►
From YouTube: Kubernetes SIG Service Catalog 2018-07-27 - Delete
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
It
seems
to
me
that
any
solution
we
come
up
with
really
can't
involve
adding
a
new
API
to
the
open
source
broker
API,
mainly
because
any
new
operation
we
add,
would
be
optional
at
best,
which
means
we
need
to
make
sure
we
work
even
when
that
operation
is
not
there
or
supported,
but
also
we
have
to
support
existing
brokers
because
they
won't
have
that
operation
there
and
and
no
matter
what
we
do.
You
know
with
any
kind
of
pre
check,
operation
or
anything
like
that.
A
A
The
broker
is
always
going
to
have
the
right
to
say
no
later
on,
and
we
have
to
be
able
to
deal
with
that,
and
so
in
my
mind
because
of
that
basic
fact,
it
guides
me
in
saying
you
know
adding
something
like
an
operation
or
some
extent.
Almost
changing
open
service
broker
in
some
way
is
almost
not
an
option
and
I
was
a
little
hesitant
to
say
not
changing,
open
source
burglary,
they
don't
get
necessarily
true.
We
may
be
able
make
some
changes
there.
A
I
just
can't
fathom
a
change
significant
enough
that
would
ever
not
allow
the
broker
saying
no
and
that's
that's
it.
That's
sort
of
like
the
first
big
undeniable
truth
in
my
head
that
I
just
cannot
get
past
making
in
my
thought
process
around
this.
The
the
other
thing
is
I,
don't
see
how
we
can
ever
allow
kubernetes
to
start
the
delete
process
on
a
resource
it
for
these
scenarios
right
because,
as
everybody
has
said
over
and
over
once
you
set
that
delete
timestamp,
it's
done
right
and
Wow
everybody
well!
A
Well,
some
people
have
suggested,
oh
well,
just
recreate
the
object
right
because
I
mean
that's
almost
a
very
inconsistent
position
to
have
because
they
keep
saying
you
can't
undo
a
delete
justifiably,
but
you
can't
undo
a
delete
because
people
are
watching
that
timestamp
and
may
react
to
it
right
and
so
the
notion,
that's
of
someone
saying
oh
well,
then
just
recreate
the
object,
because
it's
another
object.
You
know
with
a
new
delete,
timestamp,
it's
that's
cleared.
What's
the
big
deal?
A
Well,
the
actions
taken
by
that
first
object
being
deleted
because
of
that
delete,
timestamp
may
be
irreversible
and
not
things
you
want
to
have
happen
right.
So
when
we
create
this
secondary
object,
we
have
no
ability
or
no
guarantee
that
someone's
gonna
watch
that
and
say.
Oh,
oh,
okay,
well
I
delete
something
before,
but
I'm
gonna
recreate
it
because
this
object
was
not
recreated
right.
It
I
can't
fathom
that
we
can
ever
guarantee
that
happening.
A
So
that's
why
in
my
mind,
any
any
real
solution
to
come
up
with
us,
and
maybe
we
can't
find
a
real
one.
But
if
we
were
to,
if
we
were
to
claim
we
found
the
right
solution,
it
has
to
involve
not
sitting
delete
timestamp
on
the
kubernetes
resource
because
of
the
of
the
bubble
of
the
rippling
effect
of
that
delete,
timestamp
being
set
and
us
not
being
able
to
guarantee
that
we
can
roll
those
back.
A
So
those
are
kind
of
things
that
are
driving
me
in
all
these
decisions,
and
it's
it's
not
because
I'm
necessary
trying
to
be
you
know,
take
open
source
broker,
side
or
recurring
edit.
It's
more
I
can't
reconcile
these
basic
facts
in
my
head
or
I
can't
change
these
basic
assumption
to
my
head.
So
now
you
guys
gonna
know
if
you
think
I'm
way
off
on
this.
A
B
Know
I,
don't
so
on
the
I
say:
I
guess,
move
on
from
last
to
first
or
least
jump
in
right
now
in
the
last
I
I
tend
to
agree.
I,
don't
think
recreate
is
a
viable
scenario.
I,
don't
think,
that's
to
me
that
that's
not
that's
not
going
to
work
right,
I
wouldn't
even
want
to
attempt
it
Doug.
You
started
with
talking
about
backward
compatibility,
I
think
and
I
agree.
We
can't.
B
Obviously
the
the
webhook
stuff
requires
a
change
to
the
OSB
api
and
in
order
to
get
the
new
functionality
agreed
that
old
brokers
won't
pick
it
up.
