►
From YouTube: CNCF Serverless WG 2021-04-01
Description
CNCF Serverless WG 2021-04-01
A
B
A
C
C
D
Lucas,
hey
hello,
I
can't
put
my
finger
on
it,
but
something
feels
different
about
the
zoom.
I
can't
can't
just
feels
slightly
off.
I
can't
figure
out
what,
though,.
C
D
B
A
C
Has
all
kinds
of
interesting
things
in
terms
of
time
zone?
I
can
just
check
so
I
am
a
time
zone.
Nerd,
okay,
okay,
I
I
don't
mean
to.
C
C
D
C
D
D
Okay,
if
you
want,
if
you
want
to
be
associated
with
the
company
you're,
just
going
to
put
the
company
name
into
the
zoom
chat
and
I'll
make
sure
I
add
that
to
your
name
in
the
attendance
tracker
you
don't
have
to
you
don't
want
to,
but
usually
people
do
that
that
way.
Their
company
gets
voting
rights.
D
Okay,
I
assume
that's
jp
morgan.
If
I
wanted
to
think
about
the
long
version
of
it.
D
H
H
D
D
All
right
small
group
today,
but
let's
go
ahead
and
get
started
since
it
is
three
after,
although
we
don't
have
slinky
yet
I
was
gonna
lean
on
him,
because
a
lot
of
his
issues
are
coming
up.
So
let's
go
ahead
and
dive
right
into
it.
So
john,
thank
you
for
taking
care
of
some
of
your
ais
appreciate
that
community
time
anything
from
the
community
people
like
to
bring
up
that's
not
on
the
agenda.
D
All
right
not
hearing
anything,
just
let
you
know
we
do
have
a
discovery,
interrupt
call
after
this
one.
I
don't
think
there's
anything
on
the
agenda
as
of
right
now.
So
I
added
one
which
was
just
a
nagging
thing,
saying:
okay,
we
said
we're
gonna
start
testing
at
the
end
of
march.
Well,
it's
no
longer
march,
so
we
need
to
find
out
where
we
are
relative
to
interrupt
testing.
So
please
join
that.
D
If
you're
planning
on
doing
some
interop
testing,
because
we
can
find
out
everybody's
status-
okay,
coop
clan
eu,
they
did
send
that
a
note
asking
us
to
sign
up
for
our
sessions
for
the
what's
called
the
office
hours.
I
did
pick
late
times
in
the
day
for
wednesday
and
thursday,
please
let
me
know
if
you're
able
to
make
those
times
and
start
adding
people's
names
there
there's
nothing
if
you
need
to
do
in
advance.
D
It's
just
basically
show
up
and
answer
questions
that
people
might
have
if
they
decide
to
stop
by
our
booth
and
ask
questions
and
that's
about
it.
If
we
don't
get
anybody
to
sign
up,
then
obviously
we'll
start
nagging
people
but
it'd
be
great.
If
people
can,
you
know
volunteer
on
their
own,
to
do
that,
I
figured.
Maybe
two
people
would
be
good.
D
One
person
can't
handle
it
because
I
don't
think
we
get
too
much
traffic
but
still
be
nice
to
have
at
least
one
person
there
for
backup
tim
or
I
don't
see
him
on
the
call
I
was
talking
with
him
earlier
today.
He
didn't
mention
anything
to
bring
up
for
workflow
status,
so
I
assume
he
doesn't
have
anything
new
beyond
last
week.
So
keep
going
forward
all
right.
Anything
else
relative
to
you
know
any
topic
before
we
start
jumping
into
pr's
and
issues.
J
We,
you
can
add
the
link.
I
gave
you
yesterday
for
the
slide.
So
if
people
want
to
have
a
look,
yes.
J
Like
I
need
to
start
recording,
probably
tomorrow,
I
have
like
a
few
things.
I
just
added
like
a
slide.
This
morning,
I
woke
up
with
what
I
thought
was
a
good
idea.
Okay,.
D
J
D
Okay,
cool:
does
anybody
have
any
questions
if
they've
already
looked
at
it
for
remy?
Well
we're
just
going
through
the
agenda.
D
G
Yeah
I
I
commented
just
before
the
meeting
I,
let's
close
it,
it's
doing
nothing
more
than
just
adding
confusion.
D
Okay,
what
other
people
think
any
objections
to
closing?
D
Let
me
just
double
check
it,
but
I
suspect
I'll
be
happy
for
it
to
be
closed.
Yeah
just
remind
everybody.
This
is
not
changing
the
specification.
This
is
just
the
primer
and
we're
looking
to
provide
additional
guidance,
but,
as
lance
said,
it
may
actually
confuse
people
more
and
so
maybe
we'd
be
better
to
wait
until
we
had
another
concrete
example
to
leverage
that
we
wanted
to
to
focus
in
on
instead
of
one
that
ended
up
being
sort
of
a
bad
example.
Was
that
a
fair
way
to
phrase
it
lance?
G
D
D
C
Yep,
so
we
discussed
this,
I
think
three
weeks
ago
and
daniela
zuma,
and
I
had
some
discussions
internally
and
the
result
is
this
pr?
Basically,
it
just
lays
everything
out
really
explicitly
so
that
for
percent
encoding
you
don't
even
need
to
refer
to
rfc
3986,
because
it's
very
specific
about
you
must
encode
these
characters.
C
You
must
be
able
to
decode
things
that
have
been
encoded,
unnecessarily,
etc,
but
it's
still
entirely
compatible
with
rfc
3986.
As
far
as
I'm
aware
and
someone
who's
smarter
than
I
am
about
that
rfc,
please
have
a
look
for
the
second
part
of
the
encoding,
so
rfc
7230
it's
a
sort
of
interesting
reference,
because
it's
non-normative
it's
kind
of
some
fields.
Maybe
some
header
values
may
be
defined
in
this
particular
way
that
uses
double
quoting.
