►
From YouTube: CNCF Serverless WG Meeting - 2019-04-04
Description
Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
A
Alright
3
after
the
hour,
while
we
go
and
I
get
started,
we'll
catch
up
with
you
later
smuggler.
Today,
let's
see
a
is
I
know,
Rachel
said
she
had
a
chance
to
hers.
Yet
so
I'll
skip
that
community
time.
Are
there
any
issues
from
the
community
at
large
people
like
to
bring
up
alright
moving
forward?
We
have
not
an
SDK
meeting,
I,
don't
think,
there's
anything,
therefore,
to
mention
at
least
I,
can't
think
of
anything
going
on
too
exciting
with
the
SDKs
I.
A
Actually
guess,
I
should
mention
one
thing:
we
have
a
lot
of
activity
going
on
in
the
goaline,
SDK
and
I.
Think
I
made
occasionally
see
a
couple
comments,
I'm
one
of
the
other
SDKs
on
it's
that
camera,
which
one
it
may
be
the
jumper
one,
but
a
fair
number
of
the
SDKs
seem
relatively
quiet,
and
that
kind
of
concerns
me
a
little.
A
So
if
you're
part
of
an
organization
or
just
you
personally
are
working
on
whether
the
SDKs
or
have
an
interest
in
it,
try
to
make
sure
that
there
is
stuff
going
on
there
because
I'm
pretty
sure
most
of
them
are
behind
the
specification.
I
think
the
only
one
that's
probably
most
up-to-date
is
the
go
line.
One
Clement
I,
don't
know
where
the
C
sharp
line
is,
but
I
just
see
the
most
activity
going
on
the
go
line.
B
Now
I'm
I
see
I'm
a
little
bit
behind
also
because
we
didn't
we're
kind
of
in
the
interim
stage,
so
it's
kind
of
difficult
to
label
the
new
stuff
with
something
right.
But
so
that's
that's
that's
my
reason
for
it,
like
I,
haven't
discriminator
effectively
before
the
new
stuff
like
I,
could
call
it
zero
point
to
this
or
something.
But
you
know
yeah.
A
And
that's
fine
I
think.
The
only
reason
it's
kind
of
interesting
to
me
is
because
at
one
point
in
time
we
did
talk
about
possibly
doing
some
sort
of
interrupts
demo
at
coup,
Connie
you
and
I,
and
we
recently
decided
not
to
do
it
mainly
because
of
the
low
activity
we
see
going
on
and
they
SDKs
that
because
then
we
assume
people
wouldn't
necessarily
have
time
to
to
fix
up
the
SDKs
in
such
a
short
period
of
time,
and
so
it's
kind
of
a
self-fulfilling
prophecy.
A
A
Moving
forward,
let's
see
there
for
the
next
demo,
we
have
had
some
phone
calls
to
discuss
where
we
are
with
that.
I
won't
go
into
the
details
here,
other
than
to
say
I
think
we
are
making
really
good
progress.
We
have
a
fairly
well-defined
scenario.
It
is
around
the
airport,
one
that
we
talked
about
in
the
past.
We
are
gonna
steal
the
time
after
this
phone
call
to
continue
some
discussions.
A
A
We
we
do
have
a
dashboard,
that's
coming
together
route
relatively
nicely
I,
don't
think
I'm
gonna
be
able
to
show
it
during
that
meeting
after
this
one,
but
hopefully
later
today,
I
will
have
something
up
on
source
dog.
So
you
guys
take
a
look
at
so
you
can
see
how
it's
progressing
and
then
you
can
make
suggestions,
but
anyway,
keep
an
eye
out
for
those
things
going
on.
D
D
Well,
I
took
a
lot
of
the
information
and
tables
that
I've
been
presenting
in
in
a
slide
deck
and
I
consolidated.
All
of
that
into
a
word
document
that
I
was
hopeful
that
Scott
could
help
convert
that
back
to
a
Google
Doc
for
collaborative
comments.
A
A
Moving
forward,
COO
Connie,
you,
as
I
said
we
have
a
scheduled
meeting
right
after
this
one,
but
don't
have
any
topic,
so
it
might
be
very,
very
short,
I,
don't
think
anything
much
to
mention
there
nothing's
really
changed
the
the
serverless
practitioners
summit,
the
CFP
closed
last
week,
I
think
or
something
like
that.
There
was
still
going
through
reviews
of
those,
and
so,
if
you
did
submit
something,
I
expect
a
notification.
I
believe
that
about
about
a
week
or
so
we
did
not
have
too
many
proposals
actually
submitted.
A
A
Eastern,
mark
and
I
were
invited
to
join
CN,
CF
and
user
forum
of
feedback
call,
which
is
basically
just
an
opportunity
for
the
end
user
community
to
talk
to
different
working
groups
within
the
ciencia
for
different
projects
in
CN
CF
about
their
about
their
project
and
ask
questions
in
both
directions
actually
and
so
we're
scheduled
to
go.
As
I
said
next
Tuesday
11:00
a.m.
Eastern,
there
is
a
link
to
the
agenda
doc.
If
you
don't
have
access
to
that,
let
me
know
and
I'll
see
if
I
can
get
you
access
to
it.
A
Basically,
the
reason
I'm
putting
it
here
is
so
that
you
guys
can
see
the
list
of
questions
or
the.
What
we're
gonna
actually
talk
about.
So
you
should
be
able
to
see
on
the
screen
right
now.
I
was
basically
just
going
to
provide
a
quick
overview
of
the
service
working
group
as
well
as
the
cloud
events
project
and
then
lead
into
the
questions
that
we
had
for
them,
basically
just
finding
out
what
they're
doing
with
functions
and
server
lists
and
stuff
like
that.
A
What
their
pain
points
are
those
kind
of
things
if
there
are
questions
that
you
guys
would
like
to
have
us
add
to
that
list.
If
you
don't
have
access
to
the
stock,
just
drop
us
a
note
and
we'll
add
it
to
it
and
we'll,
let
you
guys
know
the
outcome
of
the
meeting
and
share
these
minutes.
If
you're
not
allowed
to
see
the
doc
okay,
any
questions
on
that
all
right
moving
forward,
then,
let's
see
b3,
okay.
