►
From YouTube: 20200811 SIG Arch Conformance
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Welcome
to
the
conformance
sub
project
meeting
of
cig
architecture,
depending
where
you
are
in
the
world,
it
could
be
wednesday,
the
12th
of
august
or
tuesday,
the
11th
of
august,
and
we
have
a
full
agenda
and
we
are
getting
recorded.
So
please
be
follow.
Our
code
of
conduct,
which
be
kind,
would
be
a
good
summary
of
that.
We
have
on
our
discussion
today
a
couple
of
points
I
put
a
few
at
the
top
just
in
case
the
third
one
goes
really
long
and
the
rest.
A
If
we
don't
get
to
them,
I'm
okay,
I'm
pushing
them
off
to
the
next
time
we
meet
the
first
one
is
improving
vendor
cncf
member
participation
and
improving
conformance
coverage.
Somebody
added
that,
but
they
didn't
add
who
that
was
so.
If
they
are
here,
I
would
love
to
yield
the
floor.
B
I
don't
know
if
it's
entirely
appropriate
for
us
to
have
this
discussion.
This
was
sort
of
in
advance
of
suggesting
that
it's
an
issue
the
steering
committee
discussed
with
priyanka,
but
for
all
the
folks
who
are
here
there
is
concern
amongst
those
of
us
on
the
steering
committee
and
those
of
us
in
sig
architecture
that
the
number
of
vendors
contributing
to
improving
conformance
test
coverage
is
one
and
we
feel
like
compared
to
the
number
of
vendors
that
have
their
offerings
certified,
as
you
know,
compliant
or
conformant
for
kubernetes.
B
That's
a
bit
of
a
a
mismatch,
and
so
you
know
I
could
see
it
as
if
I
were
somebody
who
wasn't
the
one
vendor
working
on
implementing
conformance
coverage.
I
could
be
concerned
that
you
know
coverage
might
head
in
a
direction
that
excludes
my
my
my
offering,
and
it
might
be
of
some
benefit
to
me
to
participate
in
you
know
hammering
out
coverage
upstream
so
that
we
can
sort
of
help
identify
those
those
issues
sooner,
but
I
think
you
know
priyanka
was
interested
in
having
us
kind
of
brainstorm.
B
You
know
how
how
do
those
of
us
who
benefit
from
this
program?
You
know
how
can
we
align
incentives
such
that
the
you
know,
there's
there's
more
people
contributing
here
than
a
single
vendor
and
an
iii,
and
this
is
in
no
way
meant
to
disparage
the
work.
That
ii
is
doing,
in
fact,
really
if
hibby
and
his
team
were
not
contributing
to
conformis
test
coverage,
we'd
be
significantly
further
behind
than
we
are
today.
They've
done
tremendous
work
for
us,
which
has
been
great,
and
I
don't
know-
maybe
you
could
argue
like.
Oh
well.
B
Vendors
are
pitching
in
that
way,
because
some
of
their
dues
that
they're
contributing
to
the
cncf
are
helping
pay
this
team
of
contractors
to
to
work
on
coverage
and,
like
that,
that's
understandable,
but
I
still
feel
like
we
could.
We
could
have
more
people
writing
more
tests,
so
something
to
think
about.
I
don't
know
how
much
time
we
want
to
spend
talking
about
this.
C
Earlier
we
used
to
have
a
monthly
cncf
run
conformance
meeting
too
right,
and
that
was
a
non-coding
side
of
things
where
they
used
to
discuss
or
doesn't
look
like.
We
are
having
those
anymore.
It's
probably
a
good
platform
to
attract
vendors.
B
B
So
it's
it's
not
entirely
clear
to
me
how
that's
gonna
help
drive
that
forward,
but,
like
I
do
I
could
see
you
know
if
we
decide,
we
really
need
to
dig
into
profiles
in
depth
and
we
feel
like
we're
not
getting
the
correct
audience
here
that
maybe
we
look
at
discussing
it
there,
but
you
know
this
really
is
like
if
you
care
about
how
you
know
the
cncf
sort
of
in
that,
in
that
meeting,
we
decided
that
it
was
up
to
the
kubernetes
project
to
define
what
conformance
is
like
what
it
means.
B
B
But
again
I
realize
I'm
kind
of
talking
all
all.
Stick,
no,
no
carrot
here,
I'm
not
really
great
at
brainstorming
and
being
super
creative
about
this.
So.
A
Thank
you
for
that.
I'm
aaron!
I
want
to
give
just
a
moment
for
anybody
that
looks
at
submit
via
the
chat
or
to
speak
up
real
quick.
We
did
have
that
conformance
working
group
meeting
and
there
would
be
a
lot
of
discussion,
but
I
don't
know
that
we
came
to
a
lot
of
actionable
processes
once
we
defined,
we
just
had
a
severe
lack
in
coverage
and,
most
of
our
time,
it
felt
best
to
spend
on
actually
just
getting
those
tests
written
and
improving
the
coverage.
A
As
these
discussions,
I
think,
as
aaron
was
noting,
are
leaning
more
towards
what
does
it
look
like
to
to
define
conformance
we've?
Is
this
the
space
for
that
in
this
conformance
kubernetes
beauty,
because
we
don't
we
haven't?
A
We
usually
use
this
for
for
pr
reviews
and
for
helping
to
define
the
tests
that
we
need
to
write.
We
often
have
haven't
needed
to
have
a
lot
of
space
for
defining
our
our
profiles.
