►
From YouTube: CNCF Serverless WG Meeting June 22 2017
Description
https://github.com/cncf/wg-serverless
The CNCF Serverless WG meets to explore cloud native serverless technology and concepts. This video is the Serverless WG June 22nd meeting.
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.
B
I'm
I'm
on
the
road
right
now
so
I'll
be
in
my
computer
in
about
15
minutes.
B
So
give
people
a
few
minutes
at
joining.
It
was
kind
of
going
over
the
different
sections
that
looked
on
over
the
last
couple
weeks,
yep.
D
F
B
B
So
we
kind
of
broke
it
into
a
few
different
sections
and
over
the
last
few
weeks,
we've
had
several
attendees
kind
of
contributing
to
those
sections,
and
so
what
I
thought
we'd
do
today
is
to
sort
of
stud
off
with
talking
about
the
progress
we've
made
to
date
on
on
those
sections
and
and
Doug.
If
you
don't
mind,
I'm
done
you
just
sort
of
bring
up
the
different
sections
and
talk
about
what's
needed
next
on
them,
then
I
get
from
a
computer
in
a
few
minutes,
I
can
about
more
yeah.
C
Sure,
okay,
so,
let's
see
going
through
the
various
sections,
the
first
one
is
the
terminology
section.
The
at
was
you
can
and
myself.
We
do
have
a
starting
list
in
there
right
now.
For
the
most
part,
the
list
was
taken
from
the
notes
that
we
all
gathered
over
the
last.
You
know
several
weeks
initial
description
of
each
one
was
put
in
there.
C
There
are
a
couple
that
I
wasn't
hired
to
ensure
what
people
meant
by
so
I
kind
of
some
question
marks
there
so
I'm
looking
for
some
feedback
on
those
but
other
than
that,
I
think
the
most
part
it's
in
a
position
where
I
wouldn't
mind
getting
a
few
facts
on
the
rest
of
the
group
on
ways
to
either
expand.
What
was
written
there
or
to
find
out
change
was
written
there.
If
you
think
it's
wrong,
that's
about
it!
F
G
G
So
the
scope
section
of
it
was
defined
scope
of
the
space
as
it
exists
today.
In
other
words,
today's
offerings
so
I
put
together
an
initial
outline.
I
hope
that
this
is
what
everyone
had
intended
for
this
section
to
be.
You
know,
I
wanted
to
give
a
little
bit
of
history,
I'm,
not
exactly
sure
the
the
full
history,
so
people
want
to
help
out
live
with
that
one
and
then
I
tried
to
describe
you
know
what
is
currently
available
in
public
cloud.
G
Server
lists,
open
source
server
lists
the
various
open
source
services
and
then
frameworks
which,
which
is
more
along
the
lines
of
like
chalice
and
server,
less
calm
and
Zappa.
So
it's
really
to
give
some
exposure
to
what
is
what
is
the?
What
is
going
on
within
the
community
around
serverless?
Is
that
what
the
group
intended
for
this
section
to
be?
First.
C
G
You
know
public
cloud
and
open-source
products
I
didn't
want
to
have
to
speak
for
them,
and
so,
if
they
want
to
help
guide
what
they
think
is
relevant
within
you
know,
but
I
don't
want
one
project
to
have.
You
know
three
pages
worth
of
information
and
others
not
have
any,
but
I.
Think
that
there's
a
fine,
some
commonality
of
what
we
think
is
relevant
within
each
project
and
try
to
try
to
expose
how
they
are
differentiated
or
similar
in
in
various
manners.
F
H
Ron
one
of
the
things
that
myself
and
orissa
tried
to
do
in
the
spec
document
is
actually
trying
into
what's
the
how
we
normalize
the
different
solutions.
We
actually
pointed
out
what
happening
in
AWS,
you
know
longer,
what's
going
on
in
Asia
and
other
solutions
to
try
and
instead
of
saying
you
know
which
one
does
things
better
or
the
opinion
we
could
just
try
and
collect
and
aggregate
the
different
functions
from
the
whole
district.
E
H
Because
I
think
another,
there
is
an
issue
of
the
terminology
and
a
lot,
a
lot
of
people
think
about
surveillance
or
the
API
gateway
with
some
HTTP.
And
if
you
look
at
the
the
Commission's
of
guys,
like
Azure
or
Amazon
with
lambda,
it's
a
much
broader
solution
that
address
many
different
use
cases.
So
we
wanted
to
show
that
there
you
know,
there's
our
broad
set
of
use
cases,
maybe
even
think
that
those
guys
haven't
tapped
on
yet.
