people
that
are
doing
logging?
Maybe
you
haven't
thought
about
that.
You
have
a
ten
megabytes,
binary
response
from
a
function
or
input.
Even
are
you
really
going
to
log
that
and
what.
P
I
P
Yeah
but
the
flag,
but
that's
why
you
have
an
explicit
law,
explicit
logging
API
in
many
places,
and
not
necessarily
just
saying
you
know,
everything
that
is
standard
out
is
being
logged
and
you
may
even
have
different
severity.
That's
why
I
think
once
we
dive
into
how
to
define
it,
we
may
do
a
good
job
in
defining
it
in
a
way
that
it
solves
those
problems.
A
P
So
see,
nature
is
one
thing,
but
let's
assume
we
want
a
provision
of
function
which
includes
creating
aspect.
You
know
maybe
some
code
artifacts
and
publishing
into
a
service
framework,
so
there's
no
real
standard
around
it.
There
are
a
couple
of
efforts,
you
know
sam'l,
which
is
Amazon
and
is
really
very
confined
to
an
Amazon
set
of
services
and
the
other
one
is
from
serverless
Inc.
P
That
one
doesn't
necessarily
have
a
common
you
know,
set
of
API
is
across
the
different
functions,
because
this
function
is
different.
So
if
we
come
with
a
service
pack
of
this
is
how
you
would
define
in
a
pretty
much
like
a
pods
back
or
a
deployment
spec.
Where
you
have.
You
know
the
same
deployment
across
different
deployments.
You
know
we
can
create
like
a
function,
spec
that
may
even
encompass
other
attached
services,
like
databases,
message
queues,
api
gateways,
etcetera
that
could
ease
the
burden
of
provisioning,
the
same
function
across
multiple
providers.
J
I
think
this
is
a
interesting
topic,
also
important
topic,
because
from
the
user
side
you
know
if
the
user
does
not
want
to.
You
know
to
construct
different
deployment
scripts
for
different
platforms.
So
we,
if
we
can't
define
some,
you
know
common
and
deployment
language
or
not
language
primitives.
It
will
be
easier
for
the
user
to
to
be
portable
cloud.
Service
console
is
powerful,
but
I
think
you
know
it's
I,
think
you
know
it's
not
the
deployment
it
not.
That's
not
just
involve
function.
It
also
involves
the
events.
J
B
I
Think,
given
where
we
are
in
terms
of
developments
and
in
velocity
of
the
space,
it'll,
probably
be
more
successful
to
have
a
framework
that
kind
of
sense
inside
of
those
platform
it
makes
it
makes
them
even
and
then
is
quick
to
adapt
rather
than
trying
to
it'll
make
all
the
cloud
providers
follow
a
common
standard
which
then
would
have
to
go
and
be
adapted.
I
think
I
understand
the
motivation
behind
the
bad
this
one.
The
coming
function
was
the
common
logging
model.
I
just
found
it
I
found
it
pretty
hard
to.
J
G
Layer
I've
seen
this
in
this.
What
this
is,
basically,
the
premise
of
what
Austin
is
doing
with
his
work
of
Service
Inc
is
to
be
that
common,
it's
yeah,
mph
I,
think
the
Yama
file
at
the
moment,
and
he
you
know
it
explains
me
how
many
problems
they've
been
because
every
providers
just
so
different
with
all
the
resources
you
have
to
set
up
and
they
really
focused
on
only
doing
service
functions.
Is
there
anyone
from
his
from
his
team
here,
I.
I
The
hard
reality
of
this
of,
like
you,
can
standardize
some
pieces
and
I
think
we
can
even
find
like
on
the
on
the
top
level
function.
Signature
piece,
for
instance
it
like
there's
this
I,
looked
at
the
proposal,
there's
all
the
different,
the
different
signatures.
You
can
probably
find
a
way
to
make
it
all
look
very
much
the
same
in
node.
I
But
then
then
you,
you
click
the
click
ones
and
look
at
the
next
level,
and
that
again
is
different
because
of
the
underlying
infrastructure
and
all
the
configuration
stuff
and
all
this
is
in
a
way
of
platform
dependent
and-
and
it's
also
constrained
by
by
rules.
