►
From YouTube: ☀️ GSoC CloudEvents plugin mentor meeting 2021.7.19
Description
Discussion on architectural choices for the plugin and a decision to start attending the CDF Event SIG 😊
A
A
How
can
we
or
how
are
we
thinking
of
dealing
with
maybe
transient
failures
or
just
like
simple
network
failures
and
sending
requests
for
jenkins
as
a
source,
but
also
for
jenkins,
is
saying:
how
can
we
implement
an
architecture
which
can
persist
events
and
and
also
create
a
way
for
any
source
which
is
sending
events
to
jenkins
sync,
to
really
have
that
architecture
where
events
are
not
being
lost
and
every
time
there
is
something
which
is
capturing
the
events
and
dealing
with
whatever
action
needs
to
be
taken
at
maybe
like
an
asynchronous
time,
because
I
keep
going
back
to
that.
A
One
thing
you
said,
and
that
was
really
that
really
resonated
in
terms
of
the
architecture
was
loosely
coupling
and
if
we
basically
like
design
a
system
where
it's
sort
of
like
an
rpc
call
between
jenkins
as
a
source
and
whatever
sync
and
jenkins
as
a
sync
and
whatever
source,
and
it's
very
like
dependent
and
it's
very
tightly
coupled
with
another
service
being
available,
especially
for
jenkins
and
sync
or
I
mean
jenkins-
is
a
source.
A
We
really
want
that
other
service
to
be
available
when
we
are
sending
to
do
something-
or
maybe
when
we're
talking
about
jenkins
as
a
sync,
we
really
want
the
sync
to
be
available
whenever
another
source
is
sending,
and
I
think
its
implementation
for
jenkins
is
the
sink-
is
way
more
important
than
jenkins
as
a
source,
because
you
know
someone
really
might
depend
on
taking
an
action
inside
of
jenkins
whenever
a
cloud
event
is
triggered
in
that
particular
source.
So
just
been
thinking
about
that,
there
is
actually
a
plug-in,
which
is
it's.
A
We
pops
up
like
pops
up
light
plug-in
for
jenkins,
and
it's
I'm
not
really
sure.
If
a
publish
subscribe
pattern
might
work
here,
having
something
like
a
simple
queuing
system
might
work
or
if
we,
if
we
even
want
to
you
know,
go
that
route
for
now,
maybe
just
like
to
hear
what
your
thoughts
and
that
are.
If
you
have
suggestions.
B
How
can
I
say
this,
how
much
you
guarantee
the
ordering
of
them
in
many
ways
and
and
then
how
long
you
persist
them
for,
and
I
think
that's
a
whole
series
of.
I
think
there
are
a
couple
of
different
ways
of
doing
it,
but
I
think
publish
subscribe
is,
would
be
certainly
a
good
way
to
start.
A
A
Basically,
I
like
what
in
my
head,
I'm
thinking
what
it
might
look
like
is
you
know
someone
is
creating
a
user
who
wants
to
use
jackets
as
a
sync
is
creating
a
topic
inside
of
inside
of
that
jenkins,
that
that
bus,
which
is
going
to
be
different
from
the
the
jenkins
as
a
sync
implementation
or
the
jenkins
in
the
same
configuration,
because
I
don't
really
think
that
putting
the
bus
alongside
or
like
the
bus
or
the
or
just
the
topic
configuration
with
the
actual
plugin.
A
B
B
B
No,
I
I
think,
there's
like
a
number
of
ways
in
which
you
can
deal
with
how
you
process
streams,
how
you
store
streams
like
how
streams
are
transported
like
how
are
you
directly
message,
brokers,
event
logs
and
then
how
you
process
them
afterwards,
like
there's
so
many
different
ways
to
approach
this
that
yeah,
it
is
kind
of
like
a
really
an
open
question,
because
you
want
to
leave
it
like.
Is
that
going
to
be
something?
That's
left
that
we
want
to
leave
configurable
for
an
end
user?
Is
that
something
that
we
decide?
B
I
guess
if
it's
just
jenkins,
it's
something
that
we
do
need
to
decide
on,
and
it
should
make
sense
for
us
the
icd
pipeline,
but
is
that
something
that
should
be
decided
more
at
the
level
of
like
the
event
seg
like?
I
really.
I
really
don't
know
what
have
you?
What
have
you
been
reading?
Just
out
of
curiosity.
A
Well,
I
so
all
of
the
the
usually
the
pops
up
patterns
of
design
for
any
event,
sort
of
mediator
or
just
a
middleware
any
any
service,
whether
it's
from
azure
or
whether
it's
even
a
simple
thing
like
a
kafka
implementing
something
like
that
is.
Is
that
usually
needing
a
topic
to
be
configured
and-
and
you
know
we
cannot-
I
mean
there
should
be
a
way
that
will
be
open
for
the
users
to
mention
whatever
topic.
