►
From YouTube: CNCF Serverless 10.12.17
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.
A
A
Last
time,
I
ate
suggests
going
to
trip.
My
wife
didn't
talk
to
me
for
three
days,
so
I
can't
took
that
as
a
hint
of
not
going
anywhere
for
a
little
while
she's
like
you,
joined
MasterCard,
to
stop
traveling.
Why
I've
been
gone
all
this
weeks,
I'm
like
I,
don't
know
that's
what
am
I
supposed
to
do.
I.
C
D
A
C
A
Know
I'm
I
just
got
a
team
in
India,
so
I'm
going
to
prune
A's
in
PI
the
next
month
and
then
we're
doing
some
stuff
with
Ollie
clouds
I'll
be
in
China
for
next
month
or
two.
E
A
Everyone
Hey
so
I'm
gonna
hit
about
five.
A
time
writing
it
started.
I
think
most
everyone
that
I
expected
it
beyond
is
on.
You
know:
Doug
had
a
conflict
today
and
it's
not
gonna
make
it
in
my
peak
also
I
had
a
conflict,
wouldn't
if
he's
gonna
make
it
so
there's
sort
of
a
couple
of
things
on
the
agenda
I'm
going
to
talk
through
the
first
one
is
just
in
terms
of
the
the
review
of
the
of
the
current.
What
product
about
group
so
I
think.
A
Last
week
we
see
we'd
have
about
three
weeks
that
we're
just
gonna
leave
the
document
out
there
for
you
guys
to
review
it
and
to
look
at
it.
We
accepted
the
proposed
changes
that
Sarah
made
we
removed.
The
word
absolutely
and
Chris
from
Amazon
I
think
is
on.
I
was
okay
with
that
with
us.
We
moving
that.
We
move
in
that
word
and
then
we
added
usually
for
yarn
suggestion.
A
A
Yeah
you're
good,
what
makes
they
have
till
the
end
of
end
of
this
month
to
sort
of
finalize
the
documentation
and
then
I'll
share
with
you
guys.
The
president
I
start
kind
of
presentation,
I'll
put
it
out
in
the
and
no
I
have
access
to
Google
Documents
again
at
MasterCard.
I
can
post
that
out
in
Google,
and
you
guys
can
see
you
put
comment
comments
on
it.
So.
A
This
would
be
the
second,
the
second
presentation
to
the
scene,
to
the
TOC
and
so
I'm,
going
to
try
to
focus
I
mean
they
they've
had
Chris
and
sent
out
a
reminder
to
them.
Probably
a
month
ago
now,
afterward
and
the
TOC
meetings
were
on
I
sent
out
requested
them
to
look
at
it.
Probably
a
month
before
that,
and
so
they've
had.
A
You
know
a
couple
months
to
take
a
look
at
what
we've
been
working
on
and
so
I,
don't
think
I,
don't
think,
there's
anything
new
and
what
we
present
them
other
than
what
our
next
steps
are.
So
probably
as
we're
finishing
up
the
documentation
aspect,
if
I
discuss
next
step
at
our
next
meeting
in
Doug
time
and
Mike's
able
to
make
it
we'll
talk
through,
what
do
we
want
to
propose
to
the
TOC
as
well?
That's
what
we
want
to
do
is
our
next
steps
with
the
hook,
booth.
A
It's
if
I
see
him
he's,
he
didn't
win
the
car
today,
but
we
have
somebody
who
is
working
on
that
and
I'll
have
an
update
on
that.
Probably
another
week
or
two,
the
the
initial
thought
was.
It's
is
too
large
for
a
white
paper,
but
what
we
kind
of
a
kind
of
explained
that
were
you
know
we
set
it
up
to
be
a
bunch
of
individual
components
and
sections.
So
it's
not
really
meant
to
be
one
big
large.
A
A
Yeah,
that's
a
good
question.
Kathy
I
am
was
probably
gonna
use
our
next
meeting
to
kind
of
talk
through
a
rough
draft
of
the
slides
and
and
Moke
it
might
like,
mostly
with
the
focus
on
what
do
we
want
to
ask?
