►
From YouTube: CNCF Serverless Working Group September 14, 2017
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
C
A
C
C
D
C
What
are
you
looking
at
right
now?
Are
there
slides,
yep,
you're
scrolling
through
and
quickly
yep,
perfect
yep?
Thank.
A
A
A
A
A
A
E
E
A
So
basically
been
his
focus
on
trying
to
finish
up
the
the
white
paper
itself.
I
think
the
main
focus
of
the
changes
are
really
been
on
sort
of
abstracting
it
more
making
a
little
less
container
centric
for
the
most
part
and
trying
to
get
a
crisper
definition
of,
say,
serverless
versus
faz
I
think
that's
probably
been
the
biggest
thing.
A
E
A
Right
what
I'm
going
to
get
started?
905
I
mean
actually
paste
the
agenda
link
into
the
chat,
look
and
see
that
if
they
don't
have
it
up
first,
we
have
Austin's
gonna
present.
A
quick
demo
of
the
serverless
framework
believe
is
what
it's
called
so
Austin.
You
want
to
go
ahead
and
show
your
screen
again
and
then,
let's
do
that
sure
thing.
C
Yep
all
set
okay,
I've
got
about
15
minutes
here
and
I'm
just
gonna.
Try
it
raced
through
this
pretty
quickly
kind
of
kind
of
a
lot
of
stuff
to
cover,
but
good
morning,
good
afternoon,
good
evening,
everybody
I'm
Austin
Collins,
the
founder
and
CEO
of
Cerberus
Inc.
We
make
tools
and
infrastructure
to
help
developers
and
teams
build
service
applications.
C
You
could
go
all
types
of
traditional
applications
on
it,
as
well
as
new
neat
event
ribbon
automation,
and
we
were
trying
to
be
around
and
spread
that
message
to
little
success.
Then
we
were
met
with
the
same
amount
of
excitement
as
as
what
we
had.
So
we
went
out
and
built
something
called
the
service
framework,
which
was
it's
an
application
framework
to
help
people
build
applications
entirely
on
lambda
and
now,
as
your
functions,
Google
cloud
functions
and
a
lot
of
other
servers.
Compute
providers
and
this
project
was
instantly
very
well
received.
C
I
think
it
was
the
missing
accessibility
story
for
lambda
and
how
to
use
service
compute
in
general
in
the
context
of
applications
and
I,
think
it
provided
that
developer
experience.
That
was
missing
for
that
for
that
great
technology.
So
yeah
we've
been
working
on
that
since
about
2015.
In
addition,
we're
working
on
something
called
the
service
application
platform
now
and
this
application
platform
has
three
components.
There's
the
framework,
our
original
project
there's
a
new
component-
and
this
is
our
first
piece
of
infrastructure
that
we've
that
we're
offering
it's.
C
They
called
the
event
gateway
and
is
a
very
sophisticated
event.
Router
basically
gets
data
in
the
form
of
events
to
functions
as
a
service
located
everywhere
and
lastly,
there's
a
dashboard
product
that
we're
coming
out
with
soon
we're
gonna
offer
this
platform
hosted,
is
and
you're
also
going
to
be
able
to
install
it
wherever
you'd
like
so
I'll
quickly,
go
over
the
framework
for
those
of
you
who
aren't
familiar
with
it,
it's
simply
a
command-line
interface
written
in
nodejs
and
it
all
about
developer
experience.
C
This
is
was
our
big
goal
with
the
service
framework,
and
our
mantra
in
the
early
days
was
hey.
If
you
have
to
go
look
at
the
Amazon
dashboard,
then
we
fail.
We
want
everyone
to
be
able
to
just
do
everything
and
the
command
line.
We've
been
pretty
successful
with
that
so
far,
so
you
could
you
just
simply
install
it
with
NPM.
You
could
create
a
template
and
a
variety
of
different
languages
and
deploy
something
with
relative
ease
and
have
some
type
of
outcome
or
end
result
within
a
few
minutes.
C
Our
I
think
the
the
big
reason
why
the
service
framework
became
so
successful
was
because,
when
you
look
at
what
a
service
architecture
is
on
Amazon,
it
is
like
all
these
all
these
functions.
All
these
granular
pieces
connected
to
lots
of
other
infrastructure
pieces
and
it's
a
bit
chaotic,
especially
if
you're
a
newcomer,
trying
to
figure
out
what
this
architectural
pattern
is
all
about.
It's
very
spaghetti,
ish
in
nature
and
because
all
the
pieces
are
so
small
and
you
end
up
with
a
whole
bunch
of
them,
it
just
gets
more
complicated.
C
C
Think
of
it
at
the
in
simplest
terms,
right
here
on
the
left.
There's
just
you
know,
event
sources
and
you
create
subscriptions
between
those
events,
sources
and
functions
functions
as
a
service
or
service
compute
functions,
and
we
express
this
in
our
service
framework
in
the
least
amount
of
configuration
and
this
the
most
helpful
in
in
helping
people
get
into
server
lists
and
start
using
lambda
and
cloud
functions
and
Azure
functions,
and
this
story
is
really
just
we
say
it's
a
story
of
functions
and
events
in
your
configuration
file.
C
You
simply
list
out
the
functions
that
you
have
or
that
you
want
to
deploy
and
then
each
function
as
a
events,
property
where
you
can
it's
just
an
array
of
events
that
will
go
and
trigger
that
function,
so
very
minimal
config.
You
could
build
a
REST
API
with
the
service
framework
and
probably
eight
nine
lines
of
configuration,
and
you
know
it's
it's
a
it's
a
simple
abstract
story
to
cover
up
the
messiness.
That
is
the
the
service
architecture
underneath
the
other
thing
we
do,
of
course,
is
application.
C
Lifecycle
management
for
your
service
application,
everything
from
creating
it
to
running
it
locally.
We
offer
a
great
local
experience,
which
is
traditional
in
service
architectures
in
general,
deployment
is
very
easy
and
safe.
