►
From YouTube: CNCF Serverless WG 2020-07-16
Description
CNCF Serverless WG 2020-07-16
A
B
A
Good
morning,
hey
Tommy,
Daniel,
hello,
oh,
is
this
your
first
time
on
the
call
I
apologize?
If
it's
not
I
was
here
last
week,
okay
did
I
get
your
company
affiliation
I,
though
there
was
somebody
I
missed.
Last
week,
oh
you're
Google
and
they're
mine,
like
I,
got
that
okay
slinky
how's
it
going
fine.
C
D
F
A
A
I
E
A
A
A
All
right
move
forward,
so
we
talked
in
the
last
week's
SDK
call
about
switching
back
to
having
a
call
every
other
week,
since
we
seem
to
be
running
out
of
agenda
items
and
I
posted
a
message
in
the
slack
channel
like
I
said
I
would,
and
there
was
no
objection.
So
unless
someone
raises
objection
now,
what
we're
gonna
do
is
switch
to
to
go
back
to
every
other
week
starting
next
week.
Any
objection
to
that
all
right.
Moving
forward.
All
right,
workflow
got
some
exciting
news
here:
you're
up
team.
F
Thanks
yeah,
the
exciting
news
is,
we
got
actually
accepted.
There's
a
sandbox
project
by
the
TOC
yesterday
sure
do
days
ago.
So
we've
been
busy,
we
got
our
github
organization
and
I'll
pay
paste.
The
link
I
mean
in
the
chat
and
other
news
are
we
have
a
Java
SDK?
They
have
their
own
repos.
Now
we
kind
of
copied
with
you
guys,
do
elements
of
course,
and
we're
working
on
again
moving
our
repo
structure
over
as
it
has
changed.
F
You
know
two
times
in
the
last
couple
months,
other
than
that
we're
going
on
through
an
onboarding
process
right
now,
which
I've
never
been
through.
So
there's
a
lot
of
little
things
that
TLC
is
looking
at,
making
sure
that
we
have
and
yeah
that
that's
yeah
and
also
one
more
thing.
We
changed
their
meetings
from
once
a
month
to
twice
a
month,
so
our
next
meeting
will
be
next
Monday.
F
A
F
A
F
Well,
he
asked
me
in
contact
with
him
and
he
actually
asked
me.
They
apparently
have
some
sort
of
booth
demo,
videos
for
cube
called
in
August
and
I.
Don't
really
know
what
that
means
to
be
honest,
but
if
they
were
looking
for
like
10
to
15
minute
videos,
they
can
just
play
on
some
sort
of
booth
type
of
setting
and
I
was
wondering
if
we
do
that
for
several
is
workflow.
We
could
also
add
the
cloud
events
in
there
as
well
and
I'm
sure
there
will
be
fine
with
it.
F
F
A
Okay
now,
turning
any
so,
it's
gonna
jump
into
that,
because
I
got
that
simpler
email,
like
you
got
about
this
and
I
wasn't
quite
sure
what
they
meant
by
having
a
booth,
because
I
thought
everything
was
virtual.
Well,
apparently,
they're
gonna
have
virtual
booths
as
well
and
know
that
glad
you
joined
Clemens,
because
I
was
a
lot
of
pinging
on
this
too,
since
you're
not
both
presenting
there.
A
So
basically,
there
are
three
different
options:
I
guess,
four
options.
Fourth
being
not
going
to
anything
at
all,
but
their
three
options
are
we.
You
know
we
want
to
actually
have
a
booth,
which
is
basically
is
a
virtual
booth.
Like
you
see
here
on
the
screen
and
you
can
have
slides,
you
can
do
videos,
you
can
have
downloadable
material
and
basically
I
think
that
the
point
here
is
that
people
can
just
virtually
walk
up,
see
a
video,
that's
playing
constantly
or
download
material
right.
So
it's
not
nearly
as
as
interactive
I.
A
I
A
I
So
the
the
second
option,
I
think
Matt's-
will
probably
do
the
second
option
just
because
you
have
more
control
over
it
and
it's
a
little
bit
more
focused
I.
Think
I'm
not
did.
This
is
all
new
for
us
as
well.
