►
From YouTube: Envoy Community Meeting 2017-11-07
Description
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.
A
A
D
Yeah
I
mean
just
magically
the
zoom
things
needs
you
pass
our
security
clearance
and
so
I
can
actually
use
it
now,
but
oh
fantastic
yeah
progress'
check
fixing
up
in
case
anyone
else.
Oh
yeah
no
worries.
C
C
B
B
Working
to
see
if
we
could
get
a
better
deal
there,
their
prices
are
slightly
exorbitant.
B
D
B
C
Yeah
I
mean
the
only
thing
that
they
mentioned,
which
I
can
believe
is
that
the
minute
that
you
put
the
bounty
out
there
publicly,
we
might
be
inundated
with
with
reports
that
are
not
very
interesting.
So
I
just
don't
know,
I
mean
it's
the
kind
of
thing
where
we
could
try
it
and
just
see,
and
then,
if
it's
bad,
we
could
just
pull
it.
But
I
do
tend
to
think
that
a
bounty
is
probably
the
best
way
that
we
were
gonna,
find
real
issues.
F
D
We
want,
from
the
envoy
perspective
and
I,
gave
a
talk
yesterday
about
some
of
our
growing
pains
with
Basel
and
they're
pretty
much
shared
universally,
and
there
is
actually
very
interesting
to
hear
from
what
tensorflow
and
SpaceX
were
both
seeing
large
C++
code
bases
experience
and
there's
a
lot,
a
lot
of
overlap.
In
fact,
SpaceX
probably
closest
to
us
in
terms
of
the
what
they
had
to
do
to
make
things.
D
Work
and
SpaceX
have
actually
moved
like
a
lot
of
that
code,
including
all
their
like
their
rocket
software
and
a
flight
simulation
stuff
over
to
Basel,
and
so
they're
they're
really
committed
to
it,
and
they
they
the
same
issues.
Code
coverages
us,
though
they're
similar
strategy
for
managing
external
repositories
and
dependencies.
It's
it's
nice
to
have
a
community
there.
In
fact,
I
want
to
achieve
I
can
find
the
SpaceX
guys
today
and
chat
a
little
further
yeah.
D
We
caught
out
like
we
do
the
same
thing
that
y'all
waited
for
code
coverage.
Lots
of
hacking
see
this
issue
and
you
know
okay,
it's
good
to
see
it.
There's
actually
some
amount
of
sharing
going
on
there
and
not
a
lot
of
the
you
know
independently
invention
of
the
wheel
but
I.
Think
like
the
Basel
folks
are
hearing
most
the
pain
points
and
it's
actually
across
languages.
You
know
coverage
and
working
across
repositories
and
dependencies.
It
comes
up
as
a
constant
issue.
D
There's
a
lot
of
like
really
cool,
futuristic
stuff,
coming
like
very
soon
down
the
pipe
in
ambassador,
including,
like
remotes,
builds
offloading
a
caching
over
of
build
objects
and,
and
what
execution
and
like
this
is
that
exists
in
Google
today,
but
to
Sigma
and
uber.
Yeah
are
pushing
this
as
well
as
a
real
thing
and
there's
actually
some
open
source
repositories
which
actually
contain
code
or
which
does
this
today
so
yeah
cool.
C
C
So
we
don't
have
to
talk
about
it
a
lot,
but
you
know
there's
a
bunch
of
conferences
that
where
talks
about
on
boy
would
be
very
applicable
the
next
one
that
we'll
be
opening
for
CFPs
if
you're
into
International
Travel
will
be
Kuk
on
Europe,
that's
in
Denmark
and
hey,
but
then
starting
in
the
summer
from
velocity
to
you
know,
there's
there's
all
types
of
conferences,
I
think,
there's
quite
a
few
opportunities
and
there's
a
lot
of
eagerness
out
there
to
hear
people
talking
about
this.
This
subject
and
doesn't
have
to
be.
C
You
know,
like
envoy,
mechanics
I,
mean
I.
Think
people
are
really
interested
in
hearing
real
stories
from
people
who
have
actually
deployed.
So
you
know
if
you're
out
there
and
you're
and
you're
actually
using
I
mean
just
you
know,
kind
of
a
talk
about
war
stories
or
pros
and
cons
would
be
super
useful.