What's
our
ass
back
to
the
TOC,
and
so
I
would
say
that
first
draft
will
be
sent
out
before
I'm
meeting,
which
we
had
a
weekly.
Can
this
now
we
going
back
to
bi-weekly?
Do
you
guys
know
we're
weekly
write
down
weekly
now?
Okay,
so
yes,
I'll
be
next
week.
A
A
I
B
D
B
H
G
H
H
As
you
know,
the
guy
that
makes
the
most
phone
calls
server
pushes
a
standard
and
I
think
there
are
some
deficiencies
in
one
versus
the
other
and
we
should
talk
about
and
potentially
we
can
converge
because
they're,
not
big
but
I,
think
it's
something
worth
discussing
and
unnecessarily
saying
you
know
what
we
talk
to
all
the
world
and
now
we
have
a
proposal
and
everyone
agrees
because
I
think
people
need
to
be
aware
of
what
are
the
downsides
of
one
approach
versus
them.
Yeah.
E
C
E
C
J
Is
that
to
me
Chris
months
or
Chris
a
month,
yeah
I
mean
I.
Think
a
coab
meetings
go
ahead,
said
drop
the
word
absolutely
just.
Have
it's
a
no-cost
to
an
idle
I?
Don't
think
we
need
that
strong
of
a
of
a
phrasing
in
their.
The
word
usually
seems
kind
of
ambiguous,
though
and
I
don't
know
if,
like
if
I
was
reading
this
from
there
like
an
executive
point
of
view
and
I
saw
usually
that
sounds
like
an
interesting
caveat,
where
we'd
probably
want
to
have
some
sort
of
clarification,
but
happy
to
get
rid
of.
J
E
H
One
second,
so
even
Amazon
today,
you
know
you
get
free
amount
of
implications
and
tomorrow
you
may
choose
to
give
I,
don't
know
like
a
monthly.
You
know
X
number
of
invocations
for
20
bucks,
and
you
want
to
go
beyond
that.
So
so
this
association
with
the
invocation
and
the
cost
is
a
derived
from
a
business
now,
not
necessarily
from
the
fact
that
you're
not
managing
servers.
E
It's
only
an
implementation
detail
and
there
is
a
cost
idle
and
it's
hidden
or
absorbed
so
I
think
when
you
come
to
use
an
open-source
framework,
you
deploy
your
ops
team
or
deploy
at
the
zero
cost
to
the
developer.
Now
there
is
a
capacity
in
there.
There
is
over-provisioning
happening
just
like
with
AWS,
but
we're
looking
at
technology
here,
not
not
a
business
model
or
not
a
billing
model
and
I
think
that's
too
deep,
because
it
rules
out
the
other
projects.
It.
J
Really
does-
and
it
we've
been
talking
about
this
now
for
almost
two
months
right
and
I-
think
the
thing
is
if
this
paper
was
a
fast
paper,
that
it
makes
all
sense
to
include
everyone.
But
if
you
want
to
talk
about
service,
there
are
benefits
that
extend
just
beyond
the
usage
of
the
technology
from
a
developer's
perspective.
J
Otherwise
you
start
being
able
to
very
directly
compare
this
to
any
container
framework
or
almost
any
pass,
that's
out
there
and
then
the
word
is
almost
meaningless
in
might
as
well
just
talk
about
paths
or
containers
as
a
service
type
frameworks,
so
it
just
starts
to
become
kind
of
a
meaningless
label.
So
from
from
our
perspective,
we
see
no
ability
for
you
to
separate
talking
about
what
service
means
without
talking
about
that.
The
end
impact
to
the
entirety
of
the
customer,
using
it
both
operations
and
development.
J
Is
that
there's
a
drastic
shift
in
how
you
think
about
it,
that
is
impacted
by
the
business
benefits.
It's
it's
not
just
a
hammer,
there's
a
lot
of
other
characters,
six
that
go
along
with
it
and
so
I
mean
again
I.
