►
From YouTube: AsyncAPI SIG meeting 29 (August 18, 2020)
Description
This is the recording for the AsyncAPI Special Interest Group (SIG) meeting #29.
Attendees:
- Lukasz Gornicki
- Jonas Lagoni
- Michael Davis
- Alberto Ricart
Moderation:
- Lukasz Gornicki
Agenda & Notes:
https://github.com/asyncapi/asyncapi/issues/425
Chat Dump:
00:15:24 Lukasz Gornicki: https://github.com/asyncapi/asyncapi/issues/425
00:18:06 Lukasz Gornicki: https://github.com/asyncapi/parser-js/issues/119
00:25:50 Lukasz Gornicki: https://github.com/asyncapi/asyncapi/issues/423
00:39:46 Jonas Lagoni: ==it was not possible to put code sample here because of the brackets==
00:46:01 Jonas Lagoni: https://github.com/asyncapi/generator/issues/388
00:46:42 Jonas Lagoni: https://github.com/asyncapi/generator/issues/391
A
Okay,
recording
started:
hey
everybody
watching
the
recording,
it's
a
our
usual
every
two
week
meeting
that
is
available
for
a
later
review
on
youtube
today
on
the
call
it's
it's
me,
I'm
maintaining
async
api,
tooling,
there's
michael
from
from
solace,
come
on.
How
could
I
forget
from
solace
that
also
contributes
to
the
to
the
project?
Jonas
also
contributor-
and
we
have
alberto
from
snadya
from
the
nuts
project,
also
support
us
here
with
questions.
A
A
A
A
Everything
is
implemented
in
the
parser
and
in
the
generator
we
are
with
parser
on
release
candidate
number
two
and
generator
release
candidate
x,
but
for
sure
less
than
ten
it
will
not
be
more
than
10.,
and
now
we're
gonna
just
leave
it
for
the
community
to
try
out
for
for
two
three
weeks
just
to
make
sure
there
are
no
bugs
nothing
that
we
have
to
quickly
fix.
That
will
break
the
api,
but
but
other
than
that
there
will.
A
There
should
be
no
delays
and
that's
that's
it
if
it
comes
to
releases,
so
we're
pretty
pretty
good
if
it
comes
to
first
major
releases
for
those
two
packages
now
another
topic
is
is
something
I
I
wanted
to
brought
up
something
that
didn't.
Let
me
sleep
something
that
I
noticed
when
I
started
working
on
all
those
features
for
parser.
A
And
the
the
only
way
you
can
the
and
it's
done
in
the
way
at
the
moment,
this
way,
because
the
assumption
is
that
the
the
user
of
the
parser
will
later
on
decide
which
errors
they
want
to.
They
want
to
silent,
but
looking
on
on
the
type
of
driver,
now
the
there's
an
issue,
I'm
gonna
link
it
as
well
in
the
chat.
A
A
So
that's
one
thing
to
make
it
easier
to
silent,
for
example,
some
some
errors,
but
I
also
I
was
all
the
time
thinking
when
I
was
adding
all
those
new
errors.
I
was
thinking
about.
Why
not
have
a
kind
of
warnings
as
well,
so
do
a
validation
that
is
useful
for
the
user,
but
it's
not
required
by
the
spec,
but
it's
still
useful,
but
it's
up
to
the
user
to
decide
if
he
wants
to
have
this
additional
validation
or
not.
A
So
there
would
just
be
some
option
like,
for
example,
fran
mentioned
to
me
once
that
there
could
be,
for
example,
a
option
like
silent.
So
you
could
silent
the
the
additional
validations
and
I'm
asking
about
this
because
there
was,
I
introduced
two
validations,
so
I'm
checking
whenever
you
have
a
variable
in
the
server
I'm
checking.
A
B
No,
I
think
that
makes
perfect
sense.
You
know
that's
a
good
example,
because
I
was
having
trouble
imagining.
Why
would
you
accept
an
invalid
async
kpi
document,
but
that's
that's
a
perfect
example
of
why
you
would
you
know
if
you're
just
doing
the
basic
kpi
just
for
documentation
or
whatever
don't
really
care
about
code
generation
yeah
we
could
relax
some
things.
B
A
C
A
A
C
B
C
I
mean
I
would
still
expect
it
to
be
an
error
from
I
mean
either
you
forgot
that
there's
still
a
parameter
in
the
channel
or
you
simply
forgot
to
add
the
parameter
type
and
either
one
of
those
use,
k
or
either
one
of
those
scenarios.