So
we
don't
really
talk
about
that
a
lot,
but
if
you're
interested
in
that
kind
of
thing
you
can
reach
out
to
me,
I
have
a
pretty
good
idea
of
what
the
conference's
are.
G
C
So
everything
that's
in
one
point:
5.0
we're
hoping
to
get
done
that
the
major
push
for
this
month
is
going
to
be
to
get
Docs
and
validation
around
the
v2
API.
So
you'll
see
a
bunch
of
stuff
happening
in
that
area.
I
think
that's
about
it.
I,
don't
I.
Don't
have
anything
else
so
happy
for
us
to
answer
questions
or,
if
there's
anything,
that
anyone
else
wanted
chat
about.
F
C
G
C
E
D
Literally
today,
you
should
see
a
commit
go
in
today
to
play
an
API
which
enables
importing
the
repository
from
let's
proto
production,
validator
repo,
which
includes
the
annotation
proto
validator
proto,
and
that
basically
defines
all
the
annotations
and
they're
basically
they're,
at
least
as
powerful,
I
think
as
JSON
schema,
and
you
can
actually
start
throwing
these
Perthes
for
annotations
on
two
protos
and
it
will
build
and
release
validate
that
syntactically
we're
using
the
annotations
correctly.
So
that's
going
in
today.
The
I
also
modify
the
proto
doc
tool
to
allow
us
to
interpret
the
optional
vs.
D
required
annotations
and
spit
out
some
useful
output
there
and
then
I
think
you
know
the
actual
validation
stuff
is
a
bigger
piece
of
work,
actually
not
doing
the
C++
template
thing
and
getting
all
the
cross
repository
build
stuff
working,
but
let's
at
least
start
out
getting
the
docs
there
and
throwing
the
annotations
into
the
photos
and
a
lot
of
the
way
there.
You.
E
C
I
mean
we
are,
we
are
using,
go
and
Python
internally,
mostly
so
yeah
it
was.
It
was
go
first,
but
I
think
they're
very
open
to
supporting
multiple
languages,
which
is
great
but
yeah,
I.
Think
once
that
gets
in,
and
you
know
I
think
towards
the
end
of
this
month.
It's
gonna
be
a
large
group
effort,
just
to
randomly
like
where
we
are
gonna
need
a
lot
of
help
to
go
in
and
kind
of
fix
up
the
docs.
C
But
it
still
means
that
there's
a
lot
of
texts,
that's
in
the
v1
Doc's
that
probably
literally
needs
to
get
copied
into
the
data
plan.
Api
repo.
So
it's
just
gonna
be
like
you
know.
One
thing
that
we
might
consider
you
know
is
kind
of
doing
it
over
over
email
and
kind
of
scheduling.
But
maybe
in
two
or
three
weeks
we
could
do
like
a
doc
bash,
where
we
just
get
a
bunch
of
people
into
slack
for
like
two
hours
and
just
like
read
the
docs
and
like
fix
Doc's,
basically
yeah.
E
By
just
simply
copying
some
of
the
dogs
from
the
main
envoy
thing
into
the
data
brain,
a
piece
with
that
actually
will
have
dogs
for
both
the
configuration
and
architecture
Docs
or
in
one
place.
It
just
becomes
easy
so
that
you
can
always
publish
talks
on
that
rapport,
either
than
flying
to
link
from
multiple
places.
G
E
C
C
But
that,
but
that
still
gets
confusing
it's
like
if
you
want
to
write
architecture
dogs,
it's
like
do
you
have
paragraphs
that
are
like
alighted
from
the
final
dioxide
I
mean
it's
I'm,
not
I'm,
not
opposed
to
moving
all
of
the
docs
they're
one.
One
downside
to
that
would
be
that
we've
made
this
theoretical
claim
that
the
data
plane
API
project
is
is
multi
projects
and
if
we
start
moving
all
the
docs.
Therefore
envoy
that
becomes
less
neutral,
less
believable,
so
you
know
I
could
probably
go
both
ways.
