►
From YouTube: CNCF Serverless WG Meeting - 2019-03-28
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
A
A
A
B
B
A
C
D
B
A
D
B
A
A
A
B
Me
behind
stepped
away
for
a
sec,
okay.
Well,
let's
go
back
to
attribute
so,
regardless
of
the
correctness
of
the
quote
site,
I
would
like
an
official
from
down
from
on
high
ruling,
because
I
can
change
the
SDK
to
be
compliant
and
that's
no
problem
or
it
can
leave
it
the
way.
It's
currently
doing
and
it'll
only
talk
with
my
SDK
and
people
that
didn't
read
the
spec
as
well
as
the
c-sharp.
An
SDK
did
so
so
Scott.
How.
D
D
A
We
know
how
to
serialize
the
types
that
we
know
about
right
strings.
Don't
have
cording
stuff
like
that.
You
said
there
are
some
edge
cases
that
caused
concerns,
in
particular
that
someone
could
just
put
a
number
there
explain
to
me
why
that's
a
problem,
because,
even
if
they
put
just
a
number,
if
we
define
it
a
string,
it
gets
rid
of
the
string.
D
All
these
engines,
anything
anything,
that's
Dana
had
just
be
added
as
an
extension
was
defined
well
known
or
unknown
to
the
the
to
the
envelope
anything.
That's
there.
We
basically
rely
on
type
inference
or
we
rely
on.
You
know,
type
information.
In
being
sorry
for
eight
computers,
we
rely
on
the
imputation
type
system
because
that's
their
right
and
there
you
say
this
is
a
string,
and
this
is
a
map
and
there's
an
ins
and
blah
blah
so
that
has
that
and
for
for
HTTP
since
HTTP
the
header
is
a
string.
D
You'll
have
to
rely
on
you,
you
know
a
string
being
coded
because
otherwise
you
can't
identify
that
as
spring
or
at
least
you
can't
tell
them
apart
from
a
string
that
contains
only
numbers
and
you
you
probably
don't
you
don't
necessarily
know
what
to
do
with
with
curly
braces
or
with
with
square
brackets.
So
you
have
to
rely
on
yes,
I
think
Neverland
type
inference.
F
G
A
D
A
No
no
well,
but
then
book
that
gets
into
the
second
case,
though,
which
is
even
if
it's
defined,
someone
may
not
know
about
the
specification
and
therefore
it's
unknown,
and
what
does
someone
do
with
them?
I
receive
it
right
now.
In
those
particular
cases,
though,
I
would
say
that
they've
almost
falls
into
the
category
of
is
an
implementation
detail,
meaning
the
receiver
is
going
to
decide
how
to
take
this
particular
string
in
essence
and
decide
how
they're
going
to
store
it.
A
B
Because
if
two,
if
two
receivers
that
are
trying
to
consume
binary,
HTTP
encoding
attempt
to
consume
the
same
stream,
they
they
need
to
use
the
same
rules,
and
so
they
do
need
type
inference.
I
think
now
that
I
think
about
a
little
more
I
think
the
rule
that
could
be
future
proof
is
that
for
every
for
everything,
that's
in
the
core
envelope
they
don't
need
quotes
because
they
have
strict
types
defined
by
the
specification
for
every
extension
and
unknown
or
unknown.
They
use
the
quoted
JSON
mentality,
yeah.
D
But
now
you've
now
you've
bolted
down
the
extent
of
the
course
Bank
for
eternity.
Well,
that's
I've
already
started
altered,
yeah,
so
everything
so
so,
if
you,
if
used
that
rule,
then
an
extension
property,
let's
say
the
thing
we've
discussed
with
the
the
partition
key
right.
D
If
that,
if
that's
something
that
becomes
so
important
that
we
think
it's
part
needs
to
become
part
of
the
quorum
flow,
then
it
will
be
quoted
as
the
extension.
But
then,
as
soon
as
this
marches
into
the
core
envelope,
it
will
be
no
longer
quoted,
which
means
that
all
code
that
is
using
it
will
now
have
to
be
effectively
rewritten
adapted
for
it
to
be
not
encoded.
With
calls.
A
A
H
A
So
yeah,
but
my
point
is
anybody
who's
actually
going
to
use
that
attribute
has
to
know
two
different
things.
One
is
how
the
attribute
is
defined
right.
It
may
look
like
a
number,
but
it's
actually
a
string
that
kind
of
stuff
and
that's
part
of
the
specification
of
that
attribute,
and
they
just
have
to
know
that
by
understanding
it.
Otherwise
they
can't
really
use
that
data
for
anything
meaningful
anyway
and
two.
A
D
I
can't
ignore
that
and
simply
say:
well,
you
just
deal
with
it
as
a
string
on
our
most
okay
with
them.
Sorry
for
the
core
properties,
but
as
soon
as
it
gets
to
extensions,
then
the
middleware
will
generally
the
middleware
will
not
know
about
extensions,
except
that
the
ones
that
it
cares
about
angling
and
then,
and
so
for
that
we
definitely
the
inference
and
then
for
me
the
question
that
for
me,
then
the
compatibility
question
is:
is
one
of
when
I
go
and
promote
something
that
is
currently
an
extension
into
the
core
spec.
