►
From YouTube: WebAssembly Filters Meeting (June 14th, 2021)
Description
WebAssembly Filters Meeting - June 14th, 2021
Join the community at https://layer5.io/community
Find Layer5 on:
GitHub: https://github.com/layer5io
Twitter: https://twitter.com/layer5
LinkedIn: https://www.linkedin.com/company/layer5
Docker Hub: https://hub.docker.com/u/layer5/
A
Let's
get
up
and
going,
this
is
our
kind
of
a
kick-off
meeting
for
a
set
of
discussions
and
some
work
around
web
assembly
filters,
the
focus
in
context
of
envoy,
and
so
thanks
all
for
coming,
there's,
12
or
15
or
so
of
you
on
the
call,
I
suspect,
we'll
have
a
couple
of.
A
A
There's
you
know
understanding
now
that
when
we
sent
this
out
that
that
it
would
go
out
to
so
many,
I
guess
I
hadn't,
given
a
lot
of
consideration
as
to
much
of
a
formal
kickoff
other
than
just
what
had
been,
and
so
I'll
kind
of
explain
the
backstory
here
that
there
had
been
well
for
about
a
year
now
some
amount
of
doings
with
web
assembly
filters
in
the
community.
A
Some
of
those
that
work
had
been
demonstrated
at
last
year's
dockercon,
and
so
it
was
demonstrated
with
console
and
the
fact
that
it
uses
an
envoy-based
data
plane
and
is
capable
of
you
know
having
filters
loaded
into
it,
and
so
so
about
a
year
ago
there
was
an
effort
to
create
a
sample
application
and
show
that
people
can
you
incorporate
business
logic
into
their
data
plane
and
offload
kind
of
business
logic
into
their
data
plane,
and
that
that
that
particular
sample
app
was
doing
that.
A
Working
with
and
interacting
with
filters
and
some
of
the
roadmap
focus
for
what
we
think
might
come
and
some
of
that's
dependent
upon
your
input.
All
of
you
that
are
on
the
call,
and
so
we've
got
a
couple
of
topics
queued
up
before
we
jump
into
the
first
topic,
is
anybody
anybody
have
topics
to
add
or
clarifying
questions
or
comments
about
the
the
scope
of
the
meeting
or
the
kind
of
part
of
the
goals?
A
So
so
you
know
don't
be
shy.
I
guess
is
one
thing
number
two
is
that
some
of
you
haven't
seen
what's
the
best
way
to
to
go
about
this,
so
so
we
definitely
have
some
specific
topics.
Some
of
the
topics
may
interest
each
of
you.
Some
of
you
might
be
keen
to
just
be
apprised
of
what
mesherie
is
able
to
do
and
how
it
looks
to
parlay,
with
each
of
the
service
meshes
that
it
manages.
A
B
Right
and
here
yeah,
I
haven't
been
to
any
of
the
meetings
for
a
long
time,
but
I
saw
this
web
assembly
is
a
topic
dear
to
my
heart.
So
I
I
I
thought
I
popped
in
to
see
what
this
was
all
about
and,
and
I
see,
there's
a
filter
hub
mentioned
on
the
on
the
medium
units.
So
is
filter
hub
part
of
measuring
or.
A
Yeah,
that's
a
good
yeah.
I
it's
an
ill-formed,
it's
intentionally,
under-formed
concept.
It's
it's!
Yes
is
the
general
answer.
Although
open
for
feedback-
and
you
know
and
and
being
influenced
like
it's
been
the
you
know
the
notion
that
if
measure
is
going
to
help
manage
some
of
these
artifacts
that
are
you
know,
of
a
service
mesh
but
also
can
be
independently,
you
know,
maintained
and
built
and
shared
that
well,
there
does
need
to
be
to
be
some
place
for
distribution.
A
It
doesn't
have
to
be.
One
place,
would
be
many
places
and
so
yeah
to
date.
That
concept
hasn't
been
taken
very
far.
Currently
measuring
will
allows
you
to
load
your
filters
and
it'll
it'll.
Let
you
load
those
from
the
cli
or
from
the
ui.
A
It
will
connect
to
github
if
you
want
it
to
and
parse
through
your
registry
identify
filters,
it
will
let
you
load
those
filters
up
from
a
remote
http
source,
just
a
generic
http
endpoint,
and
so
it
just
has
very
basic,
like
crud
operations
for
being
in
an
artif,
a
filter
repository
like
an
artifact
repository
in
that
way,
and
that's
about
that's
about
the
extent
of
it.
Thus
far
it's
like
so
yeah,
I
would
love
so
and
it's
been
a
while.
Yes,
very
nice.
