►
From YouTube: CNCF Serverless Working Group 2020-01-09
Description
CNCF Serverless Working Group 2020-01-09
A
A
Hello,
oh
hey,
call
call
on
Jesus
Clemens,
hey
got
Scott
classier
there.
Yes,.
A
Hi
hello,
9:00
to
5:00
is
that
John
Michell.
That's.
A
G
I
J
H
He
this
is
Ryan
Horne.
He
he
works
with
me
and
it'll,
be
you
know
stand-in
when
I'm
not
available.
Okay,.
A
K
D
D
D
A
A
A
A
Wow
there
we
go
three
after
okay,
let's
started,
let's
see
a
eyes
one
worth
mentioning.
Is
this
one
right
here
mark
and
can
I
are
having
some
offline
discussions
to
figure
out
what
we
want
to
do
going
forward
in
case
you
guys
haven't
noticed.
Not
only
is
there
a
cig
app
now
in
the
CNC
F
there's
also
a
cig
runtime,
it's
a
per
vote,
so
it's
becoming
even
more
muddied
in
terms
of
where
we
might
fit
whether
it's
went
under
one
of
those
or
a
new
cig
or
what
so
we're
started.
A
Some
discussions
there
in
terms
of
what
kind
of
repose
we
want
to
bring
back
to
the
group
here
to
move
forward,
but
just
one,
but
you
guys
know
that
this
isn't
dead.
We
are
having
some
discussions
there.
Obviously,
if
you
had
some
feedback
you'd
like
to
provide,
let
us
know
and
I'll
pull
you
into
those
discussions.
It's
obviously
not
a
very
clear
answer.
A
Alright
:
forward,
we
do
have
a
redesign
of
the
websites
Luke
like
Aaron's
last
name.
Unfortunately,
he
is
a
CNC
F
guy,
though
he
did
a
whole
bunch
of
different
are
put
together
new
design,
which
I
thought
looked,
really
really
cool
and
I.
Think
that's
gonna
be
easier
to
manage.
The
old
format
was
a
little
bit
funky.
So
please
take
a
look
at
it.
If
something
looks
wrong,
please
let
us
know
we'll
get
it
fixed,
but
I
think
it
looks
really
good
anyway.
A
One
make
sure
you
guys
are
aware
of
that
koukin.
A
couple
things
here,
I
believe
by
I,
want
to
say
January
20th,
you
have
to
get
back
to
them
in
terms
of
what
sessions
you
want
to
host
they're.
Unlike
previous
times
working
groups,
only
get
one
session
so
I
believe
it's
a
35
minute
session,
full-blown
projects,
I,
think
incubator
or
graduate
status,
can
have
up
to
two
sessions,
typically
for
an
intro
and
deep
dive.
So
we
have
two
different
ones
that
we
want
I'm
open
to
anything
here.
A
But
given
it's
only,
you
know
two
months
away:
I
wasn't
sure
how
much
time
we
or
how
much
material
we
have.
They
actually
filled
two
different
sessions:
I
won't
I'll
open
the
floor
and
our
people
to
have
comments
on
this
but
we'd.
Obviously
you
don't
have
to
make
a
decision
today
about
this.
We
have
until
the
20th
to
decide,
but
is
there
anybody
who
has
any
comments
one
way
or
the
other
about
ideas
around
this.
A
L
A
A
Okay,
I
think
about
it.
Let
me
know
we
can
add
words
to
the
list.
I
do
like
that
idea,
though
at
least
that
would
allow
us
to
fill
two
sessions
for
sure,
so
that'd
be
good.
Okay,
so
as
I
think
about
it
and
we'll
talk
about
it
again
next
week,
so
this
practitioners
summits,
they
are
planning
on
having
one
I
believe,
is
on
no
I'm.
Sorry
go
ahead,
no.
I
Yeah
I
sorry
I
had
a
time
finding
the
unmute.
The
one
thing
that
we
talked
about
in
San
Diego
was
a
little
bit
of.
This
is
one
thing
with
just
the
code.
The
other
one
was
sort
of
kind
of
the
data
formats
and
what
do
the
events
look
like
and
so
forth,
because
I
know,
there's
been
a
lot
of
confusion
like
what
fields
go
where
and
so
forth,
and
we
talked
about
I
think
something
like
office
hours.
A
Yeah,
you
know
what's
interesting
at
the
last
coop
con
I.
Remember
there
being
a
whole
bunch
of
food.
This
were
is
almost
like
an
ask
me
anything
kind
of
thing
for
different
projects:
yeah
I
need
to
find
out
and
I'll
take
the
AI
to
do
this.
I
need
to
find
out,
if
crowded,
now
that
we're
at
the
incubator
stage,
whether
we
get
one
of
those
because
that
might
handle
the
office,
our
steamy
plugin
yeah,.
I
A
Into
that,
and
that
will
help
us
decide
whether
we
want
to
do
a
booth
kind
of
ask
me
anything
thing
or
whether
we
want
to
include
that
as
part
of
one
of
our
sessions
okay
hold
on
a
second.
My
phone
is
ringing.
Here
we
go
cool.
Thank
you.
They
any
other
comments
on
the
core
ideas.
Before
you
move
on
okay
practitioner
summit,
it
will
be
March,
30th,
I,
believe
the
CFP
for
that
will
start
I.
