►
From YouTube: CNCF SIG Network 2020-01-16
Description
CNCF SIG Network 2020-01-16
A
A
B
B
E
So
I
might
repeat
this
a
time
or
two
and
Matt
may
as
well,
but
this
SIG
is
just
just
forming
just
getting
underway,
and
so
today
it
is
well
as
we
in
the
call
today
it
will
be,
it
will
be
launched
yeah.
We
will
have
done
some
earnest
considerations,
so
I'm
excited
for
today's
agenda.
This
is
this
is
great
by
the
way
my
Nick,
my
name,
is
Lee.
E
If
I
haven't
spoken
to
you
before,
Matt
Klein
is
on
the
call
as
well
so
bye,
hi
Matt
is
the
TLC
liaison
for
the
sig,
so
some
respects
the
sponsor
for
the
sig.
If
you
will,
ken
Owens
is
also
co-chairing
this
sig
network,
along
with
me,
so
as
we've
begun
to
establish
practices
for
the
meetings,
if
you
would
not
dissimilar
from
many
other
open
source
calls,
they
are
on.
The
meeting
minutes
are
a
collaboration,
much
like
much
like
the
work
that
you
guys
do
out
in
open
source
land.
E
Do
we
know
we've
got
representation
today
from
project
contour.
It
was
just
having
an
interaction
with
folks
representing
SMI
and
I,
mean
I,
see
some
of
them
on
the
call
now
very
good,
and
some
of
those
folks
may
even
be
representing
both
of
the
project,
I'm,
not
sure,
but
we're
covered
either
way,
so
good,
it's
a
very
good
five
after
hopefully
that
sets
so
pretext.
Many
of
you
are
familiar
with
these
types
of
presentations
and
and
reviews.
Any
additional
context
mat
that
you
might
want
to
set
before
we
invite
the
contour
team
to
present
I.
E
C
Hi
everybody,
so
this
is
kind
of
the
first
time
were
we're
looking
at
this
a
new
way
of
basically
presenting
projects
to
the
different
SIG's,
so
I'm
not
sure
I
know
what
to
expect
from
from
from
butene.
Has
anybody
had
a
chance
to
look
at
our
PR
that
we
file
for
contour?
That
has
some
of
the
relevant
details.
C
E
And
actually,
let
me
apologize
up
front
Michael
you'd,
ask
this
question
of
mere
a
few
days
ago
about
you
know,
items
to
bring
and
things
to
repair
and
I
didn't
get
back
to
you.
We
actually
have
a
something
of
a
template
for
the
things
that
we're
looking
for.
I
expect
that
of
what
you
said.
I
expect
you
you've
got
those
things
that
that
that
we
were
that
we're
looking
for
and
one
review
I'll
send
out
that
template
post
after
the
call,
because
it'll
be
really
after
the
call
yeah.
C
That'd
be
great:
if
you
can
email
me,
the
template
I'll
make
sure
that
we
basically
provide
all
the
information
that
that's
basically
I've
been
asked
for
us.
So
I'm
going
to
go
through
this
presentation.
We'll
have
a
few
team
members
here
as
well.
We
have
a
Steve
sloka
who
have
David
Cheney
and
Tim
hunter
later
different
folks,
we'll
talk
a
different
area.
Since
you
know
they,
they
have
expertise,
ins,
but
to
hear
it
from
the
actual
folks
the
build
certain
components
of
the
architecture
but
I'm
assuming
you
can
all
see
my
screen.
We.
E
Can
and
actually
Michael
to
two
quick
things
to
set
context
for,
and
just
help
provide
clarity
on.
One
is
that
you
know:
Kathy
had
a
lot
of
things
as
I
say
here
like
unless
I'm
unless
I'm
mistaken,
or
rather
it's
my
understanding
that,
yes,
that
projects
that
go
into
incubation
stage
at
that
point
do
go
under
the
full
due
diligence
for
projects
so
kind
of
this
presentation
and
there's
a
number
of
other
items
for
due
diligence.
I.
E
I'm,
the
top
of
our
link
to
that
due
diligence
know
about
incubation
stage
in
it
and
I
think
it
applying
at
that
time.
The
second
note-
and
this
is
I,
think
other
than
Ken
who
might
be
on
the
call
an
hour
at
some
point.
He
might
be
one
of
the
only
others
that
recollect
this
show
that
were
around
at
the
time,
but
prior
to
CN,
CF
SIG's
there
and
many
of
you
have
been
in
in
some
other
work
work
groups.
One
of
them
was
the
networking
working
group
and
C&I.
E
C
And
we
meet,
we
meet
all
of
these
incubation
state
requirements.
I'll
talk
a
little
about
the
customers
later
on,
but
we
need
all
of
the
all
of
the
Commons
who
I've
seen
this
before
many
times
since
I'm,
also
trying
to
get
Harbor
to
the
graduation
stage,
so
we'll
create
a
document
and
work
with
you
guys
to
produce
the
due
diligence.
Talk
unsetting.