B
A
Yeah
next
time,
we'll
also
try
to
we'll
provide
some.
So
I'm
going
to
put
a
future
there's
a
couple
of
future
topics
here.
So
one
of
those
is
sort
of
a
demo
and
review
of
mesherie's
current
capabilities.
A
A
Yeah
and
so
the
contrib
there's
a
couple
of
contributors
that
have
made
the
web
assembly
capabilities
that
mesri
has
today
possible.
So
one
of
those
is
garage
chata
who's
on
the
call
navindu
podocat
is
it
has
no
electricity
at
the
moment,
there's
there's
a
handful
of
folks
who've
been
doing
that,
but
good
okay,
so
of
the
this
other.
The
first
discussion
here
so
samir
and
rudraksh
sameer
has
a
use
case
for
a
specific
behavior
that
he
would
like
out
of
a
filter,
and
this
has
to
do
with
and
so
samir
smear.
A
C
Yeah,
so
this
is
basically
from
a
like.
You
know.
We
are
also
using
a
micro
service
based
application
deployment
in
our
organization,
so
we're
in.
Currently
we
are
utilizing
the
capability
of
azure
api
management
service
to
do
this
particular
thing
or
the
thing
which
is
mentioned
here
basically,
but
why
we
do
not
want
to
use
that
is
because
of
like
you
know.
The
api
is
going
down
most
of
the
time,
because
this
is
this
is
a
behavior
of
azure.
So
there
are
some
uncertainty.
C
So
most
of
the
time
we
see
like
you
know,
this
apim
doesn't
work
very
efficiently.
C
So
that
is
why
we
have
like
you
know
when
we
planned
our
infrastructure
to
move
from
as
your
web
apps
to
like
you
know
the
kubernetes
workload.
Then
we
have
decided
to
go
with
services
capability.
Then
we
have
thought
of
various
use
cases
or
how
we
can
get
most
out
of
the
service
message.
That
is
where
we
come
across,
like
you
know
the
filter
capability
of
istio,
but
currently,
like
you,
know
what
I
have
been
doing
so
I
have
been
like
you
know,
beating
around
the
boost
in
envoy
proxy.
C
So
that
is
where
I
just
wanted
to
give
a
thought
like
if
we
can
achieve
this
from
the
like
necessary
point
of
view
because,
finally,
like
you
know,
this
is
something
which
is
very
close
to
the
heart
of
isto
and
when
we
talk
about
the
mercedes
which
does
manage
and
which
is
the
capability
to
like.
You
know,
visualize
how
your
service
needs
to
be
so
if
we
can
bring
in
this
kind
of
feature
so
which
is
basically,
we
have
the
jwt
token
wherein,
if
you
see
we
have
couple
of
information
which
comes
along.
C
So
there
is
a
header
part
wherein
it
has
a
particular
kind
and
from
there
there
is
a
like.
You
know,
yeah.
So
there
is
a
provider
who
actually
the
issuer,
who
is
from
the
odd
zero
and
when
we
decrypt
this
particular
thing?
Okay,
so
we
find
this
particular
header
and
payload.
So
what
exactly
we
are
trying
to
do,
because
we
have
to
do
some
of
the
processing
in
one
of
the
microservice
we
nee,
which
needs
my
email
id
as
well.
C
As
my
like,
you
know,
subject:
okay,
so,
which
has
to
be
added
to
the
header
and
passed
on
to
the
microservice
and
just
remove
the
alg
or
type
or
kind
from
the
header
part,
because
that
is
of
no
more
use
and
just
append
this
and
pass
the
request
to
like
no
next
virtual
service,
or
you
can
say,
like
next
kubernetes
service,
so
that
it
can
do
its
own
work
and
also
we
are
again
fine
graining
into
like.
You
know
how
to
get
only
the
domain
name
from
the
email,
id
and
process
it.
D
So
I
guess
in
the
interest
we
can
handle
the
roots,
I
mean
where
exactly
we
want
to
handle
some
of
the
headers
and
when
the
chart
is
received
basically
I'll
talk
about
this,
I
don't
know
that's
about
how
we
care
about
the
roots
as
of
now.
So
when
the
short
comes,
the
filters
role
here
would
be
to
decrypt
it
and
extract
the
headers
from
it.
B
D
C
C
So
if
you
can
see
my
screen
just
let
me
yeah.
C
Okay,
so
I
have
achieved
this
so
far,
so
this
particularly
does
the
authentication
okay.
So
it's
basically
take
the
issuer
which
we
have
given
and
the
jwk
is
uri.
Okay
and
after
that,
what
happened?
