►
From YouTube: CNCF Serverless WG Meeting 2020-10-29
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
A
Yeah,
it's
it's
been
a
little
windy
up
here
in
raleigh,
but
nothing
too
bad
the
power
flickered
very,
very
quickly
once
but
nothing
too
bad.
B
Yeah,
that's
good.
We
had
like
60
miles
per
hour
winds
here
this
morning
it
was
yeah
and
my
dog
is
she's,
so
scared
of
the
any
sort
of
storm.
So
she
decided
to
like
jump
with
me
and
wake
me
up.
A
Yeah,
we're
supposed
to
my
wife,
was
telling
me
that
she
saw
a
news
report
that
said
yeah
we're
potentially
supposed
to
get
60
mile
an
hour
winds
up
here
as
well,
but
I
don't
think
I've
got
anything
quite
yet,
and
my
only
worry
is
that
I
have
two.
Actually
I
guess,
including
this.
I
have
three
phone
calls
where
I
kind
of
need
to
talk
a
little,
and
so,
if
I
lose
power,
that's
going
to
make
it
harder
and
one
of
them
is
with
a
customer.
So
that's
that's
not
going
to
be
good.
C
C
A
A
A
A
All
right,
hey
eric.
E
A
A
C
Yeah,
I
don't
hear
anything.
A
Okay,
okay,
perfect!
Thank
you!
It's
just!
It's
amazing!
You
can't
hear
that
all
right!
Lou
are
you
there
hi
hi
doc,
good
morning,
hello
good
morning
and
there's
someone
else
flying
by
oh
lance.
A
So
oh
heads
up
there's
a
storm
sort
of
on
the
east
coast
someplace
so
far
my
power
has
been
good,
but
if
I
for
some
reason
just
vanished,
it's
not
because
you
guys
pissed
me
off
or
anything
it's
just.
I
lost
power,
so
someone
else
would
need
to
kind
of
take
over.
At
that
point,.
A
F
Hey
clements,
I
I
have
I'm
so
narrowly
missing
the
out
on
being
able
to
present
something
to
you
all
guys.
All
you
guys
today
could
get
the
whole
work
done.
You
had
homework.
F
I
had
self-imposed
homework
with
the
replication
model
for
the
schema
registry,
but
I
am,
I
just
got
distracted
and
then
I'm
missing
an
hour
of
work
on
it.
So.
A
I'm
late
all
right
all
right,
one
last
attendee
and
then
we'll
get
started.
Scotty
there
howdy
howdy,
all
right.
Let's
go
get
started.
I
think
I
got
everybody
all
right
community
time
anything
from
the
community.
People
want
to
bring
up
all
right
moving
forward.
So
I
can't
remember
if
I
mentioned
this
last
week
or
not,
but
I
did
manage
to
get.
A
I
think
some
fairly
good
times
for
kubecon
office
hours
again,
if
you
yeah,
I'm
not
remember,
I
did
mention
it
if
you'd
like
to
volunteer
just,
let
me
know
and
I'll
add
your
name
to
the
list
as
we
get
closer,
the
nagging
will
increase.
So
please
think
about
that.
A
Sdk
calls
scheduled
for
this
week
as
of
right.
Now
there
is
nothing
on
the
agenda.
If
we
don't
have
anything
by
the
end
of
this
call,
we
will
cancel
interop.
So
I
did
do
some
minor
updates
to
the
interop
dock
itself.
Please
take
a
look.
If
you
get
a
chance,
nothing
major.
I
think
the
biggest
thing
there
is.
A
I
realized
that
we
probably
should
have
a
common
service
that
multiple
people
can
implement
so
that
we
can
do
some
testing,
because
if
it's
just
a
random
service,
it's
gonna
make
it
harder
to
to
do
any
kind
of
testing.
If
you
don't
know
what
the
heck
you're
actually
subscribing
to.
So
I
just
did
a
quick
little
ping
service.
Please
take
a
look
if
you
get
a
chance.
A
I
was
wondering,
though,
since
we're
supposed
to
be
doing
something
real
next
week
in
terms
of
interop,
whether
we
should
have
a
time
for
us
to
get
together
to
talk
like
maybe
tomorrow
or
something
or
maybe
monday,
just
sort
of
get
our
way.
He
was
planning
actually
coding
something.
What
do
people
think.
A
A
A
A
A
Yeah
and
we
have
to
have
the
the
interop
slack
channel
if
you're,
not
part
of
that,
let
me
know
and
I'll
add
you:
okay,
cool,
so
same
zoom
as
normal,
all
right,
cool
anything
else
relative
to
kubecon
or
discovery
interop
that
we
need
to
talk
about.
B
Yeah
just
quickly,
we
moved
our
website
to
netlify.
So
now
we
don't
have
the
silly
redirects
to
the
github
pages
anymore.
We
moved
our
team
meetings
now
to
weekly
meetings,
so
it
will
be
every
monday.
Now,
instead
of
every
other
weeks,
we
should
make
it
clear
and
just
doing
overall
improvements.