C
G
E
E
H
E
That's
right
further
lambda
and
green
grass
think
in
terms
of
email
functionality
similar.
Just
why
is
the
green
grass
is
lightweight,
which
is
because
it's
running
out
there
on
the
edge
right
or
IOT
site,
but
in
terms
of
the
step
function,
it's
much
broader
scope
than
lambda,
because
it's
multiple
functions
and
as
far
as
I
know,
there
are
other
you
know
sorts
to
because
it
is
that
cannot
satisfy
some
customers
use
cases.
So
there
are
people
who
are
people.
H
So
Ken
here
thinks
that
function
is
not
a
necessarily
specific
to
lambda.
You
can
actually
invoke
workloads
in
VMs
or
who
external
clients
I
think
it
is
very
close
to
London
and
we
need
to
see
what
we
want
to
include
in
the
scope
of
a
function.
I
totally
agree
with
your
you
know.
General
line
that
were
close
are
certain
and
enhancer
I
said
that
we
want
to
know
where
or
to
grow
the
line.
H
C
C
E
Yeah
I
think
it
probably
is
good
to
include
this
document.
You
know
I
saw
there's
an
open
lambda
and
there
you
know
we
can
just
add
a
section
to
describe
that.
Maybe
at
that
link
inside
this
document,
if
they
are
links,
they
are
multiple
link.
Oh
I,
think
probably
way.
You
know
somewhere
like
that.
Well
I
think
help
like
that,
and
then
people
get
comments.
C
H
H
E
C
Well,
what's
interesting
is
I'm
scoring
this
document
and
I
really
like
the
way
it's
laid
out.
I
think
it's
I
love
all
the
bulleted
lists
of
the
sections
for
each
one.
It
kind
of
reminds
me
of
the
scope
document
that
mark
is
putting
together
because
it
does
kind
of
talk
about
each
of
the
various
projects
or
many.
A
H
You
do
want
our
separate
just
because
the
points
and
catalogs
before
you
know
like
what
we're
building
is
also
designed
to
be
a
binary
data,
run
on
a
Raspberry
Pi
and
not
necessarily
with
a
fully
loaded
situation
or
also
for
death
and
death.
You
may
want
to
run
your
function.
You
know
with
somewhere
mocking
the
framework
and
so
I
think
certain
kubernetes
is
a
way
to
orchestrate
a
bunch
of
function,
but
it's
not
necessarily
the
only
way
else.
I,
don't
think
the
one
I
89,
the
other
organizations.
H
C
G
You
know
something
that
occurs
to
me
is:
perhaps
positioning
needs
to
go
before
the
scope
just
in
terms
of
flow,
but
if
we
can
decide
that
later
on,
in
other
words
being
able
to
describe
how
the
differentiation
is
between
server
lists
and
has
et
cetera,
would
set
the
stage
for
then
going
it
deeper
into
what
serverless
is
all
about.
I
like.
E
I
agree
with
that
I
secretly,
we
can
go
into
deeper
and
I
think
this
ultimately
is
more
awesome.
It's
more
under
you
know,
describing
the
different
alarm
abstraction
level
right
like
containerization
Lisa,
it's
the
lowest
level
and
then
we
have
passed
and
then
now
have
service.
But
yeah
is
I.
Agree
with
you
know
the
comments
we
can
go
into
deeper.
C
E
G
F
C
C
A
C
Is
the
next
step
right?
Do
we
just
stop
there?
Do
we
look
to
start
some
more
technical
project
over
here?
Do
we
look
for
some
level
of
interoperability
or
standardization
to
happen
someplace?
You
know
we
haven't
figured
that
out.
Yet
we
have
a.
We
do
have
a
documented
list
potential,
what
we
call
phase
two
work
items
and
that's
where
we're
going
to
start
gathering
ideas
for
potential
next
steps
for
us.
I
H
You
know
what
one
of
the
key
things
we
want
to
see
a
serve
as
an
outcome
is
a
you
know
when
the
other
writes
a
function.
Who
knows
what
the
function
is,
what
we
define
also
when
someone
operates
a
function.
If
someone
is
building
now
a
new
development
tool
or
feels
to
to
know
what
what
it
is,
so
we
try
to
say
from
the
different
implementation.
You
know
what
does
Amazon,
for
example,
defines
as
a
context
or
event
for
Asia
or
or
others
and
I
welcome.
H
You
know
IBM
and
others
to
start
adding
things
that
you
think
are,
and
then
you
know
when
will
when
will
you
know
something
that
potentially
we
can
do
in
the
future?
We
go
into
a
knows
if
their
position
is
a
dirty
word,
but
maybe
create
a
way
for
people
to
write,
functions,
works
and
then
work
in
different
frameworks.
