►
From YouTube: CNCF Serverless WG 2020-02-28
Description
CNCF Serverless WG 2020-02-28
A
A
A
A
B
E
F
B
G
B
H
B
D
I
H
D
B
Okay,
cold
son,
make
sure
all
right.
Nothing
new
with
AI
is
a
believe
community
time.
Okay,
anything
from
the
communities
that
people
want
to
bring
up
lesson
on
the
agenda,
all
right,
cool,
all
right,
ku,
con
tu,
first
of
all,
I,
don't
know
if
anybody's
been
paying
attention,
but
last
I
heard
the
conference
organizers
will
still
planning
on
having
the
conference
despite
the
the
virus
and
all
the
other
stuff
running
around.
A
That
again,
it
has
some
local
color
on
this
go
for
it.
I
live
very
close
to
the
Dutch
border
and
we
have
a
there's.
A
my
neighbor
in
district
is
the
one
that
is
currently
causing
all
the
noise
for
for
Germany
and
is
right
by
the
Dutch
border.
So
it's
I
don't
want
to
make
you
any
anybody
panicky,
but
it's
like
it's
not
so
unlikely
that
this
is
going
to
jump
into
the
Netherlands
and
is
also
going
to
jump
into
into
my
Chris
this
trick
and
then
who
got
good?
B
G
H
B
Ahead,
sorry,
the
elbow
boom
yeah
I
was
thinking
fist
bump,
but
yeah
like
elbow
bumps
is
usually
you
have
some
clothing
that
you
know,
although
okay,
we're
really
digressing
here
but
now
I'm
starting
to
think
of
the
Rocky
Horror
Picture
Show,
and
maybe
that's
going
way
too
far-
nevermind,
okay,
but
speaking
of
booths,
we
did
get
confirmation
that
we
have
a
booth
reserved
for
us.
It's
booth,
11.
B
If
you
click
on
that
link,
you
can
actually
see
more
information
about
what
the
other
booths
are
doing
if
you're
interested
we
are
in
the
afternoon
sessions
for
all
three
days.
So
please,
if
you
are
able
to
sign
up
for
some
booth
times
by
putting
your
name
in
here,
I
did
add
when
the
two
sessions
are
just
so
you
guys
have
that
for
easy
reference
in
case
you
are
planning
on
speaking
one
of
those
or
joining
one
of
those.
So,
for
example,
I
think
Scott's,
no
I
think
it's
got
session
on.
B
Wednesday
might
require
several
people,
because
it's
a
more
of
a
lab
kind
of
thing
and
he
wants
people
to
help
out
there.
So
think
about
that
as
you
go
forward
anyway.
Please
add
your
knees
when
you
get
a
chance
also
when
you're
talking
about
the
booth.
This
is
the
text
that
they
gave
me.
The
most
interesting
thing
here
is
to
make
sure
that
you
do
not
talk
about
products
there.
That's
a
I!
Think
that's
an
obvious
thing.
You
should
talk
about
the
technology
and
the
spec
itself,
but
if
would
the
interesting
thing
to
me?
B
Was
this
little
bit
right
here
where
they
said
you
know
if
they
hear
you're
doing
a
sales
pitch,
don't
you
know
they
right?
They
deserve
the
right
to
kick
you
out
or
so
that
was
kind
of
interesting,
which
makes
sense,
but
it
seemed
kind
of
brutal.
So
keep
that
in
mind,
if
you
are
studying
it
for
booth
duty,
okay,
I
think
trying
to
think
that
anything
else
here,
no
I
think
everything
else
is
old
news.
So
any
other
questions
or
comments
relative
to
coop
con.
B
K
Can
I
just
make
a
quick
comment?
Yes,
so
hello
I
mean
janitors.
People
call
me
Jojo
and
I'm
very
interested
in
surveillance
like
the
Panthers
fabulous,
but
bones
and
I'd
like
to
my
first
meeting
of
the
working
group
for
surveillance.