A
So
one
point
in
time:
different
people
have
mentioned
the
possibility
and
I
myself
included
of
possibly
getting
to
0.3
and
time
for
coop
Connie.
You
I
do
believe
it's
still
technically
possible.
We
may
have
to
speed
things
up
a
little,
but
in
terms
of
the
roadmap
that
we
have
listed,
4:03
I
think
we're
making
it
I
actually
paste
it
here
into
the
dock.
I
think
we're
making
some
really
good
progress
on
the
optional
attributes.
A
I
think
we're
slowly
getting
those
behind
us,
Sakura,
lated
ones,
I
believe
we've
actually
addressed
the
one
security
issue
that
I
think
Jem
opened
up
our
PR.
The
gem
opened
up
I'm,
not
actually
aware
of
any
security
related
issues
in
our
backlog.
So
if
you
think
I'm
missing
one
or
if
you'd
like
to
raise
one
to
get
out
of
the
list,
please
do
it
sooner
rather
than
later.
So
we
can
try
to
make
sure
we
address
those
in
time.
A
A
E
A
And
now
I
think
that
was
gems
PR
that
went
and
recently
to
address
that
I
believe
so
so
Evan.
You
may
want
to
take
a
look
at
gems,
PR
the
merge
recently
and
make
sure
it
addresses
your
comment,
I'm
pretty
sure
it
does,
but
other
side
Afiya.
There
are
other
sphere
related
issues,
please
open
up
those
sooner
rather
than
later,
so
we
could
try
to
address
those
relative
to
practical
use
issues.
I
know
that
we
had
the
the
size
limit.
One
I,
don't
believe
anybody
has
really
done
much
with
the
other
ones
out
there.
A
A
I,
don't
think
we
have
time
this
I
do
a
deep
dive
discussion
about
it,
but
I
want
to
point
people
to
the
document
that
we're
supposed
to
have
the
discussion
around
that
so
I
think
that
will
address
most
of
those
things
you
mentioned
Devon,
but
some
of
the
other
things
that
you
don't
think
they're
covered
someplace
and
there's
no
issue
for
it,
just
open
it
up.
Well,
there
is
an
issue
and
you
think
I've
vacant
date,
I've
ignored
it
by
mistake.
Ping
me
and
I'll
make
sure
it
gets
added
to
the
agenda
all
right.
A
All
right
moving
forward
may
7th
you're
going
to
have
our
yearly
status.
Slash
review
with
the
CN
CF
TOC
I
have
not
started
working
on
the
presentation
for
that.
Yet
I
will
just
get
that
started
fairly
soon
and
what
I
do
I'll
make
it
available
for
you
guys
to
review
and
comment
on
any
and
all
review
would
be
very
much
appreciated.
A
Someone
mentioned
the
possibility
of
us
going
to
incubator
around
the
same
time.
I
checked
the
loot,
the
graduation
criteria
to
go
to
incubator
and
I.
Think
this
little
bit
of
text
right
here
is
probably
the
biggest
hurdle
for
us,
which
is
documenting
that
we
have
at
least
three
independent
and
users
now
have
an
open
issue
to
the
TOC
to
define
what
end
user
means
in
our
particular
case
right.
So,
for
example,
does
that
actually
mean
cloud
provider
or
is
it
a
user
of
the
cloud
provider
I'm
going
to
try
to
get
some
clarity
around
that?
A
Okay,
so
I
said
I,
don't
want
me
to
say
to
discuss
it
here,
just
when
I
bring
it
off
the
killer's
attention
to
think
about
all
right
before
we
jump
into
the
PRS.
Are
there
any
other
high-level
topics,
people
like
to
bring
up
that
I
may
have
forgotten
to
mention
all
right,
let's
jump
into
PRS,
then
all
right,
so
this
PR
whoops
11
things
are
getting
in
my
way
all
right,
so
this
PR
has
been
out
there
for
a
very
long
time.
A
James
I
believe
his
name
is
James
Roper.
He
wanted
to
add
an
event
key
to
our
spec,
mainly
for
transfers
like
Kafka
up
more
than
anything
else.
We
went
back
and
forth
quite
a
bit
on
here
and
where
we
ended
up
was
rather
than
adding
a
partitioning
type
of
property
to
the
mate
spec
itself
or
defining
an
extension,
and
we
decided
to
make
it
so
that
the
definition
that
buddy
Razoo
proposed
actually
hasn't
changed.
It's
still
used
for
defining
a
casual
relationship
or
for
doing
some
sort
of
grouping
of
events.
A
The
biggest
thing
that's
changed
recently.
Is
this
section
right
here
and
hope?
Let
me
and
I
like
that.
So
that's
next
read
it.
This
pick
a
sentence
that
start
that's
highlight
a
little.
The
biggest
point
here
is
that
this
property
actually
can
change
if
the
event
travels
through
multiple
hops,
because
I
believe
Clemens.
You
were
very
concerned
about
the
fact
that
it
may
be
impossible
or
it
may
the
business
logic
at
each
hop
may
actually
change
the
buckets
for
categorization
purposes
yeah,
and
so
this
allows
for
that
to
actually
happen.
A
G
A
I
believe
it's
I
believe
the
intent
is
separate.
Events
that
may
have
some
sort
of
relationship
in
terms
of
the
bucket
in
which
they'll
be
placed
into
at
the
destination
so
resemble
I
believe
helping
to
classify
what
queue
they
may
go
ahead,
go
into
at
the
destination.
I
don't
have
that
right.
Clemens
is
just
one
example.
A
Jc
Cathy
was
asking
about
this
phrase,
multiple
events
and
whether
it's
the
same
event
replayed
or
whether
it's
completely
independent
events
and
I'm,
saying
it's
independent
events,
yes
and
the
relationship
that
we're
doing
here
is
for
as
an
example
to
figure
out
which
cue
to
put
an
event
into
when
it
reaches
its
final
destination.
Yeah.
B
There's
there's
there's
typically,
some
level
of
correlation
between
those
events,
because
partitioning
is
is,
is
creating
buckets
that
are
then
easy
to
process
from.
Basically,
you
have
a
torrent
of
events
that
comes
it
in
one
side
and
then
you
want
to
go
and
pre
partition
it.