We,
you
can
invoke
it
pretty
easily
through
our
command
line,
see
logs
rollback
functions,
rollback
your
entire
application
and
remove
it
there's
a
ton
of
stuff
we
offer
just
in
the
CLI
and
I
think
we
do
this
pretty
well,
we
try
and
bake
in
all
the
best
practices
that
we've
learned
over
the
years.
C
You
know
right
into
this
automation
the
use
cases
that
we
see
with
the
service
framework
number
one
backends
ap
is
number
two
data
processing,
number,
three
kind
of
random
event:
urban
automation
and
overall
there's
just
a
ton
of
stuff
people
have
built
with
this
framework.
It
is
incredible.
Every
single
day
we
see
someone
trying
to
fit
some
odd
job
into
a
laminate
function
and
make
it
work,
and
it's
and
it's
pretty
fascinating,
I
think
you
know
once
you
have
a
computer
service
that
is
this
convenient.
C
It
really
brings
all
users
and
use
cases
to
the
table
good
at
the
end
of
the
day,
these
qualities
as
your
administration
or
auto-scaling
at
paper
execution.
These
qualities
are,
are
addictive
and
we
see
people
trying
to
make
as
much
stuff
as
they
can
with
it
using
our
framework,
so
some
examples
of
that
over
here
on
the
left,
we're
multi
language
and
multi
provider.
C
Now,
so
we
offer
lots
of
different
languages,
Java
Python
node
and
we
have
I
think
eight,
eight
or
nine
different
provider
plugins
now,
so
you
could
really
deploy
service
functions
all
over
the
place
and
our
goal
is
to
continually
to
offer
a
uniform
experience.
For
you
know,
writing
your
functions,
one
way
and
deploying
them
to
multiple
providers.
C
We've
done
a
pretty
good
job
of
that
in
service
framework
version,
one,
although
it's
been
it's
been
a
challenge
at
the
same
time,
because
the
landscape
has
moved
so
fast
and
it's
just
hard
to
build
that
perfect
abstraction
on
top
of
this
rate
of
innovation,
but
we,
but
we
have
a
lot
of
work
plan
for
this
a
lot
of
stuff
coming
and
we're
going
to
do
this
a
lot
better
in
the
very
near
future.
Here
many
plugins,
we
have
fantastic
community
they've.
C
C
Lastly,
we
just
hit
19,000
stars
on
github
in
a
couple
years.
I
think
that's
something
that's
pretty
good
I
think
we're
definitely
one
of
the
fastest
growing
open-source
projects
out
there.
We
have
a
ton
of
enterprise
users,
all
the
work,
we've
done
to
bacon
best
practices
and
focus
on
and
focus
on,
providing
a
safe
service
application
lifecycle
management.
Experience
has
really
appealed
to
the
enterprise,
especially
now,
because
all
we
hear
these
days
is
service.
First,
a
lot
of
these
companies
that
come
to
us,
they
say
look
time
to
market,
reducing
operational
overhead
expenditures.
C
But
a
lot
of
these
users
will
try
and
work
around
these
limitations
as
much
as
possible
just
to
just
to
take
advantage
of
those
underlying
those
underlying
qualities,
and
only
when
they
can't
get
around
these
limitations
then
they
start.
Then
we
see
them
starting
to
fall
back
on
other
technologies,
but
yeah.
The
framework
has
been
great
for
enterprise
in
general
because
it
really
helps
them
operationalize,
serverless
development.
It's
a
it's
a
it's
a
great
tool
for
spreading
across
your
developer
teams
and
giving
them
some
safe
guidelines
to
so
that
they
do
so.
C
They
use
best
practices
when
they're
building
these
applications
and
largely
I
think
we're
a
focal
point
of
the
servlets
movement.
I
think
we
definitely
had
something
to
do
with
a
lot
of
this
activity
and
hopefully
we're
I
think
we're
a
big
inspiration
for
it
and
if
you
ever
want
to
see
serverless
trends,
you
know
what's
happening
in
this
in
this
movement
in
this
community.
C
Our
github
repo,
all
of
our
open-source
projects,
are
just
a
great
place
to
to
observe,
observe
all
that
the
event
gateway
is
our
new
project
I'm
going
to
actually
do
a
demo
of
it
if
time
permits,
it
is
a
highly
productive
event.
Router
for
functions
as
a
service
and
productivity
is
just
the
key
theme
that
we
focused
on
with
this
event
gateway.
First
off,
it
is
simply
an
HTTP
server
and
you
could
send
all
type
of
data
to
it,
but
it
always
expresses
data
in
the
form
of
events
right
now.
C
It
just
accepts
requests
via
HTTP
and
whatever,
depending
on
the
shape
of
your
HTTP
request,
it'll
express
it
as
in
events,
no
matter
what
so
you
could
send
any
custom
data
you
want
into
the
event
gateway
and
it'll
it'll
express
it
as
an
event
or
you
could
send
an
actually
structured
event,
schemas
into
the
event
gateway
and
then
it'll
just
take
that
event
and
use
it.
The
second
thing
it
does:
is
it
routes,
events
to
all
functions
as
a
service
right
now
we
support
lambda.
C
We
support
as
your
functions
Google
Cloud
functions,
but
we
have
a
lot
of
other
providers
that
we
plan
to
support
in
the
near
future,
and
it's
has
a
very
flexible
design.
You
know
we've.
We
spent
a
lot
of
time
looking
at
the
service
architecture
and
trying
to
identify
the
key
pain
points.
A
lot
of
them
can
be.
A
lot
of
these
pain
points
can
be
categorized
generally
under
there's
just
a
lot
of
infrastructure
components
in
this
architecture,
and
it
started
to
get
us
thinking.
C
You
know
along
the
lines
of,
can
we
build
a
new
piece
of
infrastructure
that
seeks
to
abstract
a
lot
of
these?
Other
types
of
infrastructure
at
a
higher
level,
and
so
that
you
can
actually
focus
on
less
infrastructure
as
a
result,
because
this
is
one
of
the
cultural
tenants
of
the
service
movement
in
general.
So
this
event
gateway
seeks
to
do
just
that.
