►
From YouTube: CNCF Serverless WG Meeting - 2019-03-07
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
B
Alright,
it's
three
after
we're
gonna
get
started:
let's
eat
I'm
Safari,
alright,
I,
don't
see
Rachel
on
the
calls.
I
can't
nag
her
about
her
AI.
Look
into
that
offline,
alright,
ok,
community
time,
I,
don't
see
I've
been
on
the
call,
so
I'm
not
quite
sure
what
to
do.
With
his
suggestion
for
a
topic,
then
I
think
the
same
person
who
may
have
pinged
the
group
during
a
slack
session
asking
for
people
who
may
want
to
present
that
Ukrainian
conferences.
B
You
may
want
to
look
for
him
on
slack
if
you're
interested
in
doing
that.
So
just
mentioning
that
lets
moving
forward,
SDK
workgroup
I,
don't
believe
anything
is
going
on
there
other
than
if
you
are
part
of
the
group,
that's
working
on
SDK.
We
are
hoping
to
do
some
sort
of
interoperability,
demo
showcase
thingy
at
pukani,
you
but
I
think
the
biggest
hurdle
in
doing
that
is
we
want
people
to
try
to
show,
inter
alia,
across
all
the
various
versions
of
the
spec,
not
just
the
very
latest
version.
B
So
we
may
need
a
lot
of
updates
from
the
SDK
authors
out
there.
So
please,
if
you
guys,
do
have
an
SDK.
Please
mention
your
interest
in
participating
in
that
in
the
SDK
slack
Channel,
so
we
can
start
organizing
something
there
Scott
or
mark.
Can
you
guys
think
of
anything
else,
SDK
or
later
than
are
forgetting
to
mention.
D
D
I
think
I
can
take
a
pass
at
it
and
then
maybe
Doug
can
fill
it
in,
but
we
we
had
a
meeting
with
Heathrow
or
Heathrow
was
in
attendance
and
we're
still
trying
to
figure
out
exactly
how
arcus
and
the
airport
can
be
worked
in,
so
that
vendors
can
participate
in
a
demo.
That's
actually
shows
what
we're
trying
to
show
at
a
like
a
a
coop
con
like
event,
so
we're
still
trying
to
figure
out
exactly
what
the
demo
should
be.
D
E
E
Specific
conditions
and
actions
that
would
be
part
of
the
workflow
and
it
the
the
main
guy
it
was
all
about
being
role
based
I
think,
is
what
we
agreed
on,
so
that
the
attendees
could
select
a
particular
role
that
they
wanted
to
play
in
a
collaborative
workflow
process.
That
involved
some
manual
tasks,
tasks
that
the
participants
would
perform,
combined
with
automation
through
micro
services
and
and
and
all
of
that
would
all
the
interactions
would
be
through.
B
Okay,
any
questions
on
that
all
right
cool.
So
thank
you.
Duggan
looking
forward
to
reading
the
northeast
end
last
night
and
then
the
malays
phone
call
sounds
like
it
should
be
fun
all
right.
Moving
forward,
kakan
you
the
last
time
we
had
a
conversation
or
a
meeting
about
that.
Most
people
were
busy
with
other
stuff,
so
there
has
been
a
lot
of
progress
there,
but
people
aren't
stressing
too
much
yet
so
we
decided
to
take
a
little
bit
of
break
and
come
back
in
a
couple
weeks.
B
The
one
that
I
will
point
out,
though,
is
there's
work
planning
on
being
done
up
in
the
white
paper
and
landscape,
Docs
service
working
group,
I.
Think
Scott
and
two
other
people
came
here
who
are
table
leads
on
that.
So,
but
if
you
are
interested
in
making
updates
to
those
docks,
please
reach
out
to
me
or
or
Scott
and
we'll
get
you
in
contact
with
the
other
folks.
So
we
can
coordinate
the
activities
there.
All
right,
let's
see
nothing
kook
on
China
are
then
I
request
the
sessions.
B
E
F
F
In
addition,
so
there's
that
the
content
type
which
says
this
is
what
is
in
there
and
the
data
content.