C
C
It
wouldn't
entirely
surprise
me
if
there
are
existing
implementations
that
do
double
quoting
so
any
implementation
parsing
an
http
message
should
handle
double
quoting,
but
I've
said
do
not
have
to
handle
comments,
because
I
don't
think
we
expect
any
header
fields
to
include
comments
within
the
attribute
that
aren't,
if
something's
got
its
own
comment,
format
within
the
attribute
value,
fine,
but
no
header
comments
which,
frankly,
I've
never
seen
in
http.
Anyway,
that's
the
sort
of
tl,
dr
of
it
I've
gone
into
specifics
about
it's
utf-8.
C
Lost
the
lost
the
plot
non-vmp
characters
in
this
particular
way.
You
must
reject
things
that
have
encoded
surrogate
pairs
as
two
separate
utf-16
code
points.
I
think
I've
been
reasonably
exhausted,
but
it's
very
easy
to
think
that
when
not
doing
so
so
particularly
anyone
who
knows
a
load
about
unicode,
please
let
me
know,
especially
if
I've
misused
the
the
characters.
Sorry,
the
the
terms
I
have
a
c-sharp
unmerged,
pull
request
that
implements
this.
C
C
I
really
want
you
to
encode
these
things,
in
particular,
double
quote
and
space
and
for
decoding
it
tends
to
be
within
url
decoding,
which
will
also
decode
plus
as
a
space,
and
I
don't
think
we
want
to
get
as
far
as
saying
that
http
headers
are
kind
of
urls
when
they're
not
so,
I'm
quite
happy
for
the
c
sharp
code
to
be
used
as
sample
code.
It
is
it's
larger
than
we
might
like,
but
frankly,
a
few
hundred
lines
once
per
language,
I
think,
is
kind
of
reasonable
okay
kind.
D
Of
sales
pitch
all
right,
a
quick
question
for
you
before
I
open
up
to
the
floor,
is
it
your?
Is
it
your
assumption
that
what
you
wrote
here
from
I'm
sorry,
I'm
full
blown?
My
words:
is
it
your
assumption
that
you
did
not
change
the
desired
semantics
of
what
we're
trying
to
do,
even
though
I
know
it's
technically
a
change
from
what's
there
today
is
this?
Do
you
think
this
is
in
line
with
what
we
intended
to
mean
in
the
original
spec?
C
I
had
found
it
clear
what
was
intended.
I
possibly
wouldn't
have
raised
the
issue.
I
I
think
there
will
be
subtle
differences
in
terms
of
I'm
requiring
a
few
particular
characters
to
be
percent
encoded.
So
my
guess
is
that
existing
implementations
may
well
not
present
encode
spaces.
C
A
bunch
of
implementations
won't
then
double
quote
context,
attributes
that
start
with
a
space
so
that
space
will
be
lost
in
many
cases.
So
I
think
it's
strictly
better
than
and
when
and
compatible
with
in
terms
of
it
won't
start
losing
information
that
was
previously
not
being
lost
already
is
my
expectation.
C
D
Okay,
so
I
think
it's
then
important
that
the
in
particular
the
sdk
folks
in
the
group
look
at
this
very
carefully
to
see
what
kind
of
impact
does
it
have
on
their
code.
If
any,
I
think
that
should
help
us
decide
one
if
we
want
to
go
this
direction
and
two
if
this
is
actually
a
breaking
change
and
if
so,
do
we
want
to
do
it
now
or
not.
You
know
later
or
that
kind
of
thing,
so
big
decisions
in
front
of
us.
C
I'm
I'm
in
under
no
illusions
that
this
is
about
to
be.
You
know
rubber
stamped
through
and
I'm
expecting
quite
a
lot
of.
D
Discussion;
okay,
so
obviously
it's
fairly
large
and
you
need
some
reviews,
so
I'm
not
gonna
even
think
about
asking
for
any
kind
of
approval.
Today,
however,
does
anybody
have
any
immediate
questions
for
john
that
they'd
like
to
ask
now
before
the
group
has
a
chance
to
actually
look
at
it
in
more
detail.
C
D
Okay,
cool
all
right,
any
other
comments,
chat
or
voice;
okay,
not
hearing
any,
then
please,
everybody
take
your
time
to
review
it
very
carefully,
because
I
said
this
could
have
been
big
implications
going
forward,
especially
if
it
unintentionally
does
a
breaking
change
that
we
didn't
mean
to.
C
D
Right
and
and
and
we
have
a
situation
in
the
past
where
we
put
something
in
the
spec
and
it
just
was
not
what
we
what
we
intended
and
so
technically
it
was
a
breaking
change,
but
we
considered
that
to
be
more
in
the
typo
category
more
than
anything
else,
because
we
did
not
mean
it
to
be
to
come
out
the
way
it
did,
and
so
we
we
this
may
be
in
that
category.
So
we'll
have
to
say.
C
And
presumably,
there's
the
technically
breaking
change.
You
know,
theoretically,
if,
if
there's
an
sdk
implementation
that
we're
not
aware
of,
then
it
might
be
a
breaking
change,
but
actually
all
the
existing
sdks.
It
wouldn't
be
a
breaking
change
that
could
influence
as
well
as
influence
us
as
well.
Presumably,
yes,
agreed.
Okay.
D
I
Sorry
I
didn't
raise
the
end,
but
it
will
be
a
breaking
change,
because
one
one
application
built
with
the
previous
version
of
the
sdk
and
the
new
version
between
the
new
version
of
the
sdk
is
going.
They
are
not
going
to
talk
each
other.
C
I
would
hope
that
so
anything
that
already
does
percent
decoding
should
be
able
to
handle
whatever
the
new
sdk,
outputs
and
the
and
the
new
sdk,
parsing
and
http
message.
