►
From YouTube: CNCF Serverless Working Group 2020-04-02
Description
CNCF Serverless Working Group 2020-04-02
B
A
A
E
F
F
I
A
Yeah
I
guess
in
this
day,
where
most
people
are
doing
video
conferencing.
Now
it's
hard
to
tell
whether
something's
going
wrong
it's
because
you
forgot
the
unmute
or
it's
a
problem
with
your
laptop
or
a
problem
with
the
WebEx
system
itself
with
them.
You
know
the
VC
system
itself.
It's
too
many
things
to
go
wrong
in
that
thread.
K
A
A
G
M
A
It
would
help
I
come
off
mute
as
far
as
I
know.
Everything
seems
to
be
going.
Okay,
I
mean
yeah,
but
everybody
at
home.
It's
hard
to
know
exactly
how
bad
it
is.
Then,
what
you
may
hear
on
the
news
and
the
news
doesn't
sound.
You
know
that
bad
occasionally
hear
about
stuff,
but
you
know
I
mean
there's
still
food
in
the
grocery
stores,
not
honest
of
everything,
but
it
doesn't
seem
too
bad.
C
A
It's
interesting
that
you
mention
that,
because
I
was
hearing
something
on
the
news
yesterday,
it
was
saying
that
I
don't
care
of
it.
They
were
talking
specifically
about
Seattle
or
just
Washington
in
general,
where
it
seemed
like
their
curve,
was
either
plateaued
or
to
starting
to
go
down
or
something
like
that,
because
it
wasn't
they
making
the
news
as
much
anymore
after
that
nursing
home
yeah,
we
stopped
testing
well.
That
will
do
it
too.
C
A
C
A
N
G
A
A
I,
don't
know
sometimes
it's
questionable
yeah.
This
is
the
dangerous
part
about
me.
Sharing
my
screen.
You
get
to
see
all
my
wonderful
typing
errors
and
it's
scary,
all
right,
hey
Google,
the
spelling
of
that
please
mark
humor,
I
love
it:
okay,
beautiful,
let's
get
on
to
the
fun
stuff:
okay,
Klaus
there,
yes,
I'm
there,
okay
Lancey
there.
A
L
A
G
A
Somebody
but
see
hi,
that's
Dustin!
Probably
oh
man
well
got
it.
Okay,
anybody
else,
all
right,
let's
jump
into
it.
Thank
you
guys
appreciate
that
alright
status,
we
have
no
action
items.
That's
a
nice,
not
exciting,
woohoo,
alright,
community
time
anything
from
the
community.
People
want
to
bring
up.
That
is
not
an
agenda.
Thomas
yeah.
B
B
A
Much
easier
answer:
we
don't
version
them
at
all.
That's
meant
to
Mesa
kaliba
a
moment,
the
wiki
page,
more
than
anything
else
in
the
sense
that
anybody
can
add
an
extension
to
the
list
as
long
as
it
here's
to
ever
criteria
we
have,
if
we
even
have
any
and
that
and
they
can
get
updated
anytime,
it's
just
a
spot
for
people
to
put
things
to
share
with
the
rest
of
the
community
and
they
have
no
official
standing
whatsoever.
A
G
P
A
Good
question
because
it's
been
a
while
and
we
have
a
lot
of
newcomers,
so
normally
what
happens
is
new
pr's
or
PRS
that
go
through
a
significant
change
have
to
be
updated,
were
created
by
about
Tuesday
evening
that
way,
people
have
at
least
two
days
to
review
it
before
this
phone
call.
If
they're
updated
in
that
shorter
window
than
that,
then
we
may
bring
it
up
on
this
call,
but
we're
not
going
to
do
any
kind
of
official
vote
on
it,
because
people
need
more
time
to
review
it.
A
A
Thursday's
call
with
the
hopes
that
during
the
week
between
now
and
next
Thursday
people
will
review
it
offline
and
you'll
be
able
to
make
comments,
because
it
would
be
kind
of
mean
for
people
to
wait
until
next
Thursday's
phone
call
to
mention
a
whole
bunch
of
nitpicky
things.
When
you
could
have
made
those
edits,
you
know
before
Tuesday
does
that
help
totally
works?
Thank
you.
Okay,
thanks
for
question,
all
right
anything
else
for
the
community.
A
I
was
gonna:
ask
you
about
that.
I,
wasn't
sure
how
you
wanted
to
proceed
on
that,
because
both
I
think
all
three
of
those
are
tagged
as
still
a
work
in
progress,
and
normally
we
won't
bring
up,
pour
requests
on
the
phone
call
unless
they're
basically
ready
to
go
because
we're
usually
short
on
time
with
just
a
one-hour,
and
that's
that's
why
I
brought
the
mobile
in
last
week's
call,
because
I
knew
we'd
have
extra
time
not
to
talk
about
those.
