►
From YouTube: Serverless Working Group Meeting 2019-09-26
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
Serverless Working Group Meeting for September 26, 2019
B
B
B
B
B
B
B
B
B
B
B
B
A
A
A
A
A
A
A
A
A
A
A
A
C
A
E
E
B
A
That's
that's,
ok,
I
think,
I
think
this
will
be
relatively
short
because
we
don't
have
a
lot
of
PRS
there's
one
ongoing
discussion
where
I'm
not
sure
that
Alan
is
gonna
show
up,
but
I
can
probably
talk
to
its
closest.
You
know
and
there's
a
there's,
someone
with
on
the
phone
9
to
5
6,
9,
9
0,
to
7.
D
B
A
A
B
G
G
G
B
B
A
Dock
turn
off
print
mode,
III
I
can
barely
so
I,
don't
think
I'm
even
enter
in
this
thing,
so
I
can
probably
make
a
proposal
and
then
someone
has
to
go
and
accept
it
and
I
don't
want
to
break
anything.
I
I'm,
just
I'm,
just
gonna,
go
make
annotations
in
this
document
which
will
later
accept
by
by
someone
and
then
no
plea,
I,
don't
break
anything
alright,
so
but
I
hope
you
can
see.
A
So
Doug
is
out
right
now,
I'm,
not
sure
what
he's
doing,
but
he
asked
me
that
I
take
over
the
call
for
this
week
and
I
think
next
week.
Someone
else
who
is
doing
the
call
so
I
think
it's
really
relative
to
be
quick,
because
we
don't
have
a
lot
of
PRS
and
we
have
a
few
issues
and
then
there's
going
to
be
the
SDK.
A
call
is
going
to
be
afterwards
but
so
question
community
time.
Anybody
have
anything
of
importance
to
raise.
A
F
F
A
Sorry
yeah,
you
might
be
okay,
so
but
I
had
commented
on
this.
It's
relatively
easy.
We
have
affected
a
left
over
in
the
average
human
is
that
data
is
represented
as
basic
c4,
which
is
senseless
in
in
in
Avro,
because
ever
understands
bytes.
So
it's
it
just
changes
that.
So
that
makes
more
sense.
Now
we
can
make
this
a
little
bit
more
complicated,
but
that's
what
the
correction
is.
A
I
would
not
take
this
right
now
and
it
would
still
comment
because
data
contents,
I
this
data
content,
type
and
I-
would
probably
go
and
prefer
if
this
was
not
using
the
content
type
at
all,
but
would
use
it
something.
That's
the
data
field
would
have
a
simple
value,
so
I
would
I
think
this
is
a
patch
that
we
should
take
personally,
but
I
would
not
take
that
right.
Now.
A
No
it's
late,
so
let's
go
and
take
a
snip
next
week
so
that
we
see
that's
the
only
PR
I
think
that's
in
the
list.
That's
going
to
take
a
look
at
the
list
per
se.
I.
A
A
Effectively,
we
have
not
made
clear
that
the
context
that
all
context
attributes
need
to
be
of
the
the
in
inside
the
type
system
and
Allen
has
been
concerned,
because
that
is
missing
there
specifically
for
extensions
that
would
cause
it
to
operability
issues.
So
this
is
effectively
tighten
up.
Tightening
up
that
text.
A
So
I
think
the
rule,
the
the
thing
I
think
the
rule
that
he
wants
to
add
here
is
effectively.
The
extension
attributes
must
also
adhere
to
that
rule.
That
extension
afternoon,
specifically
also
must
follow
the
canonical
string
encoding
to
avoid
ambiguity
with
other
stringing
Kony's
alpha
at
personally
I.
Don't
have
a
strong
feeling
about
this.
I
I
H
A
So
so
I
think
his
point.
His
point
was
at
the
beginning,
as
we
started,
this
discussion
is
that
interoperability
isn't
possible
because
the
extension
is
not
the
excel.
The
extensions
are
not
to
not
define
to
be
within
the
type
system
and
I
think
that's
the
the
the
goal
of
the
fix
is
to
make
sure
that
they
are
and
I
think,
that's
the
that's
the
objection
and
yes
but
I
see
these
extensions
are
just
contracts
after
its
context,
attributes
and
I'm
not
sure
that
that's
true,
okay.