We've
gone
through
a
couple
different
iterations
of
that,
and
I
want
to
make
sure
that
sorry.
This
is
we're
kind
of
confusing
topics:
the
the
cncf
member
participation.
C
B
Good,
I
think
walter
walter
made
a
great
point
in
chat
that
pushing
this
via
six
is
probably
one
approach
to
take,
so
those
vendors
who
participate
in
six
would
be
able
to
contribute
that
way.
Yeah
I
I
could
see
us.
I
could
definitely
see
us
heading
in
that
direction.
I
think
you
know
hippies
team
has
sort
of
been
really
solidifying
their
tooling,
and
so
we
could
get
to
the
point
where
we
use
that
to
provide
guidance.
B
I
also
feel
like
the
usual,
so
so
we've
done
a
good
job
of
making
sure
that
we
carry
that.
We
don't
like
make
the
situation
worse
going
forward.
Anybody
who
needs
to
who
wants
to
promote
their
feature
to
ga
has
to
have
conformance
tests
for
it.
B
So
I
think
we've
done
good
in
that
regard,
as
some
of
hippies
recording
can
show,
but
as
far
as
retroactive
coverage,
that's
a
really
good
question,
and
I
know
that
there
are
a
number
of
folks
within
the
project
who
are
kind
of
talking
about
maybe
really
focusing
on
our
testing
for
the
next
couple
of
months
as
we
look
to
improve
just
the
overall
reliability
and
stability
and
coverage
of
our
testing.
B
So
it
may
be
that
you
know
we
decide
to
improve
the
reliability
of
kubernetes,
we're
going
to
say
that
sigs
make
improved
performance,
testing
coverage
or
priority.
B
There's
been
discussion
of
things
like
not
allowing
any
new
feature
work
to
land
until
we
feel
like
sigs
have
done
their
due
diligence,
which
again
sounds
like
I'll.
Stick,
no
carrot,
but
I
think
mobilizes
potentially
more
of
the
project's
contributors.
So
it's
a
really
good
point.
Walter.
D
Aaron,
I
I
have
a
question.
This
is
vladimir
yeah,
so
based
on
what
you
just
described,
which
is
other
efforts
going
on
to
make
sure
that
cigs
define
whatever
their
compliance
is
prior
to
having
some
feature
being
moving
forward.
What
else
do
you
see
is
needed
to
to
to
enlarge
the
the
the
volume
of
of
test
coverage.
B
I
think
what
I
what
I
threw
out.
There
was
just
sort
of
an
idea
off
the
top
of
my
head
and
we
still
sort
of
need
more
discussion
and
consensus
around
the
cats.
Like
that's
the
direction
we
would
want
to
have
ahead
and
then
like
how
we
would
how
we
would
measure
that
work
and
dv
divvy
it
up.
B
I
don't
know
if
that
really
answers
your
question,
though,
what
what
more?
Beyond
that
I
mean
the
the
the
ideal
outcome
is
the
number
of
tests
that
are
written
and
then
ultimately
promoted
to
conformance
that
come
from
authors
other
than
the
ii
team
grows,
like
that's
the
that's
the
successful
metric.
D
B
Yeah
I
mean
we've
definitely
had
this.
I
I
would
say
a
lot
of
the
discussion
here
from
the
strategic
level
has
centered
around
like.
So
what
is
the
most
important
thing
to
cover?
First,
given
the
limited
set
of
resources
that
we
have
available
to
us-
and
I
feel
like
we've
done
a
pretty
good
job
of
that,
but
it
would
be
cool
if
more
resources
were
available,
so
we
could
accomplish
more,
rather
than
continuing
to
operate
in
an
extremely
constrained
mode.
B
Not
to
my
knowledge,
that
also
sounds
like
a
good
idea.
B
It's
sort
of
the
situation
where
I
feel
like
our
end-to-end
tests
are
not
the
most
user-friendly
thing
to
experience.
I
know
again,
he
and
his
team
have
been
making
a
lot
of
progress
on
building
up
knowledge
and
how
to
educate
others
in
that
process,
and
so
that
could
be
something
that
we
leverage.
E
Part
of
my
thought
is
also
that
if
we
have
workload
authors
who
want
their
workloads
to
work
anywhere,
be
it
you
know,
operator,
like
operator,
you
know
tools
or
anything
else
where
they're
trying
to
to
sell
their
workload
or
just
generally
have
it
work
anywhere
yeah.
Then
they
have
a
lot
to
gain
by
trying
to
make
sure
the
conformance
tests
cover
all
of
the
api
surface
that
they
use.
A
Agree:
one
of
the
approaches
we've
used
in
helping
to
make
sure
that
our
mock
test
or
the
attempts
before
we
write
a
test
that
make
our
coverage.
Is
we
deploy
api
snoop
as
a
audit,
sync
and
inject
that
into
a
database,
to
allow
us
to
see
what
our
ede
binary
is
hitting,
but
it
could
just
as
easily
be
used
on
an
existing
cluster
to
see
what
a
particular
workload
is
hitting.
A
So
for
this
isolated
workload,
when
you
select
all
of
the
components
part
of
that
from
a
very
discrete
level,
not
only
which
apis
you're
hitting
but
the
specific
way
you're
hitting
it
from
an
from
an
analysis,
point
of
view,
rather
than
kind
of
going
through
the
code.