So
it's
easy
to
go
and
process
them,
so
there's
an
implied,
implied
relationship
between
those
events
like
the
same
origin,
etc
so
that
you
are
able
to
then
process
them
in
context
and
that
in
that
infrastructure,
where
they
landed,
they
are
in
about
the
right
order.
B
F
B
B
That's
following
that,
where
you
have
kind
of
pre
partitioning
for
of
that
torrent
of
events
for
an
initial
set
of
process
stage
of
processing,
and
then
you
route
those
events
further,
then
all
of
those
events
would
have
the
same
partition
key,
which
means
they
would
all
land
in
the
same
partition
in
the
downstream
Kafka.
And
that's
not
what
you
want.
You
want
to
have
some
some
reasonably
even
distributions,
which
means
you
have
to
go
and
figure
out
a
different
partition
key
there
different
pipe
criterion
to
go
and
distribute
them
fairly
across
those
downstream
partitions.
G
I'm,
just
thinking
no,
because
quite
some
time
ago
we
discussed
the
correlation
of
events
right
so
I
think
this
is
kind
of
like
there's
some
overlap
there.
This
is
specific
to
the
partition.
I'm.
Just
thinking
you
know,
for
the
correlation
of
you
know
different
events
for
different
use
cases.
It
will
be
other,
you
know,
other
keys
or
other
identifiers,
so
just
three,
why
we
find
this
one
and
not
the
others,
so.
B
They
don't
give
you
they
give
you
the
general
direction
of
where
you
want
to
throw
an
event
like
the
broad,
but
these
partitions
typically
take
events
from
very,
very
many
different
contexts
so
that
you
can
go
and
process
them
in
parallel.
So
if
you,
the
the
use
case
for
this
is
like
Kafka
or
event
hubs
where
you
get
a
torrent
of
offensives,
that's
coming
in
one
end,
and
then
you
have
a
hundred
different
via
a
hundred
v
ends
which
in
parallel
pulled
the
data
out
each
for
its
partition
and
go
and
work
through
it.
B
What
matters
is
that
each
stream
is
made
in
a
way
that
it
lands
in
that
partition,
so
that
that,
in
the
end,
everything
that
is
coming
from
that
particular
whatever
origin
or
context-
that's
labeled
with
a
partition
key
shows
up
in
that
in
the
the
time
of
arrival
order
in
that
stream
and
gets
pulled
out
in
time
of
arrival
order
in
that
stream
from
that
partition.
But
that's
about
the
only
separation
if
you
get.
G
B
B
Then
you
probably
have
the
correlation
that
you
want,
so
you
can
use
this
key
to
create
that
correlation
and
and
and
I,
don't
think
the
the
the
language
that
we
have
here
speaks
against
that
it's
defining
a
casual
relationship
grouping
between
multiple
events,
and
so
you
can
certainly
for
the
use
case
that
you
described
what
half
a
year
ago
or
a
little
bit
longer.
You
can
certainly
use
that
field.
G
So
I'm
just
thinking
and
that,
if
we
want
to,
if
we
would
like
to
make
this
very
generic,
probably
we
should
call
it
correlation
key,
but
we
would
like
this
to
be
just
you
know
a
partition
to
stream.
Then
we
can
quote
petition
key
and
then
we
might
need
to
define
other
keys
for
other
use
cases
and
also
the
description.
I
think
it
would
be
better
to
add.
You
know
clarification
on
this.
G
Maybe
give
some
examples
how
we
are
going
to
use
this
otherwise
I
feel
it's
not
very
clear.
I
know
this
thing
in
the
extension,
but
still
for
people
to
correctly
use
it.
It
would
be
good,
you
know,
to
add
more
clarification
or
we
we
need
to
decide
whether
we
would
like
to
make
this
very
generic
or
specific
to
a
category
of
use
cases.
G
E
E
So
what
I
wanted
to
ask
was
I'm,
not
gonna.
Let
you
answer
again.
That's
her
is
this:
we
were
talking
about
semantics
versus
I,
don't
know
functional
are
keys,
so
the
use
case
for
this
was
actually
spinning.
The
stream
are
using
this,
as
the
key
in,
for
example,
Kafka
right.
So
this
is
not
the
semantic
key
or
subject
because
I
think
the
subject
field
is
the
semantic
version
of
this
which
we
were
talking
about
last
week,
yeah.
B
E
B
So
the
the
reason
why
I
was
arguing
against
making
this
core
part
of
the
spec
and
initially
also
argued
for
this
being
element
of
the
transport
but
I'm.
Okay,
with
this
being
an
extension,
is
that
ultimately,
the
consumption
strategy
for
these
for
the
events
is
dictating
what
the
partitioning
model
needs
to
be.
So,
ultimately,
the
partitioning
key
is
something
that
the
the
you
choose
space
on
the
needs
of
the
consumer.
B
It's
not
necessarily
something
that
the
publisher,
when
they
initially
create
the
event,
will
know
about,
but
there's
a
further
step
and
that's
I
believe
that
further
step
is
really,
when
you
hand
off
the
events,
the
produced
event
to
the
transport
infrastructure
that
you
then
go
and
figure
out
how
you
construct
from
the
given
metadata.
That's
already
there
how
you
construct
the
partition
key
so
to
satisfy
the
needs
of
the
the
partitioning
model
that
you
chose
for
consumption
and
then
once
you
and
then
for
these
multistage
processing
models.
B
When
you
are
effectively
pre
partitioning
a
stream
using
Kafka
into
let's
say,
16
partitions,
and
then
you
pull
on
one
of
those
partitions.
But
you
further
want
to
go
and
subdivide
that
stream
for
processing.
Then
you
will
probably
go
and
pick
up
those
events
and
then
you
will
choose
and
they
will
all
effectively
have
the
same
sets
of
keys.
At
least
they
law.
They
will
hash
the
same
way,
so
they
will
they
would
that
would
naturally
lend
in
the
same
partition.
B
If
you
happen
to
have
a
situation
where
you
have
like
16
partitions
here
and
then
downstream,
you
have
another
16
partitions
and
if
they
all
end
in
partition
1
they
are
again
are
going
to
land
in
partition
1,
which
means
you
no
need
to
go
and
change
your
strategy
in
terms
of
how
you
create
that
partition
key
so
that
you
get
even
distribution
downstream.