Our
issue
section
has
become
quite
large.
A
A
That's
right
and
I'll
teach
you.
Maybe
you
can
comment
on
what
he's
done
here
and
what
your
concerns
are.
A
E
Yeah
sorry
yeah,
you
prob
button
problems
it's
late
night
yeah,
so
he
switched
from
really
kind
of
focusing
on
vendor
implementations
of
security
protocols
and
went
ahead,
and
it
was
more
general
mostly
is
a
word
smithing
sort
of
a
complaint
that
what.
E
Oh
it
I
he
expresses
that
the
the
concerns
of
security
aren't
necessarily
a
part
of
the
core
interest
of
the
spec,
and
I
I
don't
think
that's
quite
capturing
what
I
would.
I
guess
the
position
of
the
group
is
on
it,
and
that
would
be
more
like
these
things
are
of
core
importance
and
we
need
to
support
them
being
possible.
E
There's
a
need
for
us
to
do
that
in
a
way
that
doesn't
king
make
there's
a
lot
of
difference
in
opinions
on
how
to
solve
these
security
problems
or
whether
some
people
want
to
solve
particular
security
problems
and
there's
a
lot
of
different
solutions
so,
and
and
not
only
that
as
these
solutions
evolve
over
time,
and
so
we
want
to
allow
for
the
ecosystem
to
continue
to
react
to
changes
as
they
shift
preferences,
and
I
didn't
feel
like
it
really
captured
that.
So
that
was
the
rough
bit
of
my
comment.
A
Okay,
but
you
said
it
was
mainly
more
wordsmithing
than
anything
else
right,
yeah,
okay,
okay,
I
don't
see,
I
need
still
on
the
call
aside
from
actually
let
me
go
ahead
and
show
you
so
hear,
eric's
comments.
You
guys
can
read
that
if
you
want
I'm
wondering
from
anybody
else
on
the
call
are
there.
General
concerns
are
people,
okay,
with
the
direction
that
anish
was
going
here.
F
I
think
that
captures
what
we
were,
why
we
scoped
out
the
security,
the
security
aspects,
okay,
because
we
didn't
want
to
sink
the
ship
by
by
pulling
in
security
right.
F
I
don't
think
that,
and
I
think
that
precludes
that
we
might
have
a
an
extension
that
will
define
it,
so
I'm
still
having
the
I'm
still
having
the
vague
notion
without
having
thought
about
it,
deeply
more
deeply
that
we
can
go
and
and
do
a
mapping
of
s
mine
to
cloud
events
where
we
would
effectively
use
s
mime
to
encode
the
body
of
a
message
and
then
would
have
exactly
like
an
email
have
additional
attributes
which
would
then
have
the
the
required
hints
that
we
have
about
algorithms
etc.
F
But
that
would
be
an
extension
that
doesn't
require
any
changes
to
the
core
spec
so
yeah.
So
I
think
this
really
reflects
what
we
what
we
want
to
do
and
that
we
didn't
want
to
have
the
w
start.
Our
security
nightmare
kind
of
dragged
into
the
house.
A
Okay,
in
that
case,
I'm
going
to
assume
that
anish
will
respond
to
you
eric
at
some
point
in
time.
You
guys
can
go
back
and
forth
on
the
word
smithing,
but
it
sounds
like
general
direction
is
okay,
so
we
just
need
to
wait
for
this
issue
or
this
set
of
questions,
or
these
set
of
questions
to
be
resolved.
A
A
End
call
I
figured
that
there
was
this
typo
yep
and
it's
clear,
which
is
a
typo,
should
be
data,
schema
any
objection
to
approving
this
one
I
apologize.
I
would
have
normally
just
accepted
this
one,
but
since
I
didn't
notice
it
till
last
night,
I
decided
to
wait
for
the
call.
A
All
right
this
one,
like
manuel,
I
was
going
through
some
of
the
interop
stuff
and
I
came
across
the
you.
This
is
into
the
discovery
spec.
I
believe
that
in
discovery
I
was
looking
at
the
url
field
and
it
says
it's
an
absolute
uri
to
reference
the
service.
A
However,
as
I
was
reading
that
because
it
had
been
watson
to
actually
play
with
this
stuff,
I
couldn't
remember
for
sure
if
it
was
the
url
to
the
service
definition
in
the
discovery
endpoint
or
the
url
to
the
service
itself,
and
this
sentence
that
I
have
highlighted
here
is
a
little
bit
ambiguous
related
to
that.
So
I
just
made
it
clear
that
reference
to
this
service
within
the
discovery
endpoint,
I
think
everything
else
is
basically
the
same.
A
Cool
perfect,
thank
you
all
right.
Next
and
clemens,
I'm
glad
you're
on
the
call,
because
I
think
this
is
mainly
in
the
discoveries
and
the
actually.
I
think
I
do
hit
the
subscription
spec,
so
this
one
again
was
because
of
the
interop
stuff.