This
is
the
first
time
we've
done
anything
like
this.
We're
also
concerned.
If
we
did
a
booth
about
how
much
content
we
would
actually
have
to
put
up
there
and
and
I
don't
know
how
much
feedback
we
could
get
from
it.
A
Okay
and
then
I
see,
the
third
option
is
to
do
a
combination
of
both,
and
so
my
question
for
the
group
here
was:
is
there
any
interest
in
doing
either
of
these
two
options
or
three
options?
I?
Guess,
because
it
is,
if
you
look
down
here
now,
it's
a
little
a
small
print,
but
I
want
an
each
one.
If
in
online
there
is
a
fair
amount
of
time,
commitment,
I,
don't
know
if
you
have
to
necessarily
do
the
entire,
so
all
the
hours
listed
for
for
all
the
days
or
just
a
subset.
A
But
it
is
a
commitment
from
folks
and
I
know.
That's
not
being
there
physically
makes
it
harder
to
make
a
commitment
because
you
get
pulled
off
on
emails
and
other
phone
calls
and
stuff.
So
if
people
do
want
to
do
this,
then
I
think
they
have
to
be
prepared
to
make
it
a
real
live
commitment
and
not
to
say
y'all.
Do
it
and
then
not
show
up
god
rather
say
no,
then
happy
people
be
disappointed
us
and
not
meeting
our
objectives
and.
A
Point,
yes,
okay,
so
any
questions
about
what's
going
on
here
before
I,
ask
the
question
of:
do
people
want
to
do
or
not
people
understand
what
we're
sort
of
talking
about
here
and
here's
the
list
of
materials
and
stuff
these
for
the
booth?
Okay,
so
let
me
ask:
do
we
have
anybody
who
wants
to
who
think
this
is
a
good
idea
and
is
going
to
volunteer
to
sign
up
for
either
of
these
two
options?
I
believe
I'll
have
to
let
them
know
by
the
end
of
today
or
tomorrow.
It's
a
fairly
quick
decision.
A
I'm
torn
I
gotta
be
honest
with
you,
I
I
kind
of
like
the
idea
of
having
something
just
to
keep
our
name
out
there,
but
I
honestly,
gotta,
say
I,
don't
know
if
I
had
the
time
to
put
into
it
to
make
it
worthwhile
I
mean
I.
Might
I
might
be
able
do
an
hour
here
or
there,
but
if,
if
it's
that
random
I'm
not
sure
how
we're
gonna
gets,
you
know
the
impact
we
you
want.
Take
it
worthwhile
I.
J
I
Feel
like
we're
a
guinea
pig
versus
the
MDF
or
a
lot
right
now,
so
I
think
and
I
have
not
gotten
confirmation
from
Katie.
Yet
I
think
we
have
to
tell
her
today
if
we're
gonna,
do
anything
my
assumption
for
the
zoom
so
option
two
would
be
that
we
would
be
able
to,
and
this
again
an
assumption
we
would
be
able
to
pick
times
and
then
would
be
on
the
schedule
like
so
for
the
physical
one.
I
When
we
had
a
maintainer
booth,
we
could
tell
them
when
we
were
gonna
staff
it
and
it
was
on
the
schedule.
So
my
assumption
is
it
will
be
the
same
type
of
thing,
so
you
can
say
we're
gonna
have
this
zoom
call
or
whatever
from
this
time
to
this
time,
and
they
would
be
on
the
schedule.
I,
don't
think
we
have
to
have
an
open
zoom
call
for
that
whole
entire,
extended
time,
I,
don't
think,
but
again
we'll
find
out.
I
can
respond
and
let
you
know
what
Katie
comes
back
with.
Okay.
I
A
A
A
Okay,
tell
you
what
I
don't
mind:
volunteering
I'll
raise
my
hand
to
do
an
hour.
Maybe
I
could
probably
get
maybe
doing
an
hour
a
day,
especially
since
it
means
pretty
early
in
the
morning
here,
which
means
it
won't
overlap.
Another
call
I,
don't
mind
we
get
up
early
and
so
tell
you
what
so
ginger
will
pick
it
back
on
your
idea.