Think
what
we've
seen
what
I've
yet
to
see
is
anyone
come
up
with
a
better
example
of
this?
That
gives
you
a
thing
that
defines
service
that
isn't
just
defining
fast.
That
also
includes
back
in
there's
a
service
and
even
potentially
database
as
a
service
type
service
layers.
Again
that
isn't
just
fast.
J
E
I,
don't
I,
don't
agree,
I
feel
like
what
we're
trying
to
do
is
trademark
the
term
service
just
for
the
cloud
infrastructure
providers.
You
do
have
all
of
these
costs.
You
do
have
all
of
these
provisioning.
You
do
have
cost
at
rest.
It's
just
that.
You
hide
it
now.
If
an
ops
team
on
enterprise
team,
maybe
Kathy,
set
up
a
data
center
and
they
gave
their
ops
team,
they
gave
it
developers
one
of
the
open
source
projects.
They
would
have
server
list
with
all
these
attributes.
E
H
Alex
I,
don't
necessarily
think
it
just
cloud
I
think
it's
again,
it's
an
operation
model.
If
someone
goes
and
packages
fair
solution
with
a
self-service
portal
answerable,
you
know
server,
sealed
approach,
you
know
think
like
a
juror,
ezra
stack
is
a
hosted
platform.
You
know
if
it's
sealed
and
you
know
one
is
sort
of
touching
and
tuning
the
two
at
their
knobs,
and
I
know
about
the
containers
underlying
architecture.
Then
it
is
service,
even
if
it's
on
pram,
I
think
the
difference
is.
Is
someone
aware
of
the
underlying
technology
when
it's
deploying
this
functions
right.
H
There
is
one
point
here
again
about
the
capacity
you
know
most
people
today,
this
cell,
for
example,
you
know,
storage
like
dropbox
and
others.
They
they
give
you.
You
know
for
20
bucks
you
get
hundred
gig
and
tomorrow,
even
amazon
may
choose
to
you,
know
for
20
bucks,
you're
gonna
get
thousand
invocations.
That
actually
violates
this
statement.
Right
now,
because
you're
saying,
let's
assume,
I
paid
20
bucks
for
a
thousand
invocation
I
haven't
used.
H
J
So
I'm
curious
how
you
guys
define
this
differently
than
any
organization
for
the
last
10
or
so
years.
That
is
built
up
a
sophisticated
enough
paths
internally,
where
the
developers
have
no
visibility
to
the
underlying
infrastructure,
because
I
could
rattle
off
about
50
different
large-scale
enterprises
that
have
built
out
effectively
what
you
guys
are
defining
that
does
not
look
like
a
functions
as
a
service
or
be
like
service
from
the
complete
operational
benefit
of
it.
So
again,
we're
coming
back
to
what
people
have
called
my.
G
J
Are
you
talking
about
where
people
are
running
their
own
things
on
their
own
iron
or
even
in
other
people's
clouds,
or
whatever
does
not
remove
that
benefit
from
an
organization?
And
it's
a
thing
that
businesses
spend
millions
of
dollars
a
year
across
hundreds
of
peoples
of
time
to
do
so?
It
doesn't
really
change
them.
H
J
H
And
we
doing
that
I'm
not
entirely
aligned
with
Alex
on
his
definition,
I'm
saying
all
the
things
that
you
just
mentioned.
You
know
the
operational
aspects
and
all
that
they
don't
necessarily
to
tie
tie
in
to
how
much
you
pay
per
month.
Is
it
you
know
a
cap
or
is
it
per
invocation
or
you
get
a
freebie
thousand
invocations
and
then
you're
starting
to
pay
all
those
things
are
a
business
model
related.
J
Model
when
you
talk
about
no
Kyle's
for
Idol,
that
is
not
a
pricing
model
that
is
a
tenant
of
all
of
the
various
services
that
we're
discussing
now,
you
guys
can
say
that
I'm
representing
lamda
and
that's
fine,
but
if
you
look
across
the
industry
at
things
that
look
like
serverless,
so
the
other
functions
of
the
service
providers,
things
like
with
what
fauna
DB
does
things
like?