So
we
have
done
authorization
policy
thing,
so
we're
in
only
text
request:
principles,
okay,
so
which
is
already
done,
and
the
challenges
we
are
facing
is
here
okay,
so
I
am
literally
due
to
something
it
is
not
working,
and
this
seems
to
be
a
bit
complicated
for
me,
but
I'm
literally,
I
have
no
idea,
so
I'm
just
trying
out
various
things.
C
So
here
what
we
have
done.
So
this
is
the
envoy
function
which
I
have
used
with
the
request
handle,
and
then
I
am
fetching
this
from
the
body.
So
this
I
didn't
find
it
in
envoy,
like
you
know
and
boy
documentation,
and
I
do
not
know
whether
this
works
or
not
and
and
we
have
tried
it
didn't
work
first
time,
so
we
have
not
tested
after
that.
A
C
C
That
is
basically
these
three
things,
so
these
three
we
do
not
need
after
the
authentication.
So
once
this
comes
to
my
ingress,
then
it
one
second
once
this
comes
to
my
ingress.
So
now
this
type
jwt
has
been
handled
by
this
particular
thing,
and
and
it's
done
the
validation
this
one.
So
after
that
I
don't
need
those
things
so
now.
What
I
need
is,
I
need
something
which
is
in
my
body
here
has
to
be
made
available
here,
and
this
needs
to
be
deleted
from
this.
A
So
yeah,
and-
and
this
is
this
is
an
this-
is
an
example
of
the
contents
of
the
jwt.
C
Yes,
this
is
after,
like
you
know,
after
you
decrypt,
this
is
the
one
which
you
get.
So
how
did
I
do
that?
Just
I
went
here
say
wt
dot.
C
C
E
So,
basically,
how
you
can
manipulate
it
in
what
do
you
say
what
wasm
does?
Is
it
works
directly
with
http
packets?
So
I
I
just
had
a
question
to
clarify.
Like
this
token,
is
it
present
as
the
payload
for
the
request.
C
E
E
C
E
No,
no,
you
misunderstood,
so
you
have
http
headers
and
you
have
http
payload.
This
is
about
jot
tokens
themselves.
Yes,.
E
E
So
essentially,
your
ideal
case
would
be
to
handle
this
through
the
filter
itself,
right
like
through
the
ui.
Sorry.
C
Sorry
yeah,
so
what
I
wanted
to
do
is
I
wanted
to
write
a
filter,
something
something
like
this.
This
is
the
envoy
filter
so
because
I
do
not
understand
much
about
how
does
it
work?
C
C
C
E
E
Right
so
when
I
so
yeah
talking
speaking
about
that,
so
I'm
guessing
that
http
packet
after
the
author,
whatever
is
done,
does
the
authentication
service
generate
a
new?
What
do
you
say
a
request
or
forward
the
same
packet
so.
C
C
B
C
So
user
first
connect
to
the
front
end,
and
then
it
goes
to
the
auth
service,
which
is
authorization,
and
this
authorization
or
xero
provides
us
a
jwt
token
and
forwarded
this
to
istio
ingress
and
from
there
it
happens
like
you
know,
we
validate
the
request
with
jwp
roots
then.
Secondly,
what
we
have
to
do
is
extract
the
values
from
jwt
body
and
adds
it
to
header
and
pass
it
to
the
application.
So
this
is
how
it
happens.
E
C
E
My
concern
was,
you
were
directly
trying
to
manipulate
the
job,
so
I
was
just
thinking
how
to
since
again
the
sort
of
end
goal,
at
least
for
the
trade
quote-unquote
rate
limit
filter
for
now,
is
supposed
to
be
extensibility.
E
How
would
specifically
manipulating
the
jot
header
make
sense
so
yeah
your
use
case,
I
guess,
does
make
sense,
but
okay,
I
I
guess
I'll
come
to
that
in
my
topic.
A
Okay,
nice
yeah,
so
smeared
subsequent.
A
C
Exactly
yes,
yes,
that,
currently
what
we
are
doing,
we
are
using
the
path
best
route,
but
yeah.
That
is
again.
We
have
to
be
because
our
apim
currently
does
that
so
depending
upon
the
value
here,
it
will,
like
you,
know,
send
you
to
the
particular
like
you
know.
Let's
say
my
email
service.
Is
there
one
of
the
like,
you
know
microservice.
If
it
sees
an
email
in
the
header,
then
it
will
send
this
request
to
the
email,
microservice.
A
Nice,
okay,
and
so
so
at
the
moment
that
subsequent
routing
is
already
handled
by
your
microservices,
yes
via
istio,
so
we
yeah-
oh
okay,
you
already
have
existing
istio
config
that
that
okay,
very
good
yeah,
so
so
yeah
that
routing
isn't
happening
as
as
logic
inside
of
your
microservice,
but
rather
as
configuration
of
the
service
mesh.
