►
From YouTube: CNCF Serverless WG Meeting - 2018-03-29
Description
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.
B
A
H
A
I
A
A
K
A
D
B
A
Sean
here
cool,
thank
you
that
really
everybody
well,
you
guys
are
early
today.
I
said:
there's
Ryan,
are
you
there
hey.
B
A
A
A
And
just
so
you
guys
know
you
know,
obviously
you
don't
mind
if
you
do
it,
but
you
don't
technically
have
to
add
your
last
name
and
company
unless
you
really
want
to
in
there
as
long
as
your
name
is
unique
enough.
That
I
know
which,
like
for
example,
which
Dan
I'm
referring
to
I'll,
get
you
on
the
attendee
tracker,
but
obviously
you're
still
free
to
add
the
other
things
you
want.
It's
just
not
mandatory
I.
A
M
G
A
A
Alright,
with
that
I
think
we're
good
to
go
cool.
Thank
you
guys,
but
Abu.
So
action
items,
I
believe
there's
nothing
really
say
here
other
than
to
remind
people
to
please
complete
your
eyes.
When
you
get
a
chance
reminder,
we
do
have
the
face-to-face
coming
up
the
coop
con.
It
will
be
an
official
meeting
so
far
18
people
there
didn't
say
they're
going
to
show
up.
Obviously
you
can
still
show
baby.
If
you
don't
do.
Doodle
poll
and
I
will
set
up
a
proposed
topic
doc.
Let
me
get
closer
usual
spiel,
so.
M
A
That's
no
fun
if
we
can't
exclude
people
what's
the
fun
of
that,
okay,
so
actually
I
guess
I
should
have
actually
formally
asked
that
question.
Is
there
any
objection
to
making
it
an
official
meeting
as
long
as
we
can
get
a
dialing,
probably
zoom,
and
the
assumption
there
is?
If
we,
for
some
reason
have
technical
problems
and
we
don't
have
zoom,
then
we're
gonna
have
to
say
it's.
Not
an
official
meeting.
Is
everybody?
Okay
with
that.
A
Objection
all
right,
so
no
objection
cool.
Thank
you
all
right.
Let's
move
on
to
the
fun
stuff
PRS.
First,
one
up
is
Clemens
PR
for
use
of
scenarios.
I
hear
that's
going
really
slow
to
her,
so
I
believe.
Last
time
we
agreed
that
we
were
gonna
basically
vote
on
this
one.
This
week
people
should
have
had
more
than
a
week
through
you.
It
so
far.
Are
there
any
actual
I
guess
I
should
ask
since
Clemens
isn't
here
actually
Dan
knows
her
offer
actually,
who.
C
A
A
Going
west,
okay,
cool!
Thank
you
guys
very
much
and
keep
in
mind.
I
should
have
said
this
before
that.
One
of
the
goal
here
on
the
call
is
to
try
to
make
forward
progress
as
best
we
can.
As
with
any
PR,
we
can
always
accept
changes
later
on
we're
going
for
heading
in
the
right
direction.
Not
perfection
mm-hmm,
all
right!
Next,
one
Sarah
contributor
list
is
there
anything
you'd
want
to
say
on
this
one
before
we
ask
the
question.
M
A
A
B
A
M
A
So
this
one
is
just
adding
a
link
to
Clemens
video
that
he
sent
out
an
email
about
I,
live
on
the
22nd
and
a
link
to
the
PowerPoint
itself,
all
right
any
questions
on
that
all
right,
any
objection
to
accepting
it.
So
let
me
go
ahead
and
close
this
okay,
any
objection,
cool
moving
forward,
then
one
two
three,
so
this
one's
been
out
there
for
about
a
week,
unchanged,
I
believe
so
this
one
is
mine.
A
It
basically
takes
the
position
of
having
source
be
defined
as
a
URI
and
that
collapses
source,
ID
source
type
namespace
and
the
source
object
all
into
just
this
one.
You
are
as
I
mentioned
in
the
PR
itself.
The
assumption
here
is
that
people
will
open
up
follow-on,
PRS
to
potentially
extract
information
if
they
think
it's
worthy
of
being
it's
a
separate
attribute,
but
this
at
least
gets
us
a
baseline
at
which
to
work
from
any
questions
or
comments
on
this
one.
E
A
A
A
B
A
M
Well,
I
just
want
to
say
that
like
I
would
really
love
it
if
we
prioritized
changes
to
required
attributes
that
people
think
are
necessary
to
hit
our
point.