I
would
expect
to
be
able
to
handle
whatever
the
old
sdk
put
out
as
well,
because
there's
this
business
about
you
have
to
be
able
to
parse
double
quoted
stuff,
but
you
should
never
produce
double
quoted
stuff
and
it
may
well
be
you
know,
what's
on
the
screen
right
now,
doesn't
show
all
of
that.
C
It's
a
long
pr,
I'm
afraid
but
yeah.
So
slinky.
Please
don't
misinterpret
me
saying.
No,
I
don't
think
it's
breaking
in
that
sense
as
a
genuine,
I'm
sure,
you're
clearly
wrong.
I
suspect
it's
best
to
make
concrete
examples.
If
you
put
a
concrete
example
in
the
pr
saying,
imagine
an
old
sdk
that
does
this
and
the
new
sdk
does
that,
and
this
is
the
attribute
value,
etc.
I'm
really
really
happy
to
work
through
all
kinds
of
things
like
that.
D
Do
you
have
a
set
of
of
existing
http
headers
that
may
be
influenced
by
this
pr
and
if
so,
maybe
add
that
as
a
comment
in
here,
so
people
are
very
aware
at
least
the
ones
that
are
that
you
know
about
that,
may
be
impacted
by
this.
C
I
don't,
I
have
a
bunch
of
tests
in
the
new
implementation,
but
I
don't
have
we
would.
We
would
need
to
collect
samples
from
across
multiple
sdks.
I
guess
so.
If,
if
anyone
else
has
examples,
maybe
from
their
existing
integration
tests
and
things,
that
would
be
fantastic
and
we
can
just
validate
as
we
go
well.
D
C
That's
what
I
what
I
meant
by
validation
as
in
you
know,
this
is
what
we
would
currently
do
and,
let's
validate
that,
it
doesn't
break
okay,
yeah,
please
feel
free
to
dump
loads
of
examples
in
the
pr
okay.
D
C
So
we
had
a
question
actually
in
the
c-sharp
repo
massively
coincidentally
saying:
how
do
we
go
about
versioning
cloud
events?
What
happens
if
I
want
to
make
a
breaking
change
in
the
schema?
What
changes
are
breaking
etc?
C
And
I
gave
my
personal
view,
but
we
agreed
that
it
would
be
good
to
have
something
in
the
primer
moved
it
over
into
the
spec
repo.
Three
weeks
ago
we
discussed
it
and
I
certainly
went
away
with
an
impression
of
what
the
group
expected
to
or
wanted
to
see
in
the
primer,
whether
I
accurately
remembered
and
then
accurately
expressed
it
is
a
different
matter,
but
the
basic
idea
is:
this
is
only
guidance.
It's
not
requirement.
C
It's
saying
that,
once
you've
published
something
with
a
particular
cloud
event
type,
then
you
probably
shouldn't
unless
it's
clearly
labeled
as
beta
or
you've
got
other
documentation
somewhere
else
saying
stuff
isn't
stable,
yet
you
shouldn't
make
breaking
changes
within
a
type
feel
free
to
version
your
types
in
any
particular
way.
Data
schema
is
more
an
informational
thing,
so
just
changing
the
data
schema
uri,
whilst
breaking
stuff
isn't
enough
notice,
because
you
may
well
still
have
old
consumers.
C
C
K
C
I
think,
if
I
understood
your
suggestion
correctly,
I
think
there
there
may
well
be
some
kind
of
cloud
events
that
are
handled
very
dynamically
and
will
always
be
handled
very
dynamically
and
making
breaking
effectively
making
breaking
changes.
But
advertising
that
within
the
data
or
within
the
data
schema
is
fine,
but
I
suspect
in
most
cases,
if
people
expect
json
with
a
field
called
name,
and
you
suddenly
change
that
to
title.
Then,
however,
you
advertise
that
they've
still
got
code,
that
that's
expecting
it
to
be
called
name
and
writing
code.
That
is
extremely
dynamic.
C
If
they
don't
know
that
it's
going
to
be
changing
from
name
to
title,
how
can
they
work
out?
What
to
do
with
it?
I'm
certainly
happy
to
maybe
have
a
few
more
options
for
versioning
and
pros
and
cons,
but
I
I
suspect
that
if
we
don't
provide
any
guidance,
a
lot
of
people
will
either
not
think
about
versioning
at
all,
which
I
suspect
is
going
to
be
fatal
or
come
up
with
a
zillion
different,
not
quite
compatible
ways
of
doing.
D
D
Any
other
questions
comments.
Obviously
I
think
people
need
time
to
look
this
over.
Providing
guidance
is
always
interesting
when
it
comes
to
areas
where
you
have
lots
of
choices
right,
you
got
to
be,
you
have
to
be
careful,
so
I
think
people
need
to
think
long
and
hard
about
this.
C
C
I've
kind
of
that's
my
understanding,
but
I
can't
point
to
anything
that
says
that
it
makes
sense
to
me,
but
may
well
be
completely
wrong,
so
we
may
just
want
to
strike
that
paragraph
edit,
it
etc,
and
those
with
more
history
in
why
things
are
designed
that
way
may
well
have
a
fit
when
they
when
they
see
what
I've.
C
D
All
right,
cool
two
interesting
pr's
require
lots
of
thinking
all
right.
In
that
case
john,
I
think
those
are
your
only
ones
that
are
up.
D
Much
for
obliging
me
and
doing
those
first
sure,
all
right
in
that
case,
let's
go
back
into
the
top
start
with
the
oldest
one
first
slinky.
This
one
is
yours.
I
think,
last
time
you
asked
us
not
to
merge
it.
I
think
you
may
have
made
some
minor
changes
here,
but
is
there
anything
you
want
to
draw
people's
attention
to
on
this
one
before
I
ask
if
there
are
comments
or
questions
on.
I
I
Yeah
well,
for
me,
this
is
good
to
go.
Okay,
I
I've
so
so.
Well,
let
me
give
some.