Okay,.
C
A
Yes,
nice
all
right,
you're,
good,
rudraksha
and
pranav,
and
you
know
anyone
else,
any
other
clarifying.
You
know.
Questions
for
samir
outside
of
I
think
samir
will
get
a
sample
http
request,
but
other
than
that
early
thoughts,
rudraksha
pranav,
it
sounds
like
an
early
thought
from
you
is
great.
The
use
case
makes
sense
of
the
filter
that
you've
been
focused
on
this.
It
doesn't
fit
that
mold
necessarily
but
the
use,
but
it's
a
great
use
case,
and
so.
E
I
cannot
promise
amid
that
he
won't
have
to
code
anything
at
all.
That's
my
what
does
it
take
so
he
might
need
to.
What
do
you
say?
I
mean
I
guess
we
can
provide
functionality
and
then
okay
I'll
just
get
into.
When
I
talk
to
about
my
topic,
I
guess
it
will
be
a
little
more
clear.
A
Good
that
sounded
confident,
okay,
so
yeah
the
the
goal
there
is
that
you
actually
that's
the
goal.
Basically,
is
that
you
don't
you
don't
need
to
do
any
coding
that
there
would
be
a
filter
that
addresses
an
extensible
filter
that
addresses
this
use
case
that
allows
you
to
make
those
configurations
from
from
measuring
so
that
you
can
supply
the
logic
that
you
need
to
from
from
a
familiar
interface,
a
friendliness
file,
so
that
is
yeah,
yeah,
yeah
cool,
all
right,
fair
enough.
Let's
move
along
okay.
F
E
Up
hello,
good,
so
essentially
I
I
have
been
alluding
to
this
for
a
while,
and
I
apologize
for
my
questions
taking
up
so
much
time,
but
the
main
thing
that
okay,
when
I
started
off
on
this,
of
course
I
was
completely
new
and
I
didn't
really
have
a
grasp
of
the
topic,
but,
as
I
you
know,
continued
developing
the
filter
itself,
I
understood
what
it
was
and
more
of
its
capability
and
the
sort
of
what
they
say
holy
grail
for
a
filter
would
be
one
filter
to
rule
them
all
and
that
probably
won't
be
possible.
E
But
as
of
now,
there
is
just
one
rate
limit
filter.
One
file:
you
can
pass
it
a
configuration,
you
know,
define
the
paths
you
want
parts
you
have
and
as
well
as
the
rule,
you
want
to
apply
per
path
now
it
sort
of
you
know
started
off
as
that
and
then
it
needs.
Then
I
say:
okay,
fine,
since
it's
a
filter.
Now,
why
not?
I
spoke
lee
also
about
this.
E
A
couple
of
days
back
and
in
the
end
goal
would
be
to
make
some
sort
of
library
so
that
people
can
sort
of.
You
know
create
their
own
filters,
are
you
know
using
certain
predefined
functions
or
you
know,
ways
to
manipulate
http
packets
from
the
library?
But
you
know
this.
E
The
thought
of
a
library
was
because,
since
it's
what
do
you
say
image
hub
right,
the
goal
is
to
have
a
hub
where
people
can,
you
know,
sort
of
create
your
own
filters
and
since
one
filter
cannot
address
each
and
every
what
do
you
say
feature
that
a
user
may
want
or
each
and
every
manipulation
someone
would
want
to
perform
on
the
packet.
E
A
library
is
sort
of
an
end
goal
kind
of
there's
a
question
mark
on
that.
But
alongside
that
there
is
also
the
you
know.
The
the
filter
itself,
that's
meant
to
you,
know,
handle
a
lot
of
use
cases
for
that
as
well.
The
configuration
format
also
needs
to
be
set
such
that
you
know
it's
easy
enough
and
it's
also
viable
to
construct
that
config
and
pass
it
along.
E
So,
for
example,
for
the
again
I
coming
back
to
the
rate
limit
filter
once
you
define
that
path,
let's
say
path,
pull
as
the
rule
rate
limiter.
You
can
then,
have
you
know,
n
fields
such
that
you
can
define.
Okay,
fine,
if
you
find
this
particular
token
or
this
particular
key
in
the
jot
that
comes
in,
then
you
can
set
the
limit
to
10
or
20
or
30
and
like
that
you
can
have
n
such
values.
E
So
currently
it
is
crafted
for
image
hub
and
currently
image.
Hub
only
has
three
plant
types
as
a
personal
enterprise,
and
I
guess
one
more
with
a.
I
guess:
a
thousand
pull
requests
or
something
like
that
and
the
idea
is
you
shouldn't
be
limited
to
three.