One
milestone
I,
like
I,
think
that
this
is
a
kind
of
a
good
way
to
deal
with,
like
a
lot
of
things
that
with
that,
but
anytime,
we
start
out
would
be
nice.
I
just
feel
like.
We
have
really
crowded
working
group
meetings
and
I'd
love
to
focus
on
the
required
days.
Yep.
A
And
just
let
you
know,
one
of
the
reasons
I
did
put
this
one
fairly
high
is
because,
as
we
start
talking
about
other
app,
it's
I
wanted
to
give
people
a
home
in
which
to
put
extensions
like
this.
So,
for
example,
if
we
start
talking
about
a
napkin
that
someone
says
they
need
I
want
someone
to
be
able
to
say
well,
that's
nice,
but
it
should
really
be
an
extension,
and
it
should
go
into
this
document
there.
That
way,
there's
no
ambiguity
about
where
it
would
go
if
it
does
get
into
that
bucket.
A
Without
this
document
there
we
were
sort
of
left
in
this
medalist
state
of
you'll
get
to
find
someplace
else.
I
wanted
people
to
feel
like
if
they
wanted
to,
they
could
have
a
home
for
it.
That
was
the
reason
I
kind
of
prioritize
this
one.
That
way,
if
there
is
any
objection
or
a
concern
at
all,
I
have
no
problem
with
deferring
this
to
later,
but
that
was
my
thinking.
A
J
Doug
this
is
Austin
I'm
gonna.
Second,
what
Sara
said,
of
course,
that
is
super
important.
At
the
same
time,
I
do
like
this
idea
and
I
recognized
the
intent
and
I
think
we
actually
came
up
with
this
at
the
end
of
last
year
and
it
simply
just
fell
through
the
cracks
we
got
lost
in
the
shuffle
I'm
glad
to
see
this
come
back.
I
think
it
will
solve
a
lot
of
problems.
Okay,
all
right.
A
E
A
A
A
All
right
cool!
Thank
you
very
much.
Next
Jason
serialization,
okay.
So
this
one's
open
up
six
days
ago,
this
one
takes
a
first
pass
at
defining.
What
a
cloud
event
would
look
like
as
realized
in
Jason
I
believe
Thomas
had
a
suggestion
for
a
slight
change
to
there,
which
I
hope
we
make
a
little
doing
a
follow-on
PR,
but
so
the
bulk
of
the
PR
has
been
unchanged
and
in
basically
six
days
the
only
thing
I
did
change
today
was
based
upon
someone's
comment.
A
I
made
it
clear
that
non-mandatory
properties,
as
defined
by
the
spec,
are
not
required
to
appear
in
the
Jason
I
thought
that
was.
It
was
a
implicit
but
I
trying
to
make
it
explicit.
But
if
they
do
appear
here,
then
they
have
to
adhere
to
the
description
described
above
and
then
I
just
added
a
quick
example
because
sometimes
examples
help
people
understand.
What's
going
on
there,
any
questions
on
this
one
well.
M
I
I
think
that
this
is
I
like
that
sort
of
work
on
the
JSON
format
and
I.
Don't
disagree
with
it.
I
just
is
this
gonna
cause
that
every
time
somebody
proposes
an
attribute,
we
have
to
change
it
in
two
places.
A
M
J
J
I
So
I
think
some
of
the
fields
here,
since
we
have
the
you
know:
New
Year's
cases,
if
plan
just
I
mean
just
commit
it
right.
There
might
be
some
question.
I
saw
some
question
about
some
of
the
fields.
Without
you
know
they
should
be
there
or
not.
So
this.
A
B
B
A
A
A
M
E
Obviously,
yes,
yeah
I'm
I
am
a
huge
advocate
of
this
use
case
and
I
can
talk
a
bit
about
like
why,
for
example,
sampling
is
not
averaging
or
like
that
I
am
not
100%
sure
like
I've
been
ambivalent
about
whether
this
is
a
top-level
field
or
an
extension.
I
think
that
Ben's
concern
was
that
you
know
like
as
a
provider
of
an
observer
ability
system.
This
needs
to
be
something
that
is
not
just
like
the
few
hipster
services
might
support
being
observable.
L
A
I
actually
had
a
conversation
with
him.
I
can't
remember
it
was
last
week.
He
definitely
did
indicate
to
me
that
he
may
be
pulling
back.
