►
From YouTube: CNCF Serverless WG Meeting - 2019-01-31
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
Even
those
later
on
and
sits
on
my
head
right
now,
1a
Clements
action
item
was
to
make
sure
there
was
a
call
set
up
today
to
discuss
PR
218
and
the
topics
around
that
I
believe
it's
at
2:00
p.m.
Eastern
11:00
a.m.
Pacific.
It's
using
the
same
zoom
call
as
this
one.
So
if
you
want
to
join
in
the
conversation,
please
remember
to
join
that.
A
Rachel's
action
item
result
is
on
the
agenda
today,
so
we
do
that
later,
all
right.
Community
time
is
there
any
community
related
topics.
You
would
like
to
bring
up
I'm,
just
anybody
new
on
the
call
so
I'm
guessing
no
okay,
not
hearing
a
name.
Kid
move
forward:
okay,
SDKs
subgroup!
We
did
have
a
call
right
for
this.
One
I,
don't
think.
There's
anything
earth-shattering
worth
mentioning.
I'll
just
check
the
notes
here
very
quickly.
A
A
B
In
short,
in
short
terms,
there
was
a
issue
raised
of
the
c-sharp
sdk
were
the
c-sharp.
Sdk
did
not
does
not
magically
turn
a
object
graph,
it's
adjacent
in
binary
mode,
and
that's
intentional,
because
the
assumption
is
that
you
bring
either
a
readily
stringing
coded
data
or
a
binary
encoded
data
and
that's
being
put
into
the
payload,
because
the
encoder
that
we
have
is
not
really
for
the
payload,
but
it's
for
the
envelope.
B
A
Sounds
good
Thank,
You
Hans
say
that
was
the
SDK
I
don't
see
Kathy
on
the
car
and
I.
Don't
think
anything
happen
with
the
work,
though
subgroup
something
that's
in
there.
So
Scott
on
the
proposal
for
the
next
demo,
like
I
apologize
I
did
not
get
in
his
look
at
this,
even
their
updates
here,
but
is
there
anything
you
would
like
to
mention
here?
I
think
some
people
may
have
commented
on
here.
C
A
A
A
A
We
should
probably
start
thinking
about
who
is
going
to
be
giving
a
talk
during
our
intro
and
deep
dive
sessions,
and
you
know
what
the
various
topics
were.
Alright
I
think
the
intro
is
probably
pretty
much
what
we've
done
in
the
past.
This
is
for
newbies,
who
pretty
much
know
nothing
about
us.
We
could
probably
use
a
lot
of
the
same
material
there
I
do
think.
A
There's
gonna
be
more
opportunity
for
what
sort
of
look
for
new
stuff
or
invention
in
the
deep
dive
section,
because
we'll
have
newer
things
in
spec
to
talk
about
or
potential
other
exciting
things
just
to
mention
there.
So
I
think
that
might
be
where
the
more
interesting
area
lies,
but
we
should
probably
first
start
talking
about
one
who's
going
and
then
we
have
that
set.
Who
would
he
would
want
to
actually
be
part
of
those
presentations?
A
That's
an
interesting
question:
what
other
people
think
I
guess
there's
two
aspects,
so
one
is:
do
we
think
it's
worthy
enough
put
it
that
way,
but
then
also
I
would
need
to
check
with
Dan
Khan
and
the
CN
CF
guys
to
know
whether
sandbox
projects
are
allowed
to
be
that
that
prominently
featured
it's
a
different
issue.
But
what
do
people
think
is
this
something
you
guys
would
want
to
pursue.
F
A
Can
you
set
up
a
brownie
table
tea
and
put
brownies
on
there
and
sell
lemonade
the
same
thing
so
tell
you
what
Scott?
Let
me
let
me
if
it's
worth
at
least
asking
whether
it's
even
possible
I
suspect
that
the
folks
have
mentioned.
Is
there
already
partly
correct?
It
probably
would
require
a
fair
amount
of
money
which
I
don't
know
if
anybody's
willing
to
put
forward,
but
the
other
aspect
that
I'd
actually
concerned
about
is
this,
like
I
said:
were
they
enough
to
actually
be
there
24/7
on
the
bond
the
booth
floor,
yeah.