Encoding
now
says-
and
this
is
how
this
is
encoded.
So
if
it's
a
string
based
encoding
as
it
is
with
Jason,
you
say
this
is
basically
four
encoded
binary.
You
basically
Claire
with
that
that
the
data
field,
if
it's
a
string,
then
you
know
it's
the
basics
before
encoded
binary.
That's
what
that's
for,
and
it's
optional
means
you
only
provided
when
and
I
I
just
missed,
missed.
F
B
So,
as
we
talked
about
obviously
there's
some
syntactical
things
that
need
to
be
changed
here,
but
from
a
semantic
perspective,
what
do
people
on
the
call
think
about
this?
It
sound
like
it's
headed
in
the
right
direction
and
I
didn't
see
any
complaints
about
it
in
the
PR
itself
over
the
last
two
weeks,
or
so.
D
B
B
F
This
is
effectively
the
an
explanation
of
something
that
we
haven't
had
yet
in
the
spec
and
that's
how
the
this
text
set
is
actually
layered
on
top
of
each
other.
So
there's
the
base
spec.
We
have
extensions
that
go
and
layer
on
top
of
the
base,
spec,
there's
format,
encodings
and
then
we
have
the
transport
bindings
and
then
effectively.
F
F
D
B
B
B
I
believe
he
said
it
in
the
general
direction
of
saying
that
ID
and
source
are
the
unique
aspects
of
cloud
events
and
in
particular,
I,
was
talking
about
how,
if
those
are
the
same,
the
receiver
can
then
shoot
them
as
duplicates
if
it
wants
to
decent
to
something
logic,
while
their
discussion
is
still
going
on,
I
wanted
to
get
a
sense
from
the
group
here
in
terms
of
what
you
guys
thought
about
this
general
direction.
I
know
there.
B
B
B
G
C
B
B
B
C
B
B
Okay,
in
that
case,
I
wish
Christophe
was
here,
is
I
really
want
to
talk
about
his
minnow
supported
thing.
B
B
Taking
my
chair
offers
sec
I
think
this
is
a
fairly
serious
change.
The
spec
meant
by
serious
I,
don't
mean
good
or
bad.
Just
it's
a
is
a
significant
change
and
I
would
really
really
like
people
to
take
a
look
at
it
to
make
sure
that
they're,
okay
with
this
direction
or
if
they
have
concerns
to
raise
those
concerns,
I
I,
don't
want
this
one
to
be:
silence,
equals
consent
kind
of
thing.
This
is
this.
This
is
a
pretty
big
change
in
my
opinion,
but
maybe
I'm
wrong.
B
H
H
G
B
I
The
idea
is
that
he
wants
to
kind
of
guarantee
to
a
user,
that's
submitting
a
event
in
their
kind
of
native
format
that
when
it
gets
then
serialized
and
sent
across
the
wire,
it
will
successfully
proceed
through
a
series
of
middleware
and
end
up
at
the
end
consumer,
predictably,
and
and
having
some
some
Headroom.
There
helps
ensure
that
if
the
encoding
increases
the
bit
does
okay
thanks.
I
B
F
Still
think
it's
an
extreme
amount
of
effort
for
the
the
reading
coding
case
and
also
now
now
you're.
Actually
you
with
the
single
events
you
can
violate.
You
know,
I,
don't
know
how
many
I
haven't
counted
the
rules,
but
there's
a
lot
of
rules.
You
can
now
violate
that.
You
all
need
to
be
aware
of
of
the
quotas
that
you
have,
which
could
be
a
little
frustrating.
G
F
But
I
prefer
that
risk
to
do
all
that
math
and
having
to
enforce
20
rules,
which
could
be
pretty
frustrating
if
you
just
have
that
one
long
string
field
and
have
a
monstrous
longer
eye,
which
probably
even
includes
a
token
or
whatever,
whatever
we
need
there,
and
then
you
can't
do
that
because
you
know,
even
though
you
are
under
the
limit
for
the
event
size,
that
one
particular
field
is
constrained.