I
didn't
get
the
indication.
It
was
because
of
not
being
able
get
this
through
or
other
changes.
It
was
more.
He
just
he
was
busy
in
there
a
small
start-up
on
this
particular
change.
He
did
seem
to
have
the
feeling
that
I
apologize
for,
if
I'm
for
speaking
for
him,
but
this
was
my
interpretation.
What
he
said
was
that
his
general
sense
was.
A
The
working
group
would
probably
not
accept
this
as
a
top-level
thing
and
that
showing
up
as
an
extension
I
think
was
okay
with
him,
but
I
have
to
go
back
and
double-check
in
my
notes,
or
our
slack
chat
that
we
had,
but
that
was
not
going
to
be
a
required
thing
at
best
was
to
be
optional
but
probably
fall
into
the
extension
bucket.
Well.
M
N
M
A
What
I
think
my
mind,
one
of
things
to
consider
is
something
that
appears
in
the
spec,
even
as
optional
cannot
be
easily
removed
without
a
version
bump.
That's
something
to
think
about,
and
the
other
thing
is
when
people
have
talked
about
the
well
defined
extension
document
that
we've
just
approved
a
lot
of
you
love
you
that
has
an
experimental
place
so
if,
for
example,
people
aren't
sure
with
how
they
feel
about
something
like
this,
it
could
go
into
that
particular
bucket.
A
N
You
know
I
could
give
you
just
another
example
is
like
better
size.
You
know
maybe
I'm
getting
a
batch
of
events.
It's
at
least
the
same
level
of
importance
as
having
a
sample
rate
because
then
I
when
I
decipher
the
message.
I
know
if
it's
an
array
or
individually.
So
if
I,
if
that's
the
level
of
things
that
we
want
to
ratify
here
and
I,
could
come
with
a
dozen
that
are
the
same
level.
I
rather
keep
it
minimal.
Yes,.
A
M
We
should
actually
address
the.
How
are
we
going
to
have
multiple
teams?
You
know
who
are
all
aligning
on
a
attribute
which
needs
to
be
produced
by
one
team
and
consumed
by
another.
You
know
like
going
to
work
and
they.
This
is
exactly
why
I
proposed
this
governance
model
of
having
multiple.
If
multiple
active
members
are
like,
we
want
this
thing
and
it
will
help
us
interoperate.
Then
I
think
that's
a
really
strong.
M
You
know
argument
towards
doing
something
rather
than
just
saying
anytime,
somebody
thinks
maybe
it's
not
required
or
high-value,
then
we
just
throw
it
into
the
extensions
bucket
and
like-
and
this
is
and
I
think
that
this
conversation
is
really
important.
It's
just
frustrating
to
have
this
conversation
before
we
have
aligned
on
the
required
attributes.
So.
A
I
I
It's
not
needed
I
think
if
it
helped
with
interoperability
and
sported
by
you
know
several
teams
I
think
you
know
we
should
consider
it
it's
an
important
attribute,
because
there
are
other
fields
that
are
optional
too,
which
are
part
of
these
attributes,
but
either
we
have
a
good
guideline
of
how
we,
you
know,
decide
I.
Think.
B
A
So
if
we
want
to
change
the
governance
model,
that's
fine
parrots
air
has
a
PR
and
we
should
talk
about
the
governance
change
in
that
PR
and
that
and
that's
fine,
but
right
now
in
front
of
us
we're
looking
at
this,
and
there
are
that's
in
my
mind,
there
are
two
different
proposals
on
the
table.
One
is
accept
the
PR
as
it
is,
and
I'm
not
hearing
a
lot
of
support
for
accepting
that,
but
I
could
be
wrong.
The
other
proposal
is
to
accept
the
definition
as
defined
here,
but
move
it
into
the
extension
dock.
A
M
Think
we
should
table
this
until
we
figure
like
I
would
like
to
hear.
I
actually
think
then
makes
a
good
case.
I
would
like
to
hear
that
multiple
vendors
would
like
this
as
is,
and
then
I
really
want
it
in
the
spec.
Otherwise,
because
if
it
helps
have
those
vendors
participate
in
order
to
have
alignment
on
the
naming,
then
I
would
advocate
for
it.
M
A
A
L
A
Okay,
so
tell
you
what
I
will
make
a
comment
in
this
PR
that
says
it's
good
this!
Do
it?