So
just
introducing
myself
my
first
time
here,
I'm
going
to
be
at
tubecon
and
we're
going
to
be
speaking
at
the
rejects
about
open
third
service
as
well.
So
just
wanna
join,
what's
happening
on
the
same
chef
services,
working
group,
Oh,
Gerry,
saying
hi,
yep,.
B
L
B
So
I
don't
forget
all
right:
okay,
moving
forward
cuckoo
comer
done
with
okay
SDK
call,
so
you
don't
have
a
call
last
week
whether
we
do
four
so
I,
don't
think.
There's
anything
to
mention
other
than
I
did
create
the
rust
repo
and
the
folks
are
currently
working
on
doing
that.
Merging
of
those
to
rest
projects
that
are
out
there
today,
I'm
assuming,
even
though
we
didn't
take
an
official
vote,
is
there
any
objection
to
us
pulling
them
in
as
a
as
a
new
SDK
repo,
okay,
I,
didn't
think
so?
B
Cuz,
no
one
raised
an
objection
before
but
I
know
we
didn't
do
it
a
formal
vote
so
so
I
make
that
official
we
do
technically
arms
are.
We
are
supposed
to
have
a
call
today
so
immediately
after
this
call
will
have
the
SDK
call.
Unfortunately,
if
the
call
runs
past
the
top
of
the
hour,
I
will
not
be
able
to
join
it,
so
someone
also
have
to
be
a
note-taker
at
that
point.
Just
a
warning
for
a
quick
call.
B
Let's
see
Kathy
is
not
on
the
caller
believed.
Do
we
have
anybody
else
from
the
workflow
subgroup?
Who
wants
to
give
status?
I?
Don't
see
anybody,
okay
moving
forward,
then
let's
talk
about
discovery,
API
is
so
Clements
might
pick
on
you
to
go
first,
since
I
think
you
said
you
need
to
leave
her
a
little
bit
early,
maybe
bring
us
up
a
data
on
anything
that
might
have
happens.
You
know
I,
think
you're
on
vacation.
A
A
Colin
Colin,
it's
more
a
little
bit
further
down
to
the
sections,
but
we
have
the
configurations
for
the
transports
there
yeah
so
Colin
added
further
information.
We
need
formats.
So
that's
the
this
there's
subject:
name
the
cosmos
full
mode
and
then
which
are
all
clients
like
configurable
settings
that
equals
similar
settings
we
have
making
in
20
and
then
480p.
A
We
clouds
contributed
that
we
talked
about
that
in
some
detail
and
were
fairly
happy
with
that
set,
and
then
we
also
added
the
HDPE
things
which
are
all,
and
all
of
these
are
effectively
transport
specific
configuration
parameters
for
pushing
out.
So
you
have
this
eviction
management
subscription
manager
is
tasked
with
taking
event
and
moving
it
elsewhere,
push
and
so
that
these
are
the
parameters
for
the
respective
transports
that
we
have
called
events
as
standard
price
boards,
so
we're
gonna
go
and
configure
those
and,
of
course,
every
implementation.
A
If
that
wants
to
support
a
proprietary
protocol
can
add
their
own
as
much
as
you
can
add
your
own
transport
bindings
to
cloud
event
and
the
proprietary
fashion,
if
you
wanted
to
so
those
are
the
ones
the
protocols
that
we
have
in
the
in
this
set,
so
that
that's
what
that
is,
and
then
Ryan
had
a
proposal
that
he
has
not
put
into
this
document
yet.
But
that's
the
plan
for
filtering
and
we've
and
he's
getting
some
variations
of
this
right.
J
J
A
So
you
know
if
you're
using
ATP,
you
go
and
create
a
subscription
of
sorts
and
then
pull
data
towards
you
with
MQTT.
You
have
a
spread
gesture
and
again
you
pulled
events
towards
you,
I'm
meditating
connections.