Think
it's
scheduled
to
start
on
Monday.
A
A
Okay,
as
I
said,
see,
people
should
start
on
Monday
I.
Believe
so
look
out
for
that.
If
you
want
to
submit
something
question
for
you
guys,
I,
don't
know
how
many
people
gonna
be
able
make
it
there,
but
I
was
wondering
whether
we
should
have
a
face-to-face
meeting
there.
I
was
thinking
if
nothing
else,
it
might
be
good
to
do
some
face-to-face
discussions
around
the
subscription
and
discovery.
Spec
I
find
face-to-face
meetings
tend
to
move
a
lot
quicker
than
offline
discussions.
I
want
to
see
what
you
guys
brought.
N
Hi
doc
this
is
Kathleen
hi.
So
for
the
face
to
face,
can
we
include
the
workflow
discussion?
I?
Think.
A
We
can
definitely
include
that
yeah
we
can.
We
can
see
if
we
can
find
a
couple
hours
and
split
up
the
time
sure
we
could
probably
do
that
if
we
need
to
yes,
okay
I
mean
okay.
Let
me
turn
this
around
sound
right.
Hearing
anybody
jump
up
and
down
too
much
except
for
Clements.
Is
there
any
objection
to
us
formally
saying
we're?
Gonna
have
a
face-to-face
there,
so
we'll
be
a
formal
meeting
and
stuff
like
that.
It's
not
there's
going
to
be
an
ad-hoc
get
together,
because
I
think
for
our
governance
rules.
A
A
Now
I
must
alright
cool.
Thank
you
guys.
Alright
moving
for
I
guess
I
should
ask.
Is
there
anybody
or
any
other
things
you
late
to
coop
comment,
I'm,
forgetting
to
mention
you,
okay,
next
decay,
I
think
we
do
have
a
meeting
scheduled
for
today.
I
never
actually
did
send
out
an
invite.
Oh
wait,
I
think
we
did
degree
the
next
one
would
be
today.
Do
we
have
a
need
for
a
call
today?
Do
it?
Does
anyone
have
any
topics.
D
I
have
a
question,
yes,
which.
A
N
So
we're
happy
working
under
what
clothes
back
actively
last
few
months
and
we
are
thinking
about
starting
on
the
work
group
aspect,
discussion
meeting
the
weekly
meeting
on
soon
so
I
have
created
a
a
doodle
poll
for
people
to
specify
the
time.
So
if
any
one
of
you
are
interested
in
that,
you
know
very
welcome
to
join
on
the
meeting
so
that
we
can,
you
know,
work
together
to
move
this
into
the
right
direction
and
also
to
make
the
spec
very
generic
I
can
post
that.
You
know
cool
yeah,
yeah.
N
Yeah,
so
we're
planning
to
start
the
meeting
either
meetup
this
month
or
maybe
starting
next
month,
depends
on
you
know:
people's
feedback
counter.
You
know
on
the
time.
A
Okay
moving
forward,
then
Thank
You
Kathy
not
going
to
go
into
too
many
of
the
issues
here.
I
think
I'm
trying
to
work
through
people
offline,
there's
just
two
things
I
wanted
to
mention.
One
is
this
one
then
Clements
I
was
hoping.
Maybe
you
could
take
a
look
at
it?
They
have
a
question
about
the
is
your
bus.
A
M
A
That
thank
you,
sir
and
alright,
so
we
can
either
blame
or
thank
I
think
it
was
Klaus
for
this
one,
but
he
noticed
that
our
documents
still
say
they're
working
drafts,
so
obviously
I
messed
up,
so
the
question
I
have
for
you
guys
is
how
do
we
want
to
fix
this?
Obviously
I
could
do
a
an
amend,
commits
and
move
the
tags
for
read
1.0
release.
So
no
one
ever
knows
about
the
screw-up.
A
Another
option
is
create
another
commit
for
the
1.0
branch,
but
don't
do
anything
else.
Another
one
is
to
start
the
process
of
creating
a
v1
101
I'm,
not
too
through
with
101.
Just
for
this.
What's
obviously
a
syntax
error,
but
I
wanted
to
see
what
it,
what
you
guys
thought
and
whether
you
guys
have
any
opinions
on
how
to
address
this
problem,
because
I
think
I
think
we
do
need
to
remove
that
text.
One
way
or
another.
A
A
J
A
A
Okay,
not
doing
any
moving
forward
alright
discovery
and
subscription
spec
work.
So
before
we
start
talking
about
the
spec
itself,
I
wanted
to
talk
about
sort
of
a
high
order,
governance,
kind
of
issue
or
process
kind
of
issue.
Let
me
get
down
to
the
air.
Alright,
so
I
started
writing
up
some
text,
because
I
think
we
need
to
update
the
TOC
on.
What's
going
on
with
our
work,
you
have
nothing
else
to
at
least
let
them
know
that
were
headed
in
this
direction
or
amber
starting.
A
This
work,
however,
as
I
was
writing
it
up
and
I
was
talking
to
Marc
about
this.
We
realized
that
there's
a
basically
high
worded
question
for
the
group,
and
that
is
this
new
spec.
They
were
working
on.
Is
it
a
cloud
events
project
or
is
it
a
service
working
group
project
because
that's
going
to
change
how
we
present
this
to
the
TOC?