So
all
of
these,
because
you
know
the
limit
the
limit
is
always
the
limit
we
set
here
is
arbitrary
of
64k.
B
B
B
B
I
was
wondering:
okay,
based
upon
the
comments
in
this
issue,
I
think
between
Evan,
gem
and
least
one
other
person
in
there
I
put
together
this
very
rough
outline
of
proposal
just
to
get
the
ball
rolling
this
and
then
Eric
and
I
you
you
wanted.
Maybe
do
some
wording
Smith
wordsmith
on
it,
but
I
was
wondering
what
people
thought
about
this
general
direction.
G
B
B
All
right
is
there.
Anybody
who
wants
to
volunteer
to
turn
this
into
a
formal
pub
core
request
I'll
do
that
nobody
else
wants
to
excellent.
Thank
you.
Jim
I
appreciate
that.
G
B
B
B
J
J
Middle
Baron
can
somehow
modify
event,
attributes
and
and
or
if
there
are
any
rules
for
this
and
well
over
the
discussion
you
can
see
this
has
been
opened
already
in
August.
So
it's
an
old
issue
already
I
think
we
are
the
people
who
disgusts
kind
of
agreed
that
there
are
a
lot
of
cases
and
it's
different
difficult
to
define
very
strict
rules
on
this.
J
But
that
may
be
some
note
and
the
primer
would
be
good
to
emphasize
that
there
are
certain
restrictions
on
the
attributes
like,
for
example,
what
we
had
a
few
minutes
ago
with
sauce
and
ID.
That
should
also
be
kept
when
someone
in
the
middle
is
somehow
modifying
attributes
or
maybe
another
one.
If
the
time
attribute
is
updated,
like
Evan
posted,
then
it
should
be
assumed
that
it's
a
new
event
and
then
also
the
ID
should
be
another
one.
Things
like
this:
okay.
K
B
B
B
Next,
one
I
think
this
one
actually
Eric
this
year,
yeah,
okay,
yeah!
We
don't
need
the
cover
this
one
right,
Eric.
This
is
already
carved
by
the
other
one.
We
talked
about
agreed,
okay,
cool!
Thank
you.
Let's
see,
okay,
this
one
did
you
do
all
right,
Adam
this
one's
yours?
Would
you
like
to
introduce
this
issue.
L
Sure
so,
to
give
some
background,
I
work
on
the
Canada
of
eventing
beast,
which
uses
cloud
events
as
kind
of
the
Interop
between
all
of
our
different
servers
and
I.
Looked
at
the
spec
written
up
a
curl
command
that
followed
it
exactly
and
then
realized
that
none
of
the
libraries
we
used
actually
worked
with
it
and
track
down
the
problem
to
the
spec
specifying
in
HTTP
binary
mode
there.
Any
string
and
a
header
value
should
be
surrounded
by
double
quotes,
but
none
of
the
implementations
I
came
across,
which
concretely
were
only
two
distinct.
L
Go
implementations
actually
respected
that
they
always
interpreted
the
quotes
into
the
value
itself
rather
than
removing
them,
and
the
some
people
have
talked
to
basically
said
this
looks
like
a
spec
bug
rather
than
an
implementation
bug.
I
just
wanted
clarity
on
which
one
it
is,
and
if
it
is
a
spec
bug,
we
should
go
back
and
fix
it,
because
all
the
current
examples
basically
state
these
have
to
be
present
and
have
to
be
removed,
but
no
implementation,
I
can
find
actually
does
so.
Okay,.
B
So
my
question
is
for
Clemens,
because
I
think
you
may
have
actually
been
the
person
to
write
this
up
and
I
think
this
came
around
because
of
some
language
in
the
spec
that
talks
about
doing
adjacent
encoding
on
values
and
stuff.
It
was
it
really
your
intention
to
include
quotes
as
the
headers
or
in
matters
itself.
No.
D
B
You
don't
be
really
cool
is
if
there
was
some
way
and
marked
down
to
do
like
an
include,
and
that
way
we
could
have
test
cases
that
actually
run
on
the
on
some.
