►
From YouTube: Contour Community Meeting - August 18, 2020
Description
August 18, 2020
News
Roadmap update PR. tl;dr we are pushing the roadmap out for two months while we refactor to allow us to deliver the features in a more sustainable way.
Kubecon Meet the Maintainer Sessions:
Wednesday, August 19 • 17:00 - 18:00
Thursday, August 20 • 20:00 - 21:00
Kubecon Contour BoF - Thursday, August 20 • 13:00 - 13:35
Office Hours:
Overview Page: https://github.com/projectcontour/community/wiki/Office-Hours
Cal Link: https://calendar.google.com/event?action=TEMPLATE&tmeid=Nm9qN2lrMm1lcnNxc29pMWlyYWoxY2s3YjFfMjAyMDA4MjBUMTcwMDAwWiBzdGV2ZUBzdGV2ZXNsb2thLmNvbQ&tmsrc=steve%40stevesloka.com&scp=ALL
Discussion
Setting up local development environment [Chad Cravens]
Contour architecture (envoy API, go-control-plane, dag, etc.) [Chad Cravens]
Issue 2287, Service APIs implementation [Daneyon Hansen]. How to proceed with the following:
List of issues that block doing some implementation of the Service APIs behavior.
List of issues that need to be implemented to actually deliver the v1apha1 Service APIs behavior.
Issue 2187, add an operator for Contour. Cluster-addons is a k8s-sigs project for managing addons (e.g. CoreDNS) using operators [Daneyon Hansen].
A
A
A
A
So
I've
got
apr
here
with
the
oh.
This
is
the
old
one
I
had
I
needed
to
reopen.
It
changed
the
main
branch
to
maine,
but
yes,
I've
opened
a
pr
here,
just
to
tell
everybody.
I've
pushed
out
the
the
delivery
dates
for
a
number
of
a
number
of
deliverables,
because
we're
spending
some
time
refactoring
in
order
to
actually
be
able
to
do
some
of
this
stuff,
steve
raised
a
good
point
that
that
xds
upgrade
has
to
be
finished
by
december.
A
So
I
put
it
ahead
of
some
of
the
other,
some
of
the
other
things
one
of
the
this
is
one
of
the
big
things
that
we've
needed
to
do
a
bunch
of
refactoring
to
be
able
to
do
right
now.
The
xds
v2
stuff
is
yeah
sort
of
has
tendrils
very
deep
into
contour
and
part
of
what
we're
doing
with
the
refactoring
is
trying
to
drop
some
api
surface
around
some
of
that
stuff.
A
B
B
Proxy
objects,
tcp
proxies
all
those
different
things
process
through
them,
come
up
with
validation
rules,
meaning
is
it
valid?
Is
it
not
valid?
Does
it
match
the
service
and
kubernetes
do
all
of
that
like
business
logic,
and
then
that
gets
built
into
the
dag
and
then
once
you
have
that
that
graph,
you
can
walk
that
graph
and
come
up
with
envoy
configuration
one
of
the
things
problems
we
have
today
is
that
that
is
all
baked
into
one
single
file
called
builder.go,
which
is
super
difficult
to
maintain,
as
you
add
more
things.
B
So
as
we
look
to
add
service
apis,
it
gets
just
more
and
more
complex
and
more
fluff
in
there.
So
I'm
working
on
trying
to
break
that
apart
into
pieces,
so
we
can
make
it
more
componentized,
so
we
can
have
ingress,
be
a
section.
You
know
once
the
ingress
v1
gets
support.
We
can
put
that
in
you
know,
proxy
will
have
one
and
then
the
new
service
apis
will
have
another
another.
B
C
C
C
Which
is
kind
of
like
a
http,
a
envoy,
http
layer
thing
then
that
internal
and
then
it
has
a
one-to-one
relationship
between
the
http
layer
concept
of
a
service
and
the
way
it
collects
endpoints.
So
we
need
to
break
that
because
for
auth
we
need
a
envoy
cluster
which
can
be
backed
by
more
than
one
kubernetes
service.
So
we
need
to
kind
of
break
this
implicit
one-to-one
relationship
between
the
way
endpoints
are
constructed
and
the
way
the
internal
service
model
works.
