►
From YouTube: WebAssembly Filters Meeting (Nov 01st, 2021)
Description
WebAssembly Filters Meeting - November 01st, 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
So
welcome
everyone
to
the
watson
filters
meeting,
so
this
is
a
meeting
that
is
particularly
focused
on
initiatives
at
layer,
5
which
focus
on
web
assembly
filters,
and
we
have
a
couple
of
topics
that
are
from
last
week.
So
couple
of
action
items
from
last
week
and
people
have
been
working
on
it
over
the
week
and
to
start
us
off
scientin.
Would
you
like
to
give
us
an
update
on
the
github
action
for
measuring
patterns.
B
So
actually
I
was
working
on
this
vr.
I
think,
but
this
pr
has
a
lot
of
changes
which
I
have
done
in
my
own
repository.
So
basically
I
have
that
peer
has
got
a
lot
of
mistakes
over
there,
some
mistakes,
so
I
have
already
corrected
those
in
my
local
changes.
B
So
right
now
this
I
have
already
run
a
workflow
based
on
this.
My
current
action,
like
which
I
have
in
my
this
in
my
local
fork,
so
I
have
run
a
current
action.
So
it's
basically
working
like
the
setup
job
is
doing
its
work
now.
The
next
one
is
the
message
serial
pattern
action.
So
first
it
is
checking
for
the
kts
and
it's
it
is
connecting
my
ads
and
it's
also
deploying
the
kubernetes.
So
I
am
telling
so
this
is
the
mini
cube.
B
I
am
I'm
being
using
ubuntu
2004
in
my
github
workflow,
so
yeah.
So
basically
over
here,
the
micro
is
being
connected,
and
then
I
am
also
installing
the
machine.
So
my
measure
is
also
getting
installed
with
the
message
details
current
version,
so
it's
pulling
up
the
layers
and
after
this
I
think
it's
all.
It's
also
getting
a
visual
in
localhost
901,
but
the
problem
I
am
facing
is
that
I
am
trying
I
am
currently
trying
to
I'm
currently
trying
to
do
this
action
mystery
serial
pattern
apply,
and
I
am
using
this.
B
So
if
you
see
this
is
the
current
file
which
I
am
trying
to
deploy
in
my
measuring,
but
the
problem
that
is
coming,
I
I'm
still
not
able
to
figure
out
this
rbc
error,
like
it
says
no
matches
for
kind
and
my
filter
and
versioning
this
thing.
So,
basically,
right
now,
I
am
checking
out
this
error.
I
think
after
this
error
is
solved.
I
can
successfully
deploy
this
pattern
to
my
machinery.
C
So
since
then,
did
you
install
stu
in
the
overflow.
C
Mesh,
actually,
that
error
is
coming
because
you
did
not
initiate
the
installation
of
mish
officio,
so
these
custom
resources
are
probably
not
registered.
So
that's
what
the
it
is
complaining.
I
think.
B
C
And
so,
as
you
can
see,
the
type
is
onward,
filter
istio
so,
and
that
is
available
only
in
the
is
the
adapter.
So
for
this
particular
thing
only
actually
you
don't
have
to
do
it
manually.
You
can
do
it
in
the
pattern
file
itself.
We
can
say
in
the
pattern
files
that
these
things
depend
on,
so
your
workflow
can
stay
generic
while
pattern
file
would
get
specific
that
okay,
this
thing
is
going
to
rely
on
istio.
So
what
you
can
ask
your
pattern
file
in
any
pattern.
C
A
Quick
question:
like
what
all
configurations
are
we
exposing
to
the
user
right
now
in
the
action.
B
So,
basically,
right
now,
I
am
just
allowing
this
this
thing,
like
only
the
url
thing,
because
I
I
haven't
figured
out
about
like
giving
a
file
itself,
but
I
have
figured
like
I
have
almost
done
this,
giving
through
a
url.
A
Yeah
yep
so
yeah,
so
ideally
like
url,
so
we
should
be
able
to
deploy
with
url
a
pattern
name
or
a
pattern
file,
so
that
should
be
configurable
in
the
action.
I
guess.
A
And
like
do
we
also
I'm
not
sure
about
this,
but,
like
maybe
you
should
use,
this
should
also
be
able
to
configure
what
mesh
they
want
to
deploy
like
basic
the
basic
they
should
have
the
control
of
their
measuring
ctl
config
file,
yeah,
like
that,
can
be
done
like.
A
All
right
anyone
else
have
any
other
questions
or
suggestions.
D
A
I'm
not
sure
about
what
you
mentioned
like.
There
is
a
separate
branch
which
has
both
the
functionalities.
E
Yeah
so
earlier,
when
I
was
writing
e
to
e
test
for
mystery,
I
just
updated
the
action
a
bit
to
have
more
changes
so
that
you
can
also
use
in
cluster
and
out
of
cholesterol.
So
the
master
branch
doesn't
have
these
changes
right
now.
