►
From YouTube: CNCF Serverless WG Meeting - 2019-06-13
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
C
A
D
A
Talk
a
little
bit
about
that
last
time.
It
I
think
it
might
further
well
demo
seemed
to
work,
which
was
good
I
can't
be
anything
else
worth
mentioning,
but
got
a
good
feedback
from
this
server.
This
session
I'm
sorry
the
service
working
group
session.
In
terms
of
why
people
may
be
resistant
to
your
server
list
and
stuff
like
that,
if
you
check
back
in
the
meeting
minutes
from
last
week,
there
should
be
a
pointer
to
the
for
the
to
the
notes
that
I
took
on
that.
You
can
read
up
on
that.
A
Okay,
yeah
go
back
and
check
the
meeting
minutes
last
week
of
the
line.
Let's
see,
STK
work
didn't
have
a
meeting
that
anybody
want
to
mention
anything
needs
to
cake
stuff.
I
know
I
actually
knows
quite
a
bit
of
activity
going
on
and
the
repos
there.
So
is
there
anything
anybody
wants
to
mention,
okay
and
for
then
kook
on
China
is
in
I.
Think
just
under
two
weeks,
I
have
put
together
some
initial
drafts
of
some
slides,
so
you
guys
could
take
a
look
at
those.
A
The
most
part
of
copy
of
what
we
did
in
that
the
European
cucas
and
just
reminder,
if
you
have
a
ten-point
up
for
the
demo,
please
give
it
running.
I
am
planning
on
showing
the
demo
there
soon
any
questions
are
coming
on
it
all
right
cool.
Thank
you.
So,
in
terms
of
incubator
status,
last
week
we
talked
about
how
they
give
us
definition
of
end
user,
which
is
three
users
of
products
that
when
did
cloud
events,
so
high-level
question
for
people
you
want
to
go
forward
with
incubator
status.
A
I
did
check
and
basically
all
it
means
is
we
get
on
the
TLC
agenda
and
we
present
our
case
for
why.
We
think
we
meet
the
criteria
and
the
biggest
beta
criteria
is
meeting
the
three.
Oh,
the
three
independent
end-users,
so
you
would
need
to
actually
name
names,
basically
of
at
least
three
different
people
who
are
willing
to
say.
Yes,
we
use
cloud
events
in
this
particular
product.
I,
don't
think
that'd
be
a
problem
based
upon
who
I've
heard
so
far
supports
it.
F
A
So
there
is
no
requirement
for
us
to
be
at
one
point:
oh,
we
did
check
on
that
and
I
believe.
The
only
real
difference
in
terms
of
being
a
sandbox
versus
incubator
is
obviously
bragging
rights,
cuz
you're
a
little
higher,
but
we
change,
but
we
do
get
the
ability
to
get
more
as
we're
looking
for
more
of
a
marketing
normal
marking
material
available
to
us.
So,
for
example,
we'd
be
allowed
to
have
our
name
mentioned
in
keynotes
and
stuff
like
that
right
now
they
they
really
frown
upon
them.
A
F
G
Ya,
since
CN
CF
is
a
very
happy
marketing,
happy
organization,
yet
getting
I
higher
up
on
the
flagpole
and
getting
more
attention
from
the
marketing
machine
is
not
bad
and
I.
Don't
think,
that's
necessarily
a
new
maturity
statements
either
so,
whether
we're
at
0.3
or
1.0.
For
me,
those
making
their
friends
that
we
should
just
go
and
climb
the
ladder
as
quick
as
we
can.
Okay.
A
D
A
D
A
A
A
H
A
A
E
A
All
right
well,
thank
you,
guys
very
much
I'll
sort
the
process
and
of
pushing
that
out
the
door.
All
right,
0.1,
discussion,
I,
try
to
summarize
where
we
are
relative
to
the
pro
requests
and
keep
in
mind
that
this
is
categories.
Categories
are
based
upon
my
initial
take
on
where
the
different
issues
and
PRS
fall
into
right,
either
required
for
1.0.
We
should
try
to
do
one
point
know
what
is
that
necessarily
required
versus
definitely
a
post
1.0
thing?
A
Hopefully,
you
guys
I
had
a
chance
to
look
at
that
categorization
that
I
did
and
assuming
it's
more
right
than
wrong.
I
think
this
is
the
current
layout
of
where
things
lie.
So
we
only
have
basically
14
things
in
front
of
us.