Also
your
template
questions
so
we'll
do
that
within
the
next
week
or
so.
Okay.
C
So
so,
let's
talk
a
little
bit
about
contours,
so
our
website
is
project
concluded
IO
where
I
know
like
I
mentioned
earlier.
An
open-source
communities,
ingress
controller
will
build
on
top
of
envoi
agent
service
proxy,
and
we
are
basically
built
to
support
both
dynamic
configuration
updates,
who
work
really
well
with
multi
teams.
I
want
to
provide
English
delegation
as
well,
as
you
know,
making
sure
that
that
delegation
is
done
in
a
very
secure
way.
C
G
That
can
be
done
a
number
of
different
ways.
Typically,
you
have
a
load
balancer
in
front,
so
in
some
sort
of
cloud
environment,
but
again
in
a
different
type
of
environment.
Just
as
long
as
you
can
attract
traffic
this
this
will
work
functionally
well,
but
in
any
case,
traffic
hits
some
sort
of
load,
balancer
and
then
get
sent
to
envoy
and
then
that
riding
decision
happens
there
at
that
envoy
layer.
G
G
C
All
right
moving
on-
and
you
know
just
give
you
guys
a
little
bit
more
visibility
here,
reviewer
or
to
view
control
deployment
in
a
kubernetes
environment.
You
see
that
you
know
we
have
an
invoice.
Daemon
said
that
that
deploys
the
invoice
part,
and
then
we
also
have
a
control
deployment
that
basically
handles
everything
from
the
third-generation
jobs
to
the
secrets
are
mounted
into
the
pods
to
secure
the
the
communication.
C
C
All
right
so
from
a
project
overview,
we
start
control
started
at
HAP
deal
either
near
November
of
2017,
but
they'll
be
over
two
years
ago,
and
we
released
to
our
wonder
doe
release
in
November,
two
thousand
nineteen,
essentially
signaling
to
the
community
that
were
entering
a
stable,
backwards-compatible
API
release
of
contour.
From
an
implementation
standpoint,
we
have
five
plus
product
implementations
that
are
basically
some
our
commercial
products
like
the
ones
that
VMware
is
publishing
like
essential
PKS
and
a
couple
of
others
that
are
coming
down
the
line
right
now.
C
We've
just
completed
an
integration
so
that
key
native
kind
of
contour
is
an
ingress
controller.
4K
native
and
flagger
has
also
done
an
implementation
with
contour
to
provide
it
as
an
ingress
controller
within
their
product
portfolio.
We
don't
have
entirely
all
the
complete
statistics
in
terms
of
the
contributing
organizations
or
since
we're
not
in
sin
CF,
yet
we
don't
have
access
to
their
stats.
So
some
of
this
data
needs
to
be
gathered
manually
and
we're
working
on
basically
gathering
all
of
that,
but
you
have
over
100
plus
community
members.
C
Looking
at
some
of
our
high
level
project
statistics,
we
have
over
two
and
a
half
two
point:
1000
github
stars,
80,
plus
contributors
on
github,
with
319
Forks
and
close
to
2,000
clones,
and
the
clones
are
done
by
almost
200
individual
github
users.
We
have
four
main
trainers.
Two
of
them
are
here
on
the
call,
David
and
and
Steve
who
had
42
releases,
there's
no
clear
way
for
us
to
track
downloads.
C
Yet
because
of
the
fact
that
we
have
also
some
testing
engines
are
basically
pulling
images,
but
we're
not
working
to
getting
a
number
for
for
everyone
as
well,
who
have
480
slack
members
a
ton
of
slack
messages,
but
some
of
them
are
archived
in
it's
very
hard.
I
get
a
phone
number
over
500
to
the
followers
2,000,
who
meets
had
10
blogs
and
I'm,
trying
to
get
a
list
of
all
our
Cuba
contacts
that
we've
had
and
envoi
contacts.
C
But
it's
about
seven
in
the
last
couple
of
years,
who
had
one
point,
two
thousand
p
rs.8000
github
views
and
close
to
2000.
It
had
unique
visitors.
So,
as
you
see,
we
have
a
community
and
I'm
gonna
show
you
guys
we're
gonna
produce
some
charts
as
part
of
the
due
diligence
document.
That's
gonna
show
that
there's
a
stable
number
of
commits
are
happening
over
the
last
two
years
on
contours
or
the
both.
C
The
momentum
of
the
project
has
been
stable
and
increasing
over
time,
and
also
going
to
show
that
the
number
of
contributors
have
also
increased
over
that
same
time
span.
So
that's
to
show
you
that
the
project
is
in
good
health,
it's
alive
and
well
funded
by
different
organizations,
mainly
VMware,
and
it's
here
to
stay
any
questions
on
this.
If.
E
This
is
this
is
fantastic
by
the
way.
This
is,
if
there's
no
small
work
not
only
doing
this,
but
but
then
summer,
I
think
at
all.