A
A
But
there
is
that
you
know
middle
middleware,
which
is
sitting
in
between
receiving
all
of
those
topics
and
then
based
on
maybe
the
filters
or
based
on
the
topic
or
just
based
on
here
for
canadian
eventing.
They
have
they
have
triggers,
which
is
which
they
say
that
it
it
represents.
A
desire
to
subscribe
to
events
from
a
specific
broker,
so
for
them
trigger,
is
something
which
is
sort
of
like
sitting
at
that
level
between
broker
and
the
actual
whatever
is
going
to
be
working
with.
A
That
will
event
so
they're
filtering
based
on
the
they
have
different
brokers,
maybe
configured
with
you,
know,
different
sort
of
maybe
metadata
or
whatever
a
broker
name
a
broker
type.
I
think
it's
usually
broker
name
yeah.
It
is
broken.
B
A
So
so
what
so?
It's
basically
like
the
trigger
is
sort
of
sitting
at
that
broker,
and
it's
so
when
we
have
that
trigger.
It's
basically
saying
that
I
want
to
receive
event
from
this
particular
broker,
which
is
so
there
is
a
default
broker
and
I
think
they
call
it
default.
So
there
is
a
default
broker
and
then
there's
broker
that
that
someone
can
create-
and
you
only
maybe
as
a
publisher,
only
send
events
to
that
broker
say:
let's
call
it
k
native
trigger
or
no
no,
let's
say
it's
like
a
pipeline.
A
Tecton
pipeline
run
broker
and
I'm
sending
my
events
to
this
broker
and
then,
as
a
trigger,
I've
said
that
as
soon
as
this
broker
receives
any
event
I
want
to.
I
want
to
subscribe
to
to
that,
so
that's
actually
creating
a
trigger
like
subscribing
to
event
from
a
particular
broker,
and
then
you
know,
as
as
I
have
subscribed
I
want
to.
A
B
Broker
is
so
so
so
is
the
trigger.
So
there's
the
broker,
which
is,
I
guess,
like
some
sort
of
message,
queue
message,
brass
and
then
the
trigger
is
the
logic.
Is
that
right?
It's
the
logic,
is
the
logic
of
what
you
do
with
the
different
events
or
what
can
be
done
like
it's
selling
to
the
subscribers
like
it
has
more
knowledge,
more
control.
More
sense
of
I
don't
know,
I
guess
more
sense
of
context
in
many
ways
like
this
thing
happen,
it's
not
I'm
not
just
giving
information.
B
A
That's
what
like
that's
what
I
understand
of
like
trigger
is
just
saying
that
okay,
I
do
want
to
when,
when
an
event
is
coming
into
a
particular
broker
with
a
particular
name.
I
am
I
I'll
send
this
to
a
subscriber,
but
there's
there's
so
there's
filtering
inside
of
triggers,
so
you
can.
You
can
like
filter
with
different
triggers,
but
I
wouldn't
really
say
that
the
the
the
trigger
itself
has
a
lot
of
information,
but
it
does
have
information
for,
like
let's
say
the
I'm
actually
looking
at.
A
Something
one
of
them
they're
they're
like
filtering
and
how
they're
doing
it
so
so
they
have
like
the
broker
and
then
they
have
filters
inside
of
that
broker,
which
will
create
that,
like
that
contact
or
that
knowledge
that
you're
talking
about.
So
you
know,
even
when
you
have
a
broker,
which
is
subscribing
to
a
particular
which
is
basically
sitting
between
that
publish
and
subscriber
of
two
different
kinds,
and
like
publisher,
says
that
you
know
I
have
all
of
these
events,
all
of
which
will
go
to
broker
a
even
inside
of
that
broker.
A
You
will
have
like
trigger
scan
filter
based
on
here.
They
have
attributes,
so
we
can
also
have
attributes
of
you
know.
The
c
type
should
be
so,
and
so
the
c
source
should
be
so,
and
so,
if
that
makes
sense,.
B
I
didn't
know
if
that
helped.
You,
though,
but
I
think
this
is
a
really
interesting
pattern.
A
Yeah,
I'm
just
thinking
that
if
we
are
creating
that
tight
coupling
between
you
know
jenkins
as
a
sink
and
any
any
such
any
sort
of
publisher,
you
know
whatever
that
source
is
it's
it's
very
tightly
coupled
then
the
whole
idea
of
communication
being
asynchronous
or
being
you
know,
failure
resistant
is
sort
of
like
the
whole
idea.
Is
it
it's?
This
loses
the
whole
thing.
A
So
I'm
not
again,
I'm
not
really
sure
because
they're,
not
just
you
know,
there's
women
pops
up
patterns
too
there,
as
you
said,
there's
so
many
ways
we
can
implement
this,
and
the
only
one
thing
that
I
did
look
into
was
the
the
pumps
of
light
plug-in
inside
of
jenkins,
which
is
like
a
puffs
up.
