►
From YouTube: Spec 3.0 meeting (March 29th 2023)
Description
B
C
A
A
So
I've
we
don't
have
anything
for
the
agenda,
so
if
any
of
you
folks
have
anything,
you
want
to
bring
up
so
far.
Otherwise
I
think
I
can
give
a
rough
update.
B
C
Yes,
one
thing
so
hiker
was
saying
on
the
channel
today
that
he
wanted
to.
He
wanted
to
bring
on
the
topic
of.
We
should
not
be
using
avra
for
headers
or
defining
headers
that
doesn't
make
sense,
and
so
I
replied
on
the
channel,
but
I
just
want
to
continue
the
conversation
here.
So
I
replied
saying
that
it
makes
sense,
because
Kafka
users
use
avra
for
both
things
right
for
for
headers
and
for
and
for
payload.
C
C
He
meant
that
that
yeah
like
if
we
got
if
we're
I,
mean
I'm
speaking
for
himself
so
I'm,
not
sure
if
he
meant
that
or
not
but
I'm
I'm.
Assuming
that
his
concern
is
that,
if
ever
is
possible,
then
protobuf
should
be
possible
there
as
well
right
or
might
be
possible
as
they
as
well
there
and
and
yeah.
That's
that's
not
usually
the
case.
I
I've
never
seen
that
so
so
yeah
or
might
be
I,
don't
know.
C
In
any
case,
what
doesn't
make
sense
is
that
we
Define,
for
instance,
server
variables
and
channel
parameters
and
all
this
stuff
with
with
any
schema
language
that
doesn't
that's
an
Overkill
to
me,
like
that's
too
much.
That's
something
specific.
These
things
don't
go
beyond
the
async
API
right,
so
not
unlike
headers
and
payload.
These
things
are
going
Beyond
this
in
kpi
scope
right,
but
General
parameters,
it's
something
for
a
Singapore,
nothing
else,
and
so
everybody
apples
something
this
is
for
us.
A
Yeah,
so
you
talk
you're,
basically
talking
about
you,
want
it
as
simple,
like
not
having
two
schemas
for
one
for
headers,
basically
and
one
for
payloads
and
having
two
different
sets
right.
A
A
C
Of
today,
yes
of
today,
schema
format
only
affects
the
payload,
but
not
the
headers
yeah.
Yes,.
C
A
C
Or
yeah
you
can
have
enorms
I
mean
for
headers
I
will
give
it
the
same.
Let's
say
priority
as
in
payload
headers
and
payload
to
mirror
to
something
in
terms
of
like
they.
They
both
should
be
first
class
citizens
right,
so
headers
and
payload
could
be
defined
with
abroad.
So
so
we
should
allow
people
to
have
schemas
full
schemas
in
the
headers
and
so
on.
That's
completely
fine.
What
I
mean
is
that
for
several
variables
and
for
parameters,
Channel
parameters,
those
two
things
in
my
in
my
opinions
will
not
be
called
schemas.
C
You
know,
like
I,
mean
not
cold
schemas.
We
can
call
them
schemas,
but
these
will
be
using
like
super
simplified
versions
of
schema
like
enums
default,
the
type,
maybe
not
even
the
type,
because
yeah
in
the
end,
everything
is
a
string.
Once
you
are
a
channel
parameter,
everything
is
a
string
once
you
are
in
a
server
variable.
Everything
is
a
string,
so
it's
it.
It
reaches
a
point
where
everything
is
Extreme
but
yeah.
That's
debatable!
Maybe
type
is
it's?
Okay,
it's
okay
there,
but,
but
that's
it.
C
Maybe
an
innum,
maybe
a
default,
and
in
the
description
and
a
type
and
that's
it
and
I,
don't
think
I
don't
know.
Maybe
you
can
have
minimum
and
maximum
in
case
there's
a
number.
C
That's
it
I,
don't
think
anyone
else
will
have
any
other
needs,
like
any
other
needs
for,
for
that
specific
value.