So
how
would
you
like
to
proceed
on
those?
A
N
I
want
to
proceed
first,
having
a
try
and
try
to
create
a
draft
implementation,
because
if
we
without
an
implementation
with
them,
we're
not
sure
where
we
are
going
so
I
want
to
first
try
to
implement
it
and
I
want
to
present
an
implementation
show
how
it
works.
If
people
likes
how
it
works,
I
can
go
through
this
back
and
fix
it,
but
I
think
that
that
depends
on
so
also
on
the
fact
that
we
want
to
add
this
in
an
order:
optional,
HP,
multiple
events.
N
A
A
Okay:
well:
what
if
we
do
this?
What
if
we
periodically
remind
people
on
this
phone
call
about
the
PRS
out
there?
It's
just
so
you
get
some
initial
feedback,
but
then
you
get
closer
to
feeling
that
your
implementation
is
is
good
enough
for
people
to
actually
start
playing
with
it.
Why
don't
we
at
that
point
point
people
to
the
implementation,
and
then
we
can
raise
the
PR
issue
back
up
to
the
forefront
and
and
possibly
put
it
on
the
agenda
for
a
more
formal
discussion.
How
does
that
sound?
That's.
N
C
A
Just
just
a
process
thing
would
you
became
like
in
that
question
to
the
PR
itself.
Oh
I
already
have
oh
okay,
okay,
because
because
I'd
rather
not
dive
into
a
deep
discussion
about
that
PR
here,
only
because
that's
not
the
normal
process
we
follow.
I
was
okay
with
bring
it
up
just
to
draw
people
to
finish.
Do
I,
so
so
Francesco
relative
to
your
question
about
merging
the
stuff,
because
I
get
the
sense.
That
may
not
be
a
quick
answer.
Even
though
I
asked
for
the
group's
feedback.
A
N
A
Cool
all
right
anything
else
for
the
community
time.
Okay,
next
up,
SDK
call
I,
don't
think
we
have
anything
to
discuss
there,
but
we
do
have
a
call
after
this
one.
So
stick
on
the
line,
if
you
guys
want
to
join
that
team
or
I,
don't
see
him
online,
no
he's
not
online,
so
I,
don't
think,
there's
anything
up
day
from
the
work
flow
group
out
there.
So
I
was
gonna,
say,
though
no
I,
don't
think,
there's
anything
to
say:
okay
koukin.
A
So
hopefully
everybody
knows
that
they've
rescheduled
for
August
13th
to
the
16th,
which
is
over
a
weekend
which
is
going
to
be
interesting.
My
question
for
you
guys
is:
do
we
want
to
still
plan
on
having
a
face-to-face
meeting
there
I
know
a
lot
of
that
is
gonna
be
dependent
on
how
many
people
actually
plan
on
going
and
I
suspect.
Most
people
don't
know
for
sure
yet.
But
let
me
ask
a
general
question:
if
it
looks
like
the
virus
stuff
is
you
know,
tapering
off
and
people
are
starting
to
travel
again.
E
G
A
It's
just
too
early
to
say
right
now:
yeah,
okay,
okay,
well,
tell
you
what,
since
I
would
hate
to
say
no
to
those
folks
and
then
lose
the
room.
Let
me
go
ahead
and
say
yes
and
I'll
make
sure
they
understand
that
we
may
end
up
dropping
it,
but
at
least
then
we
will
have
them
available
if
it
turns
out
a
whole
bunch
of
people
do
go
yeah,
so,
okay,
I'll,
just
I'll
just
hold
on
to
the
room.
Okay,.
E
E
A
Yep,
okay,
okay,
yep,
okay,
will
do.
Thank
you,
everybody
all
right.
It's
Charter!
First
up
as
I
mentioned
I
believe
last
week
we
were
working
on
a
charter
and
I
made
it
available.
Hopefully,
everybody
saw
the
emails
that
went
out
about
it.
I
think
it's
some
review
comments.
Thank
you
for
everybody
that
did
that
in
terms
of
process.
I
believe
the
next
step
is
for
me
to
open
up
a
pull
request
to
the
TOC
to
formally
request
the
news
say,
get
created
at
that
point
in
time.
A
I'm
sure
they'll
have
lots
of
commentary
about
the
Charter.
So
do
not
assume
that
this
is
the
final
version.
I
suspect
it
will
get
changed
based
upon
feedback.
We
get
from
people.
Having
said
that,
though,
hopefully
people
had
a
chance
to
review
it,
and
if
you
had
any
major
issues
you
would
have
already
raised
it
and
so
I'm
hoping
everybody
on
this
call
basically
says
yep,
it's
okay
to
send
the
PR.
A
But
let
me
just
ask
the
formal
question:
is
there
any
concern
about
the
general
direction
of
the
PR
that
we've
warranted
holding
off?
I'm?