So
it's
it's
both
a
pub/sub
as
well
as
an
api
gateway,
and
then
it's
going
to
expand
into
a
few
other
areas
in
the
future
as
well.
C
But
basically
you
can
use
this
to
expose
API,
endpoints
and
all
HTTP
requests
coming
in
will
be
expressed
as
HTTP
events.
You
could
send
in
structured
event,
schemas
in
a
pub
sub
fashioned
directly
to
it,
and-
and
it's
also
going
to
have
a
lot
of
integrations
built-in,
with
third-party
SAS
providers
and
other
tooling
and
infrastructure,
does.
B
C
We
have
a
lot
of.
We
have
a
lot
of
support.
We
have
a
lot
of
other
features
plan
for
that
in
the
future.
You
know
right
now.
We
did
just
want
to
focus
on
HTTP
at
this
time,
because
our
community
is
very
HTTP
focused.
So
in
the
future
you
know
we'd
like
to
offer
people
some
other
ways
of
doing
that,
but
right
now
we're
just
focused
on
this.
Also,
if
you,
if
you
have
other
questions,
I,
think
a
few
members
of
our
team
are
here
in
the
chat
room
feel
free
to
write
them.
B
I
have
a
couple
of
questions.
First,
you
showed
the
yeah
meal
and
I've
also
looked
at
it
in
the
past.
They
serve
associate
a
function
with
the
within
event
and
I
would
sort
of
look
at
the
problem
from
a
different
perspective
is
the
event
needs
to
be
pushed
into
a
function?
Not
necessarily
the
function
consumes
an
event
like
no
most
programming
paradigm.
You
go
to
an
s3
you're
saying
which
function
that
I
want
to
trigger.
B
When
something
happens,
don't
necessarily
go
to
the
function
say
where
do
I
want
to
get
you
know,
because,
because
one
of
the
things
that
it
could
happen
is
that
you
want
to
change
that
the
function
that
serves
that
event,
you
know
we're
the
version,
for
example,
from
production
to
beta
the
other
way
around.
So
what
do
you
think
about
that?
Is
that
the
right
abstraction,
where
the
event
is
coupled
with
the
function
and
not
serve
as
a
standalone
mapping,
yeah.
C
C
But
it's
you
know
it's
something:
we're
continuing
to
experiment
with
and
I'm
just
going
to
race
through
the
the
rest
of
the
presentation,
real
quick,
so
I've
only
got
15
minutes
and
then
I
think
we
should.
We
should
definitely
have
a
longer
chat,
but
in
the
meantime,
if
you
have
any
other
questions,
just
enter
them
right
into
the
the
chat
box,
yeah
so
again,
event
gateway.
A
big
goal
is
just
to
simplify
the
service
architecture
in
general.
C
These
are
the
things
that
we
we
saw
to
to
make
easier
or
to
eventually
replace
with
a
single
piece
of
infrastructure
that
is
API
gateway.
As
your
your
pub
subsystem,
some
type
of
streaming
data
system-
and
this
is
this-
is
what
we're
going
for
in
the
long
run
it.
Lastly,
a
big,
a
big
goal
of
ours
was
just
to
decouple
logic.
In
D
couple
teams,
microservices
serverless
functions
as
a
service.
C
They
have
great
potential
to
to
help
teams
move
with
more
autonomy,
to
allow
them
to
push
things
into
production
without
affecting
the
system
as
a
whole
and
not
be
blocked
as
much.
In
the
event,
the
event
driven
and
a
design
really
helps
this
a
lot.
Actually
that's
between
just
the
pub/sub
model.
So
we're
trying
to
use
that
to
allow
teams
to
to
author
functions
and
own
them
and
be
able
to
push
them
into
production
without
affecting
anything
else.
So
because,
right
now,
one
of
the
big
problems
that
we
saw
were
yeah.
C
You
could
have
service
functions
and
they're
independent
units
of
deployment,
and
that
comes,
and
that
offers
a
lot
of
autonomy,
but
there's
still
a
lot
of
other
infrastructure
dependencies
that
were
shared
that
would
still
result
in
blocking
situations,
particularly
like
an
API
gateway
and
Amazon's
API
gateway.
The
design
of
it
is
is
such
that
someone
has
to
kind
of
be
in
control
of
the
API
gateway,
and
if
you
want
to
push
stuff
into
production,
you
have
to
ping
that
person
and
say
hey.
C
Can
we
deploy
this
on
the
API
right
now
and
that
gets
complicated
if
you're
a
big
team,
all
working
on
the
same
on
this
same
back-end,
but
in
but
with
the
the
way?
The
immediate
way
is
designed.
You
don't
run
into
that
situation.
There's
no
big
shared
resource,
there's
just
simply
pub/sub
subscriptions.
C
What's
next
for
us
we're
releasing
our
service
platform,
it
includes
a
hosted
version
of
the
event
gateway
along
with
a
lot
of
other
interesting
team
collaboration,
features
and
we're
very,
very
interested
in
service
application,
function
and
event
standards.
This
is
something
we've
been
working
on
for
a
long
time
when
we
look
at
what
our
community
wants.
They
want
access
to
to
everything
they
want
to
be
able
to
use
any
platform,
any
provider
and
any
other
managed
service
on
those
providers
to
best
solve
their
problems.
C
So
a
common
request
for
us
is
hey
we're
using
Amazon
lambda
and
we're
using
Amazon
in
general,
because
everybody
is
using
Amazon,
but
we
really
want
to
use
Google's
machine
learning
stuff.
It's
there
a
way.
We
could
stick
some
Google
cloud
functions
over
there
and
have
all
these
functions
kind
of
work
together
in
some
type
of
consolidated
service
architecture,
and
that's
is
very
much
of
interest
to
us
because
we
don't
view
the
service
architecture
as
a
provider
specific
architecture.
C
C
We
look
at
the
time
here.
Maybe
I
can
do
this
in
about
five
minutes.
Let's
see,
we
have
a
event
gateway
example:
project
at
the
event,
gateway
can
be
run
locally.
It
works
really
nicely
locally.
Thanks
to
the
service
framework,
there's
a
great
built-in
integration.