So
yeah
this
will
change
and
it's
it's
really
driven
by
the
consumption
model
rather
than
the
publishing
model.
E
Yeah
exactly,
but
that's
thanks
for
the
clarification
just
because
I
think
the
subject
being
the
semantics
key
and
this
being
the
functional
partitioning
key
I
think
is
Rea
like
core
missing
pieces
from
the
spec
at
the
moment.
So
I
am
very
happy
about
this.
But
to
put
the
carries
point,
yeah
I
mean
correlation
is
a
really
big
by
to
take
so
I
wouldn't
make
this
generic
exactly
because
of
the
because
this
has
a
well-defined
use
as
being
a
partition
key,
and
you
wanted
to
change.
E
A
Don't
see
what
its
hand
up
so
let
me
just
speak
in
here:
Cummins
I
think
the
description
you
just
gave
about
how
you
know
the
star
nut
can
read
the
exact
words
you
use,
but
it
sounded
really
nice,
where
you're
talking
about
how
something
like
source
is
really
meant
from
for
the
producers
point
of
view.
In
terms
of
you
know,
this
is
telling
you
sort
of
like
where
it's
coming
from
kind
of
stuff
or
subject,
whereas
this
is
really
more
for
consumption
by
the
receiver
to
figure
out
what
they're
supposed
to
do
with
it.
A
I
can't
really
like
word
to
use,
but
what
do
you
think
for
you
to
rephrase
that
the
text
we
could
put
that
in
here,
because
I
think
that
distinction
in
terms
of
the
role
this
thing's
supposed
to
play
would
really
help
clarify
why
this
is
an
SLA
meant
to
be
a
generic
correlation
ID,
you
know
to
have
a
property.
Did
that
make
any
sense.
Yeah.
B
A
That'd
be
great
and
and
I
guess,
I
think
where
the
vet
techs
actually
goes
into
here
or
the
primer,
because
typically
specifications
are
relatively
concise,
so
there's
maybe
better
text
for
the
primer.
That
way,
you
know
we
can
ramble
a
little
to
make
it
perfectly
clear
and
just
think
about
don't
know
where
go.
E
E
A
F
B
A
A
D
A
Problem
thanks,
Evan,
alright,
moving
forward.
All
right
thing
do
I
believe
the
gentleman's
name
all
right.
So
given
our
previous
agreements
for
proprietary
spec
references,
this
PR
adds
a
reference
to
the
rocket
MQ
transport
binding.
So
it
just
replaces
our
placeholder,
with
the
reference
and
I
did
check
that
URL
is
good.
So
this
one's
relatively
easy
and
it's
the
first
one.
We
have
any
comments
or
questions
on
this.
A
Any
objection
to
adopting
this
PR
or
merging
it
cool.
Thank
you
guys
very
much.
All
right,
Clemens
I,
don't
debating
changes
of
this
one.I,
so
I
think
this
one
is
ready
to
go,
but
I
did
have
a
question
for
you.
Where
is
it
not?
Here,
Evan
made
a
comment
talking
about
a
constraint
section
and
I
want
to
make
sure
that
this
stuff
down
here
in
your
mind,
covers
Evans
comment.
A
A
B
It
is,
it
is
proposed
because
of
a
leak
on
way,
so
so
he's
been
Eastman
trying
to
implement
an
intermediary
that
dealt
with
cloud
events
and
he
found
and
that
think
that
included
some
transcoding
of
the
content
and
he
found
that
look
at
just
a
data
and
the
content
type
or
data
content
type.
He
could
not
figure
out
how
a
non
string
causa
non
string
content
was
encoded
because
he
could
not
tell
between
whether
a
string
that
was
basics
before
really
contains
no
binary
or
whether
he
should
interpret
that
as
string
like
he
do
now.
B
It
wasn't
clear
for
him
to
go
and
figure
this
out
and
there's
there's
precedent
for
this.
So
HTTP
solves
this
using
and
in
mime
also
solves
this
using
the
concept
encoding
field.
So
this
is
now
called
data
content
encoding
because
it's
only
related
to
the
data
field
and
if
present,
that
field
as
specified
holds
really
holds
the
value
basics.
H
H
H
Can
I
don't
know,
hold
this
up.
H
I
B
I
decode
that
so
in
the
in
our
structured
encoding
mode,
where
we
have
everything
in
a
JSON,
then
data
is
there's
no
binary
option
for
for
data
like
there's
no
way
to
represent
anything
but
a
string.
Just
because
that's
that's
how
Jason
represents
things
so
then
you
are
effectively
forced
it
forced
to
put
something
into
that
string.
That
carries
the
binary
and
that's
usually
basics
before
all
right,
but
but
it
might
also
be.
It
might
also
be
something
completely
different,
so
you
might
use
percent
encoding
or
anything
else.
B
I
A
A
H
So
the
the
real
actually
I'm
for
me
is
to
read
the
like
prayer,
because
my
question
about
subject
is
how
it's
different
from
the
things
we
talked
about
before
specifically
topic:
I!
Think
that's
what
we
call
it
before
and
I
failed
to
read
to
go
back
and
read
all
of
our
prior
on
that
one.
Like
the
conversation
we
had
then,
but
that
was
that
was
the
like
concern
I
raised
because,
like
we've
brought
up
topic
before
like
no,
we
don't
want
to
do
that.
So
in
subject
came
up.
I
was
like
well
what
has
changed?
H
Does
every
mmmm
so
I
read
through
this
I
read
through
this
PR
by
itself,
I
haven't
read
through
the
through
the
like
links
to
the
prior
art,
and
this
by
itself
seems
ok.
Does
like
I
I
feel
like
does
everyone
else
to
remember
the
conversation
that
we
had
about
topics
and
like
why
we
why
we
rejected
that
originally
I
do
go.
A
B
It
I
see
I
think
the
we,
the
major
the
major
disconnect
we
all
had
was.
That
topic
is
so.
It
was
about
topic,
not
so
much
about
subject
and
I
think
we
were
arguing
together
for
subject
and
topic
and
as
a
joint
thing
and
I
think
the
hang-up
on
topic
was
that
topic
is
a
as
a
concrete
concept
in
middleware
or
people
felt
bolting.
