►
From YouTube: CNCF TAG App-Delivery 2022-05-04
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
B
C
D
D
A
A
A
Okay,
as
it's
already
five
minutes
late,
I
think
we
can
start.
Yes
welcome
to
today's
tag
meeting
today
we
have
a
few
agenda
points.
I
think
they
are
a
bit
less
than
we
than
we
had
initially.
So
I
I
think
we
had
a
presentation
of
this
yeah
which
is
not
taking
place.
A
No,
so
it
was
not
me,
but
nevertheless
it's
let's
talk
to
him
and
when
he
wants
to
wants
to
present.
D
So
last
I
was
the
only
one
in
the
last
meeting,
so
I
shared
some
stuff
with
myself
and
I,
but
I
was
working
on
this
evergreen
charter
thing
and
I
put
a
link
in
there.
It's
kind
of
to
say:
if
we
had
people
these
are
the
things
we
would
do,
but
so
take
a
look
see
what
you
think.
I
think
it's
on
the
I
think
it's
the
right
stuff,
but
we
need
more
support
to
do
it.
So
we
know
about
that
issue,
thomas.
D
Kind
of
it
was
never
like
officially
accepted,
but
yeah.
It
makes
sense.
In
fact,
that's
why
I
put
on
their
kcp,
which
is
an
interesting
project
coming
out
of
red
hat
right
now
around
multi-tenancy,
but
yeah.
I
don't
know
we
kind
of
need.
We
kind
of
need
someone
to
drive
that,
like
I
said,
I
was
only
one
of
the
last
meeting
and
robert
and
me
were
the
only
ones
in
the
one
before
so.
E
Yeah,
I
think
the
the
multi-tenancy
our
tenants
white
paper
is
is
it's
like
I
said
a
couple
of
times:
it's
a
really
interesting
project
to
do.
At
the
same
time,
we
just
we
just
need
to
kick
started
and-
and
we
haven't
been
able
to
do
that
due
to-
I
guess-
a
lot
of
actors
on
a
lot
of
people's
place,
so
anything
that
we
can
do
to
get
that
ball.
E
Rolling,
I
think,
would
be
interesting,
but
at
the
same
time,
since
we're
talking
about
doing
operator
white
paper
part
two
or
version
two,
maybe
we
have
the
the
the
other
white
paper
a
little
bit
on
the
on
the
back
burner
and,
for
instance,
for
from
from
my
own
sake,
I
would
like
to
just
even
be
like
a
shadow
on
on
the
operator
white
paper
just
to
kind
of
get
into
that
zone.
A
E
We're
going
to
discuss
that
and
this
one's
over,
while
the
the
multi-tenant
multi-tenancy
wiper
base
a
little
bit
more
substantial,
it's
something
from
scratch
and
we
need
to
establish
a
group
of
people.
That's
doing
it
and
everything
like
that.
So.
A
From
my
perspective,
it
would
make
sense
if,
if
the
cooperative
delivery
working
group
would
work
a
bit
more
on
the
on
the
multi-tenancy
white
people,
because
I
think
this
could
this
could
get
people
into
the
into
the
working
group.
Because
it's
a
topic
which
is
very
interesting
for
many
people
and
therefore
this
we
should.
We
should
have
kind
of
a
focus
there.
E
Always
well
yeah
yeah.
I
agree
it's
it's.
It's
a
really
good
projects
to
have
under
the
corporate
delivery
working
group.
It
just
happens
to
be
mostly
me
and
josh,
and
I
you
know
I
it's
kind
of
hard
to
to
you
know,
split
the
tension
between
I'm,
both
in
the
get
ups
working
group
and
incorporated
working
group
and
and.
C
E
If
we,
if
we
we
can
start
looking
into
it,
we
can
start
planning
stuff
but
like
attack
it
after
the
white
paper.
The
operator
white
paper
at
least,
has
you
know,
beginning
like
began
to
roll,
at
least,
if
nothing
else.
A
F
By
the
way
on
this
topic
getting
interested
in
getting
more
reach,
I
think
one
of
the
problems,
I
think,
is
less
than
people
that
we
don't
have
necessarily
interest
in
it.
I
think
there
are
enough
people
out
there
who
have
actually,
but
the
problem
is
also
very
much
the
people
most
likely
really
don't
know
about
it.
F
This
was
an
action
item
that
I
took
away
a
while
back
and
we
have
the
possibility
to
publish
on
the
cncf
blog,
so
we
can
use
cncf
means
to
publish
and
push
things
out
and
we
should
actively
start
using
it
like
security
and
others
are
using
it
because
there's
enough
people
out
there,
that
would
have
an
interest
to
contribute.
I
think
right
now
they
just
don't
learn
about
it.
F
Having
some
information
there,
showing
people
how
they
can
contribute
and
so
forth,
will
help
to
get
more
visibility
and
more
people
than
right
now,
because
I
think
right
now,
as
we
are
a
small
group
I
mean
it
was
there's
always
obviously
ups
and
downs
on
size.
It's
just
a
bit
hard
to
get
people
always
interested
in
this.
So
that's
also,
I
think
what
we
should,
at
least
for
one
of
those
initiatives
start
to
do
have
a
follow-up.
There
was
amy
that
would
help
us
as
well.
E
If,
if
you,
if
you're
at
least
then
you
know
got
word
on
how
we
can
do
a
blog
post
or
something
like
that,
I
I
could
I
could
write
one
for
the
multi-tenant
white
paper
and
like
a
call
to
action
there.
F
Yeah,
I
was
just
also
traveling,
but
I'm
working
on
it.
F
I
was
already
doing
it
so
thanks
for
for
the
offer
I'll
get
back
to
you
as
soon
as
I
have
more
details
on
this
one,
but
I
think
that's
that's
really
the
key
on
on
how
to
do
it,
and
if,
honestly,
if
we
don't
get
like
being
able
to
publish
it
on
the
cncf
blog
like
we
can
easily
even
take
that
information,
put
it
on
a
nicely
formatted,
even
github
page
and
share
it
out
or
that
we
don't
use
medium,
which
the
cncf
doesn't
want
us
to
do,
and
you
even
have
it
available
there.