Sorry
any
concern
with
the
Charter
as
proposed
that
would
warrant
it
not
getting
PR
to
relatively
soon,
because,
obviously,
some
minor
things
can
be
adjusted
later
through
the
PR
to
internal
PR
process.
Anybody
have
any
concern
with
what's
in
there
today,.
C
A
L
A
C
C
C
M
C
A
A
K
A
So
we
already
had
the
PR
for
this
one.
No,
so
we
don't
need
worry
about
that
one,
these
two
as
of
right
now
correct
me
if
I'm
wrong
everybody,
but
I,
think
our
current
agreement
was
as
of
right
now
we're
going
to
keep
them
under
the
cloud
events
project
and
at
some
point
in
the
future.
If
we
think
either
they
deserve
their
own
project
or
it
gets
too
big
to
be
under
one
umbrella,
then
we
were
going
to
think
about
splitting
it
out,
but
we
didn't
want
to
make
that
decision.
Prematurely
is
the
way
I.
P
A
A
Okay,
as
I
said,
don't
assume
this
is
written
in
stone.
We
can
still
make
edits
this
through
normal
PR
process,
but
I
will
now
move
forward
with
PR
in
this
to
the
TOC
and
trying
to
get
on
their
agenda.
I
suspect
that
I
have
to
put
together
at
least
a
slide
or
two
presentation
on
this
when
I
get
that
ready,
I'll
share
it
with
you
with
you
folks,
so
you
could
review
it.
I,
don't
expect
it
to
be
very
large
or
anything
like
that,
but
I
will
want
your
feedback.
P
A
A
F
E
So
one
of
the
unfortunate
parts
of
the
open
API
of
open
API
per
se
is
that
it
is
biased
towards
the
path
and
that's
where
you
anchor
everything
so
I
need
to
go.
So
there
was
a
there
was
a
there
was
desire
expressed
to
also
have
the
the
IDE
in
the
path
for
the
get
operation
and
that's
I'm,
indifferent,
whether
that's
a
query
string
or
whether
it's
it's
in
the
past,
but
because
I'm
in
difference,
I,
don't
think
having
the
the
IDE
and
the
query.
String
is
worse
than
having
it
in
path.
E
I
think
both
of
those
are
just
officers
that
are
ok,
so
I
think
I
will
have
to
go.
Do
a
little
and
refactoring
to
go
and
affect
the
factor
out
the
operation
per
se
and
then
I
hooked
that
up
into
two
places
and
since
I
haven't
done
that
and
done
that
part.
But
I
I
don't
want
to
right
the
one
of
the
other
way
and,
as
I
said
in
my
in
this
comment,
fit
in
particular,
you
know
having
a
question
mark
ID
equals
one
or
subscription
slash.
E
E
A
Q
E
E
E
E
E
G
A
E
A
A
Proprietary
specs
right
Jerry,
so
this
was
from
Jim
and
I,
don't
believe
he
was
able
to
make
the
call
today.
I
think
we
actually
review
this
twice
already
now,
but
he
went
through
some
wording.
Changes
based
upon
some
people's
suggestions.
I'll,
give
you
guys
a
chance
just
to
quickly.
Look
it
over
I,
don't
think!
There's
anything
controversial
here!
This
is
just
additional
information
or
people
want
to
add
their
own
proprietary
binding.
A
C
See
I
threatened
to
submit
this
last
week
after
we
met
this,
takes
the
the
single
the
single
query,
API
and
breaks
it
down
into
more
restful
api.
So
there
are
different
entities
now
provider
type
and
producer.
They
have
their
own
paths
and
resource
IDs
and
query
strings
for
search.
So
I
think
this
is
a
much
more
restful.
Api
I,
certainly
like
it
better
I,
have
I
saw
a
dog
play
a
bunch
of
comments
in
last
night
and.
M
C
C
A
Okay,
thank
you
for
doing
the
work
on
here
and
I
apologize
for
waiting
until
you
know
Wednesday
evening
to
my
comments.
However,
I
will
say
that
it
would
be
really
nice
if
everybody
else
got
a
chance
to
review
this,
they
would
I.
Don't
it
disturbs
me
that
I
was
the
only
one
that
comment
that
I'm
there,
because
I'm
pretty
sure
given
a
local?
Some
of
you
guys
are
that
you
have
plenty
of
things
to
say
and
as
soon
people
were
just
busy
so
I'd
rather
not.
A
You
know,
think
about
merging
this
thing
until
at
least
we
have
evidence
that
other
people
have
reviewed
it.
So
please,
when
you
get
a
chance,
look
it
over
yep
cool,
Thank,
You,
Ryan,
okay,
but
at
this
point,
if
people
have
to
have
done
a
quick
scan,
is
there
anything
at
the
high
level
people
like
to
bring
up
on
the
call
right
now?