Let
me
poke
around
a
couple
of
these
things,
not
because
the
answers
to
them
or
are
necessarily
gonna
matter,
one
way
or
the
next,
but
it's
just
a
good
point
of
note,
and
that
is
of
the
forks
and
the
clones
are
those
indicative
of
the
number
of
like
in
the
case
of
clones.
Is
that
sort
of
the
number
of
downloads?
If
you
will
is
that
how
the
project
installs
that.
C
That
could
be
an
indication
of
folks
that
just
want
to
basically
clone
the
project
and
look
at
the
source
code
viewed
in
their
own
repo,
maybe
make
some
more
changes
and
and
specifically
on
the
forks.
It's
that
we
we
know
at
least
of
one
or
two
organizations.
I
have
fought
contour
just
because
I
wanted
to
add
some
of
their
own
things,
and
then
they
have.
They
want
to
basically
push
them
upstream
later
on
some
of
those
organizations
don't
want
to
be
named
publicly
that
they're
either
using
or
contributing
to
contour.
C
E
C
Yeah,
so
sorry
for
other
projects.
The
way
these
done
is
that
if
they
have
Dockery
mate
on
docker
hub,
they
basically
grab
the
dollar
number
for
their
from
there.
They
post
that
I
don't
find
that
to
be
reliable,
which
is
why
I'm
not
doing
something
like
that.
Other
folks
have
a
Google
or
an
AWS
bucket
that
they
post
their
binaries
and
that's
how
people
download
them
and
they
post
that
in
general
downloads
is
if
you
can
find
the
true
number,
which
is
very
hard
if
an
indication
of
deployments.
C
But
it's
you
know
if
someone
tells
you
they
figure
the
golden
formula
for
these
are
likely
lying.
There
is
no
easy
way
to
get
that
we
don't.
We
don't
have
telemetry
that
it
pulls
back
to
us
how
many
folks
have
installed,
or
something
like
that.
So
since
that
doesn't
exist,
it's
very
hard
to
actually
pinpoint
a
number
we'll
give
you
guys
a
rough
estimate
on
the
number
of
downloads
of
our
binaries,
but
I,
don't
know
how
realistic
or
how
much
stock
you
can
take
into
that
I
mean.
C
Why
I
didn't
want
to
put
it?
I
was
trying
to
do
figure
out
a
better
way
to
provide
that
number,
but
in
absence
of
that,
will
eventually
go
with
docker
hub
and
I'll
I'll
put
a
caveat
there,
which
is
we
just
want
to
be
honest
here.
We're
like
I'll,
give
you
an
example
for
hardboard.
One
of
our
images
has
1
million
downloads,
I
know
for
sure
I
don't
have
1
million
deployments,
so
those
numbers
like
absolutely.
E
Telemetry
is
a
pervasive
problem
for
mobile.
You
know
usage
analytics
for
most
open
source
projects.
It's
a
it's
a
shame
that
it's
not
just
more
commonplace
of
the
four
maintainer
--zz
are
there:
how
can
you
speak
to
their
organizational
affiliation
and
then
their
focus
or
responsibilities
around
particular
components
of
of
contour
yeah.
D
D
Sorry
DevOps,
just
that
me
not
all
those
all
those
things.
I
spent
a
lot
of
time,
working
with
a
bunch
of
load,
balancers
I
can't
because
I
mainly
most
my
previous
roles
were
they
to
be
VC
type
type
of
e-commerce,
become
a
site,
so
lots
of
HTTP,
stuff
and
I've
worked
with
almost
all
of
the
almost
all
the
popular
open
source
web
servers
that
the
apache
nginx
like
a
Cherokee,
all
of
them
yeah
Michael
should
I
speak
for
the
country.
D
We
generally
work
on
individual
individual
feature
streams
because
were
remote
team
tends
to
break
down
better
that
way
we
have
at
the
moment.
James
is
not
on
the
call,
focusing
mainly
on
our
integration
tests.
We
are
working,
we
testing
is
insisting
is
always
challenged,
but
James
is
very
passionate
about
improving
in
doing
integration
tests
so
that
we
can
be
more
sure
about
when
we
get
new
releases
envoy
brought
rather
than
just
it
seems
to
work.
Can
we
really
be
very
very
sure
about
about
that?
It
works.
D
Steve
is
focused.
Our
locks
on
individual
features,
like
the
the
features
focused
towards
the
towards
the
customer
he's
been
working
on
things
like
upstream
features
right
looming
I
know,
is
one
that
he's
really
wanted
to
work
on
for
a
long
time
and
will
land
it.
Let
it
soon
and
Nick
is
the
kind
of
point
point
person
on
ingress
feature
is
what
key
close
with
the
sig
sig
services
oversight,
guys,
because
one
of
our
goals
is
that
pilot
part
of
the
thing
that
came
out
of
adding
our
HP
proxy
support
is
contour.
D
Has
the
ability
to
use
multiple
to
consume
multiple
ways
of
describing
ingress
since
the
IDS,
like
we're
not
closely
tied
to
any
one
ingress
document,
and
that
gives
us
the
ability
to
interact
new
ones.