A
D
B
It
may
be
possible
yeah,
I
feel
like.
If
there
was,
I
mean
what
it's
the
law
of
open
source
right,
the
less
friction
there
is
the
more
likely
you're
you
are
to
pick
up
contribution
right.
So
I
would
I'd
say:
there's
there's
a
decent
amount
of
friction
involved
in
writing.
End-To-End
tests,
I've
been
tangentially
involved
in
the
testing
commons
sub
project
of
testing,
to
try
and
either
like
break
up
the
e2e
framework,
or
maybe
reorganize
it
or
maybe
look
at
using
something
else
entirely.
B
I
feel
like
there
are
some
folks
from
vmware
who
are
helping
out
there,
but
it
also
seems
to
be
kind
of
a
bandwidth
constraint
thing.
So
I
think
like
if
somebody
had
an
end,
a
testing
framework
that
they
really
liked
it
could
make
it
could
make
people's
lives
a
lot
better.
B
B
Priyank
is
not
around
so
thanks.
I
appreciate
the
time.
A
Thanks
for
your
time
on
that
everyone,
I
wanted
to
put
this
topic
at
the
top
real,
quick,
because
it'll
be
really
quick
and
I
can
hopefully
get
it
merged.
A
We're
trying
to
reduce
our
denominator
for
the
number
of
things
we
need
to
hit
and
21
of
our
open
api
endpoints
in
the
proxy
area
will
be
absolutely
unable
to
be
reached,
and
this
is
because,
within
the
code
there's
a
301
move
per
minute.
Reader
direct-
and
you
can
see
here
via
our
curl
when
we
go
in
and
hit
anything
for
a
pod
space
proxy
without
any
flashes,
it
just
gets
a
permanent
redirect
and
that
is
located
in
the
code.
I
think
there's
actually
the
link
somewhere
to
that.
So
we
were
going
to.
A
The
links
at
the
very
top
thanks
for
that,
so
there's
there's
the
302
redirect
so
just
a
quick
look
at
the
underlying
code.
If
it's
missing
any
part
of
the
query
path,
it
just
goes
ahead
and
pins
that
slash
and
puts
the
query
part
on
there.
So
you
can't
hit
those
end
points.
A
Any
questions
or
discussion
on
that
I'm
confused.
Why
can't
people
test
them
like?
Maybe
I
just
don't
know.
I
think
I
miss
totally
this
code
right
here
I'll
make
it
real
big
on
incoming
http
request.
If
there
is
it's,
there's
a
link
to
it
like
a
query
part,
it
will
rewrite
the
header
and
add
the
slash
and
then
give
a
302
redirect
for
where
it
wants
you
to
go.
A
G
A
F
So
there's
with
sorry
proxy
with
path
is
what
it's,
the
api
endpoint
that
it
gets
redirected
to
and
that's
what
is
actually
getting
used.
Okay,.
G
Proxy
without
path
is
not
possible.
Okay,
sorry
yeah
that
that's
totally
reasonable
to
say
due
to
the
design
of
the
system.
You
do
not
have
to
implement
the
operation
proxy
without
path,
because
it
is
impossible
to
invoke
now.
I
guess
we
could
say
we
do
require
you
to
do
a
redirect,
which
seems
reasonable,
because
that's
what
a
cube
api
server
does
today.
So
aaron
I
see
aaron
hodding,
so
I
would
be
totally
fine
saying
the
the
conformance
definition
of
that
operation
is.
It
will
give
you
a
move
permanently
to
slash.
G
Although
it's
funny,
because
now
that
you
mentioned
that
I'm
like
terrified-
because
I
I
vaguely
remember
this
issue-
and
I
like
oh-
I
was
gonna
come
back
to
this
like
five
years
ago
and
then
forgot
so
now.
This
is
bubbled
back
up
and
I'm
gonna
go
see
something
horrifying.
That's
gonna!
Make
me
sad.
Thank
you.
It
sounds.
C
G
F
F
Issue
with
grafana
breaking
without
slash.
G
A
Yay
for
a
long
family
history
in
our
community
and
and
I
look
forward
to
ensuring
that
everybody
in
the
world
when
they
have
kubernetes,
is
we'll.
Do
a
301
redirect.
A
So
that's
that's
that
one
and
then
I
think
the
this
is
one
that's
going
to
get
a
lot
of
discussion
here,
so
I'm
going
to
go
ahead
and
close
all
my
other
windows.
So
it's
just
down
to
here.
This
is
our
our
thoughts
on
conformance
on
our
last
meeting
bella
mark
john
had
said:
I'm
going
to
put
together
a
document
thanks
clayton,
for
that
I
make
one
mistake:
support
it
forever.
A
A
So
we
have
this
profiles
kept
that
we
got
through
in
the
provisional
status
and
I'm
sure
if
I
should
try
to
walk
through
this
together
or
what's
the
best
way
to
generate
discussion,
because
I
think
we
we
any
conversation
when
I
have
on
the
purpose
of
profiles
as
we
go
through
these
three
paragraphs.
H
H
If
I'm
wrong,
hey,
we've
got
a
lot
of
areas
where
we
got
to
build
up
just
tests
that
we
don't
need
to
put
in
a
conformance
profile
that
would
just
go
into
the
base
and
until
we
get
the
base
really
flushed
out
we're
probably
okay,
deferring
the
discussion
on
profiles.
I
mean
that
was
my
recollection
erin.