There
I
have
a
service
web
application
that
I'm
looking
at
now
it
has
a
front-end.
C
It
also
has
a
back-end
that
is
entirely
service
and
service,
and
the
backend
is
comprised
of
a
few
different
server
list
services
and
each
one
of
these
is
simply
like
a
project
that
has
different
functions
in
it.
So
I
could
open
up
the
the
users
service
here,
and
this
is
just
a
service
framework.
Can
figuration
file
you'll,
see
that
there's
one
function
in
here?
That's
listing
on
an
HTTP
event.
If
I
go
over
here,
I
have
the
front
end.
C
In
the
event,
gateway
example,
project
I'll
quickly
CD
into
these
services,
and
you
could
see
all
these
different
groupings
of
functions
in
this
project
here,
I'm
going
to
CD
into
the
the
templates
risks
which
actually
doesn't
have
any
functions
but
I'm
going
to
to
do
this
I'm
going
to
start
with
this,
because
we
just
want
to
observe
real
quick
events
actually
going
into
the
gateway.
So
serverless
framework
has
a
great
new
command
called
serverless
run
it's
going
to
emulate
both
the
event
gateway,
as
well
as
all
your
service
functions
across
cloud
providers
locally.
C
I
just
initialized
it
right
here
and
it's
all
connected
to
this
front-end
application.
So
as
soon
as
I
start
hitting
as
soon
as
I
start
making
HTTP
requests,
you
can
see
them
start
to
show
up
here
in
my
terminal,
so
the
so
every
HTTP
requests
by
default,
if
it
does
not
include
a
special
header,
is
expressed
as
an
HTTP
event.
And
that's
what
we
see
here
yeah,
you
could
see.
Headers
query
parameters,
the
body
all
that
stuff.
C
Now
one
very
cool
thing
about
all
this
is
that
we
have
all
these
different
projects
with
different
groupings
of
functions
and
the
way
that
the
service
framework
is
designed
is
such
that
you
can
go
in
and
actually
add
all
these
functions
to
your
local
session,
so
I,
just
logged
in
I
just
seen
it
into
that
users
project
which
had
one
function
and
I
ran
serverless
run
and
that
actually
added
it
to
my
existing
running
session
here.
So
you
can
see
function.
C
Users
register
was
loaded,
the
event
gateway
also
registered
that
function
and
created
a
subscription
for
it
to
the
HTTP
event.
Let
me
clear
this
real,
quick
I'm
going
to
enter
my
email
address
here
and
hit
register
now
something
a
little
different
happen.
The
HTTP
event
came
in,
we
can
see
the
body
of
it.
You'd
see
my
email
address
right
here
and
then
it
triggered
then
service.
The
service
framework
trigger
was
triggered
to
this
function
via
HTTP,
request
and
started,
running
the
function
and
that
function
actually
then
emitted
another
event
back
into
the
event
gateway.
C
So,
as
I
was
talking
about
before
in
the
event
gateway,
you
could
receive
web
HTTP
requests.
So
it
functions
like
your
API
gateway,
but
you
could
also
send
structured
data
into
it
as
well
to
use
it
as
a
pub
step
and
in
the
actual
code
of
this
users
register
function.
We
register
a
user
right
here
and
then
we
emit
a
user
registered
event.
C
So
all
in
a
single
flow,
using
like
this,
this
one
piece
of
infrastructure
we
received
a
HTTP
request
process
to
function,
emulated
the
end
of
this
lambda
function
locally
and
within
that
function
we
also
sent
another
event
back
into
back
into
the
event
gateway,
and
you
can
see
all
that
listed
out
right
here
now
D
and
what
this
is
is
basically
the
beginning
of
an
adventure
bid.
Workflow
and
I
could
log
in
to
a
few
other
services
to
start
building
on
to
this
event,
driven
workflow.
C
So,
for
example,
I
have
an
e
mails
service
here,
which
has
one
function
that
respects
to
the
user
registered
event
and
sends
the
person
a
welcome
email.
So
I
just
added
that
to
my
session,
you
can
see
the
said.
Welcome
email
function
was
added
to
my
session
and
a
subscription
was
created
for
it
on
the
event,
gateway,
I'll
run
through
the
process,
one
more
time
again,
it's
the
DP
event
that
came
in
it
triggered
the
users
register
function
that
then
sent
out
another
event
back
into
the
event
gateway.
C
This
is
this
is
the
beginning
of
a
law
of
a
larger
event,
driven
workflow
and
the
cool
thing
is
what
you
don't
get
with
the
service
architecture,
as
we
know
it
right
now.
Is
this
awesome
local
experience
where
you
could
kind
of
gradually
introduce
pieces
of
logic
into
it
to
are
working
with
those
all
locally?
Of
course
this.
This
has
a
few
other
neat
features.
Once
the
user
session
has
been
created,
we
actually
incorporated
our
SDK
into
the
front-end
of
this
web
application
and
we
built
it
right
into
react.
C
So
if
you
know
react-
and
you
know
redux-
these
are
event-driven
front-end
frameworks
and
they
and
there's
a
lot
of
conceptual
consistency
with
what
we're
trying
to
achieve
with
the
event
gateway.
So
we
built
our
it.
We
put
our
SDK
right
into
right
into
redux
and
we're
actually
able
to
tap
into
all
client
side.
Events
that
are
happening
and
selectively
send
those
to
the
event
gateway.
C
So
when
a
user
session
is
created,
this
app
is
actually
set
up
to
start
receiving
user
activity
events
whenever
the
user
is
going
around
and
doing
something
on
the
front
end
of
the
application,
and
these
events
are
being
sent
to
the
event
gateway.
You
could
set
up
server
list
functions.
Do
all
types
of
things
with
this
to
create,
like
a
personalized
experience
for
your
user
to
just
look,
save
analytics.
C
Anything
like
that
and
the
last
piece
of
functionality
I'll
show
off
here
is
I'm
going
to
bring
in
a
a
CRM
function,
and
this
is
designed
to
react
to
a
user
created
event.
It's
going
to
simply
add
the
user
to
a
CRM
system.