Otherwise,
we'll
just
take
the
reviews.
Offline.
C
A
E
It
does
and
what
why
key
value
matter
and
that
this
that
the
specification
really
is
about
mapping
the
value
of
the
key
and
the
header
section
of
the
Kafka
message
and
then
so
I.
That's
some
context,
setting
that
I
thought
was
missing
and
then
I
specific,
an
and
leaves
the
rest
of
the
document
alone,
because
that's
actually
talking
about
the
value
which
is
not
introduced
a
reference.
E
E
Then
it
was
really
meant
so
I'm
just
simplifying
this
and
saying
the
key
of
the
Kafka
message,
and
so
that's
now
really
just
the
key
that
you're
setting
if
you're
sending
a
from
a
message
by
an
API
that
is
populated
by
key
mapper
function
and
that
might
and
is
not
further
defined
here.
It's
what
that
is.
It's
in
really
implementation,
specific.
E
But
then
every
implementation
is
free
to
provide
a
callback
or
you
know
whatever
the
mechanism
is
to
go
and
take
something
from
the
message
or
take
something
from
the
environment.
You
know
create
a
some
compute,
some
some
compute
function,
which
goes
and
takes
extracts
pieces
from
the
payload,
whatever
else
and
just
generates
a
key
and
then
goes
and
sensitive
as
the
key.
That's
what
the
abstraction
the
key
metric
extraction
should
be,
but
the
key
point
is
that
key
is
not
part
of
the
cloud
event.
E
N
Yeah
so
first
question
is
in
the
SDK.
Should
we
treat
this
change
as
a
part
of
the
one-point
toes
back
so
when
I
parks,
the
message
should
I
do
that
in
the
101.
One
point
goes
back.
That's
super
good.
First.
Question
second
question
is
the
partition
key
after
its
it's
used
as
the
key
of
the
kafka
messages
should
be
also
written
as
an
extension
attribute
sure,
so
it
should
be
double
the
message.
N
E
This
is
what
this
is.
What
was
what
I
think
what
the
original
intent
was
so
I
see
this
as
a
clarification
rather
than
a
change,
so
the
key
attribute
was
always
meant
to
be
really
the
key
of
the
kafka
of
the
kafka
message.
I
think
that
was
that
was
a
mistake
and
because
it
may
it
actually
never
made
sense
for
this
to
be
a
key
attribute
if
the
Kafka,
if
the
Kafka
API
doesn't
know
about
this,
so
that
was
just
that.
E
I
think
that
was
just
honest
mistake
and
then
the
and
then
the
the
text
at
blind
132
is,
unfortunately,
was
wishy-washy
in
you
know
it
should
use
the
extension
value.
The
only
value
that's
defined
in
that
spec
is
the
partition
key,
and
so
therefore,
if
you
are
creatively
reading
this
text,
then
you
would
figure
out
that
you
know
the
value
of
that
extension
is
the
value
you
should
use
for
that
key
and
so
I
think.
E
Q
H
H
E
A
A
Of
course
sure
yeah
this
one
was
too
new
to
merge
today
anyway,
so
I
was
gonna,
give
people
another
week
to
review
it
but
I,
but
since
you
were
the
one
that
raised
the
additional
the
original
issue,
I
want
to
make
sure
that
you're,
okay
with
it.
So
that's
good,
excellent,
okay!
So
everybody
please
look
this
over
and
we'll
look
to
merge
in
next
week,
assuming
that
the
implementation
test
of
it
actually
works,
no
need.
We
did
that
for
all
of
ours
for
all
issues,
all
right.
A
Okay,
before
we
jump
to
this
issue
here
and
between
the
graph,
q1
I,
just
make
sure
I'm,
not
forgetting
anything.
Are
there
any
topics
that
I'm
missing
that
are
on
the
long
lines
of
you
know
we
need
to
make
a
decision
on
something
like
that.
You
know
need
to
be
done
immediately.
Otherwise,
we'll
jump
to
this
might
still
make
sure.
I
might
be
mrs.
anything
since
we're
in
the
free
time
space.
Now.
C
So
I
think,
a
couple
weeks
ago
we
were
talking
about
the
design
for
discovery.
Api
and
Jim
had
suggested
that
we
look
at
grabs.
You,
though,
so
I
did
that
when
and
implemented
a
implementation
of
the
API
yeah
I
started
from
knowing
like
nothing
about
graph
QL
I
found
it's
really
nice
to
work
with
I
think
it
actually
solves
the
problem
quite
elegantly
because
of
the
relationships
we're
trying
to
model
here
in
thinking
about
different
ways
that
somebody
might
come
at
discovery
and
how
they
might
work
very
so.
C
C
C
C
Sorry,
that's
the
results
from
that
group
geez.
So
the
next
one
be
like
okay.