This
to
middleware
was
inappropriate
and
that
source
was
a
more
appropriate
way
to
express
where
the
event
came
from
now.
B
If
you
go
deep
down
in
Theory
land-
and
you
know
forget,
forget
about
the
existence
of
software
for
a
moment
and
topic
as
word
per
se
kind
of
gives
you
that
abstract
grouping
of
events
in
to
you
know
share
topic,
but
I
think
the
holdup
was
that
topic
is
used
in
in
since
it's
used
so
pervasively
in
middleware.
It
felt
that
it
was
inappropriate
to
kind
of
you
know,
create
that
strong
association.
H
B
Those
two
things
together
should
be
unique,
because
the
subject
really
just
points
into
a
inner
structure
of
the
source,
but
I
don't
think
that
is
something
that
contributes
to
the
uniqueness
of
the
event
per
se.
I
think
the
the
event,
the
events
identity
is
scoped
to
the
emitter
and
that's
the
source.
I'm.
A
A
A
A
G
B
B
Let's
say
temperatures
and
co2
levels
you
could
model
it
so
that
you
can
get
the
state
events
for
the
temperature
and
the
co2
readings,
and
you
could
go
in
and
model
that
on
the
subject,
which
means
you
subscribe
to
get
to.
You
subscribe
to
get
all
the
events
that
that
sensor
gives
you,
but
you
can
go
in
dispatch
on
the
or
filter
on
the
subject,
and
one
of
the
subscribers
may
only
be
interested
in
co2
and
the
other
subscriber
may
only
be
interested
in
in
temperature
readings.
Oh.
B
G
Okay,
so
it's
for
the
so
it's
for
convenience
of
you
know
to
to
decode
on
the
source
information
right,
because
the
source
of
they
are
trying
to
find
out
how
we
define
the
source,
the
source
itself.
If
I
remember
it,
let
me
take
a
look
at
this.
It
said
its
itself
is
it's
a
has
its
own
structure.
A
So
Kathy
I
don't
know
if
this
helps
not
but
I
tend
to
look
at
this,
as
the
subject
is
the
entity
on
which
the
event
occurred
or
the
derive,
but
it's
the
entity
which
the
event
is
about
right.
So
when
you
create
a
new
blob
in
a
blob
store,
it
is
the
blob
itself
that
way
you
can
easily
filter
to
know.
Oh
I
just
want
to
get
events
about
this
one
particular
image.
That's
in
the
in
the
data
store.
E
Sorry,
I
don't
understand
our
color
sensor,
because
I
don't
think
any
sensor
would
have
a
subject
because
they
are
so
simplistic
and
the
whole
information
would
be
source.
But,
for
example,
if
you
think
about
the
first
example,
is
really
good
they're
actually
like.
If
you
have
a
storage
storage
container
and
you
get
events
from
that
storage
container,
where
the
source
is
that
storage
container,
then
the
subject
wouldn't
be
a
part
of
this
source.
E
B
E
Doesn't
need
to
be
a
part
of
the
source
if
you
look
at
the
example,
the
source
is
this
storage
container
that
holds
files
and
the
subject
is
the
actual
file
name
that
is
being
modified,
so
it
might
be
something
inside
the
source.
This
is
the
point.
If
you
have
a
database,
you
have
a
table.
The
source
is,
for
example,
the
database
at
the
table.
Concatenated
the
subject
might
be
the
ID
of
the
role
that
was
modified
for
yourself.
A
A
A
A
Not
calm,
okay,
so
going
back
to
this
one.
Do
it
does
anybody
have
any
other
questions
or
concerns
on
this
PR
and
I
know?
Sometimes
it
may
not
seem
like
it,
but
it
is
perfectly
acceptable
to
raise
your
hand
and
say
I'd.
Just
like
more
time,
I
mean
eventually
we
will
say
times
up,
but
if
you
need
more
time,
then
don't
hesitate
to
speak
up.
A
You
clemens
for
that
one
as
well.
All
right.
We
have
nine
minutes,
yep
nine
minutes
left.
Unfortunately,
we
don't
think
we
have
a
whole
lot
of
time
to
dive
deep
into
something.
So
whatever
I
would
do
right
now
is
drop
people's
attention
to
a
couple
of
key
discussions
that
are
going
on.
In
particular,
this
one
about
resolving
the
quote
issue:
definitely
not
ready
for
approval,
because
there's
a
lot
of
discussion
going
on
over
there,
but
we
need
to
get
this
one
resolved
relatively
quickly.
A
A
So
they
can
know
when
something's
been
replayed
and
I
believe
right
now
the
spec
pushes
us
towards
source
and
ID
in
combination
being
that
unique
identifier,
some
people
have
advocated
that
we
need
to
add
type
to
the
list,
but
if
we
do
that,
obviously
we
need
agreement
from
the
group,
but
then
I
think
that
also
is
going
to
impact
the
definition
of
what
I
deem
it.
Because
right
now,
ID
is
meant
to
be
pretty
much
best
thing.
A
A
So
what
I
did
is
I
started
a
document
here
with
the
current
version
of
the
spec
and
I
believe
David
Allen's
PR
that
tries
to
link
ID
and
source
as
they
make
identifiers
into
that,
to
start
to
get
a
discussion
going
around
what
these
attributes
actually
mean
going
forward,
because
we
have
to
resolve
these
questions.
So
please
take
a
look
at
that
document.
There's
some
questions
in
there
to
get
a
better
feel
for
what
people
are
thinking
around
these
particular
fields,
and
so
please
comment
in
that
PR.
H
E
A
Drop
it
oh
okay,
I'll
double
check
that,
but
I
want
to
make
sure
I
didn't
put
something
in
there
to
talk
about
how
we
have
to
talk
about
the
relationship
between
them.
You
know
it's
not
used
for
uniqueness,
because
I,
because
that
this
document
isn't
just
about
uniqueness
checks,
it's
about
what
is
a
semantic
meaning
of
all
four
attributes
and
I'll
double
check
and
make
sure
that
nothing
in
the
document
suggests
that
anybody
wants
to
use
subject
for
uniqueness.
So
thank
you
now.
