►
From YouTube: WebAssembly Filters Meeting (Oct 25th, 2021)
Description
WebAssembly Filters Meeting - October 25th, 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
Welcome
everyone
to
the
awesome
filters
meeting
today
is
the
25th
of
october,
and
we
have
some
topics
lined
up
mostly.
This
call
would
be
more
of
a
discussion
on
what
what
we'll
be
working
on
in
the
future,
so
around
awesome
filters
in
in
measuring
and
its
surrounding
projects.
A
So,
let's
start
off
this
meeting
by
looking
at
the
cloud
native
day
web
assembly
so
lee
like.
Would
you
like
to
start
this
off
yeah.
B
So
so,
in
this
meeting,
we've
been
talking
a
lot
about
mastery's
ability
to
manage
web
assembly
filters
and
that's
good,
and
we
will
talk
about
that
again
today.
We
also
have
spoken
or
had
other
people
come
and
present
on
webassembly
in
general,
more
specifically
webassembly
toward
filtered
envoy
filters,
and
so
you
know
the
last
time
we
met
that
we,
we
just
said
hey
it's
probably
time
to
invite
some
more
folks
to
come
and
share
with
us.
B
So
we
can
learn
as
we
go
here
and
and
so,
given
that
there
was
just
a
recent
cloud
native
day,
wasn't
some
of
those
speakers
will
have
fresh
material,
and
you
know
fresh
things
to
share
and
so
yeah
that
that
schedule
makes
a
good
place
to
look
for
some
of
those
topics.
B
Levindu
called
out
one
on
wassi
and
neural
networks,
which
to
me
seems
quite
relevant
to
our
to
our
focus
and
unfortunately,
the
two
people
that
presented
that
wassi
nnn
wassi
neural
network
they're,
both
from
intel
and
intel
partners
on
some
of
the
projects
here.
So
while
both
of
their
names
aren't
familiar
getting
them
to
come
present
is
probably
easy
enough.
B
B
A
Yeah
yeah
like
for
me,
since
we
were
looking
into
some
machine
learning
capabilities
around
measuring
one
of
the
like.
We
should
be
able
to
leverage
the
the
system
in
which
the
machine
learning
model
is
running.
So
vaasi
neural
networks
comes
with
such
a
capability,
like
you
can
utilize
the
inbuilt
cpu
like
the
cpu
architecture
more
effectively,
so
you
can
use
the
full
capacity
of
the
system
in
which
you
are
running.
So
if
it
has
a
gpu,
then
we
can
natively
run
it
on
the
gpu.
So
that
is
what
I
was
thinking.
B
Nice
yeah
do
you
the
venue:
do
you
want
to
reach
out
to
the
so
this
is
is
relevant
for
smp
as
well.
Let
me
reach
out
to
sunku
and
ritika:
ask
if
they
know
these
folks
and
can
provide
an
introduction
and
see
if
we
can
get
them
to
come
and
probably
present
in
this
call.
You
know
because
it's
the
most
relevant
here,
but
but
it's
relevant
to
smp
as
well.
B
That's
a
good
one.
Hopefully
yeah
it
wouldn't
just
yeah
anyway.
I'll
withhold
some
of
my
questions
on
it,
but
it
seems
there's
another
one.
Just
this
general
one
on
developing
and
debugging
awesome
filters.
So
do
who
was
this
given
by
oh.
B
Okay,
or
I
mean
who
is
the
presentation
given
by
yeah,
we
don't
want
to
have
this
one,
but
but
there
are
some
other
good
ones
to
call
in
that.
First,
one
is
a
good
suggestion
we
should
follow
up.
Does
anybody
else
have
suggestions
on
topics
about
webassembly?
A
So
I'll
share
the
link
to
the
wassie
awesome
event,
so
you
can
check
those
out.
The
event
took
place
a
couple
of
months
back,
so
there
are
recordings
as
well.
So
you
can
check
this
out
those
out
as
well.
B
So
yeah,
so
I
dropped
this
on
on
obvi.
So
actually
this
might
be
more
of
a
there's,
a
design
needed
here.
What
what
have
we
concluded?
Uekash.