C
So
this
is
a
big
refactor
and
basically
a
rewrite
of
the
way
we
handle
endpoints.
So
one
of
the
benefits
is
that
it's
actually
all
explicit
now,
like
you,
can
follow
the
code
directly
by
by
reading
code.
You
don't
have,
it
doesn't
depend
on
the
kind
of
special
generation,
special
creation
of
cluster
load
assignment
names,
which
everything
kind
of
depended
on
before.
A
D
Yeah,
I
guess
what
I'm
looking
for
is
help
from
the
community
or
collaboration
with
the
community
on
how
we
can
start
listing
out
what
needs
to
be
done
in
preparation
for
adding
service
apis
and
then,
as
I
follow
on
to
that
or
whatever
we
know
exists.
You
know
what
work
needs
to
be
done
to
add
service
apis
support,
and
so
that's
essentially
what
those
two
sub
bullets
are.
B
Yeah,
I
think
so
I
think
yeah,
adding
splitting
apart
the
builder
and
making
it
more
more
flexible,
we'll
make
it
more
straightforward
to
add
in
those
those
bits
so
you'll
know
you
know,
there'll
be
an
ap.
An
interface
you'll
have
to
implement,
which
right
now
is
very
simple,
but
you'll
have
to
you
know,
to
add
those
bits
in
there.
So.
C
Think
the
way
I
see
what
steve's
working
on
is
that
you'll
end
up
being
able
to
have
like
a
ingress
builder,
a
http
proxy
builder,
a
service
apis
builder
and
or
maybe
service
apis
will
have
multiple
builders,
so
you
can
kind
of
configure
them
at
the
top
level.
That's
that's
kind
of
where
you're
aiming
at.
Isn't
it
steve.
B
What
we'll
have
to
decide
on
is
how
we
reconcile
those
together,
because
today
there's
a
notion
of
like
http
proxy
objects,
take
precedence
over
ingress
objects.
So
if
we
want
to
have
that
hierarchy
or
not
in
contour
or
if
contour
should
handle
things
separately,
meaning
if
you
had
ingress
objects
sitting
alongside
service
apis
objects,
what
should
happen?
I
guess:
should
you
process
both?
Should
they
get
merged
or
not?
Some
of
those
kind
of
rules
we'll
have
to
think
about,
but
yeah,
absolutely
yeah.
A
A
That's
a
serial
process
where
you
know
it
goes
through
and
does
x
object,
then
y
object
and
z
object
and
it
runs
through
and
does
the
ingress
objects?
Let's
see
it
grows
with
a
capital
I
and
then
it
does
http
proxy
afterwards.
So
if
it
finds
the
same
config
from
h2
like
the
same
v
host
from
http
proxy,
the
hp
proxy
config
will
overwrite
whatever,
wherever
it
has
inc,
whatever
it
has
in
common
with
ingress.
A
E
A
The
way
contour
has
you
know:
one
contour
is
equals
one
english
class,
and
so
you
we
need
to
make
the
decision
whether
or
not
we're
going
to
keep
that
behavior
or
if
we're
going
to
do
something
more
clever,
a
kind
of
lean
towards
keeping
the
thing
that
one
contoured
deployment
equals
one
ingress
class,
regardless
of
how
the
ingross
class
or
gateway
class
is
configured.
There
are
some
tricky
semantics
about
how
you,
if
you've
got
one
contour
looking
for
an
ingress
class.
A
A
C
Yeah
config
is
still,
I
think,
is
a
thing
too.
So
when
we
right
you
know
right
now
we
have
the
config
map,
which
is
bound
to
the
contour
deployment,
but
in
a
ingress
class
gateway
class
world
we
can
bind
the
config
to
the
class
well
that
at
least
that's
how
that's
sort
of
supposed
to
work
yeah.
So
I
think
so
I
think
there's
a
lot
of
things
to
think
about
about.
I
think
we
need
to
kind
of
step
back
and
think
about
with
these
things
about
what
is
the
right
deployment
model
like.