F
A
A
Okay,
so
we
have
I
reached
out
to
them
to
the
people
of
the
other
of
the
other
working
groups,
but
I
think
this
was
too
you
know
too
short
noticed.
Does
anybody
know
something
more
about
the
github
working
group
or
the
kios
engineering
working
group.
E
I
I've
had
a
a
a
bit
of
a
month
of
not
being
able
to
do
anything
so
I
haven't,
I
haven't
been
in
get
up
so
ending
up
to
get
out
working
group
meetings
lately,
but
I
haven't
really
seen
anyone
complain
about
anything.
So
I
think
I
think
it's
business
as
usual,
people
are
just
getting
ready
for
qcon.
I
think
that's
that's
that's
about
it,
but
obviously
yeah.
I
don't
think
there
is
on
any
big
major
issues
or
something
like
that.
E
They
are
working
a
lot
with,
like
the
the
trying
to
break
up
stuff
like
who
is
a
member
of
open
get
ups
and
what
what
kind
of
member
the
you
know
tears
are
there.
There
was
some
that
was
a
little.
E
Some
people
were
not
happy
with
the
founding
member
here,
which
is
kind
of
weird,
but
there
has
been
a
little
bit
of
someone
who
was
not
happy
about
how
people
got
credit,
which
I
find
kind
of
weird
like
I,
I
had
a
crayon
as
a
a
member,
not
a
family
member,
even
though
I
was
there
from
almost
the
start,
and
I
don't
care
because
obviously
the
bigger
names
they
pull
their.
They
have
a
lot
of
started
and
all.
But
some
don't
see
it
like
that.
I
guess
so
so.
F
Yeah,
oh
and
also
josh,
to
not
did
the
charge
here
the
one
from
the
github
certain
group
on
what
their
plans
really
are,
because
if
they
really
want
to
shut
it
down,
then
they
should
have
a
plan
to
do
it
and
want
to
do
everything
as
part
of
the
project.
E
Yeah
they're
they're
they're
working
on
it.
That
was
the
idea
from
the
start,
having
the
working
group
as
a
short-lived
thing
to
get
the
open,
get
ops
product
up
and
running,
and
now
the
open
ghettos
project
is
in
the
cmdf
and
there's
a
website
there's
you
know
all
these
kind
of
things
around
it,
but
it's
been
it's
been
lagging
behind
when
it
comes
to.
You
know,
transferring
certain
documents
and
so
on
and
so
forth.
E
E
A
Perfect,
yes,
then,
we
have,
I
think,
two
project
updates,
so
one
from
the
feature
project
and
the
second
one
I
think
chris
is
also
here
now.
We
will
hear
some
presentation
of
the
mystery
project
so
who
of
you
both
wants
to
start.
B
I
I
won't
do
it
today.
I
I
saw
it
in
vietnam,
sorry
for
for
interrupting.
I
saw
it
in
that.
I
had
to
cancel
it.
I
had
to
cancel
the
previous
meeting
and
I
was
planning
to
reschedule
today
for
another
for
for
another
meeting,
so
so
yeah,
sorry
for
sorry
for
the
mix-up,
but
yeah
I
would,
I
would
prefer,
if
we,
if
we
were
to
do
it
next
time,
is,
is
that
okay,
with
everyone.
F
I
mean
I
was
it's
like
today,
not
that
many
people
here,
so
we
can
then
later
on,
move
it,
but
the
last
time
it
was
always
last
minute,
cancellations
and
then
usually
we
move
other
things
to
the
next
meeting
so
like
a
short
heads-up
then
would
be
appreciated
and
that
we
could
really
do
it
in
the
next
meeting.
That
would
be
great.
F
It's
fine!
If
you
want
to
move
it
again,
that
that's
fine
too,
but
we
can
definitely
otherwise
start
with
open
githubs.
And
then
you
can
decide
whether
you
wanna.
C
Get
started
yeah,
so
there
there
is
a
lot
of.
There
is
a
lot
of
open
openness.
Like
always
said
I'll,
go
ahead
and
introduce
myself
since
I
think
this
is
the
first
time
this
is
the
first
time
I've
attended
one
of
these
meetings.
I'm
todd,
I'm
an
engineer
at
dynatrace,
I've
been
working
on
open
feature
which
I
think
has
been
mentioned
at
a
few
other
meetings,
so
I'll
give
like
a
a
quick
refresher
on
on
what
it
actually
is
and
then
give
a
an
update,
including
a
demo.
C
I'm
gonna
try
and
go
through
this
stuff
fast,
but
please
cut
me
off
if
you
think
I'm
taking
up
too
too
much
time.
So
just
a
reminder.
What
open
feature
is
it's
a
vendor
neutral,
open
standard
for
feature
flags,
including
sdk,
implementations
for
various
languages
as
well
as
some
cloud
native.
You
know
kubernetes
native
components.
C
Over
the
weekend
we
got
a
sudden
bump
in
activity.
Actually,
our
our
specification
repository
went
from
like
25
stars
to
like
100
and
something
so
the
project's
getting
some
interest
recently
and
it's
growing
pretty
quickly.
We
have
a
pretty
good
vendor
participation,
I'll,
go
ahead
and
share.
My
screen
now
so
this
might
have
been
something
that
has
been
brought
up
before,
but
I'll
revisit
it.
So
now
we
have.
C
We
have
this
interested
parties
file
in
our
community
project
and
it's
just
you
know,
listing
a
whole
bunch
of
people
who
are
involved
in
the
product
are
in
the
project.
At
this
point
you
can
see
we
have
a
good
representation
across
a
bunch
of
different
vendors,
many
of
which
are,
you
know,
sas
feature
flight
vendors,
so
flagsmith
launched
darkly.
C
You
may
have
heard
of
harness
unleashed,
there's
a
unleash,
there's
a
whole
bunch
of
them
as
well
as
we
have
some
some
interested
parties
who
are
building
their
own
feature,
flagging
systems
and
have
so
ebay
is
a
really
good
example
of
that.
So
it's
good
to
have
their
contribution
as
well.
This
list
continues
to
grow
a
really
cool,
diverse
representation
of
interest.