I
A
D
A
C
D
A
A
H
I
had
a
question
about
data
content,
encoding
yeah,
if,
let's
say
you're,
trying
to
translate
a
an
HTTP,
structured,
encoded
message
into
a
binary
encoded
message.
Would
you
expect
for
the
body
to
be
basic,
C,
4
encoded
and
then
have
the
normal
header
set
or
where's
that
not
a
concern
like
if
you
preserve
that
for
that
translation?
If
you.
A
H
A
A
You
read
it,
but
so
so
basics
we
have
now
made
basically
for
a
feature
of
Jason,
so
underscore
basic
C
force
is
a
feature
of
the
JSON
binding,
which
means
as
if
you
leave
it
in
underscore
basics,
before
it's
effectively
a
flag.
It's
a
it's
an
attribute.
If
you
will
on
the
data
field.
That
says
this
here
is
Jason
and
this
is
Jason
string,
but
you
ought
to
treat
it
as
a
binary,
so
you
read
it
in
and
it
becomes
a
binary
in
your
in-memory
structure.
A
A
A
So
arguably,
the
MVP
mapping
is
too
wide
in
numbers
in
terms
of
numbers
and
bits
and
then-
and
here
and
then
Alan
raises
this-
sets
because
we're
now
going
to
1.0.
He
just
wants
to
make
sure
that
that
we're
dressing
this
the
spec
uses
32
bits
for
integers.
So
those
two
things
are
related.
If
we
make
the
race
bigger
than
the
mismatch
with
HTP
is
gone,
so
he
is
that
he's
like.
Shall
we
make
this
64
bit?
So
let's
take
a
look
at
that
briefly.
A
Here's
that
so
I
think
was
being
raised
here
is
the
question
of
since
we're
not
talking
about
bits
here
where
well.
Actually
we
would
say
this
is
within
the
range
of
the
sign
32
bit
at
third,
the
villager,
whether
we
shall
expand
that
to
64
bits
and
whether
whether
we
shall
leave
that
at
32
does
anybody
have
it
believe.
I
Yeah
I
think
it
would
probably
be
dangerous
to
extend
them
to
64
bits,
because
a
very
high
proportion
of
the
data
shuffling
back
and
forth
around
the
internet
is
in
JSON
and
in
JSON.
The
only
thing
you
can
expect
to
be
interoperable
is
64
bit
floats,
which
means
you
really
only
have
53
bits
of
integer
and
essentially,
if
you
require
it
to
support
64
bits.
That
means
you
probably
can't
transmit
it
in
JSON,
unless
you
encode
it
as
a
string
which
is
Nikki
Tim.
A
A
B
A
Me
and
Jason
support
for
converse,
okay
and
then,
and
then
I
I,
don't
think
the
mathletes
Thank
You
P
long
hurts
because
nqp
is
actually
pretty
clever
about
how
it
encodes
things.
So
if
the
number
is
short,
the
the
encoding
will
be
shorter
so
and
that's
mostly
between
Ellen
and
me,
I
would
say
for
the
for
the
MVP
mapping,
so
I
don't
have
any
problems
with
the
with
the
issue
that
he
raised
there.
So
let's
go
and
also
make
a
comment
on
this
one.
F
A
Bit
with
bit
sizes
are
something
that
that's
a
decision.
That's
pretty
heavy
in
terms
of
you
know
back
and
forth,
back
and
forth
compared
in
compatibility,
you
can
always
go
longer,
but
you
can't
really
go
shorter,
so
yeah
I
think
that's
a
that's
the
thing
that
Alan
race,
specifically,
he
called
that
out.
He
said
this
is
something
we
should
discuss
before
we
go
Jiwon
and
I.
Think
we
just
did
that.
Okay,.
A
No
good
well,
then,
then,
I
have
a
few
people
that
I
think
we're
done.
Then
there
will
be
an
SDK
call
right
after
this.
If
everybody
is
here
for
the
SDK
Paul,
which
kind
of
seems
like
we
couldn't
start
that
right
away
after
that,
otherwise
I
wish
you
a
great
day
Christian
Lexi
Don,
that's
joined
hello.
You
got
my
name
right.
Thank
you.
A
I
think
that's
it.
So
thank
you
very
much
for
showing
up
and
the
next
meeting
we'll
have
a
few
people
missing,
because
there's
a
there's,
a
holiday
in
Germany
and
doggies
are
not
going
to
be
here,
but
we
still
have
business
to
close
out.