I
was
doing
so
first
of
all
here
in
the
discovery
spec
in
the
subscription
config
section,
it's
a
map
of
key
value
pairs
where
I
believe
the
value
is
one
of
the
types
conforming
to
the
the
cloud
event
type
system.
Okay.
A
So,
rather
than
putting
value
here,
I
decided
to
change
it
to
to
the
word
type
just
to
make
it
clear:
it's
just
not
some
random
value,
okay
and
then
just
a
simple
typo
here.
So
this
isn't
your
part
yet
clemens,
but
for
everybody
else,
are
you?
Okay,
with
changing
the
word
value
to
type
just
to
make
it
a
little
bit
clearer.
A
Any
concerns
with
that
okay,
now
clemens
to
get
into
your
stuff.
Okay.
First
of
all,
I
was
assuming
that
id
needs
to
be
unique
within
the
scope
of
the
substitution
manager
and
then
back
up
a
little.
This
is
the
id
on
the
subscription
object
itself
yeah.
So
I
was
assuming
that
the
id
needs
to
be
unique
within
the
scope
of
the
subscription
manager
and
it's
immutable.
F
A
F
You
literally
have
like,
if
you
have
one
subscription,
if,
if
you
have,
if
you
have
a
subscription,
the
subscription
has
one
filter.
If
you
want
to
filter
on
something
something
yet
different.
A
F
Yes,
so
one
of
those
things
is
wrong,
so
I
the
intent,
the
certainly
the
intent
that
I
had.
Oh
I
I
know
what
discrepancy
comes
from.
We
were
working
on
this
to
with
a
few
folks,
and
I
didn't.
I
think
I
didn't
do
the
the
language
piece,
the
filter
language
piece.
I
forgot,
who
did
this
volunteers?
F
A
A
Okay,
cool,
okay,
we'll
circle
back
around
to
the
other
filter
stuff.
Next
was,
I
didn't
see
a
spot
to
put
okay,
let's
circle
back
up
so
up
here
we
have
subscription
config
with
the
new
list
of
name
value
pairs.
Okay,
I
did
not
see
a
config
section
in
the
subscription
resource.
Now
there
were
two
ways
to
go
here:
one
was
these
just
appear
as
top
level
entities,
but
I
got.
I
was
worried
that
that
would
conflict
with
the
extension
stuff.
A
So
I
thought:
okay,
maybe
there's
just
an
oversight,
and
you
guys
forgot
to
put
a
config
section
to
put
the
key
value
pairs
from
the
subscription
or
from
the
subscription
config
thing
in
the
discovery.
Endpoint.
A
A
A
A
F
A
It
depends
on
what
you're
setting
up
doesn't
it
so,
for
example,
if
I'm
subscribed
so
okay,
let
me
give
a
different
example.
Maybe
one
that's
easier
to
follow.
Let's
say
I'm
subscribing
to
a
cloud
object,
storage
and
I
need
to,
and
I
need
to
specify
what
bucket
to
watch.
Yes,
that
would
be
a
config
data.
F
That
is,
that
is
the
subscription
scope.
That
is
what
you
subscribe
to,
that.
That
is,
that
is
our.
That
is
what
we,
what
we
think
of
as
as
a
source.
A
F
Well,
we
need
to
define,
we
need
to
figure
out
how
let
me,
let
me
go
into
this
room.
F
E
A
F
All
of
those
so
the
the
subscription
there's,
I'm,
I
don't
think,
we've
we've
done
a
good
enough
job
here
to
be
maker,
make
it
concrete
with
the
what
the
scope,
how
we,
how
we
define
what
the
scope
is
for
the
the
subscription
manager
really,
but
the
subscription
the
the
the
subscription
operation
should
specify
the
scope
at
which
you
are
subscribing,
and
I'm
assuming
it
does
in
some
way.
Hang
on.
F
Scope
means
what
is
the
ob?
What
is
so
the
event
producer
we're
a
little
bit
washy
washy
here
three
two
four
one
is
on
of
the
original
document
is
the
create
operation
should
be
supported
by
compliant
event,
producers.
It
creates
a
new
subscription.
So
the
assumption
here
the
the
implied
assumption
is
that
you
effectively
the
event
producer
has
a
exposes
that
interface
or
someone
acting
on
behalf
of
the
event
producer
and
that
kind
of
implies,
and
that
needs
to
be
clear
that
that
you
have
a
subscription.
F
You
have
a
subscription
endpoint
where
the
endpoint
already
reflects
the
scope,
the
scope,
meaning
what
are
events
raised
for
so
meaning
you
would
walk
up
to
an
endpoint
that
already
reflects
your
storage
container,
and
then
you
know
it
has
that
encoded
in
the
uri
that
you're
talking
to
and
then
on
that
uri
on
the
container
itself,
you
asked
to
be
notified.
F
That's
the
that's
the
inherent
assumption
that
that
exists
there,
that
it's
not
that
it's
not
explicit
in
that
spec,
but
that's
kind
of
what's
in
that
that
was
at
least
that
was
in
my
head
when
I
wrote
it.
But
apparently
I
didn't
write
that
down.