Right,
like
value
in
in
my
channel
parameter,
it's
super
complex
to
validate
and
or
super
complex
to
Define.
No
must
probably
I
mean
in
in
my
whole
career.
All
I've
seen
is.
C
This
is
just
Dynamic
value.
This
is
a
dynamic
string
and
it
means
that
that's
it
I
don't
care
it's
a
string.
It's
not
even
constrained
constrained,
or
at
most
it's
constrained
to
have
a
few
set
of
values,
and
that's
it
like,
for
instance,
here
you
can
have
the
channel
for
the
US
for
Europe
and
for
whatever
so
there
there
you
can
put
us
EU
and
whatever
other
values,
but
that's
it.
So
that's
all
I've
seen
in
my
whole
career
that
you
really
have
that
simple
kind
of
sets
on
the
channel
Maybe.
C
A
C
A
C
C
Yes,
yeah
could
be
a
constraint
as
well.
If
you
think
about
it,
it's
a
constrictor.
This
is
super
simple
constraint,
yeah,
it's
a
super,
simple
constraint.
It
is
not
like
I,
don't
think
we
should
be
supporting
in
Channel
parameters
or
server
variables
like
all
of
any
of
not
you
know.
If,
then
else
you
know
not
to
mention
Avro
protobuf
and
all
this
stuff
right.
That
would
be
way
too
much
right
for
those
specific
things
that
doesn't
make
sense
to
me.
It's
just
it's
too
much.
C
I
went
too
far,
maybe
for
the
Simplicity.
For
me
to
December,
this
is
yeah:
it's
a
scheme,
it's
only
there,
but
it's
it's
offered
way
too
much
capabilities.
And
now
it's
a
problem.
A
But
don't
you
think
it
will
also
on
the
other
side,
if
you
change
what
a
parameter
or
how
a
parameter
is
defined
and
even
a
hitter,
don't
you
think
that
it
also
complicates
the
spec
itself
to
be
more
complex,
to
understand
and
remember
where
what
fits,
what,
for
example,
that
you
need
to
remember
that
parameters
can
only
be
five
keywords,
let's
just
say
five
keywords:
instead
of
having
the
full
just
a
Json
schema,
even
though
that
you
don't
use
all
of
those
keywords,
we.
C
Or
most
of
them
it's
we
make
this.
This
keywords
the
same
as
in
Json
schema.
That's
it
sure.
So
you
don't
have
to
you.
Don't
have
to
remember
them.
A
C
Yeah,
that's
what
I
mean
that
we
should
make
two
things.
The
same,
maybe
give
a
little
bit
more
power
to
the
to
those
like.
Maybe
you
can
have
minimum
maximum
if
it's
a
number,
but
they
cannot
by
any
mean
being
an
object
or
an
array,
because
how
how
is
it
supposed
to
be
serialized,
then
there
are
many
ways
to
see
realize
an
array
to
shrink.
C
So
if
we
support
arrays,
then
we
should
be
saying
how
this
will
be
serialized
to
string,
because
the
channel
is
a
string
in
the
end
right,
so
it
has
to
be
serialized
as
a
string
at
some
point.
So
if
type
is
array,
then
we
also
have
to
specify
and
the
way
is
utilize.
It
is
like
this
because
there
are
multiple
ways
to
serialize
a
an
array
so
and
so
I
will
not
even
do
it
at
all,
because
why?
Why
would
you
have
want
to
have
a
an
array
and
a
channel
name
right?
C
An
Channel
address.
C
C
A
B
A
C
C
That's
the
only
thing
that
we
yeah,
because
they're
I
mean
no.
Others
is
no
system
in
the
world.
Supports
I,
don't
know
headers
as
an
array
or
headers
as
a
bowl
in
it's
always
a
map
of
string
string
string.
Actually,
it's
always
a
mapped
of
strings
and
strings.
Those
string,
values
might
be.
Numbers
means
your
license
string
but
yeah.
It's
like
okay,
fair
enough.
Let's
just
give
people
the
chance
to
do
it
to
use
the
string
to
any.