A
So,
for
example,
if
it's
a
cloud
events
project,
then
I
don't
think
we
need
the
TOC
as
approval,
because
the
cloud
event
prods
the
kind
of
event
project
itself-
can
decide
whatever
it
wants
to
do,
and
but
we
do
need
to
make
the
TOC
aware
it
since
we
are
CEO
and
see
if
project.
So
that's
more
just
a
fYI
kind
of
a
thing.
A
This
is
something
we
want
to
work
on,
but
we'll
get
the
permission
to
start
it
up
as
a
new
project,
not
as
a
stand
box
project
yet,
but
just
as
a
new
piece
of
work,
that's
sort
of
an
exploratory
kind
of
a
thing,
and
then
in
that
sense
we
kind
of
kind
of
feel
like
we
need
to
get
their
permission
right.
So
I'd
kind
of
need
a
sense
from
the
group
here.
Do
you
guys
view
this?
M
J
E
You
know
we
just
saw
our
cloud
events
if
you
will
and
it
turned
out
to
be
useful
and
was
widely
adopted
by
lots
and
lots
of
other
services,
including
SNS
and
car
wash
logs
and
some
people
in
alexa
and
so
on
and
so
forth.
So
the
empirical
evidence
would
suggest
that
if
we
were
to,
you
know
specify
such
as
saying
the
notion
it.
Probably
it's
used
probably
wouldn't
be
restricted
to
cloud
events
in
practice.
B
Interesting
so
I'm
tempted
to
clarify
that.
Do
you
think
that
the
word
this
project
is
currently
would
be
able
to
be
a
sandbox
project
on
its
own,
or
should
it
should
we
and
I'm
gonna
coin
coined
term
here?
Should
we
incubate
it
within
cloud
events
for
some
period
of
time
and
then
split
it
out?
I'm
just
curious
your
opinion
on
that.
Oh
sure,.
E
A
A
No,
they
asked
for
people
who
view
this
as
sort
of
a
crowd
events
type
project
at
least
initially.
Does
that
mean
that
you
think
that
cloud
events
is
a
required
piece
of
it
or
it's
just
an
optional
thing
that
people
can
use?
For
example,
you
can't
filter
on
cloud
event
properties,
but
it's
not
a
requirement
that
that
be
a
first-class
thing
in
your
in
your
interaction
with
the
event
producer.
When
you
do
your
subscribe.
M
That's
you
know
broadly
useful
for
all
kinds
of
events
sources,
so
I
would
rather
I
would
rather
have
something
that's
specific
to
cloud
events,
but
covers
all
transports
rather
have
something
that
is
just
HTTP
and
covers,
although
the
entire
world,
in
terms
of
what's
the,
what
the
format
is.
Okay,.
A
J
I
think
it's
also
possible
to
write
a
spec
where
you
can
use
different
parts,
but
not
everything.
For
example,
in
Guelph
you
have
these
different
authorization
flows
like
four
different
ones,
and
you
can
use
one
you
can
use
another
or
you
don't
have
to
use
all
at
the
same
time.
So
I
think
the
same
principle
could
be
applied
here.
The
people
can
pick
which
part
I
use
and
then
it's
properly
declared,
which
part
I
use,
and
everyone
understands
what
part
of
the
spec
is
supported.
A
Okay,
so
it's
I
want
to
make
sure
I
understand
that
properly,
though
you
are,
you
are
leaning,
it's
not
like
you're,
leaning
more
towards
what
Mark
and
Clemens
are
saying,
basically
start
out
with
Cee
being
sort
of
scoping
function,
but
it's
not
necessarily
a
hard
requirement
that
you
have
to
actually
use
it.
I
mean.
J
What
I'm
trying
to
say
is
we
can
marketing
the
whole
thing
as
cloud
events.
Well
then
say:
this
part
of
the
subscribe
project
is
called
cloud
event
subscriptions,
so
you're,
specifically
supporting
only
that
part
or
product.
If
you
want
to,
for
example,
or
you
only
support
the
whatever
we'll
call
it
the
event
format,
that
would
be
the
spec
we
currently
have.
Okay,
all.
N
So
I
have
question
here,
so
my
understanding
of
the
subscription
discovery,
ace
should
be
Amy
needs
to
be
based
on
some
format
right
or
some
definition
of
the
events.
So
now
that
we
already
defined
the
crowd,
events
I
think
that's
naturally
part
of
the
cloud
event
scope,
because
if
we
want
to
discover
or
subscribe
to
another
on
definition
of
events,
we
first
need
to
have
I
mean
that
the
definition
or
the
format
definition
or
the
attribute
the
finish
of
that
type
of
event.
Right
so
I
think.
Naturally
it
falls
under
that's.
My
understanding.
A
H
I
think
this
is
a
Ryan
I
think
what's
going
through,
my
head
is,
if
we
start
building
out
more
more,
you
know,
features
Rebecca,
a
lack
of
a
better
term
that
is
coupled
to
cloud
events.
Then
using
those
independently
is,
you
know,
more
difficult.
They
have
to
buy
into
the
whole
ecosystem
right.
So
there's
something
to
be
said
about
you
know
having
making
it
easier
to
adopt
individual
pieces,
even
if
they
in
some
ways
can
be
coupled
or
made
better
if
they
were
using
flood
events.
H
There's
30
be
said
about
abstracting
it
out
so
that
they
are
not.