You'd
have
to
have
a
broker.
That
was,
you
know.
We'd
first
have
to
get
the
extension
into
OS,
bapi,
saying:
okay,
pre-check,
here's
is
the
validating
pre-check
webhooks
and
the
brokers
would
have
to
be
revved
and
they'd
have
to
register
the
URL
for
it
and
that
kind
of
thing,
and
then
yeah
I
admit
there
is
there
are
cases
there
are.
There
are
definitely
scenarios
where
you
invoke
the.
B
Even
if
you
invoke
a
pre-check,
you
may
later
fail
when
you're
actually
doing
the
actual
deletion.
In
my
mind,
that
scenario
is
somewhat
limited
and
for
a
lot
of
the
part
can
be
worked
around
and
I
guess
I'm
willing
to
put
up
with
a
resource
that
gets
wedge
and
requires,
perhaps
administrative
assistance
to
deal
with
it
I
think
there
can
be
follow-on
work
that
would
help
with
wedged
problems.
B
A
A
How
does
that
reconcile
in
your
head
with
the
feeling,
with
the
fact
that
we
both
agree
that
you
probably
don't
want
to
allow
the
delete
timestamp
to
be
set
so
because
to
me
that
means
okay,
if
we
don't
want
a
lot
of
times
have
to
be
set,
but
you
want
to
allow
some
sort
of
admin
and
reconcile
thing
in
your
mind.
Is
the
reconcile
thing
involved
talking
to
the
broker,
to
reconcile
it
or
taking
some
action
on
the
kubernetes
side?
I'm,
not
quite
sure
where
the
action
be
taken
in
your
mind,
you.
B
Know
I
think
so,
I'm
still
pretty
much
along
the
lines
of
you
say,
delete
on
the
kubernetes
side
and
you're
done
with
it.
It's
it's
going
to
go
away,
one
way
or
another
and
the
admin
action
pretty
much
takes
care
of.
You
know
the
admin
is
responsible
for
ensuring
that
any
combination
resources
are
cleaned
up,
but
he
also
you
know
he
has
the
he's,
got
the
big
hammer
on
the
on
the
broker
side
and
he'll.
Do
whatever
is
necessary
to
clean
up
the
broker
side
or
I?
B
A
See
I
say
that
that's
something
that
I'm
really
struggling
with,
because
what
that
kind
of
implies
is
is
two
different
things.
One
is,
we
can
add
a
web,
we
can
add
a
pre-flight
check.
Type
of
operation,
which
is
still
very
wary
of
his
existing
brokers,
will
not
work,
but
also
you
you,
you
remove
the
ability
for
the
broker
to
say
no,
then,
because,
even
if
he
says
no
you're
gonna
say
well,
no
we're
gonna
go,
get
a
hammer
and
do
it
through
some
other
path
and
and
that's.
B
A
B
A
B
A
C
B
C
E
Had
discussed
commanding
one
and
eight
they're
kind
of
different
enough
that,
like
you
know
if
I'm
gonna
say
at
every
single
point
well,
we
can
do
this
or
we
can
do
that
or
we
can
do
this,
so
we
could
do
that
or
we
could
do
this,
so
we
could
like
they're,
not
really
the
same
anymore
after
okay,
after
sort
of
having
gone
through
them
at
least
having
gone
through
one
a
couple
of
times.
You
know:
we've
reached
a
point
where
it's
like:
it's
not
really
compatible
with
eight
in
many
ways.
E
Yesterday
and
I'm,
like
I
made
a
list
of
like
all
the
differences
and
I
was
like.
Well,
these
are
difference:
yeah,
yeah,
gotcha
yeah.
The
main
difference
is
like:
what
are
we
calling
the
flag
for
one
versus
eight,
like
sure
they're
the
same,
but
that's
not
really
we're
using
it
differently.
It's
now
a
counter.
Instead
of
a
flag,
it's
got
different
semantics
based
on
what
do
you
do
to
it?
So
they're
not
really
the
same
thing
anymore.
In
my
mind,
okay,.
B
A
E
A
See,
what's
interesting
to
me
is
the
from,
in
my
mind,
the
real
issue
of
to
isn't
the
the
lack
of
transparency
or
people
watching
things
they
shouldn't
be
doing.
It's
only
that
I
agree
with
you,
I,
don't
think
most
people
watch
resources
that
closely
and
because
we
own
these
resources,
we
could
be
very
blunt
and
say:
don't
watch
you
know,
don't
don't
expect
normal
things
to
happen.