Then
the
platform
makes
and
that's
not.
For
instance,
in
Azure
the
functions
team
seems
to
decide
right
there.
Their
management
API
has
a
certain
shape
because
that's
the
same
thing
across
the
hundreds
of
azure
services,
so
that
there
are
Liberty
to
go
and
break
out
of
that.
A
Let's,
let's,
let's
move
on
those
I,
don't
dive
too
deep
into
each
one
of
these,
yet
I'm,
just
thinking
render
stands
with
the,
but
the
proposal
here
is
that
the
high-level,
but
we
have
one
more
to
go
through.
Then
we
can
talk
about
next
steps
and
how
we're
going
to
resolve
what
to
work
on
and
talk
about
that
process.
So
you're
on
you
want
talk
about
the
last
one.
I
think
that
one's
yours
as
well.
P
G
It
needs
to
be
some
parameters
around
that
when
you
benchmark
something
on
on
a
cloud,
I
mean
compare
that
to
an
open
source,
France
there's
a
very
big
difference,
but
perhaps
at
least
a
common
ground
should
be
that
they
are
all
running
on
kubernetes
or
they
all
have
the
same.
Api
I
saw
some
of
the
the
benchmarks
there
from
the
link
and
they
weren't
done
with
kubernetes.
It
was
a
rocket
socket,
I
believe
on
local.
G
P
Agreed
and
if
you
look
at
how
speck
vert,
for
example,
works-
it's
not
even
those
even
from
a
sense
of
an
ROI
okay,
so
it's
like
for
a
given
cost.
What
do
you
get
so
it
doesn't
allow
you
to
normalize,
also
the
cloud
provider
solutions.
So
there
are
different
ways
and
not
saying
that
we
know
the
answer
I'm
just
saying:
let's
sit
together
and
define
how
we
measure
it.
So
customers
can
say:
okay.
A
G
G
A
That
works
are
people
ready
to
I
guess
you
could
also
didn't
have
a
conversation
within
that
issue
too.
So
you
don't
have
to
worry
about
the
moas
thing.
So
what
about
that
idea?
I
like
that
one?
What
if
we
open
up
an
issue
for
each
one
of
these,
and
then
people
can
vote
and
comment
on
the
issue
and
have
a
little
back
and
forth
and
when
things
seem
to
die
down,
then
we
can
look
at
it
and
say:
okay
time
to
resolve
the
voting.
L
A
Any
other
suggestions:
okay,
okay,
in
that
case,
are
there
any
other
things
that
you
want
to
bring
a
relative
to
this
welted
to
this
whole
idea
of
picking
out
the
next
work
stream
to
work
on
kind.
G
Of
relative
to
that
relative
to
the
0.1
work
stream
and
the
demo
that
was
shown
at
cuba,
cotton
is
anybody
able
to
say
which
event
sources
have
implemented
cloud
events
to
some
extent.
I
know
that
event
grid
from
as
your
house
and
we've
used
that
I
heard
that
AWS
had,
but
I
I
couldn't
find
any
anything
about
that
online.
N
G
H
I
G
G
G
A
Let's
take
that
discussion
offline
if
we
can
yep
okay
cool
in
that
case,
with
a
whopping
ten
minutes
left
I'm,
not
sure
we
have
time
to
dive
too
deep
into
a
really
meaty,
PR
and
I.
Think
most
of
them
actually
do
involve
a
fair
amount
of
discussion,
except
for
the
first
one,
which
I
think
is
strictly
syntactical
and
it's
the
idea
of
moving
the
data
out
from
under
the
context
attributes
into
a
another
section.
So
it's
presented
as
a
sibling
to
all
the
other
attributes
which
are
metadata
about
the
event
itself
was
weren't.
A
I
A
A
A
Like
to
I,
don't
want
to
rush
it.
Okay,
so
me,
that's
the
question,
then:
is
there
any
objection
to
adopting
this
PR
all
right
cool?
Are
there
any
other
PRS
people
can
think
of
on
this
list
that
are
relatively
small,
otherwise
I'd
like
to
defer
the
PD
discussions
for
later?
We
have
more
time
because
I
think
I
only
have
like
eight
minutes
left
I,
don't
think,
there's
anything
small.