Here.
C
I
also
wanted
to
just
give
an
update
on
the
actual
specification.
So,
as
I
said,
a
big
part
of
open
feature
is
a.
You
know
a
language,
a
vendor-neutral
multi-language
specification
for
a
feature,
flag,
sdk,
essentially
so
a
lightweight
api.
And
then
you
know
you
can
pretty
much
include
in
any
package
without
any
dependencies
and
then
you
can
implement
providers
for
any
vendor
using
the
sdk.
D
C
C
Yes,
that's
one
of
the
primary
goals
so
essentially
yeah
the
primary
goal
from
like
the
client
sdk
perspective
is
exactly
what
you
just
said
to
completely
abstract
the
source
of
any
feature
flag.
I
I
mean
I
get
you
call
this:
the
source
of
truth
for
any
future
flight
or
any
future
flag
evaluation
to
completely
abstract
that
and
then
present
a
consistent
api
across
all
those
vendors.
C
All
those
vendors
or
implementations
like
I
said
they
might
not
all
be
vendors,
so
I'll
demo,
a
little
bit
of
that
today,
but
you
got
it.
That's
that's
it.
So
this
is
the
spec,
but
you
also
are.
D
C
Right,
yes,
so
so
we
have
we're
basically
going
to
have
multiple
implementations,
so
basically
one
for
every
language.
Well,
one
for
every
target
language.
Our
initial
target
languages
are
go
dotnet,
java
and
node
and
when
I
say
node,
I'm
mostly
talking
about
server-side
javascript.
C
So
the
the
service
side
more
or
less
what
the
what
the
providers
communicate
with-
and
this
might
be
made
more
clear
when
I
do
the
demo
it
is-
is
really
mostly
out
of
scope
of
the
specification.
C
So
within
our
within
our
goals,
we
do
want
to
develop
a
cloud
native
provider
which
would
talk
to
a
kubernetes
native
service
which
would
maintain
a
you
know
a
contract
and
features
that
are
compatible
with
open
feature,
but
the
cloud
native
components
themselves
aren't
explicitly
defined
in
the
spec,
because
the
spec,
at
least
in
terms
of
the
multi-language
support
in
the
sdk,
the
part
of
the
spec
we're
talking
about
now,
is
completely
agnostic
of
where
the
source
of
truth
for
any
of
this
data
comes
from.
C
It's
it's
your
job
as
a
spec
implementer,
sorry
as
an
api
implementer
to
simply
define
the
abstractions
necessary
to
communicate
with
that
flag
control
plane,
be
it
a
kubernetes
native
implementation
which
we
will
implement.
I
guess
that's
the
reference
implementation
you're
talking
about
or
a
vendor.
C
But
the
specifics
of
how
that
particular
cloud
native
feature,
flag
operator
and
associated
service
work
is
not
something
we're
going
to
specify
here.
We
may
develop
an
additional
specification
for
that
provider
for
that
flight
control
plane,
but
it's
kind
of
out
of
the
scope
of
the
you
know
what
you
mentioned
first,
which
is
the
vendor-neutral
api
that
you
actually
use
across
various
languages
to
call
a
flag
evaluation.
I
hope
that
makes
sense.
E
C
Every
vendor
provides
almost
every
vendor,
provides
sdks
themselves.
A
lot
of
the
providers.
I
mean
you
could
have
a
provider
that
calls
a
rest
service,
but
at
the
end
of
the
day,
a
lot
of
the
providers
are
going
to
wrap
vendor
feature
flag
sdks,
which
themselves
have
the
awareness
of
the
of
the
upstream
flag
control
plane.
If
you,
if
you
follow
me,.
D
F
Yeah
actually,
a
lot
of
this
is,
as
I
was
involved
in
the
project.
Actually
a
lot
of
us
inspired
by
how
open
telemetry
work.
Think
of
like
the
sdk
for
feature
flagging
as
the
sdk
on
the
api
would
see
an
open
telemetry
and
then
that
the
provider
interface,
which
start
will
show
here
as
small
as
what
the
exporter
does
in
open
telemetry
to
have
like
this
similarity.
C
Yeah,
yeah,
and
and
and
to
your
point
it
may
be,
it
may
be
great
to
have
a
separate
specification
really
dealing
with
how
to
generally
implement
a
cloud
native
provider.
Yeah.
D
A
C
No
problem
so
yeah
this
the
specification
is,
is
still
alpha
and
it's
it's
pretty
incomplete,
but
we
do
have
the
main
parts
of
the
application
author
facing
api
defined
as
well
as
the
providers
and
the
provider
is
really
that
adapter
layer
that
we're
talking
about
that
allows
you
to
connect
with
various
vendors.
So
I
I
won't
dive
too
much
into
the
spec
here,
because
it's
not
very
exciting
feel
free
to
look
at
it.
If
you
want
it's
under
active
development.
C
Obviously
I
will
do
a
quick
demo,
and
this
is
the
latest
demo.
We
have
it's
not
quite
published
yet,
but
what
you're
going
to
see
here
is
basically
a
ui
for
a
sample
sas
app
landing
page
software
as
a
service
landing
page
and
it's
going
to
take
advantage
of
multiple
features
within
our
current
sdk
implementation
within
our
alpha
sdk
implementation
and
I'll,
try
and
run
through
that
as
quickly
as
I
can
but
feel
free
to
stop
me
or
ask
questions.
C
Okay,
so
we
have
this,
you
know
very
contrived
landing
page
that
you're
seeing
here
and
it
has
a
it-
has
a
a
bunch
of
different
features
on
it
that
are
connected
to.
In
this
case
I
launched
this
using
the
launch
darkly
provider,
so
we
have
three
other
providers
with
vendors
currently
now
in
our
in
our
sample
demo
repository,
but
I'm
just
gonna.
You
pick
on
launch
darkly
this
time.
It
could
be
anything
else.
It
could
be
flagsmith
or
cloudbee's
future
management
application,
but
I'm
going
to
just
do
some
basics.
C
So
you
know
the
marketing
department
is
constantly
changing
the
name
of
the
app
you
know
you
can
go
into.
This
is
the
launch
darkly's
future
flag
management
console.
