►
From YouTube: 20190423 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
A
B
B
We
are
debugging
that
there's
a
recursive
loop
in
there
causing
it
not
work
and
the
other
thing
I
realized
was
the
functionality
itself
is
pretty
minimal
in
this
er
that
I
ported
we
need,
probably
more
than
that
like
when
an
issue
is
opened
against
theory
of
conformance.
We
need
to
bring
that
in
the
transport,
not
just
the
pr's
and
wild
talk.
While
we
were
discussing
about
the
project
about
the
PR
lifecycle
itself,
it
mandated
to
have
some
kind
of
standard
around
what
columns
and
what
states
that
the
ours
can
have
like
you
know.
A
Think
the
the
I
don't
really
care
about
some
of
the
other
columns
that
can
be
done
manually.
I
think
the
inbound
triage
is
the
most
important
so
that
it's
not
lost
in
the
noise
and
that's
the
one
where
it
requires
human
intervention
like
because
of
reviewer,
is
actually
reviewing
these
things
anyways
along
the
lifecycle.
They
it
doesn't
matter
if
the
automation
doesn't
automatically
promote
to
the
other
columns.
I
think
the
only
thing
that
matters
is
that
is
that
it
gains
visibility
to
this
group
as
soon
as
the
issues
created.
B
A
So
if
you
get
that
feature
in
that,
that
I
think
was
the
the
missing
gap
for
automation.
Once
we
have
that,
then
I
think
it's
really
easy
for
us
to
run
through
the
project
forward
every
single
time
and
as
we
assign
people
and
do
things
that
person,
whoever
is
the
assignee,
can
give
it
update
and
we
can
move
the
cards
along
the
Reds,
the
issues
and
PRS
along
the
board.
C
A
D
So
I
was
gonna.
Just
add
that
to
the
list
of
things
to
talk
about
the
agenda,
so
I
was
wrong
in
the
number
it's
closer
in
the
40s,
but
it's
around
seven
point
something
gigs
but
doing
the
investigation.
There
were
a
few
things
I
found
there.
There
was
already
some
I
mean
there's
some
images
that
are
getting
reused,
but
only
two
or
three
times
it's
kind
of
unclear.
What
functionality
they're
really
trying
to
do
so
I
I,
do
think,
there's
some
centralization
we
could
do,
but
out
of
the
seven
and
a
half
gigs
of
images.
D
Two
of
them
are
the
vector,
math
stuff
for
doing
like
graphics,
work,
and
you
know
I,
don't
use
that
test
image.
One
of
them
is
I.
Think
both
of
them
are
around
two
gigs
each
and
there's
actually
zero
tests.
That
use
like
the
v2
of
the
image.
But
whoever
is
working
with
those
images
says
like
now:
we
can
phase
towards
using
means,
so
I,
don't
think
the
answer
is
necessarily
like
getting
rid
of
one
of
those
right
because
Soviet
five
gigs,
but
probably
the
solution-
is
gonna.
Look
more
like.
D
Splitting
up
the
list
of
images
that
are
there,
it's
good
that
it's
centralized
in
a
single
place
in
a
single
file,
but
maybe
we
do
need
to
still
look
at
how
we
can
break
that
out,
so
that
it's
in
more
useful
segments
so
like,
if
you're,
really
just
wanting
to
run
conformance
tests.
These
are
the
images
you
need
and
then
we
can
focus
on
maybe
having
you
that
as
small
as
possible.
D
It's
a
little
bit
difficult
to
try
and
it
can
be
either
a
manual
process
just
what
images
are
getting
used
in
specifically
the
conformance
test.
What
I
did
was
I
made
a
custom
build
of
the
tests,
and
every
time
you
called
the
functions
to
get
the
image
name
from
that
centralized
location.
It
threw
out
a
log
line
and
then
I
grabbed
the
logs,
and
it
looks
like
there's
only
maybe
14
images
that
are
getting
used
in
conformance
tests,
which
seemed
pretty
reasonable.
D
E
A
A
Yeah
I
think
I
think
we
could
probably
deep
them.
I
honestly,
don't
see
why,
for
the
conformance
test,
I
honestly,
don't
see
why
there
isn't
just
one
to
be
honest,
just
tearing
one-one
test
to
rule
them
all.
I
do
know
that
some
of
the
tests
do
weird
things
were
like
pulls
down
in
genetics,
right
and
I.
Don't
I
think
he
gives
smash
that
into
a
single
container,
with
with
a
from
line
right.
D
A
A
Next,
up
on
the
list
was
I
put
this
here
when
I
was
going
through
stuff.
Is
that
there's
a
label
we
called
lifecycle
active?
We
use
this
sensical
stroller
cycle
to
denote
when
things
are
being
worked
on,
so
it
doesn't
really.
We
don't
really
need.
We
have
like
an
in
progress
in
an
in
review.
My
guess
is
that
if
we're,
we
don't
really
need
two
columns
life
cycle
active
would
note
the
label
life
cycle.
A
When
it
will
go
through
life
cycles,
stale
and
then
life,
and
then
life
cycle
rotten,
so
that
that
happens
over
180
days,
total
90
days,
the
first
one,
then
I
think
tiny
again
for
the
second
one.