For
one
point,
though,
and
I
know
what
Karen
look
was
last
week,
somebody
asked
for
a
little
more
concreteness
relative
to
when
we're
actually
gonna
ship
1.0.
A
The
specifications
and
documents
do
final
reviews
and
stuff
and
start
working
on
the
the
try
1.0
and
once
we
feel
like
we're
done
with
the
testing
period,
then
we
can
decide
whether
we
should
say
no
we're
gonna
ship
right
now,
even
though
we
haven't
finished
all
those
18
open
issues
or
we're
going
to
wait
to
resolve
those.
So
it
gives
it
a
little
bit
of
time
to
sort
of
do
a
little
deeper
analysis
on
those
18.
So
basically
I
guess
the
meta.
A
What
I'm
trying
to
suggest
here
is
we
really
push
hard
to
say,
try
to
resolve
these
14
issues
and
PRS
41.0
within
the
next
say
two
to
three
weeks
and
then
see
where
we
are,
after
that
relative
to
these
outstanding
try
ones,
but
that's
just
an
initial
thought.
I
had
on
the
process,
anybody
else
want
a
voice,
an
opinion.
A
So,
in
order
to
make
that
happen,
I
think
the
biggest
issue
I
have
is
we
have
six
unassigned
issues?
The
four
that
are
signed
are
these
four
people
here,
whether
you
four
know
it
or
not?
I
did
to
sign
those
four
to
you.
Okay,
mainly
because
I
think
you
guys
were
the
ones
to
open
the
issue.
So
I
tagged,
you
is
working
on
it,
so
I'll,
just
six
that
are
there.
I
would
really
appreciate
it
if
you
guys
in.
E
A
You
necessarily
come
up
with
a
poor
request
yourself.
It
just
means
you're
going
to
do
the
the
driving
and
pushing
now
you're
gonna
call
it
to
get
her
with
a
finish
line,
and
that's
why
it
could
also
mean
to
suggest
we
close
it
and
do
nothing
with
it.
But
I'm
looking
for
six
volunteers
and
I
won't
take
up
time
on
this
call
to
do
it,
but
please
you
know
if
you're
on
the
call,
please
look
at
those
six
and
try
to
volunteer
for
them.
A
Okay,
if
I
don't
start
seeing
people
volunteering
over
that
school
day
is
also
doing
some
nagging
myself,
but
I
really
preferred
not
that
the
neck.
Okay,
anything
else
relative
to
one
point
hope
you
want
to
discuss.
I
guess
I
should
mention.
There
are
four
PRS
tag:
41.0
they
need
to
be
updated,
I
know,
I,
think
Clements,
you
know
one
or
two
I
think
I
have
one
I
can't
reveal
any
other
one
is,
but
if
you
do
a
PR
out
there,
that
needs
to
be
updated.
Please
take
a
look
at
it.
Yeah.
A
I
think
we
have
a
meeting
scheduled
for
next
week
after
this
call
yeah.
So
we
should
definitely
discuss
it.
Then
if
we
don't
get
it
resolved
before
then
all
right
anything
else.
Someone
point
nope.
You
want
to
talk
about
all
right
moving
forward,
then
so
during
my
review
or
categorization
exercise
for
those
three
groups,
I
think
came
across
these
four
issues
that
I'm
proposing
that
we
close
either
you
can
easily
no
longer
apply
or
based
upon
the
sudden,
read
the
group.
A
I
didn't
think
people
are
going
to
buy
into
it
in
particular
the
last
two
I.
Don't
think
people
wanted
a
method
attribute
because
that
sort
of
exposed
the
transport
layer
these
fighters
have
in
the
past
it
seems
like
they're
everywhere
you
go,
we
seek
you
sound
awfully
close
petition
key
example
or
an
example
essentially
redefine,
so
it
seemed
like
a
duplicate
I,
don't
think
it's
in
some
architecture,
talk
anymore,
I
think
we're
pretty
much
underway,
and
our
primer
covers
a
lot
of
that
and
I
haven't
heard.
A
Anybody
complain
about
the
fact
that
we
do
the
Plus
Jason
thing
and
our
mind
types
so
I'm,
assuming
that
we
could
probably
close
that
one,
and
if
someone
does
decide
it
is
an
issue
they
could
reuse
to
the
issue
they
want
to
do
so
anyway.
Those
that
was
my
reason
for
those
for
any
comments
or
questions
on
those
four
issues.
A
Okay,
cool
in
that
case
this
reminder,
if
you
believe
it
would
close
one
incorrectly,
we
can't
reopen
it,
don't
feel
like
we
can't
the
only
usually
we
do
require
a
little
bit
of
a
bar.