C
A
So
yeah
so
daniel,
that's
probably
that's
probably
a
pretty
hard
blocker
to
actually
doing
much
to
do
encoding,
because
the
first
thing
we're
going
to
need
to
implement
is,
you
know,
gateway,
classes
and
and
we're
kind
of
stuck
on
implementing
the
egress
v1
ingress
classes
at
the
moment
because
of
this
exact
problem,
so
I
don't
steve,
do
we?
I
know
you
were
working
on
the
ingress
cosby
one.
There
is
an
issue
for
the
v1
ingress
class
right.
B
A
Will
go
hunting
for
those
before
I
go
back
to
bed
today
and
you
know
link
link
them
into
the
2287
so
that
we
can
try
and
get
some
discussion
happening
there.
I
think
the
key
thing
here
in
my
mind
is
you
know
the
service
apis
and
ingress
are
sort
of
specifically
built
to
allow
you
to
do
this
sort
of
config.
D
Right
so
the
data
plane
in
open
shift
4x,
which
that
data
plane
is
h
a
proxy,
so
that's
deployed
and
fully
managed
by
an
operator
which
kind
of
gets
me
to
the
last
bullet
point
of
today's
discussion.
If
we
have
time
and
happy
you
know,
I
think
mikhail's
on
a
call
here
too,
so
we're
happy
that
there's
time
to
spare,
we
could
jump
into
all
sorts
of
details
on
that.
D
D
And
you
know
there
was
like
three
or
four
different
things
that
were
rattled
off,
and
so,
if
we
can
just
get
links
to
whatever
those
issues
or
prs
that
everyone
was
talking
about,
that
would
be
really
helpful
so
that
we
can
at
least
just
go
to
2287
and
start
seeing.
Okay,
here's
all
the
stuff
that's
in
flight,
or
that
needs
to
be
addressed
because
I
you
know,
I
want
to
start
like
looking
into
that
stuff
and
helping
me
come
up
to
up
to
speed
and
then
hopefully,
here
in
the
near
future.
D
I
feel
comfortable
comfortable
enough
that
I
could
start
helping
out.
But
at
this
point
I
just
feel
like
you
know,
I
started
looking
at
what
was
it
something
adopt?
A
controller
runtime
cache
I
started
looking
at
that.
I
still
need
to
go
back.
Look
at
it,
some
more,
but
I
guess
long
story
short.
If
we
could
just
you
know,
for
the
time
being
use
2287,
you
know
all
the
knowledge
that
you
guys
have
just
start
putting
whatever
you
think
needs
to
be
addressed
in
preparation
for
service
apis.
A
Okay,
I
think
that's
an
entirely
reasonable
request.
I
will
put
as
much
as
I
can
think
of
stephen
james,
please,
please
add
in
your
own
ones
as
well,
and
then
that
way
that
way,
hopefully
in
a
day
or
two,
we
can
get
that
get
that
you
know
and,
as
I
said
I'll
create
another
issue
to
for
us
to
be
able
to
talk
out
the
the
larger
issue
about
the
the
various
class
objects
on
both
english
class
and
gateway
class.
A
Well,
yeah,
I
mean
the
more
we
can
unblock
you
so
that
you
can
get
started
on.
You
know
on
doing
that,
the
better,
I
think,
that's
one
of
the
reasons
why
we
decided
to
sort
of
stop
a
lot
of
the
feature.
Work
and
start
start
do
do
a
bunch
of
refactoring.
Is
that
you
know
all
of
us
were
finding
that
we
were
stuck
and
not
able
to
easily
hand
off
any
of
this
stuff,
because,
because
of
everything
had
fingers
into
everything
else,.
D
Yeah-
and
I
think
it's
a
great
opportunity
to
you
know,
get
things
right
before
adding
more
complexity
with
service
apis
and-
and
you
know
it's
just
a
useful
tool
for
me
to
start
exploring
the
code
and
not
just
you
know,
for
contour,
but
you
know
I
need
to
brush
off
a
lot
of
dust
with
envoy
and
the
go
control.
Plane
is
a
new
project
for
me,
so
I
spent
some
time
looking
at
that,
and
so
my
intent
is
to
spend
a
lot
more
time.
D
You
know
and
hopefully
come
up
to
speed.