C
I
I
personally
think
the
path
of
least
resistance
is
going
to
be
to
do
what
we've
said
that
we're
gonna
do
I,
don't
think
it's
very
hard
and
then,
if
it
becomes
very
painful,
we
could
potentially
consider
doing
something
else.
But
I
think
that
once
we
can
kind
of
just
link
the
docs
in
and
do
that,
cross-linking
I
actually
think
it'll
fall
together
fairly
fairly
quickly.
Okay,.
D
I
mean
hopefully
that
should
be
dominated
the
week
and
then
basically
any
Aaron
who
wants
to
hack
on
this
I
think
you
know
one
of
the
issues
we
have
right
now
in
the
a
plain
API
repose.
The
comments
is
like
a
lot
of
these.
That
I
put
it.
There
were
based
on
the
v1
Docs,
but
they
didn't
have
any
of
the
iris-t
markup
or
anything
like
that.
That
me
to
sort
of
go
in
and
clean
all
that
up
and
make
sure
it
looks
essentially
what
it
looks
like
today.
But
you
know
from
the
v-chip
flavor.
C
Well,
that's
what
I
was
thinking
is
that
what
we
could
do
and
I
can
I
can
organize
this.
You
know
like
we
can.
We
can
get
volunteers
over
email
and
what
I
was
thinking
we
could
do
is
maybe
make
a
spreadsheet
or
like
we
could
assign
one
person
to
one
small
section
of
the
docs
right
so
so
and
then
like
we
can
just
spread
it
out.
Then,
basically,
that
person
will
be
responsible
for
getting
that
section
of
the
tops.
C
C
C
Yeah,
it's
mostly
yeah
I
mean
so
I.
Think
people
who
are
doing
any
work
on
new
features
definitely
will
have
to
add
annotations
and
if
we
add
this
doc
tag,
which
is
the
not
implemented
yet
tag
like
I,
think
probably
what
we
need
to
do
is
there's
just
like
there's
a
bunch
of
cleanup
to
do
so.
It's
like
there's
documentation,
for
example,
that's
in
data
playing
API
right
now
in
markdown.
There
probably,
for
example,
needs
to
go
into
the
architecture
Doc's
right
so.
C
D
C
Time,
yeah
I,
like
I,
think
if,
if
you
can
just
focus
on
basically
getting
us
so
that
we
can
build
the
docks
and
cross
link
and
basically
add
text
about
XDS
and
add
examples.
If
you
want
to
do
that
in
markdown,
that's
fine!
It
would
be
easier
if
it
was
in
rst,
but
it
doesn't
really
matter.
Leaving
I
mean
it
just
doesn't
matter
like
I
tend
to
agree
that
you
spending
your
time
doing.
C
Just
the
actual
mechanics
is
it's
better
and
then
I
think
we
can
put
together
a
spreadsheet
and
basically
just
assign
people,
small
sections
who
will
go
through
and
apply
annotation
and
basically
most
of
it
honestly
is
going
to
be
manually
copying
the
rst
text
from
the
v1
options,
including
markdown,
into
data
playing
API
so
that
when
it
when
it
renders
back
it
basically
looks
the
same
right,
I
think
it's
mostly
a
manual
and
then
once
it's
mostly
there.
It's
just
getting
people
to
read
through
the
docs
and
like
see
if
there's
any
errors,
basically
yeah.
D
C
You
know
I
think
one
thing
that
would
maybe
be
helpful
and
something
that
we
could
think
about
too.
Is
we
probably
need
to
have
a
little
contributor
guy
for
data
plane,
API,
which
basically
says
you
know,
for
example,
when
you
add
a
new
feature,
do
you
need
to
think
about
annotations?
You
should
add
the
not
implemented
tag
until
it's
implemented.
C
Right
like
it
doesn't
have
to
be
super
extensive,
but
that's
something
that
you
could
probably
help
with
too,
and
even
you
know,
even
if
it's
a
small
paragraph
but
like
a
little
example
on,
you
know
like
how
did
how
to
use
the
tool
to
do
X
or
something
like
that
like
just
to
make
sure
that
people
know
how
to
build
the
docs
locally.
How
to
cross
like
the
docks,
locally
I
think,
that's
probably
all
that's
required,
and
then
we
can
get
people
to
help
out.
