►
From YouTube: CNCF Serverless WG 2020-07-09
Description
CNCF Serverless WG 2020-07-09
A
A
A
A
C
C
A
D
F
F
A
A
A
A
A
A
A
A
Welcome
all
right:
let's
get
started
here,
community
time,
okay,
anything
from
the
community
that
people
like
to
bring
up.
That's
not
on
the
agenda,
all
right,
I'm,
not
hearing
anybody
SDK
calls.
We
have
one
scheduled
after
this
week.
I'm
sorry
after
this
call
nothing
on
the
agenda
as
of
right
now.
So
if
you
have
anything,
please
add
to
the
agenda
doc,
otherwise,
we'll
be
very
short
call.
A
One
thing
we
probably
should
talk
about,
though,
is
we
originally
scheduled
that
to
happen
every
week
is
for
a
while
there
we
had
a
whole
flurry
of
activity
going
on
and
lots
of
things
to
talk
about,
I.
Think
if
I
remember
correctly,
the
last
two
or
three
weeks
really
haven't
had
a
whole
lot
to
talk
about
and
I'm
kind
of
wondering
whether
we
should
switch
back
to
every
other
week.
A
We
don't
need
to
discuss
that
right
here,
but
think
about
that,
and
maybe,
as
the
only
topic
for
the
SDK
call
later
on,
let's
see
Timur
is
not
there.
Anybody
from
the
SDK
I
was
right
from
the
workflow
I.
Don't
see
anybody,
so
another
thing,
I,
don't
think
is
anything
update
there
other
than
I
will
say
that
every
now
and
then
he
pings
me
to
tell
me
some
good
news
about
some
other
company
getting
involved
and
wanting
to
collaborate
with
them.
A
So
it
doesn't
seem
like
there
is
more
and
more
attention
popping
up
around
the
work
flow
spec.
So
if
you've
had
any
interest
in
the
past,
you
may
want
to
pop
that
back
on
your
radar
to
take
a
look
at
it
because
it
does
seem
like
it's
getting
more
traction
which
is
kind
of
cool
and
okay.
So
before
we
jump
to
the
pr's,
is
there
anything
on
the
agenda?
People
need
to
add
that
I
forgot
to
add
all
right.
A
I
Yeah
sure
so
this
was
something
that
I
call
member
who
raised
an
issue
for
because
the
json
schema
definition
was
not
entirely
aligned
with
the
specification
document,
so
in
the
spec
it
basically
allows
the
data
construct
to
be
any
Chason
object,
whereas
the
our
original
schema
just
had
objects
and
strings.
So
all
I
did
was
add
all
the
other
sort
of
potential
Jason
types
that
could
be
present.
I
A
D
D
A
A
Everybody
next
sear,
this
one
is
this
is
grant
grant
is
not
on
the
call
and
neither
is
Kristoff
so
okay,
I'll
try
to
summarize
this
one
as
best
I
can,
let's
see
so
grant,
was
bothered
that
the
specification
talks
about
two
different
types
of
modes,
structured
and
binary,
and
then
he
noticed
that
in
the
HTTP
spec
we
then
talked
about
batched
and
he
was
looking
at
that
as
another
mode.
So
he
was
bothered
that
there
seemed
to
be
an
inconsistency
where
were
totally
talk
about
two
valid
modes,
but
then
we
introduced
a
third
one.
A
Someplace
else
excuse
me,
so
we
thought
we
really
should
add
some
text
in
here
that
says,
oh
by
the
way
transports
can
define
other
modes.
Well.
Christoph
pointed
out
that
there's
that
batch
isn't
really
another
mode,
it's
just
a
different
flavor
of
structured.
So
what
he
added
recently
and
I,
we
helped
with
the
wording
yesterday
on
this,
is
to
make
it
clear:
that's
the
transports
themselves
may
modify
technically
either.
One
of
these
message
structures
right,
maybe
add
some
sort
of
wrapper
as
necessary
for
the
transport
or
in
the
batch
case.
A
It
can
actually
add
more
than
one
events
into
the
message
itself
it.