So
as
we
get
girl
fingers
for
you
invader
to
English
the
one
I'm
gonna
aggressively
to
we
have
the
ability
to
integrate
those
at
a
real
reason.
The
low-cost
and
so
part
of
the
Clinton
we've
said
to
to
Bowie
Bowie
in
those
folks.
Is
we
want
to
be
your
first
user?
We
want
to
be
your
get
beat.
E
E
C
Absolutely
so
today
the
diversity
is
mostly
VMware,
but
there
were
a
very
open
and
welcoming
community
who
have
we
have
community
meetings
multiple
times
a
month
which
are
open
and
not
flexible
time
zones,
so
other
folks
can
come
in
and
contribute.
For
example,
much
more
from
from
the
key
Native
community
just
came
in
and
added
Kennedy
support
into
contour
in
a
matter
of
about
three
four
weeks
of
work,
just
engaging
with
our
community
and
getting
that
I've
been
running.
C
So
the
flagger
folks
did
the
integration
with
very
minimal
interaction
with
us,
but
we
are
welcome
more
maintainer,
see
if
other
folks
wanna
come
in
and
contribute
to
the
project
to
have
a
flexible
governance
to
enable
them
to
come
in
and
contribute
and
have
a
seat
at
the
table
and
I
want
to
mention
one
more
thing
before
we
leave
this
slide
on
the
Cuba
contacts
still
sloka
and
one
other
person
from
our
team.
The
presentation
like
Cuba
Conda
had
almost
10%
of
cubicle
attendees
signed.
C
D
Want
to
just
chime
in
on
contributions
like
this
is
something
that
I'm
very
person
very,
very
passionate
about
and
part
of
the
way.
All
the
way
to
do
that
is
like
I
have
a
very
strong
policy.
We
do
everything
and
github
we
try
and
keep
as
much
as
possible
in
the
open.
We
we're
trying,
sometimes
it's
very
easy
to
fall
back
to
old
habits,
but
actually
trying
to
have
like,
but
we
have.
We
have
a
contour
channel
inside
being
wet,
but
we
actually
try
and
use
the
public
one
on
the
kubernetes,
like
even.
F
D
Like
I
like
develop
a
chit
chat,
just
like
talking
back
and
forth
they're
like
what's
this
baggage
break
this,
do
you
see
these
things?
We
try
and
do
as
much
as
possible
in
the
open,
recruiting
contour
contributors,
it's
hard
keeping
them
is
about
10
times
harder
than
that.
So
everybody
that
we
can
make
feel
welcome
is
crucial.
C
All
right,
so,
let's
move
on
the
customer
profiles
really
quickly.
So
so
contour
is
being
used
both
in
production
pre-production,
staging
many
different
customers.
We
have
a
github
customer
testimonials
link
up
on
top
there's
only
one
testimonial
there.
Our
problem
is
a
lot
of
folks
know
if
customers
are
using
English
controls,
I.
C
Talk
about
what
they're,
using
for
the
front
door
of
their
kubernetes
clusters
publicly
to
have
a
major
financial
institution
that
has
basically
made
come
to
the
default
English
controller
for
all
the
communities
clusters
and
they
have
a
lot
of
them.
One
of
the
leading
online
marketplaces
that
uses
contour
in
production
today,
key
native
I
mentioned.
C
Provide
an
option
for
a
country
to
be
an
English
controller
in
the
product,
portfolio
Cillian
and
included
in
link
there
where
they
talk
about
contour
and
then
Flagler
also
has
an
implementation
of
contour
as
the
ingress
controller.
Furthermore,
Adobe
had
a
presentation
at
cubicle.
2019
was
a
landing
talk,
so
referencing
that
they're
big
users
of
contour
and
they
talked
about
the
architecture
and
how
they've
implemented
it
within
their
infrastructure,
so
they're
also
big
users
of
ours,
and
we
work
with
them
directly
and
open
up
for
questions
now
since
I
know
we're.
C
E
F
It's
something
that
we
talked
about,
I
think
it's
certainly
an
option.
I
I
think
from
my
perspective,
speaking
more
from
the
from
the
envoy
side
of
things,
is
it
it
would
be
complicated
both
in
the
sense
that
we
don't
have
any
process
yet
from
the
project
perspective,
to
take
a
project
like
this
in
we've
done
some
project
adoptions,
but
this
would
be
of
a
different
type
and
scale
than
before
so
we'd
have
to
develop.
You
know
all
those
procedures
and
to
be
totally
Frank
and
honest.
F
The
larger
problem
is
that
it
would
be
politically
quite
complicated,
just
because
contour
has
quite
a
few
competitors
and
then
all
who
use
envoy
and
then
you
know,
there's
some
question
of
what
that
would
look
like
within
within
the
envoy
or
in
terms
of
picking
say
a
default
and
gars
controller.
So
my
advice
to
the
contour
team
was
that
a
direct,
CN
CF
donation
makes
more
sense
right
now.
I
I
think
I
think
it's
just
simpler.