A
A
B
A
Okay,
so
yeah
we
should
probably
first
let
me
to
gather
the
list
of
people
and
then
we
can
have
another
call
or
a
different
discussion
to
figure
out
what
we
want
to
do
there,
because
one
of
the
things
that
I'd
like
to
do
is
give
an
opportunity
for
newer
folks
or
other
people.
To
present
there
was
a
lot
Clemens
you
and
I
presented,
so
we
so
may
be
best
if
we
stood
back
a
little,
let
some
of
the
guys
do
it
if
they
wanted
to,
but
we
can
figure
that
out
later,
but
we'll
see
so.
A
I'll
put
my
name
in
there
too,
because
I'm
willing
to
talk
if
necessary,
but
I'll,
say
okay,
so
think
about.
If
you
guys
wanted
to
potentially
talk
and
you'll,
be
there,
let
me
know
otherwise
what
I'll
do
is
I'll
set
up
another
call
not
in
a
not
too
distant
future,
because
people
need
to
start
making
planning
arrangement
and
stuff
like
that
and
get
permission
and
obviously
being
one
of
the
speakers
can
help
with
approvals
from
your
company.
So
I'll
set
up
a
call
for
us
to
talk
about
that
all
right,
Oh.
A
Know
the
structure
yep.
Yes,
we
will
definitely
have
at
least
the
intro
and
deep-dive
whether
there's
an
additional
session.
There
I
think
that's
up
to
whether
you're
speaking
whether
someone's
submitted
a
call
for
also
around
it
and
whether
it
gets
accepted
or
not.
Now,
having
said
that,
I
just
remembered
gondii,
a
blank
Chris
Chris
Anna
check
pasted
a
message
to
our
slack
channel,
saying
that
they're
thinking
about
doing
some
sort
of
service
conference
thing
I
can
really
like
phrase
he
used
at
coop
Connie
you
yeah
I
can
write
if
it's
a
half-day
or
full-day
thing.
A
I
think
it
may
be
a
full-day
thing,
so
that
may
be
another
opportunity
for
us
to
present
what
we're
doing
there.
I
don't
know
exactly
what's
going
on
so
I
may
be
completely
wrong,
but
that
may
be
an
opportunity
there
as
well.
So
I
get
more
information
about
that
as
we
go
forward.
So
maybe
another
opportunity.
A
A
Okay,
now
I
know
that
in
the
in
the
section
of
our
of
our
bar
repo,
we
do
have
sort
of
an
open
source
section
with
list
of
open
source,
implementations.
I,
know,
I,
think
there's
at
least
one
or
two
sort
of
cloud
event:
verification
tools
in
there.
So
there
might
be
a
good
starting
point
for
us
to
consider
but
I'm,
assuming
you're,
probably
thinking
of
something
a
little
more
formal
than
just
someone
puts
a
thing
out
there
right.
A
C
A
H
A
Cool,
thank
you
all
right.
Moving
forward
on
to
pr's
papi!
Oh,
isn't
that
on
the
call.
However,
we
talked
about
this
one
last
week
and
I
think
we
were
generally
okay
with
it.
But
there
was
one
question
because
I
think
he
had
on
oh
just
refresh
his
memory.
What
he
wanted
to
do
was
to
modify
the.
What
is
this
the
swagger
duck
to
give
a
minimum
length
to
all
of
our
strings
and
at
one
point
he
actually
had
a
constant
here
of
0.2
and
I
came
here
who
it
was,
but
somebody
pointed
out.
A
That
means
you
can't
support
anything
but
0.2
and
that's
very
problematic
for
forward
and
backwards
compatibility.
So
he
ended
up.
Removing
that
I
think
other
than
that
there
was
I
was
the
only
change
he
made.
So
are
there
any
other
questions
or
concerns
about
this
one,
because
we
almost
improved
this
last
week,
except
for
that
one
comment.
A
I
B
Did
the
implementation
of
it
and
the
only
thing
and
that's
one
of
the
things
I'm
going
to
go
and
put
into
the
SDK
notes-
is
that
the
new
media
type
that
were
defined
a
year
for
that
format
is
as
there's
a
risk.