D
D
But
then
you're
effectively
Fernet
from
the
function,
but
in
ATP
in
ATP,
so
some
so
my
perspective
on
this
is
affecting
the
MVP
mapping
of
JMS
in
MVP.
The
underlying
type
system
of
of
ATP
allows
for
four
properties
to
be
tight.
They
can't
can't
be
complex
types,
but
they
can
be
any
of
the
Valentina
P
types.
Does
what
that
just
okay.
D
You
can
give
me
numbers
and
you
can
even
have
dates
and
you
have
constraints,
but
that's
that's
permitted.
The
what's
constraining
QP
is
that
the
property
names
need
to
be
symbols,
but
the
rest
of
the
value
space
is
unconstrained.
You
can
have
lists,
you
can
have
maps,
but
they
can
be
any
any
of
the
other
side
so
that
that
is
that
is
type
safe.
It's.
A
G
A
That's
I'm
kind
of
wondering:
can
we
Punk
on
this,
because
HTTP
doesn't
have
this
problem
right
today,
everything's,
basically
a
string
to
them
without
actually
saying
it's
a
string.
It's
just
a
bunch
of
characters,
I'm
wondering
whether
at
the
sea
at
the
cloud
of
that
level,
we
actually
need
to
worry
about
this,
or
we
just
say:
nope,
everything's
a
string
and
if
you
all
understand
the
types
and
then
fine
you
can
convert
it.
But
that's
outside
the
scope
of
this
back.
If
that's
outside
of
caught
events,
processing,
yeah.
D
D
The
goal
was
basically
with
the
type
system
to
distinguish
between
strings
and
then,
which
was
most
of
most
people
we
had
and
times
and
timestamp,
as
a
special
case
of
that,
because
that
is
a
spring
expression.
It's
just
a
specially
formatted
string
expression.
So
the
time
stamp
is
in
fact
you
drive
from
spring,
and
then
we
only
we
effectively
had
the
the
binary
as
a
special
case,
because
that
might
be
expressed
in
string.
A
D
D
The
fact
that
the
reason
why
we
have
why
we
have
binary
is
really
to
just
accommodate
binary
in
data
in
the
any
type
like
binary
is
not
never
used
directly,
but
just
what
attributes
per
se,
which
we're
now
on
debating
we're
not
talking
about
data,
we're
just
talking
about
the
attributes
for
those
spring
and
spring,
whereas
x
would
be
sufficient,
and
so
we
could
go
and
plug
pump,
basically
punt
on
that
problem.
If
we
get
rid
of
the
it's.
D
A
B
D
D
Then,
and
we
have
I
think
we
have
extensions
or
do
we,
then
then
then
the
question
is:
can
we
can
really
represent
that
accurately?
Like?
Can
we
encode
that?
Because
the
map
is
effectively
like
we
need,
we
would
have
to
go
distinguish
between
the
strain
and
the
map?
If
we
say
okay
maps
are
legal
for
attributes,
then
that
gets
us
on
earth.
I
was
that
out
of
that
as
well?
Well,.
A
What
if
we
did
this?
What
if
we
said
if
you're
going
to
define
an
extension
that
is
of
a
different
type
and
map,
is
a
good
example?
Then
you
need
to
define
the
proper
serialization
for
that
particular
attribute
as
long
as
it
can
be
sterilized
as
a
string
in
the
end,
and
if
that
means
for
this
particular
extension,
you
want
to
be
sterilized
in
JSON
format.
Then
that's
the
rule
for
that
big
R
attribute.
B
D
D
And
if
we
say
we're
not
allowing
Maps
but
like
if
you
want,
if
you
need
five
fields
in
your
extension,
well
bring
five
fields,
don't
bring
a
map,
and
then
so
that's
a
thing
that
that
that's
the
thing
we
could
say
and
then
we
can
strike
the
integer
and
say
well
if
you
need
to
have
a
number
make
a
rule,
but
ultimately
the
only
way
you
can
go
in
and
code.
That
is
straight
and
then
then,
in
that
case,
then
it's
easy.
D
G
F
D
B
A
D
G
A
L
I
N
L
B
A
K
E
N
E
E
A
C
A
P
A
A
And
Justin
this
maple
we
lost
them,
I'll
be
back
so
Justin.
I
think
this
may
be
your
first
time
in.
If
so,
can
you
do
me
a
favor
and
add
your
company
affiliation
if
you
want
to
have
one
into
the
agenda
next
to
your
name.
Just
for
the
attendance
tracker,
I
put
a
link
to
the
document
that
I'm
editing
into
the
chat.
If
you
can
see
that
all
right,
let
me
do
one
like
look
and
then
we'll
get
started.
I
Cathy!
Are
you
there?
N
A
All
right,
let's
go,
get
started
three
after
the
hour,
a
eyes,
I,
don't
think.