So
we
have
a
definition
of
what
website
consensus
among
multiple
people.
What
does
it
mean
event
or
what
does
a
context
to
contribute?
But
this
is
a
basic,
not
know.
H
Obviously,
every
vendor
will
have
his
own
interpretation
and
extension.
It
can
make
the
user
life
simpler.
So
we
try
to
categorize
a
function.
Event
sources.
You
know
the
life
cycle
of
a
function.
What
do
you
do
with
a
function?
There
are
small
things
that
you
know
whether
people
didn't
just
for
a
time
to
market
or
I.
Guess
you
saw
them
like
you
know.
When
you
publish
a
function,
you
don't
have
branching
today
in
any
of
those
solutions.
H
E
Is
section
two
to
the
end
of
the
document,
which
is
the
user
facing
application
specific
specification
model
I
think
that
you
know
are
it's
important
it
we
or
it
will
be
good
if
we
can
provide
some
generic
configuration
directives
for
functions,
events
and
the
workflow
and
so
I,
something
there
also
I
think
you
know
I
some
comments
to
the
are
to
the
previous
session,
because
I
I
would
all
that
I
would
not
like
to
just
change
it.
I
want
you
know
to
to
see
what
you
know.
E
A
E
For
the
function,
invocation
ties
I
some
comments,
I'm,
not
sure
whether
this
is
right
so
I
just.
H
Go
like
this
I
just
I,
just
so
worse
on
them
and
learn
to
comment
because,
like
you
know,
really
mentioned,
there's
already
synchronous
and
asynchronous
and
streaming
is
quite
different
than
asynchronous
because
has
an
ordering
that's
also
Amazon's
limitations
considered
told
they're,
like
you
know
the
definition
of
when
a
function
fails
or
retries,
or
all
that
is
quite
different.
Among
this
you
know
something
like
scheduled.
Exam
is
more
like
an
asynchronous
design,
Chinese
visa,
but
there
are
some
sample
yeah.
E
So
I'm
just
thinking
how
I
we
can
rearrange
this.
You
know
some
of
the
points
because
it's
just
out
there
functioning
location
types
right,
discipline
of
asynchronous
the
other
couple
into
from
that
perspective,
invocation
and
fall
into
that.
But
in
terms
like
you
know,
each
function
is
we
do
retry
or
not
through
retry
or
that
right,
but
that's
also
in
that
has
it's
also
close
to
you
know,
synchronous
or
asynchronous,
so
I
think
you
know,
I
was.
E
H
That
is,
there
is
a
major
difference
between
asynchronous
and
streaming
and
right
so
I
think
it
does
at
least
those
three
models
now
within
streaming
say:
okay,
CDC
stream
for
chemistry
or
a
dynamodb
stream.
Although
they
have
some
subtle
differences
or
you
know,
maybe
just
screaming
a
CSV
file,
although
they
have
subtle
differences,
things
fall
under
the
same
category
extreme,
but
they're,
not
a
synchronous
event
like
an
email,
which
is
something
that
doesn't
have
anything
to
do
with
other
events.
But.
E
C
E
Are
they
will
be
streamed?
There
will
be,
you
know,
other
things
right,
but
I
just
feel
lucky.
Today
they
are
a
southerner
to
synchronous,
a
signal,
so
they
have
synchronous
asynchronous
inside
the
stream.
If
the
stream
consistent,
you
know
we
can
take
this
apart
and
then
I'm
not
saying
you
know,
we
should
not
talk
about
this
I'm.
Just
thinking
you
know
found
out
it's
multiplayer
right.
We
might
need
move
this
to
another
separate
section
which
is
in
our
arm
other
elements
of
the
function
or
other
aspects
of
the
function.
H
E
H
E
Yeah
I
think
that
that's
really
important
one
I
think
that
can
be
specified
either
in
the
you
know,
in
the
user
configuration
in
the
user
configuration
so
how
you
are.
A
E
H
Not
asserting
it's
actually
not
related
to
the
event,
although
amethyst
or
Asia,
certain
bundle
triggers
and
data
sources
into
the
same
definition
and
just
look
at
the
link
that
I
put
there
there's
more
details
in
there
Asia
website
and
the
notion
the
general
idea
is
has
in
many
cases
of
function
require
access
to
data.
In
other
words,
it
serves
very
minimal
function
and
instead
of
handwriting
all
the
you
know.
Every
time
you
go
to
function,
you
regenerate
the
connection
to
a
database
or
someone
put
it
in
a
ambiguous
way.