What
zero
does?
No
cost
for
idle
is
very
much
a
foundational
aspect
of
those
platforms
of
the
space
of
how
customers
come
to
consume
those.
J
J
This
is
where
we
100%
disagree,
so
I
mean
Alex.
You
have
a
fast,
it's
a
wonderful
fast
product,
but
it
is
a
fast
operationally.
It
does
very
little
to
reduce
the
burden
on
a
majority
of
people
in
an
organization
they
have
to
care
about
scaling,
servers,
securing
servers
and
managing
servers.
It
does
not
remove
that
from
their
ability.
So
if
we
go
up
to
a
point
just
above
here
where
we
talk
about
there
being
no
management
of
servers,
it
doesn't
change
the
fact
that
there
is
someone
inside
company
X
that
has
to
manage
servers.
J
E
Like
you've,
just
moved
it
to
your
operations
team
in
AWS,
that's
the
only
thing
and
if
somebody
did
the
same
thing
and
they
took
one
of
these
really
popular
open-source
projects
and
they
made
a
hosted
version
of
it,
then
guess
what
that
is
that
now
service?
You
know,
I
think
that
we
really
got
a
struggle
here,
because
a
definition
is
someone
else
is
looking
after
it
can
be
any
of
the
software
in
this
document.
As
long
as
someone
else
is
hosting
it.
That's
now,
service
I
think
that's
a
very
that's
a
real
stretch.
J
Facebook
eBay
Google
internally
in
Amazon,
outside
of
AWS
companies
like
Etsy
in
New,
York
tumblr,
any
of
the
big
organizations
that
be
interfacing
for
a
long
time.
Their
developers
have
no
concept
of
the
servers
that
they
run
on
so
again,
what
is
the
thing
that
you're
defining
that
is
different
than
all
of
that
stuff?
J
E
C
E
J
You
need
to
read
the
document
worker
understand
so
as
your
souland
screen
here
we
have
split
up
developer
and
provider
as
kind
of
two
different
types
of
personas
that
are
involved
here,
and
so
yes,
the
provider
does
not
see
the
benefits
to
this,
whereas
the
end
consumer
does.
So,
if
you
look
at
it
again
like
I,
think
that's
what
everything
you
just
said
is
basically
in
the
doc
the
way
that
it's
defined
right
now
increase.
C
That's
actually
a
way
that
there
are
two
perspectives
and
maybe
the
if
we
consider
cost
or
pricing,
or
we
consider
this
paragraph
from
from
those
two
perspectives.
It
might
be
that
we,
we
collectively
spend
a
lot
in
a
bit
more
easily.
I.
Think
if
you
look
at
something
from
the,
if
you
look
at
costs
from
an.
C
J
J
Think,
oh,
this
benefits
half
my
company,
but
not
the
other
half
of
my
company,
so
I'm
all
for
clarifying
information
here
I
think
it's
very
different
to
remove
the
aspects
around
the
fact
that
you
don't
pay
for
idle
the
service
that
is
de-facto
one
of
the
things
that
you
see
every
analyst
and
many
blog
posts
and
other
people
talking
about
it.
There's
absolutely
a
space
for
faz
that
you
host
and
one
yourself
it
just
shouldn't,
be
known
as
serverless
facet,
you
self-hosted
faz,
so
I
think
obvious.
E
Intention
from
all
the
projects
on
this
paper
and
I,
don't
think
we
have
that
on
this
call,
even
if
you
look
at
the
web
page
from
minami
forculus
or
the
one
for
efficient,
they
both
say
serverless
frame
service
functions
and
for
us
to
take
that
away
from
them.
I
think
we
should
at
least
get
everyone
to
agree
to
consensus
on
that
paper
and
submit
PRS
to
everyone's
reaper.
Saying
change
your
website,
because
you
can't
use
that
term
anymore
now,
in
effect,
your
trademarking
this
or
removing
it
from
these
projects,
but.