I'll
send
them
a
note
saying
if
possible,
we'd
like
to
do
option
two,
we
just
don't
know
the
exact
hours
yet,
but
chances
are.
I
A
Okay,
now
going
back
to
Tim
or
your
question:
what's
your
thoughts
on
this?
Did
you
want
to
do
something
obvious
now
that
you're
a
full
fledged
project?
Actually
there,
let
me
back
up.
I
was
kind
of
assuming
that
we
would
do
this.
I
said
ginger.
Do
you
know
is
this
for
this?
Is
this
for
projects
or
working
groups-
I'm
guessing
it
might
just
be
for
projects
as.
I
A
A
F
F
My
initial
thought.
Maybe
we
can
join
forces
together
to
kind
of
create
something
together.
If
not,
then
that's
fine
as
well,
whatever
you
guys
decide
but
yeah.
If
you
guys
want
to
put
together
like
I,
pulled
two
videos
to
make
him
have
something
longer
or
whatever.
That's
fine
with
me:
I'm
open,
okay,
I'm
sure
we're
open
for
everything.
A
A
Yeah
I'll
ask
too
cuz
because
I
think
if
we
can
merge
the
two,
it
may
be
easier
for
us
to
man.
The
zoom
call
right,
because
I
think
we
could
probably
get
people
on
our
side
to
at
least
answer
some
very
high-level
questions
about
the
workflows
back
and
likewise
the
workflow
folks
could
probably
answer
some
cloud
events
question.
A
If
they
really
had
to
I
know,
you
guys
are
Chinese
kind
of
enter
than
the
covers
as
well,
so
in
terms
of
getting
coverage
for
more
hours,
I
think
combining
forces
might
be
better,
but
only
if
they're,
okay
with
that,
okay,
so
I'll
reach
out
to
them,
and
we
can
talk
offline,
but
I,
think
we
have
to
get
an
answer
back
today
or
tomorrow
at
the
latest
kind
of
a
thing:
okay,
okay,
okay,
any
other
questions
or
comments
on
that
topic,
all
right
before
I
jump
into
PRS
and
stuff.
Are
there
any
other
topics?
F
I'm
sorry
Doug
me
again,
I
thought
so
much
in
this
meeting.
I
just
wanted
to
kind
of
tell
everybody
and
again
it's
this
redhead
thing.
We
have
these
things
called
tech
Tuesdays,
it's
some
sort
of
video
that
can
be
from
thirty
to
one
hour
long
and
it's
an
interview
style
thing
where
people
just
go
on
there
and
and
learn
about
some
different
technologies,
and
there
is
opportunities
in
August.
F
A
A
D
C
A
Okay
sounds
good
perfect.
Any
any
I
guess
I
should
ask
any
questions
at
this
point
in
time
for
Francesca
this
may
have
mana
bring
anything
up
or
do
I
just
take
it
off.
Fine,
okay,
I'm
gonna,
hear
any
questions
so
we'll
do
it
offline?
Okay,
this
one
very
easy
just
didn't,
merge
it,
because
no
one
buffets
that
and
lgt
em.
I
just
noticed
we
missed
through
missing
the
PHP
and
rust
SDKs
to
read
me
any
objection
to
adding
those
all
right.
I
didn't
think
so,
obviously
feel
free
to
lgt
em
easy
stuff
like
this.
A
Alright,
this
one
okay,
so
unless
we
call
I
introduced
the
idea
of
adding
a
unique
identifier
per
service,
that
is
immutable
so
that
subsequent
query
you
can
know
whether
this
is
the
exact
same
service
or
not,
because
maybe
everything
metadata
has
changed.
But
obviously,
if
this
one
field
can't
change-
and
you
know
it
is
the
exact
same
thing
just
with
different
metadata
now
Klaus-
you
asked
a
question
about
down
here.
This
is
the
interesting
one.
A
K
A
I
want
to
make
sure
I
understand
the
scenario
so
I
think
what
you're
saying
is.
Let's
say
you
have
a
a
discovery
endpoint
that
people
turn
it
directly,
but
then
that
same
discovery
endpoint
is
available
through
something
like
an
aggregator
or
something
like
that.