You
can
go
ahead
and
flip
a
switch.
That's
going
to
change
the
name
of
the
app
so
you'll
see
here
that
the
the
application
name
changes
according
to
that
oops.
I
flipped
the
wrong
switch
here.
C
So
I'll
turn
off
this
new
welcome
message
and
you'll
see
this
string
change.
I
mean
there's
nothing
too
too
crazy
about
that
here,
but
you
can
see.
This
is
a
very,
very
basic
boolean
flag
right.
It
has
no
context
awareness,
no
awareness
of
who
I
am
or
anything
like
that,
and
the
flags
get
increasingly
complex,
so
another
one
I
can
do
is
go
ahead
and
change
the
value
of
a
color,
so
I'll
go
ahead
and
make
this
red.
C
So
this
is
our
like
an
overall
style
change.
I
mean
you
can
see
a
whole
bunch
of
components,
change
color.
This
is
this
is
pretty
basic
stuff,
but
obviously
most
future
flag.
Vendors,
in
addition
to
boolean
flags
that
turn
something
on
or
off,
they
can
also
have
payloads
that
define
specific
values.
So
you
know
I
can
change
this
color
to
any
number
of
colors
to
blue
green
or
whatever
I
have
defined
here.
C
Things
get
a
little
bit
more
interesting
when
we
involve
targeting
so
I'll
show
you
what
that
looks
like
and
again
this
is.
This
is
common
to
basically,
all
vendors
I'll
go
to
installation
instructions
here.
So
this
says:
if
my
platform
and
platform
is
an
attribute
that
exists
for
any
user
of
this
site
is
equal
to
linux,
then
I'm
going
to
show
a
certain
value
in
this
case.
It's
the
value,
as
implied
here
with
the
flag
name,
it's
the
value.
That
shows
how
I
can
download
this
this.
C
This
feature
how
I
can
download
the
app.
So
in
my
case,
because
I'm
on
a
linux
distribution,
it
says
install
our
killer
app
package
using
your
distributions
package
manager,
so
this
is
basically
pulled
from
standard
headers
that
chrome
sends.
You
know:
chrome
sends
a
lot
of
telemetry
headers
about
what
kind
of
operating
system
you're
on
various
details
about
the
host
and
we
have
defined
this
flag
so
that
it
looks
at
those
particular
values
and
then
can
actually
intelligently.
C
You
know
respond
accordingly.
If
I
was
on
a
windows
system,
this
would
say
download
our
setup
file.
C
In
addition
to
rules
like
this
that
are
based
on
particular
attributes
of
the
user
and
evaluated.
Based
on
that,
you
can
do
splits,
so
you
can
do
kind
of
a
b
testing
this
way.
So
you
can
say
you
know,
fifty
percent
of
users
are
going
to
have
this
experience.
Well,
a
lot
of
fifty
percent
are
going
to
see
this
experience.
C
So
now,
instead
of
using
launch
darkly,
what
I'm
going
to
be
doing
is
using
basically
just
a
file,
a
json
file
provider.
So
this
is
obviously
not
practical,
but
it's
really
good
for
demo
for
purposes
you
never
want
to
deploy
with
with
you
know
something
like
a
file
provider,
it
would
be
pretty
redundant,
but
now
you
can
see,
I
have
direct
control
over
these
flags
just
based
on
these
files,
this
file.
So
if
I
change
this
is
writing
to
a
file
in
my
local
file
system,
again
just
for
demo
purposes.
C
But
if
I
change
a
value
here,
you
see
it
immediately
affects
the
result.
So
I
didn't
change
any
application
code
at
all
and
I'm
switching
between
you
know
a
launch
darkly
feature
flag
control
system
to
one
defined
in
the
file
system.
I
could
have
just
as
easily
switched
to
cloud
b's
feature
management
solution,
flagsmith's
feature
management
solution
or
any
of
the
other
providers
we
support.
C
One
of
the
other
features-
that's
probably
interesting
to
show
here
is
some
of
the
extension
and
integration
points.
So
one
of
the
cool
things
we
can
do
comes
out
when
we
use
this
fibonacci
demo.
C
So
this
just
calculates
fibonacci
using
a
particular
using
a
particular
algorithm,
and
you
can
see
here
we're
using
a
recursive
algorithm,
which
is
hopelessly
inefficient.
I
can
change
that
to
a
quicker
algorithm.
That's
not
really
the
important
part
I
wanted
to
demo
here.
C
The
important
part
is
the
way
that
this
package
is
set
up,
that
this
demo
is
set
up,
we're
treating
the
fibonacci
code
like
a
third-party
library,
so
we
have
no
direct
control
into
into
any
of
the
fibonacci
code,
but
the
fibonacci
library
does
use
the
open
feature
api,
so
you
can
see
they
have
their
own
deflect.
They
have
their
own
flag
defined
here,
they're,
not
exposing
any
of
it
to
us
as
the
consumer
of
this
library
and
they're,
not
they're,
not
implementing
any
provider.
C
All
this
third-party
library
has
done
is
referenced
open
feature.
The
zero
dependency
api
again
think
the
open
telemetry
api,
which
has
no
dependencies
really
it's
it's
just
a
just,
an
api
layer
that
defines
some
basic
interfaces.
So
that's
what
this
third
party
library
has
done
and
we
can
basically
the
because
the
way
open
feature
works.
We
can
do
a
couple
things.
C
We
can
take
control
over
that
flag
value,
basically
inject
our
value
into
it
from
our
application
code,
but
we
can
also
inject
additional
context
into
it
without
having
any
without
having
any
first
level
awareness
without
having
having
any
first
party
awareness
between
this
library
and
and
our
code.
So
we
don't
have
to
configure
this
library
in
any
way
to
to
know
about
our
user
or
our
context
or
anything
like
that,
but
you
can
see
that
it
does.
It
knows
you
know
what
what
platform
I'm
on.
C
You
can
see
that
being
locked
here.
It
has
my
local
ip
address,
so
basically
we're
similar
to
how
you
would
see
an
open,
telemetry
baggage
in
the
open,
telemetry
context
are
injected
using
techniques
like
thread,
local
storage
or
continuation
storage.