I
would
still
expect
it
to
give
me
I
mean
I
would
still
expect
it
to
be
an
error,
but,
of
course
again
it
depends
on
where
you
want
to
draw
the
line
so
to
speak.
B
A
B
B
A
A
A
That
will
make
github
tokens
that
are
used
in
the
workflow
instead
of
read
only
for
the
forks
and
they
will
be
able
to
actually
write
as
well,
which
means
we
can
now
start
adding
to
the
prs
some
custom
validations
some
some
github
actions
that
have
to
update
the
pr
with
some
information
like,
for
example,
if
we
would,
for
example,
use
a
a
validation
of
the
of
the
pr
titles.
If
they
follow
conventional
comments,
there
could
be
a
a
github
action
used.
A
A
So
to
get
rid
of
this
responsibility,
we
can
introduce
this
github
action,
additional
workflow
that
will
validate
the
dpr
title,
but
something
more,
let's
say
for
a
discussion
and
like
I'd
like
to
hear
your
opinion
on
that
a
lot.
It's
something
that
I
noticed
in
the
in
the
kubernetes
community.
Whenever
so,
it's
part
of
the
of
the
rci
system,
which
is
brow
and
the
approach
they
have
there,
is
that
nobody
manually
merges
pull
requests.
A
So
it's
all,
let's
say
automated
with
pro
there's
a
special
plugin
that
does
it
and
it's
doing
it
basing
on
the
on
a
comment
to
the
pr
provided
by
a
maintainer.
So
if
a
maintainer,
I
think
it
makes
a
comment:
okay
to
test
and
not
okay,
to
test
it's
ready
to
merge
or
something
like
this
and
then,
if
the
pr
has
all
the
checks,
green,
it
can
be
merged.
A
A
This
means
that
actually
every
contributor
could
be
revoked
to
read
only
so.
We
don't
have
a
risk
that
some
maintainer
left
the
organization
and
we
have
like
write
access
assigned
to
someone
that
doesn't
work
on
async
api
anymore
and
also
we
could
easily
invite
people
to
the
organization
give
them
rights
to
label
prs.
A
So
more
more
people
working
on
asking
api
would
be
able
to
merge
the
prs,
so
they
would
not
like
like
two
or
three
maintainers
would
not
be
a
bottleneck
to
actually
merge
apr
because
they
would
be
able,
after
some
approvals
and
and
and
a
and
a
usual
checks,
they
would
be
on.
They
would
be
able,
on
their
own,
to
to
merge
apr
with
a
with
a
comment
or
a
or
a
label
to
the
pr.
A
A
A
So
that's
one
of
the
reasons
why
you
don't
want
to
have
it
done
automatically,
but
we
can
also.
We
could
also
do
it.
This
way
that
we
can
like
what
is
in
kubernetes
again,
they
have
do
not
merge.
I
think
comment
and
basically
on
this
comment,
the
bot
adds
a
label,
do
not
merge,
so
we
could
do
it
like
this
that
by
default,
you
have
all
approvals
on
all
and
all
checks
we
merge.
A
D
C
A
Okay,
then,
we
would
have
to
put
higher
priority
on
introducing
code
owners
because
we
could
then,
with
code
owners,
specify
then
and
then
block
on
the
pr
that
the
let's
say,
counting
approval
is
only
from
a
people
that
are
listed
in
the
code
owners
file
in
the
repo.
That's
a
github
functionality,
yeah.
A
Good
point:
that's
something
that
has
to
be
checked
because,
from
the
permission
level
I,
as
far
as
I
remember
it's
a
to
add
a
label.
You
need
to
have
at
least
I
think
triage
permission
level.
It's
a
pretty
fresh
few
months
old
permission
level
that
you
can
have
on
a
user
in
a
github,
so
it's
triage
and
then,
as
a
triage
user.
You
can
add
labels.
A
C
D
D
It
depends
on
the
client
or
whether
the
client
does
encoding
and
decoding
globally
or
just
for
per
request.
If
you
have
clients
that
are
actually
doing
multiple
subscriptions,
the
threaded
clients
can
be
not
thread
safe,
meaning
that
one
subscription
could
be
processed
in
one
thread
while
another
one
gets
processed
in
a
different
one.
But
that
depends
on
the
client.
I
believe
that
java
one
is
able
to
do
something
like
that
on
the
ones
that
I
work
on
javascript.
Obviously,
we
we
only
have
a
single
thread
at
any
one
time,
so.