So
you
should
nevertheless
show
up,
even
though
some
of
us
are
not
and
that's
it
SD
cake,
SDK
folks,
just
stay
on
the
line
and
I
would
say,
or
if
you
insist
that
we
do
this
in
half
an
hour,
we
can
do
that.
Do.
B
A
A
A
A
A
J
H
A
Yeah
and
and
I
so
I've
done
I've
done
a
little
bit
of
go
exercises
exercising
between
whenever.
Well,
you
know,
because
we
might
my
scope
and
that
work
is
growing
a
little
bit
so
I
need
to
be
well.
I
will
not
gain
proficiency,
probably
I
should
be
able
to
grow
and
we
read
the
code
well
and
so
I've
been
taking
some
courses
etc.
So
I
know
I,
understand
better
where
you're
coming
from
and.
H
A
A
You
know
original
sense
of
see
in
the
way,
with
lots
and
lots
and
lots
of
modern
elements
and
more
safety,
but
it
clearly
has
a
very
sea
feeling
to
it
like
from
when
I
am
first
Lauren
see.
That
seems
like
a
very
natural
upgrade
path,
also
for
mindset,
perspective,
etc
and
kind
of
fixing
all
the
problems
that
C
had
without
going
through
the
terrible
horrible
thicket
of
barbed
wire
that
is
C++
yeah.
H
A
A
A
A
H
H
A
It's
been
so
for
me:
it's
been
fairly
uneventful,
except
that
I've,
so
I
figure
doing
about
a
bunch
of
conditionals
around
some
constructs
because
we
made
we
so
we
didn't
make.
We
didn't
we'd
make
some
changes
towards
30.3
that
were
structural
and
then
we
started
cutting
mat,
support
now
towards
below
and
I
think
that
was
the
biggest
significant
change
and
then
we
had
also
rules
around
simplifications
around
need
a
constant
side,
etc.
A
A
The
strongly
typed
in
the
strongly
type
representation
of
the
event
there
I'm
effect
attracting
the
latest
version
and
that's
what
you
need
to
program
against
and
if
you
want
to
go,
and
this
is
kind
of
towards
1.0,
you
know
you
pick
the
SDK
and
the
SDK
has
a
particular
version
that
it
is
biased
towards,
and
you
take
those
bits
and
use
those
the
the
internal
corset.
This
called
the
cloud
events
attributes
and
that's
the
thing
that
does
the
serialization
and
does
the
validation.
A
That
thing
just
operates
on
a
dictionary
and
that
that
thing
then
gets
kind
of
loaded
up.
If
you
will,
by
the
by
these
strongly
typed
Shem,
that's
around
it.
So
I've
been
doing
like
almost
of
the
upgrade
work
happened
in
there
and
I
didn't
find
yeah
for
me
that
wasn't
very
complicated.
It
was
mostly
just
it
was
cutting
code
in
in
the
sense
of
you
know,
stashing
it
away
if
it's
not
1.0
or
rather
than
rather
than
making
anything
more
complicated.
So
I
think
we've
been
doing
some
simplification.
H
Actually,
founded
so
in
the
primer,
there's
a
there's
language
that
says
the
specification
places
no
restrictions
on
the
type
of
extension
attributes.
Maybe
they
meet
maybe
simple
types:
strings
and
integers,
complex,
structured
or
undefined
collection
of
attributes.
How
about
how
about
you,
you,
fellow
PR,
to
go.
H
I've
been
highlighting
things
that
I
need
to
follow
up
on,
so
I
have
several
things
that
I
found
that
are
slightly
inconsistent,
but
I'm
this.
This
is
incorrect.
Right,
like
we
don't
allow
multi-level
extensions
into
the
payload
anymore.
We're
certainly
not
the
payload,
be
the
attributes.
I
can't
say
like
food
bar
dot.
Baz
is
a
string.
No.
A
A
So,
basically
out
of
the
development
experience
would
be
great
to
go
and
make
a
take
a
pass
through
the
farmer
specifically
and
see
where
there's
stuff.
That
needs
to
be
patched
up,
and
it
could
be
that
there's,
even
you
know,
duplication
between
paragraphs
or
that
you
know
one
paragraph
doesn't
really
flow
with
another
with
another
one.