F
Yeah
right
so,
and
that's
also
so
that's
also
how
so
the
subscription
you're
exactly
so
that's
what
the
subscription
url
is
effectively.
There
is
a,
and
this
is
the
interaction
with
the
disc,
with
the
discovery,
with
the
discovery,
respect
that
means
still
need
to
go
and
resolve.
So
this
is
a
good
discussion
to
have
effectively.
The
question
is
what
is
discoverable
in
as
a
as
an
entity
inside
of
a
system.
So
are
we?
Are
we
publishing
in
discovery
all
the
containers
of
that
store
of
of
of
the
storage
system?
F
We
should
because
it's
the
if
the
containers
are
the
ones
which
are
right,
which
are
raising
the
events.
Then
then
those
are
the
ones
that
shall
be
that
shall
be
published
and
they
should
have.
They
should
have
subscription
urls.
If
we
say
we
take
the
entire,
so
containers
or
buckets
or
whatever
they
are
right,
or
we
can
say
we're
gonna
allow
subscriptions
and
we're
gonna
publish
as
a
service,
the
storage
account
or
you
know,
whatever
other
entity
we
have,
then
that
thing
will
be
will
be
published
and
that
has
a
subscription
url.
F
So
I
think,
ultimately,
the
things
that
we
put
into
the
discovery
catalog
are
defining
what
the
scopes
are
on,
which
you
can
go
and
subscribe
and
then
and
you
can
go
and
filter
those
down
and
that's
why
we
have
all
the
metadata
there
right
and
then,
once
you
have
found
the
exact
object
that
you're
interested
in
in
subscribing
into
you
know,
knowing
all
of
the
events
that
it
has
and
all
those
things
and
that's
kind
of
what
the
discover
api
needs
to
give
you,
then
you
just
go
to
that
subscription,
url
and
say,
and
that's
where
you
issue
your
issue,
your
your
your
post
request
to
go
and
create
the
subscription.
F
F
F
There's
so
you
have
a
list
of
protocols
here.
This
is
in
discovery,
so
you
have
a
list
of
protocols
here
which
are
supported
by
that
by
the
by
the
publisher
or
sorry
by
the
whatever.
The
middleware
is
so
by
the
publisher
in
the
simplest
case,
but
by
the
middleware
in
the
delegated
case,
or
you
can
go
and
choose
from
so
you
have.
Let's
say
this
says
http
and
the
subscriber.
F
No
now
knows
that
http
and
if
we
want
to
list
https
https
or
if
we
want
to
say
if
it's
oh
https
only
which
I
would
prefer
this
is
subscriber
now
picks
one
of
those
protocols,
but
the
subscriber,
if
we're
assuming
a
push
model,
will
so
let's
do
that.
F
First,
the
subscriber
will
then
it
it
knows
what
its
parameters
are,
because
it
exposes
a
web
hook
and
that
web
hook
has
a
certain
set
of
parameters
that
need
to
be
configured
on
this
on
the
subscription
endpoint
to
for
the
subscription
to
be
established.
So
it
knows
its
own
uri.
It
knows
the
the
the
authentication
key
that
needs
to
be
set
and
whatever.
So
all
those
parameters
is
something
that
the
discovery
service
can't
know
and
they
can't
be
configured
because
it's
really
about
the
thing
about
the
subscriber
that
wants
to
go
and
subscribe.
F
Then
we
need
to
assume
that,
based
on
the
uri,
the
client
has
already
permissions
because
of
course
we
can't
tell
the
clients,
you
know
credentials
or
or
permission
data
in
here,
unless
you
want
to
go
and
embed
that
in
into
your
eye
using
some
kind
of
token.
F
But
but
if
you,
if
you
say
you
know,
here's
a
here's,
a
service
and
the
service
allows
you
to
to
obtain
these
these
following
events
and
they're
being
exposed
for
mqtt,
then
your
subscription
url
would
be
an
mqtt
topic,
and
that
implies
already
that
you're
going
to
use
the
mqtt
protocol
directly
and
not
the
subscription
api
to
go
and
obtain
those
events.
So
the
subscription
url
effectively
also
expresses.
A
A
Okay,
okay,
that
that
I
understand
okay,
you
want
to
kill
that.
However,
it
seems
to
me
yeah
and
it
seems
to
me
that
there
will
be
some
event-
producers
that
may
want
additional
configuration
right.
I
mentioned
one
for
the
ping
servers
in
terms
of
interval
and
you
question
whether
that's
a
real
event
source,
but
then
I'm
I'm
trying
to
steal
ideas
from
the
k-native
world.
So,
for
example,
another
one
I
think
from
the
k-native
world
is
you
can
subscribe
to
github,
and
one
of
the
parameters
you
could
pass
in
is
a
is
a
key.
A
In
essence.
That's
then
echoed
back
on
each
on
each
notification,
so
you
could
verify
that.
Yes,
this
actually
is
coming
from
github
as
opposed
to
some
rogue
entity.
Now,
that's
not
necessarily
a
transport
level
thing,
even
though
sure
it
appears
as
an
http
header.