You
know
on
this
file
basically,
and
then
at
the
end
they
can
do
a
curl
with
it,
but
then
also
sucks
it
into
the
markdown.
You
don't
do
it
some
weird
copy
and
paste
thing
that
gets
outta.
Sync.
Yes,.
B
I'm
wondering
if
it
would
be
worthwhile
to
make
a
note
not
in
the
old
version
of
the
spec
but
somewhere
in
our
documentation.
That
says
we
noticed,
there's
a
typographical
error
in
the
previous
versions
of
the
spec.
While
we're
not
updating
it.
The
implementers
should
be
aware
of
the
the
error
and
you
know
basically
remove
the
quotes.
Yes,
so
I
know.
B
F
B
K
K
My
preference
before
one
dot,
oh,
is
that
we
treat
these
all
as
Veda
and
they
won't
necessarily
be
interoperable
in
the
same
way
with
most
of
the
patchy
projects.
I
worked
with
on
in
the
past.
Thrift
is
an
example
that
that
it
was
annoying
when
they
would
switch,
but
we
knew
what
we
were
doing
and
we
knew
that
we
chose
a
a
beta
version,
free
one
dot
over.
D
L
My
preference
is
someone
trying
to
use
the
spec.
Is
that
at
the
very
least,
there's
a
note
in
the
spec
itself
saying
this
isn't
accurate.
You
should
do
this
instead,
because
looking
at
an
older
version
of
the
spec
has
a
lot
of
our
things
omit
point.
One
I
shouldn't
have
to
have
kind
of
arcane
knowledge
for
oh
ignore
this
particular
part.
I
want
to
be
able
to
look
at
the
spec
and
immediately
see
what
is
correct.
L
B
I
guess
in
terms
of
old
versions
of
the
spec
I
think
there'd
be.
There
are
two
options
there.
If
we
decide
to
do
something,
one
is
actually
fake.
This
thick
the
spec
proper
and
create
a
zero
dot
one
dot,
one
zero
zero,
two
two
and
zero
3.1,
so
they're
funky
in
there.
But
you
know
I
meant
how
did
op
three
of
the
minor
version
number
or
add
a
note.
It
says
we're
not
happy
in
the
spec
but
remove
the
quotes.
Who
was
a
typo,
but
we
did
want
to
think
on
the
burden
of
modifying
this.
B
F
B
F
It
because
I
actually
make
the
difference
between
between
cloud
events,
fields
that
are
encoded
and
they
have
the
quotes
if
they're
strings,
and
then
they
have
effect
they're
using
JSON
encoding
all
the
way,
which
is
then
correct
with
quotes
because
I
have
other
fields
which
are
not,
which
are
mapped
if
every
natively,
like
content
type,
is
not
explicitly
explicitly
to
an
HTTP
field,
and
then
that
is,
of
course
uncoded,
because
that's
what
the
norm,
the
normal
format
for
content
type,
is
so
my
question
again.
What
from
that
curl
example
what
that
bug
come
from,
but.
F
D
F
F
If
we
want
to
allow
for
arbitrary
extensions
to
pass
through
HTTP,
we
need
to
allow
for
new
numbers,
because
we
just
added
them
and
for
Strings
to
be
encoded
which
II
to
me,
which
means
we
need
to
make
the
difference,
because
otherwise
we
can't
decode
them
anymore.
So
we
made
the
rule
to
say
we're.
Basing
we're
basing
this
on
Jason
and
so
Jason
is
the
way
how
those
strings
that
the
header
value
is
to
be
interpreted.
You
run
it
through
you,
you
read
those
through
the
JSON
decoder
and
the
infer
type
is
what
you
get.
D
F
B
Understand
that,
but,
like
I
said
you
get
to
remove
your
technical
hat
for
a
second
and
look
at
it
from
somebody
looking
at
this
with
the
first
set
of
eyes,
they're
gonna,
look
at
this
and
say:
oh
someone
by
mistake,
but
quotes
around
stuff.
Obviously
these
are
all
strings
type.