That's
it
that
says
the
current
status,
because
I
don't
have
rush
people
so
say
the
current
status,
and
let
people
know
that
we're
probably
make
a
decision
next
week
and
what
the
various
options
are
so
be
prepared,
how's
that
yeah,
if.
A
So
Sarah
I
have
to
admit
I'm
very
confused
you,
you
start
off
the
converse
that
basically
the
phone
call
talking
about
how
you
want
to
be
able
to
go
through
things
in
essence,
faster.
All
right
summarize,
you
know
I'm
being
kind
of
blunt
with
it,
but
basically
you
want
to
be
able
to
move
faster,
and
yet,
when
we
try
to
move
faster,
you
want
to
talk
about
governance,
about
moving
faster,
which
talk
about
governance,
doesn't
necessarily
us
faster
I.
Would.
M
A
M
M
What
I
would
like
on
the
agenda
for
next
week
is
either
people
have
PRS
to
the
required
attributes,
or
we
all
accept
that
it
is
sufficient
and
we
are.
We
have
the
required
attributes
that
would
allow
us
to
be
moving
on
to
making
samples
and
and
validating
that
those
required
attributes
work.
I.
Yes,.
A
M
M
A
L
About
this
have
a
leave,
a
comment
on
that
PR.
We
don't
add
it
to
the
agenda
for
next
week,
because
I
suspect
it
will
take
more
than
one
week's
worth
of
work.
And
then
we
move
on
to
the
conversation
about
the
next
PR
about
Thomas's,
pure
understanding
that
will
have
to
change
it.
In
light
of
the
things
we
just
accepted
right.
A
I
A
H
Side
I
feel
like
what's
missing,
is
clear,
clear
criteria
for
what
it
means
to
be
a
core
attribute
in
the
first
place,
and
so
we
have
these.
Discussions,
like
my
proposal,
would
be
that
a
core
attribute
is
something
descriptive,
generically
like
the
source,
the
type
content
type
so
on,
but
any
any
attribute,
that's
related
to
the
context
in
which
an
event
is
being
used
or
emitted
sampling
batching.
Those
types
of
things
should
be
extensions.
If
anything
and
not
core
attributes.
H
K
H
And
within
the
core
attributes
there
are
required
and
optional.
Yes,
I
think
what
I'm
suggesting
is
that
if
something
is
like
sample
type,
is
clearly
not
going
to
be
relevant
in
in
all
systems.
It
was
like
an
extension
as
opposed
to
the
content.
Type
is
relevant
in
all
systems.
Go
mark,
I,
think
this
is
a
really
good.
All.
I
Have
another
suggestion
so
in
the
extensions
ill
I
mean
it's
mentioned,
it's
not
officially
supported
by
this
workgroup,
something
like
that.
Somebody
like
that
I
think
that
what
why
we
put
that
if
people
spend
time
discussing
that
and
Grille
mate
I
think
you
know,
you
know
why
we
know.
We
said
that
you
know
it's
not
officially
supported
by
this
work
group.
Okay,.
A
I
G
G
A
A
A
E
So
I
wanted
to
like
my
theory
about
like
why
some
of
these
fields
have
been
contentious.
Is
that
when
you
look
at
the
trees,
you
lose
the
forest,
so
I
want
to
at
least
be
able
to
discuss
the
forest
with
everyone,
and
then
we
can
like
if
we
see
that
some
subset
is
non-controversial
and
some
stuff
is
controversial,
I'll
certainly
break
it
up.
I
just
want
to
break
it
up
according
to
actual
controversy,
as
opposed
to
just
pre-emptive
and
cautious
pieces.
E
I
think
the
examples
might
be.
The
best
way
in
the
description
might
be
the
best
way
to
navigate
this
sorry,
I
wonder
where
to
go
yeah,
okay,
so
there's
a
couple
motivating
factors
here:
one
is
I
was
trying
to
bring
back
some
of
my
original
inspiration
of
using
the
sto
vocabulary
for
what
they
call
respect.
I
think
they
call
it
source
as
well
and.
E
Also,
some
of
my
confusion,
I
had
had
with
name
space
which
we
already
you
know
now
just
removed
I
think
there's
two
things
that
ultimately
need
names
facing
the
source
and
the
event
type.
So
I
had
to
cut
that
and
put
it
into
the
source
where
Authority
and
path
or
break
basically
re
breaking
up
the
URI
for
source
I,
don't
want
to
say
mind
if
we
want
to
instead
merge
Authority
and
path
and
keep
it
as
one
thing
if
we
specify
that
the
or
I
must
be
a
subtype
of
the
URI
spec.