You
should
be
able
to
pass
in
a
config
that,
can
you
know
let
you
define
10
000.,
so
the
ui
also
should
be
able
to
handle
that
another
thing
is
as
sami
spoke
about.
E
You
know
routing
to
specific
applications,
another
capability
that
you
know
that
should
be
there
that
can
be
passed
to
this,
like
you
know,
to
the
applying
part
of
what
you
say,
the
adapters
that
would
be
actually
pulling
the
current
list
of
services
and
then
selecting
which
service
to
apply
that
filter.
For
so
that's
another
thing.
That
would
be
really
like.
I
think
it's
integral
to
sort
of
you
know,
get
a
role.
Currently
it's
again
just
limited
to
image
up
for
those
of
you
who
sort
of
aren't
familiar
with
image
hub.
E
It's
basically,
just
a
image
hub
itself
is
a
sample
application.
It
aims
to
mimic
docker
hub
in
the
sense
that
you
can
sort
of
pull.
You
know
quote
unquote
images
when
you
actually
pull
nothing
really
happens.
It
just
sends
a
request
and
then
the
counter
goes
up
on
your
limit.
You
know
image
pull
limit,
so
the
goal
is
to
be
able
to
adapt
these
filters
for
any
application,
any
service,
and
so
far
I
guess
the
ui.
It's
still
under
active
development.
A
Yeah
thanks
thanks
for
that,
thanks
for
introducing
everyone
to
the
image
hub,
filter,
so
couple
ques
and
by
the
way
everyone
there's
there's
context
for
the
image,
hop
application
and
the
the
filter
that
goes
along
with
it.
Pranav
and
rudraksh
have
been
making
it
extensible,
which
is
great
extensible,
so
that
it
doesn't
have
hard-coded
values
in
the
filter
and
that
extensible
filter
is
ideally
configured
from
any
of
mashri's
clients
or
any
client
of
measuring.
I'm
the
two.
A
The
two
for
measured
right
now
are
mesh
ctl
and
measure
ui,
and
so
before
we
get
into
measure
ui
pranav
the
the
status-
and
this
is,
I
guess,
for
rudraksha
as
well,
but
the
status
of
the
measuring
adapter
for
istio
there
there's
an
update,
but
there's
a
is
there's
an
outstanding
pr.
I
think
that
will
tie
off
a
lot
of
the
initial
work
that
you
guys
have
done
and
that'll
make
that
will
sufficiently
have
removed
hard
coding
from
the
filter
is
that
is
that
an
accurate.
D
E
So
again,
the
main
thing
I
guess
naveen
do
is
looking
into
this,
and
I
was
supposed
to
actually
have
a
meeting
with
him
was
how
to
get
the
actual
filters
from
the
front
end
and
the
config
as
well
to
the
back
end.
There
is,
I
think
he
can
comment
on
this,
but
I
believe
the
grpc
calls
made,
and
there
is
a
particular
section-
okay
I'll
I'll.
Just
let
him
talk
about
this.
A
So
yeah
is,
although
I
think
what
you're
alluding
to
is
like
just
the
configuration
of
the
filter
itself
so
and
that
we
just
need
a
common
format
for
capturing
that
config
and
way
of
exchanging
the
configuration
info.
A
There's
there's
kind
of
two.
Actually
you
want
to
explain
the
the
concept
of
filter
configuration
from
the
two
pers
who
crashes
now
the
venue.
Maybe
if
this
is
familiar
enough
to
you,
do
you
want
to
explain.
A
E
Yeah
the
filter
itself,
currently,
no,
no,
no
changes
over
there.
You
know,
since
sameer
has
requested
for
manipulation,
so
I
guess
adding
http
headers
and
deleting
them
that
could
be
added
to
the
filter
itself.
So
I
guess
you
could
add
that
as
a
rule
type,
if
that
makes
sense,
you
might.
A
E
Existing
filter
can
what
you
say.
I
think
it
can
be
done
so
he
wants
all
the
so
he
can
specify.
I
guess
in,
like
I
said
right,
you
can
input
for
the
rate
limit
thing
itself.
You
can
input
n
values
so
similar
to
that
you
can
input
n
headers
that
you
want
removed
and
n
headers
that
you
want
added,
but
I'm
just
thinking
of
how
to
take
a
header
value
from
a
and
then
put
it
into
sorry
from
the
body
to
the
header.
E
A
G
Nothing
super
exciting,
so
we
have
kept
the
code
alive
in
some
private
branches.
It
was
last
working
on
console,
1.9
or
yeah
1.9,
so
it
probably
needs
some
minor
updates
to
110..