If,
in
the
future
you
know,
we
want
to
eventually
move
it
under
the
envoy
or
guy
I.
F
E
E
J
You
Emily
I
did
want
to
let
you
guys
know:
I
did
post,
it
has
been
updated.
It
was
like
a
class
that
day
it
was
three
months
ago.
There
is
a
due
diligence.
We
view
template
out
in
the
TOC
and
github,
and
so
it
does
kind
of
give
you
a
rough
idea
of
what
the
TOC
is
looking
for,
going
to
the
different
levels
and
I'm
sure
it's
going
to
change
as
the
SIG's
can't
looked
at,
but
at
least
there's
a
gives
you
a
rough
idea
of
what
was
being
done
at
the
time
for
intelligence
and.
C
J
I
Had
a
quick
one,
so
it's
probably
to
David
I,
don't
know
whoever
can
answer
it.
Please
answer,
of
course,
shortly
no
need
to
go
deep.
So
when
you
started
the
project
like
two
years
ago
or
a
little
bit
more
than
two
years
ago,
what
were
the
gaps
that
you
identified
that
are
out
there
in
the
open
source
landscape
in
this
domain?
And
how
would
you
compare
this
to
today's
state
of
this
landscape?
Are
there
any
new
projects
coming
up,
something
that's
overlaps
with
the
goals
that
we
had
back,
then?
D
They're
from
two
parts
to
this
question,
the
first
part
is
with
with
all
due
respect
to
Matt
two
years
ago.
The
kind
of
stand
out
here
that
Envoy
that
we
all
know
today
wasn't
as
clear
two
years
ago.
So
this
was
a
little
bit
experimental,
but
also
almost
immediately
as
I
started.
The
project
I
realized
how
good
a
fit
the
Envoy
XDS
api's
were
for
the
declarative,
the
kind
of
declarative
nature
of
the
kubernetes
api.
It
was
a
very
neat
fit.
A
A
The
surface
area
of
contours
API
as
it
were,
the
document
description
is
much
smaller,
partly
because
the
orig,
the
the
de-facto
ingress
specification,
has
evolved
over
many
years
that
you
can
see
it.
So
the
contour
project
has
stayed
really
true
to
keeping
things.
You
know
very
good
defaults
and
not
even
exposing
nods.
You
know
the,
for
instance,
ayah
always
uses.
A
E
K
L
L
Okay,
I'm
gonna
sue.
My
everybody
can
see
this,
so
let
me
give
the
high
level
of
what
SMI
is
all
about
first.
So
the
idea
is
that
service
nations
are
awesome
and
starting
to
really
proliferate,
especially
at
Keuka
in
San
Diego.
We
saw
kind
of
a
baby
of
us
about
thing
at
mesh,
console
Connect
coming
out,
sto
linker
D,
we're
all
got
really
great
innovation
in
this
space,
and
it's
introducing
some
interesting
constraints
on
users
and
implementers
and,
to
be
honest,
I
think
that
at
this
point
in
time,
we've
got
a
really
standard
feature
set.
L
So
the
idea
was
to
come
at
this
from
a
definition
perspective
and
talk
a
little
bit
about
how
we
can
produce
an
interface
that
all
of
the
service
specialists
can
interact
with
and
then
all
the
integrations
can
go
on
top
of.
So
let's
talk
a
little
bit
about
what
SMI
covers.
These
are
the
three
major
pieces
that
we
saw
from
users
that
they
wanted
from
a
service
mesh,
the
traffic
policy,
aka
access
control
and
identity
telemetry.
L
Those
are
the
golden
metrics
that
every
Ari
wants
and
traffic
management,
which
would
be
a
flagger
style
canary
rollouts,
on
more
complicated
solutions
there
and
we
really
focused
as
a
project
on
those
three
things
more
than
anything
else.
So
why
does
it
because
system
need
it?
Well
number
one
we're
starting
really
hard
to
be
provider.
Agnostic
benefits
on
both
sides.
Here
the
users
get
to
have
choice.
The
integrators
get
to
integrate
against
a
single
API
and
work
across
all
of
the
backends
and
the
service
mesh
implementers
don't
need
to
go
and
dream
up
new
api's.
L
They
can
go
and
use
what's
kind
of
been
suggested
in
as
best
practices
for
the
ecosystem,
so
that's
kind
of
where
we're
going
there.
This
is
a
better
picture,
I
think
kind
of
what
I'm
talking
about
there,
which
is
you've,
got
asked
to
an
ecosystem
on
the
top
of
the
service
missing
smashed
in
your
face
and
then
on.
The
bottom.
You've
got
all
of
the
really
great
service
measures
that
have
come
along
and
provided
fantastic
functionality
there.
L
So
going
back
to
those
three
use
cases
we
actually
have
technically
four
api's
today,
traffic
metrics
is
the
most
obvious
one
that
builds
off
the
policies
that
are
put
together
by
the
metrics
and
custom
metrics,
API
and
kubernetes,
and
then
adds
the
gold
metrics,
which
would
be
success.
Rate,
throughput
and
latency.