C
Well,
I
think
I
I
propose
that
this
is
how
the
structure
of
the
config
could
be.
I
mean
the
metadata
that
we
should
be
storing
along
with
filters
and
yeah.
That
was
the
last
that
we
talked
about
and
in
the
entire
thing
the
entire
solution
was
in
our
github
issue.
I'm
not
sure
if
the
github
issue
is
still
open
or
not,
but
yeah
there
is
where
I
was
proposing
that
this
is
the
all
of
the
metadata.
C
Exactly
even
I
don't
remember
what
the
metadata
or
what
all
I
was
suggesting
at
that
point
of
time,
but
the
issue
was
about
that.
These
are
the
materials
that
we
should
be
storing
and
these
are
the,
and
this
is
the
way
we
should
be
storing
them.
So
that
was
the
last
thing
that
we
talked
about.
B
And
if
I
yeah,
I
don't
recall
not
or
like
it
seems
like
yeah
after
after
we
review
it
seems
like
we
should
probably
go
forth.
It
might
be
that
there's
a
contributor
needed,
though
from
the
community,
I'm
probably
a
golan
contributor
right
to
be
able
to
work
through
this
there's
something
that's
a
little
bit
of
a
challenge
like
to
the
extent
that
people
are
using
measuring
ctl
to
import
a
filter,
a
binary
and
excuse
me,
and
if
they
also
import
a
configuration
file,
the
question
becomes
well:
do
we?
B
How
do
we
want
them
to
package?
That
is
that
supposed
to
be
in
one
zip?
Should
they
import
two
files?
Should
they
point
to
a
directory,
and
then
the
utility
will
recurse
the
directory,
the
zip
or
the
tar
or
the
zip
makes
more
sense,
because
you
can
then
reference
it
as
a
single
location
for
a
remote.
You
know
remote
url
as
well,
but
that
also
means
then
defining
well
not
a
standard
and
not
not
a
spec
person.
Well,
some
respect
is
spec
right
because.
B
B
B
As
an
alternate
to
pointing
to
a
tar
file
or
a
zip
file,
it
could
be
that
we
peop
we
require
that
people
point
to
a
service
mesh
pattern
that
has
this
configuration
that's
being
proposed
and
that
measury,
as
it
interprets
patterns,
reads
through
any
other
urls
that
it
needs
to
to
go
retrieve
them.
B
That
would
might
circumvent
some
work
for
us,
because
we
have
to
define
something
else
and
kind
of
support.
Some
other
structure
it
might
reinforce
the
surface
use
of
service
mesh
patterns
is
like
the
thing.
It
also
might
mean
that
if
someone
provides
a
pattern
that
happens,
that
so
happens
to
call
to
a
filter.
C
Glad
I
thought
of
that
suggestion.
I
think
the
proposal
was
that
we
will
have
everything
in
one
part
and
file,
while
so
the
patent
will
have
an
http
url
from
which
you
can
import
version,
and
it
will
also
have
the
dynamic
configuration,
but
what
it
did
not.
I
think
so.
One
thing
that
was
left
out
was
the
json
schema
of
that
custom.
The
json
schema
of
the
json
schema
like
the
custom
config
that
will
get
inside,
though
so
that
was,
I
think,
the
thing
that
was
left
out.
B
B
B
Yeah,
I
suppose
on
one
hand,
if
what
people
are
typically
applying
as
configuration
supplying
as
configuration
to
a
filter,
is
a
string,
some
string
or
set
of
strings
of
something
that
is
that,
would
it
be
harmful
to
wrap
it
in
a
json
which
might
have
like
I
mean
like
wrap
it
externally.
B
I
don't,
I
don't,
think
it's
harmful
right
to
wrap
it
externally,
because
you'd
still
say
well,
here's
whatever
blob
they
provided
it's
just
that
you're,
adding
some
metadata
to
make
sure
that
it's
like
well,
they
provided
it
on
this
date,
and
this
is
the
version
number
they
actually
provided
eight
times
now.
B
So
there's
a
you
know,
eight
versions
of
it
and,
and
you
know,
and
whatever
else
there
is
like
it's
okay,
so
was
that
really
just
okay
that
json
that
metadata
json
of
json
was
that
any
different
than
is
it
gonna
get
confusing,
but
the
the
metadata
that's
tracked
about
a
filter,
a
binary
today,
it's
things
like
from:
where
was
it
originally
loc
updated
like
what
was
its
location
or
what?