There's
anything
here,
I
should
mention.
We
had
a
discussion
during
the
SDK
call,
which
is
right
before
this
one
about
the
hetero
value
issue
that
I
believe
Adam
opened
up
a
little
while
ago.
I
think
we
had
a
really
good
discussion
and
Scott
took
the
actually
had
to
write
up
a
proposal
so
expect
to
see
something
there
soon.
They
don't
need
to
go
in
that
now,
but
just
let
you
guys
know
we
did
I
think
made
some
good
progress
there.
A
A
All
right,
my
forward,
then
SDK
call
as
I
mentioned.
We
just
had
one
thirty
minutes
ago.
There
probably
is
anything
worth
mentioning
there
other
than
I
know.
In
the
past
we
talked
about
potentially
having
some
sort
of
SDK
interrupt
type
of
thing
go
on
at
KU,
Connie
you,
given
everybody's
workload
and
stuff.
We
don't
actually
see
that
it
happening
things
could
change
if
people
get
some
free
time,
but
as
of
right
now
we
are
not
planning
on
doing
it.
A
B
A
A
A
A
And
I
apologize
as
we
have
been
having
these
phone
calls.
Doug
has
been
doing
a
really
good
job
of
producing
documents,
to
sort
of
explain
the
eight,
the
current
thought
process
and
I
guess
we
have.
We
probably
should
start
sharing
that
with
a
broader
community,
because
I
know
this
Google
Doc
right
here
doesn't
have
the
very
latest
so
we'll
take
the
action
item
after
the
call
after
this
one
to
start
making
sure
that
the
documents
are
in
a
place
that
everybody
could
actually
see.
A
A
A
A
A
A
Nothing
relative
to
cook
on
China
so
I
believe
we're
into
the
peer
review,
and
just
do
one
quick
check
here
or
something:
okay,
no
new
votes
have
come
in
okay,
so
into
the
PR
review
stuff.
So
last
week
we
started
votes
on
the
mid-size
type
of
PRS
are
out
there
and,
if
that's
like
mistake
and
I
believe
this
is
the
current
status
of
the
votes,
is
there.
A
A
In
that
case,
I
think
it's
clearer,
I,
don't
know
the
numbers,
but
the
two
definitely
wins
over
one
for
that.
So
this
is
the
one
that
we're
going
to
go
with.
I'll
enter
in
the
vote
results
into
the
PRS
themselves,
but
is
it
before
we
move
on,
though
I
think
the
next
steps
here
it's
for
Clemens
to
potentially
do
some
wording
changes
on
this
one
I
think
some
people
add
some
comments
last
time.
But
aside
from
that,
are
there
any
other
discussion
points
people
want
to
bring
up
relate
to
this.
A
Q
P
I'm
really
sympathetic
with
that
and
I
honestly
sorry,
I'm
second
I
should
talk
this
much,
but
I
think
that
we
have
interoperability
constraints
in
either.
One
of
these,
and
so
like
just
speaking
for
myself,
I
would
have
rather
had
more
conversation
and
and
I'm
really
open.
If
someone
else
wants
to
keep
the
conversation
going
about
like
how
interoperable
work
with
either
of
these
constraint.
P
Okay,
I
guess
for
me,
I
was
thinking
that,
like
like
I
have
I
I
can
read
the
text
right,
I
can
read
the
text,
I
understand
what
it
means,
but
then
it
still
seems
like
there
are
some
unsalted
like
there
are
some
outstanding
questions
for
me
about
how
this
all
like
it
seems
like.
We
have
an
interoperability
problem
either
way,
because
it's
a
should
it's
a
must
like
we'll,
still
have
problems
and
like
if
middleware
modifies
things
like
how
do
we
like?
How
do
we
make
sure
that
an
event
can
get
all
the
way
through?
P
P
A
Well,
that's
what
I'm,
trying
to
figure
out
is
I
I'm,
not
sure
this
phone
calls
the
best
place.
I
have
that
conversation,
because
I
feel,
like
we've
kind
of
gone
around
in
service
for
a
while,
but
if
people
do
feel
like
we
may
have
made
the
incorrect
decision,
then
maybe
we
should
have
a
separate
discussion.
I
I.
P
Of
these
would
have
been
like
okay,
and
this
one
seems
easier
to
implement
to
me,
but
there
are
a
lot
of
outstanding
questions
and
like
there
and
we
had,
we
were
starting
to
ask
those
questions
at
the
end
of
the
call
again
chat.
There
was
like
a
very
lively
chat
going
on
and
all
of
those
questions
are
still
outstanding
in
my
mind.
So.
P
A
F
The
problem
for
me
here
is
that,
as
far
as
I
am
concerned,
last
time
it
was
being
Christophe
in
chat
and
how
that
conversation
ended.
As
he
said,
these
these
proposals
end
up
with
different
constraints
and
I.
Don't
think
there
were
any
questions
open
between
me
and
Christophe
in
that
matter,
II
was
just.