J
Have
washed
terms
for
decades
right,
DevOps,
big
data,
a
cloud
these
are
terms
have
been
abused
by
everybody
and
anybody
that
wants
to
be
part
of
and
in
an
industry,
interesting
trend.
I
mean
that's
just
going
to
happen.
It
doesn't
mean
that
we
shouldn't
have
a
paper
that
represents
what
a
lot
of
people
the
industry
is
going
in.
J
B
H
E
Dan
specifically
told
me
that
there's
there
were
three
or
four
projects:
none
I
think
only
one
of
them
is
on
this
call.
He
have
a
vested
interest
in
proposing
to
be
to
be
considered
as
incubator
or
inception
projects
by
the
CN
CF.
Those
are
the
ones
that
I
think
he
should
at
least
ask
to
to
join
or
I.
Don't
know
why
they're,
not
here,
and
actually
we.
E
F
E
J
E
J
Sure,
but
if
it
came
out
and
this
document
when
and
when
it
said,
okay,
this
is
what
services.
Now
everyone
sees
this
document
and
that's
what
services?
How
big
we
deal?
Is
it
just
say
you
know
what
you're
right
we're
just
fast:
we're
not
service.
How
damaging
is
that
relate
to
the
products
that
people
have
built
that
have
their
own
benefits
and
use
cases
and
so
forth.
I.
A
A
Know
it
I
think
this
technology
is
so
only
enough
that
I,
don't
think
that's
wave
up,
and
he
knows
in
confidence
so,
but
we
do
want
to
try
to
use
terminology
that
is,
is
accepted
across
a
group
of
individuals.
I
know,
I,
don't
know
if
I
don't
think
we're
gonna
solve
solve
this
specifically
today
in
this
call,
but
I
think
if
we
just
maybe
write
out
both
both
perspectives
as
two
different
competing
definitions.
Today,
I,
don't
I
think
that
might
solve
some
of
the
concerns.
I
missed.
A
C
A
Right,
so
yes,
if,
if
whatever
isn't
being
captured
in
this
pie,
if
you
guys
want
to
suggest
that
over
the
next
week,
we
can,
you
know
sort
of
try
to
close
laughs
next
week.
If
that's,
except
for
the
one,
if
that's
not
acceptable,
then
we
can.
After
we
talk
with
the
TOC,
we'll
see
if
the
TOC
wants
to
even
continue
with
the
white
paper
in
its
current
state
or
not
so.
H
Think
that
term
usually
is
because
I
am
a
choose
to
make
a
fixed
cost
per
month
or
I
may
may
not
associate
the
invocation
directly
with
with
the
cost.
Give
you
you
know.
100Th
invocations
per
month
for
free
is
usually
we
do.
I
do
agree
on
the
turn
on
the
fact
that
service
is
more
of
a
hosted
version
versus
fast
and
and
all
the
things
about
zero
operation,
but
the
term
that
I
must
be
charging
a
certain
pricing
model
seems
to
me
pretty
limiting
yeah.
A
G
Yeah
I
think
the
concept
of
no
cost
when
I
do
that?
That's
that's
what
you
put
there,
but
in
terms
of
you
know,
say
absolutely
I
also
have
some
reservation,
because
you
know
when
your
warm
start.
There
are
some
resources
there
right
and
also
in
the
I
thinking,
Amazon
right
the
step
functions
and
there's
some.
You
know
resources
there,
which
you,
which
are
you,
but
you
do
not
charge
right.
G
F
J
J
H
J
I
wouldn't
say
necessarily
SLA
potentially
because
then
maybe
that's
like
bucketing
too
much,
but
like
some
sort
of
additive
like
platform,
capability
or
feature
set
or
something
like
that
which
in
this
case
you're
considering
an
SLA
one
of
those
things.
But
it
doesn't
really
have
to
be
just
that
as
a
potential
space.
It's
good
for
me.
H
J
B
C
It's
not
unimaginable
that
in
the
future,
service
providers
wouldn't
necessarily
change
their
their
pricing
model
around
when
you're,
when
functions
are
not
running
whether
or
not
you've
got
them
warm
or
they're.