That's
what
pool
means
and
of
course,
if
you're
choosing
amp,
you
can
see,
there's
a
certain
set
of
filtering
capabilities
that
exists
in
that
broker
and
if
you
use
ATP
that
there's
a
certain
set
of
filtering
capabilities
that
exist
in
that
broker,
maybe
be
standardized,
maybe
custom,
and
so
we're
not
going
to
touch
those.
A
And-
and
it's
really
like,
if
you
use
any
of
that
sort
of
infrastructure
you're
effectively
compliant
with
this
subscription
model,
because
you're
using
native
capabilities
of
a
broker
I,
don't
think
we
need
to
go
in
and
start
standardizing
into
an
Coupee,
and
we
don't
need
to
start
standardizing
into
n
kpt,
because
people
who
are
using
these
sorts
of
brokers
know
what
they're
doing
now.
What
what's
different
is
is
if
we
create
relationships,
then
I
kind
of
have
these
multi-hop
deliveries.
Where
you
might,
you
might
want
to
get.
A
You
might
want
to
deliver
to
NPT
clients
which
need
to
pull
from
a
DD
broker,
but
somehow
we
need
to
get
that
event
into
that
entity
broker.
And
that
is
the
sort
of
subscription
that
we're
covering
here
so
you're,
creating
a
routes,
a
push
route
effectively
from
subscription
manager
into
the
DT
broker,
the
event
that
gets
delivered,
incident
and
PTT
broker,
and
then
the
clients
basically
use
NTT
ways
of
getting
of
getting
to
that.
A
A
You
could
arguably
do
a
prefix
and
a
suffix
offer
for
the
subject
field.
That's
that's
possible,
but
that's
the
simplest
set.
So
that's
a
simpler
circular
base
you
just
anchors
on
those
three
fields
and
then
so
that's
the
basic!
That's
what
we
call
the
simple
dialect
and
then
we
will
have
extended
dialects
or
we're
proposing
that
there
might
be
extended
dialects,
which
then
also
allow
in
a
matching
against
arbitrary
other
properties,
but
based
on
implementation
experience
that
we
have
and
then
also
that
Ryan
has
said
they
have
because
of
cost
of
this
matching.
A
We
want
to
keep
this
simplest,
so
they're
really
simple
and
super
constrains,
and
that's
the
thing
that
everybody
must
implement
effective
because
lines
and
then
we'll
leave
effectively.
The
definition
are
further
dialects
to
extension,
similar
to
the
way
and
we've
done
extensions
for
for
cloud
events
per
se.
So
if
someone
wants
to
once
to
come
and
say,
were
we're
gonna
we're
gonna
define
a
sequel
dialect
to
match
events,
that's
great
and
the
way
this
and
the
way
this
ought
to
show
up
how
the
the
filter
dialects
are
being
communicated.
A
Let's
see,
that's
the
idea
we
have
about
this
multiple
dialects,
there's
some
negotiation
happening
through
ways
of
discovery,
and
we
might
have
a
discovery
mechanism
for
HTTP
as
well.
We
haven't
thought
about
how
this
looks
in
HTTP
just
yet
I'm
at
the
protocol
level,
but
I
can
imagine
it
we're
gonna
have
a
potentially
an
optional
extension
way.
You
can
basically
walk
up
to
the
description
manager.
Endpoint
do
an
options
request
and
then
you
get
a
header
with
all
the
supported
dialects
for
for
subscriptions,
which
is
similar
to
how
an
GP,
for
instance,
will
do
it.
A
That's
what
we
had
that's
what
we
are
Safari
in
terms
of
getting
to
a
complete
draft
I
think
this
section
is
good
in
shape.
I,
don't
I,
don't
know
how
that
what's
your
idea
for
how
the
I
mean
there's
stuff
missing,
but
did
you
think
we
should
need
to
break
this
out
and
just
a
separate
document?