E
The
other
branch
has
them
it's,
because
there
were
some
things
that
I
was
not
sure
if
someone
would
really
need
when
using
the
action
like
having
control
over
what
version
I
mean,
if
they
want
to
deploy
stable
or
edge,
and
things
like
that,
so
yeah
I'll,
probably
remove
them
and
then
update
master
so
that
the
action
supports
both
in
cluster
and
out
of
cluster
deployments.
A
Yep
makes
sense,
and
so
cyan
then,
like
astrudrax,
updates
this
this
branch
and
maybe
merges
it
with
the
master
branch.
Maybe
we
can
propagate
those
configurations
across
other
actions
as
well,
so
that,
like
we
remain
consistent
and
like
users
have
will
have
more
configurability
of
their
action.
A
All
right,
thank
you.
Thanks
for
the
update,
so
the
next
next
step
is
utkarsh
like
we're
able
to
check
in
with
the
cypress
testing
yeah,
actually.
C
I
did
I
initially
forgot
to
reach
out,
but
I
then
did
reach
out
to
him.
I
haven't
yet
got
a
response
regarding
the
updates
any
if
there
are
any
so,
I'm
not
sure
yeah.
Basically
right
now,
I'm
not
sure
I
don't
have
updates
from
mario
at
the
moment.
A
D
Yeah,
I
think,
don't
think
that
we
should
wait
for
either
of
those
two
gentlemen.
So
vishal
has
been
here
for
a
little
over
a
year
and
I
think
he
made
he
made
one
attempt
to
contribute.
We
went
up
with
a
smackdown
code
in
ci
at
one
point
and
he's
volunteered
to
help.
D
I
don't
know
15
20
times,
and
so
I
wouldn't
he's
not
gonna
do
anything,
and
I
wish
he
were
on
the
call
so
that
I
can
tell
give
him
that
feedback
like
he
needs
to
do
what
he
says
he's
going
to
do
for
mario
yeah.
We
shouldn't
give
up
hope.
He'd
been
you
know,
he's
knowledgeable
in
this
area,
he'd
been
gung-ho
to
in
joining
the
community.
He
joined
a
lot
of
the
calls
and
and
his
engagement
had
abruptly
stopped.
D
So
I
wonder,
if
he's
you
know,
okay
or
I
wonder
if
something
hasn't
changed
for
him,
so
we
should.
We
should
try
again
I'm
I'll
I'll,
take
the
baton,
I'll
go
reach
out
to
him
and
see,
if
he's,
what
he's
up
to
outside
of
that,
though,
if
there's
anyone
else
on
the
call
that
would
like
to
try
their
hand
at
some
functional
testing
using
cyprus.
D
There
are,
you
know
yet:
there's
a
couple:
we've
given
a
presentation
about
so
rudolfo
ramir
rodolfo
martinez
vegas
is
a
maintainer
in
the
community.
He's
within
he's
focused
on
ci,
and
so
he
has
brought
cyprus
as
a
tool
she's
a
good
individual
to
engage
with
and
really
the
the
like.
I
think
there's
about
only
one
thing
that
needs
to
be.
D
You
need
to
bear
in
mind,
as
you
were
right
of
functional
test
in
cyprus,
and
that
is
just
that
you
want
to
try
to
make
it,
as
least
as
anti-fragile,
because
just
like
unit
tests,
these
types
of
tests
have
a
way
of
becoming
a
sustaining
burden
over
time.
D
So,
if
you
imagine,
if
any
of
you
are
familiar
with
like
front-end
based
testings,
some
of
the
tools
in
the
past
will
like
record
the
position
of
where
your
mouse
is
on
the
screen,
which
that
means
when
you
move
something
when
you
redesign
the
site,
you've
got
to
go
update
the
test,
because
that
mouse
position
is
no
longer
relevant
or
others
have
been
they've,
been
various
approaches.
The
approach
that
cyprus
is
using,
I
think,
is
more
broadly
adopted
now,
which
is
nice.
D
It
uses
just
a
unique
key
to
identify
to
id
a
given
element
on
the
dom
and
to
go
interact
with
that
to
do
to
invoke
whatever
funk
whatever.
If
it's
an
on
click
or
an
on
hover
or
whatever
it
is,
that's
appropriate,
so
just
try
to
try
to
make
them
so
that
they're
not
fragile.
This
is
the
thing
there's
prior
art,
though,
in
there
abby
that
you'll
be
able
to
use
as
a
base
for
creating
another
test.
D
Yeah,
you
know
another
one
would
be
like
abby
if
you
do
get
in
there
and
have
success
with
a
test
or
even
if
you
just
become
really
familiar
with
the
existing
tests.
This
is
a
potential
topic
to
present
at
like
a
community
meeting
to
just
educate
people
on
what
cyprus
is.
Why
we,
why
we're
using
it?
How
it's
helpful?
A
Oh,
thank
you.
Thank
you
thanks
lee
and
ravishek.
So
moving
on
to
the
next
topic
rudraksha,
could
you
give
us
an
update
on
the
seeding
of
filters
from
the
last
discussion
and
I
think
you
have
some
areas
where
you
need
help
so
could
you
also
mention
that.
E
E
E
This
is
the
updated
spec.