You
know
that
if
I
just
want
to
use
the
filtering
piece
or
the
subscription
piece
or
discovery
piece
I,
don't
I
can
use
that
with
you
know,
a
system
that
might
already
exist
that
isn't
using
events
versus
having
to
go
in
and
you
know
buy
into
the
whole
ecosystem
is.
H
A
Okay,
I
believe
then
the
majority
released
majority.
If
not,
everybody
seems
to
be
leaning
towards
starting
out
a
CD
Projekt
and
we
may
choose
to
work
it
off
or
pull
it
out
later
or
at
least
for
right
now
using
cloud
events
as
the
scoping
function
might
be
nice
or
it's
not
like
people.
Aren't
they
with
that.
A
So
is
there
any
objection,
then,
with
heading
down
that
that
path
and
I'll
work
with
mark
to
write
up
the
text
or
the
ideas
that
we're
going
to
be
then
present
to
the
TOC
and
give
you
guys
a
chance
to
look
it
over
before
we
actually
could
talk
to
them,
I
think,
probably
a
week
or
so
anyway,
and
then
they'll
take
a
lot
of
you
get
on
their
agenda.
Any
objections
with
that
all
right,
cool.
A
Okay.
So
that's
what
we
decided
all
right
moving
forward.
You
just
make
sure
one
thing
here:
agenda,
oh
okay!
So
at
some
point
we
don't
think
it's
a
decided
today,
but
at
some
point
we
need
to
decide
what
we're
gonna
call
the
spec.
So
we
can
actually
well
refer
to
it
by
the
same
thing:
I'm,
not
sure
clouds
description
is
the
right
one,
cuz
I'm,
not
sure
accurately.
This
covers
everything.
Maybe
it
doesn't.
This
cover
discovery,
and
it
does
depend
of
your
point
of
view.
Don't
have
to
decide
today
think
about
it.
A
If
you
have
alternative
ideas
for
a
name,
please
speak
up
someplace
in
the
doc
or
email
or
someplace.
We
can
start
thinking
about
it
and
possibly
take
a
vote.
If
there's
other
ideas
out
there,
I
don't
want
people
to
think
that
cloud
subscriptions
is
the
definitive
name.
You
can
obviously
change
that
next,
all
right,
so
moving
forward,
I
believe
the
current
scope,
basically
encumbers
encompasses
two
things,
obviously
discovery
and
then
subscriptions
I
every
now
and
then
over
the
last
couple
weeks
would
start
adding
a
little
bit
of
content
here
and
I.
A
Think
Klaus
actually
added
some
stuff
down
here
in
terms
of
event
domains.
What
I'd
like
to
do
is
I.
Don't
think
this
phone
calls,
let's
say
obviously
the
best
place
to
start
having
wordsmithing
discussions
or
writing
text
for
the
spec
I
think
we
should
do
that
offline.
However,
I
do
want
to
make
sure
that
we
aren't
gonna
just
sit
back
and
hope
that
someone
else
takes
the
pen
if
we
do
that,
no
one's
gonna
take
the
pen.
A
I
thought
might
be
interesting
here,
but
is
there
somebody
who
would
like
to
take
a
little
more
of
a
formal
pen
and
say:
okay,
here's
an
initial
proposal
for
the
list
of
metadata
that
you
would
get
back
if
you
were
to
query
and
event
producer
for
the
list
of
events
that
they
are
going
to
generate.
There's
somebody
who
wants
to
take
a
first
stab
at
that
be
on
with.
K
K
A
Okay,
Thank
You,
Kathy
I
have
some
opinions
here.
I
was
wondering
if
you're
gonna
speak
up
there,
Scott
okay,
so
unless
somebody
else
wants
to
jump
up
and
down,
but
this
let's,
let's
keep
it
relatively
small,
but
can
I
assume
that
maybe
Mike
since
you
were
the
first
one
to
step
forward
you'll,
take
the
action
to
reach
out
to
the
other
folks.
N
A
H
A
A
N
C
A
Okay,
obviously
keep
in
mind
that,
just
because
your
name
is
not
on
its
list
does
not
mean
that
you
should
not
or
think
that
you
cannot
just
put
your
random
thoughts
and
text
into
these
doctors
into
this
document
itself.
These
are
just
the
people
that
I
want
to
be
able
to
lean
on
and
poke
and
nag
it
as
appropriate,
as
we
keep
moving
forward
to
make
sure
we
keep
making
forward
progress.
Okay,
but.
A
A
Can
you
guys
do
me
a
favor
I
know
for
some
of
you
guys
that
you're
on
slack
and
we
could
look
your
slack
profile
to
get
your
email
well,
I'm,
not
sure
everybody's
on
slack.
So,
if
you
can,
can
you
guys
you
know
these
four
people
here
and
the
guys
appear?
Can
you
make
sure
that
your
email
address
is
in
here,
so
that
Mike
and
Clemens
know
how
to
reach
you
guys
offline.
N
How
EverQuest,
how
should
we
do
this?
Should
we
click
on
this
back
and
start
to
putting
our
thoughts
or
should
we
start
I
mean
a
meeting
for
a
discussion
and
then
how
should
we
do
this
so.
A
A
Okay,
with
that,
let's
turn
this
back
around.