Very
hot
or,
like
some
of
this
is
subject
to
change
as
well.
If
the
entire
world
was
mostly
clean
functions,
you
can
bet
the
price
there's
any
updates
around
how
things
are
priced
and
that
may
have
I.
J
A
E
A
Then
I'm
and
think
what
we'd
like
to
do
a
lot
cuz
have
them
provide
some
like
a
paragraph
that
kind
of
describes
where
they
throw
like
they
differentiate
from
what
we
have
in
here
and
we'll
just
add
that
and
and
try
to
say
that
you
know
that's
what
just
one
view,
and
we
can
talk
about
that
next
week.
If
that
makes
sense
to
have
that
view
in
here
or
not
know,
if
it
modifies
your
change
in
you
the
thoughts
we
have
so
far,
I.
J
Mean
I'd
like
to
see
text
me
it.
That
explains
how
points
wanting
to
don't
define
server
lists
right,
like
I,
want
to
hear
what
the
story
is.
That
makes
it
different.
Then
again,
what
paths
has
been
for
many
large
organizations
for
a
very
long
time,
and
what
kind
of
the
key
differentiator
is
to
define.
A
A
Cool
and
then
I
think
it.
The
only
other
topic
I
had
on
the
agenda
was
to
talk
about.
There
was
some
discussion
I
get
to
the
last
minute.
I
couldn't
make
around
defining
a
specification
and
I
think
we
should
hold
off
on
any
kind
of
specifications
until
I
have
to.
After
we
talked
to
the
TOC
there'll,
be
one
of
the
items
I
want
to
bring
up
with
them.
You
know
they
want
to
see
us
create
some
kind
of
a
specification,
so
I
think
it
might
be
too
hard
of
a
word
but
more
like
hey.
H
I
think
we
can
start
breaking
small.
You
know
specification
on.
You
know,
let's
not
call
them
specification,
but
you
know
the
events
is
one
and
maybe
I've
seen
you
know.
Some
of
the
challenges
talked
about
in
service
courses.
All
the
you
know
serve
debugging
and
instrumentation
and
services.
So
maybe
we
can
come
up
with
some
approaches
to
do
that.
You
know
so
yeah.
We
can
go
point
by
point
and
creating
those
mini
specifications
or
even
reference
implementations
of
setting.
A
G
G
Actually,
there's
like
there's
a
issue,
you
know
in
the
open
source
yeah.
What's
that
track
in
the
open
source
thing
it
does
not
work.
Can
would
you
like
to
scroll
down
to
the
accession
yeah
I
just
want
to
sure
that's,
not
my
own
problem
when
I
click
that
it
does
not
work.
You
know
the
comparison
of
the
open
source
projects.
G
A
H
G
A
H
H
L
H
H
There
are
also
some
services
where
there
is
no
message,
because
otherwise,
our
platform
services
that
just
you
know
like
s3,
generates
an
event
you're,
not
taking
triggering
the
event
through
a
message.
It's
not
it's
a
platform
service
that
generates
an
event
and
Sri
has
his
own
closed
schema
and
you
cannot
influence
externally.
So
we
need
to
distinguish
between
those
two
elements.
A
H
Maybe
you
can
spend
just
a
couple
of
minutes
on
the
points
of
the
events
that
I
wanted
to
clarify.
You
know
the
Darda
to
proposal
is
essentially
our
Austin
and
the
one
that
I
came
with
I.
Think
on
the
metadata
types
of
metadata
were
mostly
aligned.
You
know
it
just
beats
I
think
the
key
there
friends,
which
is
also
what
Mark
was
digging
into.
H
Is
that
the
way
we
see
that
the
message,
but
it
needs
to
be
the
transport
body
and
then,
if
you
want
to
add
more
stuff,
you
added
through
the
headers,
and
that
means
that
you
can
have
you
know
I
could
have
protobuf
or
something
like
that
running
efficiently
and
the
way
that
the
open
events
thing
work
is
that
you,
you
have
two
layers
of
serialization,
because
the
message
is
being
serialized
as
another
JSON
object
within
the
original.