So
like
one
of
the
things
with
storage
unit,
that's
what
I'm
interested
in
so
you're
able
to
say.
Okay,
give
me
some
more
details
right.
So
it's
that
same
thing,
I'm
still
asking
for
providers
but
I'm,
giving
in
a
query,
string
and
I'm.
C
Saying:
okay
now
expand
that
to
the
types
that
are
available
so
I
can
see
that
I've
got
two
different
blob
storage
providers
that
are
providing
the
same
event
types
and
if
you
go
back
to
the
bug
one
more
time,
I
think
the
next
query
down
there
is
the
okay
I
want
to
subscribe
to
events
from
that
cloud.
Storage
provider
for
the
I
think
create
event
type.
What
do
I
need
right
so
being
able
to
query
through
that
I.
C
Did
not
tackle
subscription,
API
I
think
we
would
need
to
have
a
unified
graph
QL
schema
for
listing
subscriptions
and
creating
subscriptions.
There
is
a
mutate
API
here.
Other
concerns
like
AWS
and
threading
through
identity
is
totally
doable
with
the
the
context
and
the
way
graph
QL
queries
are
executed.
C
Q
So
this
is
awesome,
thanks
for
putting
that
example.
Together,
I
think
it's
always
helpful
to
see
something
working,
I
really
like
graph
QL,
but
just
to
play
devil's
advocate
a
little
bit.
I
am
asking
myself
what
value
it
actually
provides
on
top
of
the
HTTP
API
and
it's
a
different
question,
because
we're
both
introducing
a
dependency,
a
technology
and
something
that
lots
of
folks
might
not
be
familiar
with
so
I
just
want
to
make
sure
we're
being
judicious
about
what
we
introduce
and
make
sure
it's
actually
solving
a
problem.
C
Yes,
all
right
so
I
think
so
from
a
consumer
perspective,
I've
heard
a
lot
from
like
I
spoke
with
some
thoughts
and
the
Google
developer
relations
org,
and
they
were
actually
really
excited
to
see
this
because
they
say
like
they're
developers,
that
they
talk
with
really
like
graph
QL
api's
and
find
them
easy
to
work
with.
So
remember,
I'm
a
client
perspective.
That's
at
least
one
data
point
in
the
positive
as
far
as
the
server
libraries',
so
I
linked
into
the
bug.
C
C
A
Okay,
so
my
hands
up
next
I'm
Christoph
I'll,
go
to
you
so
I
I
have
some
other
concerns
too.
What
Ryan
mentioned,
which
was
you
know,
becoming
dependent
on
one
particular
technology,
even
though,
obviously,
no
matter
what
we
pick
or
depend
on
something,
and
at
least
rest
or
HCP
seems
to
be
one
that
it
Barry's
comfortable
with
and
graph
QL
being
depend
on.
That
worries
me
a
little
from
that
perspective,
but
my
question
for
you
Ryan
is:
do
you
see
this
as
replacing?
A
C
J
I
Yeah,
maybe
I
can
give
some
insights
to
graph
girl
usage,
so
we
so
I
had
a
colleague
who
was
at
the
conference
for
Facebook
announced
graph
go
and
he
was
instantly
sold
and
he
hardcore
pushed
adoption
on
it.
So
we
started
using
it
internally,
also
making
a
public
and
for
the
first
times
we
didn't,
have
had
customers
interested
in
not
a
lot
of
usage,
but
over
the
last
year
or
two.
I
We
are
still
in
a
place
and
time
very
offer
both
for
everything.
We
do
well,
I.
Think
and
let's
say
two
years
ahead.
I
could
absolutely
see
us
offering
new
features
on
in
graphical
if
they
had
option
for
us
goes
on
like
this
and
yeah
as
far
as
using
the
technology
for
us,
it's
real
its
young
technology.
We
don't
have
all
the
technology
in
place
to
do
all
the
tracking,
so
there's
some
things
to
be
relearned,
but
it
definitely
has
its
nice
adventures
also
from
an
api
provider
perspective.
In
my
opinion,
yeah.
I
E
Well,
a
graphical
operates
over
a
database,
and
what
we're
doing
here
is
we're
trying
to
we'll
ultimately
trying
to
build
a
Federation,
which
means
which
means
I
think
this
is
great
as
a
layer
that
allows
you
to
go
and
navigate
a
space,
but
we
need
to
go
and
get
those
stuff
together
and
we
need
to
go
and
and
effectively
create
a
Federation
of
how
we
get
that.
How
do
we
get
that
data
into
that
space
that
we're?
How
do
how
are
those
arrays,
actually
populated?
E
And
so
while
this
is
a
great
way
to
go
and
navigate
the
graph,
you
have
to
go
and
assemble
the
graph
first,
so
I
think
I
think
the
discovery,
the
discovery
and
and
and
publishing
mechanism
that
we
need
to
that.