Let
me
provide
some
context
on
why
I
opened
so
much
vrs.
So,
let's
make
it
easier,
so
I
went
through
this
week
and
I've
implemented
the
whole
language
in
java
in
the
sdk
java,
and
there
is
an
open
pr
for
it
and
the
end
at
least
I
checked
this
morning
again
and
it
implements
the
full
spec,
including
all
the
various
casting
rules,
the
operators,
of
course,
the
old
built-in
functions
and
so
on.
I
B
B
I
Yeah
I'd
like
to
go
through
all
of
this
very
quickly,
so
then
I
found
out
some
minor
things
like
the
division
by
zero
can
return
a
better
default
value.
There
are
some
things
that
are
not
very
well
clarified,
like
the
in
semantics.
The
function
overloading
the
concat
should
accept
the
delimiter
and
all
kind
of
small
things.
So
I
I
kept
this
splitted
in
order
to
avoid
one
giant
pr
and
fix
all
the
different
things.
I
So
if
people
is
not
doesn't
agree
with
some
some
of
the
specifics
we
can
discuss
about
it
so
yeah,
that's
pretty
much
the
thing
so
general
copy
editing,
at
least
in
my
opinion,
can
be
merged.
If
people
is
okay
with
it,
so
there
is
an
order
for
merging
things.
Yes,
we
need
to.
We
need
to
merge
the
general
copy
editing.
First,.
I
D
Okay,
we'll
tell
you
what,
let's
start
with
the
first
one
so
on
this
one.
Anybody
have
any
comments
or
questions
on
the
the
initial
grammar.
D
D
I
know
that
there
are
based
upon
our
previous
conversations.
It
led
me
to
believe
that
there
is
not
a
single
sql
standard.
There
may
be
a
common
pattern
out
there,
but
there
is
no
single
sql
standard
out
there
that
everybody
says
yes,
this
is
the
one
and
only
one
there
are
slight
variants
out
there.
However,
in
your
opinion,
these
changes
that
you're
making
here
like,
for
example,
how
to
deal
with
divide
by
zero.
D
I
The
ladder
okay
so
well
for
for
the
in
so
the
in
pr.
I
Let's
say
I
did
some
what
made
most
sense
for
me,
but
it
doesn't
work
the
same
as
oracle
sequence,
so
I
tried
it
on
oracle
sql
and
it
doesn't
work
the
same
so
yeah.
That's
what
I'm
saying:
let's
go
one
by
one
and
find
out
what
people
think
about
it.
I
So
this
one,
this
one
is
basically
so
you
know
that
the
language
is
total,
so
it
always
returns
a
value
even
if
the
parameters
are
invalid.
But
if
the,
if
one
of
the
parameters
is
invalid,
it
depends
to
the
error
list
the
error.
So
in
this
case,
when
you,
when
you
divide
by
zero,
it
makes
far
more
sense
to
return
infinite,
plus
infinite
and
minus
infinite
more
than
returning
g,
and
in
this
case
plus
infinite,
is
the
maximum
value
of
in
32..
D
I
In
the
implementation,
in
the
implementation,
when
you
execute
evaluate
it
returns,
it
returns
a
type
that
contains
both
the
return
value
and
the
error
list.
If
the
error
releases
is
is
not
empty,
then
it
means
that
an
error
at
and
then
things
went
bad
right.
Okay,
so
that's
that's!
That's
at
least
how
I
implemented
that's
unless
the
implementation
right.
D
I
D
D
J
Time
for
me,
I
don't
understand
how
he
can
return
and
raise
an
error
at
the
same
time,
but
so
it's
maybe
just
me.
I
I
can
I
can
show
you
the
implementation,
so
it's
easier
to
to
understand,
at
least
at
least
what's
my
idea
behind
this.
It
may
be
wrong,
but
I'll
give
you
a
link.
J
Yeah
and
I'm
not
a
big
fan
of
taking
max
integer
because
it
represents
infinity,
but
what
happened
if
we
go
to
64
bytes
or
I
don't
know
in
20
years
in
128,
bytes
might
change.
No,
I
don't
know
I,
but
so
that's
I
don't
understand
the
full
context,
probably,
but
it's
just
just
that
sentence
as
a
programmer.
I
find.
D
So,
on
cases
like
the
slinky,
would
it
make
sense
to
first
get
clarity
and
a
and
a
concrete
decision
on
how
to
deal
with
errors?
Meaning
are
we
going
to?
Are
we
going
to
stop
processing
on
an
error
and
only
return
error,
or
are
we
going
to
return
a
value
and
an
error,
because
I
think
that's
a
pretty
high
level
decision?
Isn't
it
yeah.
I
That's
that's
well!
That's
that
that
was
the
that's
the
point
of
our
writing.
In
case
of
an
error.
We
don't
mandate
what
to
do.
Okay,
so
the
implementation
can
decide
to
do
whatever
it
wants.
So
it
can
decide
to
stop
the
execution.
It
can
decide
to
continue
the
execution
with
the
default
value
or
maybe
there
are
some
errors
that
are
more
important
somewhere
else
that
are
less
important.
I
Okay
and
we
are
not
mandating
anything
around
that.
Then
a
filtering
implementation
might
decide,
or
maybe
the
the
what's-
the
subscription
spec
might
decide
to
say
hey
if
there's
an
error,
it's
just
has
to
fail
and
we're
good
okay,
but
the
language
itself
is
subtracting
from
this,
and
it's
just
saying
every
function
is
total,
so
everything
works
on
something
there.
May
there
might
be
the
side
effect
of
appending
to
the
error
list,
but
still
there
is
a
return
value,
always.
J
J
It's
I
don't
understand
why
we
make
mandatory
to
return
a
value,
because
I
saw
someone
else
putting
a
comment
where
I
don't
understand
the
value
of
having
like
minus
max
integer
compared
to
zero,
and
I
must
say
that
I'm
with
him
on
that
one
because,
like
obviously
I
don't
really
care,
that's
mathematically,
it's
impossible
to
do
that.