You
know
within
the
envelope
JSON
object,
so
that
may
be
a
very
nice
thing
for
low.
H
You
know:
low
fidelity,
low
frequency
type
of
events,
but
probably,
if
you're
doing
stream
processing,
that's
not
something
you
would
consider.
So
one
way
is
to
say
we'll
choose
one
or
the
other.
The
other
ways
say
you
know
what
maybe
you
want
to
do
that
in
the
serialization
and
all
that
you
go
with
this
packaging
format?
For
the
event,
you
want
something
more
efficient.
You
go
with
the
other
approach,
I.
G
H
H
H
Yeah
this
is
where
I
listed
the
two
forms
of
you
plants
or
platform
events
for
lack
of
a
better
name
and
sort
of
external
events
which
have
to
be
sort
of
transferred,
somehow
or
in
protocols,
and
then
in
serve
list.
The
two
things
the
that
message
and
protocol
the
protocol
is
not
something
new.
It's
existing
protocols
that
you
basically
lay
those
messages
on,
and
the
second
one
is
an
example
of
an
API
and
I
gave
there
a
JSON,
yes
or
a
go
language.
You
know
interface
example,
this
essentially
parses.
H
H
The
nice
thing
about
about
it
is
that
actually,
if
you
just
do
an
HTTP
boss
post,
it
complies
with
it,
no
need
to
do
anything,
it's
only
you
know.
So
from
the
function
perspective,
maybe
you
know
existing
functions,
don't
necessarily
comply
to
the
API.
You
can
create
a
shim
layer
that
translates
the
message
into
a
common
API.
Very
lightweight
and
project
can
change.
We
can
decide
over
time
to
adopt
this.
You
know
this
thing
of
describing
some
schema
to
the
event.
H
But
if
you'll
go
for
it,
you'll
see
that
there
is
not
too
many
definitions.
It's
essentially
just
the
few
headers
give
you
things
like
that.
They
are
exist,
exist
in
every
user,
villus
like
an
ID,
timestamp
and
few
other
fields.
You
could
see
out
there
and
most
of
them
are
optional.
Besides,
don't
there's
only
thing
to
which
are
not
optional.
One
is
the
ID,
because
every
event
has
to
have
some
sort
of
an
IT.
H
E
H
H
So
I
I
listed
those
differences
in
this
last
paragraph,
okay,
I
think
Austin
is
taking
it
from
a
product
that
they
have,
which
is
essentially
an
HTTP
message.
Router,
okay
doesn't
feed
streaming,
it
doesn't
feed
many
other
use
cases.
It
doesn't
fit
cases
where
you
know,
let's
assume
I'm
streaming
streaming,
it
database
records
from
phosphorus
and
they
come
in
an
ad
informant
of
phosphorus
I'm,
probably
not
gonna
marshal
them
the
way
Austin
suggested
in
its
document.
So
it's
very
good,
for
you
know
Web
books
and
those
kind
of
things.
H
E
Applications
so
with
Austin's
example
in
his
message
in
slack
he
said
that
he
had
Amazon
Google,
Microsoft,
Oracle
and
IBM
are
all
in
agreement
with
his
specification.
Now
I,
don't
know
if
that
was
an
exaggeration,
but
if
he's
got
that
much
backing,
maybe
maybe
we
should
also
look
at
what--he's
theory
know
again
that
we're
gonna
be
Steinbeck.
E
H
Right
but
I
think
we
should
also
look
at
technical
merits.
Not
just
you
know,
sir
marketing
campaigns.
Okay,
I
think.
If
we
go
and
present
to
everyone,
that's
involved
and
we
can
say:
look
you're
going
to
pay
a
huge
performance
penalty
every
time
you're
going
to
the
marshal
an
event.
Is
that
okay
for
you?