E
That
includes
the
authority
component,
but
these
are
basically
source.
Authority
source
paths
are
very
explicitly
defined
as
portions
of
a
URI
that
make
it
an
absolute
reference
in
practical
use.
I
find
this
very
useful
that
the
path
is
something
that
generally
the
you
know,
the
C
code
inside
the
software
has
decided,
whereas
the
Authority
is
probably
based
on
some
config
about
the
deployment.
E
L
I
L
I
E
L
I
Yeah
I
think
this
car
is
the
information,
but
you
know
for
application,
for
how
do
we
know
which
one
like,
for
example,
the
previous
example
a
example
right?
You
have
those
paths
you
have
the
so
here
you
have
the
difference.
You
have
sauce
path,
sauce
label
event
time.
How
could
a
plication
know
which
field
which
one
will
be
used
to
correlate
all
those
different
events
from
different
events
or
sit
together?
Yes,.
I
E
I
think
that's
in
some
ways
intentional
that
the
its
decoupling
information
so
I
wanted
to
make
sure
that
the,
for
example,
the
IOT
sensor,
didn't
need
to
understand
the
way
that
any
given
application
would
be
correlating
the
data
and
that
it
could
provide
non
structured
information
where,
for
example,
the
the
application
that
allows
a
maintenance
staff
to
make
sure
that
something
is
working
could
look
at
a
particular
house,
or
maybe
the
vendor
itself
of
certain
boilers
could
make
sure
that
they're,
not
overheating.
I.
E
I
A
E
A
But
yeah
I
mean
obviously
not
extensions
to
the
person
putting
them
there
or
they
may
not
be
extensions.
The
person
put
in
there
but
from
the
spec
perspective
it's
sort
of
a
open
space
where
you
can
stick
stuff
that
you
need
to,
but
make
it
clear
that
it's
route,
that
these
aren't
just
general
extensions,
they're
extensions
that
are
related
to
the
source.
M
Specifically,
my
reading
of
this
is
they:
they
are
things
that
the
sort
the
source
is.
The
sources
would
be
consistent
in
how
they
advertise
these
labels,
so
they
they
get
to
define
them,
but
then
they
don't
like
just
arbitrarily
use
them
like.
You
know
what
it
like.
The
idea
that
they're
consistent
for
a
particular
source
so
that
the
consumer
actually
says.
Ok,
this
source
has
this
set
of
labels,
and
you
know:
maybe
they
wouldn't
always
be
there,
but
they
need
I
mean
I,
don't
need
to
overly
specify
it,
but
the
idea
is
they're.
M
I
Your
point
is,
you
know,
for
different
vendors
if
they
have,
the
same
type
of
you
know
saw
source
I
was
a
spree
time
same
same
type
of
sensors,
but
those
sensors
produced
by
different
vendors.
You,
like
a
standard
way
of
you,
know
how
they
define
all
these
source
paths
and
the
labels,
so
the
consumer
can
know
how
to
use
them
where
I
have
a
standard
way
of
using
it.
Is
that
what
children
well.
M
Just
that
I
mean
I
think
that
I,
what
I'm
saying
is
that
I
think
that
I
see
that
that
I
see
patterns
where
sources
provide
kind
of,
like
some
sis
consistent
metadata
of
their
own
devising
right.
That,
then,
is
useful
to
the
consumer,
but
I
think
that
sort
of
you
know
just
kind
of
extends.
This
notion
so
I
think
that,
like
extensions,
it's
one
way
of
putting
it.
M
M
This
is
precisely
addressing
the
the
need
of
filtering,
so
this
is
when
there
is
certain
metadata
for
the
event
which
needs
to
be
kind
of
like
on
the
outside
of
the
envelope
that
then
the
router
it's
totally
the
source
decides
this
right.
The
source
is
like
hey
there's
some
of
my
met.
I
have
some
metadata
that
I
think
would
be
useful
for
routing,
and
the
source
has
a
place
to
promote
that.
Is
it
Thomas
I'm
articulating
that
I
want.
E
To
be
very
careful
with
the
word
source,
because,
especially
with
hosted
software,
there's
multiple
actors
involved,
it's
a
very
practical
example
that
happened
to
me
in
the
past
was
you
know
during
Google
I/o
last
year,
Google
is
rolling
up
some
new
software
and
they
really
really
didn't
want
to
accidentally
roll
out
the
software
and
break
a
demo
of
Google
I/o.