This
is
a
bug
risk
here,
which
is
not
too
grave,
but
it's
an
easy
buck
to
make,
because
the
cloud
event
applications
are
called.
B
But
that's
really
the
only
thing
I've
found
otherwise
from
a
factoring
perspective.
Since
we're
now
move
that
into
the
into
Jason's
thing,
the
Jason
now
jason
has
exactly
matching
one
let
they
should
be.
Does
that's
really
where
it
lands
is,
that
was
distinguishing
the
messages
by
the
the
media
type
and
then
really
dealing
mostly
with
the
Jason
decoding
and
not
with
making
this
HTTP
specific.
Really,
the
HTTP
specific
piece
is
touching
the
evaluating
the
content
type.
So
that's
right,
that's
the
right
model.
So
for
me
that
is
good
and
we
should
be
merging.
A
A
Yeah,
it's
funny
as
we're
talking
to
me
cake
all
about
this.
It
I
can
see
the
fact
that
the
the
start
of
it
messes
I,
sorry,
it
starts
with
application
cloud.
Events
can
I
say,
be
a
good
or
a
bad
thing
right
cuz.
If
you
just
want
to
find
all
cloud
events,
Jesus
went
to
be
nice
to
only
at
the
search,
the
beginning
part
of
the
string.
A
H
A
I
I
tried
to
make
this
a
part
of
the
adjacent
format
and
that
I
said,
but
I'm
not
100%
sure
what
I'm
trying
to
say
there
is
the
Jason
event
format
and
there's
the
Jason
batch
event.
Format
and
those
are
technically
completely
separate
things.
I
mean
logically
they're
very
similar,
but
when
we
say
Jason
event
format
we
mean
the
non-veg
one
remember
say:
Jason
badly,
read
format,
we'll
rename
the
other
thing
one
thing
to
make
it
even
clearer
will
be
to
split
it
up
into
two
files,
but
I'm,
not
sure.
I
A
I
Personally,
I
don't
want
to
force
everyone
to
use
the
best
mode
if
we
take
the
function
as
a
service
as
an
example,
you
have
the
model,
you
get
an
HTTP
request
that
should
run
on
a
single
instance
of
a
function
process
the
event
and
then
be
done.
So
you
kind
of
have
the
model
that
a
single
HTTP
request
should
model
to
one
event.
Once
you
make
a
better
of
events,
then
this
model
breaks.
So
this
is
one
good
reason
why
I
don't
think
I
want
to
force
everyone
to
do.
A
It
literally
says
that,
but
the
problem
is
I
think
people
will
very
quickly
confuse
the
batch
Jason
format
with
the
normal
Jason
format
and
think
they
have
the
support
both
and
so
I'd
like
to
vent
this
sentence
just
to
make
that
clear,
I'm,
not
gonna
change
the
semantics
I
just
wanted
to
get
clear
that
we're
only
talking
about
the
one
format.
That's
okay,
sure!
A
A
Just
just
make
sure
I
can
every
sometimes
tell
whether
it's
I
agree,
I,
don't
need,
say
anything
or
I,
just
don't
give
a
crap,
so
that's
it.
So
this
is
good.
Thank
you
for
speaking
up
okay.
So
let's
go
with
this,
so
I
I
believe
that,
let's
see
where
is
it
I
think
there
are
Co
spots
that
have
to
be
data.
Content-Type
I
think
those
are
obviously
very
like
typos
kind
of
things.
A
What
I'd
like
to
do
is
treat
the
tweak
to
this
sentence
as
editorial,
but
so
I
like
to
at
this
point
is
ask
for
approval,
conditioned
upon
that
editorial
tweak
and
then
once
I
get
say,
2lg
Tam's
offline.
We
can
then
merge
it
in,
but
that's
all
predicated
on
the
fact
that
you
guys
approve
this
as
it
is
today
with
the
agreement
of
those
minor
editorial
tweaks.
So
is
everybody?
Okay
with
that.
B
B
A
B
Which
is
which
is
effectively
if
we
wanted
to
if
we
want
to
go
and
send
time
series
data-
and
we
say
the
time
series-
is
all
encoded
s-
s
with
called
events
metadata,
which
means
every
single
time.