What
continues
to
be
its
location?
Potentially,
when
was
it
imported?
When
was
it
last
updated?
Like?
B
C
So
that
the
material
that
you
just
talked
about
is
something
that
we
store
alongside
the
pattern,
because
everything
this
entire
thing
is
represented
as
a
button.
Alongside
that
we
store
location
and
update
and
create
and
those
things
and
who
created
it,
the
the
material
that
I
was
proposing
to
store
other
than
that
was
in
it
would
be
in
the
patent
file.
So
not
other
fields,
it
would
be
in
the
patent
file,
but
regarding
that
where
the
filter
can
be
dropped
in
so
it
would
act
as
a
not
that
we
won't.
C
We
won't
allow
the
user
to
modify
that,
but
the
idea
is
that
if
we
just
park
it,
so
let's
say
we
have
official
filters
and
we
know
where
to
drop
them.
We
know
how
to
configure
them,
so
we
can
have
default
default
setup
there
and
the
user
can
change
them,
but
they
would
be
default.
They
can
just
click
on
apply
with
a
custom
conflict
like
adjacent
or
maybe
in
some
cases
there
is
no
config,
you
just
deploy
it
and
it
just
works
out
of
the
box.
C
B
Yeah
and
then
that
makes
sense
to
me.
It
also
supports
the
notion
that,
as
a
user
goes
to
use
mesh
map
that
they
can
meshmap
will
need
to
track
what
well
it
needs
to
understand
what
objects
on
the
canvas,
what
nodes,
what
node
types
on
the
canvas
are
allowed
to
receive
a
filter
right,
because
that
yeah,
because
it
because
of
th-
and
this
that's
where
this
metadata
would
also
come
in.
B
Okay-
and
I
don't
know,
if
there's
anyone
on
the
call
that
who's
a
that
this
is
well
suited
for,
but
if
there
is
for
anyone
who's
on
the
call,
it's
the
last.
It's
the
link,
it's
the
issue
that
navendu
is
sharing
and
oops.
B
So
yeah
take
a
look,
if
not
we'll
just
tag
this
with
the
will
fix,
which
davindu
already
did,
which
will
get
the
stale
bot
off
of
its
back.
B
Okay,
if
there's
no
immediate
takers,
good,
let's,
let's
move
on
we'll
we'll
we
partial
to
talk
about.
If
no
one
reaches
out
today
to
you
about
it,
then
we'll
just
have
to
talk
about
how
to
get
it
queued
up.
A
All
right,
the
next
topic
is
github
actions
for
western
filters.
B
Oh
cyprus,
testing
of
what
was
it:
oh,
the
github
action.
Oh
sorry,
well,.
B
This
would
be
the
ability
to
so
so
this
is
yeah.
This
is,
I
think,
if
you,
if
you
make
this
generic,
this
is
the
ability.
This
is
a
github
action
for
not
for
performance
testing,
not
for
smi
conformance
testing,
but
for
measuring,
like
patterns
and
applying
a
pattern,
and
it
might
so
happen
that
the
pattern
that
you're
dealing
with
includes
a
webassembly
filter.
B
It's
like
the
reason
that
we
don't
is
because
of
actions
themselves,
how
it
is
that
when
you
configure
an
action
and
expose
and
and
people
go
to
use
it,
they
need
to
configure
what
they're
going
to
do.
You
can
have
a
generic
one.
It
just
means
that
it's
an
uncurated
experience
like
it's
unopinionated
and
a
little
a
little
more
difficult
to
wield
so
having
one
dedicated
to
just
the
use
of
patterns
might
make
sense.
B
And
and
jubril
like,
if
it's
not
that
we
would
that,
I
don't
think
anyone
here
would
push
back
on
a
generic
one.
You
know,
let's
do
it
and
then
for
other
situations
we
were
able
to
spoon
feed
a
bit
more.
B
B
So
that's
fantastic
like
like
it's
it's
going
to
be
used
and
and
like
it.