So
if
we,
but
that
would
be
visible
in
the
project
board
too
as
well
and
our
project
for
is
pretty
small
by
most
other
boards
comparison
yeah.
G
I
guess
the
thing
I
was
getting
at
was
like
on
the
windows
board
in
progress
means
someone's
actually
working
on
it
and
then
there's
not
a
label
for
that.
Whereas
if
an
issue
is
just
active,
but
nobody
is
looking
at
it,
then
the
fact
that
it's
in
the
backlog,
rather
than
in
progress,
is
helpful.
Yeah.
A
H
A
The
the
stale
will
be
added
to
it
after
90
days.
That's
like
a
90
day
window
I
don't
want
to
create
another
label,
but
if
we
want
to
talk
with
test
infra
about,
you
know
changing
stale
to
be
one
month
cycle
and
then
rotten
to,
because
right
now
is
180
days.
That's
a
long
time
because
we're
giving
people
plenty
of
leeway
but
I
think
if
we
wanted
to
shorten
that
window,
I'd
be
supportive
of
that.
But
that's
that's
a
test
and
for
broadly
Australia.
A
I
I
I
There
are
still
two
images
left
that
are
missing
from
this
poor
request.
I
still
have
to
add
done,
but
that's
mostly
because
they
are
very
different
and
I
have
one
other
commit
which
adds
five
other
images
which
are
outside
of
the
communities
test,
images
folder,
but
they
are
still
using
performance
and
other
tests
that
are
necessary
to
be
run.
A
We
don't
we
don't
know
to
go
through
the
minutiae
of
all
of
it.
I
think
what
I
want
to
do
in
this
meeting
is
assign
a
couple
of
people
to
take
a
look
at
it.
I
know
that
DIMMs
has
had
a
lot
of
experience.
Working
with
the
Googlers
to
get
images
in
place
in
the
multi
arch
manifests
in
place
and
I
know
John's
working
on
the
image
of
deduplication,
so
I
think
they
seem
like
two
good
candidates
to
review.
Tim's
and
John.
Does
that
seem
fair.
E
E
E
I
E
A
A
There
was
two
items
that
were
followed
up
in
the
in
review:
I,
don't
see
Quinton
on
the
call,
but
he's
been
on
this
issue
for
a
very
long
time
and
we
haven't
actually
gotten
feedback
premier
dims
I
know:
I
saw
that
I
think
both
you
and
I
commented
on
this
one.
Do
we
want
to
just
take
this
one,
or
do
we
want
to
poke
Quinton
one
more
time,
I
think.
A
If
you
look
at
the
board,
two
of
the
three
are
actually
assigned
to
you.
The
one
that's
on
me
was
the
LG
team
was
cancelled
and
we
were
supposed
to
talk
about
it
here.
Aaron
had
requested
that
we
talk
about
here
was
the
ad
each
need
to
verify
Rinna's
none
options
in
pod
security
context
and
I'm,
trying
to
recall
the
details
of
why.
E
A
I
was
just
curious
like
where
he's
getting
his
priorities
from
because,
like
they've,
been
spending
a
lot
of
time,
sending
me
PRS
and
reviews
for
for
go--let
stuff
I'm,
like
you
know,
we've
got
these
project
boards
and,
and
we
have
a
way
of
a
breakdown
both
here
and
as
well
as
in
the
testing
common
stuff.
It'd
be
nice
to
kind
of
channel
his
efforts
towards
more
high-value
targets,
I
think
yeah.
E
A
A
A
I
think
I'm
gonna
have
to
go
through
this
one
independently
and
get
the
context.
Swapped
back
in
history,
says
example:
tests
which
are
not
eligible
for
abortion
in
performance,
anything
that
checks
optional
condition
fields.
So
this
has
been
sisters,
reasoner
message.
A
This
has
been
common
currently
across
a
number
of
different
functions
that
there's
a
bunch
of
tests
that
actually
leverage
conditions
and
I
even
talked
about
this
with
other
people,
and
there
are
problems
currently
with
with
a
bunch
of
the
tests.
Then
and
I
don't
see
this
as
a
problem
I
see.
This
is
people
had
highly
opinionated
views
of
the
universe
and
we
kind
of
wrote
them
down
sparsely,
but
we
did
not
I,
don't
think
we
have
broad
consensus
from
siga
architecture.
A
A
A
B
The
meeting
that
I
had
with
Logan
folks
this
week
yeah
they
are
kind
of
encroaching
into
those
corner
cases
where
they're
limited
in
number.
Oh,
they
don't
have
a
good
pipeline.
I
would
say
to
to
add
or
to
convert
the
existing
case.
They're
running
into
other
areas
like
jobs
and
service
accounts,
jobs
isn't
even
well
I.
Guess
that's
yeah!
It
is
very
indeterministic
too.
There
are
some
PRS
that
are
rejected,
because
there
are
some
open
issues
there.
B
B
C
B
A
B
B
A
Think
I
think
it's
incumbent
upon
them
to
given
that
they're
employees
of
the
CN
CF
working
on
this
particular
area
to
be
here.