G
I
think
this
is
a
bad
date.
You
know
on
very
good
discussion,
I
think
wait.
We
really
need
to
sort
out
this.
You
know
what
is
the
relationship
was
a
uniqueness
so
that
people
are
we
bring
people
on
the
same
page.
Otherwise,
I
think
you
know
when
people
just
read
the
document
could
be
sometime,
people
might
be
confused,
yeah.
A
And
in
fact
Clements
I
don't
know
if
you
saw
it
I
bet
a
comment
on
yours,
I
think
it's
your
subject,
PR,
where
you
had
an
example
in
there
where
you
actually
start
talking
about
the
relationship
between
subject
and
source,
I,
think
and
I
thought
the
description.
There
was
really
really
useful
for
someone
to
understand
how
the
various
properties
work
together
in
in
conjunction
with
it's
I
thought.
I
might
be
useful
to
take
that
example
and
pull
it
out
from
under
the
subject
attribute
and
make
it
a
more
higher
level.
A
B
A
E
On
just
a
spree
line
to
the
uniqueness
it
occurred
in
the
current
wording
of
the
spec.
So
I
have
this
problem
where,
if
you,
if
you
create
two
events
for
the
same
occurrence,
with
two
different
types,
that
say
there
are
different
versions,
version,
1
and
version
2,
do
they
have
the
same
idea
or
not
in
the
current
spec
I,
don't
mean
like
after
that
discussion.
It
might
change,
of
course,
so.
A
E
A
A
A
A
Maybe
that's
done
again.
Okay,
all
right
in
that
case,
I
believe
we're
done
for
today.
I.
Thank
you
guys
very
much
and
if
you
want
to
hang
on
or
er
you're
interested,
either
the
demo
or
cloud
event
or
Kuk
on
EU
discussion,
that's
gonna
happen
on
this
zoom
channel
right
after
this
call
in
three
minutes.
Otherwise,
you're
free
to
go
hi
guys.
E
A
Stuff
yeah,
what
I
liked
about
the
subject,
one
in
particular,
is
as
Rachel
mentioned.
You
know
we
have
this
whole
big
discussion
before
and
I
think
we
were
getting
ourselves
sort
of
mixed
up,
because
you're
trying
to
incorporate
too
many
things
all
at
once
and
so
I
like
the
fact
that
we
sort
of
landed
on
a
single
string,
meaning
source
and
then
slowly
or
maybe
pulling
it
apart
as
we
come
up
with
concrete
use
cases
and
clear
justification
for
why
has
to
be
pulled
out
and
I.
A
E
That
I
forgot
to
bring
up
a
point.
I
was
meeting
up
to
Nicole,
but
I
said
now
so
the
topic
there's
the
subject.
The
old
versus
new
I
think
the
topic.
One
is
that's
the
part
where
I
was
saying
that
we
don't
do
routing
information,
but
we
do
do
routing
information
because
someone
was
suggesting
to
have
to
respect
this
sentence
that
we
don't
do
routing
information
in
the
context
attributes.
But
we
actually
do
say
like
specifically
that
we
do
routing
information.
A
E
Ahead,
yeah,
yeah
and
I
think
the
difference
there
becomes
from
exactly
the
point
that
we
don't
do
topic.
We
don't
do
URL.
We
don't
do
like
the
concrete
place
where
you
put
the
event.
You
do
contextual
routing
information
and
they
are
both
routing
information
but
in
a
different
sense
or
different
scopes.
A
I
don't
know
yeah,
that's
actually
one
of
the
reasons
why
I
try
to
pick
up
this
PR
from
Thomas
and
see
if
I
can
work
it
and
that's
why.
The
reason
that
it's
kind
of
on
hold
right
now
is
because
I
feel
like
as
you're
saying
we
actually
do
have
to
talk
about
routing
information,
but
it's
different
type
of
routing
information.
It's
not
necessarily
transport
level
routing,
if
that's
the
right
phrase
for
it,
but
I
was
well.
A
But
rather
he
wants
to
be
able
to
pass
that
information
along
to
the
applications
in
a
consistent
format
or
in
a
consistent
location
and
since
cloud
event
will
be
the
thing
that
he's
passing
on
to
his
applications.
He
was
thinking
well,
why
not
include
it
in
there
right?
Why
include
some
other
extra
field?
Just
so
I
can
pass
along
the
HTTP
headers.
As
an
example,
bunny
isn't
that
the.
E
Wrong
way
around,
you
would
give
the
URL
when
you
are
putting
the
event
somewhere,
not
when
you
are
taking
the
event
from
somewhere
or
maybe
I'm
thinking
about
this
wrong.
But
when
you
need
that
information
of
where
to
put
the
event
you're
putting
it
somewhere
or
you're,
fetching
it
from
somewhere.
So
you
must
already
have
that
information.
So
there's
you
don't
need
to
give
it
to
an
application.
That
already
has
it
I.
A
Think
that
I
kind
of
say
nothing
correctly,
I
think
he's
looking
at
the
use
case
where
you're
a
function
author
and
your
function,
signature,
let's
say
it
takes
just
a
cloud
event:
that's
the
only
thing
it
takes
as
input
right.
The
function
may
want
to
know.
For
example,
what
was
the
URL
the
client
used
to
talk
to
me?
Oh
okay,
you
know
what
I
get
you
yeah,
and
so
he
was
thinking
well,
if
I'm
already
passing
in
this
kind
of
a
thing,
and
it
allows
extensions.
A
E
A
A
He
did
like
I
said
he
could
take
him
to
do
anything
he
wants.
He
just
didn't
want
to
violate
the
spirit
of
the
spec
and
that
yeah
yes,
yeah.
Okay,
all
right
so
to
the
subject
at
hand.
Could
Connie
you
wanna
mentally,
bring
up
the
dock?
I,
don't
think
anything's
happened
yet
in
there.
In
fact,
I
haven't
even
seen
anybody
open
up
issues
or
pour
requests
against
the
against
the
white
paper.
Yet
so
there's
really
nothing
going
on
there.
I
haven't
put
up
the
templates
for
the
presentations
yet
either
so
for
you
to
make
edits.