That
speaks
to
the
value
of
having
created
an
action
specific
to
that
to
smi
conformance
specif
specific
to
that
use
case.
So
it
will
see
you
since
being
built
in
right
here
and
that's
great,
but
a
generic
one,
jubril
yeah.
If
we
can
get
like.
B
Yeah
I
mean
that'd
be
great
like
because
then
people
can
just
take
existing
workflows
that
they
have
and
like
slice
in
just
one
messy
ctl
command
here,
maybe
somewhere
else
another
one.
Here,
it's
good
because
yeah
at
some
point
we
can't
define.
We
can't
write.
We
shouldn't
write
15
actions.
It
should
we
should
end
up
with
a
generic
one.
B
Yeah
and
so
they're
actively
trying
to
smi
is
actively
trying
to
pull
in
the
github
action
now
keith
maddox
is
the
smi
contributor.
That's
actively
working
on
that.
So
if
you
should
see
his
activity
in
under
the
smi
spec
repo
under
the
controller
sdk.
B
B
Okay,
yeah,
as
they
go
to
use
it,
they
have
to
include
a
couple
of
secrets.
I
think,
in
order
to
transmit
the
conformance
test
back
securely
and
so
ludkarsh
made
a
tokens
command
like
so
I
think
I
think,
they're
okay
to
go
use.
Okay,
I
don't
know
of
any
open
issues
that
would
prevent
them
from
using
it.
B
Yeah,
so
anita
or
adina
adina,
I
was
just
on
a
call
with
anita,
so
apologies
yeah,
so
actually
as
we're
gonna
do
this.
That
means
that
we
probably
need
two
more
repos
and
that
adina,
to
the
extent
that
you
were
to
go,
take
an
attempt
at
one
for
webassembly
filters.
Well,
I'm
sorry
for
web
assembly
filters,
but
but
actually
it's
for
it's
for
patterns
it
just
so
it
like.
If
your
pattern
included
a
webassembly
filter
great
than
it,
then
you
know
it
would
work
so
I'll.
Go
I'll.
B
Take
an
action
item
to
go,
create
two
repos
real
quickly
and
make
sure
that
jubril
and
adina
both
have
administrative
privilege
over
the
repos.
Just
because
up
front
we
don't
need
a
lot
of
rigor
around.
B
Pull
request,
reviews
and
that
kind
of
a
thing
I
think
having
administrative
access
to
the
repo
is
helpful
to
the
extent
that
part
of
the
at
least
for
me,
as
I'm,
trying
to
test
out
a
github
action,
it's
like
I
want
to
go
invoke,
I
get
it.
I
want
to
go
start
the
action
and
if
I
don't
have
at
least
write
permission,
then
I
can't
just
go:
kick
it
off
so.
B
I
don't
know
that
there
are
open
issues
for
either
of
the
two
of
these
requests.
I
could
be
wrong
so
adina
and
jibril.
Actually
it's
a
if
the
two
of
you
open
up
issues,
that's
a
great.
That
would
be
good,
so
we
would
tr
track
this.
Also.
It
would
be
a
great
way
of
you
having
to
make
sure
that
you're
you're
mentally
grocking
the
goal,
because
you
have
to
write
it
down
and
say
what
you're
going
to
try
to
do,
and
so
you
might.
B
B
B
From
f5
he
was
going
to
work
on
some
cyprus
tests
he's
not
on
at
the
moment.
Gosh
is
that
who's?
Who
remembers
this
yeah,
the
nice
guy
from
f5
that
was
new?
He
was
he
lives
in
mexico
mark
anyway,
I'll
find
his
name,
but
but
he
was
going
to
work
on
making
sure
that
we
have
some
new
cypress
tests,
ones
that
would
cover
webassembly
filters.
B
B
I
think
fixing
one
of
the
cypress
tests
in
a
different
area,
so
I
think
he's
into
the
rhythm,
but
he
probably
needs
a
good
description
of
what
would
be
done,
okay
and
crush
like
I
actually
don't
know
that,
like
because
of
the
changes
that
are
going
on
around
webassembly
filters,
it
might
not
be
the
best
one
for
him
to
chase
it
might
be
that
like
instead,
just
going
after
one
for
patterns
might
be
better.