I
don't.
I
don't
view
it
as
a
as
a
transport
level
thing,
because
it's
not
meant
to
influence
the
transport.
It's
just
an
extra
header.
That
happens.
F
A
F
F
Let
me
just
and
I'm
just
I'm
I'm
sitting
between
a
thousand
specs
that
I'm
that
I'm
currently
writing.
So
I
need
to
go
and
remind
myself
so
yeah
so
in
in
in
three
two
two
one
in
http
we
have
headers.
F
And
in
three
two
two
two
mqtt:
we
have
user
properties
and
in
nqp
we
have
you
can
set
link
properties
in
kafka.
We
you
can
extract
the
partition.
I
think
we
didn't
set
the
we
you
could
have
also
we
could
have
also
added.
We
probably
should
add
the
the
application
properties
there.
So
there
are
ways
to
go
and
set
these
values
in
these
various
transports.
F
So
if
you
want
to
go
and
set
a
set
so
yeah,
so
if
you
want
to
set
a
header
that
you
want
to
have
played
back
to
you,
then
you
can
do
this
with
that
with
the
http
settings.
F
So
what
is
so?
What
is
the
so
so
given
given
that
construct?
How
does
that
correlate
to
the
so
the
configuration
that
you
had.
A
So
to
me,
I
I
view
this
stuff
in
here
hp
settings
headers
in
particular
as
different
from
the
subscription
config,
because
this
is
saying
I
as
a
user,
I'm
sorry,
I,
as
a
receiver
of
the
event,
want
to
manipulate
the
http
request
coming
in
right.
I
want
to
add
this
specific
header
and
then
and
and
that's
a
very
sort
of
lowest
level
thing.
Okay,
I
view
these.
A
I
view
these.
Where
is
it
I
apologize,
I
good
guy.
Where
is
it
sorry?
I
viewed
the
subscription
config
as
something
different.
This
isn't
going
to
the
low
level
of
saying
hey.
I
understand
the
transport
because
I'm
a
complete
geek
and
I
need
to
add
this
http
header.
This
is
saying:
github
has
required
that
I
pass
in
a
key,
an
api
key
of
some
kind
or
a
token.
I
guess
that's
what
I
called
now.
What
github
chooses
to
do
with
that
token
is
completely
up
to
it.
Okay,
in
this
particular
case.
A
Yes,
it's
going
to
echo
it
back
to
me
as
an
http
header,
but
that's
sort
of
an
implementation
choice
right.
It
may
choose
to
do
something
else
with
this
bit
of
information.
So,
for
example,
I
may
be
passing
in
whenever
I
don't
want
to
go
there
I'll
confuse
it.
A
F
Okay,
but
but
how
does
that?
How
does
how
does
that
materialize
in
the
in
the
flow
back
to
the
to
the
subscriber.
A
It
may
or
may
not
right
in
this
particular
case.
I
agree
with
you
that
it
does
kind
of
materialize
itself
back
as
an
http
header,
but
I
assert
that
there
are
other
configuration
options
that
are
that
don't
necessarily
manifest
themselves
as
like
a
separate
property
in
the
event
that
gets
sent
to
you
it
just
it
tweaks
the
way
the
producer
is
going
to
talk
to
you
so,
for
example,
the
ping
source
is
an
example
right.
It's
only
going
to
talk
to
me
every
five
seconds
versus
10
seconds.
F
A
So
my
assumption
was
that
a
a
person
who
wants
to
subscribe
will
look
at
this
field
and
say:
oh
there's,
a
key
called
interval
and
it's
of
type
integer.
Now
what
that
actually
means,
I
don't
think
we
can
express
here.
They
have
to
go.
Look
at
the
documentation
to
say:
oh
interval
means
how
often
to
send
the
ping
right
and
that,
but
they
can
at
least
programmatically
now
once
they
decide
they
actually
want
to
use
it.
F
A
Yes,
I
was
asserting-
or
I
was
assuming
what
you
guys
had
written
was
there
would
be
something
like
a
config
bucket,
with
the
name
of
with
a
set
of
name
value
pairs,
where
the
key
would
be
the
config
name,
and
the
value
would
be
the
value
that
you
wanted,
that
you
want
that
you're
trying
to
set
to
so.
In
my
case
it
would
be
interval
colon
five
for
a
ping.
Every
five
seconds.
A
A
F
So
I
have,
I
have
a
use
case.
I
have
a
use
case
in
my
head
that
comes
from
a
scenario
that
I
have
a
scenario
for
your
functionality.
I'm
not
convinced
of
your
of
your
scenario,
but
I
have
one
okay.
F
Let
me
let
me
just
give
you
that
briefly:
okay,
cool
yeah,
so
in
opc
ua
for
everybody,
who's,
listening
industrial
standard
standards
for
industrial
stuff
talking,
they
have
a
notion
of
subscribing
to
very
fast
streams
of
data
where
you
want
to
do
a
bit
of
aggregation
before
you
want
this
or
you
want
to
only
have,
or
you
want
to
go
and
configure
whether
you
want
to
have
all
events
or
whether
you
want
to
have
keyframes
that
partial
events
or
keyframes.