We
can
inject
contextual
data
into
third-party
libraries,
flag
evaluation.
C
I
know
that's
a
bit
of
a
complex
example.
You
can
kind
of
ping
me
after
if
you'd
like
to
hear
more
about
it,
I'm
going
to
move
on
from
it
because
it
gets
a
little
bit
esoteric,
but
one
of
the
other
really
cool
things
that
this
kind
of
pattern
allows
us
to
do.
One
of
the
other
cool
features
that
our
api
allows
us
to
do
is
without
adding
any
dependencies
into
the
kind
of
the
api
def.
C
The
api
package
like
we
were
mentioning
before
we
can
do
some
first
class
extensions
and
integrations,
so
one
of
them
would
be
open,
telemetry
itself.
So
I'm
going
to
switch
this
back
to
recursive,
because
it's
a
little
bit
better
for
this
demo.
C
We
have
a
concept
of
hooks
which
allows
you
to
add
arbitrary
behavior
to
any
flag
evaluation
globally.
So
again,
whether
that
flag
exists
in
first
party
code
or
third-party
code,
you
can
add
arbitrary
behavior
around
them,
and
this
is
a
great
place
to
add
telemetry
to
add
additional
logging.
C
So
I'm
going
to
go
ahead
and
show
you
what
I
mean
we
have.
This
sample
open
telemetry
hook,
and
basically
this
uses
some
semantic
conventions
that
we
recently
proposed
to
the
open
telemetry
specification
that
that
basically
provide
additional
metadata
about
provide
metadata
about
a
flag
evaluation.
So
basically,
we
can
configure
this
open
telemetry
hook
in
our
first
party
code
and
flags
existing
anywhere
in
our
code
are
going
to
have
this
additional
behavior
added
around
the
evaluation,
and
we
have
specific
life
cycle
hooks
that
we
can.
C
Basically,
you
can
basically
define
in
this
hook.
Object
lifecycle
stages
is,
is
what
we
call
them
and
you
can
put
arbitrary
behavior
there.
So
what
that
allows
us
to
do
practically
in
the
case
of
this
hotel
hook
is
inject
telemetry
around
feature
flagging,
so
I'm
just
going
to
go
ahead
and
run
this
one
more
time
to
make
sure
I
have
recent
data
and
I
go
to
zipkin.
C
You
know,
trace,
trace
viewer,
just
going
to
run
a
query.
Now
you
can
see
there's
a
couple,
long
queries.
This
is
our
four
second
fibonacci
query
and
you
can
see
you
know
we
executed
this
request.
The
interesting
part,
as
you
can
see
at
the
bottom
here,
there's
a
feature
flag,
open,
telemetry
spam
and
you
can
see
all
kinds
of
interesting
stuff.
You
can
see
the
provider
that
was
used
to
to
get
this
fan.
You
can
see
the
actual
evaluated
value
got
so
it's
recursive
the
the
key
of
the
flag.
C
So
this
is
just
one
possible
extension
that
we
that
our
api
kind
of
allows
people
to
define.
So
you
can
get
additional
observability
into
flag
evaluations
and
really
how
they
impact,
how
they
impact
performance,
especially
if
you're
doing
a
b
testing
and
that
kind
of
thing.
So
you
can.
You
can
implement
arbitrary
behavior
here,
whether
it's
logging
even
transformation
of
the
flag,
values
themselves.
You
can
use
hooks
for
that
sort
of
stuff.
So
what
the
context
injection
stuff!
C
I
was
showing
you
earlier
as
well
as
on
these
hooks
to
add
arbitrary
behavior
behavior
to
a
flag
evaluation.
This
is
stuff
that
no
feature
flag,
sdk
currently
offers
at
all
and
for
good
reason
right.
So
you
need
some
kind
of
standardization
to
even
do
the
kind
of
telemetry
stuff
we're
talking
about
right.
C
So
you
need
a
layer
where
you
can
define
some
basic
structures
and
then,
and
then
you
know,
make
sure
that
they
conform
to
make
sure
that
they're
in
sync,
with
a
telemetry
specification
like
hotel,
for
example,
and
so
that's
why
it
becomes
important
to
have
an
open
standard
and
why
standardizing
this
stuff
really
unlocks
a
lot
of
additional
features
and
richness
and
feature
flagging.
C
So
I
think
that's
a
high
level.
That's
a
high
level
demo
of
basically
where
we
are
at
right
now.
Oh
one
thing
I
can
do
just
to
make
things
clear,
I'll
show
you
the
launch
dark
week
provider
so
again
like
josh
was
asking
about.
This
is
basically
the
a
really
simple,
very
simple
implementation
of
the
mapping
layer.
That,
essentially,
you
know,
allows
you
to
use
abstracts
launch
darkly
from
any
of
your
flag.
C
Evaluation
calls
any
of
the
evaluation
that
an
application
author
would
be
doing
it
just
implements
this
feature
provider
interface.
You
know,
which
has
to
be
defined
in
every
language
implementation
and
then
provides
a
bunch
of
methods
for
actually
resolving
those
flag
values
from
that
that
source
of
truth.
C
D
Yeah,
I
have
a
question
one.
This
is
pretty
cool.
You
know
it's
cool
to
kind
of
standardize
and
make
some
conventions
around
these
apis,
and
I
really
like
that.
You
know
you're,
starting
with
the
shims
for
the
providers
and
the
sdk
side,
because
we
can't
really
influence
the
api
providers
first,
you
know-
or
at
least
it
would
be
harder,
but
my
question
is:
is
this:
should
this
be
generalized
to
more
just
any
configuration
management
like?
How
does
why
something
specific
for
feature
flags
rather
than
you
know
like
a
spring
config?
D
C
I
suppose
that's
not
a
good
question
I
mean
generally
generally,
I
suppose
that's
a
good
question
yeah.
Thank
you.
I'm.
C
My
apologies,
I
meant
to
say
not
a
bad
question.
I
suppose
that's
not
a
bad
question.
I
think
I
think
that
feature
flags
probably
represent
a
a
super
set
of
functionality
that
you
that
that
you
can't
just
do
with
basic
configuration
so
to
go
back
to
an
example
which
maybe
might
make
what
I'm
saying
a
little
bit
clearer.