Event
record
is
a
cloud
event
per
se,
then
for
in
our
industrial
data,
etc.
Those
batches
can
become
very
large
and
then
it
really
makes
a
lot
of
sense
to
use
a
effectively
a
table
format
where
you
have
headers,
like
close
your
eyes.
Imagine
CSV
but
different
differently
coded
and
so
there's
ways
to
do
that.
B
That
requirement,
because,
because
I
think
it's
real
I
think
it's
a
real
case,
but
then,
if
we
want
to
set
these
fairly
large
batches,
then
and
I
think
we're
going
to
work
coming
to
the
minimal
maximum
size
thing,
then
we
should
probably
have
any
clothing.
That's
that's
more
efficient
and
it's
actually
really
good
at
these
use.
I'm
serious
batches
so
get
that
much
is
a
Commons
of
probably
a
I
bug
issue
that
we
should
do
it
and
investigate,
but
I
think
it's.
A
B
I,
don't
think
it's
a
I,
don't
think
it's
a
batch
of
individual
records,
but
it's
a
it's
a
different
kind
of
it's
a
time.
It's
it's
a
batch,
but
it's
not
a
you
know,
take
an
event
in
a
particular
shape
and
now
send
multiple
of
those
but
you're,
literally
starting
with
I,
want
to
send
site
time
series
data
and
here's
about
your
records
right.
A
So
let
me
ask
this
question:
is
there
anybody
on
the
call
who
disagrees
that
this
is
a
separate
issue,
not
not
passing
judgment
on
whether
it's
a
good
or
a
bad
idea,
but
just
whether
it
should
hold
up
this
particular
PR
itself,
because
I'm
hearing
for
at
least
from
gem
and
now
hand
Clemons
that
they
view
this
as
a
separate
issue?
And
it's
not
necessarily
blocking
one
for
this?
Is
there
anybody
on?
The
call
would
disagree
with
that
assessment.
A
I
I
The
mqp
message
is
actually
above
256
or
whatever
our
limit
will
be,
or
you
still
have
to
accept
it,
because
the
event
contained
in
that
message
is
longer,
but
if
you
make
it
smaller
than
that,
so
if
you,
if
you
would
also,
if
you
take
a
different
event
and
serialize
it
and
compete,
and
it
would
turn
out
it
is
below
250
kilobyte.
There
is
also
no
guarantee
that
is
being
accepted,
because
if
you
would
serve
as
Jason,
it
may
turn
out
to
be
bigger.
I
So
if
you
are
a
middleware
that
wants
to
guarantee
that
you
can
send
you
end
on
what
you
should
always
do
is
measure
it
in
trace
and
if
you're
just
concerned
about
like
rejecting
messages
so
that
you
protect
your
own
self,
then
you
can
take
an
sort
of
arbitrary
limit
that
will
fit
any
256
kilobyte
or
whatever
limits.
Jason
I
hope
that
made
some
sense.
I
struggle
a
bit
with
that.
I
A
B
These
numbers
always
totally
arbitrary,
which
makes
it
kind
of
hard
to
go
and
argue
for
or
incident,
because
you
know
our
64k
as
are
as
arbitrary
as
the
256k
here,
so
having
a
rational
argument,
for
it
is
for
or
against
it's
kind
of
difficult.
The
reason
why
we
made
it
smaller
is
to
force
everything.
That's
everything
is
PII
and
everything.
That's
you
know
a
binary
file
to
force
that
either
into
a
callback
model
or
to
force
a
claim
check
pattern.
So
we
it's
it
wasn't
it's
it's.
B
That's
there
for
all
of
us
are
in
one
region
and
then
the
second
is
to
literally
keep
the
payload
size
small,
so
that
you're
effectively
forcing
everything
that
requires
access,
control
to
view
with
that
size
limitation.
We're
making
it
fairly
obvious
that
you
should
turn
around
and
go
to
back
to
the
place
and
so
affected
that
the
the
place
that
raised
your
your
event
to
get
at
stuff.
That
requires
access
control
because
that's
one
of
the
things
from
a
privacy
perspective
and
I'm.