So
it's
a
different
flavor
of
sort
of
massaging
of
the
events
themselves
for
a
particular
transport
so,
but
either
way
you
still
have
the
basic
structure
of
it's
either
binary,
basically,
just
the
raw
data
versus
it's
a
structured
mode
where
the
caught
event
attributes
themselves
appear
inside
the
message
body,
as
opposed
to
like
HTTP,
headers
and
stuff
like
that.
Okay.
So
hopefully,
this
text
in
here
makes
it
clear
that
transports
can
define
theirs,
which
they
do
today
in
HTTP.
A
I
I
That's
that's
the
only
so,
for
instance,
when
you
we
add,
you
know
the
request
to
add
batching
or
streaming.
Is
it
then
a
requirement?
All
trans,
all
HTTP
transport
implementations,
support
those
other
modes,
or
do
they
become
optional,
that
that
was
the
only
it's
not
really
related
to
this
spec,
but
it's
related
to
all
these
sort
of
quirky
transport
things
that
have
been
popping
up
recently.
Yeah.
A
A
So
I'm
not
seeing
it
jump
out
at
me,
but
let's
do
we
take
that
one
offline,
because
I
think
that
might
be
a
good
clarification
in
general
for
some
specification
someplace.
Maybe
it's
the
main
spec
to
make
it
clear
that
this
that
it
is
optional
I
mean
the
fact
that
it's
welded
that's
made
for
the
spiking
of
the
finite,
but
it's
not
clear
whether
you
have
to
support
it.
So
you
know
you
mind
if
we
take
that
offline
and
do
it
as
a
separate
issue.
No.
A
In
the
yeah
that
we
want
wonderful,
you
can
open
that
up.
It
remind
us
to
go
back
and
double
check
that,
because
I
believe
the
intention
was
something
like
Bosch
is
not
required,
but
I
have
this
vague
recollection
that
we
do
require
them
to
support
the
in
general,
binary
and
structure
that
you
know
the
simple
version
of
those
I
do.
A
E
Yeah
so
I'm
I
think
Grant
opened
this,
partly
because
of
my
confusion
as
I
was
doing
some
really.
The
implementation,
so
I
think
I
think
my
confusion
was
still
that
there
was
there's
this.
E
A
A
E
A
A
A
Okay,
this
one
I
think
this
one's
mine.
Let's
see
if
I
were
what
I
did
here,
okay,
so
this
one
I
was
mainly
clean
up.
Basically,
I
decided
to
move
the
governance
stock,
which
was
all
by
its
the
SDK
governance
stock,
which
is
in
a
directory
all
by
itself
into
our
community
directory
to
live
it
with
all
the
other
governance
related
documentation.
For
example,
our
main
main
governance
stock
is
in
here
now
as
well.
So
I
put
this
one
in
there
and
just
call
the
SDK
governance.
A
Certainly
I
didn't
do
any
textual
changes
in
here,
I
just
renamed
the
the
PR
and
maintainer
guidelines,
because
these
are
actually
for
the
SDKs,
so
just
rename
those
to
quit
SDK
in
front
of
it
to
make
it
really
clear.
It's
about
those
and
then
the
main
readme
I,
just
added
more
information
to
actually
point
to
these
documentation
to
these
new
documents
that
are
in
there
does
it.
Obviously,
change
any
of
the
specs
just
moving
things
around
in
the
repo
more
than
anything
else,
any
questions
or
concerns
about
those.
A
Well,
maybe
we
should
make
the
name
immutable,
but
then
I
was
with
somebody
pointed
out
that
maybe
that's
not
so
great
either,
because
what
if
there's
a
typo
in
there
or
for
whatever
reason
they
need
to
change
that,
but
it
is
still
technically
the
exact
same
service.
So
that's
when
I
defaulted
after
thinking
about
it
and
said:
okay,
let's
just
introduce
an
ID,
that's
globally
unique
and
it's
a
UUID
and
it's
basically
just
uniquely
identifies
the
service.
The
value
must
not.
A
A
A
B
So
I'm
just
wondering
because
I
think
in
kubernetes
we
have
similar
concept,
that
you
have
a
name
and
your
ID
for
objects,
and
it
means
a
different
thing
slightly.
So,
even
if
you
recreate
something
with
the
same
name,
it
will
then
have
a
different
UI
Jesus
that
intended
to
be
something
similar
here.