C
This
is
a
very
forgive
the
terminology
like
dumb
configuration,
it's
a
static
configuration,
so
it
doesn't
have
any
knowledge
of
who
the
user
is.
It
doesn't
have
any
rule
sets
to
find.
This
is
a
very
basic
one
that
essentially
is
just
reading
a
file,
and
so
that
would
map
to.
I
think
what
you're
saying
really
when
you
talk
about.
C
You
know
on
the
various
future,
flagging
vendors
and
kind
of
a
separate
sas
service
or
management
platform,
or
even
cloud
native
components
and
crds
for
defining
feature
flags
if
you're
not
doing
some
kind
of
dynamic
evaluation,
whether
it's
you
know
based
on
like
a
random
split
or
a
distribution
based
on
percentages
or
something
like
that
or
based
on
the
identity
of
the
user.
I
I
think
that
stuff
just
becomes
pretty
basic,
so
I
think
there's
there
are
lots
of
solutions
for
that.
C
It's
really
the
the
you
know
the
dynamic
understanding
of
the
user
and
and
rule-based
evaluation
split
percentage-based
evaluation,
a
b
testing.
I
think
that's
where
a
lot
of
the
value
is
and
that's
why
it
makes
sense
to
kind
of
to
kind
of
target
that
a
lot
of
people.
Frankly,
when
they
first
hear
about
future
flags,
they
think
just
in
terms
of
boolean
on
or
off
static
configuration
that
has
no
dynamic
input,
and
then
they
ask
questions
like
why?
C
C
D
C
Okay,
I
think
that's
all
I
have
this
this.
You
know
I'm
not
a
ui
developer,
but
this
this
ui
and
this
kind
of
demo
is
very
shortly
going
to
be
pushed
into
our
project
as
well
as
a
big
readme
around
it.
So
you
can
play
with
this
stuff.
If
you
want
to
it's
not
there
yet,
but
in
the
next
couple
days
it
will
be
so
you're
kind
of
this.
This
meeting
is
the
first
people
who've
seen
this
new
fancy
demo
but
yeah.
F
Yeah
thanks
todd
again
we'll
share
with
them
with
us
as
well
for
the
recording
and
good
questions
here,
so
we're
almost
on
top
of
the
our
still
chris,
as
you
are
here
and
obviously,
unfortunately,
random
bit
longer
here.
Do
you
still
want
to
provide
us
like
a
short
intro?
What
missed
is
yeah.
B
Yeah,
you
know,
let's,
let's
do
like
five
minutes
intro,
let's
say
because
it's
still
very
very
early
in
any
case,
so
I'm
my
name
first
of
all,
is
chris
psaltis,
I'm
the
co-founder
of
mrio
and
we're
building
an
open
source,
multi-cloud
management
platform
there.
So
over
the
next
months,
we
we
are
thinking
of
contributing
the
the
project
that
we've
been
working
on
the
past
few
years
to
cncf
or
some
other.
B
Station
so
that
we
can
improve
the
the
governance
of
the
project
and
work
with
like
a
more
transparent,
more
community-oriented
way.
So
I
wanted
to
show
you
a
little
bit
more
about
what
it
is,
so
you
can
get
an
idea
of
what
I'm
talking
about
but
like
if
I
can
share
my
screen
for
a
second.
B
So
yeah,
okay,
so
this
is
the
cli
interface
of
the
platform.
There's
also
a
web
ui,
but
this
is
one
of
the
things
that
we're
working
on.
So
we
we're
trying
to
make
it
us
like
we're,
taking
inspiration
from
cube
cdl,
let's
say
so.
Let
me
check
once
okay,
so
I'm
now
targeting
the
production
installation
of
of
this.
So
practically
what
you
can
do
is
that
you
can.
B
You
know
one
in
alibaba,
one
in
ec2,
google,
vmware
installation
of
vsphere
vulture
docker
host,
so
we
support
about
20
different
vms,
so
on
one
hand
we're
integrating
integrating
with
native
apis
coming
from
its
cloud
provider
and
then,
on
the
other
hand,
we're
exposing
the
missed
api,
let's
say
so
practically
what
you
could
do,
for
example,
is
like
get
a
list
of
all
the
all
the
vms
or
the
clusters
or
the
containers
that
we
have
running
across
the
platforms
that
you
have
connected,
or
maybe
you
know
get
just
what's
on
vulture
or
I
don't
know
like
search
only
for
for
machines
that
are
stopped,
and
then
you
know.
B
B
Or
you
can
just
want
to
check
out-
what's
the
name,
for
example,
and
the
cost
of
this
vm
and
then
export
it
on
csv,
so
you
can
get
like
a
report
for
what
you're
spending
like
or
things
like
that
and
but
they're.
You
know.
Obviously,
you
can
create
new
resources
like,
for
example,
in
machines.
B
You
can
do
something
like
you
know:
tests
test,
minus
minus
cpu
two,
and
what
this
will
do,
for
example,
right
now,
is
that
it
will
go
over
all
the
all
the
accounts
that
I
have
connected
to
to
mist
and
it
will
decide
where
it
makes
more
sense
to
provision
something
like
the
the
parameter
here
for
the
decision
is
the
cost,
so
is
aws
zipper
or
is
azure.
What's
the
the
cheapest
image
I
can
find
with
two
cpus
and
I
will
spin
it
up
there.
B
Obviously
you
can
be
more
specific
about
that,
but
this
is
really
helpful
when
you
just
need
to
spin
something
up
quickly
there.
There
are
many
types
of
resources
that
we
support.
I
already
mentioned,
you
know
compute
types
like
containers
container
hosts,
hypervisor,
hosts
vms
managed
clusters.
Let's
say:
ks
gke
things
like
that,
but
also
you
know
standalone
comparing
these
clusters
red
hat
opposition
things
like
that.
Both
you
know
public
and
private
types
of
infrastructure,
but
there's
also
networking
object,
storage,
blobs
stored,
like
sd.