G
We
just
have
held
off
on
that,
because
we
want
to
make
sure
that
it's
clearly
messaged
between
or
as
being
an
experimental
feature
and
not
something
that,
like
our
enterprise,
customers
and
the
commercial
product
expect,
that
is,
you
know
fully
supported
that
they
can
rely
on
for
production
use
because
we
don't
obviously
wasm
is
still
pretty
young
and
new
and
rapidly
changing.
We
don't
want
people
to
implement
functionality
and
then
something
changes
and
they
have
feel
like
the
rug's
been
pulled
out
from
under
their
feet.
G
So
it's
just
more
of
a
support
thing
from
our
standpoint
of
how
do
we
again
keep
you
know
our
enterprise
customers
very
clear
on
what
the
expectations
are
for
that
feature.
A
G
Yeah,
I
do
think
they're.
The
good
thing
is.
We
have
some
other
use
cases
internally
that
look
like
they
might
be
solved
for
through
wasn't
filters,
so
there's
some
other
needs
why
this
might
end
up
being.
Finally,
mainlined,
rather
than
just
you
know,
there's
some
interest
in
the
community
from
a
few
folks.
A
Yeah
as
we
go,
yeah
makes
makes
sense
as
we
move
forth
in
here
on
various
use
cases
and
envisioning
ourselves
with
those
if
it
next
it
might.
You
know
if,
if
we,
if
we
get
some
of
your
time
next
time
we
meet
if
there
is
or
if
you
just
want
to
send
it
asynchronously,
but
if
there's
pre-experimental
guidance
as
to
you
know,
like
suggested
tooling,
or
a
suggested
approach
for
or
if
there
are
things
that
we
can
be
testing
out
of
the
experimental
approach.
A
You
know
you
know,
guide
us,
because
either
that
might
help
get
you
guys,
some
feedback
or
or
will
know
like
hey,
don't
try
this
on
110..
You
know
stay
focused
on
one.
Nine
use
this
mechanism
for
insertion.
G
Yeah
yeah,
I
can.
I
can
look
at
sharing
some
of
that
with
you,
the
you
know,
the
change,
the
things
that
have
changed
from
the
last
time
you
tried
it
is
now
we
had
a
more
first-class
way
to
configure
those
filters
using
console
service
configuration
the
config
entries
and
things
we
also
had.
I
think
we
were
looking
to
support
loading,
the
wasm
filters,
not
just
from
disk
but
even
from
a
central
http
source.
G
I
believe
that
code's
working,
I
have
to
go
check
that
as
well,
but
a
couple
nice
things
I
mean
we,
it
was
definitely
more
featureful,
so
I
can
maybe
get
that
together,
put
that
in
the
branch
of
my
personal
github
and
and
maybe
publish
a
docker
container
or
something
make
that
available
to
you.
A
A
That
yeah,
those
are
those
sound.
The
features
that
you
just
mentioned
are
can
make
things
a
lot
more.
A
Cool
okay,
yeah,
okay,
I
imagine
the
sensitivity
around
managing
the
road
map
for
for
this
type
of
a
feature
for
considering
how
young
it
is.
A
Okay.
Well
so
utkarsh
isn't
here,
maybe
that
all
and
the
vindu
was
on
his
mobile
and
he's
his
powers
out.
So
this
particular
conversation
here
is
it's
kind
of
two-parted
and
it's
real.
The
first
part
is
relatively
straightforward
and
this
is
to
say
that
anytime,
that
there's
a
filter
that's
to
be
loaded
into
a
service
mesh,
there's
service,
mesh,
specific
configuration
that
needs
to
be
supplied
along
with
the
the
filter
where
to
load
it
where
to
get
the
filter
from
where
to
load
it.
A
A
And
so
the
format
by
which
that
configuration
is
captured
is,
I
think,
the
the
first
one
of
those
is
relatively
straightforward
for
measuring
to
capture
and
utkarsha
is
the
individual
working
on
that
I
mean.
I
don't
know
that.
There's
a
lot
of
questions
there,
like
that's
defined
by
the
service
mesh
and
so
mastery
will
just
expose.
A
An
interface
to
configure
the
service
mesh
for
the
second
class
of
configuration
information,
that's
specific,
to
a
filter.
Well,
there
probably
we
probably
need
further
discussion
about
how
that
configuration
information
is
bundled
with
the
binaries
and
how
those
are
potentially
stored.
Alongside
the
binaries,
that's
configuration,
and
then
it's
really
that
type
of
config.
A
It
would
be
ideal
if
mescheri
is
able
to
give
give
users
a
pleasant
user.
You
know
a
user
interface
to
visually
update
the
filters
config.