Let
me
know
if
yours
was
different,
but
that's
what
I
remember
hearing
you
say
a
lot
and
I
think
I
think
the
interesting
thing
was.
H
You
know
waiting
for
the
point,
a
hippie
hacker
for
when
somebody
shows
up
with
a
test
or
a
group
of
tests,
and
they
for
some
reason
feel
like
we
can't
put
it
in
the
base
right
so
have
we
had
the
person
show
up.
That
said,
I've
got
a
group
of
tests
and
half
the
vendors
are
freaking
out
about
my
tests.
H
We've
done
the
diligence
and
I
can't
get
to
say
above
50
of
the
of
the
distros
wanting
to
put
this
in
the
base,
and,
oh,
my
goodness,
if
I
can't
get
you
know,
if
I'm
only
at
about
50
percent,
you
know
maybe
there's
the
cloud
provider
version.
Maybe
there's
the
the
lighter
version
of
kubernetes
and
we've
got
a
we've
got
a
partition
between
our
our
wonderful
community.
H
B
It
does,
I
feel,
like
I
see
william
unmuted,
but
I'll
just
say
a
few
words
since
you
mentioned
me
just
to
be
clear
about
my
position.
My
position
was
that
I
cared
about
core
conformance
coverage
and
I
didn't
particularly
I
mean
if
you
all
wanted
to
continue
to
discuss
profiles,
that's
fine,
but
that
didn't
seem
to
help
move
the
ball
forward
on
core
coverage,
and
so
I
think
we're
at
like.
I
did
just
make
a
plea
that
we
could
use
more
help,
making
corey
coverage.
B
Don't
get
me
wrong,
there's
more
to
do,
but
an
example
of
a
user
banging
on
my
door
could
probably
be
anybody
who
wants
to
run
a
stateful
workload,
can't
actually
verify
or
validate
that
their
stateful
workload
will
run
on
any
kubernetes
cluster
according
to
the
cncf,
because
we
have
no
way
of
testing
out
storage
in
a
binding
way
right.
So
I
think
storage
is
one
of
those
profiles.
That
makes
a
lot
of
sense
to
me,
but
I
think
you
know
there's
also
been
a
discus
again,
I
think,
sort
of
from
the
user
perspective.
B
I
I
don't
have
a
whole
lot
to
add
there
other
than
I
guess
you
know
it
yeah
it
does.
It
does
feel
like
a
lot,
has
matured
and
and
sort
of
changed
in
the
last
three
years
and
and,
as
you
said,
aaron
there
are
some
more
restricted
offerings
out
there.
So
maybe
there's
a
good
user
angle
to
to
say
you
know
to
be
able
to
test
a
workload
and
know
that
it's
going
to
run
everywhere,
for
example,
if
it
doesn't
need
certain
privilege,
access.
G
So
it's
kind
of
interesting
that
you
brought
that
up
aaron
because,
like
I
think
we
were
waiting
for
someone,
so
it's
kind
of
the
if
you
think
that
nobody
cares
that
you
exist,
try
missing
a
mortgage
payment.
G
I
think
that
I
I
have
to
say
that
there's
been
times
where
I've
been
tempted
to
say
we
should
put
something
in
as
a
test
of
the
emergency
broadcast
system
of
conformance
and
see
who
fails
on
one
of
these
dimensions,
just
from
the
perspective
of
like
as
an
experiment,
but
that
would
create
a
bunch
of
churn
that
we
might
do
more
productively.
G
G
We
were
always
waiting
for
something
else
to
show
up,
nothing
else
ever
did
and
we've
very
carefully
tried,
not
to
rock
the
boat.
I
think
these
are
the
two
most
reasonable
art
things
that
anyone
has
articulated
in
five
years.
That
is
a
meaningful
improvement
for
users
that
benefits
people
and
at
worst
we
will
be
like.
Okay,
like
we're
doing
this
with
this
intent,
oh,
we
suddenly
caused
a
whole
bunch
of
people
to
become
non-conforming,
let's
stop
and
take
their
feedback
and
iterate,
and
as
long
as
we
have
that,
I'm
not.
G
H
I
G
Open
shift
up
for
the
like,
we
can
take
out
the
thing
that
makes
the
test
process
root
on
production
clusters
in
order
to
be
able
to
run
that
test
and
be
a
great
step.
I
don't
I
don't.
I
haven't
heard
someone
articulate
another
one
yet,
except
possibly,
the
like
exec
is
not
allowed
on
a
lot
of
production
clusters
and
we
allow
that
to
be
disabled
and
exec
actually
is
one
of
the
most
useful
tools.
G
H
G
H
Your
storage
example,
what
would
be
interesting
is
if
there
was
a
proposal
to
move.
You
know
some
of
the
storage
tests
in
it
would
be
very
curious
to
see
what
is
the
percentage
of
of
of
the
distros
out
there
that
break
so?
Is
it
five
percent?
Is
it
twenty
percent?
Is
it
50,
because
that
helps
you
to
to
to
define
you
know
where
you're
at
you
know?
Is
it
well?
H
H
What
it
sounds
like
is:
maybe
there's
just
two
profiles:
there's
the
more
fully
functional
one
that
can
handle
storage
and
then
maybe
there's
and
when
you
call
that
you
know
a
fancy
name,
and
we
could
pick
the
fancier
name
that
that,
from
a
marketing
perspective,
whether
it's
cloud
provider,
level
or
enterprise
level
or
whatever,
is
the
appropriate
name.