That
just
says:
if
there's
new
information
right,
if
there's
nothing
new,
you
just
want
to
open
it
just
because
that
might
be
a
hard
sell.
But
if
you
have
new
information
that
makes
it
so
we
should
revisit
our
previous
decision.
Then
that's
how
usually
enough
reason
to
reopen
all
right.
All
right
move
forward.
A
A
So
it's
a
Kafka,
Transport
binding
I,
don't
want
to
go
through
it
completely
right
now,
but
I
did
want
to
ask
Clements
in
particular,
because
I
think
you
have
done
the
most
recent
review
of
it.
Do
you
have
anything
in
here
that
you
think
is
worthy
of
bringing
up
for
fuels,
attention
I,
think
mostly
issues
or
sorry
of
the
comments
you
made
were
relatively
minor
or
so
tactical
in
nature.
More
than
anything
else,
are
there
any
high-level
issues
or
concerns
you
think
might
be
worthy
of
bringing
up
to
the
group
at
this
point?
No.
G
I
think
I
think
the
the
this
has
been
updated
with
a
hint
for
the
callback
mechanism
like
how
the
partition
key
is
being
created
and
that's
sufficient
I
think
the
yeah
that
one
and
then
maybe
the
prefixes
are
a
little
long,
but
that's
cosmetic,
but
otherwise,
not
just
sorry.
The
only
since
the
CAF
got
messages.
The
message
structures
are
relatively
simple
and
doesn't
even
have
a
notion
of
contents
that
we're
introducing
that
with
a
custom.
G
A
H
I
mean
thanks
thanks
Clemens
and
Doug,
for
making
the
updates
today.
I
had
one
question,
probably
for
you:
Clemens
is
I
put
in
the
key
attribute
section
there
I
don't
actually
define
any
precedence
on
whether
a
partition
cage
should
override
the
use
of
a
key
extractor
or
whether
they're
mutually
exclusive
or
not
I,
kind
of
presumed
and
I
guess.
This
is
bit
of
a
gray
area.
H
I
presume
that
the
partition
key
would
take
precedence
if
one
was
provided
and
then
falling
back
to
the
partition
key
extract
a
bit
I,
don't
really
know
what
the
semantics
should
be.
This
just
my
presumption
I
think
Donna.
He
mentioned
that
on
the
the
other
partition
key
P
R
as
well.
So
what
are
your
thoughts
on
that
since.
G
Partition
so
I
would
always
I
would
I
would
construct
it
in
a
way
that
you
always
have
a
function
and
and
one
in
one.
The
default
function
looks
for
that
key
that
key
and
it
makes
that
the
partition
key,
and
then
you
can
write
another
extractor
function.
That
then
goes
and
looks
at
difficult
criteria,
and
that's
that's
how
I'm,
probably
you
want
to
do
that?
Yep?
That
makes
sense,
which
means
which
means
the
specification
per
se
will
always
talk
about
the
extractor
function.
Yeah.
G
And
then
you
can
effectively
point
to
the
extension
and
say
here's.
This
is
what
this
is.
What,
if
our
going
to
do,
but
I
don't
don't
be
prospective
and
you
might
go
and
add
an
example,
expand
that
a
little
bit
more
and
add
an
example.
Here,
that's
is
how
it
would
approach
it
like
that.
There's
there's,
no
preconceived
notion
of
what
that
PK
should
be,
is
always
a
product
of
function
and
and
the
proposal
is
that
that
there's
a
function
that
looks
at
that
key.
That's
it.
Okay,.
G
G
The
function
is
the
mechanism
and
then
there's
multiple
implementations
and
one
of
the
default
implementations
is
to
rely
on
their
yeah,
otherwise,
otherwise
I
would
otherwise
I
would
just
go
and
tighten
up
the
the
prefix
because
I
think
we
shortened
it
also
up
to
so
you
see
II
and
HTTP
I,
think
I
I'm,
not
sure
when
I
already
did
the
Ferengi
p.m.
in
so
go
ahead
and
do.
A
Okay,
is
there
any
reason
to
think
that,
once
those
edits
are
done,
people
will
need
more
time
or
I
will
have
an
objection
just
want
to
get
these
things
out
sooner
rather
than
later,
okay,
so
yeah
so
Neil.
If
you
can
make
those
changes,
you
should
be
able
to
pretty
much
approve
this
one
really
quickly.