B
Just
explaining
to
you
that
the
product
rationale
here
from
a
privacy
perspective,
since
pops
up
is
generally
a
thing
where
you
are
raising
an
event.
And
then
you
have
a
bunch
of
parties
being
able
to
subscribe
without
having
differentiated
access
control.
You
probably
don't
want
to
include
too
much
detail
into
that
event
that
you're
raising,
but
just
effectively,
including
just
enough
information
with
that
event.
B
For
that
other
party
to
know
whether
it's
relevant
for
them
and
if
they
need
to
have
further
detail,
they
just
turn
around
and
walk
up
to
the
the
center
and
there
they
run
into
an
access
control
gate.
And
if
there's,
if
they're
not
authorized
to
see
whatever
personally
identifiable
data,
etc,
then
they
won't
get
at
it.
So
we
have
basically
those
considerations.
We've
been
landing
at
64k
being
a
fairly
reasonable
limit
when
I.
J
Yeah
I
do
hey
I'm
masquerading
as
glad
we're
at
one.
So
I
agree
with
what
what
comes
is
saying.
I
guess
I'd
like
to
hear
from
some
of
the
people
that
were
arguing
for
larger
messages,
originally
I
believe
they
sort
of
had
I
only
use
cases
where
you
know
claim
checks
may
not
work
for
them.
So
I
I
understand
you
know,
closes
point
I,
just
wanna
make
sure
whatever
limit
we
put
on,
you
know
is
going
to
support
those
use
cases
that
people
were
concerned
about.
J
K
About
that,
I
will
say:
I
don't
think
raising
it
works
early,
because
if
it's
a
minimums
for
its
size
and
here
on
pipelines
supporting
all
the
Bossier
want
any
middleware
in
the
middle
of
it
supporting
only
version
one
year
long
and
be
guaranteed
the
smallest
smallest
size.
That
was
very
specified
in
your
major
version.
J
A
A
So
so
I
sort
of
raised
my
hand,
then
to
things
that
run
through
my
mind,
is
I'm
nervous
about
the
first
thing
about
to
say,
because
I
actually
like
having
strong
words
in
there,
but
the
must
they're
familiar
of
building
perspective.
Obviously
a
must
is
the
right
way
to
go.
However,
on
something
like
this,
where
we
know
there
are
some
systems
out
there
that
can
only
support
things
that
are
smaller.
A
J
A
J
A
A
B
Groups
to
use
cloud
events
and
I
know
that
there
will
be
limited
constrain
areas
where
we
can
probably
just
barely
fit
a
cloud
event
even
into
a
transfer
frame
and
having
a
recommendation
here
is
for
know
that
the
general
abroad,
everybody
is
in
the
clouds,
the
area
of
having
recommendation.
There
is
certainly
better
than
a
must
where
it's
then
becomes
fairly
impossible
to
go
through
compliant
implementation.
For
these
cases,
at
the
edge
in
embedded
systems
and
and
those
and
those
folks
are
relatively
humorless
when
it
comes
to
normative
language.
I
Chris
Berg,
in
that
case
firmly
then
just
say,
must
be
64
kilobyte
then,
because
I
think
then
I
like
the
strong
language
and
then
having
a
lower
limit,
then
basically
having
or
how
would
I
say
so,
I'm
the
guy
who
sense
events
and
I
need
to
make
sure
for
my
customers
that
they
arrive.
If
they
don't
arrive,
then
I
am
the
one
that
has
the
problem.
I
B
The
reason
why
it
must,
but
why
really
hard
rule
is
it's
difficult?
Is
that
the
further
you
look
down
into
the
into
different
systems?
The
smaller
your
sizes
gets
that
you
can
support
and
I
think
there's
still
applicable,
atif
comm
events
in
that
case,
so
let's
say
if
you
wanted
to
go
and
make
loud
events
work
on
TSN,
which
is
time-sensitive
networking,
which
is
a
if
nobody
has
heard
about
that.
It's
a
way
to
create
an
Ethernet
setup
which
gives
you
guaranteed
hard
real
time
over
Ethernet
with
an
extra
layer
of
hardware
and
software
pieces.