That
is
very
inclusive
and
it'll,
be
very
curious
to
see
you
know
if
you
break
it
into
those
two
profiles,
the
one
with
storage
the
one
without
and
then
how?
H
What
is
the
the
percentage
of
the
number
of
distros
that
fit
into
it?
Is
it
80
20
can
80
handle,
storage
and
20
can
that'll
help.
You
get
a
good
feel,
I
think,
for
you
know
the
interoperability
and
the
confusion
that's
going
to
occur
by
now
having
two
profiles
instead
of
one.
It
helps
us
to
to
get
our
our
hands
around.
You
know
how
big
a
complexity
this
is
by
by
trying
to
embrace
those
those
new
storage
and
into
the
conformance
clayton.
So
one
option
would
be.
G
I
don't
think
it's
unreasonable,
potentially
for
us,
in
a
major
release
to
slim
down
the
scope
of
what
is
officially
conformance
to
a
subset
base,
keep
the
extra
ones
in
there
and
report
something
in
a
non-judgmental
way,
which
gives
people
a
time
and
a
process
and
then
set
a
timeline
for
feedback
based
on
that
over
a
couple
reasons.
G
For
instance,
if
we
added-
and
maybe
it's
like-
not
just
one
like-
we
could
do
like
something
more
specific
like
I,
you
know,
I
could
think
maybe
one
other
example
which
is
more
of
an
infrastructure
one
which
is,
I
know,
platforms,
don't
allow
people
to
create
namespaces,
because
if
you
can
create
namespaces,
you
can
bypass
any
specific
protections.
People
put
so
there's
a
there's
another
factor
there
of
we
don't
have
the
infrastructure
to
do
that
in
the
current
test
framework
and
people
just
grant
access
the
set
of
considerations.
G
If
we
had
those
options
and
we
reported
them,
we
could
pick
one
two,
three
four
smaller
groups,
simply
flag
them
and
say
these
are
being
considered
for
inclusion
in
a
more
advanced
profile.
Please
include
whether
you
pass
in
a
way
that
we
can
just
mechanically
gather
the
data
without
causing
extra
work
for
cncf
submitters.
H
And
it's
really
cool
that
you
mentioned
that
clayton,
because
that
process
you
talked
about
the
process
of
getting
people
to
level
up
and
figure
out.
What
the
issues
are,
I
think
that's
been.
Our
big
concern
is
that
process
has
always
been
the
piece.
That's
never
been
sort
of
written
down.
I
think,
if
you
go
look
at
john's
proposal,
I
do
not
think
that
process
isn't
there
and
I
think
that's
like
that's
like
the
linchpin.
H
That's
like
the
key
thing
that
that
makes
this
work,
because
that
goes
back
to
the
spirit
of
march
2017,
because
we
were
all
in
the
room
in
austin.
That's
exactly
what
we
were
doing.
Let's
go,
go
look
at
these
tests
and
go
run
them
and
let's
go
see
how
many
each
distro
passed.
So
I
think
you're
getting
us
back
to
the
spirit
clayton
of
what
we
wanted
to
do
many
years
ago
and
that's
really
really
cool.
G
But
with
aaron's
caveat
that
people
are
going
to
actually
need
to
do
something,
and
so
maybe
that's
actually
so
erin.
This
might
actually
be
a
good
opportunity
to
both
have
a
little
bit
of
a
carrot
and
a
little
bit
of
the
stick,
and
this
would
actually
be
a
fairly
significant
event
with
which
to
create
a
desire
to
engage.
G
B
Yeah,
I
agree,
I
mean
I
kind
of
whatever
I
always
figured
when
profiles
got
serious.
Vendors
would
come
back
to
the
table,
so
yeah.
B
That
that's
true,
too,
so
I
hear
your
point
brad
about
like
concern
over
fragmentation.
I
I
share
that
concern
deeply,
but
I
also
don't
necessarily
know
whether
or
not
we
should
be
beholden
to
the
the
momentum
of
what's
there.
I
I
definitely
so
so
for
this.
The
storage
example
is
certainly
more
complicated,
which
is
why
I
think
that
we,
when
we
were
talking
about
this
originally
we
were
talking
about
the
base
and
privileged
example.
It
seemed
like
it
was
pretty
much
a
no-brainer.
B
It
seemed
like
there
was
consensus
on
that
and,
like
the
reason
we're
here
talking
about
it
now
is
because
it
wasn't
quite
as
straightforward
as
we
anticipated,
and
so
I
think
like
rather
than
dig
into
the
logistics
of
exactly
how
the
crank
would
be
turned.
You
know
with
that
concern
of
fragmentation
in
mind,
it's
important
for
us
to
think
about
what
is
the
appropriate
shape
and
scope
of
profiles
so
that
they
do
make
the
most
sense,
and
so
we
don't
have
you
know
overly
stratified,
overly
fragmented
things
that
serve
nobody.
B
So
I
think
I
just
like
I
made
up
some
examples
off
the
top
of
my
head.
I
sort
of
trust
like
clayton
and
john's
judgment,
a
lot
more
when
it
comes
to
other
ways.
We
could
frame
this
or
think
about
this
further
down
in
the
document.
H
But
what's
interesting
is
like
clayton
said:
we've
had
this
topic
for
say
five
years
and
after
five
years
of
study
and
opportunity
for
investigation
he's
come
up
with
two,
so
you
know
the
the
concern
of
the
we're
going
to
end
up
with
12
profiles
and
everything's
going
to
be
a
crazy
nightmare.