My.
A
Intention
right,
this
was
yes.
If,
if
somebody
does
something
like
they,
they
first
have
a
service
called
foo,
and
then
they
change
its
name.
So
it's
now
called
bar
and
then
someone
else
comes
along
and
creates
another
service
called
foo,
even
though
it
shares
the
same
name
as
the
original
service.
It
is
a
complete
different
one
because
it
will
have
a
different
UUID,
so
I
think
in
conceptually
I
think
it
is
similar
to
cover
Nettie's,
but
I
definitely
did
not
model
that
after
kubernetes.
A
I
know
we're
gonna
rathole
in
this,
but
sure
why
not?
It
drives
me
nuts,
because
you
can
actually
get
into
a
weird
state
where
things
are
referenced
by
other
things,
by
fine
just
names,
sometimes
things
you
can
get
a
bad
reference
to
a
new
object
that
you
didn't
mean
to
actually
reference.
It
gets
really
really
ugly,
sometimes,
but
anyway,
back
to
this
one,
hey
are
there
questions
comments
by
anybody.
I
have.
C
C
A
A
Okay-
in
that
case,
this
one
now,
this
one
was
just
open
today,
so
we
can't
approve
it,
but
I
just
want
to
talk
to
people's
attention.
Slinky
open
this
up.
We
were
missing
some
documentation
in
the
SDK
governance.
Talk
about
how
new
maintainer
z'
get
added.
You
know
what
kind
of
voting
process
there
is
how
the
nominated
stuff
like
that.
So
this
is
his
first
initial
pass
at
it.
Obviously,
this
matters
most
to
the
SDK
folks,
so
at
least
those
people
please
review
when
you
get
a
chance,
but
obviously
anybody
in
their
wider
group.
A
If
you're
interested
go
ahead
and
review
it,
it
seems
fairly
straightforward.
We
want
to
tweak
some
of
the
wording
just
because
English
is
not
his
first
language.
Aside
from
that,
it
seems
like
it's
fairly
consistent
with
what
I
see
in
most
communities
anyway.
Take
a
look
at
that
when
you
get
a
chance.
You
also
add
a
little
section
here
about
how
to
update
this
document
itself.
I
think
that
was
missing
as
well
and
I
think
the
rules
he
defined
here
are
consistent
with
how
he
updated
the
the
main
governance
doc
itself.
A
Anyway,
any
comments
on
that:
okay
review,
that
when
you
get
a
chance
and
I
believe
gem,
we're
not
ready
to
talk
about
protobuf.
Yet
are
we
nope,
yep,
I,
think
so?
Okay
and
I
believe
Francisco
still
wants
to
hold
off
on
these
two
okay,
so
I
think
two
weeks
ago,
or
so
we
talked
about
possibly
reviewing
where
we
were
relative
to
the
implementations
of
the
discovery.
Spec
I'll
raise
my
head.
A
Yeah,
if
we
have
time
demo,
be
really
kind
of
cool
to
see
something
in
action
like
like,
like
you
in
the
last
week,
I
didn't
get
a
chance
to
actually
work
on
mine,
so
I'm
hoping
maybe
this
weekend
to
be
able
to
work
on
it
and
then
yeah
me.
We
can,
you
know,
show
each
of
our
clients
talking
to
each
other's
servers
or
something
like
that.
That'd
be
kind
of
cool
all
right.
Any
other
topics
I
mean
any
other
discussion
around
the
Interop
or
implementation
side
of
things
here.
A
A
D
A
And
then
well,
hi,
hello
and
Vinay
woopsy.
Spelled
you
wrong
I'm
here
now
you
there!
Yes,
that's
that
good,
okay
I
got
you
Doug!
Thank
you.
Did
I
miss
anybody
else
for
the
roll
call
all
right,
in
that
case
non
SDK
folks,
you
are
free
to
leave.
Thank
you
for
joining
and
we'll
start
the
SDK
call
in
just
about
a
minute
herself.
Thank
you.
Everybody.
Oh.
E
A
A
G
A
G
There
was
a
proposal
to
rename
the
SDK
from
cloud
of
incest
DK
to
just
pure
cloud
events
and
kind
of
following
some
of
the
guidelines
that
slinky
did
put
together.