I
know
there's
changes
going
on
there.
B
So
you
know
in
terms
of
like
continuous
integration
that
functional
testing,
if
you
are,
you
know
just
to
say
so,
because.
B
Because
more
tests
more
coverage
there
is
needed,
and
there
aren't
that
many
people
here
that
know
cyprus.
So
it's
I
don't
think
it's
very
difficult,
but
it
does
you
know,
I
don't
think
it's
horrifically
difficult.
It
does
take
an
intelligent
individual,
a
little
bit
of
time
to
familiarize
and
like
get
started.
It's
really
impactful
because
we're
not
we
don't
have
great
functional
test
coverage,
and
so
the
shaw
this
is.
B
This
is
one
of
those
things
like
actually
vishal.
It
would
be
good
as
you
review
the
coverage.
That's
there,
which
you
can
see
inside
of
the
repo
you
can
go.
You
can
go
to
mostly
any
of
the
open,
pull
requests
and
see
what
cyprus
tests
are
being
run
and
get
familiar
with.
The
existing
ones
and
you'll
readily
see
that,
like
there's
a
lot
of
other
functional
areas
of
mastery
that
are
not
being
tested.
B
And
so
for
sure,
if
you
could,
if
you
could
do
that
and
come
back
and
suggest
like
hey
that
you
think
we
should
do
one
here,
please
do
mario
and
rodolfo
and
delusional
optimist.
Rudraksh
parikh.
Those
three
in
particular
have
been
kind
of
focused
on
that.
So
inside
of
the
ci
channel
yeah,
please
please
go
suggest
the
way.
B
And
so
good,
well,
I'm
just
having
a
heck
of
a
time
just
kept
going
through.
I
just
keep
going
through
the
next
topics,
and
so
assuming
we're
okay,
there
gaurav
he
was,
he
did
make
a
pr,
or
at
least
a
prior
pr
on
the
import
options
so
megana,
since
you
are
on
the
call.
D
Last
time
I
talked
with
him,
I
think
at
that
point
the
awesome
filters
meetings
were
stopped
for
a
for
a
while
right.
So
it
was
that
time
and
he
did
he.
He
wasn't
sure.
I
think
if
the
work
needs
to
be
continued
or
not,
then
he
started
working.
I
didn't.
I
didn't
talk
to
him
after
that.
Actually,
so
I'm
I'm
not
sure
either
regarding
it.
I
was
planning
on
talking
to
him
about
it.
D
He
he
he
wasn't
working
on
it.
I
think
so
or
he
stopped
for
a
while
or
I'm
not
exactly
sure
about
it
either.
He
was
a
little
confused
about
the
github
imports.
So
that's
what
he
mentioned
to
me.
B
D
That
that
that's
what
I
know,
actually
I
don't
have
a
update
regarding
it.
B
Oh
okay,
yeah!
No,
no,
no
problem!
Yeah!
Let's
take
let's
refresh
real
real
quick,
because
because
I
think
we're
almost
done,
if
not
done
the
whole
like
complicated
github,
like
oauth
interaction
with
github.
That
can
wait
like
that.
We
can
have
that
as
a
as
a
secondary
focus
after
we're
after
we've
confirmed
that
we've
got
a
uniform
experience
here.
So
if
we
take
a
look
at
applications
when
someone
goes
to
by
the
way
the
applications,
these
are,
for
the
most
part
kubernetes
manifest
just
different
workloads
that
people
will
provision
on
kubernetes.
B
As
someone
goes
to
import
one
we
have
looks
like
we
have
two
buttons
right
now,
I'm
just
I'm
just
refreshing.
So
we
have
this
button
that
if
you
want
to
import
one
using
a
direct,
url,
okay,
I
think
this
pr.
I
think
this
was
recently
changed
to
not
use
this
color
and
I
think
so
I'm
running
v05.67,
but
I
think
this
has
changed.
So
at
least
the
styling
is
different.
B
B
Now
this
is
the
experience
that
might
change
a
little
bit
like.
I
don't
think
we
have.
These
are
the
requirements
that
we
were
just
looking
at
from
utkarsh
earlier
on
the
call
it's
like
okay.