A
I
haven't
seen
any
discussions,
so
I,
don't
think
is
any
for
us
to
discuss
other
than
a
lot
of
us
have
action
items
to
do
something
in
some
space
here.
So
let
me
just
open
it
up
and
say:
does
anybody
have
anything
that
they
wouldn't
like
to
bring
up
for
a
discussion
point?
Otherwise,
this
part
of
the
meeting
will
end
and
we
could
switch
over
to
the
demo.
G
A
I
I
think
for
the
intro
and
deep
dive.
We
already
have
people
who
have
volunteered
to
present
for
the
service
summit.
I
know
you
and
I
were
talking
offline
and,
if
our,
if
the
proposal
that
this
group
put
forward
does
get
accepted,
then
yeah
I
think
you
and
I
can
talk
on.
That.
One
is
that
okay,
oh.
A
Yeah,
okay,
I
know
when
you
and
I
talked
before
we
were
talking
about
how
Scott
might
have
wanted
to
do
a
talk.
I
wanted
to
join
that
talk
for
the
serverless
summit,
but
I
talked
to
him
offline
and
because
he's
already
talking
at
camera,
which
one
of
these
he
started
talking
during
the
coop
concessions,
already
I
talked
to
him
and
he's
okay
with
with
you
talking
in
his
place
for
the
server
this
one.
Okay,.
G
A
A
A
Crud
image
really
weird
for
some
reason
want
to
open
up
your
mp4
file.
It
keeps
the
plane
to
open
up
in
OpenOffice,
as
opposed
to
QuickTime
and
I
could
fly
thick
enough
for
my
it's
very
weird
good.
A
A
My
intern
has
agreed
today
has
time
to
look
at
this
and
he's
already
started.
I'm,
hoping
that
maybe,
before
the
end
of
today,
I'll
be
able
to
share
with
you
guys
religion
of
what
he
has.
It
looks
pretty
good
so
far
and
then
once
you
have
that
in
place,
we
should
be
able
to
start
getting
people
to
hook
up
their
functions
to
it
and
then
so
Doug
I
can
see
your
your
share
now.
So
all
that
you're
going
to
talk
to
whatever
you
have
here.
D
Right
so,
like
I
said
in
the
call
earlier,
as
I
I
tried
to
put
in
a
single
document
all
the
key
discussion
points
that
we've
had
about
this
and
and
then
incorporate
relevant
tables
from
the
those
PowerPoint
decks.
So
I
started
out
with
just
a
narrative
that
was
really
intended
to
be
almost
attendee
facing
as
an
introduction
to
the
demo.
So
it
kind
of
can
go
through
that
and
see.
D
On
the
dashboard,
so
the
whole
display
that
you
know
the
proposed
display
where
it
had,
the
graphics
really
could
be
looked
at
as
a
as
a
grid.
You
know
we're
positioning
icons
and
images
and
texts
in
just
grid
position,
so
I
kind
of
put
together,
it's
just
a
simple
layout
for
where
the
static
images
would
be.
D
You
know,
based
upon
you
know,
well
the
the
graphic
that
was
in
the
deck.
So
you
know
it's
a
starting
point
and
then
what
the
from
there
included
a
link
to
the
to
the
mp4
file
in
the
video
that
shows.
You
know
how
all
this
animation
actually
or
how
the
micro-services
would
generate
events
that
would
be
Lissa
reacted
to
by
the
dashboard
app
to
turn
on
and
off
those
dynamic
images
and
text.
So
then,
this
section
here,
I'd
again
I
just
put
this.
This
is
a
table
of
you,
calm,
micro
services
or
processes.
D
D
D
So
this
the
rules
section
here
is,
is
looking
at
what
is
in
each
of
those
attributes
that
is
coming
in
it's
looking
at.
You
know,
what's
in
the
type
attribute,
what's
in
the
in
the
data
elements
or
attribute
which
the
data
in
these
cases
have
attribute
value
pairs,
one
or
more,
that
are
associated
with
the
subject,
which
is
an
which
is
an
object
of
type
with
the
value
of
what's
in
the
type
right.
D
D
D
You
know
in
either
the
role
of
a
supply,
retailers
carrier
supplier,
but
it's
your
your
roles
or
base
or
sorry.
The
way
your
going
to
take
action
is
based
on
your
offering
and
you're
offering
could
be
a
product
or
a
service
in
this
and,
if
you're,
a
a
retailer
that
is
providing
a
you,
know:
coffee,
drink,
you're,
going
to
have
a
product-
that's
like
you
know:
ID
P,
1,
P,
2,
P
3,
and
it's
going
to
you.
D
Isolate
the
the
events
that
you
would
take
action
on,
based
upon
a
location
that
your
service,
that
you're
serving
within
that
airport
exam
environment.
So
again,
this
is
how
we
can
scale
it
from
you
know.
If
there's
only
four
crowd
event
participants,
then
you
can
scale
it
down.
But
if
there's
up
to
I
think
this
one
includes
twelve
different
participants,
each
with
a
different
service
offering
that's
either
going
to
be.
D
You
know
the
retailer
product
or
it's
going
to
be
the
inventory
of
the
some
of
the
cups,
the
supplies
that
feed
into
that
that
finished
product
or
your
offering
as
a
service
transport
service
which
can
be,
you
know
transporting
as
Freight
the
the
replenishment
of
the
the
cups.
But
you
could
also
extend
it
into
being.
E
A
A
Just
just
let
you
know
as
of
right
now,
because
we
have
limited
real
estate
for
a
demo.
The
current
the
current
plan
that
I
have
the
intern
working
on
is
there.
As
you
said,
there
are
three
basic
roles
that
that
the
cloud
event
working
group
participants
could
could
fill
one
is
there
are
a
coffee
shop?
The
second
is,
they
are
a
supplier
of
coffee
cups
over
like
a
warehouse
and
then
the
third
role
is
the
transport
between
the
warehouse
and
the
coffee
shop
and
I
figured
between
those
three
different
roles.
D
D
A
D
It's
right
and
and
and
so
we
had
both
location.
We
both
talked
about
both
location
and
you
know
that
size
matrix.