L
Traffic
split
is
the
ability
to
do
Canaries
with
a
poor
constraint
like
flagger
doing
the
canary
rollout
itself.
Then
we
have
traffic
specs
that
lets
you
explain
how
traffic
looks.
The
idea
here
is
that
these
are
requirement
of
access
control
so
that
you
can
do
per
route
access
patrol
as
well
as
doing
on
a
service
by
service
basis,
so
that's
kind
of
where
we
are
today.
L
Let's
talk
about
who
we've
done
this
with
I've
been
having
a
really
fantastic
time,
working
on
SMI
and
myself
because
of
how
many
great
partners
we
have
that's,
really
a
cross
industry
thing
and
we've
been
getting
a
lot
of
fantastic
feedback
from
pretty
much
everybody.
Both
service
providers
and
folks,
like
you,
cos
who
are
building
solutions
on
top
of
service
meshes
for
their
users.
L
So
all
of
this
leads
into
goals
and
non
goals.
The
primary
goals
are
being
agnostic,
making
sure
that
it's
vendor-neutral
and
solving
real
world
problems
for
users
both
where
users
are
end
users,
as
well
as
implementers
on
top
from
an
ecosystem
perspective
and
the
service
measures
themselves.
An
on
goal
is
to
implement
a
service
match
I,
don't
think
any
of
us
want
to
do
that,
we're
already
building
those
as
projects.
L
We
don't
require
implementation
of
specific
SMI
api's,
so
the
important
part
here
is
that
we're
not
being
prescriptive.
If
a
service
mesh
wants
to
just
support
traffic
splits,
that's
all
they
need
to
do.
In
fact,
we
have
a
long
conversation
that
lis
is
a
partner
in
SMI
about
compliance
and
providing
users
visibility
into
what
is
and
isn't
supported,
so
that
it's
a
incremental,
slow
thing.
Instead
of
this
big
bang
requirement
that
everybody
needs
to
implement,
but
also
we
don't
want
to
restrict
what
it
is
to
be
a
service
mesh.
L
L
Cool,
so
a
quick
technical
overview
is
kind
of
three
parts.
We've
got
some
kubernetes
series.
This
goes
back
to
SMI,
being
kubernetes
centric,
while
we're
having
conversations
about
how
to
bring
in
the
rest
of
the
real
world.
We're
sticking
really
hard
to
our
guns
on
kubernetes
being
the
one
true
way
moving
forward,
we
have,
unless
my
provider
to
opt
on
the
api's,
there's
a
go
SDK
for
folks
to
use,
there's
extension
api's
to
build
on
top
of
things,
and
the
resources
are
obviously
configurable.
L
The
thing
that
I
like
to
bring
up
the
most,
however,
is
that
SMI
isn't
actually
just
spec
project.
We
actually
have
quite
a
bit
of
software
and
components
in
it.
We've
got
the
SMI
metrics
extension
API
server
that
actually
works
for
SEO
and
League
review.
Today,
we're
moving
in
a
container
of
functionality
into
SMI.
It's
a
common
pattern
that
all
of
the
service
meshes
are
adopting
at
this
point
and
really
should
just
share
the
implementation
and
then
the
whole
point
behind
that
is
that
we
can
go
innovate
on
functionality.
L
That's
unique
and
interesting
to
our
implementations.
Instead
of
just
doing
the
same
patterns
over
and
over
again,
my
selfish
goal
here
is
that
once
we
start
having
these
common
patterns
of
software
we'll
be
able
to
go
and
get
those
patterns
to
be
smoother,
for
example,
sidecars
being
a
first
class
citizen
inside
of
communities.
L
So
here's
links
to
the
community
and
related
repositories
website
github
we're
doing
meetings
there.
All
public
they've
been
fantastic
for
figuring
out
where
everybody's
going
there
and
I
think
that's
it.
Let's
see
so
benefits
of
CNC
a
conclusion.
The
biggest
one
for
me
is
the
association
with
kubernetes
and
other
CNCs
projects.
L
The
other
big
part
of
this
is
that,
as
part
of
being
vendor
neutral,
we're
able
to
go
and
get
more
community
contributors
and
speed
up
the
adoption
of
both
API
spec
itself
and
software
components.
Inside
of
that
and
then
finally,
CN
CF
is
the
elephant
in
the
room
when
it
comes
to
cloud
native
and
kubernetes
in
particular,
since
we've
pitched
our
pitch
our
horse
to
that
part,
and
so
that's
all
a
big
part
of
that
dress.
L
L
L
We
want
it
to
be
simple,
so
that
users
can
understand
it
and
we
can
go
and
provide
end-user
benefits
as
well
as
implementers,
and
we
want
to
be
as
ecosystem
friendly
as
possible
with
me,
call-to-action
being
at
the
end
of
this
meeting
will
have
the
PRI
and
would
love
all
of
the
discussion
and
any
questions
you
have
that
we
don't
answer
here
to
go
into
that.
Pr.
E
Wonderful,
this
is
gay,
okay,
just
a
quick
time
chefs.