Let's
call
that
you
know
the
brad
paranoia
maybe
is
unfounded
right,
maybe
that's
unfounded.
If
we're
really
after
five
years,
we're
kind
of
down
to
storage
and
base
versus
privilege,
and
as
someone
who
has.
G
G
Jordan
did
highlight
in
that
last
cig
arch,
some
of
the
areas
where
we're
still
weak,
like
we
had
a
good
effort
at
upgrade
like
we
used
the
upgrade
tests
and
we've
added
others,
but
like
some
of
them
are
very
specific
to
the
distro,
which
is
kind
of
like
a
weird
nuance,
but
like
it's
important
that
this
stuff
is
available
during
rolling
upgrade
like
the
fact
that
service
proxies
with
load,
balancers
and
local
endpoint
policy,
don't
actually
offer
graceful.
G
Draining
and
termination
is
something
that
we
should
have
caught
in
cube
a
long
time
ago,
because
it's
kind
of
embarrassing
that
you
can't
actually
run
an
aha
app
through
service
load,
balancers
and
use
the
service
traffic
local
policy.
To
get
like
these
are
things
that
we've
learned.
We
have
a
vehicle
to
put
the
tests
in,
but
we
don't
have
any
vehicle
to
drive
it
out
and
if
we
can
come
up
with
a
process
that
works
on
the
first
profile.
G
Maybe
profiles
isn't
the
right,
maybe
they're
like
little
bumps
on
the
on
the
growing
conformance
suite
and
it's
just
a
bunch
of
check
boxes,
and
maybe
there's
maybe
it
serves
as
we
spend
most
of
our
time,
adding
check
boxes
that
people
are
not
passing
and
everybody
kind
of
takes
a
couple
releases
to
get
up
to
speed.
That
would
be
is
just
as
good
an
outcome
because
the
end
result
here
is:
we
want
people
to
to
have
the
capacity
to
change,
to
be
to
improve
conformance
to
improve
the
say
their
systems.
G
I
think
a
lot
of
people
don't
know
what
tests
they
aren't
passing
because
they
get
conformance
running
and
they
don't
get
anything
else
running
and
that
I
get
that's
a.
I
don't
want
that
to
seem
like
an
aversion.
There
is
a
lot
of
stuff
to
go.
Make
work
in
cube
that
regular
tests
don't
cover.
So
there's
also
that
factor
I
I
don't
think,
there's
anything
that
I
think
it's
the
right
time,
basically
to
start
adding
the
bumps
on
the
the
profile
and
start
growing
it
in
that
dimension.
H
Right
and
then
the
only
request
was
for
those
bumps
was
if,
by
chance
really
the
bump
could
be
something
that
feels
like
it
fits
in
the
base,
we'll
just
add
it
to
the
base,
don't
make
it
a
new
profile.
That
was
right
because
it
sounded
like
the
ones
that
you
just
described
were
ones
say
not
like
storage,
where
not
everybody
has
storage,
persistent
storage,
but
it
sounded
like
that.
The
one
you
described
was
just
a
a
level
up
one
that,
in
your
mind,
clayton.
You
thought.
H
G
I
think
and
that
plenty
of
the
things
I've
seen
like
there
are
things
that
you
know,
for
instance,
tim
in
the
early
days
of
cube
obsessed
very
deeply
over
things
with
service
proxies
and
behaviors.
Those
are
actually
inconsistent
between
cni
plug-ins,
I'm
not
saying
our
job
is
to
go
solve
cni
plug-ins,
but,
for
instance,
you
get
different
behavior
within
a
pod
if
you
curl
an
endpoint
and
it
has
zero
endpoints
versus
one
depending
on
what
your
cni
provider
is
and
the
guarantees
they've
offered
and
to
an
application.
G
G
H
I
I
think
that's
where
you
do
the
community
of
service,
because
my
fear
has
always
been
they
just
want
to.
Well,
let's
kick
the
tires
and
profiles.
Let's
go
make
it
a
profile,
but
but
if,
if
it's
sort
of
right
of
first
refusal
that
well,
let's
really
see
if
we
can
get
those
optional
check
boxes
on
a
road
map
to
get
into
the
base,
I
think
you
keep
the
the
cats
heard
it
a
little
bit
better.
G
A
We
recently
added
some
automation
to
the
cncf
kids
conformance
repo,
that
compares
the
list
of
tests
that
they
should
be
running
to
the
test
that
they
actually
do
run
and
give
feedback.
And
here
are
the
tests
that
you
were
missing
and
the
steps
that
you
weren't
following.
B
G
Yeah,
I
john
is
john
will
be
listening
to
this,
so
I'm
coming
for
you
and
know
that
we
we
love
you
and
that
we
we
want
to
give
feedback
and
do
we.
We
certainly
have
enough
to
say
there's
no
disagreement
within
this
group.
One
of
the
things
I
think
we
should
probably
do
is
do
what
we
can
to
get
that
wrinkle,
like
the
vendor.
Knee-Jerk
reaction,
I
think,
is
a
key
design
constraint
to.
G
We
should
explicitly
design
around
in
the
proposal
of
what
we
do
and
we
don't
want
to
create
a
mess
for
cncf
to
clean
up,
so
whatever
we
do
probably
should
have
that
assumption
baked
in,
and
I
think
you
know
it's
the
obvious
thing
to
do
so.