So
if
the
filter
is
doing
rate,
limiting
like
a
visual
interface
to
be
able
to
change
the
rate
limits,
you
know
per
route
or
per
per
whatever
it
does,
and
so
like
some
amount
of
commonality
in
the
description
of
that
type
of
configuration
will
help,
and
so,
as
we
as
pranav
had
alluded
to
like
a
potential
goal
of
having
a
reusable
library
to
facilitate
creating.
A
You
know
a
set
of
different
filters
to
achieve
different
use
cases.
Many
of
those
filters
are
going
to
have
common
needs,
and
one
of
those
common
needs
would
be
configuration
of
the
filter
itself
like
what
the
filter
is
doing,
and
so,
if
that
configuration
is
exposed
in
a
common
format,
that
makes
it
easier
for
some
measures
ui
to
present
input
fields
to
the
user.
A
So
I'm
hoping
that
makes
sense
and
one
of
those
that
first
class
that
that
should
be
unlocked
fairly
shortly
and
for
and
I
guess
for
rudraksha
and
pranav
that
second
class
will
one
of
the
two
of
you
give
an
example
in
json
of
the
image
hub,
filter,
config.
E
E
So
I
just
had
another
query
so
since
it's
going
to
be
in
the
end
of
basically
a
base64
encoded
value-
and
there
is
a
database
as
well
on
measuring
server
right,
so
is
I
I
I
don't
know
if
navindu
understood
me
correctly,
but
what
he
said
was
that
you
could
store
the
binary
on
the
database.
E
E
A
A
lot
of
senses
yeah
because
you
can
hydrate
hydrate
and
dehydrate,
so
to
speak,
the
yeah
that
encoded
info
and
then
and
then
prepare
for
presentation
yeah.
You
know
good
if
a
good
place
to
land
part
of
that
discussion
is
probably
probably
in
this
as
a
common
way
for
mastery
to
to
perform
that
function.
A
Yeah,
okay,
yeah,
good
suggestion.
Okay,
there's
a
couple
of
topics
we
won't
get
to
today,
but
of
the
last
one
that
we
might
so
grav.
A
H
B
H
I
have
also
raised
draft
here,
so
I
need
some
of
the
help
in
in
rectifying
that,
where
I'm
lacking
and
making
some
changes
in
that.
F
Yeah
is
that
sorry.
E
No,
I
just
wanted
to
just
if
he
had
it
running.
Could
I
see
the
ui
you
have
it
running.
B
H
H
Yeah
so
currently
I'm
working
on
adding,
so
it's
not
running
so
currently
what
I
am
working
is
when
we
click
over
the
twitter.
So
what
we
want
to
create
is
a
react.
Json
schema,
which
I
was
implementing
from
past
two
days,
but
I
was
not
able
to
achieve
so.
Regarding
that,
I
have
made
two
changes
and
implemented
in
the
implementation.
I
had
made
a
draft
pr
for
that.
A
E
Yeah
yeah
so
currently,
okay,
I
I
just
wanted
to
know
that
is
this
fine,
like
I,
and
I
just
told
god
just
go
ahead
and
do
this
so
when
you
click
on
it,
there
should
be
like
a
pop-up
that
comes
with
all
the
previous
configurations:
yeah
something
like
this.
So
instead
of
the
actual
draw
jason
over
here
at
the
yaml,
but
instead
of
the
raw
jason
you'll
get
like
a
list
of
previous
configs.
E
If
you
have
none,
there's
a
button
saying
new
config,
you
click
that
and
then
that
form
that
that
react,
json
form
that
gaurav
just
showed
that
would
pop
up.
So
that
is
the
sort
of
idea
we
I
I
had,
and
what
I
wanted
to
go
for
is
that
fine.
A
Yeah,
that's
that's
about
right.
I
think
the
mesher
ui
itself
is
is
getting
ready
for
a
bit
of
an
overhaul,
and
so
in
some
respects
it
may
not
matter
exactly.
You
know
the
way
that
you
guys
do
it.
I
would
suggest.
I
think
that
if
you
visit
one
of
the
adapter
pages,
if
any
of
them
are
running.
A
And
you'll
recall
there's
some
cards
on
the
page,
like
probably
the,
instead
of
it
being
a
modal
in
which
people
configure
the
filter,
it's
probably
similar
to
one
of
the
managed
pages
of
an
adapter
that
there's
there's
a
card
that
would
expose
the
inputs
there,
but
but
it's
really
rather
similar
to
what
you're
saying
print
off
and
yeah
the
ability
to
version
or
over
time.
That
will
be
important
right
now.