Do
we
leave
that
section
in
that
document?
Do
you
think
I.
B
Okay,
yeah,
that's
better!
Thank
you,
I!
Don't
care
about
separate
documents,
yet
I
guess
what
I'm
more
looking
for
is
to
make
sure
that
the
text
in
here
is
Morse,
peckish
right.
So
R
example,
this
section
right
here
well
I
understand
the
most
critical
things
just
get
the
information
that
you
or
get
the
bits
that
you
want
to
have
specified,
but
putting
that
into
specky
language,
so
it's
more
readable
I
think
is
really
what
I'm
looking
for
more
than
anything
else.
B
That
way,
someone
can
we
can
hand
this
to
somebody
and
say:
hey
here's,
a
rough
draft.
You
know
start
commenting
on
it
as
opposed
to
you
know,
I!
Look
at
this
section
right
now.
They
may
look
at
this
and
it's
not
gonna
be
hard
percent
sure
what
they're
supposed
to
do
with
these
things.
That's
that's
what
I'm
worried
about
I.
A
A
E
E
E
A
B
E
Was
also
on
vacation
and
came
back
to
a
pile
of
work,
but
I
wasn't
expecting
so
I'm
gonna
need
another
week
to
get
the
discovery
into
a
more
first
draft,
II
kind
of
states.
If
you
have
any
ice
companies
called
hope.
You'll,
add
comments
to
the
dock.
While
I
was
gone,
it's
great.
If
anybody
wants
to
actively
participate,
reach
out
to
me
on
slack
or
speak
up
now.
Otherwise,
I
will
probably
draft
this
either
later
today
or
tomorrow,
and
send
it
out
I'll
post
it
in
this
like
channel
when
it's
ready.
K
L
J
Sorry
this
is
Brian,
I
was
just
needed
I.
So
there
is
some
alignment.
They
needs
to
happen
between
the
two,
the
subscription
section
around
how
filters
are
advertised
and
how
that
negotiation
happens
with
discovery.
So
I
don't
know,
I'm
happy
to
join
for
a
session
there
to
talk
about
it,
or
maybe
we
can
just
revisit
it
once
we
get
this
in
better
shape
and
come
back
as
a
larger
group.
What
were
your
thoughts.
E
J
B
B
So
Klaus
you've
actually
taken
a
look
at
this
one
actually
I'm.
Sorry,
we
talked
about
this
I
think
on
last
week's
class
right,
the
last
phone
call
we
had
and
for
the
most
part
it
looked
okay,
but
didn't
Jim
had
a
comment
and
the
author
did
update
the
spec
or
is
his
changes
based
on
that
comment,
I
think
he
just
moved
things
around
slightly
or
changed
some
URL
I
think
Klaus.
You
actually
took
a
look
at
it.
Okay,.
N
At
this
single
attribute
here
the
about
the
constant
encoding,
because
there
was
about
a
remark,
I
did
in
an
other
issue
in
the
discussion,
mm-hm
I
mean
Jason
schema
allows
to
to
specify
the
encoding
and
for
data
of
base64.
It
was
clear
that
it
is
a
64
encoded.
So
that
was
my
remark
to
add
that
okay.
N
B
Okay,
does
anybody
just
wanna
checks
on
here
so
that
last
little
change
or
was
done,
Shh
I?
Don't
what
was
done
two
days
ago
yeah!
He
did
it
two
days
ago,
which
is
definitely
than
the
time
limit.
Does
anybody
need
more
time
to
look
over
this
PR
if
anybody
does
not
feel
comfortable
approving
it
speak
up
mole,
let
it
go
another
week,
but
otherwise
it's
been
out
there
for
a
little
while
so
I'd
like
to
prove
it.
If
everybody's
okay,
with
that
any
objection
to
approving
okay
cool
we'll
do
that.
Thank
you.