However,
the
great
thing
about
this
function
is
that
it
does
not
work,
it's
totally
broken
and
the
event
gateway
is
is
entirely
event-driven.
That
was
a
theme
to
begin
with,
and
we
want
to
do
it
to
be
consistent
with
that
theme,
so
we
actually
so
in
the
event
gateway.
C
We
express
all
logs
as
events
as
well,
so
anything
that
happens
with
the
system
of
the
event
gateway
or
errors.
Whenever
is
trying
to
reach
out
to
a
function,
if
there's
a
problem
there,
it
expresses
it
in
the
form
of
an
event,
so
you
can
actually
write
functions
to
react
to
system
system
issues,
so
I've
added
in
the
CRM
function,
and
it
does
not
work
and
we'll
see
that
in
a
second
but
I'm
also
going
to
add
in
a
error
handler
function
as
well.
C
And
I'll
run
through
this
whole
flow
right.
Here
it's
going
to
get
it's
going
to
get
a
little
verbose,
but
we
could
see
everything
happening
in
somewhat
of
a
linear
order
here,
so
they
use
a
so
I
clicked
register
the
HTTP
event
at
the
event
gateway
locally,
that
treatment
users
register
function
within
that
users
register
function.
Another
event
was
emitted,
that
was
the
users
registered
event.
This
triggered
this
and
welcome
email,
and
it
also
triggered
the
add
user
to
CRM
function.
C
However,
that
function
failed
due
to
an
error
and
thanks
to
our
awesome,
local
emulation
experience,
you
could
see
the
errors
pop
up
right
here
and
then
the
event
gateway
through
a
gateway
info
function,
error
event
and
we
actually
set
another
function
to
listen
to
this,
and
this
function
was
then
triggered,
and
it
simply
sends
an
email
alert
to
the
admin
of
the
system.
Saying
that
something
is
wrong
with
this
function.
You
know
you
should
check
it
out,
but
you
can
imagine
all
the
use
cases
that
come
from
this.
You
know
we
love
cloud
native.
C
We
love
serverless,
nothing
in
the
day
things
do
on
these
underlying
cloud
providers,
and
this
gives
you
a
dead,
simple
way
of
reacting
to
problems
like
what
happens
if
s3
goes
down
or
AWS
lambda
goes
down
what
what
you
know?
What
do
you
do
in
that
case?
Maybe
you
react
to
it
with
a
function
that
recreates
the
subscription
to
point
to
a
different
function
in
a
different
region
or
why
not
point
to
a
function
on
a
different
provider?
The
event
gateway
supports
cloud
functions
as
your
functions
and
a
lot
more.
C
So
you
have
that
flexibility,
but
but
yeah.
That's
it
that's
my
overview
of
everything
and
what
we're
looking
at
today,
I
think
is
version
0.1
of
the
servlets
architecture
in
the
application
experience
that's
being
built
around
it
and
in
the
very
near
future
we
have
a
ton
of
stuff
coming
out
to
improve
this
workflow
improve
the
developer.
Experience
improve
the
amount
of
use
cases
that
we
are
going
to
be
able
to
do
with
all
this
great
new
technology.
So
please
look
out
for
it.
If
you
have
any
questions,
you
can
reach
out
to
me.
C
A
Right
cool,
thank
Austin,
look,
they're,
really
cool,
normally
I
would
ask
if
there
any
questions,
but
unfortunately,
do
you
want
to
make
sure
you
don't
run
out
of
time
today.
So,
as
he
said,
you
reach
out
to
Austin
through
email
or
the
team.
Also,
as
you
mentioned,
there's
a
chat,
so
you
can
use
that
in
the
you
know,
during
the
call
right
now,
but
I
think
you're
on
you
want
to
talk
about
the
events.
Mapping
proposal
that
you
uploaded
fairly
recently
I'll
paste
a
link
to
it
in
the
chat.
B
B
So
now
trying
to
describe
the
problem
statement
in
one
diagram,
which
is
the
one
here:
okay,
the
essentially
we
have
a
cloud
platform
like
Amazon
or
erasure
or
others,
and
they
have
their
own
sort
of
API
gateways
or
brokers,
and
they
have
their
service
functions
and
there
may
be
something
I'll
refer
to
it
as
a
service
starter.
What
happens
when
the
function
is
not
really
active?
Someone
needs
to
kick
it
out,
and
then
we
have
consumer
services
that
could
be.
You
know,
web
book
that
could
be
a
on
platform
service
for
the
cloud
provider.
B
B
So
you
know
looking
at
service
comm
documentation
and
all
that
I
think
they're
doing
a
great
job,
but
you
could
still
see
that
it's
very
cloud
specific
many
of
the
definitions
so
and
also
even
when
you
go
into
Amazon
London,
you
look
at
the
structure
of
each
event,
whether
it's
a
an
event
coming
from
s3
or
dynamo
or
Canisius.
There's
no
real
way
from
the
JSON
structure
to
figure
out
what
was
the
event.
You
have
to
write
the
the
function
in
a
way
where
you
already
upfront
know
what
kind
of
event
it
was.
B
So
what
we
need
to
try
and
create,
maybe
as
a
as
a
standard,
is
a
way
for
first
defining
the
message
that
passes
into
them
the
service
function.
Potentially,
we
can
provide
some
definition
for
how
do
we
describe
this
mapping
in
a
more
generic
way,
whether
it's
who
I
am
will
file?
You
know,
maybe
to
sound
mechanisms
that
were
just
presented.
Maybe
you
know
just
think
about.
Kubernetes
ingress
is
yet
another
form
of
mapping,
because
you
have
say:
okay,
here's
an
HTTP,
endpoint
and
it
maps
to
those
specific
services.
B
So
you
have
those
and
also
Amazon,
has
those
event,
mapping
definition.
So
we
may
be
able
to
generalize
that
so,
instead
of
everyone
going
and
describing
that
in
his
own
proprietary
way,
we
can
define
that
and
also
potentially
the
way
to
produce
messages
into
a
function.