A
C
I
think
that's
the
case
for
every
for
every
system,
but
I
haven't
checked
but
I,
don't
recall,
being
able
I
mean
if
you
pass,
you
can
pass
a
number.
You
can
pass
the
word
true
or
false,
but
I
think
they
will
be
sent
as
a
string,
not
a
special
values
or
anything
like
that.
I
may
be
wrong
and
there
might
be
some
systems
who
have
which
have
more
than
than
one
type.
It's
just
not
not
just
strings.
C
A
C
We
can
have
string
to
Primitives,
but
now
object
or
arrays
or
anything
like
that,
nothing
more
complex,
and
then
you
can,
you
can
give
it
some
meaning.
Okay,
this
is
a.
This
is
a
a
number
that
goes
that
has
this
minimum
and
this
maximum
and
so
on.
C
You
might
want
to
apply
whatever
rule.
You
want
to
apply
there
whatever,
but
but
that's
it
only
Primitives,
because
they
are
they're
only
yeah
I
mean.
C
C
Yeah,
what
I'm
saying
is
that,
at
the
very
least,
I
think
a
different
list,
several
variables
and
Channel
parameters
should
not
be
schemas,
so
not
be
referenced,
should
not
be
defined
with
a
schema
object.
No
then
headers,
it's
our
Google,
but
we
can
put
constraints
there.
Whatever
format
you
use,
Avro
whatever
it
has
to
be
a
map
of
strings
to
Primitives
of
this
kind,
blah
blah
blah
and
that's
it
fixation
schema.
You
know
how
to
do
it.
If
it's
ever,
you
know
how
to
do
it.
C
You
know
if
it's
protobuf
the
same
thing,
but
the
header
can
a
header's
object
cannot
be
a
number
because
and
then
in
a
property
cannot
be
an
array
or
or
an
object.
So
it
cannot
go
deeper
than
just
one
level
properties
you
know
or
Properties
or
Fields
or
or
items
or
how
you
call
it
in
upload
and
other
forms
right.
So
because
that's
too
much
I
will
put
it
just
definition
there.
It
can
still
be
referencing
a
schema,
but
not
whatever
schema,
so
implementations
should
fail
if
they
find
something
else.
A
B
A
The
one
with
because
this
discussion
is
directly
related
to
how
we
want
to
reuse
schemas
right,
because
because
we
want
to
reuse
schemas
with
schema
format
together,
it
makes
the
problem
or
all
of
a
sudden.
Then
the
problem
of
headers
and
parameters
are
included
in
this
scope
of
the
issue.
Right,
yep.
C
Be
split
out
because
actually
Haiku
can
just
focus
on
on
saying
that
on
doing
this,
reusable
schema
formats
and
so
on
of
different
formats,
and
that's
it
that
will
that's
all
we'll
see
he
will
have
to
do
and
and
then
on
a
separate
PR.
We
just
we
just
removed
the
from
server
variables
and
from
schema
I'm
from
Channel
parameters.
We
remove
the
the
reference
to
schema
object.
C
Let
me
just
look
at
that
and
that's
it.
We
cut
there
and
that's
another
problem
anymore,
but
we
will
have
to
add,
of
course,
some
explanations
there,
which
is
completely
fine.
Some
tables
there
of
the
accepted
Fields,
just
mostly
the
same
that
we
have
on
the
on
the
server
variables
might
be
worth
to
I,
don't
know
to
extract
as
a
another
object.
B
C
Simple
schema,
object,
I,
don't
know
something
like
that.
A
super
simple
schema
object.
We
can
call
it
whatever
we
want
and
and
Define
that
sentence
on
the
headers
that
it
has
to
be
a
type
object,
object
with
Keys,
being
string
and
values
being
only
Primitives
of
this
type.
This
type
in
this
type.
A
C
A
B
A
It
won't
be
needed
anymore,
mind
writing
that
on
the
issue,
thanks.
B
C
A
C
It
hiding
slaves
for
this.
B
I
think
there
was
issue
about
parameters
because
some
time
ago,
yeah.