It's
not
novel
for
me
to
say
that
just
that.
That's
the
factor
that
to
mitigate
the
original
fear
of
profiles
is,
you
will
not
fail,
conformance
or
trademark
at
a
critical
boundary
in
the
first
release
or
the
second
release
or
within
whatever
window
we
think
is
reasonable.
B
Yeah
I
mean
to
be
clear:
the
the
original
proposal
for
vote
for
profiles
was
to
split
split,
the
existing
conformance
thing
into
based
on
privilege.
The
idea
would
be
that,
like
all
right,
let's
assume
that
that
actually
worked.
The
idea
is
like
anybody
who
passes
conformance
today
is
still
conforming
plus
a
privilege
profile.
B
So
we
didn't
view
it
as
like
having
any
negative
impact
retroactively
and
that's
why
it
seemed
to
make
more
sense,
starting
there
versus
an
additive
profile.
G
I
only
bring
that
up
just
in
the
sense
of
I
know
that
people
won't
run
the
additional
profiles
and
since
we're
lacking
the
data
about
whether
people
actually
are
going
to
pass,
fail
or
care
whatever
process,
we
pick
should
in
as
trivial
a
manner
as
possible,
because
not
everybody
uses
sona
buoy,
and
so
we
should.
We
should
just
come
up
with
a
reasonable
process
that
has
the
warm
fuzzy
for
everybody
else,
whether
it's
additive
or
not,.
B
Yeah
I
agree
I
I
will
go
back
and
revisit
the
conformance
profiles,
implementation
doc.
I
proposed
because
one
of
the
options
that
admittedly
didn't
seem
to
get
a
whole
lot
of
traction
was
like
yeah,
just
run
them
all
and
then,
like
figure
out
what
you
know,
what
profiles
you
fit
at
the
end
and
the
feedback
I
got
back,
it
was
sort
of
felt
like
that's,
really
not
intuitive
to
me
as
a
developer.
B
But
I
agree
there's
value
in
like
if
we
can
make
that
as
seamless
process
as
possible.
I
think
it
would
be
useful
to
to
have
that
extra
data
to
know
that,
like
you
know,.
G
G
It's
an
amalgamation
of
stuff,
a
place
where
a
vendor
might
be
able
to
create
value
is
looking
at
more
coherent
ways
of
us
representing
this
test
or
the
suite
of
tests.
That
could
be
sonaboo.
I
think
that's
fine,
but
like
whatever
mechanism,
that
is,
if
that
is
really
going
to
be,
the
official
way
really
does
need
to
move
under
the
responsibility.
G
Some
of
that
could
be
something
a
vendor
could
pick
up
or
help
with
I'm
just
worried,
because
every
like,
we
have
continually
turned
the
crank
on
complexity
of
the
ginkgo
stack,
and
I
I
didn't
want
to
get
into
it.
I
just
I
was
thinking
about
like
all
of
the
ways
that
all
of
the
things
we
do
are
going
to
add
complexity
and
make
it
more
error,
prone
yeah.
B
Yeah,
I
I
I
hear
everything
I've
been
down
that
road,
I'm
trying
not
to
necessarily
because
I
feel
like
there's
sort
of
the
semantic
question
of
profiles,
that's
independent
from
the
technical
implementation.
But
if
there
were
some
things
we
could
simplify
about
the
technical
implementation.
B
G
It
I
made
the
mistake
of
caring
too
much
about
this
and
we
have
like
a
whole
shim
around
it.
That
offers
a
nice
experience
and
I
still
regret
it.
So
I
I
don't
want
to
advocate
for
that,
but
yeah
some
sort
of
movement
around
like
we
don't
have
formal
sweet
definitions
like
what
can
we?
What
can
we
say?
Is
requirements
or
necks
to
haves
that
could
give
place
vendors
room
to
participate,
for
instance?
Yes,
so.
B
Maybe
this
goes
back
to
vlad's
question
earlier
about,
like
the
testing
commons
sub
project
and
making
it
easier
to
write
in
tests
in
the
first
place.
I
just
I.
B
I
appreciate
what
a
large
effort
that
could
end
up
being
making
sure
that
we
can
sort
of
bridge
between
the
old
and
the
new,
and
so
I'd
like
to
make
sure
that
I
you
know
I
anticipate
this
is
the
this
is
a
continuing
series
of
discussions
around
the
semantics
of
profiles,
and
I
I
think
it's
wise
that
we
continue
to
have
those
discussions.
B
Independent
of
of
the
you
know
technically
how
it's
all
implemented,
but
these
can
certainly
help
shape
the
requirements
of
what
we
want.
The
technical
implementation
to
accomplish,
and
the
other
piece
I
was
just
going
to
toss
in
there
is
that,
like
do
you
think
team
have
been
working
to
help
kind
of
automate
away.
Some
of
the
the
checking
of
things
on
behalf
of
the
cncf
so
like
we're
not
just
looking
to
see
like
was
the
number
of
tests
that
passed
273,
but
actually
like.
B
Was
it
this
exact
set
of
tests
with
all
of
these
test
names
that
passed,
and
we
could
probably
then
maybe
loosen
our
constraints
on
what
tests
are?
You
are
not
allowed
to
skip
to
help
with
the
implementation
of
optional
profiles
and
stuff
like
that,
so
I
feel
like
we
have
the
pieces
in
place
to
step
forward.