Next
week,
then
I
would
assume
soon.
You
won't
find
anything
major
which
would
be
really
cool.
Okay,
all
right
cool.
Thank
you
guys,
all
right,
v1
OPRS.
A
A
It
didn't
include
the
UI
reference
or
time
stamped
in
there
now.
I
know
this
actually
is
gonna
probably
overlap
with.
Hopefully
you
soon
to
come,
PR
for
ministry
that
James
Roper
mentioned
in
the
meantime,
wanted
to
at
least
get
this
one
out
there
to
see
what
people
thought,
particularly
Clements,
I.
Think
since
you
wrote
the
type
section
I
want
to
get
your
take
on
what
you
thought
about
something
like
this.
C
A
I
said
I
think
the
biggest
change
is
he
had
your
reference
and
timestamp
I.
Believe
almost
everything
else
is
the
same,
just
moving
around
of
text,
I,
think
or
anything
else
after
that,
yeah.
G
A
E
A
Okay,
but
in
the
meantime,
I
think
this
one
sounds:
ok,
your
perspective.
Is
there.
B
A
Okay,
this
one's
mine,
I,
was
always
going
through
the
spec
over
the
weekend.
I
noticed
that
the
description
of
timestamp
is
actually
kind
of
vague,
in
particular
the
use
of
the
word
of
event.
It
isn't
clear
to
me-
or
it
wasn't
clear
to
me
whether
that
meant
when
the
occurrence
happened
or
when
the
cloud
event
producer
converted
the
original
event
into
a
cloud
event,
and
so
what
I
tried
to
do
is
to
make
it
clear
that,
ideally,
it
would
be
when
the
timestamp
of
this
happen.
Any
occurrence
actually
happened.
A
However,
I
didn't
want
to
leave
it
out
for
people
who
maybe
cannot
determine
that,
but
still
wanted
to
do
the
timestamp.
So
the
receiver
could
do
some
sort
of
time-based
ordering
if
it
made
sense
for
them
to
do
so,
and
so
I
basically
said
that
you're
allowed
to
use
other
things,
such
as
the
current
time.
A
Then
things
are
kind
of
potentially
gonna
be
out
of
order,
or
at
least
inconsistent
relative
to
what
time
the
receivers
are
going
to
see
for
this
stuff,
and
so
I
try
to
make
it
clear
that
they
have
to
be
consistent
with
the
algorithm
they
use
for
determining
the
time.
I,
don't
see
what
you
guys
thought
about
this.
G
A
Well,
thank
you
Thank
You
Jude
for
the
post,
one.
Anybody
else
have
any
comments.
I
know
they're
a
couple
of
LG
teams
and
the
issue
itself
might
be
R
itself.
So
thank
you
for
those
anybody
have
any
questions
concerns
okay,
any
objection
to
approving
them
all
right
cool.
Thank
you
guys,
easy
all
right
now,
right
now,
I
would
the
normal
pattern
we
having
these
calls
is
I
usually
focus
on
PRS,
because
obviously
those
are
the
most
important
people
who
put
in
work
to
actually
make
spec
changes.
A
I
want
to
get
those
in
there
make
them
wait,
but
just
a
heads
up
because
we're
running
low
on
PRS
41.0.
In
fact,
those
are
the
only
two
of
its
they've
actually
up
in
front
of
us.
What
I
will
what
I
may
do
next
week,
depending
on
the
list
of
issues,
is
that
you
might
actually
start
discussion
some
of
the
bigger
1.0
tag
issues
just
to
so
I'll
move
the
issue
discussion
along
if
I
feel
like
it
needs
it.
A
Obviously,
if
you
think
I
included
something
a
list
or
a
by
mistake,
alright
miss
something
that
should
be
there
feel
free
to
add
it
yourself
to
the
agenda.
But
you
want
to
give
you
a
heads
up
that
we
may
start
talk
about
issues
next
week
and
not
just
for
requests.
You
know,
I
really
really
prefer
talk
about.
Plateau
I
have.
G
It
I
have
a
question
into
the
group.
Then
I
see
the
format
which
I'm
supportive
of
is.
Does
anybody
care
about
having
something?
That's
a
binary
format
like
sea
bore
and
having
aspect
like
for
this
because
that's
that's
the
most
direct
mapping
would
have
from
from
Jason
into
something
that's
binary
in
contact
without
eating
any
schema
or
if
anybody
would
be
interested
in
that.
That's
a
that's.
A
respect
still
be
interested
in
getting
into
the
specs
that.