Do
you
is
there
anything
about,
say
discovery
that
you
guys
want
to
talk
about
on
the
call
right
now,
because
we
had
basically
have
a
half
an
hour
and
I
think
we're
technically
at
the
end
of
the
agenda,
but
if
we
want
to,
we
can
use
this
time
to
talk
about
high-level
things
like,
for
example,
does
everybody
think
everybody
on
the
same
page
in
terms
of
what
the
scope
is
for
what
discoveries
in
a
cover?
A
N
A
Yeah
we
basically
took
a
vote
and
out
of
all
the
various
topics
and
functional
signature
was
one
of
them.
This
is
where
we
landed
that
this
got
the
that
this
won
the
vote.
Basically
working
on
subscription
and
discovery
API,
we
did
talk
about
potentially
having
some
offline
discussions
about
function,
signatures
in
parallel,
I,
don't
think
anything's
happened
with
that
yet,
but
in
terms
of
as
in
terms
of
first
class
next
work
item,
it
did
not
win
the
vote,
but
that
doesn't
mean
people
can't
have
discussions
offline.
A
Oops,
okay,
good
oops
answer
my
question.
They
give
so
I
got
a
question
for
you
Ryan.
When
can
you
be
a
little
elaborate,
a
little
on
when
you
talk
about
this
thing
being
prescriptive
because
in
my
mind,
I
think
if
we're
gonna
specify
some
sort
of
filtering
thing
that
in
my
mind
it
means
what
is
the
filtering
expressions
or
syntax
or
whatever
in
our
call
it
of
the
request
that's
sent
over
to
the
event
producer.
Are
you
doing
to
your
subscription?
A
H
That's
pretty
much
what
I
was
asking,
and
there
are
some
questions
that
come
out
of
that,
such
as
you
know,
if
I
specify
a
filter,
how
do
I
get
any
feedback
on
what
that
filter
matches
in
terms
of
event,
types
and
versions
when
it
comes
to
using
cloud
events,
so
I'm,
just
kind
of
thinking
through
you
know
how?
How
much
of
that
we
leave
out
to
the
system's
implementing
this
versus
being
in
the
spec
itself.
M
Prior
our
reference
AP
has
a
filter
mechanism
and
that's
on
the
source,
which
means
we
way
of
getting
things
from
and
there's
effectively
two
levels
of
selection
where
at
first
you
are
effectively
selecting
your
source
for
lack
of
a
better
term
from
the
server.
You
basically
have
a
URI,
that's
pointing
to
two,
sometimes
in
the
sir
and
then
there's
a
filter,
expression
and
the
in
capice
effect
itself
stays
away
from
being
specific
about
that.
M
What
that
is,
but
it
says
there
is
a
filter
and
then
there
are
extensions
defined
for
a
MVP
that
then
specifies
some
filter.
Some
filters
attacks.
That's
one
way
out!
We've
done
that,
because
there
are
obviously
various
ways
of
how
you
could
do
this.
You
could
do
writing
ax.
You
could
do
a
sequel,
you
could
do
all
kinds
of
ways
of
how
to
define
a
filter,
and
that
seems
to
work
fairly
well,
so
I
would
I'm
not
sure
whether
we
should
bold.
You
know
the
one
and
only
way
into
the
into
the
specification.
H
I
think,
generally,
you
know
I
think
you
know,
as
we
dive
into
see
the
specifics
of
this,
we'll
probably
have
more
detailed
discussions
around
how
we
expose
it,
but
I
think
your
original
answer,
Doug
was
answered,
answered
my
comment,
which
is
yeah,
but
it
it
will
be.
There
will
be
some
specification
of
how
filters
will
work.
They
won't
just
be
you
know
a
filter
field
and
left
up
to
the
the
systems
employing
to
us
to
figure
out
what
that
actually
means
and
I.
Take
that
total
free.
F
Well,
I
think
that
I
also
wrote
this
in
the
questions
on
the
comments
before
holiday
season.
I
wonder
if
this
is
really
gonna,
be
a
contract
between
producer
and
subscriber
or
I
mean
maybe
yes,
so
because
I
mean
we
have
the
cloud
event
spec,
which
enables
us
to
send
events
across
multiple
hops
to
do
routing
and
everything
so
should
release
be
subscription
and
discovery.
Be
then
a
thing
directly
between
the
consumer
and
the
producer
that
seems
to
be
a
bit
asymmetric.
F
F
Suppose
you
will
have
producers
and
consumers
and
authorizations
for
them,
and
and
maybe
what
I
would
call
event
catalogs
where
you
maintain
a
list
of
what
is
available
so
and
in
this
scenario
a
consumer
would
query,
probably
the
eventing
domain
for
event,
types
or
sources
available
in
that
domain
or
the
domain
could
also
organize
the
routing
to
other
domains
and
maybe
cache
the
catalogs
of
other
domains.
That's
just
I
mean
right
now,
just
an
idea
to
have
this.
This
notion
of
the
eventing
domain,
so
a
consumer
would
would
do
subscriptions
via
the
domain.
A
F
F
E
A
Its
kind
of
I
was
asking
question
because
I
guess
in
my
mind,
I
always
kind
of
assumed
that
we're
gonna
define
this
subscription
type
model
and
and
even
a
discovery
type
model
where
the
client
who's
talking
to
these
things
doesn't
know
or
care
whether
they're
talking
to
a
in
essence
single
event
producer
or
an
entire
set
of
event.