We
do
have
a
fair
bit
of
time
for
some
questions,
maybe
the
first
one
that
will
help
drive
some
additional
questions
or
or
how
or
how
much
homework
or
how
deep
to
dig.
If
you
will
is
it
well,
as
maybe
a
question
to
the
SMI
team
do
are?
Is
there
a
consideration
around
entry
into
sandbox
or
entry
and
as
an
incubation
project.
E
K
I'd
love
to
take
it.
Thank
you.
I'll
catch
the
ball
yeah.
We
had
originally
reviewed
the
graduation
criterias
posted
on
the
CN
CF
github,
and
we
felt
that
we
could
be
considered
for
incubator,
but
obviously
were
at
the
you
know,
coming
under
the
auspices
of
whatever
the
CN
CF,
cig,
Network
and
TOC
decides
either
sandbox
or
incubation.
G
K
F
F
So
I've
actually
like
to
talk
about
that
point,
but
but
just
because
I
think
of
those
general
concerns
and
just
how
early
it
is
I
guess
my
advice
would
be
I.
Think
sandbox
will
be
completely
non-controversial.
I
think
incubation
may
be
more
controversial
it.
It
might
happen
like
I'm,
not
sure
how
people
would
vote
or
what
people
would
think.
But
I
do
think
that
just
given
what
has
happened
particularly
again
with
kubernetes
ingress
over
the
last
several
years,
there's
probably
going
to
be
some
reservations
about
this
type
of
type
of
spec.
M
N
H
M
So
you
know
I
I
had
I
recommended
incubation.
Looking
at
you
know
the
adoption
in
the
last
several
months
and
how
there
has
been
several
meshes
just
adopt
SMI
in
production,
but
I
totally
agree
with
you
Matt
in
that
you
know,
sandboxes
is
non-controversial
and
I.
Don't
think
that
the
team
really
cares
one
way
or
another
I
think
that
what
we're
really
after
is
that
vendor-neutral
home
we
want
to
have
you
know
a
common
touch
point
to
talk
to
all
of
the
meshes
and
the
ingress
be
to
people
and
any
other
tooling
in
this
space.
E
Just
maybe
just
yeah
my
council
would
be
it
would
be
the
same
in
you
know
in
part,
adding
to
what
each
of
you
said
also
in
part
based
on
I.
Think
well,
maybe
before
I
say
a
couple
of
other
things.
I
would
say
that
there's
another
question
to
be
asked
and
that
would
also
bear
weight
on
some
of
the
particulars
of
the
requirements
around
the
different
levels.
E
So
the
question
is
and
I'm
in
most
respects
asking
number
these
questions
for
the
fourth
being
very
familiar
with
SMI
myself,
but
but
kind
of
for
the
public
record
here
and
for
the
going
through
the
process,
but
is
well.
Let
me
be
long-winded
about
this
and
say
there
is
some
prior
presses
system.
So
it
has
long
been
said
that
the
CNC
F
is
not
a
standards
body.
It
was
an
intended
necessarily
produce
internet
standards
per
se.
E
That
said,
the
difference
between
and
the
value
between
standard
and
a
specification,
adoption
and
dotting,
T's
and
crossing
eyes,
and
he's
saying
we
could
have
a
long
conversation,
but
all
of
these
things
rather
the
there
is
some
precedent
for
it
spent
a
lot
of
time.
Inside
of
the
service
working
group,
as
did
Ken,
knew
who's
on
the
call
as
well
and
helping
navigate
that
fine
line
successfully.
E
K
Mean
I
think
that's
hard
very
difficult
to
answer
if
it
if
it
appears
to
go
towards
standards,
sure
you
know
if
we
tee
it
up
that
way,
but
you
know
as
a
specification
and
providing
value
in
this
ecosystem.
The
CNC
F
ecosystem,
specifically
I,
think
we
want
to
at
least
focus
on
that,
and
you
know
I've
heard
similar
things
about
specifications.
We
got
cloud
events
we've
got
tough,
which
has
recently
got
even
graduated
I
think
having
more
specifications
in
the
CNCs.
K
Well,
you
know
if
the
TOC
is
favorable
to
that
behavior
I
think
you
know,
it'll
only
make
it
better
for
subsequent
specifications
to
find
a
good
home
in
the
CNC
F
I
do
want
to
quickly
address.
One
thing
that
Matt
brought
up
was
the
ingress
lowest-common-denominator
value
proposition
here
and
I.
Think
it's
interesting
because
ingress
specifically
and
even
off
the
back
of
contour,
you
know
those
kind
of
abstractions
were
very
early
ingress
specifically
and
if
you
go
and
take
a
look
at
ingress
v2.
K
Actually
it
takes
a
lot
of
what
contour
did
and
some
of
what
SMI
did
in
terms
of
extensibility.
That
is
not
driven
by
annotations,
which
was
one
of
the
big
gripes
for
ingress
v1,
and
with
that
extensibility
kind
of
framework,
you
can
actually
have
deep
integration
and
provide
more
value
other
than
lowest-common-denominator
and
I.