F
So
you
want
to
go
and
and
modify
the
subscriber
the
the
the
subscription
somewhat
relative
to
the
event
stream.
That's
coming
on,
so
it's
not
full
on
aggregation,
but
it's
kind
of
some,
some
modification
of
the
events
that
you're
getting
so
either
you're
getting
sparse
ones
or
you're
getting
key
frames
and
you
have
a
keyframe
keyframe
interval.
F
So
that
seems
to
be
so.
From
that
perspective,
it's
it's.
It
seems
to
be
a
mechanism
that
works.
We
I
just
find
the
I
just
found
the
the
the
you
know,
the
modification,
the
modification
of
the
producer.
I
just
found
that
a
little
weird
as
an
example,
but
at
least
it
helped
me
remember
the
the
keyframe.
A
F
Yeah
so
so
their
scenario,
their
scenario
is
that
they
have
a
lot
lots
of
events
and
sometimes
the
data
for
an
event
would
be
300
different
fields
and,
of
course,
it's
terrible
to
go
and
translate
them
all.
If
they're,
not
all
change
change
that
much
so
they
would
only
transfer
the
delta
with
each
event.
Whenever
there's
a
change
and
then
every
10
seconds
they
would
go
and
force
one
full
event
to
be
sent
so
that,
if
you,
if
the
subscriber
comes
late
into
the
stream,
that
eventually
they're
going
to
get
all
the
data.
F
E
F
A
Okay,
let's
go
and
go
to
the
queue
scott
you're
up.
First,
I
think.
D
D
All
triggers
all
right,
because
you
wanna
you
wanna
find
out
which
of
the
like.
Why
are
you
getting
these
events
and
maybe
there's
some
decoration
that
can
happen
for
that
particular
request
to
subscribe
to
already
existing
event,
sources
not
produce
an
event
source
because
of
a
subscription
which
is
kind
of
backwards.
A
A
A
Yes,
if,
if
you're
talking
about
k
native
so
that's
interesting,
because
because
I
could
see
an
argument
that
says
the
the
pink
source
is
just
weird
and
that
doesn't
fall
into
this
model.
I
don't,
I
don't
agree
with
it,
but
I
can
understand
that
argument
to
say
I
can't
pass
in
the
api
key
for
github.
That
seems
like
we're
broken.
D
Well,
because
you're
talking
to
a
delegate
and
the
delegate
needs
to
be
created,
I
think
if
we,
if
we
removed
k
native
from
this
picture-
and
we
were
just
talking
to
github-
then
I
think
you
would
use
the
subscription
config
to
pass
in
all
the
api,
keys
and
stuff
it
I
mean
like
so.
This
is
where,
like
k
native
becomes
a
little
fuzzy,
because
ping
source
is
a
creation
of
a
thing.
D
It's
good
in
the
github
case.
I
would
expect
that
to
be
translate
so
so
there's
some
service,
that's
exposing
the
discovery
api
on
the
k-native
cluster
that
translates
into
maybe
the
the
simple
case
would
be
a
github
source,
a
new
channel
for
you
and
a
subscription.
A
You're
using
canadian
terms,
I'd
rather
leave
key
native
out
of
it.
Then
right
if
I
was
talking
directly
to
github
and
I
wanted
to
subscribe
and
github-
requires
that
I
pass
in
an
api
key.
I
think
you
said
you
agree
that
it
would
be
passed
in
under
the
subscription
config
block.
D
A
D
It
gets
more
complicated
because
the
discovery
api
could
expose
subscribing
to
a
github
web
hook,
but
it
could
also
expose
the
ones
that
are
created
already
for
you
that
are
available,
because
we
we
proxy
all
of
the
existing
web
hooks
at
the
moment.
So
if
you
look
at
that
you,
you
would
see
several
instances
of
github
sources
that
have
been
configured
and
then
maybe
this
generic
one
that
helps
you
make
all
the
mechanics
of
trying
to
get
those
events
to
you.
A
D
A
Don't
leave
but
leave
kane
out
of
out
of
this,
because
I
feel
like
you're
you're,
mixing
up
setting
up
subscription,
meaning
talking
to
the
event
producer
itself,
you're,
mixing
that
up
with
the
idea
of
a
canadian
specific
thing
that
says:
oh
you're
already
getting
events
into
the
system.
You
just
want
to
get
them
routed
to
another
service.
D
I
think
this
is
exactly
why
I've
always
pushed
back
on
k-native.
Is
it
doesn't
transparently,
look
like
github,
it's
always
a
k-native
thing,
and
so
in
that
case,
the
discovery
api
could
expose
a
straight
github
subscription
api,
which
would
take
a
token
and
then
there's
these
k-native
github
things
that
don't
take
a
token,
because
they're
already
created.
A
And
that's
fine!
The
work
we're
talking
about
here,
I
think,
is
the
former,
which
is
we
are
just
talking
about
how
to
subscribe,
to
github
directly,
not
to
some
some
github.