E
So
Google
cloud
platform
has
the
concept
of
labels
built
in
and
if
I
just
annotated
my
project
with,
like
a
certain
key
and
certain
value
that
said,
I
want
to
opt
out
of
changes
during
this
time
window.
Then
someone
else's
software,
who
you
know
the
IT
enforcement
policy
at
Google-
would
change
the
behavior
for
my
specific
Google
Cloud
Storage
bucket,
and
that's
like
the
type
of
thing
where
it's
it's
not
even
the
in
the
hosted
service
case.
For
example,
it's
not
even
the
person
who
created
Google,
Cloud
Storage.
E
That
knows
too
too
inject
a
specific
label.
They
have
given
a
platform
for
the
operators
of
the
system
as
a
whole,
whereas
experiments
it's
very
often
that,
for
example,
the
router
or
the
source
has
a
non
canonical
field.
They
understand
they
publish
and
they
submit
where,
as
labels
are
left
to
the
developers
or
operators
of
this
cluster.
N
Yeah,
you
know
I
two
comments,
one.
If
I'm
looking
at
the
hosted
service
example,
I,
wouldn't
classify
something
like
Bardot
JPEG
as
a
source
I
think
it
served
the
thing
we're
talking
about.
It
runs
I
started
to
think
that
generated
the
event,
and-
and
the
second
is
that
labels
doesn't
have
to
be
confined
to
the
source.
There
may
be
labels
that
are
injected
throughout
the
way.
So
why
not
generalize
the
notion
of
labels.
E
N
If
it
sounds
to
me
like
we're
confusing
the
term
sure
source
once
again,
because
the
if
you
take
the
example
I
published
with
the
s3
topic
that
comes
from
SMS,
that
comes
from
s3
etcetera,
you
know
the
source
is
not
in
the.
In
this
event,
the
source
is
not
read
the
object
that
was
generated
in
a
desperate
bucket
he's
either
there's
an
S
service
or
the
s3
service.
The
reports
about
something.
M
M
It
says
labels
associated
with
the
resource
that
emitted
the
event
right,
allowing
filtering
or
routing
based
on
non-hierarchical
ike.
I
think
that's
pretty
clear
that
it
is
exclusive
of
something
that
is
like.
Oh
I'm
going
to
add
something
that
says
this
was
transmitted
SNS,
which
doesn't
preclude
having
a
as
we
discussed
in
last
call,
doesn't
preclude
having
some
middleware
that
access
the
source
and
says
no
actually
I
know
that
this
I
Oh
t
device
was
in
a
particular
location,
so
I'm
going
to
annotate
it
because
I
am
making
my
IOT
device
more
efficient
right.
N
It's
actually
like
putting
the
attachment
name
in
the
email
as
source
versus
the
you
know,
from
in
the
email
as
source
and
the
second
again
I
round
the
labels.
If
there
is
a
need
for
that,
I
would
argue
that
labels
needs
to
be
generic,
not
specific
to
source,
because,
as
a
intermediate
point,
I
may
want
to
inject
label
because
I've
inspected
the
message
and
I've
noticed
something
and
I
may
want
to
label
it.
N
H
A
H
A
Yes,
so
to
me,
I
tend
to
agree
with
that,
but
I
think
you're,
saying
they're
marking
that's
one
of
the
reasons
that
I
proposed.
The
idea
of
a
single
uri
is
because
ultimately,
the
source,
I
guess
the
consumer,
because
I
got
to
understand
it.
You
know
they're
going
to
decide
what
data
to
include
and
stuff
and
and
the
mini
start
splitting
these
things
out.
You
then
have
to
get
agreement
on
how
many
different
ways
you
split
it.
A
What
does
each
thing
actually
mean
and
each
source
may
actually
have
their
own
way
of
interpreting
those
things
or
their
own
set
of
data?
They
want
to
pull
out,
whereas
if
you
just
say
it's
a
URI
and
the
receiver,
because
they
know
they're
talking
to
me-
is
going
to
be
forced
to
extract
the
things
that
I
tell
them
to
do
extracting
the
right
way
to
extract
it.
A
Then
we,
as
a
spec
author,
don't
have
to
get
into
that
business
of
doing
that
for
them
or
trying
to
find
a
normative
way
to
define
how
to
split
it.
You
sort
of
let
the
URI
producer
define
it
themselves.