B
But
you
can
literally,
if
you
say,
I,
want
to
have
the
message
a
message
over
at
this
other
place
in
one
microseconds,
it
happens
in
exactly
one
microsecond
and
it's
a
the
TSM
layer.
Does
that
but
TSM
is
not
IP,
it's
not
TCP.
It's
none
of
those
things
you
basically
just
get
transport
frames
and
the
transport
frame
you
get
to
use
is
one
and
a
half
K
which
fits
and
column
ends,
especially
if
we
have,
if
we
had
a
binary
encoding,
for
instance,
of
a
quarter
buff.
B
But
then
you
know
you're
literally
constrained
to
to
whatever
else
you
could
put
into
that
payload
which
might
be
sufficient
or
is
sufficient
for
some
some
real-time
applications.
So
there's
there's
further
you
further
down
at
the
edge
you
go,
the
smaller
the
sizes
get
and
which
makes
it
really
hard
for
us
to
kind
of
make.
B
A
Look
people
think
it
seems
like
we're
hearing
it
on
two
possible
choices:
change
the
must
who
is
strongly
recommended
or
potentially
reduce
the
size
now
or
is
there
another
option
out
there
I'm,
not
hearing
anybody
complain
too
much
about
this
direction
in
general.
It's
just
tweaks
at
this
point.
It's
guy
named
big
tweaks,
but
we're
important
tweaks,
but
still
I'm,
not
hearing
anybody
object
to
this
general
direction.
So
as
another
option,
people
can
think
of
a
poor
point
of
discussion.
I
I
You
could
more
or
less
explicitly
say
while
you
support
and
then,
if
you
build
up
the
system
you
plug
in,
you
have
to
make
sure
that
you
plug
in
in
the
right
part,
so
it
doesn't
make
sense
to
plug
in
a
producer
who
does
one
megabyte
events
into
a
decent
system
that
only
consumes
one
kilobyte.
So,
instead
of
having
one
set
limit,
you
can
sort
of
choose
or
Clara
where
you
have
well.
It
also
adds
a
lot
of
complexity,
I'm,
not
sure
if
it's
perfect,
but
that's
one.
A
I've
heard
I
heard:
different
people
have
been
reactions
to
profile
some
people,
probably
okay
with
it,
and
other
people
go
screaming
from
the
room,
and
you
mentioned
profiles
because
they
think
back
to
the
days
of
the
WS
star
standards,
because
around
security,
when
you
had
different
profiles-
and
that
was
great-
everybody
was
compliant
with
their
own
profile.
It's
like,
he
had
really
thought
I
was
older.
I
was
like
what
I
meant
remembers.
Those
days.
H
A
A
L
Would
suggest
maybe
I'll
give
you
one
more
give
me
more
time,
one
more
week
for
people
to
think
about
this
I
think
this
transport
and
even
consumers,
other
ones
that
you
know
I
need
to
be
compliant
with
to
support
this
size
right.
This
is
a
basically
the
size
that
need
to
be
supported,
as
the
event
goes
along
the
way,
because
all
the
way
to
the
eventual
event
consumer,
so
all
the
middleware
and
and
the
final
even
consumer,
need
to
support
the
size.
A
Okay,
well
so,
okay,
so
Kathy,
you
expressed
an
opinion
or
a
preference
in
there.
So
thank
you
for
that,
but
then
you
also
suggest
that
maybe
give
people
a
little
more
time
to
think
about
it
and
I
like
the
idea
of
getting
little
people
a
little
more
time,
mainly
because
I
don't
think
this
is
critical.
It
has
to
go
in
today,
cuz
I,
don't
think
this
is
necessary
going
to
change
the
implementation
code.
A
A
We
could
look
at
sort
of
start
up
a
vote
if
nothing
else,
maybe
at
a
bare
minimum,
a
sort
of
sense
of
preference
for
which
direction
we'd
like
to
go
and
see
how
that
takes
us,
like
I,
said
I'd
like
to
try
to
avoid
too
much
of
a
formal
vote
on
this.
We
don't
think
this
a
contentious
and
I.
Don't
think,
there's
a
there's,
an
obvious
answer
out
there.