We
could
define
how
we,
how
it's
done
and
potentially
even
put
a
restful
interface,
just
like
what
you
just
seen
in
front
of
those
you
know,
even
in
Kafka
or
in
front
of
Canisius
or
any.
B
So
if
you
needa
store
the
uniform
access
to
any
service
you,
you
may
take
something
a
bit
slower,
but
you
still
have
this
uniformity.
So
this
is
the
general
idea.
The
the
other
thing
is
within
the
cloud
platform,
especially
in
the
open-source
community,
where
we
build
those
things,
we
can
also
try
and
standardize
those
components
of
how
they
died.
B
Gateway
talks
to
the
service
function
or
how
we
serve
ignite
those
functions
from
certain
states,
Europe,
ok
and
then
I
went
into
each
one
of
those
components
within
the
document
and,
for
example,
and
gave
more
details
on
those
just
you
know
there
are
different
messaging.
What
we've
just
seen
right
now
is
just
this
model,
but
I,
don't
think
that
we
want
to
serve.
B
So
what
does
HTTP
have
have
like
a
destination?
Your
I?
You
know
we
may
need
it
some
type
and
version
and
maybe
who's
the
region,
the
region,
maybe
another
service,
or
maybe
a
host
address
or
an
IP
address,
and
then
we
have
you
know
headers
and
body.
The
body
may
need
to
be
described
not
necessarily,
but
we
can
say
this
is
a
JSON
body,
so
I
know
how
to
interpret
it,
and
maybe
if
it's
a
structure
about
the
you
know
like
a
JSON
file,
maybe
we
want
to
have
an
optional
schema.
B
So
that's
the
general
idea
that
and
this
by
the
way,
all
of
that
can
map
into
HTTP
requests
all
that
maps
to
AMQP.
Some
of
those
advanced
things
could
be
sort
of
thrown
into
header.
Just
like
you
know,
in
HTTP
content
type
is
not
necessarily
a
you
know
a
separate
field,
it's
just
yet
another
specific
header.
So
all
the
things
that
are
beyond
the
body-
and
you
know
maybe
source
and
region,
could
be
done
on
that
way.
Another
important
thing
to
think
about
is
how
do
you
authenticate
I?
B
Think
it's
a
critical
issue,
one
dive
into
it
right
now,
and
so
that's
a
general
idea
is
maybe
if
we
can
define
this
is
at
least
the
standard
of
how
a
message
looks
like
or
how
to
interpret
a
message
that
book.
The
sender
knows
how
to
generate
this
message
in
this
format
and
the
receiver
knows
how
to
receive
it
and
then
maybe
just
like
what
we've
seen
right
now.
I
meant
a
mechanism
where
I
could
just
tell
the
service
go
and
send
this
message
into
this
endpoint
and
some
are
castration
layer.
B
You
know
I
also
have
details
on
like
a
workflow
here
that
people
can
look
at
you
know,
example
use
case
actually
two
use
cases,
one
with
the
Gita
book,
another
one
which
are
scheduling
service
that
you
could
see.
You
know
basically,
a
form
of
mapping
definition
that
translates
into
orchestration
that
in
service
guys
case
could
be
the
actual
gateway,
but
maybe
potentially
going
and
telling
the
serve
the
source
and
the
destination.
You
know
how
to
act
together,
basically
telling
the
source.
This
is
your
URL
endpoint
for
the
service
that
you
requested.
You
know.
B
Maybe
these
are
the
credentials
you
know
doing
something
similar
on
the
listener
side,
and
then
you
serve
established
a
service
okay.
So
there
is
a
lot
of
content
here.
Also
listed
references
I,
don't
you
know,
Microsoft
has
a
product
in
the
space
called
events
out
the
confluence
or
the
Kafka
maintainer
is
have
something
called
connectors,
which
is
also
sort
of
the
same
thing
around
Kafka.
You
could
feel
it
with
one
type
of
data
and
fan
it
out
with
another,
so
we
can
look
at
all
those
and
I
don't
know.
B
A
So
you're
on
one
of
the
questions
I
have
when
I
was
kind
of
glancing
over
this
and
I
apologize,
I
didn't
get
a
chance
to
look
at
it
in
depth
is,
from
a
definition,
point
of
view
or
abstract
point
of
view.
I
think
I
like
what
I
see
there
in
terms
of
just
defining
the
concepts
and
stuff
like
that.
But
in
terms
of
next
steps
for
this,
do
you
actually
see
this?
You
know
heading
down
the
path
of
defining
a
standards.
You
know
it's
a
nerdy
event,
format
and
stuff
like
that.
B
B
C
Hey
you're
on
this
is
Austin
here
yeah
we,
as
I
mentioned
in
slack.
We
reviewed
the
document
it
it
looks.
It
looks
great
actually
I
think
it's
a
great
starting
point
and
we
have
a
very,
very
similar
vision
and,
and
we've
been
focused
on
a
lot
of
these
concepts
for
a
while,
and
it
looks
like
you
just
shared
the
document
with
us,
so
we're
definitely
ready
to
start
hashing
it
out
with
you
and
whoever
else
wants
to
wants
to
participate.
C
You
know,
building
this
out
and
getting
traction
and
getting
adoption
is
going
to
be
a
huge
challenge.
We've
been
working
with
a
lot
of
providers
on
this
problem
for
like
over
a
year.
It's
going
to
be
it's
not
going
to
be
easy,
but
we're
we're
very
interested
in
heading
down
this
path
anyway,
to
see
where
it
leads.
C
B
I
think
the
nice
thing
is
that
people
that
don't
necessarily
you
know
the
market
leaders
and
don't
necessarily
participate.
We
can
always
write
a
proxy.
You
know
something
that
takes
a
standard
rescue,
API
and
maps
it
into
their
service
yeah.
It
is
part
of
the
kit
and
then
eventually
they
say
you
know
what.
Why
don't
we
just
go
in
and
join
the
effort.
B
D
D
B
I
had
this
exchange
with
the
Alexus,
you
can
even
read
the
details,
but
I'm
not
trying
to
standardize
another
pops
up
and
especially
alexis
is
sort
of
the
originator
of
AMQP,
probably
a
sensitive.