We
are
looking
for
the
four
different
things:
okay,.
P
I
might
have
different
questions.
How
about
as
a
action
I
am
I
tried
how
about
I
write
up
either
like
a
dog
or
I?
Think
it's
I
think
it
docks
a
bad
idea,
then
like
something
in
github,
because
it's
like,
we
probably
don't-
want
the
artifact
of
this
conversation
to
like
live
in
the
repo
and
try
to
describe
the
problems
and
like
a
potential
solution.
P
A
So
so,
normally
I
think
once
we
have
a
vote
on
something.
This
is
behind
us,
however,
because
I'm
hearing
that
there
may
be
some
people
on
the
call
who
feel
like.
Maybe
we
ended
the
discussions
prematurely,
let's,
let's
sort
of
bend
the
rules
a
little
here
and
say:
okay,
Rachel
go
off
and
create
that
dock
and
let's
see
where
that
discussion
leads,
and
we
won't
nicely
rush
through
merging
any
player
at
this
point
in
time.
Sorry.
P
A
Yeah
but
I
also,
suddenly,
let
me
put
it
this
way
and
the
reason
I'm,
okay
with
not
merging
it-
and
this
is
just
my
opinion.
You
know
you
guys
can
say
I'm
wrong.
That's
fine!
The
reason
I'm
not
kind
not
to
merge
it
yet
is
because
I
don't
want
people
in
Si
coded
up.
If
we
decide
to
go
different
direction
and
I
don't
get
the
sense
that
while
it
is
a
very
important
issue,
I
don't
think
it's
necessary
for
somebody
to
implement
the
spec
to
test
out
the
rest
of
it.
A
P
Q
P
A
I
did
say
that
earlier
I
think
Clemens
had
some
work
to
do
to
tweak
the
PR
a
little,
and
now
people
can
go
review
it
for
more
stringent
wordsmithing
kind
of
stuff.
But
my
point
was:
let's
say:
that's
all
done
by
next
week:
I
don't
want
to
necessarily
try
to
force
a
merge
in
next
week
of
a
rewrite
of
that
PR.
If
Rachel's
document
has
had
some
good
discussions
in
there,
that
was
my
only
point.
Look
I
did
and
I
did
I
didn't
want
people
to
have
an
expectation
says:
oh
Doug,
we
took
a
boat.
A
P
A
B
Behind
me,
this
is
just
prettier
dot.
Io
is
a
linter
tool,
it
does
a
very
opinionated
version
of
markdown
formatting
and
they
made
no
content
changes
except
for
in
the
I
think
that
contributor
dot,
MD
file
was
implemented
in
a
way
that
was
not
marked
down
compatible,
so
I
had
to
reformat
it
a
little
bit.
What's.
O
A
B
A
Hey
mark
had
to
come
in
there
about
they
didn't
modifying
the
make
file,
but
then
you
said
something
other
than
response.
B
E
P
B
B
So
I
have
a
workflow,
that's
like
local
to
me,
but
I
use
prettier,
and
it
does
this
command
that
you
see
here.
This
is
the
one
that
just
runs
everywhere,
but
I
used
a
version
of
this.
That
only
looks
at
your
difficult
for
your
current
come
commit
and
it
runs
lint
on
that.
So
I
always
check
in
stuff.
That's
Linton
vehicle.
P
A
A
A
This
one
is
mine,
just
added
some
texture,
primer,
I,
think
yeah.
Okay,
this
the
booklet
right
here
basically
just
wanted
to
add
some
texture
primer,
explaining
some
of
the
rationale
or
ideas
behind
the
idea
itself
for
the
ID
attribute
itself,
basically
pointing
out
the
biggest
thing
here
is
that
the
ID
is
meant
to
be
unique
per
cloud
event
within
the
scope
of
a
producer,
so
that,
for
example,
if
a
single
occurrence
generated
two
separate
cloud
events,
those
two
cloud
events
get
separate
IDs.
They
do
not
share
the
same
ID.
A
If
you
want
to
do
some
sort
of
correlation
and
stuff
like
that,
look
at
some
other
have
to
be
to
do
it,
but
that's
not
what
ID
is
for
the
ID
is
strictly
meant
for
uniqueness
across
cloud
events
from
a
single
provider
and
as
a
result
of
that,
what
I
did
is
down
here.
I
replaced
database
commit
ID
with
just
a
UUID,
because
my
concern
was
that
someone
can
interpret
this
as
one
commit.
Id
me,
I
should
generate
more
than
one
cloud
events.
A
Therefore,
it's
okay
to
use
that
commit
ID
and
more
than
one
caught
event
that
I
didn't
want.
Be
able
to
to
make
that
mistake
and
I
don't
want
to
get
into
that
kind
of
prose
in
the
spec
itself.
I'd
rather
leave
that
for
up
here,
which
is
what
I
did
that's,
why
I
just
changed
it
to
ID
or
UID
I'll,
leave
that
there
for
a
second,
for
you
guys
who
made
out
of
it
chance
to
read
it.