L
Yes
sounds
good,
I
have
a
question,
is
it
what
options
are
you
do
mean
two
options
different
size,
or
do
you
mean
that
one
option
is
to
set
aside?
The
other
option
is
like
I
mentioned
before
you
use
some
profile
use
some
messaging
sync
up.
You
know
under
I
mean
this.
A
L
A
A
G
Know
if
I
ot
producers
would
are
much
more
likely
to
user
and
system
selected
events
setting
a
lower
limit
that
might
in
fact,
cloud
producers
where
we
have
higher
limits
might
not
be
worth
it,
but
if
there
is
actual
IOT
interest
for
this.
Obviously,
a
lower
limit
is
good.
It's
kind
of
a
mean
comment,
but
we
kind
of
have
to
take
this
in
consideration
to
like
our
entirety
and
edge
producers
most
like
more
likely
to
use
on
proprietary
event
system
and
even
format.
Obviously
we
do
want
it
for
probability.
A
So
maybe
we
can
use
this
week
to
try
to
reach
out
to
some
of
our
colleagues
and
friends
and
see
if
they
have
an
opinions
on
this
and
we
can
bring
those
to
the
next
week's
call
that
fair
okay,
but
with
that
what
I'd
like
to
do,
is
to
move
on
and
then
see.
If
we
can
wrap
this
up
next
week's
gay
people
we
could
think
about
it,
but
with
a
short
time
that
we
have.
What
I'd
like
to
do
is
ask
Christophe
to
quickly
introduce
his
claim,
check
thing
and
then
Rachel.
A
Maybe
you
could
give
us
a
minute
or
two
overview
of
your
PR
as
well.
Both
of
those
are
relatively
new,
so
I
wasn't
gonna
push
for
a
vote
or
anything
like
that.
Today
got
a
big
one
at
least
get
the
ideas
out
there
we'll
talk
about
them,
so
you
can
start
thinking
about
it
so
crystal.
Maybe
you
could
summarize
this
one
yeah.
I
So
this
is
bit
the
Cleary
claim
check
pattern.
So
there's
a
new
attribute
next
to
the
data,
that's
called
a
no
ref
and
it's
basically
a
reference
to
where
you
can
also
get
the
same
data,
so
you
can
have
both
a
at
the
same
time
or
only
one,
and
there
are
basically
free
use
cases
for
it
that
are
outlined
here.
The
first
one
is:
the
content
is
too
large,
basically
what
we
discussed
before.
So
if
it's
Balad,
you
put
it
into
a
different
place
and
then
the
consumer
can
retrieve
it.
I
In
that
case
it
would
probably
or
it
could
be
public
and
then
the
two
other
cases
are
more
security
related.
The
first
one
is
that
you
want
to
verify
that
the
data
hasn't
been
tampered
with,
so
you
would
retrieve
it
again.
Maybe
it's
supplicated
and
I
would
check
that.
It's
really
the
same,
and
and
you
trust
that
source
and
the
other
is
that
face
also
talked
about
today.
You
have
some
personally
EDI
identifiable
information
in
there.
You
don't
trust
all
middleware
between
it,
so
only
a
trusted
consumer
can
retrieve
it.
I
So
you
don't
put
the
data
into
your
message.
You
want
to
put
in
the
data
ref
and
then
only
the
trusted
consumer
has
to
secret
to
feel
it
yep
and
then
I'm
being
pretty
open-ended
I'm,
just
saying
it's
a
URI
reference
and
then
have
fun
with
that.
There's
also
well
on
purpose.
There
is
nothing
inside
the
message
so
that
can
authenticate
at
that
point.
The
idea
is
that
in
one
way
or
the
other,
the
secret
has
to
be
pretty
sure
yeah,
that's
the
overview,
I
think,
okay,.
A
B
I'm
not
sure
whether
that
that's
a
first
class
thing
and
by
that's
not
just
in
inside
the
data,
because
the
event
could
be
not
only
about
one
context
but
about
many
and
then
your
for
many
aspects
of
a
context.
And
then
you
have
you
know
the
distinctions
between
a
thing
where
you
can
only
have
one
of
your
I
versus
a
payload
where
you
that
need
to
and
then
and
then
you're
breaking
it
out
into
it.