But
then,
if
you
start
getting
into
the
carrier
and
what
carrier
is
going
to
be
selected
to
do
the
transport
and
what
supplier
is
a
it's
hard
to
carry,
you
could
certainly
do
you
could
do
that
to
a
certain
extent,
by
just
focusing
on
size
but
I
think
a
location
in
the
location
that
a
carrier
was
servicing
and
a
retailer
was
servicing
it.
D
A
And
that's
definitely
something
that
we're
taking
into
account
I.
Think
location
was
going
to
be
one
of
the
sort
of
the
filtering
criteria
that
type
of
stuff
and
I
think
size
that
the
cup
was
getting
another
filtering
criteria.
So,
depending
on
how
many
participants
we
have
in
there
we'll
play
with
those
variables.
But
yes,
it
yeah.
D
A
D
Okay,
so
then
so
yeah,
so
this
becomes
a
key
reference
table
for
those
micro-services
that
needs
to
be
transportable
and
distributed
for
last-minute
updates.
Same
thing
was
starting
inventories
right.
Each
of
those
retailers
is
going
to
have
initial
stock
to
fulfill
the
drink
orders,
and
then
you
have
suppliers
that
have
their
starting
stocks
and
then,
when
those
stocks
are
all
depleted,
then
you
could
start
generating
those
out
of
stock
notifications.
D
It
talks
about
each
of
those
somewhere
in
those
cloud
service
nodes.
You
have
to
be
able
to
generate
those
events
that
reflect
a
new
connection
coming
in.
So
that's
just
showing
you
know
what
what
how
to
populate
those
attributes
of
the
cloud
events
related
to
connections,
user
interface
for
the
attendees,
that's
going
to
be
able
to
able
to
generating
generate
originating
objects.
That
starts
the
orchestration,
so
there's
there's
the
orders
and
how
an
order
would
be
represented
as
a
cloud
event.
So
that's
in
here
and
that's
it
so.
E
A
We're
cutting
out
a
little
because
there's
some
static
but
I,
think
you're,
asking
yeah,
I
think
I
think
there's
a
lot
of
static
on
your
line.
However,
maybe
you're
moving
something
close
to
the
microphone
Jude.
Were
you
asking
about
whether
dashboard
will
be
dynamic
enough
to
allow
people
to
join
at
the
last
minute.
F
I
Like
you
know
just
broadcast,
like
you
know,
a
ping
saying:
hey,
I'm,
Mike
Microsoft
I
provide
a
latte
to
this
location,
and
this
is
my
class
of
beverage
right
and
then
the
the
a
pod
node
can
dynamically
generate
this.
It
will
basically
at
runtime
and
then
the
dashboard
app
can
just
look
this
up
for
a
real-time
view
of
what
kind
of
lottie's
are
available.
Do
we
have
any
chocolate
rings
or
you
know,
and
so
on
and
so
forth?
Yes,.
A
That
was
my
assumption
as
well.
Yes,
everything
everything
will
be
dynamic,
ed
to
the
point
where,
if
we
don't
get
enough
people
registering
for
the
proper
roles,
the
demo
won't
work
properly
right,
because
no
one
will,
for
example,
refill
the
us.
You
know
refill
a
coffee
shop
with
the
appropriate
cup
source
or
whatever
right,
no
one.
We
have
no
one
registers
to
be
a
supplier,
so
you.
D
I
A
And
that's
why
there's
there's
part
of
me
that
thinks
at
some
point
we
may
need
to
when
somebody
actually
register
to
the
system.
We
may
need
to
have
to
tell
them
what
they
can
actually
do
in
terms
of
okay,
they're
registering
as
a
supplier,
but
we
may
need
them
to
support
small
cups
instead
of
large
cups,
because
if
everybody
registers
are
small,
but
we
have
coffee
shops
that
are
doing
small
and
large,
that's
a
problem
right.
So
with
their
anything,
they
need
to
be
a
little
bit
of
a
negotiation
during
the
registration
process
too.
A
I
I
D
A
A
I
So
this
is
from
the
view
of
a
passenger,
so
if
the
retailer
has
enough
stock,
the
passengers
sojust
orders
using
the
dashboard
are
saying
to
conclude
our
coffee,
XYZ
and
the
retailer
tour,
the
coffee
to
update
its
own
inventory
and
then
delivers
the
coffee
right.
That's
that's
I
think
it's
a
direct,
simple
use
case
right.
I
Okay,
so
then
we
come
from
the
case
to
where
the
retailer
has
no
stock,
because
coffee
has
returned,
like
you
know,
less
than
minimum
stock,
say
either
10
or
0
for
that
matter.
So
the
retailer
will
then
tell
it
to
tell
the
supplier
that
dude
I
do
not
have
coffee
and
the
supply
of
will.
Then
in
turn
tell
the
carrier
dude.
Here's
your
coffee,
X
and
deliver
to
the
retailer
Y,
which
was
your
and
then
an
event
from
the
carrier,
will
go
directly
to
the
retailer,
which
says
your
story,
although
some
Cali.
A
Yeah
I
think
in
my
mind,
that's
actually
the
the
way
I
think
it
is
supposed
to
work
based
on
our
previous
questions.
The
only
thing
I
would
add
to
it
and
does
actually
change
anything
other
than
it's,
not
that
the
carrier
sends
an
event
dude.
This
is
you
know.
Here's
your
coffee,
rather
an
event
is
sent
when
the
carrier's
truck
arrives
at
the
retailer,
but
that's
a
minor
twist
on
it.
I
I
A
So
later
this
evening,
I
was
going
to
generate
a
dummy,
little
function
that
would
generate
all
the
adventure
he
needs
to
write
once
you
feel
like
that's
working
well
enough,
so
that
it's
not
embarrassing
and
it
shows
a
general
flow,
then
we'll
open
it
up
and
you
guys
can
all
start
generating
events
yourselves
from
your
systems
and
we
can
start
integrating
it
and
see
how
well
things
work
or
don't
work.
Nice.
A
A
I
I
A
All
right
cool
in
that
case
will
sync
back
up
again
on
Monday
or
through
the
cloud
event
demo
slack
general.
So
please
keep
an
eye
out
for
that
because,
as
things
become
available,
I'll
probably
post
information
into
that
slack,
Channel
all
right:
okay,
cool
thanks,
guys,
we'll
talk
again
on
Monday.