I'm
sorry,
some
some
k
native
implementation,
choice
that
may
or
may
not
expose
different.
You
know
set
of
configuration
parameters.
A
D
Okay,
yeah
and
then
in
that
case
I
think
the
subscription
config
would
include
things
like
what
tokens
are
required.
A
Right
and
so
then
I
would
go
back
to
clemens,
say:
okay,
if
if
we
agree
that
that
is
a
valid
scenario
right,
where
the
the
person
doing
the
subscribe
to
github
wants
to
pass
in
or
needs
to
pass
in
a
token,
and
that
appears
as
a
subscription
config
thingy
from
the
discovery
endpoint.
Where
does
that
appear
in
the
subscribe.
H
Yeah,
just
when
I
hear
you
talking
about
the
what
the
subscription
config
is
and
like
basically
that's
gonna
define
the
object
that
you
send
when
you
try
to
create
the
subscription
for
what
I
understand
isn't
like
right
now.
It
seems
a
little
bit
limited
like
limited,
don't
don't
we
should
just
say
the
subscription
config
is
just
like
a
csun
schema
that
define
what
you
can
pass
as
a
config.
H
So
if
you
want
to
pass
like
complex
object,
you
can
and
then
basically
it's
up
to
the
subscription
endpoint
to
define
what
you
need
to
pass
and
then
it's
still
like
super
flexible.
So
when
we
implement
we,
we
maybe
refine
it
but
to
keep
it
open
for
now,
because
I
have
the
feeling
that
we
create
limitation
in
the
spec.
A
H
Sorry,
yeah
again
my
volume.
I
really
need
to
fix
that
in
my
mic,
yeah
the
I
was
just
saying
that,
basically,
if
we
want
to
keep
it
open,
we
should
just
put
a
decent
schema
as
subscription
config
that
describe
how
like
what
type
of
object
you
need
to
pass
to
when
you
subscribe
and
then
the
config
needs
to
match
the
subscription
config
schema.
F
F
So
putting
an
extra
schema
there
or
trading
creating
complex,
complex
objects
in
that
place
seems
to
be
a
little
much
for
my
taste.
H
But
for
me
like,
it's
really
depend
on
the
people
who
do
the
implementation,
so
I
would
agree
with
you
that
the
best
the
best
is
to
keep
it
simple,
but
yeah.
F
F
How
much
do
we
want
really
want
to
go
and
specify
concretely
so
there's
one
thing:
there's
so
here's
here's
one
piece
that
is
that
that
might
be
contentious
in
terms
of
of
whether
you
want
to
go
and
bolt
this
down
as
a
as
a
first
class
thing,
but
which
would
probably
be
helpful
and
would
probably
be
something
that
most
people
will
use,
and
that
is
where
would
we
put
a
refresh
token
or
why
would
we
declare
that
we
want
to
have
a
refresh
so
a
free
press
token
and
an
oauth
authorization
sts,
because
ultimately
we
will
have
to
have
one,
and
we
will
have
to
have
that
in
this
game
and
that's
a
parameter
that
we
have
to
go
and
and
declare
we
will
have
to
go
and
and
have
a
way
to
pass
a
refresh
token
in
the
respective
url
of
the
oauth
sts,
which
can
go
and
and
trade
that
refresh
token
for
an
access
token.
F
But
then
the
question
is:
do
we
do
we
want
to
go
and
pass
this
in
in
these
subscription
country
parameters
which,
where
they
would
go
right
now,
because
they
are
not
protocol
dependent,
they
can
be,
they
can
be
used
with
with
you
know,
multiple
protocols
and
they
will
be
mapping
to
different
constructs
in
the
protocols
or
do
we
say
no,
no,
no,
we're
going
to
go
and
make
that
the
first
class
thing,
because
my
inclination
is,
you
would
have
to
go
and
say
you
would
have
to
go
and
say
in
the
in
the
discovery
spec
here
you
would
have
to
go
and
say
no.
F
This
is
an
oar
thing
and
and
by
the
way,
this
is
an
all
thing
where
you
have
to
go
and
you
have
an
sds
endpoint
where
you
have
to
go
and
trade
in
your
refresh
token,
and
that
should
probably
be
in
the
air
and
then
there
needs
to
be
some
url
that
you
have
to
go
to
to
go
and
obtain
the
refresh
token
that
you
can
then
go
and
configure
with
your
with
your
with
your
subscription.
F
H
I
think
it
became
really,
even
if
I
do
agree
that
it
might
be
like
the
common
case.
For
me
as
a
as
a
spec.
We
need
to
let
it
open
and
it's
easier
to
say
for
now
and
maybe
like
a
new
version,
will
basically
start
adding
those
kind
of
constraints
but
yeah.
I
would
just
see
it
as
like
easier
to
make
it
open
and
then
maybe
close
some
stuff
down
later
and
the
nice
thing
with
the
grayson
schema
is
then,
if
you
want
to
do
it
to
build
a
ui
on
top
of
it.