J
D
I
It
raises
an
error,
but
maybe
in
that
case
this
that
particular
error
can
be
just
we
can
just
forget
about
it.
So
that's
that's
an
error
that
I
can
ignore.
So
that's
the
use
case
where
I
can,
where
I
might
want
to
say
hey.
There
are
some
errors
that
I
just
want
to
ignore,
because
I
don't
care
about
them
or
maybe
there
are
somewhere
or
maybe
I
just
want
to
maybe
in
the
filtering.
I
always
want
to
ignore
errors,
or
maybe
there
again
there
are
some
errors.
I
That
might
be
not
interesting,
so
we
can
ignore
them.
They
are
accessing.
Accessing
a
value
is
one
of
those
errors
where
you,
you
might
actually
want
to
say
yeah.
I
don't
care,
if
there's
an
error
when
accessing
a
value,
because
that
value
is
not
there.
I
can
show
you
I
I
I
stumbled
upon
across
this
side
issue
today,
while
I
was
testing
one
of
the
expressions
from
from
the
spec
wording,
so
this
one
this
particular
one.
I
I
Okay,
this
one
okay,
okay
yeah,
so
this
one
is
I'm
checking
if
there
is
first
name
and
then
if
there
is
a
and
last
name
are
equal
to
my
name
and
surname
or
if
there
is
a
subject
within
a
with
my
name
and
surname.
When
I,
when
I
try
to
address
an
extension
or
an
attribute,
that
is
not
there.
I
I
So
when
I
try
then
to
if
I
try
to
evaluate
this
specific
expression
giving
to
it
giving
to
it
an
event
which
doesn't
contain
first
name
and
last
name,
this
is
this
expression
will
fail
and
or
better
will
return
a
value,
but
it
will
raise
an
error
saying:
hey
you're,
trying
to
address
first
name,
but
first
name
is
not
there.
I
I
There
is
not
no
concept,
because
if
you
introduce
no
concept,
it
starts
to
be
more
complex,
so
there
is
no.
There
is
no
null
concept
and
this
one
doesn't
raise
an
error.
You
see
it's
the
answer,
checks
if
it's
not
failed,
but
if
you
remove,
with
extension,
the
the
line
1,
a
and
107
this
one
fails
because
it
returns
an
error,
saying:
hey
you're,
trying
to
address
first
thing,
but
first
name
is
not
in
the
event.
D
H
I
was
just
trying
to.
I
know
it's
my
first
time
here
and
I'm
guessing
the
middle
of
the
discussion,
but
I
think
that
by
assigning
a
value
in
an
error
and
if
people
suddenly
decide.
Oh,
let's
ignore
the
error
by
mistake,
or
whatever
reasons
they
can
start
getting
wrong
conclusions.
D
D
H
D
Yes,
I
personally
would
like
a
little
more
time
to
think
about
this
one,
not
so
much
because
of
the
value.
Although
I
do
I
do
I
do
understand.
Danielle's
comment
there.
You
know
maybe
zero
makes
more
sense
than
max
a
minute.
I
don't
know
about
that
one.
D
I
I'm
sorry
they're
choosing
to
hang
themselves
and
that's
their
choice,
obviously,
but
I
think
it's
almost
like
it's
encouraging
people
to
do
it
in
a
sense
right.
Whereas
if
you
only
return
an
error,
then
they
can
choose
how
they
want
to
handle
the
error.
Processing
right,
maybe
return,
zero
or
or
empty
string
or
something
in
whatever
case
they
have.
But
at
least
then
it's
their
explicit
decision
to
make
that
choice
and
we're
not
sort
of
encouraging
them
to
do
it
in
a
way.
D
I
Well,
I
I
want
to
add
something
yeah
I
I
would
love
to
to
keep.
I
would
love
to
do
a
distinction
here
and
the
distinction
between
how
the
language
works
and
how
is
then
modeled
in
its
interface,
okay.
So
in
its
interface
it
can
just
when
there
is
an
error,
you
can
just
always
return
an
error
and
that's
it
and
that's
fine.
I
was
planning
to
do
that
in
in
in
the
reference
implementation
too,
having
another
interface
that
when
there
is
a
there,
is
a
failure.
There
is
a
failure,
that's
it.
I
So
I
think
I
think
it
would
be
nice
to
keep
this
error
system
for
giving
the
flexibility
to
the
user,
to
choose
and
to
the
implementer
to
choose
whether
it
wants
to
implement
or
not
the
system.
D
Yeah,
I
can
definitely
appreciate
that
that
point
of
view
what
do
other
sql
languages
do
do
do
most
of
them
return,
values
and
errors,
or
do
most
of
them
only
return
errors.
How
do
they
deal
with.
D
D
I
D
Okay,
well,
okay,
I
don't
want
to
rattle
on
this.
I
think
I'd
like
to
hold
off
on
deciding
this,
and
at
least
personally
I'd
like
a
little
more
time
to
review
it,
but
before
we
move
on
to
the
next
one
manuel
has
a
question
for
you.
It
may
not
be
directly
related
to
this,
but
he
said
why
can't
there
be
the
concept
of
a
null.
I
Well,
to
simplify
things.
In
short,
when
you
bring
in
the
note
concept
you
you
start
to
have
to
deal
with
this
with
things
like
null
assertions
null
checks,
and
you
know
all
those
kind
of
things.
I
It
does
it
it
does,
but
I
think
it
has
a
different
purpose.
I
D
H
Just
on
the
concept
of
no
no,
it's
kind
of
you're,
representing
the
absence
of
value
in
the
opposition
to
like
a
default
or
zero
value.