Yes,
would
the
same
service
UUID
appear
through
both
portals,
basically
or
through
both
endpoints?
Yes,
yeah
anybody
have
an
opinion
on
that.
A
H
A
A
It
wasn't
so
much
because
the
name
thing
occurs
of
an
aggregation.
It
was
more
a
realization
that
anybody
may
have
made
a
typo
when
they
added
this
service
entry
and
I
guess
I
was
looking
through
all
the
various
attributes.
I
didn't
see
one
that
jumped
out
at
me.
That
said,
it
doesn't
matter.
If
you
made
a
typo,
you
cannot
change
this
because
it's
gonna,
it's
gonna
mess
with
people.
Everything
seemed
like
it
was
we're
in
the
risk
of
it
may
change
over
time.
A
I'm
right,
you
may
add,
or
remove
types
you
may
have
done
a
typo
in
the
name.
You
may
have
done
a
type
of
one
end
in
the
description.
I
couldn't
find
anything
that
didn't
follow.
The
category
of
someone
may
have
goofed
right
and
they
don't
want
to.
They
don't
want
their
users
to
view
this
as
a
brand
new
service
right.
They
this
want
to
update
the
metadata
about
at
the
existing
one
and
that's
and
that's
why
I
came
up
with
a
need
for
something.
A
K
G
K
K
G
Basically,
I'm
agreeing
to
it
that
would
close
that
it
was
just,
but
thinking
about
you
had
something
good
Clause
created
with
a
preset.
We
need
to
probably
moles
in
errors
such
as
communicative,
so
to
be
very
helpful,
especially
as
I've
explained
that
we
don't
have
specified
something
that
can't
be
changed.
A
A
I'm
so
I'm
torn
here,
because
I
do
think
this
is
needed
so
I.
Obviously
that's
why
I
opened
the
PR,
but
the
fact
that
I'm
not
here
speak
up
makes
me
a
little
bit
nervous
about
whether
I'm
the
only
one
I
don't
want
to
add
something.
If
I'm,
the
one
of
the
things
is
needed
all
right,
all
right,
pooping
sava,
because
they
think
yeah
it's
needed
and
you're
okay
with
it
or
do
not
people
just
aren't
sure
I
will
start
picking
on
people
if
I
have
to
and
to
get
some
opinions
on
this.
A
J
B
I
I
haven't
met,
I
have
not
thought
about
it,
I
think
I.
Think
given
where
we
are
today.
My
bias
is
sort
of
with
Klaus
that
a
you
know,
I
agree
with
your
initial
impetus
that
I
think
we
need
something
to
deal
with
those
whatever
you
are
call
it
sort
of
life
cycle
conflicts
right,
whether
it's
a
typo
or
other
thing,
I
just
haven't
thought
through
the
implications
of
you
know.
How
is
this
actually
going
to
play
through
so
I
guess
my
cop-out
is
I.
I
would
rather
have
more
time.
B
A
Fair
enough,
so
we
had
at
least
one
request
for
more
time,
so
why
don't
we
go
in
and
and
do
that
I
would
like
to
see
we
can
come
up
with
some
kind
of
decision
by
next
week's
calls.
I
think
this
one
has
been
out
there
for
two
or
three
weeks
now
and
I.
Don't
think
this
is
a
huge,
controversial
thing,
but
I
think
it's
more
a
question
of
people
needing
to
take
the
time
to
look
it
over
and
think
about
it.
Some
in
the
meantime,
maybe
Klaus
I'll
reach
out
to
you
offline
to
see.
A
H
H
A
A
A
A
E
H
E
A
So
you
use
for
sure
reach
a
URL
okay.
So
to
me,
the
URL
wasn't
able
to
be
used
as
a
unique
identifier,
because
people
could
move
where
they're
hosting
services
hosted
right.
So,
for
example,
maybe,
instead
of
today
at
being
a
URL
ibm.com
in
it.
Maybe
it's
discovery
endpoint
that
IBM
comm
right,
but
it
is
still
the
same
service,
that's
being
deployed,
it's
just
we're
just
hosting
it
at
a
different
endpoint
right.