A
F
Yeah
I
mean
I,
I
think
this
is
an.
I
think
this
is
something
where
you
would
really
def.
You
would
really
modify
the
behavior
of
the
subscription
per
se
where
you
want
to
go
and
outfit
that
with
particular
behavior
and
it's
something
that
is
specific
to
not
the
producer,
but
really
the
subscription
endpoint.
A
Well,
I
view
it
as
it
could
be
either
one
and
that's
implementation
detail
right,
because
the
configuration
knobs
here
that
I
think
that
you
could
look
at
them
as
strictly
as
modifying
the
configuration
manager,
I'm
sorry
subscription
manager
or
you
could
look
at
them
as
modifying
the
producer
itself,
but
from
the
end
user's
perspective,
I
don't
think
it
matters
to
them
all.
They
know.
Is
I'm
given
this
flag
and
I'm
and
I
can
set
it
and
magic
happens
behind
the
scenes
right.
A
D
Make
this
a
url
for
a
schema
registry
instead
of
trying
to
put
put
kind
of
like
the
half
settings
here,
just
like
the
schema
for
the
data
payload
and
cloud
events.
We
say
the
subscription
has
a
schema
it's
over
here.
If
you
would
like
to
look
at
it,
I'm
just
helping
you
find
that
host
and
its
open
api,
endpoint
or
whatever
it
is,
and
then
you
and
then
you.
F
D
Of
it,
I
think
it's
a
helper
pointer
to
the
let's
say
the
subscription
api
has
an
open
api
document.
This
would
be
the
way
to
expose
for
this
subscriber.
It
says
if
you
would
like
to
understand
how
to
hit
this
rest
in
point
or
whatever.
It
is.
Here's
my
open
api
document
that
you
can
go
look
at,
so
you
can
actually
make
it
programmatic
it
has.
It
has
one
assuming
it
has
one.
F
F
D
We
could
use
a
url
here
or
we
change
what
the
definition
of
subscription
config
means
and
one
of
those
one
of
the
known
keys
is
open,
api
or
json
json
schema
or
something,
and
then
it
points
to
a
url
that
you
can
go
next
look
up.
H
I
F
I
have
four
I
have
for
the
subscription
object
that
you
go
and
create
when
you
create
a
subscription.
I
have
an
id.
I
have
a
protocol,
a
protocol
settings.
I
have
a
sync
and
I
have
filters,
so
the
sync
is
the
url
that
you
go
to
protocol
settings,
configure
that
in
more
detail
what
the
protocol
choice
is.
I
have
the
idea
and
have
the
filter
so
base.
So
I
think
what
what.
F
Can
you
scroll
down
doug
in
what
you
added
there
yeah
try
to
sync
and
that's
the
john
f.
That's
the
config!
You
added
here
right,
yeah,.
A
Yeah,
this
is
new.
I
think
everything
else
is,
except
for
the
singular
here
most
yeah.
Well,
I
put
dialect
inside
there,
but
yeah.
Most
of
this
is
is
put
the
spec
this
stuff
right
here
is
new
and
we're
running
a
little
low
on
time.
Just
so
yeah
you
can
wrap
this
up.
Well,
not
finish
it,
but
at
least
I
have
last
comment,
I
should
say.
F
I
think
I'm
yeah,
I
think
what
you're
doing
there
is,
is
so
with
the
scenario
with
the
scenario
that
I,
with
this
configure,
configure
the
the
subscriber
scenario
with
your
key
and
keys,
keyframes,
etc.
I
think
that's
a
valid
one
and
I
think
the
solution
that
you
have
here
is
also
not
terrible.
A
Well,
I'll
tell
you
what?
Obviously
we
don't
want
to
rush
into
a
decision?
Everybody
think
about
this.
Maybe
we'll
talk
a
little
bit
more
about
this
on
monday,
if
not
obviously
next
thursday,
but
we
are
out
of
time.
I
know
people
need
to
leave
so
anyway.
Think
about
this.
The
pr's
out
there
I'm
going
to
make
some
changes
based
on
what
we've
said
already
so
keep
an
eye
out
for
another
version,
but
I
think
that
was
probably
the
biggest
change
was
the
config
thing.
A
A
Yeah
I'll
definitely
do
that.
That's
a
good
idea:
scott,
okay!
Okay,
let's
see
jim,
I
saw
you
in
chats.
Are
you
there
josh
you
there.
A
A
All
right
cool
all
right,
thank
you.
Everybody
good
lively
discussion
and
we'll
talk
again
at
least
to
some
of
you
on
monday
and
everybody
else
next
thursday.
Oh
the
musical
chat
cloud
events
sdk
nothing
on
the
agenda,
so
anybody
want
to
bring
anything
up
or
should
we
cancel
the
call.
H
I
think
we
can
cancel.
I
have
some
modifications
that
will
be
pending,
but
I
prefer
to
finish
our
discovery.
Stuff.
Okay,
based
on
that,
I
probably
have
like
big
pair
pr,
for
instance,.