A
What
it's
saying
is,
if
you
have
semantic
rules
that
are
trying
to
use
this
ID
for
some
sort
of
correlation
across
cloud
events,
that's
would
be
inappropriate,
but
it's
not
preventing
you
from
using
database
ID
if
it
fits
the
if
it
fits
the
rules
of
how
we
define
this
ID,
meaning
it's
unique
for
all
cloud
events
from
this
producer,
but
you
can
do
replays
with
this
at
its
nightly,
doesn't
block
that
because
it's
replaying
the
exact
same
cloud
event,
so
that
would
still
be
appropriate.
Does
that
make
sense
yeah.
B
A
B
A
B
A
But
I
think
I
think
that's
an
implementation
detail
in
the
sense
that
if,
if
whatever
mechanism
they
choose
to
use
to
generate
the
cloud
event,
ID
needs
to
satisfy
our
requirements
of
it
being
unique.
If
they
choose
to
use
a
field,
that's
not
going
to
be
unique
because
there's
another
guy
doing
the
exact
same
thing
from
the
you
know
and
and
get
it
spit
out
something
with
the
exact
same
source,
and
they
need
to
do
some,
some
coordination
to
figure
out.
What's
what's
going
on,
and
how
to
resolve
that
right.
H
H
A
Okay,
so
maybe
I
need
to
clarify
something
here.
My
purpose
behind
this
PR
was
not
to
change
the
semantics
of
ID.
It
was
to
clarify
what
is
currently
in
the
spec
and
I
believe.
What's
in
here
is
accurate
within
what
we
currently
have
ID
defined
as
in
the
spec
meaning
we
define
ID
to
be
unique
for
all
cloud
events
from
a
single
producer.
If
we
want
to
change
the
definition
of
ID
so
that
it's
not
unique
and
that
the
uniqueness
spans
multiple
attributes,
for
example,
type,
then
I
think
a
different
PR
I
should
address
that.
A
Yes
or
maybe
I
may
blow
up
the
scope,
yeah,
you're,
right
and
I
think
it
goes
back
to
what
Scott's
talking
about,
because
I'm
not
saying
I,
think
you're
wrong
in
your
concern,
but
I
think
what
you're
proposing
to
do
is
to
change
how
we've
currently
defined
ID
and
I'm,
not
saying
with
I,
agree
or
disagree
with
that,
but
I
think
it's
a
different
PR.
Why.
F
A
B
A
I
know
I'm
Pro
clear
that
I've
no
desire
to
resort
to
rush
this
through.
So
in
that
case,
then
let
me
go
back
here
for
a
sec.
It's
Allan's,
PR
believe
it's
this
one.
In
that
case,
that
I
think
we
may
need
to
do
is
to
go
back
and
look
at
Allen's
PR,
because
I
think
I,
don't
think
he
necessarily
took
it
upon
himself
to
change
the
definition
of
ID.
He
was
more
looking
in
terms
of
given
what's
currently
in
the
spec.
How
does
someone
establish
uniqueness
right
and
I
think
what
you're
suggesting
is?
B
A
A
Okay,
so
ID
talks
about
it
must
be
unique
within
the
scope
of
the
producer
right
and
that
and
well
we
don't
actually
have
mustard.
We
actually
somewhere.
I,
don't
say
this,
but
I
think
most
people
are
assuming
producer
equates
to
source,
which
is
why
Allan's
PR
talks
about
linking
these
two
for
uniqueness.
A
B
A
I
agree:
okay,
does
anybody
disagree
with
that,
because
my
next
Justin's
gonna
be
that
we
take
it
offline
and
set
up
a
separate
document
or
phone
call
or
something
to
have
some
additional
discussion?
So
we
don't
show
off
the
entire
hour
with
this?
Does
anybody
disagree
with
that
direction
or
we
miss
reading
the
tea
leaves
here.
L
A
Okay,
I
will
take
that
one
offline,
all
right
next
one
is
might
as
well.
B
Okay,
I'm
trying
to
member
wire
out
this.
You
are
leading
the
witness.
A
A
D
And
we
saw
that
we
ended
up.
We
ended
up
introducing
source
in
that
discussion.
I
hear
you're
going
way,
far
back
yeah.
Of
course,
I
do
because
I
also
leave
Eileen
I.
Think
I
linked
that
old,
the
old
issue
into
this.
So,
in
addition
to
what's
written
here,
there
are
some
supporting
material
that
I'm
pointed
to
that
you
might
want
to
read.
D
You,
then
want
to
know
exactly
which
file
was
created
and
you
want
to
be
able
to
generically
filter
on
this,
which
means
you
wanted
to
probably
put
a
something
filter
on
that
file
name,
which
is
Dutch
JPEG,
and
you
don't
want
to
do
this
by
cracking
the
payload.
But
you
really
want
to
have
this
in
metadata.
So
now
the
way
that
how
we
saw
that
problem
for
ourselves
and
Microsoft
is
where
we
have
it.