A
I
just
dropped
a
link
to
our
cncf
infra,
pro
config,
which
is
the
proud
cncf
that
I
o
set
up.
That
goes
against
kate's
cncf
kate's
conformance
eventually,
and
this
is
where
we
could
add
some
extra
logic
around
the
profile
and
getting
feedback
in
those
pr's,
so
that
it's
not
just
a
manual
process.
A
I
know
we
have
five
minutes
left
and
I
don't
want
to
go
through
these
fully
on
these
last
two
points,
if
possible,
but
I
just
wanted
to
put
them
out
there
for
review.
This
is
some
of
our
work
on
the
kk
gate.
There's
been
some
interest
in
that
over
time
so
that,
if
there's
any
changes
to
our
open
api
spec,
this
is
a
a
job
that
would
trigger
and
identifying
the
list
of
new
or
removed,
endpoints
and
waiting
for
the
conformance
job.
A
To
finish
so,
we
can
see
those
lists
of
hit
and
conformant
endpoints
so
that
we
can
note
a
increase
or
decrease
in
endpoints
and
if
there's
a
decrease
in
it
being
sure
to
comment
on
the
pr
and
possibly
add
a
label
to
not
do
that
this
document,
I
can
also
drop
it
in
the
case.
So
the
the
kids
conformance
select
channel
for
review
any
questions
over
that
before
I
go
to
our
next
spot.
A
Okay-
and
the
last
part,
is
just
some
update
issues.
We
I'm
not
going
to
go
through
them,
I'm
just
going
to
open
this
one
link
and
drop
it
in
here.
This
is
our
our
list
of
q3
enhancements
on
that
card.
A
Am
I
dropping
the
link,
but
it's
color
improvements,
legend
update,
stateful
url,
so
that
we
can
use
those
as
a
as
part
of
the
landing
pages,
because
part
of
this
kk
gate
is
to
have
a
a
landing
page.
So
here's
the
steps
you
need
to
follow.
A
That's
just
a
quick
feedback.
At
the
end
anything
else
anybody
wants
to
cover
for
our
last
four
minutes.
B
A
Let's
take
a
look
it's
good.
This
is
the
conformance
progress
page.
B
A
This
green
stuff-
here
you
can
see
when
we
started
the
conformance
project.
We
got
41
endpoints
that
had
not
been
covered
before
and
we
also
had
11
new
endpoints
that
came
with
conformance
coverage.
So
this
two
colors
here
is
the
important
stuff.
The
green
stuff
is
stuff
we've
done
before.
A
So
all
of
these
these
were
old
endpoints
that
were
covered.
So
this
is
the
color
we're
looking
for
since
115.
This
is
around
the
time
when,
when
we
started
writing
tests,
I
was
writing
tests.
We
have
seven
initially
for
that
release.
What
was
that
one?
That's
13.
I'll
make
this
bigger.
Since
we're
sharing
the
screen,
there's
13.,
we
actually
had
27
in
117.
A
We
didn't
realize
that
until
we
went
through
and
updated
our
con
the
metadata
thanks
for
updating
that
stephen,
we
had
10
and
118,
and
this
has
been
our
biggest
release-
30
actually
39
in
points
because
there's
an
endpoint,
that's
not
showing
up
for
some
reason,
we're
tracking
it
down,
and
we
had
a
goal
of
40
for
119
by
the
way,
so
I
think
we're
doing
pretty
good.
Also.
We
have
41
endpoints
that
came
with
their
releases.
A
If
you
click
on
these
these
links,
we
now
have
the
ability
to
list
those
endpoints
you
can
see.
This
is
all
going
to
have
119
and
119
as
promoted
with
tests.
If
I
click
on
this
area,
these
are
old
endpoints,
you
can
see,
we've
got
endpoints
that
have
not
been
tested
since
one
well
prior
to
one
five.
I
didn't
go
back
any
further
than
that,
but
this
is
the
where
those
38
input
points
come
from.
A
We
still
have
this
777
untested
so
pick
one
of
these,
and
I
think
who
did
the
we
have
a
link
for
all
of
our
the
unavailable,
endpoints
and
I'll
share
that
there
should
be
a
ui
some
here
where
shortly
where
we
see
these
are
the
endpoints,
we
will
no
longer
that
are
not
eligible
for
informants
and
that
will
reduce
this
number
up
here
from
440.
Well,
whatever
it's
not
440,
it's
probably
around
415
right
now
for
stable
all.
H
A
A
What
did
we
have
so
there
was,
there
was
a
113
had
one
endpoint
and
then
there
wasn't
any
any
activity
there,
but
I
think
it's
because
of
how
we've
kind
of
changed,
how
we
measure
this
over
time
and
that's
because
we
learned
you
know
nearly
every
release
more
about
how
to
to
grow
things
than
what
we
focused
on.
You
can
tell
like
this
114
was
where
all
the
effort
went
into
to
kind
of
start,
that
forward
progress.
A
So
aaron
don't
give
yourself
too
bad
of
time,
you're
right
there
and
okay,
let's
put
this
up
into
the
ride.
Now,
hey
look!
All
I
care
about
is
I
got
this
cat
t-shirt
out
of
it,
so
you
did
get
the
cut,
love
the
shirts,
my
friend
great
work,
everybody,
and
I
will
give
you
back
your
time-
have
a
wonderful
day.