►
From YouTube: Kyma Knative Integration WG meeting 20181220
Description
Meeting Notes: https://docs.google.com/document/d/1BZoJQ5qsSudlix8PXjykQfsJtfwC7EgUy0BP3oORDsA/edit#heading=h.ntfsragiqg5r
A
B
A
A
B
So,
as
you
probably
know,
there
was
multiple
face-to-face
meetings
arranged
for
Canadian
groups
and
we
were
basically
interested
to
join
the
eventing
on
the
event
in
group.
Had
people
from
Red
Hat,
of
course,
with
Google
right
at
pivotal,
IBM
VMware
yeah.
Mainly
these.
They
were
also
a
couple
of
guys
attending
from
Bloomberg
who
have
a
team
with
a
data
pipeline
for
machine
learning
pipeline
and
also
direct
colleagues
on
the
founder
of
that
also
was
attending
the
agenda.
I
can
I
can
add
the
agenda
meeting
notes
there
to
the
agenda.
B
B
B
B
However,
the
discussion
went
a
bit
different,
so
big
part
of
the
discussion
took
place
about
the
current
channels
and
whether
or
not
they
want
to
have
stuff
like
even
types
as
something
that
is
related
to
the
sources,
or
there
were
ideas,
also
from
Mark
Fisher,
from
pivotal
about
having
that
as
part
of
the
related
to
the
challenge
themselves
and
differently,
and
in
all
cases,
it
was
clear
that
everybody
has
his
own
interpretation
of
what
the
can't
eventing
need
to
support.
Everybody
had
different
ideas
of
our
different
interpretation
of
what
is
an
even
source.
B
For
example,
some
of
them
had
an
idea
that
it's
something
a
very
dynamic.
Some
had
the
ideas
that
it's
strongly
typed
with
some
even
types
and
stuff
like
that
and
as
an
action
item
out
of
the
long
discussion
there
was
that
we
need
to
work
on
our
even
sources,
our
East
case.
Sorry.
So
the
first
action
item,
which
also
was
followed
up
upon
yesterday
on
the
eventing
working
group
meeting,
is
that
there
need
to
be
people
working
on
the
use
cases
from
their
side.
B
They
had
someone
from
the
product
joining
their
and
they
asked
also
from
different
companies
to
have
someone
who
works
on
use
cases.
For
these
days.
Use
cases
I
would
say
the
other
items
there
is.
There
was
also
discussion
about
the
container
source,
which
is
a
generic
source
for
containers
that
probably
have
like
a
generic
will
hook.
Containers
I
would
say
so.
In
that
way,
you
can
have
a
hook
source
that
you
can
deploy
based
on
different
docker
images
and
every
docker
image
just
needs
a
configuration.
B
C
B
B
We
also
need
to
figure
out.
How
can
we
propagate
the
metadata
from
the
source
to
the
channels
whether
this
metadata
will
be
typed
or
like
strongly
tightly
coupled
to
that
source
or
be
configurable
yeah?
All
these
things
were
later
discussed.
For
me,
what
was
interesting
is
that
it
seems
that
everybody
is
always
doing
the
same.
So
almost
everybody
there
has
a
fast
platform
of
some
sorts,
have
IBM
open
risk
pivotal
with
their
ref
and
which
commercial
school
I
think
FPS
retired,
with
open
chest
something
integration,
something
on
top
of
it.
B
Jeff
specifically,
everybody
had
even
VMware
guys,
the
guy
who
contributed
Kafka
the
the
new
Kafka
channel
provisioner.
He
actually
was
in
VMware
working
in
that
platform,
that
is,
let's
say,
similar
to
Kemah
they're,
fast,
offering
and
once
they've
discovered,
K
native.
This
is
okay.
The
people
who
are
working
on
the
lower
layer
will
stop
working
on
the
product.
They
will
work
on
Kennett
of
itself
and
will
even
preach
K
native
insights
VMware,
so
he
focused
totally
his
work
on
Kenneth.
B
That
was
interesting.
The
other
also
part
is
that
they
are
very
open
to
people
contributing
and
joining
their
work,
so
they
are
really
eager
to
have
people
joining
all.
What
we
need
to
do
is
just
yeah
figure
out
something
that
we
want
to
do
their
talk
to
them.
On
slack,
then
the
working
group
meeting
and
and
and
just
yeah
joined,
contribute,
and,
as
also
you
know,
we
already
have
an
approval
for
contributions
from
our
side
on
our
private
time
as
kheema
developers.
We
just
need
to
align.
B
What
do
we
need
to
do
and
that's
it
so
this
is
also
part
that
was
probably
very
interesting
at
the
end.
It
seems
that
also
was
shared
on
the
meeting
internally
with
Villa
that
the
production
came.
It
was
probably
not
gonna
be
around
until
around
july
2019
or
q3
2019,
that's
when
they
will
have
was
probably
a
production
canadian.
A
A
B
A
A
C
So
basically,
I
want
to
see
that
I
was
looking
at
what
images
we
are
using
currently
and
which
is
covered
that
the
the
version
of
time,
which
is
that
we
are
using
for,
was
serving
an
event
think
it
is
something
that's
a
bit
older
than
what,
for
example,
Radu
built
based
on
his
pull
request
for
this
NAT
streaming
base
for
regional.
So
for
now
we
can
of
course,
upgrade
the
images
but
I
think
going
further.
C
We
need
to
maybe
come
up
with
a
strategy
that
how
we
should
decide
on
the
upgrading
images
and
what
all
things
we
need
to
take
care
of
that.