It's
not
about
that.
It's
about
how
do
you
orchestrate
a
service
I,
have
a
customer
says:
I
have
something
originating
from
this
from
a
Gita.
You
know
take
it.
A
web
book
is
an
example.
I
want
to
consume
a
Gita
web
hook
and
get
it
to
my
function.
B
You
know
the
HTTP
gave
to
in
between
or
the
message
passing
between,
it's
just
an
implementation
detail
from
my
perspective
and
maybe
I
have
higher
performance
requirements.
I'll
probably
use
a
stream
instead
of
using
HTTP,
but
the
real
challenge
that
you
know
we
want
to
in
our
platform
to
support
dozens
of
event
sources.
We
don't
have
the
R&D
for
that.
Okay,
so
we
can
say,
here's
a
asynchronous
endpoint
mechanism,
here's
an
HTTP,
endpoint
mechanism.
B
You
want
you
to
do
the
other
things,
but
now
someone
wants
to
go
and
do
integration
with
whether
it's
Gita
or
geek
lab
or
you
know,
there's
someone
the
generating
stockticker
information
as
a
service.
I
don't
know,
I
want
to
be
able
to
serve
this
tie-in
between
one
service
and
another
service,
and
this
is
just
going
look
into
Microsoft
even
have
that
nice
videos
and
all
that
they
explain
the
value
the
position
and
the
challenge.
I
think
also.
B
D
Two
other
questions,
because
the
other
one
is,
am
I
understanding
this
right
that
if
there
is
an
event,
message
structure,
that's
defined
of
something
like
the
event
router
might
in
what
would
it
always
be
the
applications
responsibility
to
include
this.
You
mitigate
the
event
message
or
wouldn't,
in
any
event,
router
or
an
API
proxy,
ever
inject
that
on
the
functions
behind
the
events
behalf
so.
B
You
can
have
some
some
sort
of
a
proxy,
but
the
the
point
is
that
the
message
doesn't
necessarily
have
to
be
structured.
So
that's
how
I
serve
you.
If
you
see
how
the
message
is
organized
is
serve
giving
you,
you
don't
have
to
do
anything.
Essentially,
if
you
take
an
HTTP
request,
it
evenly
maps
on
all
those
things
and
that
will
help
us
within
the
server
function.
You
know,
one
thing
that
we
have
in
nucleo
show
it
in
a
couple
of
weeks.
B
Is
that
the
event
class
handler
is
one
and
I
could
push
an
event
through
Kafka
Kinesis
HTTP
many
other
things,
because
they
tone
getting
normalized
into
the
same
few.
You
know
methods
of
it.
Retrieving
like
get
body
get
content
type.
You
know
etcetera.
You
may
have
auxiliary
interfaces
to
that
specific
class,
but
that
means
that
I
can
develop
a
function
once
I
can
test
it
using
HTTP
cause
I
want
to
push
it
into
production
pipeline
with
high-speed
streaming.
It's
exactly
the
same
function.
I!
H
B
In
lambda
every
event,
source
define
is
only
one
structure,
so
one
of
the
ways
to
address
that
is
by
having
this
optional
schema
that
if
I
know
what
kind
of
event
it
is
schema
name,
I
know
how
to
go
and
parse
the
body
depending
on
the
type
of
diva.
Now
it's
optional,
you
don't
necessarily
have
to
have
that
that
thing
another
thing
could
be
interesting
is
headers.
B
There
is
also
a
way
to
requirement
to
standardize
serve
the
identity
mechanism.
You
want
to
pass
the
function,
who's,
the
originating
identity.
You
don't
necessarily
meet
the
function
to
authenticate
it.
You
should
press
her
the
day
guy
gave
it
done.
You
know
SSL
or
TLS
and
authentication
and
all
that,
but
you
do
want
to
know
who's
the
guy
behind
that
message
or
who's.
The
source
endpoint
is
sent
you
that
message,
so
we
can
normalize
that
everything
that's
out
of
spec
could
be
implemented
in
headers
or
as
part
of
the
body
in
the
JSON
structure.
Okay,.
A
So
I'm
gonna
have
to
call
time
here:
cuz
I
can
I
prepare
running
it
a
little
bit
of
time,
but
I'd
like
to
suggest
is
that
people
make
comments
in
the
document
and
probably
maybe
do
a
little
follow
on
next
week
after
people
have
had
a
chance
to
actually
read
the
doc.
If
that's
okay,
but
I
really
want
to
do
now,
is
move
over
to
the
white
paper
itself.
They're
working
on,
in
particular,
Chris
I,
believe
you're
on
the
call
I.
I
A
I
H
A
I
Sorry
one
second,
okay,
the
people
see
my
screen
now.
Yep
looks
good,
that's
good
great,
so
I
shared
a
link
in
the
meeting
group
notes
with
this
here,
I'm
happy
to
share
other
places,
I
guess
dog,
you
just
posted
it
to
everybody.
You
know
this
was
kind
of
the
the
homework
that
was
assigned
to
me
a
couple
weeks
back
about
coming
up
with
some.
I
You
know
just
as
kind
of
an
example.
We
at
eight
of
us
actually
have
four
guidelines
for
what
is
survey
list,
who
went
here
with
three,
and
we
see
these
three
being
shared
by
folks
like
Microsoft
and
Google
and
others,
and
it
pretty
much
ties
in
with
kind
of
what
we
see
even
again.
Folks,
like
analysts
and
stuff
saying
what
they
consider
surplice
to
be,
and
so
you
know
feel
free
to
comment
on
here.
I
G
I
I'm
sorry,
it
was
like
because
I
can't
edit
the
doc
directly
I
was
trying
to
figure
out
the
best
way
to
do
it,
and
so
I
was
like
I'm
just
gonna.
Take
this
one
section
and
just
basically
break
it
out
and
then
let
the
rest
of
the
group
figure
out
how
it
could
be
potentially
murder,
in
my
apologies
I'm,
not
as
deep
into
document
editing
in
Google
Docs
these
days.