If
we
have
multiple
files
to
upload,
how
would
we
facilitate
that?
That's
an
open
question,
but
in
terms
of
importing,
I
think
I
think
this
is
defining
that's
done
for
patterns.
B
B
We're
done
so,
okay
and
then
yeah
so
good.
I
think
we're
concluding
that.
That
particular
meeting
item
is,
I
think,
we're
done
because
I
think
he
updated
the
colors.
I
saw
aditya
chatterjee
engaging
with
him
and
merging
some
stuff,
so
so
megan
out,
that's
nice.
B
D
B
He's
getting
in
your
way
we
need
to.
I
mean
I
I
I
was
just
talking
with
gaurav
and
actually,
I
think,
he's
doing
some
exams
and
stuff.
What
I
mean
to
say
is
we
either
need
to
have
him
step
aside
so
that
you
can
carry
this
forth
or
we
need
to
get
you
on
to
something
else
in
this
case
it,
maybe
we
should
get
you
on
to
something
else,
just
because
this
is.
B
D
Going
to
ask
for
some
if
there
are
any
open
vrs
that
I
can
work
on
like
I
do
have
some
work.
I
still
need
to
complete,
but
yeah,
I
think
I'll
be
like
once
the
exams
are
done.
I
think
I'll
be
able
to
do
it
so
shouldn't
they.
B
A
A
B
Ashish
is
not
on
the
call
right
crush.
Do
you
want
to
like
the
short
of
it
here
is
that,
yes,
I
think
filters
are
b
the
filters
from
the
wasom
filters.
Repo
are
being
excuse
me
imported.
B
And
seated,
but
that
there's
not
really
a,
but
that's
being
done,
it's
just
that
that
code
hasn't
been
released
yet
so,
if
you're
running
meshri
edge
channel
the
edge
release,
then
that
should
be
happening
for
you.
It's
actually
a
good
thing
to
you
know,
force
people
to
test
and
see
actually
push.
Do
you
mind
testing
that
you
should
see
logs
when
you
start
mesh
reservoir
on
the
edge
channel,
saying
that
there's
content,
that's
being
seeded
and
I
think.
A
B
Yeah,
I
think
it's
filters
sample
applications
and
then
utkarsh
do
you
know,
are
any
patterns
being
seated?
B
B
Now
push
when
you
do
that
test,
if
you
would
test
both
under
the
local
provider
and
then
under
the
meshry
cloud
provider,.
A
All
right,
thank
you
for
that
update.
The
last
topic
is
on
filter
hub
and
whether
we
should
have
a
a
report
template
for
creating
new
filters
or
for
people
to
contribute
new
filters
in
mercedes.
So,
as
we
showed
earlier
like
early,
if
you
could
go
to
the
awesome
filters
repo,
we
have
a
bunch
of.
A
A
So
related
to
that,
we
are
also
looking
into
a
design
filter
hub,
which
will
be
a
repository
where
which
stores
awesome
filters.
So
users
can
readily
get
filters
from
this
filter
hub
and
it
will
be
hosted
by
us
and
the
measuring
project.
B
I
wonder
if
artifact
hub
intends
to
support
things
other
than
helm,
charts.
B
Ipmi
exporter,
that's
interesting:
okay,
yeah!
I
don't
yeah!
This
is
sort
of
still
kind
of
an
open
topic.
B
Okay,
other
items
is
some
pull
requests
that
are
open
shaytan.
I
think
you
have
one
that
on
measuring
ctl,
it's
waiting
to
conclude.
B
B
And
then
I
don't,
we
couldn't
hear
you
but
I'll
get
I'll
guess
not,
but
for
adena
and
gibril
as
you
go.
B
B
A
Okay,
other
than
this
like
are
there
any
topics
for
next
week
that
we
should
follow
up
on
the
github
actions.
Is
one
item
and
reaching
out
to
the
other
mint.
The
the
intel
team
is
okay
and
okay.
So
push
will
update
on
the
status
of
seed
seed
content
in
measuring
machinery
provider.
A
A
B
A
All
right,
thank
you.
Everyone
for
joining
we'll
see
you
in
the
other
meetings
this
week,
bye.