For
you
know,
voting
on
other
issues
within
an
SDK
I
thought
this
might
be
a
good
issue
to
vote
on,
because
it's
been,
you
know
there
are
disagreements
about
it.
So
if
anyone
has
an
opinion
on
this,
I
would
appreciate
an
up
or
down
vote.
A
Laurie's
my
Hampton
I've
got
a
question
for
you.
Do
into
other
SDKs
have
a
similar
concept
to
like
a
module
where
they
have
the
same
kind
of
naming
question
in
front
of
them.
A
G
I
mean
I,
guess
you
know
for
each
one
of
the
SDKs,
they
all
have
their
own
unique
names.
I.
A
That's
okay,
it's
so
they
guess.
My
other
question
is
something
that
you'd
mentioned.
Obviously
breaking
backwards
back
breaking
existing
people
are
using.
It
is
obviously
a
question
is
what
is
the
consistent
or
what
is
the
pattern
that
you
see
in
the
JavaScript
community
around
stuff
like
this?
Do
they
normally
a
pen,
SDK
type
things
to
their
modules,
or
do
they
usually
not
do
that
kind
of
stuff
I?
Don't.
G
D
Mean
if
I
may,
it
happened
also
like
when
NPM
just
basically
released
their
organization
principle.
So
with
the
add
something
usually
what
people
are
doing
is
like
they
move
to
organization,
and
then
they
were
doing
like
a
cut
mud
to
automatically
update
euro
imports
and
change
your
codebase.
So
we
could
maybe
provide
a
cut
not
to
do
that,
but
basically
go
through
all
the
import
and
update
them
automatically.
D
Like
basically,
it
happened
with
the
organization,
so
sometimes
it
happens
like
that.
You
have
a
codemod
I,
don't
know
if
you
ever
used
that
one
that
allows
you
to
define
like
to
basically
introspect
the
JavaScript
and
typescript
and
like
modify
it
with
introspection
to
magically
update
all
the
imports
into
Australia.
G
D
A
G
A
Okay,
was
there
a
specific
action?
Lancer
looking
for
here
was
just
to
raise
it
to
people's
attention,
so
they
should
vote
if
they
ki
if
they
have
an
opinion,
the
latter,
okay,
okay,
cool.
Thank
you
all
right.
Thank
you
for
bringing
up
a
annoying
question,
but
something
it's
not
unexpected.
All
right,
Daniel.
E
Yeah,
this
came
from
a
discussion
and
brief
discussion
on
the
slack,
so
it
looks
like
for
a
bunch
of
languages,
maybe
most
languages
we're
releasing
libraries
to
our
package
managers
under
our
kind
of
our
personal
accounts
on
the
package
manager
and
I
wanted
to
ask
about
what
people
think
about
having
potentially
kind
of
a
cloud
events
account
that
making
multiple
people
would
have
access
to
four
for
the
package
manager
releases.
E
So
that's
you
know.
For
example,
someone
drops
off
some
a
maintainer
list
and
no,
we
can't
release
the
package
because
it's
under
their
personal
accounts
or
things
like
that.
This
is
there's
a
practice
that
we've
been
doing
for
most
of
like
Google's
library
releases,
where
we
have
group
accounts
that
belong
to
a
team
and,
and
so
anyone
within
that
team
can
release
libraries
I.
This
came
up
for
me
because
I'm
I
was
doing
the
first
release
of
a
ruby,
SDK
and
trying
to
figure
out
okay.
Well,
how
should
I
do
that?
E
E
A
So
my
hands
up
I'll
go
first
on
I
I.
Think
it's
a
good
idea.
I
honestly,
don't
know
how
else
is
the
case
or
handless
right
now,
I
think
most
men
are
probably
using
their
personal
account.
For
things
least,
that's
been
my
experience
from
what
I've,
when
I've
interacted
with
folks
and
like
they're
you
their
personal
accounts
for,
like
I,
don't
say
Travis
CI,
but
it
so
I
think
I.
Think
of
that
kind
of
stuff.
E
I
E
A
G
E
Just
to
clarify
for
for
Ruby
at
least
it
doesn't
have
to
be
a
single
identity.