That
probably
could
be
yeah.
A
G
I
think
I
don't
know.
If
do
you
need
special
permissions,
that
you
see
it
where
that
top
right
corner
is,
it
says
editing
if
you
get
that
drop-down
I,
don't
know
if
that
gives
you
suggesting,
and
it
gives
you
viewing
so
it's
a
jet
you
might
have
if
you
got
suggesting
on
the
other
document
and
kind
of
plug
this
in
that
might
work
yeah.
A
A
Any
other
comments
questions
before
we
move
on
and
thank
you
Chris
very
much
for
doing
this.
One.
I
Real
quick
thing
that
I
did
do
there
were
a
number
of
bullet
points
that
existed
kind
of
beneath
the
other
definition
in
the
document
that
exist
today,
where
it
was.
Thank
you
so
comparisons
to
pass
and
some
other
things
and
my
thought
was
to
keep
this
section
very
concise.
And
then
you
know
we
can
have
kind
of
the
immediate
followup
with
how
it
compares
to
pass
I
as
containers
as
a
service
etc.
I
A
Right:
okay,
okay,
okay,
let's,
let's
quickly
move
on,
we
have
five
minutes
left,
Dan,
I
kind
of
I'd
you,
the
agenda-
hopefully
that's
okay
I-
was
wondering-
can
give
a
brief
update
on
the
status
of
the
modifications
that
you
did
making
to
the
body
of
the
spec,
in
particular
around
the
this,
the
sort
of
removal
or
yeah
the
removal
of
the
containerization.
That's
things
yeah.
G
I
didn't
make
too
much
change
since,
since
we
discussed
it
last
week,
you
know
that
it
was
pretty
substantial
changes
and,
and
basically
I
could
struck
out
that
really
controversial
section
in
kind
took
a
new
stab
at
it
and
that's
kind
of
what
I
think
Chris
is
involved
the
text
to
look
like
right
now.
G
The
only
thing
I
addressed
a
couple
of
comments
to
that
the
Chris
had
are
there
on
the
side
and
I
think
one
or
two
other
people
did
as
well
and
then
the
final
thing
I
did
was
just
add
a
little
bit
of
text
to
the
abstract
about
some
outcomes.
This
we
talked
a
bit
about
the
you
know:
harmonization,
I,
guess
the
H
word
and
invest
formats
just
so
that
you
have
some
linkage
back
to
it.
Yaron
was
working
on,
but
no
substantial
changes
other
than
that
this
week.
Okay,
any.
A
Questions
or
comments
on
that,
okay,
I
think
what
needs
to
happen.
There
is
a
lot
people
need
to
go
through
when
we
do
the
doc
and
see
if
it's
in
the
right
direction
and
in
particular
Chris
I,
wouldn't
mind
if
you
could
take
a
look
and
see
if
you
think
more
work
needs
to
be
done
to
to
make
it
a
little
less
container.
Centric.
I
Sorry
was
that
for
Chris
from
Amazon,
yes,
yeah
I
started
looking
a
little
bit
at
some
of
Dan's
edits
to
this
and
made
some
comments.
I
definitely
think
it
does
seem
to
be
shaping
in
a
direction
that
makes
it
again
less
kind
of
tangle
to
the
idea
of
it
being
a
build
upon
or
a
comparison
to
container
as
service.
Okay,.
A
A
Okay,
in
last
couple
minutes
here,
I
wanted
to
discuss
that
the
conclusion
section
I
mean
though
I
know
we
don't
have
a
lot
of
time,
but
when
I
look
at
it,
I
think
there
are
three
main
things
that
I
sort
of
extracted
from
that
one
is
the
potential
to
do
so,
I
think
around
the
common
event,
format
and
you're
on
already
get
a
good
start
on
that.
But
then
we
have
the
idea
of
more
education,
and
then
we
have
the
idea
of
reducing
overlap
and
increasing
portability
and
I.
Think
it's
that
last
one.
A
That
is
probably
the
has
the
most
potential
for
future
work
for
the
CNC
F.
But,
what's
a
little
bit
unclear
to
me
is
how
you
reduce
overlap
or
enable
portability
without
getting
into
the
standards
word
and
I'm
trying
to
get
my
head
around
whether
there's
other
things
besides
standards
that
people
have
in
mind
there,
and
if
so,
can
you
add
text
the
document
with
concrete
suggestions
or
examples
of
things
you
think
we
can
do
without
actually
doing
a
standard
or
whether
people
actually
say
yeah.
This
really
means
standards.
A
We're
just
not
going
to
use
that
word,
because
it's
not
quite
clear
to
me
what
the
concrete
next
step
is
when
you
say,
I
just
want
to
reduce
overlap,
that's
good,
but
what
does
that
actually
mean
physically
right?
So
we
could
get
some
people
to
give
concrete
suggestions
on
that.
I
think
that'd
be
really
really
helpful.
Yeah.
F
The
other
thing
I
was
gonna,
bring
up,
Doug
is
I,
don't
know
if
there's
any
sub
sub
projects
sub
components
of
the
server
list
projects
out
there
that
we
want
to
take
forward,
possibly
as
interesting
projects
that
would
be
cloud
native
focused.
So
you
can
then
kind
of
take
these
projects
into
the
new
TOC.
To
kind
of
have
presentations
of
you
know
what
what
about
this
project
would
be
interesting
for
the
CNC
F
to
look
at
bringing
underneath
the
foundation.
F
C
A
Sounds
great
and
in
sort
of
the
last
30
seconds
or
so
I
suspect
we
probably
do
need
another
phone
call
next
week.
We're
getting
is
pattern
of
weekly
phone
calls,
but
I
think
it's
actually
good
is
there
any
objection
says
trying
to
do
another
call
next
week
all
right,
cool,
okay,
I'll
talk
to
Chris
and
a
check
to
get
that
phone
to
get
the
invite
sent
out
and
with
that
I
think
we're
at
the
top
of
the
era.
We
need
to
stop.
So.
Thank
you,
everybody
for
joining
we'll
talk
in
next
week.
Right
thanks.