So
it's
not
really
used
for
error
rendering,
because
here
you
know,
you
should
have
a
value
but
have
like
an
exceptional
route
and,
as
far
as
remember
like
most
of
sql,
will
throw
an
error
in
this
case
and
not
continue
to
process.
D
I
Yeah,
I
just
wanna
add
the
last
thing:
if
people
is
when
I
mean
if
we
wanna
reevaluate
the
errors,
the
way
errors
works,
I'm
completely
open
to
it.
So
I
mean,
if
somebody
has
some
concrete
proposal,
so
please
open
it
and
I'll
be
more
than
more
than
happy,
then
look
at
it.
So,
okay.
D
Yep,
that's
good!
It's
a
shame
that
clement
isn't
here,
because
I
think
he
had
some
strong
opinions
about
error,
handling
and
stuff
like
that.
So,
okay,
would
you
like
to
talk
to
this
one.
I
Yes,
so
this
is,
this
is
a
not
a
small
one.
Actually
so
this
is.
This
is
actually
saying
that
the
expression
can
return
any
type,
any
value
inside
the
type
system
and
well.
The
previous
assertion
was
that
the
expression
is
returning
a
boolean
value,
and
this
is
very
important
for
use
the
expression
language
outside
filtering
use
cases.
I
So
when
you
for
again
going
back
to
the
previous
example,
let's
say:
workflow
wants
to
use
this
expression
language
to
template,
a
specific
attribute
or
a
specific
extension
in
the.
In
the
event,
this
kind
of
use
case
can
be
fulfilled
only
if
the
expression
occurred
on
any
value
and
not
just
bullets.
D
D
I
Yeah,
so
this
one,
to
be
honest,
I'm
not
100
sure.
So.
The
problem
here
is
very
simple:
in
the
implicit
custom
rules
we
talk
about
casting
for
device
operators
by
both
unary
and
binary
operators
and
then
for
functions,
but
we
don't
really
talk
about
like
exists
and
in
so
for
lack
of
exist.
I
Of
course,
it's
let's
say
it's
natural
exist
doesn't
need
any
casting
rule
because
of
course
it's
just
taking
the
the
the
identifier
of
the
event
called
the
event
attribute
well
for
like
there
is
no
need
for
inclusive
casting
rules,
because
the
left
argument
has
always
to
be
a
string,
so
everything
has
to
be
cast
to
a
string,
but
for
in
the
the
problem
is
different.
So
for
in
actually
it's
not
very.
It's
not
even
very
well
defined.
I
To
be
honest,
so
maybe
we
have
to
reiterate
on
the
inoperator
again,
but
the
problem
is
that
for
in
there
is
some
need
to
it
did
we
need
to
understand
if
we
need
to
cast.
I
I
Well,
my
my
understanding
was
that
we
wanted
to
try
different
types,
maybe
well
well,
now
that
I'm
looking
at
it
to
be
honest,
I've
implemented
it
in
a
different
way,
so
so
yeah,
maybe
again,
maybe
we
need
to
rate
right
over
it
and
yeah.
I
would
love
to
understand
what
people
think
about
it.
D
I
I
I
That
that
makes
specific
how
in
behaves,
I
think.
I
D
In
fact,
I'd
like
to
actually
reach
out
to
some
of
the
sql
experts
that
I
have
within
ibm
to
get
their
take
on
it,
because
I
have
no
idea
to
be
honest
whether
this
is
a
great
change
or
it's
being
too
restrictive
in
some
way
or
being
too
loose.
I
I
just
have
no
clue
so
I'd
like
a
little
more
time
to
to
think
about
this.
One.
D
I
B
I
D
Cool
okay
with
that,
I
don't
think
I
have
time
to
dive
deep
into
any
others.
So
please
take
a
look
at
the
other
ones
in
particular,
if
you
know
someone
who
has
some
sql
expertise,
I
think
it'd
be
great
to
get
their
opinion
on
some
of
these
changes,
because
because
I
get
the
general
sense
from
the
group
that
we'd
like
to
try
our
best
to
to
adhere
the
most
common
use
of
sql
out
there
and
not
necessarily
to
find
something
that
that
has
almost
zero
possibility
of
leveraging
existing
code.
D
D
Okay,
in
that
case,
daniel,
are
you
there.
D
That's
good
matthew
or
matt.
Yes,
okay,
grant
you
still
there.
J
L
D
L
D
Actually,
I'm
very
glad
you
asked,
because
I
meant
to
bring
this
up
and-
and
I
completely
forgot,
I
think,
it'd
be
really
useful
if
on
the
next
on
next
week's
call,
if
you
could
give
a
very
short
little
presentation
or
something
to
show
what
people
need
to
do
in
their
commits
to
use
this
stuff
to
get
past
the
the
the
checker,
because
I
think
for
most
people.
Yes,.
D
D
G
D
J
Like
my
focus
right
now
is
cube
so
after
the
video
I
think,
I'd
be
better,
but
it
was
great
to
work
on
the
presentation
because
I
like
basically,
I
never
really
dig
enough
into
the
schema
registry,
and
so
I
had
questions
around
that
too.
I'm
not
sure
it's
complete,
because
we
are
working
more
on
the
discovery
and
subscription,
but
like
based
on
the
data
schema
that
for
me,
potentially
links
to
a
schema
registry
or
not.
J
As
I
told
you,
doug,
like
my
company,
just
got
bought
so
I
had
a
few
stuff
to
do,
but
I
expect
to
have
a
little
bit
more
time
in
like
in
the
next
week
to
code
on
that
and
like
really
focus
on
that,
but
before
that
I
really
had
to
do
the
kubecon
video.
So
I
didn't
I
picked
my
battles
and
it
was
more
on
the
presentation.
J
D
Okay,
in
that
case,
continue
to
do
this.
For
myself,
I
haven't
done
a
whole
lot
of
changes
since
then.
I
do
think
that,
what's
out
there
today
is
at
least
workable
for
some
interrupt
testing.