Producers,
that's
behind
as
you'd
call
it
here
in
a
dental
Maine
right
and
this
kind
of
actually
gets
to
a
topic.
I
was
gonna,
bring
up.
A
Kind
of
grouping,
or
how
are
things
going
to
be
returned?
They
put
it
that
way.
Right?
Are
you
getting
back
a
list
of
event
sources
and
then
there's
the
list
of
events
for
each
event
source
or
you
get
back
a
list
of
events,
and
it
includes
the
list
of
event
sources
that
you
might
be
associated
with
it
or
you
know
and
baby.
A
K
Yeah,
so
you
know
this
this
one.
We
thought
about
a
lot
about
in
in
terms
of
thinking
about
representing
all
of
the
various
events
sources
that
we
would
have
just
at
Google
as
a
provider,
for
example,
and
think
about
coming
at
it
as
the
various
event
types
that
are
available
for
us.
First,
it's
just
going
to
be
like
a
deluge
of
information
for
somebody
who
wants
to
consume
something
and
that
you
know
our
customers
typically
think
about
it
as
the
the
event
source
first.
K
So
it's
like
a
funnel
right,
the
Dilys
specific
to
most
specific,
you
know,
I,
know
I,
want
something
from
say.
You
know:
Google
compute,
engine,
ok,
weather
event,
types
does
that
have
and
thinking
about
the
funnel.
That
way,
the
thing
that
doesn't
address
is,
if
you
have
event
types
that
you
may
be
able
to
get
from
multiple
producers,
but
I
think
that's
more
of
a
special
case,
and
you
could
provide
a
pivot
for
that
functionality.
G
H
I
did
have
one
comment
just
so:
we've
seen
it
come
from
the
other
way
as
well
at
Trulia,
where
a
customer
might
want
to
receive
events
for
a
particular
source.
So,
for
example,
a
phone
call
that
they
make
through
Twilio.
They,
you
know,
are
running
their
own
state
machines
on
their
side
and
they
want
to
subscribe
to
events
from
that
one
particular
call
or
an
IOT
device
or
SIM
card,
etc.
So,
I
do
think
that
you
know
thinking
about
it
from
both
ways
are
both
equally
valid
for
us.
A
A
A
A
A
Yep
looks
like
I
see,
Falco
time
to
come
offline,
so
I'll
give
him
credit
for
that.
Lionel
you're
there
right
here:
okay
and
Eric,
ooh
I'm,
you
know
all
right
anybody
else,
I
missed
for
the
roll
call,
hey
Doug.
This
is
Christian
New
Year,
Happy,
New
Year.
Oh,
you
know
what
I
meant
to
say
to
be
in
the
co:
Happy
New
Year,
everybody
I
completely,
forget
to
say
those
kind
of
things
all
right.
Anybody
else
do
I
miss
okay.
A
In
that
case,
if
you
are
not
interested
in
the
SDK
side
of
things,
you
are
free
to
drop
and
we'll
talk
next
week
and
please
don't
forget
to
add
comments
to
the
the
new
spec
as
you
can,
and
thank
you
guys
very
much
we'll
talk
to
you
next
week
and
we'll
start
the
SDK
call
in
about
a
minute
or
so
thank
you.
I'm
good,
bye,
bye,.
J
M
G
M
The
worst
thing
was:
is
that
because
of
the
sand,
and
in
generally
protecting
the
environment
city
of
Seattle
and
all
the
the
towns
around
Puget
Sound's
do
not
put
salt
on
the
street,
and
that
is
fairly
devastating
because
it's
you
know,
there's
a
ton,
there's
a
ton
of
hills,
which
means
you
have
city
think
there's
annually
when
there's
a
little
bit
of
snow
on
the
street.
There
are
any
only
you
see,
pictures,
YouTube
videos
and
everything
of
city
buses,
just
sliding
down
the
hills.
A
A
A
E
B
L
Where
we
continue
to
push
on
this,
this
concept
of
you
can
you
can
pick
what
how
much
of
the
opinion
the
SDK
provides
you.
So
you
can,
you
can
use
it
without
the
client.
You
can
use
it
without
the
transports.
You
can
use
it
just
as
an
object
that
marshals
itself
in
and
out
of
the
format's
supported
so
I
think
to
answer
that
question
properly
would
be
I'd
have
to
list
out
all
those
layers
and
say
which
ones
are
stable
and
which
ones
aren't.
B
L
L
That
that
answer,
like
what
I
mean
it
might
be
kind
of
pertinent
to
start
up
a
go.
Sdk
kind
of
like
bi-weekly
call
for
a
quick
check-in
for
people
that
are
using
it,
have
questions
and
want
one
answers
or
need
features
or
or
have
bugs
or
issues
or
whatever.
There's
some
folks
doing.
Some
performance
improvements
in
the
going
SDK
right
now
and
the
memory
overhead
got
a
lot
reduced
or
it
will
be,
and
as
soon
as
those
PR
merges
I.
B
E
L
L
D
A
M
L
C
A
E
L
A
Let
me
ask
this:
is
it?
Is
it
a
question
of
the
document?
It
sounds
to
me,
like
might
be
a
question
of
documentation
in
terms
of
the
website
explicitly
saying
hey
this
thing's
stable,
as
well
as
some
sort
of
formal,
minimal,
1.0
type,
release,
kind
of
a
thing
right
cos:
Bobby,
we
don't
think
anything's
real
until
it's
well
at
least
1.0
I.