E
I
wasn't
really
thinking
about
versioning
so
like,
for
example,
right
now
you
may
have
what
you
say:
one
config
for
okay,
today's
I'm
just
giving
an
example.
So
let's
take
hub
itself,
so
I
mean
today
you're
feeling
like
okay,
I
want
the
rate
limits
to
be
10,
10
10
each.
You
know
everyone
should
be
equal
or
something
like
that.
Some
nonsense
like
that
and
you
just
have
a
different
version
of
the
conflict,
so
it
of
course,
would
make
the
filter
behave
differently.
E
So,
depending
on
the
situation
and
the
rule
type,
and
maybe
perhaps
some
quick
change,
you
wanted
to
make
or
test
out
something.
It's
not,
I
won't
say
versioning,
but
since
it's
a
config,
you
can
make
the
filter
behave
differently
depending
on
what
you
wanted
to
do
and
when
so
that
was
my
main
intuition.
Not
me,
yeah
versioning
is
pretty
important,
but
my
main
thing
was
having
different
behavior
with
the
same
filter.
A
Yeah,
that's
it
good
yeah
that
that's
something
we
need
to
go
right
down
into
the
design.
Spec
is
to
track
and
understand
when
when
and
where
that
a
given
filter
has
been
applied
and
what
the
configuration
was
at
the
time
that
it
was
applied,
which
I
think
is
what
you're
pointing
to
and
then
when
someone
subsequently
applies
it
somewhere
else,
if
they
up
change
that
config.
For
that
particular
instance,
we'll
need
to
track
what
that
is
as
well
and
so
yeah
out.
A
That's
those
are
considerations
to
your
point
that
are
necessary,
even
in
advance
of
versioning,
the
config
itself,
like
each
of
those
instances
in
which
it
was
applied,
they
each
would
need
to
track
their
own
revision,
history
and
so
yeah.
So
we
need
to.
I
don't
know
that
we've
got
any
of
that
in
the
design.
Spec,
that's
a
good
thing
to
go,
expand
on,
especially
in
context
of
what
you
were
just
saying.
A
Pranav
about
you
know,
persisting
the
encoded
config
in
the
database
like
yes
that
and
then
that
needs
to
be
tracked,
not
only
like
the
default
configuration
for
each
filter,
but
every
time
that
the
filter
is
applied.
What
was
the
config
for
that
application?.
E
So
yeah,
no,
so
I
I
was
thinking
instead
of
like
you
know,
having
a
default
config
to
go
with
right
off,
but
when
you're
deploying
the
filter,
you'll
get
an
option
to
apply
a
config.
So
then
I
know
it's
a
bit
tedious
as
of
now,
but
that
would
be
the
pres.
It
would
force
you
to
select
it.
If
you
don't
have
any
it's
fine.
If
you
have
something,
then.
A
A
A
couple
things
we
didn't
get
to
so
I'll.
We
should
verify
with
karsh,
I
think,
he's
probably
fairly
far
along
here.
Hopefully
we
should
make
sure
to
catch
up
with
navendu
and
piyush
on,
because
it's
not
just
the
binary
that
needs
to
be
supplied,
but
it's
the
config
as
well.
A
There
needs
to
be
a
consideration
there.
Samir
will,
I
think,
provide
maybe
a
sample
http
request
or
so.
E
He's
been,
he
has
given
me
the
request
in
slack,
I
if
sami
is
still
here,
could
I'll,
okay
I'll
just
what
do
you
say,
put
them
on
the
wasm
channel,
okay,.
A
Good
perfect
and
then
we'll
look
for
a
follow-up
on
a
number
of
these
items.
Next
time
we
meet,
but
also
probably
a
quick
review
of
meshri's
current
capabilities.
E
Also
lee
is
this
duration
of
the
meeting
fine,
or
do
we
extend
it
a
bit
more
right
now,
it's
45
minutes.
So
should
we
keep
it
in
our
or
let's
continue
one
more
time
with
45
minutes
see
if
it
fits.
A
Yeah,
we
might
want
to
extend
it
to
an
hour
if
we're
going
to
go
over
like
we
just
did.
I
didn't
realize
it
was
only
45
minutes.
Yep.
E
I
was
suggesting-
maybe
we
could
do
45
minutes
one
like
next
week
and
even
if
that
goes,
you
know
into
overtime,
then
we
will
extend
it.
That
was
my.
E
A
Okay,
so
it
was
good
I
what
what
was
this
scheduled
for?
Was
it
an
hour
or
was
it
45?
Was
it
half
an
hour
party.
A
E
A
Okay,
thank
you
all
for
coming.
It's
been
a
nice
kickoff,
see
see
you
all
in
a
week,
see
you
on
the
21st
yeah
thanks
bye.
I
guess.