A
G
It's
it's
worth
it
it's
one
of
those
things
where
I
think
that
having
a
compact
little
MLS
encoding
to
complement
Jason
as
an
option
would
be
super
useful
and
between
I
mean
there's
a
few
options
that
we
have
for
those
its
message
pack
and
C
Borat,
but
C
bar
is
in
the
IETF
RFC.
So
even
though
it's
less
popular,
it's
well
implemented
and
just
project
things
I
like
I,
like
things
that
have
no
specs
to
point
to
so
ya,
know.
There's
a
question.
Grace
Apache
thrift,
RG,
RPC
and
Apache.
G
A
G
G
Avro
is
the
Apache
Avro
is
the
format
that
the
entire
Duke
stack
is
using
for
encoding
data
and
4nf
RPC
internally,
so
there's
a
average,
both
an
encoding
format
and
a
RPC
framework
and
as
encoding
format,
for
instance,
we
use
it
in
the
product,
two
archives
that
so
effectively
our
event
table
is
like
is
Kafka
I
like
it
also
has
a
couple
go
ahead
and
we
spool
out
events
from
our
log
into
binary
packages
and
store
them
at
the
storage,
and
we
do
that,
for
instance,
using
a
pro,
so
ever
is
a
super
compact
format.
G
A
G
And-
and
it
also
has
the
advantage
and
can
carry
its
own
schema,
so
they've
email,
language
and
you
can
go
and
create
a
container
with
Avro,
where
you
package
the
schema
upfront,
and
then
you
reach
the
schema.
Then
you
have
everything
that
you
need
to
decode
the
following
binary.
So
that's
that's
also
a
thing,
so
you
can
do
it
use
you
can
use
with
embedded
scheme
or
without
embedded
scheme
as
fairly
it's
fairly
flexible.
G
D
G
A
H
H
A
Appreciate
that
I,
don't
think
meanness
I
have
to
review
this.
All
right
now
have
to
do
it
right
away.
Obviously,
I
take
it
as
post,
1.0,
I'm.
Sorry
as
a
trifle
in
point
I
believe
because,
obviously
not
nice,
they
required,
but
the
more
we
get
into
the
better.
So
you
know
obviously
fitted
in
as
you
guys
can,
but
it's
not
urgent
for
one
point:
no
any
other
things
we
want
to
talk
about
relative
to
this,
then
okay,
that
sounds
like
we
just
hit
a
fun
time
cool
all
right.
Next
one
it
is
mine.
A
A
Oh
I
think
I
wrote
this
because
I
was
gonna,
I
was
trying
to
address
issue
188
and
they
were
I
think
they
were
a
little
confused
as
to
whether
type
was
related
to
the
actual
occurrence
or
of
the
crowd
event
itself,
or
something
like
that.
So
I
just
wanted
to
add
a
little
clearer
to
here
that
this
is
related
to
the
the
occurrence
itself
and
there
actually
could
be
more
than
one
event
related
to
the
occurrence.
A
But
anyway
they
said
this
is
that
tree
contains
the
value
described
and
the
type
of
the
event
related
to
the
original
occurrence.
So
I
tried
to
tie
it
back
to
hkey
current
himself
that
it's
there's
nothing
normative
in
there.
There's
no
muss
or
anything
like
that,
but
I
tried
to
address
it
as
best
I
could,
if
you
want
I
can
do,
is
go
back
to
the
original
issue,
guys
when
I
see
that
they
were
questioning.
A
D
A
G
There
is
a
the
source,
multiple
things
can
happen
to
the
source
right,
and
so
so
you,
you
know,
you
know
state.
You
know
states
that
the
differentiation
between
different
events
is
necessary
because
one
occurrence
can
cause
multiple
events.
I
think
I
think
the
the
differentiation
between
events
is
necessary
because
the
source,
multiple
things,
can
happen
to
the
source.
G
Multiple
multiple
kinds
of
occurrences
can
happen
inside
of
the
source,
each
having
its
own,
his
own
kind
of
event,
and,
yes,
there's
a
spread.
There
might
be
a
special
case
where
a
single
occurrence
might
fire
off
multiple
different
events:
mm-hmm,
but,
but
now
what
what
you've
formulated
here
seems
to
motivate
the
type
by
a
single
occurrence
having
different
events
fired
because
of
it.
A
G
No,
no,
it's
like
you're
tying
it's
now,
it
seems
to
me
like
one
or
more
events
might
be
generating.
There's
a
rolls
own
event.