B
There
is
also
nss
service,
ralph
53,
for
example,
and
then
on
top
we
are
also
layering
some
functionality,
that's
more,
let's
say
specific,
like
for
example,
scripts.
So
you
can
upload
your
ansible
playbooks
or
your
your
executables,
like
your
bus
scripts,
your
python
scripts,
and
then
you
can
run
them
from
our
platform
to
your
target
vms,
but
you
can
also
get
you
know
direct
ssh
access
to
to
those
vms
without
having
with,
without
the
need
of
having
all
the
credentials
locally.
B
These
are
stored
and
missed
itself,
so
you
authenticate
with
mist
and
that's
it
like
so
now,
I'm
I'm
in
one
in
one
of
those
vms,
you
know
without
having
access
to
my
to
a
private
key
locally.
B
Let
me
let
me,
sir,
sir,
for
a
second,
because
you
know
this
is
this:
is
work
in
progress,
so
the
cli
I
mean
like
the
web.
Ui
is
more
like
this,
so
you
know
you
see
everything
that
I
mentioned
so
far,
we're
currently
working
on
integrating
with
helm
as
well.
So
you
can
execute
a
helm
chart
on
one
of
your
clusters.
B
And
yeah
there
are
schedules
for
executing
periodic
tasks,
taking
a
backup
rotating
some
loads
things
like
that
and
yeah
so,
and
there
is
also
like
a
monitoring
service.
I
forgot
to
mention
that
so
you
can
install
an
agent
on
your
vm.
B
We
are
working
on
top
of
the
prometheus
api
for
that,
so
you
can
plug
it
into
several
types
of
back-ends
like
prometheus
victoria,
metrics
influx
db,
so
yeah.
So
you
can
monitor
your
metrics
set
some
rules,
alerts
that
might
even
trigger
some.
Automated
actions,
like
you
know,
run
this
script
whenever
this
happens
and
yeah,
so
that's
like
in
a
in
a
very,
very
brief
period,
the
the
platform.
B
So
what
I
was
what
I
was
trying
to
to
to
get
out
of
this
demo
was:
you
know,
to
hear
your
thoughts
and
see
if
it
makes
sense
or
if
you
think
that
it
makes
sense
in
the
such
a
project
in
the
context
of
cncs,
and
you
know,
if
we,
if
it
does,
then
you
know
we
should
probably
consider
what
what
would
be
the
the
next
step
for
applying
to
to
the
sandbox,
but
at
least
at
a
high
level.
B
I'd
like
to
to
hear
your
thoughts
of
like
does
it
make
sense
in
the
in
the
ecosystem
of
cncf,
because,
as
as
you've
seen
we're
not,
you
know,
cloud
native
specific,
where
we
practically
allow
you
to
manage
any
type
of
infrastructure
containers
or
not.
We
don't
we
don't
really
care
so
yeah.
So
that's
that's
what
I
was
trying
to
get
out
of
this
discussion.
F
B
It's
it
is
90
equivalent
like
there
are
a
few
things
that
are
missing
from
the
open
source
version.
These
are,
you
know.
Mostly
enterprise
features
like
you
know,
authenticating,
with
active
directory
of
that
role-based
access
control,
things
like
that,
but
other
than
that
everything
is
is
under
on
a
path.
Cv2.
F
F
B
I
I
don't
know
if
this
is
going
to
be
like
the
end
pro
the
end
project
that
we
are
going
to
submit
to
be
honest,
so
figuring
out
what
the
the
perfect
mix
would
be
is
part
of
what
we
will
do,
but
yeah
I'm
showing
you
the
entire
thing.
Now
what
what
will
be
part
of
the
submission?
I
don't
know,
for
example,
I
don't
know
if
we
will
submit
the
web
ui,
because
it
makes
the
whole
project
more
complicated
and
much
bigger
and
difficult
to
contribute
to.
B
It
would
probably
be
like
the
the
back-end
side,
the
back-end
server
and
the
cli,
or
something
like
that.
But
you
know
this
is
still
up
in
the
air
I
mean
the
web.
Ui
is
already
open
source,
but
you
know
it's
it's
harder
to
deal
with.
It
makes
the
whole
project
more
complicated.
A
F
B
It's
certainly
it's
certain
there's
a
certain
overlap
for
sure
I
I
would
say
that.
Okay,
in
our
case,
you
don't
need
to
run
kubernetes
to
to
manage
the
infrastructure
and
you
don't
depend
on
because
you
don't
depend
on
crds
so,
but
on
the
other
hand
you
know
yeah,
so
it's
it's
some
somewhat
somewhat
related,
let's
say
yeah.
Would
I.
D
Yeah,
I
just
want
to
be
the
the
devil's
advocate,
and
this
will
clear
up.
Another
question
I
have
should
crossplane
use
this
underneath
rather
than
going
directly
to
the
cloud
providers,
what
would
it
lose.
D
Like
right
now,
they're
going,
in
other
words,
I'm
assuming
and
correct
me
if
I'm
wrong,
this
is
basically
a
shim
api
over
all
of
the
cloud
providers.
Apis,
yes,
yeah
and
crossplane
has
the
complexity
of
calling
every
you
know,
underneath,
although
I
guess
it's
using
ack,
but
what
what
are
you
losing?
You
must
be
losing
like
how,
if
I
can't
can
I
get
every
feature
out
of
the
underlying
clouds
like
I
saw
you.
B
Have
a
live
cloud
yeah!
You
know
you
don't
get
every
feature
out
of
the
underlying
clouds,
but
this
is
I
I
I'm
guessing.
This
is
a
an
issue
that
crossplan
are
also
facing.
I
mean
if
it's
not
part
of
the
sdk
or
whatever
you're
missing
it
there
as
well.
So
yeah,
I
think,
like
the
the
main.
The
main
difference
here
is:
if
you
need
kubernetes
to
run
your
management
system
or
not
like
I,
I
I
get,
I
I
think
it.
It
ends
up
being
the
critical
question.
B
So
do
you
want
your
management
to
re
to
leave
inside
kubernetes
itself
or
not?