C
C
A
A
A
But
let's
not
bring
that
up
here,
all
right
little
update.
So
in
the
last
meeting
we
talked
with
Alejandra
about
the
docs
changes
and
she
created
this
issue
and
the
website
repository
that
kind
of
encapsulates
all
of
the
different
changes
or
major
changes
to
the
spec
itself
and
then
split
them
out
into
sub
issues.
A
I've
added
all
of
these
issues
as
part
of
the
general
to-do
list
here
just
to
keep
track
of
who's
championing
what
the
Citron
status
of
it
so
I'm
I'm
shrinking
it
between
these
two
issues.
For
now.
A
C
A
All
right
so
yeah,
so
that
has
all
of
them.
We
need
to
figure
out
how
we
can
push
each
of
those
things
and
move
them
forward,
and
but
I'm
I
was
kind
of
hoping
that
Alejandra
would
be
here
today,
but
we
need
to
figure
out
how
how
we're
going
to
do
that.
A
I
also
added
two
more
issues
which
is
the
migration
guide.
So
we
need
some
kind
of
migration
documentation
on
the
website
of
moving
from
version
2
to
version
3.
A
Furthermore,
I've
added
another
to
do,
which
is
we
need
to
figure
out
which
tools
need
to
be
updated
for
version
free
support.
I
know
that
some
tools
already
have
it
I'm,
pretty
sure
Marlena
is
almost
there
might
need
a
minor
I'm,
not
sure,
but
we
have
all
of
these
tools
Beyond
this
parcel
itself
as
well,
so
we
just
need
to
figure
out
what's
needed
and
where
and
which
tools
already
ready.
A
Then
I
added
all
of
the
general
stuff
that
we
had
from
all
of
the
other
work
releases,
which
is
basically
just
about
merging
everything,
but
yeah
I
am
this
one
here
is
gonna,
be
split
out
into
so
I
have
a
specific
section
only
for
having
all
of
the
tooling
information
here,
tooling
progression,
I'm
thinking
about
adding
a
a
section
per
tool
to
State
whether
or
not
its
supports
version
free
unit,
current
state
and
what's
needed,
like
in
general.
A
If
anything
is
needed
and
then
ping,
the
code
owners
with
an
issue
or
something
and
Mark
that
as
being
complete
to
move
forward
and
to
split
out
the
responsibility.
A
Basically,
all
right
back
to
the
update,
I've
I
removed
the
entire
back
burner
section,
because
there's
no
backbone
anymore.
We
have
the
fridge
period
basically
in
place,
besides
those
two
outstanding
issues
or
three,
depending
on
how
you
see
it
with
the
headers
and
parameters.
But
that's
what
that
is
so
I
removed
that
entire
section
and
that's
basically,
the
updates
I-
did
for
the
the
for
this
issue,
and
then
we
had
didn't
Define
anywhere
that
it
was
released
for
June
I've,
defined
that
in
the
issue
itself,
that
it
is
scheduled
for
June.
A
And
that's
it
that's
kind
of
progression
update.
We
still
have
some
out
there
outstanding
issues
for
trades.
C
Magic
matek
said
that
he
hopefully
can
work
on
this
this
weekend,
but
if
not
Sergey
will
probably
jump
into
it.
If
he
can't
do
it
on
the
next
week
or
something
like
that
he's
just
giving
much
at
some
time.
A
C
Didn't
we
agree,
like
I,
mean
softly
agreed
that
it
will
be
under
schemas
existence,
schemas
sectional
components,
right
that
we
will
have
something
like
a
format
for
the
schema
format,
and
then
another
keyword
would
be
I,
don't
know,
schema
itself
and
and
inside
the
schema
keyword
there
will
be
the
schema
and
that's
it.
That's
all
we'll
have
to
do.
C
B
C
As
we,
of
course
well
not
as
long
I
mean
the
only
concern
that
was
there
is
that
we
could
then
cross-reference
different
types
of
schemas
yeah,
that's
true,
which
you
can
anyway,
do
it
now,
right
you
can,
you
can
on
payload,
you
can
have
an
abroad
that
will
reference
a
or
a
Json
schema
in
one
version,
the
reference
is
another
discount
scheme
and
another
version.