D
We
basically
have
to
know
that
we
can
break
apart,
that
your
I
at
the
pound
sign
and
then
kind
of
come
up
with
out
two
fields,
and
it
is
the
left
part
being
the
container
and
the
right
part
being
the
file
name.
But
that
seems
to
be
so
that
terms
now
out
to
be
for
for
several
people
who
have
been,
who
have
been
starting
to
implement
call
events
or
also
middleware
for
cloud
events,
and
that
duality
where
you
need
to
have
information
about.
Where
does
that
event
come
from
and
then
further
information
about
what
s?
D
What
aspect
of
that
source
has
been?
Has
that
event
been
raised
about
is
something
that
is
is
in
fact
interesting
to
people.
So
that's
why
I'm
affected
to
read
introducing
this,
even
though
we've
have
been
discussing
this
for
about
a
year
or
thing
and
then
ended
up
having
just
one
field
and
I
think
we're
now
having
starting
to
have
some
evidence.
That's
having
source
as
the
subscription
scope.
If
you
will
and
then
subject
as
the
additional
identifier
for
the
sub
objects,
that
the
event
is
really
about
makes
sense.
A
F
B
We've
had
so
in
Canada
we're
trying
to
work
on
well
on
the
eventing
side,
we're
coming
to
the
need
of
some
sort
of
registry
and
there,
if
you,
if
you're,
going
to
ask
all
of
the
consumers
to
send
to
a
central
place
and
then
ask
that
central
place.
Give
me
all
events
that
match
the
following
criteria,
and
you
probably
want
to
do
something
like
give
me
the
the
bucket
that
events
are
coming
from
and
give
me
all
the
bucket
events,
but
I
want
to
only
know
for
a
very
specific
bucket
at
the
moment.
B
We
have
no
way
to
do
that
because
we
don't
control.
Thus,
this
the
format
of
what
a
source
is,
and
we
don't
know
that
we
can
always
split
on
a
hashtag
or
a
pound
something
so
we
need
some
something.
That
means
the
the
parent
object
and
the
the
the
subject
which
this
fulfills
so
we've
actually
started
working
on
implementing
this
as
an
extension,
because
we
need
it.
F
Also
think
this
is
quite
quite
needed
for
every
single
example
I've
now
given
for
other
PRS
and
features
I've
had
to
somehow
work
around
this
issue
and
I.
Think
that
shows
that.
Well,
it's
an
issue
all.
A
I
was
gonna
get
to
that
in
a
minute,
because
I
think
Clements
missed
the
previous
conversation
I
was
gonna
say
was.
If
we
actually
liked
this
PR,
then
we
can't
merge
it
today.
We
need
to
pull
this
into
the
previous
conversation
of.
How
does
this
relate
to
ID
and,
as
you
said,
it's
got,
houses,
impact,
source
and
I
think
we
need
to
think
about
all
those
together.
First
before
we
go
about
merging
this
one.
So.
D
I,
don't
think
it
changes
source
because
the
source
is
I
can
imagine
that
what
sort
of
discussions
you
might
had
about
the
uniqueness
of
messages
or
our
formats
but
source
is
really
the
place.
The
place
that
you
were
you
for,
which
you
registered
interest
right
in
pops
up
sense
that
raise
those
events
and
then
subject
is
really
a
thing
that
is,
that
is
relative
to
that
source.
That's
that's!
D
How
I
look
at
it,
but
source
per
se
I,
don't
think
stars
as
a
definition,
how
we
look
at
it
it's
the
thing
that
raises
the
events,
it's
just
that
the
source
itself
has
some
internal
structure
and
you're
interested
in
in
what?
What
is
that
inside
of
that
source,
that
just
that
dis
events
ago
and
that's
something
that
the
source
itself
can
explain
really,
because
the
source
is
well
the
thing
that's
raising
the
events,
it's
the
outer
container.
If
you
will
can.
D
That
was,
there
were
some
brilliant
example
in
examples
yeah
actually
in
here
in
that
the
if
you
subscribe
to
pull
requests
so
you're
interested
in
pull
request.
Let's
just
look
at
the
highlighted
piece
right.
What
you
will
subscribe
on
is
on
the
cloud
events,
repositories,
pull
request,
source
right,
that's
what
that's
why
you
will
subscribe
button
and
and
when
you
get
events,
that's
how
you
will
discriminate
them
from
other
events,
because
you're
going
to
look
at
this
and
say
so:
you're
gonna
have
your
your
pull
request
notification
service.
D
The
first
half
is
I,
want
to
know,
I
want
to
identify
which
repository
they
are
associated
with.
That's
the
left
side
of
this.
That's
the
thing
you
subscribe
to.
That's
that
exactly
so
you
can
tell
from
this
which
bucket
they
belong
to,
but
the
last
part
of
that
that
one
two
three
is
not
what
you
subscribe
to,
but
that's
effectively
what
this
is.
D
D
And
the
wonderful
things
the
last
last
example
that
Scott
mentioned
is
the
email
with
the
subject
help
that's
exactly
what
that
is
right.