So
doing
some
editorial
patches
there
would
be
would
probably
be
helpful
and
the
primer
is
the
document
that
is
not
normative.
A
But
it's
really
just
the
explanation
and
if
they're
stuff,
that
it's
just
wrong
and
because
we've
made
changes
and
nobody
cared
to
track
them
in
the
primer
and
then
we
should
go
and
correct
that.
That's
it's.
The
primer
is
a
document
also
that
we
could
probably
keep
revising
and
clarifying
beyond
v1,
because
that's
not
I,
don't
think
that
I
think
that
after
that
document
is
version.
H
A
H
And
my
one
other
thing
is
eyes:
I've
been
calling,
you
know
like
HTTP
and
AMQP
and
math
things
transports,
but
the
spectacles,
the
protocol,
but
I
think
it
mixes
that
the
terminology
between
transports
and
protocols
and
transport
protocol
bindings
and
things
like
that
which,
which
is
the
preferred
M
word
to
use
because
I
want
to
change
things
in
my
side.
If.
A
You,
if
you
aim
for
perfection,
then
I
think
those
are
applications,
the
things
we
have
bindings
for
all
application
protocols,
because
the
the
transport
underneath
is
were
you
know,
TCP
or
UDP
or
quick,
and
then
we
really
mapped
to
a
probe
to
application
protocols.
They're
not
really
transports,
that's
the
purity
of
it,
but
then
what
they
do
for
us
is
they
obviously
act
as
transports.
So.
H
J
J
A
Okay,
we,
it
would
have
been
great
if
you
have
been
here
45
minutes
ago,
because
we
had
your.
We
had
your
pr's
on
the
table.
Oh
sorry,
your
issues
on
the
table
in
terms
of
business
of
integer,
yes
and
so
timbre
said
well.
We
can't
expand
it
to.
We
can't
expand
to
64-bit
because
of
the
width
of
Jason,
which
is
something
that
you
I
think
you
also
mentioned
well,.
J
Yeah,
it's
kind
of
that
was
kind
of
the
opposite,
argued
I,
don't
really
feel
strongly
about
it.
I
just
thought
we
should
make
sure
we
haven't
done
it
made
a
mistake,
usually
go
to
press
the
width
of
Jason
in
practice.
Jason
doesn't
specify
anyway,
but
in
practice
it's
the
width
of
the
integer
part
of
an
I
Triple
E
64
bit
float,
which
is
about
52
bits.
Yes,.
A
J
A
J
J
H
A
H
A
I
think
I
think
product
implementation
will
help
there.
So
if
I
can
assure
you
that
we
will
go
and
ship
product
with
cloud
events
and
then
from
what
I,
what
I
hear
to
telegraphing,
it
seems
like
they're
gonna,
do
cloud
offense
to
and
I
think
that
problem
would
solve
itself.
The
question
is
whether
that's
going
to
be
the
official
SDK
for
our
for
what
we
are
gonna
do
in
c-sharp.
A
It
looks
like
my
engineering
team
is
actually
looking
at
using
the
this
SDK
is
the
core,
which
is
a
good,
which
is
a
good
thing,
because
then
they
will
take
ownership
of
it
because
they
hate
whenever
I
write
code,
and
so
we're
probably
going
to
build
on
top
of
the
the
official
SDK
for
our
for
our
purposes
and
I
would
assume
that
in
your
world
Scott
that's
going
to
be
and
I
think
for
other
languages.
Other
languages
I
mean
we
have
a.
A
We
have
a
specification
that
people
can
build
to
and
what
we've
created
is
reasonably
compatible
with
with
existing
stacks
that
people
have
so
that
like,
if
you
want
to
send
in
a
binary
event,
you
need
nothing
but
a
regular
HTTP
clients
right.
That's
all
you
need,
and
that
was
kind
of
a
design
goal,
and
you
can
get
pretty
far
with
just
having
a
you
know:
jason
encoder
and
HTTP
client,
which
means
you
probably
don't
necessarily
need
to
have
the
sdk
for
simple
cases.
A
A
No
typescript
I
mean
none
of
those
languages
for
something
that's
more
a
little
more
sophisticated
and
c-sharp
Java
and
and
go,
and
the
expectations
of
developers
have
a
strongly
tied
experience,
etc.
And
that's
more
important,
I
haven't
looked
at
the
Java
at
the
Java
state
of
the
world.
It's
the
Java,
SDK
I.