B
Oh
yeah,
okay,
I
did
it
for
this
one,
so
this
one
so
the
distributed
tracing
apparently
has
a
rule
that
says
our
sorry
yeah.
It
has
a
rule
that
limits
the
character
set
of
what
you
can
use
for
these
values
and
equals
is
not
part
of
that
value
set
when
I
look
to
the
documentation.
So
it
seemed
completely
right
to
me,
but
does
anybody
else
have
a
chance
to
take
a
look
at
it
or
disagree
with
this.
B
Just
so
you
know
you,
the
reason
you
can't
use
commas
or
equals
I
think
is
because
they're
sort
of
like
reserved
characters,
because
I
think
this
is
basically
a
comma
separated
list
of
things.
So
therefore,
obviously
comma,
you
don't
know
whether
it's
part
of
the
value
or
a
separator
and
then
equals,
because
it's
part
of
the
name
value
pair
equal
thing.
B
B
B
Okay,
in
that
case,
this
call
is
over
if
anybody
like
to
stick
around
for
the
SDK
call
we'll
start
that
up
in
about
a
minute
or
so
assuming
we
have
anything
to
talk
about,
it
could
be
a
very
short
call,
but
if
you're
interested
please
stay
online,
everybody
else.
Thank
you
very
much
and
we'll
talk
again
next
week.
Thank.
J
B
C
Concept
to
the
Java
SDK
and
the
VP
is
need
a
little
bit
of
work
or
reworking
opinion
and
I
and
I
wish
to
contribute
to
it.
If
so,
generally,
I
found
that
generics
are
quite
painful
now,
because
when
you
need
to
write
middleware,
usually
you
don't
you
don't
need
you
don't
know,
and
you
don't
care
about
specification
version
and
the
payload
that
you
pass,
that
is
inside
the
event
so
having
to
fix
those
generics.
C
Is
it's
quite
a
problem,
and
also
there
is
this
JSON
class,
which
needs
a
little
bit
of
rework
too,
because
there
are
so
many
missing.
Some
methods
that
are
quite
hard
to
use
so
I
would
love
to
hear
the
comment
of
human
taints.
The
July's
decay
and
I'll
be
really
glad
to
work
on
it
together
with
him.
So.
C
C
B
C
B
I
think
he
was
I
think
he
was
hoping
to
join
the
cobbler
and
I
think
he
got
pulled
away
for
last
minutes.
That's
why
he's
left
the
comments
yeah
but
I
think
he's.
Definitely
the
main
person
that
you're
gonna
have
to
work
with
on
that,
because
he's
basically
has
been
driving
the
Java
SDK
for
the
most
part.
Okay,.
C
Okay,
yeah
also
also
healed
stories
that
are
good
points
like
removing
external
dependencies.
All
these
things
I
completely
agree
with
him,
and
there
is
like,
for
example,
there
is
the
I
parent
validator,
which
is
kinda
useless.
You
could
because,
like
for
example,
when
you're
quite
bitter
winter,
where
sometimes
you
just
don't
need
to
depandi,
because
if
you
don't
even
know
what
is
inside
the
event,
you
just
want
to
pass
from
one
point
to
another.
So
and
generally
that's
my
comment
about
it.
Ok.
B
B
C
No
just
I
just
want
to
let
you
know
that
we
are
working
on
this.
It's
there,
I
think
the
guys
that
are
working
with
me
did
he
manage
to
join
this
call,
but
yeah,
that's
the
the
work
is
starting.
We
have
planned
that
the
features
that
we
want
now
we
are
working
on
the
data
structure
and
then
we
will
so
the
the
plan
for
the
0.1
is
to
have
at
the
data
structure,
the
JSON,
marshalling
or
marshalling
support
for
0.3,
1.0
version
and
I.
C
B
E
M
B
E
I
E
I
C
E
B
C
B
And
I
apologize
I
I
completely
forgot
that
you
open
this
poor
request.