This
is
what
this
is.
My
messaging
system,
including
SMTP,
have
a
subject
because,
in
addition
to
the
address
whether
that
sub
comes
from,
you
still
need
to
have
some
some
sorting
criteria.
How
someone
would
figure
out
whether
it's
worth
reading
and
often
that's
the
subject,
should.
J
D
Well,
so
it's
an
SMTP
lands
technically
in
your
inboxes.
The
thing
that
you
look
at
because
that's
where
you
need
consumption,
the
subject
is
exactly
what
you
to
look
at
and
that's
also
how
your
threads
are
being
sorted
by
I.
D
J
Have
two
threads
with
the
same
subject:
they
get
threaded
separately
because
they
were
actually
sent
independently
and
you
know
helped
it's
a
great
example.
If
ten
people
send
help
over
the
next
few
weeks,
they
may
not
be
replying
to
each
other,
they
may
actually
be
asking
you
know
for
help
on
separate
topics.
I
would.
D
Argue
that's
a
tangent
of
correlation,
but
effectively.
This
is
this
is
about
how
you
tell
it's
a
subdivision
of
so
the
source
here
we
is,
the
circle
is
working
for
mailing
list
is
sending
you
events
because
you
subscribe
to
those,
and
now
you
want
to
be
able
to
tell
those
apart
and
the
way
you
tell
those
apart.
If
you
look
at
in
your
mail
in
your
mail,
application
is
using
the
subject
and
yes
there.
D
P
Have
a
question:
how
does
this
so
I
I
am
concerned
about
making
something
like
the
pub/sub
channel
or
the
Kafka
topic,
or
something
like
that
like
baking,
that
in
in
a
way
that,
like
seems
like
not
like
people,
will
be
using
not
Kafka
or
not?
Pub/Sub
and
I
want
to
make
sure
that
this
is
like
something
that
works
for
everybody
and,
and
it
reminds
me
a
lot
of
the
conversation
we
had
about
topic,
which
was
an
attribute
proposed,
I,
don't
know
like
year
and
a
half
ago
or
something
what
is
it
differently?
D
P
D
I
see
what
we're
doing
here
is
we're
we're
baking
in
a
disparate
dispatched
opera
when
you,
just
as
you
forget
to
forget
who's,
forget
whether
there's
middleware
in
play
right,
you
simply,
you
have
a
function
and
you
are
asking
10
parties
to
send
you
something
and
they
they
sent
you
something
about.
Let's
stick
with.
They
created
silence.
D
Example
right,
so
you
have
ten
parties
who
all
manage
files
and
they
tell
you
about
it
so
now
that
they
show
up
with
condiments
the
first
thing
you
will
look
at
to
figure
out
what
to
do
with
that
event.
Is
you
going
to
look
at
source
because
you
want
to
know
who
just
send
you
something
and
then
now
the
second
thing
is
you
want
to
know
what
is
that
file
that
was
great
and
whether
you
interested
in
that
file?
That
has
we
create
it
now?
D
So
you
know
you
need
to
be
able
to
apply
a
filter
on
some
metadata
of
whether
you
are
interested
in
that
file.
Is
that
following
is
that
is
about
a
JPEG?
You
might
be
interested
if
that's
about
something
else,
you
might
not
be
interested
in.
So
you
need
to
have
a
string
that
carries
that
information
and
that
is
distinct
from
the
information
about
who
sent
you
that
who
sent
you
that
data.
That's
the
decision
I'm
trying
to
draw
here.
D
P
D
P
How
about
I
take
an
action
items
to
leave
some
comments
on
this
PR
and
say
like
like
I,
like
my
main
concern
and
I
know:
I
sound
like
a
broken
record.
Is
that
I
want
this
to
not
just
work
for,
like
the
example
use
cases
that
we
are
thinking
of
the
work
for
like
more
general
use
cases
and
I
worry
that
we
are
baking
in.
You
must
be
using
pub/sub
or
something
like
that
for
it
like
in
this
one.
P
A
P
D
Yeah,
that's
not
what
yes,
okay,
I
I
understand
like
I
have
to
deepen
this
like
this
is
my
and
probably
20
years
too
far
into
this
whole
pops
up
stuff,
a
thing
that
I
don't
see
the
thing
that
you're
missing
and
I
would
like
to
see
the
thing
that
you're
missing
in
writing.
So
that
would
be
fantastic.
If
you
would
do
that.
Okay,.
P
D
And
and
one
if
you
do
me,
one
favor
as
you
as
you
think
about
this
there's
in
the
in
the
comments.
Basically
in
the
in
the
PR
itself,
I
delete
the
old
issue,
which
is
prior,
basically
ever
I,
created
an
issue
back
in
the
day,
which
was
the
subjects
but
I,
think
puppets
and
subjugated
subjects
and
or
just
subject
prior
art
and
that's
late
from
this
issue.
And
then,
if
you
scroll
further
down
kind
of
five
pages
down,
it
should
text.