D
I
don't
have
any
other
stuff
like
the
pagination
and
stuff
like
that,
but
I
do
think
I
have
at
least
a
fair
number
of
the
features
in
there
to
at
least
do
some
basic
interrupt
testing,
but
I
do
I
do
have
more
to
do
and
I,
like
you,
I've
been
very
heavily
focused
on
some
of
the
conferences
that
are
coming
up.
D
Ibm
has
a
couple
in
the
next
month
or
so
so,
unfortunately,
that's
taking
up
a
lot
of
my
time,
but
I
am
planning
on
finishing
out
at
least
to
match
some
of
the
descriptions
up
here
in
terms
of
what
we're
supposed
to
be
testing,
at
least
I
think,
what's
up
there
today,
can
please
do
some
testing,
okay
manuel?
How
about
your
side?
D
Oh
there,
we
go,
no
update,
mostly
blocked
by
internal
corporate
stuff,
still
looking
forward
to
get
a
client
going.
Okay,.
F
Yeah
I'm
trying
to
get
it
mostly
code
generated
and
I
think
it's
okay
for
the
discovery
for
subscription.
I
ran
into
that
issue
that
I
don't
have
any
public
end
points
to
to
get
the
amqp
delivery
pointed
to,
but
all
http
callbacks.
Actually,
so
I'm
still
positive
that
I
I
can
get
some
code
running
with
the
existing
discovery,
at
least
and
then
for
subscription.
I
think
I
have
to
wait.
D
Okay,
I
don't
think
we
have
anybody
is
any
of
us
in
the
call
who'd
like
to
chime
in
who
was
thinking
about
doing
some
coding.
D
Question
I
know
everybody's
really
busy
and,
and
the
coding
stuff
takes
time
obviously-
and
this
probably
is
actually
a
better
question-
asked
of
the
full
group-
not
just
this
interrupt
group,
but
let
me
start
here
and
just
put
an
idea
out
there
or
a
question
out
there.
D
I
I
do
feel
like
we
are
moving
kind
of
slowly,
not
just
in
the
code
but
on
the
specs
as
well-
and
I
know
most
of
us
have
just
been
kind
of
assuming
that
it's
because
you
know
we're
all
very,
very
busy
right
now
and
there's
a
lot
of
a
lot
of
stuff
going
on.
You
know
work-wise
as
well
as
outside
of
work.
D
Does
anybody
have
any
reason
to
believe
that
that
there's
that
this
is
a
sign
that
maybe
these
specifications
are
not
I've
picked?
My
words
right
are,
maybe
not
that
critical
or
is
it
just
the
matter
of
people
finding
time,
because
one
of
the
things
I
one
of
the
things
I
I
I'm
worried
about
is
while
we
can
look
at
these
specs
and
say
yeah,
we
see
there's
value
in
here.
Is
there
enough
value
to
actually
go
forward
with
them,
though?
Right
or
is
it
just
yeah?
D
It's
interesting
right
because
I,
with
with
cloud
events
I'll
be
honest,
it
took
me
a
little
bit
of
time
to
completely
see
the
value.
I
was
all
in
favor
from
very
beginning,
in
the
sense
that
it
sounded
like
an
interesting
idea,
but
the
value
of
it
and
to
customers
really
didn't
sink
home
right
away.
It
took
some
time
for
me
to
understand
it
better,
but
then,
once
I
did,
I
was
like
all
on
board
with
it
and
I
I'll
be
honest
with
you
with
these
specs
I'm
torn
right.
D
I
don't
know
whether
this
is
just
a
matter
of
time
and
at
some
point
a
light
bulb
will
go
off
in
my
head
and
say
yes,
this
this
is.
This
is
great.
This
is
the
next
logical
step.
We
definitely
need
this
stuff
or
are
we?
Are
we
trying
to
push
on
a
rope
here,
so
I'm
trying
to
figure
out
so
class?
You
came
off
mute
about
to
hear
your
yeah.
M
So
so
I
started
a
while
ago
and
with
this
intention
to
do
something
for
discovery
and
ran
into
quite
a
few
things,
I'd
like
to
bring
up,
but
just
didn't
find
the
time
to
do
so.
M
I
think
it
will
get
quite
important
for
what
clements
and
I
am
are
doing,
but
what
I
see
at
least
around,
where
I'm
working
is
that
people
are
really
still
busy
well
how
to
freestyle
them
digesting
the
pure
cloud
events
back,
so
I
don't
know.
For
example,
I
spent
probably
really
weeks
of
my
life
in
discussions
about
what
sources
and
and
things
like
this,
so
it's
just
that
a
lot
of
people
are
just
not
there.
J
Okay,
for
me,
I
think
the
subscription
and
discovery
seems
really
like.
I
don't
see
the
system
working
without
it
because,
but
so
I
didn't
had
much
time
with
that.
One
thing
and
the
other
is
like
that's
my
first
time
really
being
in
those
kind
of
groups
and
like
I
have
like
a
huge
respect
for
every
one
of
you
in
this
group,
and
sometimes
I
have
the
feeling
that
I'm
kind
of
a
fraud
in
that
group-
I
don't
know
so
it's
kind
of
it's
hard
for
me
to
push
because
I'm
like
I'm,
not
sure.
J
J
I'm
like
wow,
there
is
so
many
people
who
come
from
like
different
backgrounds
and
like
different
contexts
like
clemens,
like
obviously
at
asia,
he's
probably
not
talking
about
the
same
size
of
my
small
company
and
and
it's
a
little
bit
intimidating
for
me
to
to
try
to
drive
some
stuff
because
I'm
like
maybe
I'm
completely
wrong.
I
don't
know
okay.
D
D
I
I
wish
more
people
would
speak
up
because
I
want
the
interaction,
because,
even
if
people,
if
even
if
what
you
say,
is
true
right
where
okay,
maybe
you
don't
have
much
experience
with
someone
like
clements
or
something
like
that
that
doesn't
matter
right,
even
as
they
say
in
school
right,
there
are
no
dumb
questions
right.