A
A
But
basically,
when
you
do
a
query
you
can
specify
did
with
the
two-
and
this
will
query
is
something
like
this:
we
specifying
the
number
of
records
you
want
to
get
per
chunk
size,
it
was
100
and
then
the
response
could
include
somebody
headers
with
links
that
tells
you,
for
example,
how
to
get
the
next
set
or
the
previous
one.
If
you
wanted
now,
one
thing
I
did
do
was
how
did
the
option
for
the
server
to
return
and
it
expires
and
the
reason
I
did.
A
This
was
because
I
didn't
want
to
assume
that
at
this
spec
level,
I
see
this
so
at
the
spec
level.
I
didn't
want
to
mandate
whether
the
that,
whether
the
data
that
could
change
in
the
server
could
change
or
not
right,
I
wanted.
You
be
I,
wanted
the
server
to
be
able
to,
in
essence,
support
both
meaning
when
they
get
the
initial
query
up
here.
Do
they
create
in
essence,
a
result
set
for
that
particular
query
like
it's
persisted
off
to
the
side
that
way,
if
the?
A
If
the
result,
if
the
data
in
the
back-end
server
changes
the
results
that
becomes
static
right
and
in
order
to
do
that,
I
wanted
the
server
to
be
able
to
return
some
information
that
says
how
long
is
this
data
good?
For
that
way,
the
client
knows
they
better
retrieve
all
the
data
before
that
expires
time.
Otherwise
it
may
go
away
without
that,
then
you're
back
to
the
case
where
the
data
and
server
doesn't
change
very
often,
so
you
don't
need
to
worry
about
this
situation,
and
everything
could
be
kind
of
stateless
right.
A
A
Now,
that's
not
to
say
that
there
couldn't
be
other
information
included
in
here
that
that's
opaque
to
the
client,
meaning,
let's
say,
for
example,
the
server
is
in
this
scenario,
where
it's
a
static,
I'm,
sorry,
not
a
static,
a
temporary
data
set
that
gets
generated,
but
he
needs
some
sort
of
unique
identifier
on
each
subsequent
request
to
identify
which
results
set
to
that
is
being
queried
over
right.
So
they
may
include
not
just
a
size,
query
parameter,
but
maybe
some
sort
of
identifier
in
there
and
that's
okay.
The
client
doesn't
need
to
know
about
it.
A
Doesn't
you
to
care
about
it?
This
link
is
completely
opaque
to
them.
They
don't
know
anything
about
it
whatsoever.
The
only
thing
they
need
to
know
is
to
use
it,
and
if
this
expires
time
is
here,
they
need
to
finish
their
results.
That
retrieval
before
the
expires.
Time
happens.
Okay,
so
really
the
Holly
we
care
about
is
just
whether
it
exists
or
not,
and
not
right.
So
you
understand
much
more
about
it.
J
A
A
We
get
to
it
and
if
you're
quick,
you
don't
ever
need
to
even
think
about
the
expires
right,
but
if,
if
for
some
reason
you
let
you
know
that
you're
very,
very
slow
and
doing
the
retrieval
the
expires
time
here
tells
you
well
you
better
not
be
too
slow
because
that
results
that
may
go
away
at
that
expires
time.
There's.
J
J
Effectively
for
interacting
with
caches,
if
you,
if
you
use
page
forward,
you
might
be,
might
want
to
go
and
and
page
back,
and
so
your
HTTP
browser
knows
about
this
and
the
servers
and
about
this.
So
if
you
are,
if
you
have
a
ages
and
you
page
forward
and
any
page
back,
then
the
server
may
go
and
give
you
a
well
I.
Don't
have
anything,
that's
mother
modified
and
gives
you
a
status.
J
J
70
to
72
32
has
all
the
preconditions
and
last
modified
and
etags,
etc.
All
that
all
those
things
so
I
think
what
you're
trying
to
model
here
is
already
in
that
spec.
It
doesn't
end
up
in
the
in
the
your
I,
but
it
ends
up
in
headers,
but
I
think
that's
mostly
what
the
but
what
you're?
Looking
for?
Okay.
A
J
J
G
A
Okay,
I
will
definitely
take
a
look
at
that
course.