C
C
D
C
D
D
There
should
not
be
well
see
the
the
the
serialization
is
actually
set
up
on
the
connection,
so
in
that
case,
I
believe
you
would
run
into
a
problem
yeah.
So
in
that
case
I
would
suggest
that
you
actually
create
different
connections
that
are
associated
with
this
realization
that
you
want
and
then
deal
use
the
right
connection
to
do
what
whatever
you
want,
then
you
don't
have
to
worry
about
this.
A
And
that's
how
it
that's
how
it
works,
but
when
you
have
template
inside
generator
ripple,
then
the
assembling
to
the
to
the
template
is
created
in
in
node
modules,
but
dependencies
instead
of
installed
on
a
subdirectory
level.
They
are
installed
on
a
on
a
generator
level.
So
they
that's
why
they
are
in
the
root
of
the
node
module
instead
of
like
inside
the
template
inside
the
like
nested,
let's
say
in
the
node
modules.
A
A
So
the
only
way
for
me
to
like
how
do
you
say
it
reproduce
reproduce
yeah.
So
the
only
way
for
me
to
reproduce
was
to
keep
the
template
in
the
same
directory
that
the
cli
is
located,
because
I
was
only
able
to
to
what
what
was
the
word
reproduce.
Oh
yeah,
so
I
was
only
able
to
reproduce
with
with
docker.
A
A
A
A
C
C
A
C
A
A
A
A
A
D
Coming
back
to
jonas's
question:
there
is
the
ability
for
you
to
create
an
encoded
connection,
and
if
you
do
that,
then
you
can
assign
the
serializer
and
the
serializer
for
that
and
then
your
subscriptions
will
use
whatever.
That
is
for
that
encoder
connection,
so
you
should
be
able
to
do
that
fairly
easily.
I
sent
you
a
link
to.
D
C
C
D
D
Again,
if
it's,
if
it's
a
multi-credit
environment
and
you're
changing
a
connection
property
there,
it's
not
you
know,
somebody
else
could
actually
use
a
decoder
or
something
that
could
happen
for
you
right
yeah,
so
really
the
proper
way
would
be
to
say
your
connection
is
binary
and
then
on
your
callback.
Where
you
are,
you
basically
invoke
you,
you
might
be
able
to
even
use
the
serializer
that
they
have
over
there.
C
Yeah,
because
the
main
problem
is
that
when
I
generate
all
the
different
schemas
that
the
payloads
of
the
channels,
I
use
quick
type
to
kind
of
just
generate
and
basically
don't
really
care
what
is
being
generated.
But
I
know
exactly
how
I
can
I
can
convert
to
json
back
and
forth,
but
that
requires
me
to
create
these
custom
serializers
per
message,
basically,
which
makes
it
kind
of
tricky.
D
C
A
C
A
C
D
Right,
you
would
have
to
actually
cascade
that
all
the
way
down
right
I
mean
you
would
have
to
know
what
the
channel
is
expecting.
Is
data
and
then
associated
on
the
deserialization
of
the
message
right
because
you're
processing,
the
payload
so
wherever
you're,
reading
the
payload
or
setting
the
payload?
D
C
C
A
D
C
D
D
D
All
I'm
saying
is,
for
example,
from
the
net's
perspective
right
you're
coming
a
message.
The
message
has
a
subject
so,
based
on
that
subject,
you
should
be
able
to
determine
what
your
payload
is,
if,
if
you
have
any
type
of
mapping
that
you're
trying
to
handle,
so
if
you
read
the
payload
data
as
binary,
you
can
convert
it
to
whatever
it
is
and
then
hand
it
down
or
process
it
right
in
there
right.
Yes,.
D
D
If
you
go
all
raw
right,
meaning
that
you
expect
all
data
that
it
comes
in
to
be
a
vitaray,
then
you
can
decode
when
you're
processing
it
down
yourself
right.
Because
then
you
see
what
I'm
saying
the
the
the
the
subscription
is
gonna
dispatch
to
you
by
the
way.
So
then
right
there,
when
you
read
it,
you
can
then
convert
to
whatever
you
need
and
then
before
you
send
it,
you
convert
everything
to
a
byte
array
and
then
pass
it
on.
C
D
C
C
D
That
code,
wherever
you're
handing
it
down,
you
could
provide
the
appropriate
connection
for
your
publish
and
and
to
actually
handle
your
subscriptions.
But
that
means
that
your
application
must
divide
communication
between
two
different
connections
and
that
actually
has
implications
in
terms
of
message
ordering
and
other
kinds
of
things
as
well
right.
So
yeah.