So
I
can
you
know
when
we
get
to
the
point
of
adding
service
api
support
to
contour.
You
know
myself
and
potentially
others
from
the
red
hat
side.
Are
you
know
ready
to
just
you
know,
we've
we
already
have
that
baseline
knowledge
and
we're
ready
to
hit
the
ground
running
and
really
start
to
help
out
the
community.
C
D
Well
then,
this
the
work
that
steve's
doing
I'm
very
interested
in
that.
So
if
we
get
that
linked
into
287,
you
know
here
very
soon,
I'm
going
to
start
diving
into
it,
not
to
say
that
I'm
you
know
going
to
be
able
to
help
there.
But
again,
just
a
lot
of
this
is
going
to
be
just
you
know,
kind
of
bootstrapping
myself
and
some
of
the
others
from
the
red
hat
team.
B
D
Because
the
other
part
that
that
that
I'm
doing
is
just
like
putting
together
like
a
plan
on
how
myself
and
you
know,
others
from
red
hat-
can
start
contributing
and
so
of
course,
there's
a
focus
around
the
service
api
stuff.
But
you
know
the
plan
is
here
in
the
near
future
that
we're
starting
to
contribute
just
okay.
A
Sorry,
I'm
just,
I
think
I
think
chad
chan
got
the
wrong
time,
so
I
think
just
finding
out
from
him
whether
or
not
he
wants
to
come
and
go
through
instinct.
I
think
it
might
be
better
to
do
it
next
week.
Yeah
sorry
about
that
yeah
next,
yeah.
Sorry,
yeah,
no
problem
yeah!
So
do
you
so
it
looks
like
you
want
to
sort
of
go
through
setting
up
local
dev
environment
and
talking
through
some
of
the
architecture
stuff.
F
Yeah
yeah,
so
I've
got
me
and
two
other
guys
that
are
kind
of
looking
at
the
contour
source
code.
You
know
we
because
we
have
such
a
dependency
right
on
the
successful.
You
know
work
of
the
proxy
not
only
from
a
couple
things
that
we
want
to
do
is
we
want
to
help.
You
know
I
submitted
that
issue
about
the
header
modification
and
you
know
discuss
that
with
the
team
really
understand
the
architecture
we
already
started
kind
of
playing
with
it
in
our
own
branch.
F
You
know
like
working
with
the
policies
like
those
header
policies.
Looking
at
that
integration
with
envoy,
you
know
better
understanding
kind
of
how
that
works.
Learning
about
like
the
go
control
plane,
the
envoy
v3
stuff.
I
know
that
I
think
some
recent
merges
have
occurred
or
I
think
a
lot
of
work
is,
you
know,
being
done
there,
understanding
the
dag
so
that
we
can,
you
know,
help
get
some
of
these
things
past
the
finish
line.
A
You
really
would
appreciate
the
help,
but
also
we're
also
coming
from
a
place
where
we
haven't
had
a
lot
of
people
working
on
contour
before
so
we're
kind
of
needing
to
build
the
structures
to
enable
like
non-maintainers
to
work
on
it
easily.
We
haven't
been
good
at
that
in
the
past.
I
think
and
yeah
I
mean.
Obviously
I
want
to
get
better,
but
you
know
we've
got
to
yeah.
A
That's
what
part
of
that's
the
other
reason
for
the
refactoring
is
to
make
it
so
that
it's
easier,
you
don't
need
to
learn
how
all
of
contour
works
in
order
to
be
able
to
change
part
of
it
so
yeah,
I
think
yeah.
I
feel
like
going
over
setting
over
a
local,
dev
environment
and
sort
of
an
architecture
review.
Those
are
actually
great
great
topics,
steve
for
maybe
for
office.
B
Hours
sometime,
as
I
said,
there's
one
scheduled
for
this
thursday,
but
I
wasn't
sure
if
we
were
going
to
keep
it
or
not
do
a
coupon
and
it
overlaps
with
the
contour
office
hours
for
kubecon
the
second
hour
of
it.
So
I
was
thinking
we
might
cancel
that
and
just
do
the
kubecon
once
it's
such
a
crazy
time.