Thank
you
and
guess:
there's
nothing
more
discuss
on
this
one.
Then
we
go
off.
Take
a
look
at
that
RFC
and
see
if
we
can
just
lower
that
puppy
yep
all
right.
Any
other
hello
questions
comments
all
right
cool.
Thank
you,
Clements
for
the
pointer
all
right
now.
This
one
was
just
open,
I!
Think
yesterday,
so
it's
probably
too
soon.
I
was
definitely
too
soon
to
to
accept
and
Thomas
is
even
here,
oh
cool.
Thank
you.
Thank
you
class.
A
L
Yes,
I
can
I
we
talked
about
it
I
think
three
weeks
before
so.
The
main
one
issue
is
that
in
the
HTTP
protocol
binding,
we
talked
about
content
modes
and
we
have
two
modes
called
binary
and
structured,
and
they
have
exactly
the
same
name
as
the
two
message
modes.
We
define
in
the
spec,
and
that
is
very
confusing
because
one
assumes
they
are
the
same
thing
but
they're
actually
not,
but
it's
obviously
very
confusing
because
they
have
the
same
name.
So
I
try
to
be,
and
they
obviously
relate
to
each
other.
L
So
in
the
below
the
message
notes.
The
upper
part
I
said
that
for
the
binary
mode
message
usually
they're
just
used
as
they
are
and
transport
protocol
doesn't
do
much,
but
for
the
structured
role,
message,
they're,
often
embedded
and
then
I
added
some
more
details
below
what
it
embedding
can
mean.
So
sometimes
there
is
more
things
added
as
a
wrapper
around
the
message.
So
we
see
this
especially
with
Drazen,
so
you
may
add
other
things
or
you
may
batch
several
messages
together.
L
L
J
Yeah
yeah,
actually
it's
not
even
true
it
for
Kafka,
it
is
the
value
of
a
key
value
pair,
which
makes
up
a
record
and
then
480p,
it's
the
it's
the
body
of
the
Inc
P
message,
but
there's
no
envelope
there.
Okay,
if
I
sit
envelope,
I,
think
of
of
soap,
for
instance,
as
a
abstraction
that
sits
between
the
transport
container
or
a
transport
body
and
the
the
payload,
and
that
we
have
no
case.
J
I
find
sentence
153,
it
actually
could
is
conflating
is.
Could
that
is
really
conflating
two
things
so,
for
instance,
so
you
never
use
it
once
you
use
binary,
then
you
never.
You
don't
use
an
event
format
ever
because
you're
taking
the
payload
of
the
event,
and
you
met
that
straight
to
whatever
the
payload
section
is
in
your
transport
mess.
But
the
event
format
doesn't
play
a
role
here.
A
A
Just
because
the
grant
opened
the
issue
at
least
I
think
grant
opened
it
I
think
it
was
on
behalf
of
somebody
else
at
Google
and
they
were
confused
by
some
of
the
terminology
were
using,
so
maybe
a
help.
If
you
get
that
background
first,
then
to
better
understand
what
the
what
the
desired
goal
is.
What's
that
way,
okay,.
J
A
A
Anything
on
these
I
know
Gemma's
in
here.
Francesco
assume
these
are
still
on
hold
right,
mm-hmm,
okay,
all
right
so
interrupts
so
I
know,
Remy
was
working
on
implementation.
I
have
been
as
well
and
Remy
said
he
could
make
the
call
today
and
from
my
side
be
honest.
I
haven't
had
a
chance
to
do
anything
for
two
weeks
to
work
on
this.
So
there's
no
changes.
A
A
Okay,
not
hearing
any
okay,
so
obviously
you
know
we
will
try
to
keep
going
in
the
background
and
when
people
do
have
some
of
the
thing
that's
runnable
will
have
to
do
some
sort
of
interrupt
testing
or
something
when
those
lines
we'll
see
how
it
goes.
But,
at
least
from
my
point
of
view,
it
is
helping
to
flesh
out
some
issues
like
that
ID
thing
that
I
brought
up
earlier:
okay,
I'm,
technically
that's
the
end
of
agenda.
Are
there
other
topics?
People
would
like
to
bring
up.