We
need
to
find
here
and
also
the
subscription
aid
in
the
subscription
mechanism
needs
to
need
to
allow
a
publisher
someone
who
produces
events
to
make
those
known.
E
It
also
needs
to
be
able
to
go
and
have
a
effectively
let
them
go
and
produce
a
very
simple
list
of
events
that
they're
accomplishing
and
and
have
that
in
a
retrievable
and
then
similarily
for
subscription.
Ultimately,
you
need
to
go
and
walk
up
to
a
particular
party
and
say
this
is
the
subscription
I
want
to
go
and
set
up,
and
then
this
is
the
subscription
I
want
to
go
in
and
and
create.
Those
are
ultimately
very
simple
crowd.
E
Just
to
make
known
that
it
publishes
three
and
to
four
to
have
an
external
party
go
and
set
up.
A
requests
of
the
nets
can
be
that
can
be
done
with
a
very
simple
HTTP
POST
for
to
us.
Ultimately-
and
ultimately,
this
is
not.
You
know,
I,
find
it
a
little
strange
that,
because
to
portray
this
as
rest
versus
graphic
UML,
because
graphical
is
obviously
a
uses,
all
the
foundations
that
are
sitting
on
at
the
bottom
of
HTTP,
it's
just
layering
or
something
on
top
of
it.
J
I
just
wanted
to
mention
two
things
that
given
that
graph
QL
is
a
query
language.
First
of
all,
it's
actually
work.
You
don't
need
a
client
SDK
to
make
graph
Kyoko's.
You
can
simply
send
them
a
specific
class
and
parse.
The
response
is
JSON
which
and
has
a
great
thing
about
graph
Gale
compared
to
some
other
query.
Languages
like
in
a
way
er,
PC
or
table
of
RPC
and
others
and
I
also
wanted.
J
Maybe
questioned
the
statements
that
graph
QL
requires
a
database
behind
the
graphical
source
because
usually
or
it's
not
a
requirement
and
you
can
run
graph
kill
on
top
of
some
data
were
some
response.
It
can
also
be
used.
Some
other
service
call
or
something,
but
I
don't
feel
that
database
is
a
good
requirement
for
graphical
or
for
something
behind
graph
kill.
A
A
E
Think
we
need
to
I
think
we
need
to
have
a
simple
crowd.
Api
that
also
allows
you
to
get
their
lists
of
things.
That
is,
that
is
simple
to
implement
for
everybody,
because
graph
QL
critically
isn't
simple
to
implement
for
everybody,
and
then
it's
only
for,
for
you
know,
constrains
in
in
terms
of
how
much
load
you
can
take
on
for
for
dependencies
right,
because
graph
QL
is
easy
to
use.
But
it's
certainly
not
a
trivial,
a
trivial
dependency
to
take
on.
E
And
so
that's
something
that
I'm
I'm
worried
about.
I'm
worried
about
that
from
from
a
basic
perspective
about
having
too
much
weight
in
terms
of
a
of
a
dependency
that
you
have
to
implement
to
be
conforming
with
the
specification
and
then
there
certainly
there's
the
lack
of
the
luxury
of
having
a
powerful
query
language
and
it's
something
that
it
has
great
value
and
as
I
think
that's
a
layer
that
we
can
certainly
go
and
put
on
top
specifically
for
the
discovery.
Api.
But
I
think
that
it's
got
the
the
way.
E
A
C
I
just
want
to
address
this,
be
alright.
I
think
the
database
argument
is
an
invalid
one
to
build
to
build
the
REST
API.
You
have
the
same
issue
right.
The
data
has
to
live
somewhere
either
I'm
implementing
you
know
some
kind
of
SQL
query
or
some
kind
of
you
know
SQL
fetch
to
find
out
sources
that
I
have.
C
The
difference
is
like.
Am
I
implementing
the
HTTP
handler
or
am
I
getting
a
call
back
from
a
resolve
function?
So
there's
a
like
I.
Don't
think?
That's
that's
a
valid
argument.
You
have
the
same
thing
to
implement
in
both
in
both
systems.
As
far
as
making
making
events
known,
we
actually
haven't
addressed
that
yet
in
discovery
at
all,
so
we
haven't,
we
haven't
talked
about
how
my
an
IOT
device
make
itself
known
I.
If
that's
something
we
think
we
should
explore.
I.
E
Would
I
would
assume
that
if
you
allow
discovery,
then
you
also
need
to
figure
out
how
to
populate
the
discovery
database
I
mean
they
die
and
I
mean
database
really
in
the
in
in
the
conceptual
functional
sense
of
it.
I
don't
think
of
it
as
an
installed
thing,
but
it's
it's
a
you
walk
up
to
the
discovery
service
and
the
discovery
service
has
some
data
that
it
operates
on
how
that
cell
doesn't
matter.