A
lightweight
published
subscribe
pattern
for
jenkins,
but
it
is
still
it's
pretty
interesting.
So
it's
like,
I
was
just
thinking
about.
A
A
I
think
they
take
a
lot
of
thought
and
reiteration,
so
I'm
actually
excited
to
have
this
conversation
and
then
like
have
these
conversations
in
the
future.
Whatever
we
might
end
up
developing.
B
Yeah,
I
totally
agree.
I
think
this
has
got.
You
know.
The
gsoc
project
project
is
amazing
and
your
work
is
fantastic,
but
I
think
there
will
definitely
be
more
to
work
on
beyond
g-suck
and
you
you
will
be
a
great
person
to
do
it,
and
I
do
think
it
has
all
these
like
really
interesting
questions
on
architecture
and
how
you
make
a
good
event-driven
system
that
our
sort
of
questions
the
industry
as
a
whole
are
is
grappling
with
and
continually
evolving,
and
it's
right
in
the
heart
of
that
so
yeah.
A
Yeah,
in
fact,
a
lot
I
think
I
would
say,
60
percent
of
my
the
the
g
song
proposal
was
just
proposing
different
ideas
for
designing
this
event.
Sync,
because
that
more
than
the
event,
storage,
because
event
source
sort
of
sounds
pretty
straightforward,
except
you
know
trying
for
transient
failures,
but
talking
about
jenkins
is
a
sink
there's
just
so
much
that
we
need
to
keep
in
mind.
Well,
you
know
we
have
the
event
parsing
the
metadata
or
the
body
parsing
and
then
creating
filters
and
creating
a
whole
architecture.
B
B
A
Yeah,
I
have
not
really
been
able
to
connect
with
like
the
right
people
who
would
have
a
lot
of
sort
of
idea
on
this.
Regarding
the
event
architecture,
I
do
think
that
the
the
event
stick
is
the
most
relevant
for
these
type
of
questions,
and
I'm
hoping
that
you
know
it'll
be
my
first
time
today
going
to
the
meeting
and
hearing
from
them.
A
But
I
hope
that
that
can
be
a
place
where
I
can
raise
situations
around
this
and
also
have
a
communication,
and-
and
yes,
they
have
like
talking
with
you
guys
and
it
is
it
it
sort
of
gives
an
idea
to
to
be
sort
of
steered
in
the
right
direction
of
researching,
because
right
now,
it's
just
like
looking
at
a
bunch
of
different
things.
A
So
so
I
yes
and
no
yes,
because
you
know
they're,
like
all
of
the
members
in
the
community-
are
very
helpful
and
they're
also
like
guiding
for
for
this
particular
sort
of
architecture,
but
no
because
I
yeah
do
not
really
have
like
been
able
to
connect
with
or
like.
I
did
not
actually
ask
these
questions.
A
lot
before
is
just
starting
to
come
around
for
jenkins
as
a
sink,
so
I'm
hoping
once
everyone
is
on
the
call
and
also
through
the
sig
event
or
the
event
sake.
B
Yeah,
I
think
so
and
the
event
say
is
a
good
place
to
start
and
mauricio
is,
I
believe,
in
his
new
his
new
job
he's
working
on
k
native,
so
I'm
like
we
might,
I
mean
I
realized,
he's
new
he's
settling
in
and
that's
always
such
a
you
know,
there's
always
so
much
to
do
there,
but
I
think,
between
kind
of
pulling
on
more
connections
that
we
all
might
have
and
then
also
the
events
that
I
think
that
we
could
get.
B
You
know
you
speaking
to
more
people
and
it
would
be
interesting
for
you
yeah
now
I
want
to
go
to
the
event's
sake
too.
A
Because
yeah
well,
those
were
the
only
top
questions.
I
had
burning
questions.
There's
this
I'm
pretty
sort
of
set.
The
first
thing,
obviously,
is
this
doing
or
moving
the
ui
from
mobile
config
to
having
its
own
page,
and
so
I
think
this
week
is
going
to
be
doing
that
and
also
just
researching
different
methodologies
of
implementing
things,
but
but
in
a
more
async
communication
kind
of
things,
I
have
been
looking
at
a
lot
of
them
and
they
all
have
such
different
architecture.
A
I
get
so
confused,
but
but
there
is
should
be,
and
I'm
basically
like
writing
down
all
different
architectures
and
what
can
work
and
what
cannot
so.
Hopefully,
it
should
be
good
for
this
week
and
I
am
meeting
with
the
bell
later
and
maybe
I
might
able
to
schedule
something
with
mauritia
and
talk
more
with
these
things,
but
that
was
that
was
it
for
me
today.
B
Okay
sounds
good,
in
which
case
since
you
have
more
meetings
in
the
week,
we
can
give
you
the
next
half
an
hour
back
here,
I'll
stop
our
recording.