I,
don't
think
it's
I!
You
know
if
we
need
to
move
the
time
of
this
meeting,
to
make
it
more
accommodating
for
them,
seeing
how
they're
doing
a
bunch
of
the
work
so
that
they're,
guided
in
the
correct
way,
I
I
think
having
them
disconnected
from
this
effort.
B
A
A
A
F
E
A
H
B
E
H
A
A
A
A
H
H
A
A
A
C
A
B
A
I
G
A
I
A
A
A
E
to
e
proposal
verify
TCP
socket
property
of
V,
one
probe,
along
with
success
or
failure
threshold
I'm
going
to
defer
to
the
networking
people
on
this
particular
one
I'm
going
to
assign
it
to
John,
because
he's
John
moved
from
triage
to
sort
of
backlog.
I
think
he
wants
it
to
be
worked
on
so
I'm
gonna,
say
he's
not
here.
So.
A
A
A
G
A
And
the
other
PRS
her
assignment
John
right,
we
had
a
bunch
of
stuff
in
progress.
We
have
a
reasonable
backlog,
John's,
not
here
so
he'll.
He
can
update
those
two
and
these
history
knee.
If
you
talk
with
the
globin
folks,
we're
totally
good
to
go
if
executing
on
some
of
that.
If
you
have
concrete
feedback
for
us
to
review
from
the
globe
and
folks
about
the
issues
that
they're
running
into,
we
can
always
add
that
to
the
agenda
and
discuss
it
here
too
or
not.
We
don't
have
time
to
talk.
B
J
J
A
A
Is
it
marked,
as
is
it
denoted
as
serial
right
now,
although
it
might
need
to
be
so
if,
if
people
haven't
reported
it,
which
I
have
not
seen
an
aggregated
conformance
test,
failure
in
the
wild
I've
never
seen
it
in
the
wild
come
through,
and
that
usually
could
be
that
your
local?
How
are
you
running
locally.
A
A
A
B
B
B
A
These
validation,
Suites
and
profiles
like
that
I,
don't
know
where
the
overlap
or
where
the
delineation
stops
right,
because
there
should
be
the
CNC
I
should
really
own.
What
is
a
good
storage
driver?
If
so
that
way,
you
can
have
uniformity
across
different
storage
providers
and
making
sure
that,
if
I
run
something
on
Azure
and
I
run,
something
on
AWS
I
have
some
little
guarantees
a
stamp
of
approval
for
certain
storage
characteristics
and
behavior
across
those
providers.
A
A
B
A
A
A
Create
intent
test
case
coverage
for
admit
pod
by
sorted
order.
That's
the
same
wondering
right.
D
It
was
just
a
philosophy
of
whether
that's
I
think
you
didn't
want
that
to
be
tested
because
it
wasn't
API
driven
and
then
other
people
I
think
chimed
on
the
ticket.
Saying
like
that
was
the
way
to
test
it.
I
don't
know
what
the
right
answer
is
there.
A
I'm
going
to
sign
this,
for
you
I
think
the
person
that
chimed
in
on
is
like.
Yes,
it's
a
good
test,
but
not
it's
not
a
performance
test,
I
think
because
the
I
don't
know
if
you
can
guarantee
that
SSH
pre-shave
you're
between
pods
across
all
their
providers,
especially
if
they
have
security
policies
and
whatnot.
A
B
A
Already
exists
inside
of
the
by
doing
basic
level
tests
that
exist
already,
like
writing,
and
reading
pods
writing
and
reading
services
and
there's
even
like
tests
inside
of
the
conformance
suite
already.
Now
you
can
beef
up
that
characteristics
and
behavior
of
the
storage
layer
and
I
think
that's
a
reasonable
thing
to
do,
because
the
code
inside
of
kubernetes
depends
highly
on
things
like
resource
version
right
in
guaranteeing
that
your
resource
versions
are
respected,
especially
in
like
the
multi
writer
scenario.
A
K
A
B
I
K
F
A
Well,
like
key
key
areas
that
are
worthwhile
investing
in
our
watch:
behavior
resource
version,
detection
from
multiple
writers,
it
one
thing
we
do
currently
if
they
have
an
H
a
deployment.
This
would
surface
that
multiple
writers
scenario,
if
it's
not
a
CP
storage
layer,
so,
like
you
know,
I
love
this
funsies.
If
somebody
is
actually
trying
to
make
a
conformance
validation
for
something
like
SQL
well,
that
I
still
like,
because
you
have
transactional
guarantees,
but
you
can't
you
can't
have
a
chaise
support,
then.
A
K
A
Gonna,
send
this
to
me
and
I'll
try
to
get
take
as
an
oak,
we're
actually
over
time.
So
what
it?
Why
don't
I
take
a
look
at
this
in
a
send
it
for
next
time
for
follow-up
I
can
give
you
a
better
list,
because
I
helped
write
this
torch
there.
I
can
give
you
a
better
list
next
time,
although
it's
been
two
years,
you
know
I'll
try
to
get
you
more
detailed
information,
but
the
the
title
itself
is
a
little
misleading
but
we're
out
of
time.
So
we
have
to
wait
till
next
time.
Srini.