But
ultimately,
if
you
have,
if
you
have
parties
which
are
raising
events,
we
want
to
make
those
discoverable.
E
So
those
need
to
contribute
something
in
to
discover
into
the
pool
that
is,
that
provides
the
discovery
service
and
that
flow
is
obviously
also
important.
I
mean
that's
that
I
think
from
an
interoperability
perspective
that
might
actually
be
the
most
important
part
of
it,
because
we
can
have
400
ways
of
how
to
do
discovery
and
I
might
want
to
go
discover
using
a
sequel,
dialect
or
graph
QL
or
whatever,
but
I
think
from
an
interoperability
perspective.
E
What's
most
important
is
that
I
have
a
device
or
have
have
someone
who
has
a
subscription
manager
implementation
and
offers
events
up
for
subscription?
How
does
that
party
make
itself
known
to
a
service
which
provides
discovery?
So
if
we
don't
have
that,
then
we
certainly
need
that
yeah.
We
don't
have
that
yet.
A
I
Wait
maybe
it's
for
me:
Groff
killers
are
not
a
language
on
top
size,
just
the
filter,
so
Everett
forth
or
the
filter
will
languish
on
top,
like
all
the
features
and
concepts
and
semantics
behind
it
are
independent
of
it,
and
you
have
to
have
them
both
for
graphical
and
rest,
whatever
like.
If
you
talk
about
something
English
or
Spanish
or
German,
doesn't
matter
like
the
concept
below
it,
they
are
independent
of
language
and
I
feel
the
same
three
or
four
rest
aggressor.
I
A
Yeah,
okay,
but
that
I
think
we
got
to
call
time.
However,
obviously
we're
not
gonna
resolve
this
today.
I
think
it's
gonna
be
a
big
discussion.
So
please
everybody.
If
you
get
a
chance
comment
on
the
PR
itself
or
there's
an
issue.
Yeah
sorry
comment
on
the
issue
itself,
so
we
can
get
some
discussions
going
offline.
There.
Obviously
I
think
this
has
to
come
back
up
again,
so
I'm
sure
we'll
revisit
it
next
week
or
in
future
calls,
but
it
does
sound
like
it's
a
very
important
decision
point
for
us
going
forward.
A
A
A
All
right,
in
that
case,
I,
believe
we
are
done
for
today.
Thank
you,
everybody,
a
lot
of
good
progress
and
just
reminder
we
do
have
the
SDK
call
starting
right
out
to
this.
So
in
about
a
minute
or
two
we'll
kick
that
off
everybody
else,
you're
free
to
go
into
other
esterday
everybody.
Thank
you
right.
M
D
D
A
A
D
A
A
I'm
convinced
I'm,
saying
it
wrong,
I
mean
I'm
questioning
that
I
was
just
wondering
with
I,
spelled
it
right.
So,
okay,
yeah
I'm
missing
somebody
on
this
list,
two
four,
six,
eight
ten,
twelve
okay,
somebody
dropped.
Okay
with
that
when
we
get
started
so
Scott's.
Since
you
made
that
giggled
noise
arlier,
do
you
have
any
topics?
C
C
We've
decided
that
the
client
API
exposes
it
returns,
something
always
and
that
thing
can
be
tested
to
be
an
acronym
kind
of
makes
sense
for
transactional
sends
and
receives,
and
requests
and
stuff.
So
we're
still
testing
integration
inside
a
que
native
to
make
sure
the
SDK
does
things
that
a
real
production
ready
integration
would
do,
but
it's
looking
pretty
good
excellent
and
I
was
looking
at
metrics.
Let's
see
here,
we've
had
tracing.
N
C
No
I
was
looking
at
the
traffic
on
the
bat
repository
and
a
hundred
and
twenty-five
thousand
clones
in
its
city
and
how
many
125,000
clones
that's
freaking,
incredible
thing:
5k
unique
cloners
a
day
in
the
last
several
days.
It's
just
amazing,
though
I
think
we've
been
a
part
of
some
very
big
CI
CD
system.
A
C
C
A
N
N
If
anybody
is
interested
to
discuss
about
later,
at
least
if
we
can
find
who
should
discuss
about
me,
I'll,
be
really
happy
to
this
kind
of
discussion.
I
explained
what,
in
my
opinion,
is
wrong
now
and
how
I
try
to
fix
it
back.
This
is
like
a
rework
of
of
a
good
part
of
the
SDK,
so
it's
a
big,
bigger
work
so
have.
A
N
S
Yeah
I'd
be
interested
in
talking
about
this,
because
I
posted
a
similar
issue
on
the
Python,
SDK
and
I.
Think
I.
Just
look
at
this
now
I've.
Some
of
the
issues
that
I
had
with
the
pipe
current
status
at
the
Python,
SDK
or
kind
of
similar
as
well
and
so
I
was
just
gently
curious
about
also
maintenance
of
the
pipe
on
SDK.