That
is,
that
is
the,
and
now
you
try
to
differentiate
events
because
of
the
fact
that
you
have
multiple
events
resulting
from
that
occurrence.
That's
why
you
need
to
have
that
type
and
I.
Don't
think
that's
true!
A
G
A
A
A
B
A
F
B
A
B
F
A
F
A
A
Well,
okay,
I
think
Scott's
right.
We
should
probably
take
us
to
get
up.
It's
not
easy,
as
I
thought
it
was
gonna,
be
I,
don't
think
it's
matter.
Wording
change
because
I'm,
not
the
more
I
think
about
it.
The
less
happy
I
am
with
this
because
in
general,
obviously
that's
true
sources
can
emit
multiple
events
period.
A
If
this,
the
multiple
aspect
of
this
was
going
to
be
related
to
a
single
occurrence.
But
if
that's
going
to
be
up
for
discussion,
then
we
should
take
it
back
to
the
github
issue.
So
let's
defer
this
because
we
have
other
PRS
that
don't
required
necessarily
wordsmithing
okay,
but
it
is
a
good
discussion.
A
So
next
one
I,
don't
guess
I
want
to
or
try
to
approve
today,
I
just
want
to
draw
your
attention
to
it.
Some
of
us
were
talking
at
coop
con
EU
about
how
we
really
nice.
If
we
actually
gave
some
guidance
on
how
to
write
what
I
was
calling
adapters,
which
means
for
popular
events
that
are
generated
out
there
today,
how
do
you
convert
them
into
condiments
that
way
in
case
there
are
multiple
implementations.
A
Those
adapters
at
least
have
some
consistency
across
them,
so
you
can
interrupt
and
that
way,
regardless
of
which
adapter
a
particularly
receiver,
gets
their
messages
sent
through.
They
should
hopefully
get
the
same
cloud
event
as
the
receiving
end.
So
what
I
did
is
I
wrote
three
different
adapters,
one
for
github,
one
forget
lab
and
one
for
AWS
SNS
and
just
basically
put
down
what
I
thought
would
be
the
right
way
to
map
it
based
upon
the
data
that
was
being
sent
to
run
along
as
I
said.
A
I'm
not
gonna
push
to
do
this
today,
because
some
of
these
files
are
rather
large,
I
mean
the
github
one
didn't
even
show
unless
hits
it
a
load.
If,
but
please
take
a
look
when
you
get
a
chance,
in
particular
the
source
and
subject
ones
were
not,
however,
send
clear
to
me.
There
were
times
when
I
could
have
been
I
had
an
option
of
being
consistent
with
other
events
that
are
kind
of
related
to
this
versus
more
purist
in
terms
of
what
value
I
chose
there
and
I
wasn't
quite
sure
which
way
to
go.
A
I
did
comment
on
this
and
the
issue
or
in
the
PR
description
itself.
So
please
read
that
way
to
get
a
chance
anyway,
like
I
said
I'm,
not
gonna,
ask
you
to
review
it
right
now,
but
I
do
want
to
get
that
in
their
words
sooner
rather
than
later,
if
possible,
and
so
please,
let
me
get
a
chance.
Jude
your
hands
up.
D
A
So
yeah,
it
is
not
my
intent
that
every
single
doctor
in
the
world
should
try
to
push
their
specification
to
our
repo.
If
they
wanted
that'd
be
great
I,
don't
think
it
hurts
anything.
I
all
east
wanted
to
get
some
of
the
common
ones
out
there
for
two
purposes.
One
is
because
they're
very,
very
common
and
I
thought
there
actually
might
be
multiple
representations
of
these
types
of
adaptors
I
couldn't
give
up
on
I.
Think
it's
a
very,
very
popular
one.
A
I
think
they're
like
those,
but
the
other
piece
of
it
is
I
found
right
in
the
right
in
the
PR
to
be
an
incredible
learning
experience
it
sort
of
highlighted
some
of
the
problems
like,
for
example,
I.
Think
that's
the
reason
why
I
wrote
the
PR
around
time
earlier
today,
because
it
wasn't
clear
to
me
when
I
should
be
using
current
time
versus
trying
to
dig
through
the
event
that
I'm
getting
received
from
get
up
and
try
to
find
a
particular
time
in
there
to
use
and
I
realized.
A
If
I
made
that
decision
differently
per
the
event
type,
then
the
receiver
is
going
to
get
different
data
right
and
that's
why
I
pushed
for
that
consistency
aspect
that
we
just
approved
earlier
right,
so
I
thought
doing.