Things
are
slightly
different
here,
but
in
the
end,
it's
all
the
same
too.
A
E
Haven't
researched
enough
to
know
what
they're
gonna
plan,
since
this
is
an
API
server
and
I'm
doing
edits
in
the
API
server
code,
I
don't
see
any
way
for
us
to
you
know.
The
API
is
remember
that
they,
the
CRD
server,
that's
provided
by
core
is
the
CRD
server.
That's
provided
by
core
there's!
No
there's
no
opportunity
for
me
to
make
tweaks
to
it.
So
it
is
it
is.
E
E
Believe
so
I
mean
it's
an
object.
He
goes
through
well,
I
mean
information.
Controller
has
to
be
compiled
in
a
web
hook,
intercepts
whatever
it
intercepts.
So
as
long
as
you
tell
the
thing
to
send
all
of
the
particular
kinds
of
objects
to
the
web
hook,
I
would
think
it
would
right.
I
think
you
can
register
those
are
actually
registers
at
runtime
cane
yeah.
That's
right!
So
it's
just
it's
just
it's!
It's
basically
string
matching.
E
A
That
sucks
I
have
you
broached
that
idea
with
the
API
server
guys
at
all
Morgan
I
know
the
call
I'm
warning
if
we
were
as
well
to
ask
them
or
ask
cool
people
about
that,
whether
it's
even
possible,
because
I
still
don't
understand
what
the
difference
is
between
a
mutating
web
hook.
That
completely
replaces
the
entire
HDTV
body,
which
is
perfectly
allowable
versus
changing
the
HP
verb,
especially
when
a
delete
is
changing
the
verb
under
the
covers
it
just
doesn't
do
it
their
web
look
I.
E
A
That's
a
great
way
to
phrase
it:
okay,
I'm
just
I'm,
just
wondering
whether
we
made
a
mistake
in
in
discussing
this
amongst
ourselves
too
much
in
the
sense
that
we
there
are
times
we
make
assumptions
about
what
can
and
cannot
change
and
I
know,
for
example,
you
and
I'm
working
at
talks
with
what
was
that,
guy
that
other
Red
Hat
guy
did
we
talk
to
it?
I
was
in
the
car
coming
back
from
Virginia.
A
A
E
B
One
of
the
changes
that
we
could
still
make
on
Monday
or
between
now
and
Monday,
is
potentially
to
drop
a
couple.
These
choices,
if
we
wanted
to
you,
know
to
try
to
make
sure
that
we
have
a
couple
of
succinct
choices
to
vote
from
one
next
week,
Mike
good
here,
I,
don't
know
if
that
makes
sense,
but
I
guess
we
all
just
make
sure
that
you
know
if
we
got
if
we
have
four
options
here,
that
there
are
four
owners
that
are
willing
to
staunchly
support
that
choice.
I
guess,
yeah.
E
E
B
E
Well,
we've
done
that
and
we've
gotten
facts,
I'm
concerned
by
the
sort
of
the
shifting
sands
that
I've
been
making
decisions
on
you
know
six
months
ago.
It
was,
you
know,
go
do
this
than
it
was
no.
The
the
API
is
very
important
to
us
and
we
must
conform
to
the
exact
you
know,
definition
of
the
API,
so
I
went
and
did
something
else,
and
now
it's
well.
E
A
No,
it's
something
happen.
Last
night,
they
really
put
me
in
a
bad
mood
and
I
I
was
up
to
like
past
1:00
in
the
morning
and
and
I
probably
main
is
to
go
to
bed.
You
know
a
wife
turns
over
as
it
looks
at
me
and
asks
you
know
why
I
was
up
so
late.
If
I
fell,
people
suck,
that's
all
comes
down
to.
You
know
what
you
know:
who's
great
who's,
that
dogs
dogs,
oh
god,
dogs
are
wonderful.
Yes,
the
apple
of
my
eyes,
sitting
right
over
there
right
now.
Looking
at
me,
yeah.
A
B
D
B
Thanks
for
your
work
on
this
I
know,
I
understand
the
frustration,
but
you
should
know
that
there's
also
appreciation
for
all
the
work
you
have
put
into
the
prototypes
and
this
kind
of
thing
you
know,
Paul
said
it
as
much
and
I.
Don't
know
if
you've
heard
it,
but
there's
definitely
some
appreciation.
There
I
realize
it's
frustrating
okay,
yep.