H
This
is
a
way
to
define
it
as
a
cell
phone
service,
and
this
way
it
can
be
reused
across
multiple
invocations
in
a
term
more
well-defined
way
and
other
advantages
like
credentials
and
other
things
that
could
be
expected
outside
and
and
also
a
lot
of
abstractions.
So
it's
not
probably
want
to
be
the
same
function
from
Kinesis,
and
then
you
go
to
a
non-precision.
Where
you
have
Kafka,
then
you
can
change
it
through
the
definition
and
you
don't
have
to
change
the
code
for
that.
H
G
G
Ok,
Doug
I
have
a
question
for
you
with
respect
to
terminology
I
like
the
within
the
spec
I
like
the
function
invocation
types
of
you
know,
synchronous,
requests,
etc.
I
don't
know
whether
those
should
be
reflected
as
well
into
terminology
or
not,
but
you
know
not
with
with
quite
the
depth
that
it
is
here.
Yeah
the.
G
I
have
for
you
is
whether
we
should
add
things
like
a
runtime
into
the
into
the
definitions
and
also
define
what
a
framework
is,
because
I've
seen
framework
used
in
quite
different
contexts
such
as
you
know,
an
open
source
project
calls
itself
a
framework
where,
as
I've
seen,
things
like
serverless
as
being
considered
a
framework
and
they're
kind
of
two
different
things
when
it
one
is
actually
operating,
the
server
lists
I
in
infrastructure.
Well,
the
other
is
really
just
trying
to
trying
to
orchestrate
up
functions.
G
C
The
specifications
I
started
thinking
that
some
of
these
sections
in
here
should
really
be
pulled
out
into
the
terminology
section,
because
they
had
some
great
explanations
there,
but
I
think
a
lot
of
that
will
kind
of
shake
itself
out.
Once
you
get
to
the
point
where
we
start
merging
these
documents
together
and
we
want
to
remove
duplication
of
text.
C
H
G
H
By
the
way,
there
is
a
reason
reference
to
security.
You
know
personally,
look
at
the
like
data,
binding
context
and
and
function
context,
because
you
know
we
fix
all
the
definition
of
lambda
where
there
is
like
a
key
and
secret,
and
things
like
that,
so
we've
embedded,
but
there
is
no
reference
to
what
does
really
mean
like
how
do
you
conduct
the
you
know,
the
authentication?
H
If
you
also
look
at
the
Advanta
context,
one
of
the
things
that
we
put
out
there
is
originating
you
know
like
source
identity
in
order
to
think
about
like
Amazon
Cognito
or
for
someone
just
like
you
have
an
API
service,
and
you
know
the
client
identity
of
the
guide
is
both
API
function.
You
probably
want
to
pass
this
over,
but
there
is
not
a
lot
of
making
in
terms
of
how
does
the
specification
and
user
authorization
happen.
E
C
E
E
B
E
C
I'm
in
minutes
I'm,
taking
notes
and
cometh
I'm,
adding
a
remark
that
you
know
comments
are
welcome
on
all
sections
but
I'm
trying
to
indicate
whether
people
should
do
a
thorough
review
now
or
wait
a
little
while
I'm
just
want
to
put
down.
As
you
said,
Kathy
you
know
give
another
two
weeks
before
people
give
a
very
thorough
review.
Yeah
we've.
F
B
So
I
think
the
the
next
steps
are
too
long
for
everyone
and
then
we'll
start
out
in
the
email
as
well
too
long
you'll
spend
the
next
couple
weeks,
really
updating
and
commenting
and
providing
feedback
on
the
sections
adding
to
them
making
suggestions
if
we
shouldn't
do
parts
of
them
to
other
areas
and
then
in
two
weeks
we'll
meet
again.
Let's
see
if
we
can
sort
of
target
having
anything
right
from
what
orally
do
we
meet
in
two
weeks.
B
B
G
B
B
With
that
and
what
the
review
I'm
in
two
weeks
that
we
stood
off
will
probably
give
that
you
know
two
weeks
as
well
to
do
a
more
thorough
review
when
we
come
back
together.
That
planning
is
going
to
be
sort
of
looking
at
completeness
of
the
document
and
maybe
some
of
the
terminology
used,
and
how
do
we
sort
of
align
the
section
so
that
they
flow
well
together
and
not
just
independent
sections.
B
H
E
So
so
long
Terri,
so
is
it
a
showing
like
you
know,
maybe
active
I
want
to
make
sure
I
understand,
listen
in
series.