This
was
a
great
exercise
one
for
me
to
make
sure
the
spec
made
sense
for
these
attributes,
but
also
that,
as
somebody
who
actually
might
write
adapter
one,
you
can
look
at
the
spec
and
understand
exactly
what
goes
into
fields
like
subject
because,
as
I
said,
was
rayless,
it
was
encouraging.
A
Why
not
to
use
subject
versus
source
for
particular
field,
and
so
I
think
this.
If
nothing
else
can
help
educate
people
if
they
go
off
and
to
write
their
own
adapters,
even
if
they
don't
submit
it
to
us
right.
This
will
help
them
understand
what
we
meant
when
we
wanted
to
say,
fill
this
field
in
with
this
type
of
value
right
good.
We
have
examples,
but
I
think
this
type
of
example
is
even
more
useful
than
just
the
one
or
two
examples
we
have
in
the
spec
itself.
That
makes
sense
yep.
A
It's
my
double
check.
Okay,
like
I,
said:
please
review
this
when
you
get
a
chance
in
particular,
I
want
to
pick
on
Scott,
because
I
know,
you've
been
very
heavily
involved
in
the
K
native
adapters
and
I
want
to
make
sure
that
you're
okay
with
this
with
this
direction,
because
they're
definitely
some
choices
in
here
that
I'm,
not
sure
you'd,
yes,
I,
agree
with,
but
I
wanted
to
try
your
attention
tweet,
a
pic
on
your
little
yeah.
I
A
I
think
that
that's
one
of
the
things
I
would
like
to
at
some
point
have
us
discussed,
probably
initially
through
the
PR
itself.
Unless
we
run
out
lets
me
have
a
lot
of
free
time
on
these
calls,
but
I
will
mention
that
I
did
reach
out
to
get
lab
and
github
directly
to
see
if
they'd
be
willing
to
collaborate
with
us
on
this
and
in
my
mind,
if
we
can
get
them
to
buy
into
the
notion
of
saying
yeah.
Of
course
it
should
be
good,
lap
or
github.
A
A
Have
you
got
Monza's,
let's
see,
get
lab
is
definitely
interested
in
in
participating.
I!
Think
it's
more
about
our
time
right
now.
In
fact,
they
went
one
step
further
and
one
of
the
guys
suggested
that
maybe
we
should
submit
a
pull
request
to
have
get
life
supported,
natively
and
I
thought
was
really
cool.
The
github
folks
I
made
initial
contact
with
them
fix
the
Clements.
A
Thank
you
very
much
and
they
against
him,
but
Jeff
didn't
seem
against
the
idea,
but
they
also
didn't
say
yeah
yeah
yeah
either,
but
in
fairness,
I
just
talked
to
them
last
Friday,
so
I'm
still
waiting
for
them
to
get
back
to
us
and
I
suspect,
they're,
just
busy
so
we'll
see
how
it
plays
out
and
I
need
to
actually
hit
up
Tim
to
make
sure
he's.
Okay
with
this
mapping,
I
did
hear.
Obviously
I
just
sent
him
a
note,
but
he
hasn't
responded,
so
he
may
be
busy.
I
did.
A
Well
briefly,
we
talked
about
it
and
that's
another
reason
why
I
reached
out
to
him,
because
I
thought
they
may
have
already
done
this
mapping,
and
so
he
could
tell
you
where
I
went
wrong:
cool
okay,
so
anyways
I'm,
not
gonna
rush
this
through
other
than
I
do
think
it's
a
great
learning
exercise
and
I
think
it'd
be
useful
for
the
community
in
general
to
have
this
kind
of
thing
all
right.
Any
last
minute
comments
on
that
before
we
move
on
all
right.
In
that
case,
I
think
the
last
one
for
today
Eric.
J
You
know
cryptic
connection,
all
these
sorts
of
things
and
and
there's
a
lot
of
very
deep
considerations
that
we
could
go
down
there,
particularly
for
cloud
events.
It
seems
like
that,
and
particularly
in
the
discussions
that
have
resulted,
it
seems
like
that's
not
something
for
the
core
spec
to
address,
certainly
in
one
dot
Oh,
although
kind
of
what
I
asserted
here
is
that
it
is
expected
that
there
will
be
extension,
attributes
and
I
left.
G
So
I
have
questions
so,
let's
say
so:
we're
implementing
a
an
event
store
so
like
event
table
similar
to
Kafka
and
that
data
gets
persistent
there
and
we
have
a.