Otherwise
I
would
have
brought
this
up
in
the
main
call
to
draw
everybody's
attention
to
it.
Anybody
have
any
comments
on
this.
It
seems
reasonable
to
me,
but
to
be
honest,
I
know
next
to
nothing
about
Kafka,
so
I
want
to
say
so.
C
There's
so
this
is
something
that
I
brought
up
in
the
the
selectional
like
a
couple
of
weeks
ago.
Remember,
which
is
about
creating
a
new
encoding
strategy,
a
new
coding
type
for
the
HTTP
binding.
So,
basically,
now
we
had
this
batch
this
batch,
which
sends
events
encoded
in
a
JSON
array.
So
it's
a
gin
of
Jaden's
and
I
would
love
to
explore
using
multi-part
for
a
multi-part
bodies
to
to
send
multiple
events.
C
B
Actually
talked
about
that
at
one
point,
and
maybe
you
guys
can
remember:
I
could
have
sworn
one
of
the
reasons
we
pushed
back
on.
That
was
it's
a
it's
a
it's
gonna
sound
like
weird,
but
I
think
that
the
argument
was
mainly
from
Clemons
saying:
multi-part
mime
is
hard,
I
think
that
was
the
biggest
reason.
No
I,
don't
know
whether
that's
completely
true
or
not,
but
I
think
that
that's
why
I
think
I
can
remember,
can
you
can
you
guys?
Can
you
old-timers
remember
well
what
the
reason
was
that
we
rejected
that.
C
That's
not
true,
because
mucho
mucho
part
has
different
versions
so
as
different
might
mime
types,
the
form
data
allows
only
like
the
file
name
because
it's
it's
designed
for
file
uploads,
but
there
are
different
multi-part
and
there
is
basically
we
can
play
around
this
to
make
by
to
make
an
encoding
which
works.
I
mean
my
question
is:
are
you
interested
on
exploring.
B
This
so
what
what
I
would
do
is
I
would
open
up
an
issue
asking
about
it,
because
I
could
have
like
I
said
I
think
we
talked
about
this
before
and
we
we
didn't
hand
that
head
down
that
path
for
a
reason,
I
just
can't
member
for
sure
what
the
reason
is.
But
if
you
open
an
issue,
you'll
force
someone
like
Clemens
to
respond
to
the
issue
and
that
will
maybe
trigger
some
people's
memory.
In
terms
of
why
we
said
no
in
the
past
and
maybe
those
reasons
no
longer
apply,
I
don't
know.
C
C
E
C
Don't
know
I
have
no
opinion.
Frankly,
on
the
name:
I
I
really
don't
care,
I
mean
what
I
care
is
more
deficiency,
because
now
batch
is
extremely
inefficient,
because
if
you
can
send
things
like
in
binary
so
sending
the
body
it
doesn't
know,
it
removes
the
opportunity
for
doing
a
lot
of
optimizations
yeah.
B
Interesting
well,
if
you
want
to
force
that
discussion
again
again
open
up
another
issue
on
that
I'm
going
to
talk
about
it,
I
would
not
open
up.
I
would
not
work
on
a
pull
request.
Yet
until
you
get
some
discussion
in
the
issue
because
I
don't
want
to
waste
your
time,
but
it
definitely
opened
up
an
issue
to
force
the
discussion.
Okay,.
B
Cool
all
right
anything
else.
You
guys
want
to
talk
about
all
right
cool.
In
the
case.
We
are
done
thanks.
Everybody
and
we'll
talk
again
next
week.
Actually
I
should
say
for
SDK
we'll
talk
in
two
weeks.
If,
for
some
reason
you
guys
do
want
to
have
a
call
next
week,
just
just
ping
me
so
and
then
we'll
I'll
announce
it
to
the
group
that
wrote
they
will
have
a
call,
but
usually
we
don't
have
enough
topics
to
warrant
I'll
call
every
week.