I
have
a
list
of
its.
D
D
Or
no
because
there's
actually
table
Eric
there
we
go.
So
this
is
basically
that
subject
field
and
how
that
subject
field
is
looks
in
various
different
kinds
of
infrastructures.
So
that's
kind
of
the
summary
of
of
of
that.
That's
not
something
that
we're
introducing
here,
but
that
that's
something
that
is
very
common
in
the
in
the
infrastructures.
So.
D
For
me,
that's
for
me
it's
that
vast
field
for
me
is
basically
a
given
as
something
that
in
my
head
must
exist
in
a
message
and
I
understand
if
you
come
from
a
different
from
a
different
angle
that
might
not
have
to
so
so,
but
I
in
my
head,
I
have
a
hard
time
closing
that
gap
Rachel.
If
you
would
help
me
closing
that
gap
with
some
additional
tech
solution.
Ok,.
A
D
A
J
Was
slightly
concerned
about
saying
that
subject
or
that
source
was
related
to
how
the
message
was
published
specifically
because,
if
we're
ending
up
forwarding
these
messages,
it
seems
like
meeting
knowledge
of
what
the
original
publisher
was,
doesn't
have
a
good
decoupling
between
producer
and
consumer
and
so
I'd
rather
have
source,
be
some
indication
of
the
source
system.
Now
it
may
be
the
in
Clement
system.
The
source
system
is
the
pub
sub.
D
And
this
is
the
that's
the
place
where
it
goes
into
this.
This
is
where
the
philosophy
philosophical
piece
comes
in
and
that's
actually
where
I
think
where
we
landed
on
source
of
where
I'm,
actually,
where
I'm
happy
with
that
being
source,
because
topic
is
something
because
topic
is
an
overloaded
term.
So
let
me
back
up
one
the
way
how
the
way
how
topic
is
being
described
as
an
abstract
term,
which
means
is
this
is
the
general.
D
This
is
the
channel
the
logical
channel
for
which
I
deliver
events
and
where
subscribe
events
is
as
an
identifier
is
something
that
is
I.
Think
that's
still
true
that
that's
a
good
concept,
however,
since
topic
is
also
used
and
largely
mostly
used
as
a
concept
that
is,
is
a
thing
in
a
pops
up
infrastructure.
D
D
How
much
is
that
worth
that
I
put
the
initial
topic
where
I
published
that
thing
onto
into
that
event,
and
that's
something
where,
even
in
our
team
in
in
the
Avenger
team,
we're
now
actually
explicitly
into
it's
an
ax
native
schema,
adding
source,
because
we
have
now
realized
that
the
that
using
the
topic
because
of
that
overloaded
notion,
it's
probably
not
too
smart,
so
so
yeah.
So
the
the
topic
will
not
be
useful,
but
the
source
really
desi
needs
who
ultimately
put
that
event
out
and
is
a
channel.
So.
D
Agree:
B
yeah,
because
so
what
I?
What
happens
is
this
is
something
that
I've
been
reintroducing
and
we've
been
having
long
discussions
about
the
this
prior
I
wanted
to
make
me
the
pr
fairly
scope
and
not
not
edit
the
rest
of
the
spec
completely,
but
I
agree
with
you
like
so
with
with.
If
we
introduces,
then
that
needs
to
go
with
some
edits
throughout
the
rest
of
the
spec
to
basically
make
clear
what
the
relationship
between
sort
of
subjects,
okay,.
A
So
since
you
everyone
low
on
time
and
since
Clement,
you
missed
the
previous
conversation,
because
your
other
phone
call
well,
the
key
point
is
we
realize
that
we
probably
need
have
a
separate
discussion.
Vector,
combines
ID,
source,
subject
and
type
because,
based
upon
all
the
various
PRS
and
issues
that
are
out
there
right
now,
it
seems
like
we
may
be
converging
on
this
idea
of.
We
need
to
modify
some
or
all
those
four
different
attributes
in
terms
of
definitions,
our
and/or
in
case
the
subject.
A
Add
it
to
the
list
and
when
you
add
this
one,
it's
going
to
back
the
other.
So
I
took
the
action
item
to
create
a
separate
discussion
to
sort
of
have
that
broader
discussion
to
resolve
those
four
attributes
in
combination.
So
we
don't
do
one
PR,
that's
in
conflict
with
another.
We
can
do
it
all
in
one
gigantic
blob,
basically,
yeah.
N
A
A
A
A
And
Victor,
okay
did
I
miss
anybody.
Yes,
you
used
me:
oh
darn,
it
okay,
anybody
else.
Okay,
in
that
case,
I,
believe
we're
done
right.
It
W
our.
Thank
you
guys
very
much
if
you'd
like
to
hang
on
for
those
you
want
to
talk
about.
What's
going
on
at
the
demo,
prep
call
that
we
have
basically
starting
right
now,
you're
free
to
join
otherwise
everybody
else,
we'll
talk
next
week.
Thank
you
guys
very
much.
Okay,
every
talk,
everyone.