Think.
E
A
Okay,
let
me
see
if
I
can
poke
on
the
obviously
C
sharp
and
go
are
covered
by
the
guys
on
the
call
here.
But
let
me
take
the
action
item
to
poke
the
other
folks
offline
and
see
if
we
can
get
them
to
be
a
little
more
formal
with
their
releases
and
and
documentation
and
stuff
like
that
and
see
if
we
can
get
something
more
concrete
out
of
them.
A
A
By
belay,
okay,
are
there
other
STK
topics
in
general?
The
people
want
to
talk
about.
L
And
then
other
people
can
continue
to
use
that
that
client
and
it
it
makes
a
couple
opinions,
but
basically
it's
very
flexible
for
the
implementer
to
implement
exactly
what
it's
what
it
means
to
do
certain
things,
but
the
cloud
events.
But
the
specification
doesn't
really
talk
about
this
client
piece
and
so
I
was
I
was
wondering
if
there
it
is.
L
E
Can
only
speak
from
my
own,
you
know
usage,
but
that
would
be
great
because
you
know
I
were
interested
in.
You
know
generating
cloud
events
and
parsing
cloud
events,
but
you
know
using
various
different
kinds
of
underlying
transport
and
so
on.
So
a
lot
of
stuff
in
the
client
library
is
probably
not
terribly
interesting,
so
it
would
be
nice
to
have
just
the
data
and
stuff
yeah.
O
M
E
A
Just
to
have
a
slightly
contrary
and
view
I
I
would
actually
prefer
if
they
were
all
on
the
same
repo,
but
have
a
clean
separation
just
because
I
think
they
go.
They
go
hand-in-hand
so
much
that,
even
if
you
are
like
in
Tim's
world,
where
you're
only
gonna
use
one
one
side
of
it,
I
think
there
will
be
enough
people
that
will
want
to
use
the
transport
side
of
it,
that
having
them
sort
of
look
at
the
documentation
or
deal
with
two
separate
repos
might
be
more
annoying
for
people.
So
as.
A
A
Yeah,
exactly
maybe
you
think
correct,
I
kind
of
look
at
it.
As
you
know,
the
simplistic
thing
is
hey
that
you
know.
I
can
use
these
transport
things
and
handles
the
cloud
events
and
the
cars
all
this
wonderful
stuff
for
me,
but
if
I
need
to
go
around
those
and
do
with
things
more
advanced,
I
won't
have
to
one
I.
Don't
want
to
have
the
jump
over
to
separate
repo
I
want
to
look
at
it
as
more
like
an
advanced
feature
and
jumped
through
the
sea
library
itself.
Right.
L
A
F
So
doc,
you
might
remember,
we
had
this
discussion
and
issue
around
nested
events
and
I
was
wondering
so
I
had
this
action
items
were
to
finally
at
some
sample.
How
did
you
were
to
nesting
with
events
and
when
I
was
working
on?
This
I
went
with
this
wealth
problem
that,
if
I
use
the
example
of
having
of
nesting
a
structured
mode
event
into
a
binary
mode
event
that
right
now,
it
wouldn't
really
work
with
with
SDKs,
basically,
because
we
recommend
to
to
distinguish
between
binary
in
structured
mode
by
looking
at
the
content
type.
A
Because
you
can
talk
about
this
and
the
spec
right
now,
as
Klaus
said,
it
doesn't
normally
say
it,
but
it
definitely
says
it
that
you
should
look
at
the
content
like
to
decide
whether
it's
a
cloud
event
or
not.
I
was
wondering
whether
we
should
modify
the
text
to
say
look
for
the
presence
of
the
C
spec
version
header
to
know
whether
it's
binary.
Oh
no,
no.
L
C
L
M
O
M
You
send
a
structured
event.
It
is
okay
if
you
go
and
use
the
same
mechanism
as
for
the
as
for
binary
by
by
mapping
your
attributes
into
the
header.
So
that's
the
infrastructure
can't
see
them,
because
we
have
to
assume
that
the
infrastructure
has
no
visibility
into
the
payload
the
encrypted
or
not.
M
The
nesting
case,
where
were
would
be,
if
you
so
so
it
could.
So
if
you
stare
at
this
one,
that's
actually
the
this
could
serve
as
an
example.
If
you,
if
you
send
that
caught
the
the
event,
my
event
and
that
event
my
event
carries
as
payload
this
inner
event,
then
what
is
the
right
comes
in
sight,
I.
L
L
A
So,
let's
back
up
a
second
here
because
Scott
you,
you
said
something
interesting,
it's
being
a
discussion.
You
said
that
you
thought
this
was
invalid
cloud
events
I.
Is
it
or
is
it
not
because
I'm
inclined
to
say
it's
not
invalid?
It's
just
not
necessarily
the
right,
descriptive
text
for
it,
because
I
don't
think
you
have
to
necessarily
have
the
the
cool
event
type
match
the
C
type.
My
event
on
the
outside
I,
don't
think
you
have
to.
A
L
Wait
a
minute:
no,
the
outer
one
is
binary.