A
Yeah,
I
think
that's
probably
a
good
idea
that
we
that
we
do
the
contra
office
hours
and
maybe
what
maybe
we
can
do.
Maybe
we
can
just
start
the
contra
office
hours
you
and
ask
for
questions,
but
then
you
know,
if
no
one
has
many
questions,
then
we
can
start
doing
an
architecture,
review
and
sort
of
talk
through
how
how
the
different
bits
of
contour
fit
together
to
actually
end
up.
Programming.
F
A
And
so
I
would
also
say
anyone
in
the
community
who's
sort
of
interested
in
contributing
more.
The
other
thing
that
we
that
you
can
do
is
please
feel
free
to
like
on
the
pr's
that
us
maintainers
are
currently
sort
of
talking
about
how
to
talk
amongst
ourselves.
About
like,
please
feel
free
to
come
in
and
ask
questions
there.
You
know
like
if
we
are
having
a
discussion
about
something,
please
feel
free
to
say
hey.
You
know
why.
Why
did
you
do
this
or
you
know,
or
that
sort
of
thing?
A
If
it's
long
involved,
we
can
take
it
to
a
zoom.
We
can
that
sort
of
thing,
but
you
I
mean
we're
all
we're
refactoring
a
bunch
of
this
stuff
already,
so
you
know
and
you're
looking
at
those
pr's
is
probably
a
good
way
to
understand
a
bit
of
like
you
know
how
things
are
going
to
end
up
what
we're
trying
to
do
and
yeah.
So
please
feel
free
to
just
jump
in
there
and
ask
questions
and
and
that
sort
of
stuff
on
those
pr's.
A
C
F
E
E
A
I
think
that
might
be
a
good
idea
to
start
the
to
start
the
the
normal
contour
one
earlier.
I
mean
that
way
you
that
way.
We
can
go
a
bit
more
in
depth
there
and
keep
it
slightly.
We
can
probably
keep
say
either
do
the
local
dev
environment
one
first
steve
and
then
do
the
sort
of
the
broader
architecture
review
as
part
of
the
q
com.
One.
E
A
Yeah
I
mean
I
I
specifically
didn't,
because
I
wanted
to
be
able
to
have
to
tell
people
some
of
the
news
to
have
the
thing
to
be
able
to
tell
people
some
of
the
news,
but
yeah
I
mean
normally.
I
agree.
We
should
probably
cancel
the
meetings
during
coupon.
Okay.
Well,
we
are
out
of
time.
Is
there?
Is
there
anything
else
anyone
wants
to
quickly
talk
through
chad?
I'm
really
sorry
about
the
time.
F
A
I
think
we
should
we
should
definitely
still
plan
for
an
architecture
review
in
the
there.
So
there
is
a
there's,
a
contour
office
hours
session
as
part
of
kubecon,
oh
okay,
and
that
times
that
time
slot
actually
overlaps
with
the
normal
time
slot
we
have
for
the
contour
office
hours
so
steve.
Could
you
put
the
times
for
the
do
you?
Are
you
okay,
with
running
the
normal
one,
or
do
you
want
to
do
a
coupon
stuff.
B
A
I
really
appreciate
yeah
keep
pushing
us,
please
it's
easy
for
us
to
get
lost
in
the
shuffle
and
forget
to
do
things
so
really
appreciate
all
that
and
chad
thanks
for
thanks
for
being
excited,
it's
really
cool
too
exciting
to
have
some
people
to
have
some
people
coming
and
being
like.
Please,
we
want
to
help.
What
can
we
get?
There's
me
and
two
others.
So
we
got,
we
got
three
go
developers
heaps
people
awesome
so
yeah
we
really
gotta.
A
We
really
gotta
pull
our
fingers
out
and
finish
making
it
so
that
people
can
actually
contribute,
and
also
there
are
some
getting
started
docs
on
in
the
docs
website.
If
you
can
have
a
look
through
those
and
tell
us
where
they're
falling
down,
you
only
get
a
chance
to
get
started
with
the
project
once
so.
Anything
you
can
do
to
help
with
those
docs
would
be
absolutely
nice.
Absolutely.