C
C
Think
that
one
of
the
reasons
that
the
project
has
been
so
successful
is
actually
because
of
the
docks
and
I
I,
don't
want
to
compromise
on
on
dock
quality
so
and
I
I
think
you
know,
we've
done
in
general,
everyone
out
there
a
great
job
of
adding
really
nice
docks
that
I
think
if
they're
accessible
and
that
people
can
look
at
them
easily.
I
I
think
that's
great,
so
we
should.
We
should
try
to
keep
that
going.
B
Yes,
we
we
had
early
discussions
regarding
doing
kind
of
external
security
audits
having
a
professional
firm,
come
in
and
work
with
the
Envoy
team
and
do
a
public
report,
and
so
we
found
one
Matt
talked
to
them.
They
seem
to
be.
They
passed
the
mat
tests
from
as
far
as
I'm
concerned,
and
so
we
basically
contract
enough
to
do
this
work
and
there'll
be
a
public
report
published
in
January.
February
is
the
rough
time.
C
Yeah
I
mean
we.
We
are
in
general,
trying
to
kind
of
figure
out
the
best
way
of
getting
valuable
security
reports
yeah.
So
that's
the
conversation
about
bounty.
So
there's
no
there's
been
a
bunch
of
requests
out
there
for
kind
of
public
envoy
security
and
Ellison
and
yep.
No.
This
gets
very
complicated
because
we
have
some
very
large
users
of
of
envoy
who
are
doing
their
own
security
analysis
but,
as
everyone
knows
like
they're,
not
going
to
make
that
analysis,
public
yeah.
C
So
it's
you
know
it's
a
combination
of
making
people
feel
comfortable
with
with
the
software
because
it
obviously
doesn't
have
as
much
betting
is
something
like
nginx
and
you
know
getting
getting
eyes
on
it.
Even
though
we
know
that
there
are
a
lot
of
eyes
on
it,
it's
not
a
public
thing.
So
my
feeling
is,
you
know:
I
I,
don't
know
how
much
value
it's
gonna
add,
but
it
can't
hurt.
C
B
C
Cool
short
announcement
on
tap
for
two
weeks,
for
the
call
is
we're
gonna
have
Cisco
coming
in
to
discuss
envoy
with
a
technology
called
VPP,
which
is
their
user
mode.
Tcp
stack
on
top
of
DP
TK,
so
there's
been
I
would
say
a
lot
of
interest
out
there
on
doing
Colonel
by-paths,
with
with
ongoing,
so
I,
think
we're
going
to
see-
maybe
even
some
demos
by
coop
con,
though
I'm
not
I'm,
not
sure.
If
that
will
actually
transpire,
but
I
I
would
expect
that
you
know.
C
With
VPP,
especially,
they
have
two
modes,
one
of
them
is
they're
able
to
do
this.
Ld
preload
thing,
so
there's
actually
no
changes.
They
just
shim.
They
just
shim
these
api's.
The
other
thing
that's
coming
and
there's
been
a
couple
of
comments
in
different
PRS
le.
D
C
Odenie
works.
If
you
have
a
dynamically
linked
binary
right
right,
so
they
might
have
to
fix
that
right.
Oh
no,
but
I
think
they're
pre
loading
against
Lipsy,
which
are
enough.
Yet
we
do
link
to
which
we
don't
Saturday
right
yeah.
The
other
thing
that's
coming
and
I
think
is
gonna
help
this
effort.
A
lot
is
Lee's
on.
Has
that
PR
or
he's
gonna?
C
Do
the
low-level
socket
abstraction
and
I
believe
that
that
abstraction
will
let
anyone
come
in
and
basically
replace
the
low-level
connection
with
some
other
implementation,
so
I
actually
think
it'll
be
super
clean
and
that
will
allow
us
to
do
the
PPE
there's
a
couple
other
things
that
I
can't
talk
about
right
now
that
different
companies
are
actually
doing
that
may
plug
in
their
their
their
own
implementations.
It'll
allow
Microsoft
to
plug
in
a
implementation,
so
I
actually
think
that
that
abstraction
will
be
pretty
nice
and
it
should
be
pretty
clean
and
abstract.