So
so
this
is
really
possible,
so
yeah,
but
we
might
want
to
add
some
paragraph
there
saying
that
you
don't
you
cannot
do
this.
C
This
is
not
allowed
and
the
implementations
should
not
support
them.
That's
it.
We
don't
I,
don't
think
we
have
to
complicate
our
lives
beyond
that.
As
long
as
as
it's
it's
explicit
explicitly
mentioned,
then
it's
more
than
enough
I
guess
I
mean
if
I
were
a
haiko
or
haiko
if
you're
watching
the
recording
go
for
it,
make
this
proposal
like
like
this,
if
you
didn't
lose
it
already,
which
I
think
probably
it's
already
done,
and.
A
Or
request
reply,
we
don't
have
any
change
in
the
parcel
yet
or
even
a
proposed
change.
We
do
have
one
for
the
Json
schema
files
for
the
spec
itself.
C
A
C
A
A
Yeah,
just
just
an
issue,
so
we
can
track
it
here
when
it's
merged
and
introduce
changes
form
oh
yeah,
I
think
that's
more
or
less
that.
B
Not
it
then,
so
the
question
is
somehow
also
related
to
2.7
and
the
Proto
buff.
So
do
you
know
like
so
haiko
still
wants
to
push
it
for
2.7,
the
the.
A
A
He
said
he
wanted
to
push
this
forward.
At
least
I
didn't
tag
him,
so
I
don't
know
if
he
wants
to,
but
yeah
I
haven't
heard
any
further,
and
especially
not
regarding
version
2.7
or
whether
he
intends
to
do
it
for
version
free
I,
don't
know.
B
Because
we
still
have
this,
so
one
thing
is
that
we
we
know
that
support
for
referencing,
Proto
and
xsd
and
other
non-json
can
be
done
later,
but
but
we
need
to
prepare
spec
for
it
as
well.
That's
this
dollar
ref
comment
I
made
in
this
request
from
front
you're,
not
sure
if
you
recall.
C
Yeah,
we
have
to
clarify
that
clarify.
B
Description
and
making
sure
like
opening
up
opening
it
up
for
future
that
dollar
ref
is
not
the
only
keyword.
C
C
B
C
C
B
Will
open
it
up
for
more
options
in
the
future,
like.
C
Yeah,
that's
that's
why
I
wanted
to
rename
this
dollar
after
ref,
so
we
leave
it
super
like
super
open
and
generic,
and
then
we
can
just
change
it
to
whatever
we
want,
but
yeah
I.
Think
it's
better
like
this,
like
we
just
changed
the
reference
object.
We
live
it,
hoping,
as
you
said
and
and
I
think
we
should
do
it
on
3-0,
because
this
is
a
very
foreseeable
breaking
change
right.
C
Json
reference
it
uses
Json
pointer
and
you
cannot
use
it
Json
pointer
for
anything
that
is
not
Jason
right,
so
I
don't
know.
I.
C
C
In
the
future,
what
we
could
do
is
that
we
have
this
dollar
rev,
alongside
with
Evernote
kind
of
ref
or
type
of
ref,
which
could
be
I,
don't
know
the
schema
format
expected
to
come
from
this
little
ref,
I,
don't
know
like
protobuf
or
xsd
or
whatever.
So
in
this
case,
depending
on
this,
second
key
should
be
interpreted
in
one
way
or
another.
I.
B
A
B
Yeah,
of
course,
without
the
summary
and
description
because
we
don't
need
it,
but
but
yeah.
Basically,
almost
the
same
I'll
have
to
read
it
again,
but
I
think
it's
going
to
be
almost
the.
C
Same
yeah,
actually,
for
now
it
will
be
just
dollar
ref.
Without
somewhere
in
description,
it
will
be
just
a
dollar
ref,
but
it
will
not
be
the
RFC
whatever,
whatever
right
or
any
distance
schema
defined
thing
it
will
be
defined
by
as
and
then
we
Define
there
that
yeah
it
might
contain
other.