It
doesn't
have
0.3
or
1.0
versions
of
the
spec
yet
either.
Although.
A
N
A
So
this
is
something
we've
never
actually
talked
about,
so
I
think
you
know.
Obviously
we
I
think
we
need
to
discuss
this.
So,
let's
take
at
his
example,
just
to
pick
up
Java
one,
let's
say
Fabio
for
some
reason
decides
he
doesn't
care
anymore,
maybe
his
vanishes.
Obviously
we
can
make
anybody.
We
want
to
be
maintainer.
A
My
from
from
my
perspective,
my
only
concern
would
be
that
it
would
have
to
be
somebody
that
we
know
is
there
and
has
the
best
interest
of
that
project
at
heart,
right,
I,
don't
I
wouldn't
want
to
like
I,
wouldn't
want
some
random
person
who
just
shows
up
one
day
said:
hey
Fabio
is
vanished.
Here's
a
PR.
Can
you
merge
it
for
me
and
we
just
blindly
merge
it
right?
We
need
somebody
who
who
we
actually
have
some
faith
in
and
and
some
experience
with.
A
So,
in
your
particular
case,
Francesco,
you
know
even
without
reviewing
the
PR
I,
would
trust
your
judgment
based
upon
your
current
workings
in
the
go
SDK
to
say,
yeah,
you're,
probably
making
the
right
choices,
and
you
can
become
the
new
leader
of
that
that
project.
But
it's
not
always
gonna,
be
that
clear.
I
fear
right.
N
A
But,
but
if
it's
on
the
same
vein
right,
because
if
if
the
current
maintainer
Zoar
a
project
are
either
too
busy
to
do
something,
we
need
some
way
to
adjust
that
problem,
because
these
PRS
should
not
linger
forever.
I
know
I'm
open
to
ideas
here
in
terms
of
how
to
move
forward
here,
you
may
need
to
deal
with
in
AI
case-by-case
basis.
You
know
in
a
Java
particular
case,
you
say
if
Fabio
is
just
too
busy
to
look
at
PRS
and
to
comment
on
the
one
that
you
just
open.
A
N
C
N
C
E
Yeah,
so
that
actually
plays
into
the
the
point
that
I
raised
two
weeks
ago.
You
remember
the
about
designee,
which
I
haven't
solved
yet
also
because
I
was
sick
but
I
erased
about
the
code-signing
about
code-signing,
which
is
that
that
particular
issue.
But
the
question
there
is:
what
is
the
neutral?
E
What
does
the
release
pipeline
look
like
because
I
think,
ultimately,
if
we're
having
a
loose
Federation
year
of
folks
who
are
working,
who
are
working
on
a
bunch
of
as
the
case
but
ultimately
the
deliverables
that
we
have
are
binaries,
then
then
we
need
to
have
a
way.
We
need
to
have
a
neutral
pipeline
that
just
doesn't
publish.
N
C
E
C
C
Yeah
I'm
gonna
continue
looking
that,
because
I'm
doing
github
actions
for
something
else
that
I'm
working
on
so
possibly
we
could
do.
We
kind
of
apply
a
your
project.
The
new
bar
is
your
project
needs
to
have
some
sort
of
release
automation
if
it's
more
complicated
than
just
pushing
the
release
button
in
github,
yeah.
N
C
A
C
A
N
C
C
A
F
A
A
What
that
write
up
is
and
then
basically
agree
to
that
contract
in
essence
and
then
once
we
agree
to
that,
then
we
can
go,
look
at
every
single
SDK
and
apply
that
list
to
that
periodic
review
and
there
and
do
the
kick
and
kick
off
a
periodic
review
of
every
single
SDK.
And
if
some
of
them
are
meeting
those
things,
then
we
send
threatening
letters
conceptually
yeah.
C
A
A
Cuz
it
would,
it
looks
it's
kind
of
like
you
know.
I
was
a
good
or
bad
example,
but
the
Apple
Store
right
well,
some
of
them
might
call
them
dictators
they're,
trying
to
make
sure
that
people
don't
get
a
bad
experience,
because
even
if
it's
the
apps
fault,
Apple
looks
bad
because
they're
the
platform
and
I
think
cloud
events
is
the
platform
here
in
the
sense
right.
We
don't
want
to
SDK
out
there.
That
makes
us
look
bad
yeah.
A
It's
a
dead
rusty
car
in
our
long
right,
exactly:
okay,
okay,
so
I'll
take
the
action
item
to
to
come
up
with
a
rough
draft
of
the
list
of
requirements,
rules
charter,
what
guidelines
for
reviews
where
they're
going
to
call
it
and
then
you
guys
can
review
it
and
then
Scott.
You
took
the
action
item
for
the
get
up
action
stuff.
So
thank
you
for
that.
Yeah.