H
So
maybe
it's
good
for
firebase,
but
I,
don't
know
if
it's
good
for
stream,
processing
and
other
new
applications
that
are
going
into
server
lists
and
and
I
think
the
open
events
starting
point
is
in
those
type
of
applications
that
are
low
frequency.
You
know
low
fidelity,
so
we
can
say:
okay,
we're
gonna
have
two
standards.
One
is
for
a
server
back-end
applications
which
does
think
about
those
kind
of
resource
utilization
issues,
and
we
could
have
one
for
fronting
or
we
can
have
one
that,
because
there
are
two
different
things
and
respects.
H
One
is
the
metadata
fields
like
a
timestamp
etcetera.
The
both
stacks
have
roughly
the
same.
We
can
agree
that
we
call
them
the
same
names
and
they're
exactly
the
same.
Okay,
the
big
difference
that
needs
to
be
breached
here
is
the
world.
This
huge
overall
overhead
of
double
serialization
and
hard
coding
fields
versus
incrementally
adding
in
I'll,
give
you
an
example.
When
you
have
an
HTTP
gateway,
it
basically
adds
more
metadata
into
the
event.
H
So
if
you
store
everything
in
JSON
and
you
need
to
basically
the
the
capsule
a
JSON
insert
more
feels
like
the
region
of
the
you
know,
you're
not
necessarily
getting
the
idea
of
the
event
in
the
event
itself,
it
may
be
embedded
through
the
gateway
or
other
mechanism,
so
there
are
limitation
in
those
in
that
proposal.
So
you
can
say:
okay,
there
are
a
bunch
of
guys
agreeing
to
it.
So
let's
overlook
the
implementation.
E
E
E
J
The
only
involvement
that
I've
known
about
it
is
passing
conversations
with
Austin
from
service
framework
at
surface
conf
and
then
some
of
the
conversations
here.
As
far
as
I
know,
we
haven't
committed
to
any
spec
reviewed
any
spec
put
feedback
in
on
any
spec.
We
generally
like
the
idea
of
there
being
kind
of
like
an
industry
spec
that
can
act
as
a
glue
that
allows
many
different
providers
of
like
event
sourcing
to
plug
into
various
fast
solutions,
but
I
that's
about.
As
far
as
the
conversation
goes.
As
far
as
I
know,
yeah.
A
G
H
J
If
I
could
run
this
real
real,
quick,
the
thing
that
I
think
will
happen
is
that
we
can
come
up
with
a
idea
of
a
common
spec,
and
then
everyone
can
say.
Oh
we
support
the
common
spec,
but
then
people
could
still
support
what
else
they
support
slick
example
like
the
20-plus
other
teams,
at
eight
arrests
and
great
with
lambda.
We
would
never
be
like.
Oh
now,
you
must
adhere
to
this
spec,
but
I
think
in
general
we
could
be
like.
Oh
you
know,
a
given
product
is
open,
spec
compliant,
fantastic
I.
J
H
Increase
in
what
I
wrote,
every
lambda
implementation
can
fit
into
that.
The
way
that
it's
defined
and
maybe
with
some
a
small
API,
wrapper
that
just
the
message
in
figure
out,
if
it's
Canisius
or
SP,
or
something
in
generate
server
class
to
finish
and
out
of
it
because
right
now
you
don't
have
if
I'm
getting
SV
or
Canisius
or
HTTP.
You
know
it's
a
I,
greatly
message:
I
can't
decipher
that
from
the
function
itself.
There's
no,
like
you
know,
scheme
or
you
know,
like
event
type.
So
that's
the
only
thing
that's
required
is
to
decipher.
J
That's
very
I
think
like
it's
Kathy
was
saying
something
were
just
like
really
early
in
this
conversation.
So
I
don't
know
it
doesn't
seem
to
me
like.
It
makes
a
lot
of
sense
spending
more
time
on
it
today
and
right
of
time,
but
yeah
I
think
I
think
there's
a
lot
of
room
for
a
lot
of
things
to
happen,
but
I
I've
literally
had
about
four
sentences
worth
of
conversation
about
it.