We
have
a
mechanism
today
and
I
think
Kurt's
data
addressed
and
we
have
in
the
particulars
who
also
this
cue,
also
a
way
for
you
to
give
us
a
key
to
go
and
store
that
to
control
the
the
how
the
encryption
works
and
we
have
you
know,
input
authorization
and
output,
authorization
who
can
get
who
can
get
to
that
data.
G
J
J
If,
if
that
were
to,
if
the
the
writing
and
the
reading
were
being
done
by
different
parties,
then
there
would
need
to
be
some
kind
of
a
agreement
over
the
the
ways
that
that's
done
and,
and
it
could
be
that
that's
all
informally
agreed
outside
of
the
of
the
the
kind
of
specification
and
outside
of
some
sort
of
a
standardization.
But
if
the
parties
don't
necessarily
know
who
you
know
if
they're
a
right
or
what
parties
are
going
to
read
or
what,
if
they're
a
reader,
what
kind
of
what
parties
are
going
to
write?
J
G
So
I
wonder
and
I
really
wonder
what
the
with
the
weather,
this
weather,
you
are
not
trying
to
hint
at
mechanisms
that
are
that
that
kind
of
exists
for
messaging
systems
that
is
like
into
an
encryption
and
into
an
identity.
Because
that's
that's.
That
seems
where
that's
going,
because
there's
no
difference
for
at
least
not
in
my
mind,
for
whether
you
keep
a
message
on
the
wire
or
whether
you
put
that
message
into
into
a
disk
at
storage
and
do
a
stored
forward.
G
Because
the
the
wire
established
context
is
something
that,
in
in
in
in
common
pop
subsystems,
kind
of
terminates
at
the
gateway
and
then,
as
he
hand
that
out
to
two
subscribers,
you
don't
carry
that.
That
established
context
that
you
have
like
the
connection
context.
You
don't
hand
that
forward
to
the
subscribers.
J
Perhaps
another
way
for
me
to
to
say
this
is:
is
I
think
that
it's
it's
a
declaration
that
that's
an
exercise
left
for
implementers
and
that
we're
not
going
to
try
and
solve
that
within
the
cloud
event.
Spec
though,
perhaps
we
I
don't
want
to
declare
that
we'll
never
provide
support.
Isn't
it
seems
like
there's
enough
of
a
need
for
solving
this,
that
support
could
be
helpful,
but
that
we're
certainly
for
version
one
not
going
to
to
kind
of
dip.
Our
toes
and
yeah.
G
G
We've
had
twice
or
three
times
about
identity
and
encryption
because
we
have
effectively,
if
you
we
even
had
PRS
or
issues
that
were
specifically
proposing,
that
we
introduced
some
notion
of
end-to-end
encryption
or
into
an
identity,
and
we
scoped
both
of
those
out
with
you
know,
knowing
what
introducing
those
things
did
to
other
standards,
at
least
at
least
early
on,
and
then
decided
that
we
might
want
to
take
those
on
post
with
one
that's
appointed.
If
there's
enough
interest
and
I
think
I
think
it's
a
worthy
thing
to
to
ultimately
deal
with.
G
We
just
don't
have
the
necessary
infrastructure
right
now,
standardized
infrastructure
right
now
to
do
with
it,
because
we
need
to
have
key
registries.
We
need
to
be
able
to
go
and
talk
to
those
key
to
our
industry's
needs.
Have
key
rolling
and
there's
all
kinds
of
complications.
I'm
I'm
wondering
reading
that
text,
I'm,
not
sure
how
much
that
helps
me,
but
maybe
it
helps
others.
G
A
Okay,
is
there
anything
else
in
the
cards
like
the
voice
of
opinion,
we
have
about
30
seconds,
left
I'm,
trying
to
see
whether
we
should
vote
now
or
wait
I'm
inclined
to
say
we
wait
just
a
little
cuz
I'm,
not
sure
everybody's
had
a
chance
to
review
it
yet
Eric.
Would
you
be
okay
if
it
wait
till
next
week.
A
Cool
my
keys,
please,
when
you
guys
get
a
chance,
take
a
look
at
this:
it
is
it
isn't.
The
primers
was
non
normative,
but
we
do
want
to
make
sure
the
text
actually
accurately
reflects
our
current
thinking
all
right,
but
that
it
is
the
top
of
the
hour.
So
one
last
little
check
did
I
miss
anybody
in
the
attendance
list
wow.
This
is
all.