So
it's
not
a
big
topic,
but
I
think
this
might
be
important
thing
for
us
as
we
target
the
future
disease
of
kima.
So
definitely
we
need
to
update
first
thing
currently
now
to
to
the
version
where
we
can
have
this
Nats
streaming
original,
compatible
images
and
then,
after
that,
maybe
periodic
updates.
A
Okay,
so
for
now
we
will
need
to
I
guess
we
need
a
new
version
of
kinetic
for
ahtur's
work
right,
because
we
need
this
Yanis
in
a
source,
so
they
would
need,
will
not
just
need
to
bomb
the
version,
I
guess,
but
we
would
also
need
to
prepare
some
proposal
how
we
would
decide
on
bumping
cognitive
in
the
future
and
that's
what
you're
talking
about.
Yes,.
C
And
the
other
thing
is
I
also
discovered
that
the
actual
provisional
images
are
also
currently
not
there
on
the
juicier.
So
maybe
we
need
to
figure
out
that,
maybe,
along
with
Radu,
that
how
we
can
first
of
all
put
them
there
and
then
we
should
be
able
to
he
just
ate
that
we
can
use
them
in
kima
event,
Inc,
okay,.
A
C
D
Application
was
has
them,
you
need
to
get
the
real
version
of
that
kinetic,
antique
and
also
all
the
other
stuff,
so
that
for
them
to
work
on
this
POC
for
the
ACP
I
have
got
an
intermediate
version
which
is
working.
You
know
for
you
ending
and
also
for
the
controller
part
and
all
this
stuff
which
are
coming
from
kubernetes,
but
we
will
seen
and
for
kima.
We
have
to
agree
that
we
will
test
our
tentative
integration
version
and
then
we
will
install
it.
It's
our
duty
to
get
to
see
that
all
is
working.
D
D
B
C
C
C
A
A
Okay,
so
let's
jump
to
the
next
one.
I
just
wanted
to
ask
how
many
of
us
and
how
many
important
topics
do
we
have
to
discuss
next
week.
So
maybe
we
can
prepare
besides
now
if
the
topic
will
take
place
next
week
or
not
because
maybe
if,
if
there
are
no
fraction
I,
if
there
are
no
no,
not
none
of
us
next
week,
maybe
it
doesn't
make
sense
to
even
plan
it
I'm
defiantly
there
who
else
plans
to
have
something
to
say
what
do
you
think?
A
C
I'm
here,
but
at
least
for
now,
I
don't
have
a
topic.
It
might
come
up
sometime
in
the
course
of
the
week,
but
as
a
side
note,
I
want
to
kind
of
start
with
preparing
a
kind
of
a
proposal
about
how
we
want
to
adopt
the
keema
eventing
to
use
kennedy
wenting
and
what
all
the
challenges
and
what
all
the
missing
pieces
we
see
so
that
we
can
derive
some
kind
of
task
and
work
items
out
of
that.
A
Okay
and
one
more
thing
I'd
like
to
ask
you
maybe
does
notify,
is
that
planning
0.7
is
coming
soon
and
we
need
to
prepare
stuff
for
as
an
input
for
the
planning
and
I
suggest
to
prepare
what
you
preferably
would
like
to
do
on
the
next
further
next
release.
And
so
let's
talk
on
the
next
meeting
about
the
stuff
you
want
to
pull
in
or
just
our
proposed
on
the
plying.
A
A
E
So,
let's
start
by
the
two
use
cases
we
want
to
generate
a
cover,
so
the
first
use
case
is
the
use
case
that
we
currently
have,
which
is
having
kima
running
on
top
kima
running
with
Nats
streaming
included
already
and
event.
Bosses
is
working
as
usual.
So
that's
the
first
use
case
and
the
second
use
case
is
to
run
kima
on
top
of
Kay
native
and
also
can
a
tiff
eventing
and
and
have
messages
going
through
kima
2k
native
eventing.
So
we
prepared
the
high
level
design
for
that.
E
So
the
main
idea
here
is
to
have
the
event
bus
communicating
to
a
provision
ER
on
on
Canada
side
that
is
implementing
the
the
K
native
eventing
abstraction.
So
we
are
agnostic
about
that
provision,
or
it
doesn't
matter
it
would
be
Kafka
nuts
whatever.
So
we
are
just
communicating
to
a
provision
or
that's
its
vetting,
the
satisfying
the
the
abstraction
of
key
native
eventing.
A
E
Okay,
so
if
we
also
take
an
even
higher
look
on
the
high
level
design,
we
can
see
that
we
can
have
multiple
provisions.
At
the
same
time,
an
event.
Bus
can
also
send
messages
and
having
these
provisioners
also
deliver
messages
to
ever
subscribe.
So
we
thought
about
how
to
achieve
that.
We
decided
that
it's
not
the
responsibility
of
the
event
bus
to
discover
whatever
provisioners,
so
it
will
be
like
the
responsibility
of
the
publisher
itself.
E
So
when
a
publisher
publishes
a
message
to,
if
us
it
has
to
specify
three
things,
so
basically
the
source,
ID
and
the
vent
type
we
know
about,
and
the
third
thing
would
be
the
provisioner,
and
we
are
now
talking
about
even
key
native,
even
think
so.
The
first
use
case
when
keema
is
running
alone,
not
on
top
of
que
native
publishers,
will
will
only
use
the
source,
ID
and
event
line.