You
can't
have
multiple
owners
so
that
I
guess
that
is
another
potential
approach
for
package
managers
that
support
it
just
to
have
several
people
all
like
all
the
maintainer
x'
on
the
on
the
package
under
their
personal
I'm.
Just
thinking
about
the
group
idea
is
the
formalization
of
that.
But
if
that's
not
feasible,
then
babies
is
having
multiple
owners
would
be
also
a
way
to
pay
forward.
It.
A
E
It
was
yeah,
it's
it's,
the
Ruby
SDK,
because
I
wasn't
sure
what
our
what
our
practice
is,
because
I
wanted
to
release
and
know
how
did
I
do
that,
and
so
it's
kind
of
a
question
for
the
group.
You
know:
how
are
we
doing
this,
as
you
know,
as
all
languages
for
a
cloud
events?
Is
there?
Is
there
a
common
practice?
So
are
we
just
kind
of
all
doing
our
own
thing?
C
B
G
G
You
know
lots
and
lots
of
tooling
that
requires
you
know
access
to
from,
but
like
somebody
has
to
own
that
and
manage
that.
So
where
do
you
stop
the
common
thing?
Is
it
just
when
you
publish
to
like
a
repository
or
something
like
that?
Or
is
it
all
the
little
bits
and
pieces
that
go
into
actually
having
a
you
know,
CI.
A
Well
it.
It
seems
to
me
that
if
anybody
of
me-
and
if
anybody
on
these
teams,
who
sort
of
manages
one
aspect
of
the
tooling
decides
to
walk
away
for
whatever
reason,
we
don't
want
to
be
an
alert
right
and
so
I
mean
I,
agree
with
what
Mark
is
saying:
I
just
I'm
just
trying
to
figure
out
the
mechanics
of
it
right
because
I'm
like
what
Daniel
is
saying
where
they
have
a
central
place
inside
Google,
where
they
can
do
things
like
share
passwords
and
stuff.
A
We
don't
have
that
I'm,
just
wondering
how
we
actually
get
to
where
I
think
we
want
to
be
right
without
us,
without
a
shared
data
store.
There's.
A
What
do
people
think
about
that
setting
up
a
private
repo
that
only
certain
people
have
access
to,
and
we
just
stick
passwords
and
information
like
that
in
there
I
mean
I,
guess
that
that
is
one
way
to
go.
I
is
that
secure
enough
for
people
I
mean
I,
know
you're
not
taking.
This
was
a
check
stuff
into
github,
but
as
long
as
it's
a
prior
repo
is
that
okay,
I.
G
C
A
A
G
I
mean
I'll
say
that
I
was
actually
really
concerned
when
I
started
to
get
more
heavily
involved
in
the
JavaScript
SDK
and
Fabio
was
not
around
much
and
I.
Didn't
know
how
I
would
be
able
to
get
access
to
things
like
the
code
coverage
reports
and
stuff
like
that.
Fortunately,
we
managed
to
connect
on
that
and.
G
You
know
do
that,
but
it
didn't
seem
like
there
was
any
clear
you
know
common
way
to
handle
this
sort
of
stuff
in
it,
and
that
was
disconcerting
when
I
first
started
working
on
it,
yeah.
A
E
I
I
think
it's
you
know
if
you
can
find
out.
If
there
are
any
tools
available
the
great
if,
if
not,
if
it
turns
out
that
that
there
isn't,
then
maybe
someone
I'm
happy
to
do
it
could
just
kind
of
think
about
a
little
bit
right
up,
write
up
a
proposal
for
something,
maybe
simple
and
straightforward
to
start
off
with,
and
you
can
vote
on
it
or
something.
Okay,.
A
G
A
A
Okay,
I'll
tell
you
what
I'll,
let
me
paste
that
they
paste
the
topic
and
the
proposal
to
go
back
to
every
other
week
in
the
SDK
slack
channel
and
if
I
don't
hear
any
objections.
Well,
we'll
do
it
I
guess:
lazy.
Consensus
is
the
phrase.
Otherwise,
if
someone
objects
then
we'll
find
out.
Why?
Because
obviously
they're
not
on
the
call,
so
they
probably
won't
object.
So,
okay
I'll
do
that.