B
So
I
think,
in
this
context,
crossplane
would
be
let's
say
more
ideal
for
a
group
of
of
users
who
are
very
kubernetes,
focused
and
very
kubernetes,
experienced,
let's
say
missed,
would
be
more
ideal
with
in
organizations
who
are
either
making
their
first
steps
with
kubernetes
or
they
don't
have
a
lot
of
experience
in
it
or
they
don't
they
don't
just
they
just
don't
want
to
run
kubernetes
cluster
just
to
just
to
have
the
management
layer,
the
centralized
management
layer
so
yeah,
I
think,
like
there
are
several
pros
and
cons.
B
That's
that's
for
that's
for
sure.
D
B
B
D
You
know,
instead
of
azure,
you
know,
create
a
vm
and
aws
create
ec2
instance
and
gcp,
whatever
you're
saying
just
create
a
missed
vm
and
it
will
handle
it.
B
Yeah-
and
this
is
part
of
what
we're
trying
to
do
like
create
a
of
an
abstraction,
so
people
don't
need
to
deal
with
all
the
details
in
some
cases
you
want
and
we
we
can
still
provide
that.
But
you
know
in
many
cases
you
just
want
like
a
vm
quickly
or
something
like
very
quickly
without
having
to
worry
all
you
know.
Where
do
I
put
that?
Do
I
put
it
in
azure
or
google
or
aws,
so
you
know
the
and
who
cleans
it
up
afterwards.
B
So,
like
that's
part
of
what
we
do
like
we,
we
help
you
with
proactively,
let's
say
making
sure
that
people
will
not
forget
an
extra
large
instance
running
on
aws
costing
you
a
fortune
because
they
wanted
to
test
something.
F
I
think
it
is
an
interesting
project,
but
you
brought
up
like
some
points
of
view.
I
think
you
should
get.
I
mean
that
that
crossplay
and
delineation
obviously
running
on
kubernetes
and
not
running
on
kubernetes.
This
is
like
one
thing
that
you
can
do,
but
I
think
functionality
wise.
I
would
also
be
interested
in
like
what
you
provide
or
what
you
would
then
would
provide
compared
to
cross
plane
where
the
overlap
is
where
you
see
the
the
delineation.
F
Okay,
also
that,
like
the
dependency
on
an
apache
project,
nothing
wrong
with
apache
projects,
but
due
to
josh's
point
just
ask
answering
the
question:
is
this
the
the
right
place
to
live?
If
there's
like
a
dependency
there
and
and
then
also
like
how
to
how
do
you
like
going
forward,
are
planning
to
kind
of
maintain
this?
That
was
one
I
remember
when
we
talked
to
crossplane.
F
F
I
think
they're
doing
a
great
job
there,
but
in
your
case
you
build
even
an
abstraction
layer
on
on
top
of
that,
and
that
might
be
probably
even
harder
and
again,
I
assume
you
see
value
in
there,
but
that's
maybe
something
to
think
about
like
how
you
would
position
this,
how
you
want
to
keep
it
up
to
date
and
like
also
that
defining
the
features
that
like
for
me
from
from
personal
experience,
I
I
think
we
would
never
most
people
would
never
use.
F
It
does
not
have
like
any
role-based
access
control
for
creating
cloud
resources.
So
but
again,
this
is
still
for
you
to
think
about
what
you
want
to
put
in
there
and
what
you
don't
want
to
put
in
there
I
mean
the
active
directory
piece
is
something
else.
F
I
know
it's
not
necessarily
required
at
an
early
stage,
but
for
I
think,
adoption
once
you
then
go
into
incubation
and
obviously
later
on
graduation.
These
requirements
from
the
security
perspective
and
also
from.
F
Option
will
come
up
because
if
everybody
can
just
go
in,
who
has
a
login
and
create
save
50
production
instances
without
having
the
rights
to
do
so
will
make
it
very
hard
to
adopt
as
an
open
source
project
per
se.
And
still
it
should
be
obviously
possible
to
adopt
it
in
a
pure,
open
source
fashion
as
well.
But.
F
We
then
can
have,
as
you
proposed,
that
at
the
next
meeting
we.
B
Yeah,
I
will.
These
are
very,
very
good
points.
I
will.
I
will
look
into
those
and
I
will
get
back
to
you.
You
know
some
in
any
case,
this
is
not
something
there
isn't
I
mean
it's
something
that
we
are
still
thinking
about,
so
I
can
get
back
and
discuss
more
details
in
future
pro
meeting
after
after
kubecon
all
right.
F
Yeah,
let's,
let's
do
this
and
again
also
you
don't
have
to
do
it
for
sandbox.
I
still
always
recommend
looking
at
the
guidelines
for
incubating
projects,
not
that
you
have
to
fulfill
them,
but
they're
also
like
good
food
for
thought
like,
for
example,
like
relation
to
other
cncf
projects,
and
they
would
obviously
put
something
in
there
like
how
is
it
related
to
crossplane?
How
is
it
related
to
other
topics.
F
Ad
hoc
and
thank.
F
It
is,
it
is
an
interesting
problem
that
you
saw.
It
is
a
problem.
I
think
a
lot
of
people
have
as
well
and
yeah.
Let's
then
talk
again
in
a
couple
of
weeks
and
also,
if
you're
going
to
be
at
kubecon,
you
will
meet
some
of
the
people
here
as
well,
also
feel
free
to
say
hi,
obviously,
yeah.
B
Unfortunately,
I
won't
be
there
this
year,
okay,
okay,
thank
you
very
much.
Yeah.
F
So
as
we're
over
time
here
already
I'll
call
it
a
meeting
for
today
thanks
everybody
for
joining,
as
always
as
we
will
put
the
meeting
again
on
youtube,
thomas
some
follow-ups
for
us
and
also
share
it
with
the
wider
community,
thanks
everyone
and
have
a
nice
rest
of
your
day,
thanks.
A
A
Oh,
like
previously,
I
thought
that
the
time
was
changed
to
like
a
16
utc
right.
F
F
F
E
F
A
Okay,
cool
just
making
sure
it's.
It's
like
88
a.m,
in
pxp
that
particular
time
zone
right,
but.
F
A
Yeah-
and
I
think
we
should
change
this
in
the
in
the
in
our
repository
because
yeah,
it's
really
pointed
in
utc.