The
only
thing
I've
heard.
That
would
be
worthy
of
making
things
get
pulled
out
is
if,
for
some
reason,
the
the
process
of
extracting
information
becomes
a
burden
from
a
performance
perspective.
That
was
one
of
the
arguments
that
I
heard.
That
really
made
me
think.
A
E
Some
of
the
other
cases
that,
as
the
vendors
for
a
system
that
will
handle
these
events,
the
stricter
we
can
define
the
the
meaning
of
these
properties,
the
better
software
we
can
build
on
it.
So
whether
source
authority
and
path
are
joined
as
one
source
or
not.
I
would
like
to
at
least
specify
that
the
URI
must
include
an
authority
component.
This
is
fairly
important
for
knowing
how
to
actually
set
up
the
triggers
that
will
fire.
These
events,
which
I
know
is
not
currently
like
triggering,
is
not
currently
part
of
this
spec.
E
But
do
you
want
to
make
sure
that
the
I'm
kind
of
indirectly
trying
to
solve
that
problem
with
this?
As
far
as
the
labels
I
put
a
comment
in
github
asking
for
votes,
Plus
up
or
down
whether
or
not
it
should
be
source,
prefixed
or
not?
I
am
a
little
hesitant
and
putting
in
the
URI,
because
I
think
it
is
valid
that
you
know.
E
For
example,
if
I
have
a
simple
event
that
tracked
when
someone
clicked
post
on
a
URI
and
that
you
or
I
had
already
a
query
fragment,
but
the
resource
backed
by
that
URI
had
labels,
that
are,
you,
know,
internal
and
operational.
These
are
very
different
types
of
concepts
and
they
have
very
different
integrity
levels
as
well.
One
is
user
input,
1
is
developer
input,
so
I
do
think
that
these
actually
adds
a
lot
of
value
being
separate.
M
Ok,
so
we've
got
seven
minutes
left
I,
wonder
whether
if
we
say
that
if
source
labels
were
in
a
separate
PR,
we
could
have
a
separate
discussion
around
whether
they
should
be
source.
Labels
are
just
labels
and
then
I'm
curious.
Whether
people
on
the
call
like
this
breakup
of
source
authority
and
source
path,
which
I
think
is
a
clarifying
concept
but
Thomas
has
indicated
a
willingness
to
go
one
way
or
the
other.
So
I
think
that
it
might
be
good
to
hear
people's
opinions
on
that,
and
maybe
we
could
close
off
that
part.
A
I
was
actually
talking
on
mute
earlier,
so
Thomas
I
was
gonna.
Ask
you
what
how
do
you
want
to
proceed
on
this?
Is
this
something
that
you'd
like
to
work
through
just
comments
in
the
PR
itself?
Do
you
want
to
try
to
set
up
an
offline
need?
Have
a
discussion
wait
till
next
week's
phone
call
which
I'd
rather
not
do
but
I
would.
E
E
C
E
A
L
N
A
A
J
We've
only
got
a
few
minutes
left
I'm
wondering
if
we
could
just
take
a
couple
minutes
just
calibrate
on
that
on
a
possible
schedule
because
which
our
next
meeting
is
gonna,
be
in
the
month
of
April,
and
we
have
cloud
native
Khan
happening
at
the
beginning
of
May
and
we'd
love
to
announce
this
and
be
so
great
to
announce
this
there
and
we
have
to
make
sure
that
we're
moving
as
fast
as
possible
to
get.
You
know,
get
this
specification,
the
MVP
of
had
finished
and
then
getting
some
examples
and
reference
architectures
built
around
it.
J
A
So
Austin,
how
would
you
recommend
this
way
to
sort
of
I
guess
defined?
What
MVP
is
for
the
coop
gun
event?
Should
we
because
we
already
have
the
milestone
stuff
I
suspect?
That's
not
fine-grained
enough.
For
that
definition?
No,
it
is
what
we
may
be
a
wiki
set
up
for
people
to
define
what
their
requirements
are
for.
Mvp
I
thought.
A
M
A
A
Right,
so
what
if
I
move?
So
not
a
note
to
the
analyst
to
tell
people
that
on
next
week's
call
we're
going
to
try
to
narrow
down
what
the
MVP
items
are
and
so
get
your
issues
into
github
or
PRS
and
to
get
up
one
of
the
two
and
we'll
figure
out
which
things
get
the
v1
I'm.
Sorry,
these
zero
point,
one
label.