Think
the
other
areas
that
I
would
draw
to
example,
are
CSI
CN,
I
CRI
C,
star,
I
I
will
just
say:
C
star
I
interfaces
across
the
CNC
F
ecosystem
have
allowed
this
same
behavior
to
exist
and
have
X
identity.
K
Bility
points
that
provide
useful
because
I
think
storage
is
infinitely
more
complex
than
service
mesh
and
CSI
is
still
providing
value
as
an
abstraction
to
both
you
and
storage
implementers
in
the
ecosystem.
So
you
know,
ingress
I
think
is
one:
that's
people
have
a
rough
trot,
maybe
a
bad
taste
in
their
mouth,
but
I
think
ingress.
V2
is
kind
of
bringing
a
non
least
common
denominator
and
extensibility
points
which
allow
a
value
to
exist.
K
L
From
the
spec
perspective
itself
spend
a
lot
of
time,
adding
the
extensibility
points
from
my
perspective,
as
someone
who
works
on
the
could
be
on
a
regular
basis.
I
don't
want
to
be
how
all
of
my
flexibility
taken
away
to
go
and
innovate
and
do
new
and
interesting
things,
and
so
taking
a
look
at
the
ingress
spec
and
what's
happened,
and
some
of
the
other
specs
like
CNI
I,
think,
is
a
great
example
of
something
that
came
along
and
really
just
blew
up
all
kinds
of
really
fantastic
innovation
early
on
in
the
community
space.
L
E
So
so
good,
so
if
it
isn't
and
I'm
trying
not
to
make
it
too
obvious,
how
excited
I
am
that
we're
having
assess
of
my
discussion
so
so
that
this
is
fantastic,
so
good,
I
think
we
we
characterize
kind
of
standard
and
spec.
You
know
there
have
been
a
question
earlier
about
the
common
use
of
the
patchy
b2,
as
as
a
license
for
many
of
them.
E
You
they're
the
most
common
license
used
for
projects
that
enter
into
the
CN,
CF
and
sort
of
stated
preference
for
that
license
and
I
thought
I
would
say
on
the
call
here
that
of
the
the
open
web
foundation
agreement,
that's
used
as
a
license
for
SMI
that,
to
my
knowledge
there
is,
there
won't
be
contention
as
a
consideration
for
a
project
that
comes
into
the
scene.
Cf
the
OCI
uses
that
license
as
well.
E
I
understand
that
for
you
that
there's
a
choice
around
that
license,
as
it
has
different
implicit
implications
around
patents,
but
that
the
that
that's
a
license
friendly
to
the
CN
CF,
or
that
the
CF
is
friendly
to
that
license.
Good
good.
To
have
that
you
confirmed
and
said
some
Michele
I
think
you
were
following
up
on
that,
but
I
don't
anticipate
an
issue
there.
M
We
do
need
to
go
back
and
talk
to
the
ciencia
legal
team
and
staff
about
potentially
seeing
if
using
the
is
it
the
open
web
foundations
places
for
for
specs,
whereas,
like
all
the
other
projects
that
are
actual
code,
can
remain
Apache,
BT
and
I.
Think
there's
any
issue
with
that.
So
then
there
is
also
a
legal
committee
within
the
governing
board
that
we
may
also
want
to
run
this
pass.
E
Awesome
what
one
item,
if,
if
it's
easy
enough
for
you
guys
to
glean,
is
a
so
we
talked
about
I,
think
you
presented
on
community
some
stats.
If
you
would,
if
you
have
contributor
and
kind
of
maintainer
stats
and
whether
those
are
specific
to
the
four
api's
that
are
there
or
just
as
a
collective
for
the
for
the
project,
those
would
be
be
good
to
see
those
numbers.
K
K
I
E
I,
don't
disagree:
I,
yes,
I'm
somewhat
torn
between
wearing
my
my
signet
work,
my
in
my
hat
and
the
there
is
so
just
to
clarify
Michael.
If
he's
still
on
call
and
the
contour
team
of
the
template
that
I
was
referring
to
earlier.
It's
really
a
template
that
matches
up
with
the
project
proposal
process.
So
there's
a
1.2
and
in
there
it
calls
for
stats
around
community
size
and
existing
sponsorship.
Social
media
accounts
release
methodology,
mechanics
website
versions,
issue
tracking
bubble.
It
just
kind
of
goes
about.
C
E
Is
clear
to
me?
Actually,
I
was
just
calling
back
to
you
as
reference
to
for
the
SMI
team
as
well
that
it's
that
I
will
send
out.
So
can
it
send
out
a
link
to
due
diligence
which
applies
to
incubation
and
above,
but
for
sandbox
and
below
there's
also
just
kind
of
a
standard
set
of
statistics
that
the
contour
team
that
you,
you
guys
covered
well,
you've
you've
got
there,
but
feedback
for
the
SMI
team
is
to
try
to
incorporate,
incorporate
those
as
well
we'll
pop
over
a
link.