So
it's
it's
not
it's
not
clear
that
that
needs
to
be
first
class.
First
class
object
are.
A
B
B
No
I
think
I
think
actually
I
think
it's
very
legitimate
if
you
have
a
a
single
large,
so
I
have
a
large
document,
and
you
want
to
tell
the
world
about
that
large
document.
It
would
be
unwise
to
go
and
send
a
copy
of
that
large
document
to
everybody,
but
rather
you
want
to
go
and
just
inform
them
that
just
something
that
happened
with
that
large
document.
Anyone
to
point
people
to
it
so
I
think
this
is
legit.
B
A
A
J
Oh
shoot
I
just
remember
what
I
was
trying
to
say:
okay
data
is
I'm,
gonna
show
migrants.
Data
is
essentially
no
blob,
though,
isn't
it
so
me
I
to
Clemens
a
question
would
be:
how
can
it
either
be
an
opaque
blob
or
a
reference
to
something
else?
They
did
not
make
it
not
very
well
typed,
but
maybe
I
missed
the
thrust
of
what
what
you
were.
Oh.
B
Well,
if
you
prompt,
if
you
make
it
up
part
of
the
event,
payload
and
say
aspects
of
this
payload
might
be
elsewhere
and
you're,
including
your
eyes
to
those
aspects,
okay,
making
it
more
explicit.
That's
that's!
That's!
Why,
like
I'm,
not
I'm,
not
in
principal
opposed
to
you
know
having
the
payload
it's
external
and
using
that
reference
just
from
a
from
a
practical
perspective,
I'm
not
sure
I
would
use
it.
Okay,.
J
B
J
A
All
right,
okay,
so
with
that,
please
come
in
make
put
your
comments
or
your
idea
into
the
PR,
and
we
can
start
having
offline
discussions
about
that,
but
very
quickly.
Rachel,
oh,
like
you
to,
if
possible
summarize.
Your
PR
here
on
Chris
face
respects
for
proprietary
protocols
may
where
you
can
sure.
E
E
Can
do
so
by
adding
an
aspect
that
looks
like
any
other
spec
for
any
other
protocol
in
a
special
place
and
explaining
what
this
protocol
is
used
for.
The
comment
that
I
got
yesterday,
which
is
a
interesting
comment,
is
that
we
should
perhaps
have
a
higher
bar
for
proprietary
aspects
that
we
should
ask
them
to
prove
that
if
a
cloud
happens,
if
a
cloud
events
goes
in
and
then
comes
back
out,
it
is
still
in
the
same
format.
I
think
that's
a
pretty
high
bar
but
I'm
open
to
that.
E
A
B
I
read
that
comment
too
and
I
found
that
I
thought
the
arguments
fairly
convincing
and
that
is
and
I
have
I
made.
Another
comment
somewhere
else,
specifically
on
the
rocket
and
QPR
that
kind
of
goes
in
the
same
direction
and
that
is
I,
see
I.
Think
the
question
is:
is
that
spec
that
lives
in
our
repo
useful
for
anybody
outside
of
that
project?
B
E
B
B
Yeah
so
because
so
they
don't
have,
they
don't
seem
to
have
a
proper
protocol
spec,
they
literally
just
have
code,
and
so
there's
there's
nothing
that
a
binding
could
really
refer
to,
except
for
you
know,
a
version
of
the
code
and
then
there
can't
be
any
compatible
compatible
interoperable.
You
know
version
of
the
protocol
because
it's
only
documented
in
the
code.
So
then
the
question
for
me
is
what
is
the
value
of
that
map
of
the
binding?
If
that
only
refers
to
something
that
side
of
their
project,
then
it
becomes
really
just
an
effort.
A
A
And
what
about
Eric
I'm?
Here?
Okay,
jo
you,
there
Josh
Sherman
and
Lydia
lig
I
a
I'm
here!
Okay,
thank
you!
I'm,
here
age,
okay,
okay,
anybody
else,
I
missed
from
the
attendee
list,
I
think
I
got
everybody
all
right
cool.
Thank
you
guys
very
much
an
apologize
for
running
over
one
minute.
Thank
you
guys.
Next
week,.