C
Ref
is
not
strictly
just
for
Json
scheme
or
Json
structures,
and
so
on
so
I
don't
know
whatever
we
could.
We
could
put
something
there
and
we
can
even
add
an
optional
field
alongside
the
Rev,
which
will
be
type
or
something
like
that,
and
by
default
is
Json,
schema
or
sorry
is
by
default.
It's
Json,
and
if
you
don't
put
it,
it
will
work
the
same
way
as
now,
but
it
will
only
have
this
value
for
now,
but
in
the
future
we
might
add
more
values
like
product
above,
like
xsd,
whatever.
A
A
As
far
as
I
can
remember
some
of
our
discussions.
So
instead
of
having
this
one
also
include
something
like
you
can
reference
non-json
preferences,
it's
only
in
the
payload
that
you
were
able
to
to
do
that
so
overall,
in
all
other
other
areas
of
the
spec,
you
use
the
reference
up
there
in
the
schema
like
where
you
can
Define
the
schema.
A
Where
and
in
the
schema
format
there,
it
can
be
not
Json
references,
so
we
introduced
something
that
specifically
only
for
that
area
within
the
spec
and
not
for
the
rest,
because
otherwise
you
might
reference
something
like
an
XML
document
in
the
message
object
where
it's
like.
No,
that's
not
what
it's
intended
for.
C
A
C
C
C
So
yeah
and
it
will
work
because
yeah
the
type
is
telling
you
that
this
is
not
a
Json
thing.
It's
a
it's,
a
maximum
thing,
not
perfect,
but
it
will
be.
It
will
keep
adding
on
top
of
the
existing
thing
without
breaking
it
right,
but
we
need
to
do
this
change
now.
Otherwise
it
will
be
a
breaking
change
right.
C
It
will.
It
will,
because
then
people
will
put
dollar
ref
pointing
to
an
xsd
file,
and
but
there
isn't
this
other
field
here
yeah,
but
it
wasn't
there
before.
Well,
no,
it's
not
the
breaking.
It's
not
yeah,
so
the
pricking
change
there's
a
it
will
not
be
forward
compatible,
but
it
will
be
backwards
compatible.
Yes,
so.
A
C
B
I
I'll,
just
it's
trivial
task:
I'll
open
PR
tomorrow,
I,
don't
like
pro
forma.
A
B
Probably
the
same
as
we
do
with
other
stuff,
there
has
to
be
some
release,
Branch,
where
we
can
update
Docs.
So
it
doesn't
wait
for
the
release,
but
we
can
already
do
the
updates
to
deducts.
A
B
But
yeah,
let's
figure
out
we'll
talk:
Fox
Alejandra,
she's.
A
A
B
C
A
B
B
It
is
because
more
contributors,
more
often
we
release
that's.
Why
I
like
this
bug,
we
came
up
with
semantic
releases
like
before.
We
never
had
it
because
you
had
one
PR
every
few
days
and
now
like
it's
like
damn
it.
It's
like
it's
like
a
machine
that
machine
has
no
idea.
B
But
yeah,
let's,
let's
fix
them
one
by
one
but
CLI
it's
different
they're,
just
like
that's
a
different
Beast
like.
B
But,
but
sometimes
like
with
one
PR
from
Civic,
he
has
it
open
for
a
few
months
and
all
all
suddenly
at
some
point
of
time,
the
the
pipeline
stopped
working
and
you
rerun
it
as
much
as
you
want
with
clear
cache
whatever
and
it's
failing
on
installation
of
on
of
rimrath
package
so
like.
Why.
C
Sometimes
have
to
fill
in
that
feeling
with
notify
as
well
deployment
that
you
do
this.
You
have
deploy
and
clear
cache
in
the
play
and
they
usually
take
the
same
time
and.
C
It's
either
not
you
is
it
the
catch
at
all
or
it's
always
using
the
cash
I,
don't
know,
but
actually.