The
inner
one
is
structured,
but
the
content
type
in
the
on
the
headers,
as
is
supposed
to
apply
to
the
outer
piece-
and
this
is
talking
about
the
inner
piece-
and
this
is
the
only
signal
we
really
have
to
know
what
the
body
should
look
like
for
binary.
Hi
I
think
that
this
should
be
content,
type
application,
JSON
plus
cloud
events,
because
that's
the
extension
instead
of
how
to
interpret
the
entire
payload
well,.
A
A
A
Right
so
it's
almost
like
a
dual
thing
is
like
okay.
Is
this
a
binary
cloud
events?
Yes,
did
you
see
the
C
spec
version
header?
If
the
answer
is
no
it's
then
you
don't
see
that
header.
Then
you
look
at
the
cloud
event.
I'm.
Sorry,
you
look
at
the
content
type
and
say:
okay,
is
this
a
structured
cloud
event
then
did
in
this
case?
Yes,
I
just.
L
Think,
okay,
we
look
at
this
and
drop
the
my
event
attributes
yes,
because
because
of
the
fact
that
you
can
mirror
or
sorry
what
was
project
up
the
inner
pieces,
what
is
those
headers
there
that
our
cloud
event
headers
are
projected
and
we
can
be
dropped
because
we
can
just
look
in
the
body
get
the
same
data
well.
A
But
wait
a
minute:
Scott
I
I
agree
with
the
net
result.
I
think
well,
I
agree
with
your
interpretation
other
than
that
result,
but
I
disagree
with
why
it
happened.
I
think
I
think
you're
dropping
it
because
content-type
says.
Oh,
this
is
a
structured
cloud
event.
I,
don't
think
you
dropped
it
because
of
you
who
making
it
a
constant
decision
to
say.
Oh,
the
content
is
duplicated.
I.
Think
you,
you
purposely
don't
even
look
at
the
headers
because
of
the
content
type
value
I.
That's.
L
M
A
But
I
think
but
I
think
it's
interesting,
though,
because
if
the
spec
and
I
think
I
suspect
you
did
that
because
the
spec
says
look
at
the
content
type
header
to
make
this
determination
as
to
whether
it's
binary
or
not
yeah
and
I'm,
suggesting
to
you
that
if
the
spec
said
the
presence
of
the
CG
spec
version,
header
says
that
it's
binary
I
think
you
would
have
coded
it.
That
way,
and
this
may
not
have
been
an
issue
for
you
well,
but.
M
M
O
M
A
M
A
Okay,
that,
let's,
let's
let's
change
the
scenario
here
slightly-
let's
say
instead
a
content,
type,
application,
Jason
I'm.
Sorry,
instead
of
application
crowd
events,
please
Jason.
Let's
say
it
was
just
application:
Jason
right,
yeah:
the
fact
that
the
the
fact
that
the
body
looks
like
a
crowd
event
would
be
irrelevant
to
the
receiver.
If
right
now,
you
would
then
take
that
Jason
and
perhaps
based
upon
the
content
type,
do
some
routing
or
you
just
pass
onto
the
application,
and
it
would
be
expected
to
figure
out
what
to
do
with
it
right,
yeah.
A
M
L
So,
okay
go
ahead,
show
different
example.
What
do
you
take
this
exact
same
format
that
you're
showing,
except
actually
do
the
propagation
up,
where
my
event
becomes
cool
event
and
xxx
is
the
ID
and
things
like
that
the
source
is
big
cocom.
What
how
would
you
interpret
that
is
it?
Is
it
a
double
wrapped
cloud
event
or
is
it
the
promoted
piece
I
would.
A
Say
in
all
cases,
the
application
is
going
to
get
a
chunk
of
JSON.
That
looks
like
what
you
see
there
on
the
screen
and
in
the
body,
and
there
may
be
additional
metadata
carrying
alongside
that
says.
This
is
jason
and
it
and
and
that's
just
extra
metadata
to
me.
This
is
a
binary
cloud
event
and
the
payload
just
happens
to
look
like
a
cloud
event,
but
the
infrastructure
doesn't
know
or
care
about
it.
It
just
passes
it
along.
A
F
A
Think
that'd
be
perfectly
valid
for
it
to
do
so,
but
I
also
think
it'd
be
valid
for
it
to
have
the
smarts
to
know
that
it.
This
isn't
just
generic
Jason,
it's
a
crowd
of
in
Jason,
and
it
can
put
the
word
cloud.
Events
in
the
content,
type
I,
don't
think
the
presence
of
the
word
cloud
events
in
there
changes
the
processing
at
all.
It
definitely.
A
A
L
J
A
M
A
M
A
L
L
A
L
M
This
is,
this
is
famously
the
this
is
malformed
because
it
should
be
coming
its
start.
The
content-type
says
this
is
structure
and
so
now
you're
at
Liberty.
If
you
want
to
promote
the
headers
out
from
the
inner
from
the
inner
cob
events
into
those
headers,
and
then
the
infrastructure
can
see
them,
but
the
the
authoritative
event
that's
being
carried
is
the
structure.
My.
L
Recommendation
would
be
to
add
a
new
content
type,
that's
application,
JSON
plus
cloud
event,
and
if
binary
mode
would
like
to
carry
a
structured
cloud
event
it
the
body
is
first
application
JSON,
it
happens
to
be
formatted
as
a
cloud
event.
So
the
cloud
event
part
is
the
extension
now
I
see
why
you
said
that.
Thank
you.