You
know
time
ID,
those
are
all
just
strings.
Look
the
quotes.
There
must
be
a
typo.
Everybody
assumed
that
as
we're
writing
the
spec
and
we're
experts
in
this
I.
F
Mean
we
can,
we
can
drop
them.
I
need
to
effective
I
need
them
to
work
Smith
effectively
an
exclusion
clause
they're
in
the
spec
that
says,
if
so,
what's
the
distinction
now
between
a
field
that
carries
a
number
like
to
see
the
sequence,
this
our
sequence,
extension
and
a
field
that
carries
a
string.
B
Right,
I
guess
I
kind
of
assumed
that
if
we
were
going
to
fix
this,
that
we
would
basically
for
all
the
well-known
types
we
know
how
to
encode
those
HTTP
headers
already
and
you're.
Gonna
know
what
the
value
is.
So
when
you
decode
it,
you
know
whether
it's
a
number
or
a
string
already
for
extensions
in
general.
F
B
G
G
F
G
G
B
Binding
I
don't
know,
I
have
to
think
about
that
one,
because
that
was
kind
of
assuming
that,
if
you
don't
know
what
it
is
three
days
of
string
isn't
that
big
of
a
deal,
because
even
if
it's
a
JSON
blob,
you
could
still
treat
it
as
a
string
and
then
somebody
who
wants
to
actually
process
process
I
think
well.
No,
oh!
This
is
Jason
because
that's
what
the
definition
of
that
extension
means-
or
it
says
I'm
not
going
to
decode
this
Jason
I
think.
F
So
my
what
I've
been
trying
to
do,
my
intent
here
was
to
have
a
type
safe
mapping
of
the
the
information
model
of
cloud
events
onto
HTTP
headers
in
a
reversible
way,
so
that
the
the
receiver
would
be
able
to
go
and
restore
effectively
the
same
types
using
Jason
type
inference.
That
was,
that
was
the
goal
of
that
yeah.
B
B
B
B
L
B
G
Yes,
I
think
a
common
serum
in
line
with
my
thinking,
I
sort
of,
was
questioning
the
term
optional
to
handle,
which
I
think
is
what
premise
I
did
but
I
think.
If
we
can
agree
that
you
know,
middle
middle
men
have
to
pass
those
optional
things
on,
even
if
they
don't
understand
and
I.
Think
I'm
I
think
that's
really
what
the
intent
was.
G
A
Didn't
propagation
one
way
of
handling,
so
why
would
one
be
optional
to
him
with,
but
still
have
to
propagate?
How
can
you,
if
a
middleware
doesn't
support,
doesn't
understand?
There
will
be
a
header
coming
through
from
some
accession?
It
will
just
drop
it.
So
how
can
be
optional
must
propagate
that
doesn't
sound
right
to
me.
That's.
G
B
Yeah
I
think
I,
don't
know
where,
but
there's
some
conversation
about
this,
where
we
talk
about
how?
Maybe
we
should
add
text
that
says,
if
you
don't
prefix
your
extensions
with
Cee
I'm,
sorry
C,
then
you
run
the
risk
of
them
being
dropped
by
intermediaries
because
they
have
no
idea.
They're
cloud
events,
headers.
B
A
J
This
has
certainly
some
impact
on
the
PR
I
am
supposed
to
prepare
regarding
the
immutability
of
the
event
context
there
there
was
a
con
sense.
I
think
that
we
shouldn't
impose
too
strict
rules
and
I
mean
removing
an
attribute
is
maybe
an
extreme
form
of
modifying
it.
So
if
we
say
here
must
then
this
would
also
impact
this
rule.
J
B
B
F
M
B
So
I
do
feel
like,
unfortunately,
we
actually
merged
two
issues
together.
He
said
my
end.
There
too
is
Genesis.
I.
Think
the
optionality
thing
is
different
than
the
forwarding
thing
Klaus.
Would
you
want
to
take
a?
Would
you
want
to
take
the
responsibility
of
including
this
part
of
this
discussion
in
your
PR
that
you're
gonna
write
up
yeah.