I
have
the
same
attitude
here,
bringing
up
a
topic
that
you
either
want
to
get
clarity
on,
or
you
think
is
the
right
way
to
go,
and
just
even
having
the
discussion
is
moving
the
ball
forward
to
me.
J
J
Like
I'm
sorry,
if
people
don't
like
beer,
I'm
from
north
of
france,
it's
like
beer
country
in
norfolk,
but
it's
just
to
know
each
other
and
maybe
like
do
some
work
session
where
we
are
like
whiteboard,
ids
or
just
to
make
sure
we're
on
the
same
page,
so
I'm
probably
old
school
there,
but
I
think
the
now
that
I've
been
in
the
group
for
almost
a
year
and
like
tremendous
welcoming
group
like
I
mean
I
was
amazed
by
that,
like
really
congrats
dude
here,
because
I
think
you
really
I'm
really
impressed
with
your
skills
on
how
to
manage
the
community.
J
Even
if
you
don't
have
always
lots
of
feedback,
but
I
think
yeah
it
would
have
helped
to
just
meet
like
it's
same
thing
for
interrupt.
Maybe
if
we'll
just
be
sitting
in
the
same
room
for
like
a
few
hours,
it
would
have
been
probably
faster
and
with
the
time
difference
it's
hard
to
even
do
that
with
zoom.
But
maybe
we
should
do
those
kind
of
work
sessions.
J
D
That
having
a
face
to
face
not
so
much
to
meet
everybody,
although
that,
obviously
that
would
be
nice
but
having
the
face
to
face
to
have
the
whiteboard
discussions
to
hash
through
some
of
these
things
in
person
would
have
been
nice.
I
agree
at
some
point
and
that's
why
we
that's
what
we
tried
to
do
with.
D
I
think
those
you
know
those
marathon
sessions
that
we
had
a
few
weeks
ago
to
try
to
force
through
some
of
these
discussions,
and
I
think
of
having
a
having
a
well-defined
face-to-face
event
for
the
interop
would
have
been
a
wonderful,
forcing
function
as
well
right,
because
people
don't
want
to
show
up
and
travel
unless
they
have
running
code,
so
that
really
forced
them
to
do
it.
D
It's
a
little
too
easy
to
have
other
things
take
higher
priority
when
you
don't
have
a
face-to-face
meeting
to
go
to,
but
okay,
so
it
sounds
like
people
are
saying
it's
just
bad
timing,
bad
situations,
people
are
busy
with
other
things.
It's
not
so
much.
The
specs
are
are
the
wrong
way
to
go.
It's
just
it's
just
it's
just
it's
just
bad
timing,
for,
like
paraphrase
that's
what
I'm
kind
of
hearing.
J
Right
yeah
because
like
when
I
was
doing
the
presentation
like
for
me,
so
I
basically
aggregate
all
the
events
possible
in
my
company
or
trying
to
and
like
I'm
so
bored
of
like
going
and
read
the
documentation
of
github
and
then
octa
and
then
whatever
other
systems,
to
try
to
understand
how
they
do
their
web
books.
What
is
their
events?
J
So
that's
my
envision
and
I
think
it's
it's
some
neat
things
to
have,
so
I
I
don't
think
we're
on
the
wrong
path.
Okay,.
M
Just
a
thought
that
there
are
different
degrees
of
interoperability
and
I
think
we
so
the
standard
so
far
is
nice
because
everybody
can
use
the
same
tooling,
like
the
sdks
and
so
on,
but
with
discovery
and
subscription.
It's,
I
think,
another
degree
of
interoperability,
because
you
really
can
also
link
different
infrastructures
at
some
point,
and
I
think
that's
just
something
that
takes
more
time
until
also
the
the
thinking
evolves
and
those
scenarios
emerge.
M
D
Okay,
yeah,
okay,
maybe
I'll
just
try
to
shelve
my
worry
for
a
little
while,
it's
just
I,
I
guess
the
whole
kubecon
event,
maybe
step
back
and
look
at
this,
because
I
was
looking
at
the
the
charts
that
I
put
together
for
the
kubecon
europe.
M
So
I
plan
on
currently
collecting
some
ideas
to
increase
that
the
primer
we
started
for
discovery
and
and
subscription
in
a
similar
way,
the
scenarios
claimants
already
outlined
in
the
schema
registry.
I
think
it's
still
a
pull
request
right
with
the
authorities
and
the
different
replications
and
things
like
this
yeah.
So
my
ideas,
perhaps
we
then
again
start
also
discussing
a
bit
more
discovery
and
things
like
this.
Okay.
D
Okay,
well,
I
don't
want
to
talk
anymore
about
this.
I
mean
it's
just
so
kind
of
winding
at
some
point
I
guess,
but
I
just
wanted
to
to
make
to
find
out
what
you
guys
are
thinking
about
it
if
at
all,
so
thank
you
for
that.
Okay,
in
that
case,
in
terms
of
going
back
to
the
interop,
then
I
guess
the
best
we
could
do
at
this
point
is
I'll,
maybe
offline,
poke,
some
folks,
I
know
clemens
was
supposed
to
be
working
on
some
stuff.
D
I
think
scott
is
working
on
some
stuff
I'll
poke,
both
those
folks
offline
to
see
how
they're
doing-
and
maybe
we
can
quickly
revisit
this
again
on
next
on
next
week's
call,
even
though
we
don't
have
a
interrupt
call
scheduled
for
next
week,
maybe
on
the
main
call
just
ask
people
to
update
their
status
in
this
page
right
here.
Just
so,
we
can
see
where
people
are
again,
because
I
think
we
may
need
a
little
bit
of
a
nagging
reminder
for
folks.
So,
okay,
all
right
any
other
topics.