I
will
everyone
else,
and
so
an
under
gave
given
to
we're
giving
given
two
weeks
to
finish
finish
the
first
draft
and
then
we're
going
to
have
some
star
reviews
for
among
this
work
group
people
right
and
then
welcome
to
higher,
coming
to
send
to
get
to
solicit
more
feedback
from
the
whole.
C
E
Good
yeah,
then
I
guess
probably
needs
some
time.
You
know
to
combine
them
together
because
of
the
flow
on
and
in
up
each
back,
but
I
think
it
makes
sense
to
combine
them
on.
B
And
we
sort
of
lay
them
out
in
the
would
be
sort
of
the
look
in
the
meeting
minutes.
We
kind
of
laid
out
each
section
in
order
and
we
make
like
like
Doug
Mitch.
We
may
end
up
combining
or
flipping
some
of
those
around
and
breaking
them
apart
and
adding
additional
sections.
So
it's
it's
not
a
hundred
percent
set
in
stone,
but
the
idea
is
because
sort
of
at
the
end
one
full
document
with
different
jackers
Oh.
B
Only
other
thing
I'm
going
to
bring
up
today.
Doug
is
you
know,
based
on
on
the
work
we've
done
so
far.
Did
anyone
have
a
strong
feeling
on
what
we
should
propose
to
the
TOC
in
terms
of
server
lists,
and
we
want
to
kind
of
bring
that
up
after
we
get
the
position
document
a
little
bit
more
complete
in
two
weeks.
H
Yeah,
you
know
my
agenda
in
general,
not
specific
to
this
project.
We
could
run
enable
more
collaboration
between
projects
because
typical
services,
there
are
a
lot
of
moving
parts,
but
you
have
ABI
gateway,
you're,
the
runtime
portion,
your
integration
with
CI
ACB
pipelines
in
situation
with
orchestration
and
right
now.
Every
project
is
the
implements
all
those
things.
Some
do
a
better
job
on
one
component
and
knowledge
good
job
on
the
other
component.
H
So
by
try
and
define
sir
the
layering,
we
can
probably
drive
more
collaboration
across
open
source
projects
and
each
one
will
sort
of
add
is
own
component.
That's
one
thing,
I
think
the
second
thing
is
terms
of
user.
You
know
usability
being
able
to
run
function
on
multiple
platforms
and
put
the
guys
may
be
the
one
participate
in
this.
You
know
working
group.
We
can
create
shims
that
are
thin
enough,
so
you
can
route
it
around.
H
G
I
agree
with
those
comments
and
I
also
say
that
if
we
were
in
the
future
able
to
double-click
down
into
like
the
event,
data
and
metadata
is
such
that
we
could
provide
guidelines
for
service
offerings.
Open-Source
service
offerings
to
be
able
to
send
events
into
the
various
systems
that
would
allow
for
more
more
open
source
services
to
then
be
able
to
interoperate
with
the
functions
of
the
service.
Another.
H
Area
is
a
function,
lacks
items
are
like
submitting
a
function,
creating
a
function,
publishing
etcetera,
think
it
will
create
a
common
model
that
people
could.
You
know,
in
some
cases,
just
create
new
logic
to
existing
frameworks
or
just
write
natively
to
that
we
enable
people
to
do
the
back
to
editors.
You
know
test
utilities
cetera
as
work
against
one
framework,
instead
of
creating
new
startups
that
all
they
do
is
just
creating
change
for
all
the
different
frameworks.
B
That
was
sort
of
like
you,
only
a
lot
of
what
I
was
thinking,
the
other.
The
other
thought
I
had
was
around
some
of
the
looking
at
some
of
the
projects
that
are
out
there
and
trying
to
determine
if
we
wanted
to
reach
out
to
those
projects
and
instead
talking
to
them
about
you.
Looking
at
joining
the
foundation.
H
H
B
B
B
B
B
E
I
have
a
yeah,
so
when
I
joined
I
thought
might
my
karma
calendar
if
this
event
has
been
changed
on
and
then
so,
I
tried
calling
from
the
numbers
with
all
three
members
and
and
the
u.s.
number
eight
does
not
work
when
I
call
through
this
link,
the
unique
it
eventually
works.
Okay,
so
could
clip
consumers?
Okay,
are
you
the
one
to
send
this
outlook
with
I
mean
basically
having
the
Google
image
imitation.
B
E
Yeah
I
think
it
would
be
great
if
at
this
meeting
into
the
the
latins
lot
of
this
page,
according
in
for
all
those
information
and
also
I,
don't
know
why
he
showed
up
on
my
issues
shows
that,
on
my
calendar,
this
event
has
been
changed.
Maybe
do
mind,
send
me
out
on
that
invitation
again,
so
that
a
classic
tip,
admiring.