Yes,
I
just
did
changes
few
seconds
ago,
mainly
I
was
confused
about.
How
would
we
exactly
be
because,
because
the
challenges
that
we're
facing
here
is
not
overwriting
the
changes
that
users
make
and
still
making
sure
that
they
get
all
new
updates
and
yeah.
So
I
was
stuck
with
this.
I
don't
know
for
a
day,
but
I
just
updated
it
right
now
and
I
don't
I
don't
have
that
table
designed
yet
in
mystery
object
mode.
Yet
here
is
how
it
would
look
and
yeah.
E
E
D
Yeah,
okay,
well
so
part
of
what
redrock
is
having
to
deal
with
here,
isn't
necessarily
easy
that
bit
about
how
to
allow
people
to
customize
the
software
and
in
this
case
a
filters
configuration.
D
And
so
what
they'll
have
to
do
when
they
want
to
use
one
of
those
filters
in
accor
in
like
and
apply
a
different
config
is
that?
Well,
if
their
config
is
externalized
and
not
necessarily
directly
affiliated
with
that
the
official
filter,
they
might
have
config
that's
buried
in
a
pattern
file
so
that
they
may
be
versioning
that
and
storing
it
separately
from
the
filter.
D
Now,
okay,
when
there's
yeah-
and
you
know
what
it's
going
to
take
a
little
while
to
walk
through
all
the
use
cases
here
verbally,
but
we
can,
we
can
quickly
arrive
at
a
couple
of
different
ways
in
which
to
to
avoid
it
and
still
let
people
use
the
filter.
D
So
it
can
be
inconvenient
if,
if
the
user,
if
we
provide
the
user,
I'm
a
collection
of
seed
content
and
then
we
say
well,
no,
like
you,
can't
touch
that
directly,
like
you've
got
to
make
a
copy,
and
you
know
it's
not
as
okay,
if
we
for
fill
for
for
filters
specifically
if
the
binary
and
the
configuration
provided
at
runtime
to
that
binary,
if
if
we
just
treat
the
filter
as
like
the
official
like
the
filter
in
some
sample,
config
is
like
the
official
thing
in
that
sample
config.
We
don't
really
care
about
like.
D
If
people
can
orthogonally
like
inside
of
a
pattern
file
track
their
own
config
separately,
then
it
becomes.
D
The
concern
isn't
really
there.
Now.
I
think
I'm
oversimplifying
and
being
naive
as
I
state
this,
I
don't.
I
don't
know
that
we
can
verbally.
You
know,
digest
this
on
the
call.
Okay,
fine.
So
so
that's
one
approach
is
like:
don't
let
the
user
make
copies?
I'm
sorry,
don't
let
the
user
use
it
directly,
but
an
alternative
approach
is
to
just
go
ahead
and
let
them
modify
the
bejesus
out
of
the
thing
and
when
they
upgrade
is
to
provide
the
a
new
new
seed
content
that
is
versioned
for
that
new.
D
That
next
version,
so
you
know
the
the
worst
like
yeah
it's
interesting
because
for
a
pattern
like,
like
the
the
the
evolution
of
the
pattern,
may
build
in
experience
over
time
that
that
makes
it
a
better
pattern
and
people
would
say,
wait
a
second
like
I
want
to
benefit
from
that.
That
would
have
been
valuable.
D
D
For
that
I
think
that's
kind
of
specific
to
it.
I
mean
sometimes
for
the
filters
as
well.
If
they're
running
more
efficiently
people
may
you
know,
have
the
same
complaint
anyway.
It's
really
difficult
to
get
to
make
every
all
the
use
cases
happy
and
satisfied
without
building
a
fair
bit
of
versioned
life
cycle
management
across
in,
like
the
concept
of
extension
points
as
a
concept
really
revolves
around
these
same
problem
statements
as
well,
it's
like.
D
It's
the
same
thing
as
like
you're
trying
to
provide
the
content.
Content
is
an
extension
point
in
some
respects,
like
you
have
a
platform
as
capabilities
people
can
bring
new
content
or
if
a
platform-
and
it
has
an
area,
an
extensible
area-
that
if
you
want
to
bring
in
a
couple
new
hyperlinks
or
like
change
the
logo
out
in
the
corner,
go
ahead
and
provide
your
hyperlinks
or
provide
your
logo
and
what
you
try
to
do.
Is
you
try
to
carve
out
apis,
basically
extension
points
that
are
basically
apis.
D
That
say
whatever
you
want
to
do
inside
of
here,
we're
not
going
to
change
or
but
we're
going
to
be
ignorant
of
it.
Of
course,
there's
some
constraints
here,
so
you
can't
do
bitcoin
mining
off
of
the
you
know
this.
You
know
like
so
sorry,
so
stepping
back
there's
two
approaches,
one,
don't
let
them
make
changes
to
it,
just
version
it
every
release
and
so
give
them
the
new
the
new
versions
and
if
they
so
that
I
might
have
two
complaints.
The
one
complaint
is
hey.
D
I've
got
I've
upgraded
like
three
times
in
a
row,
so
I've
got
like
three
copies
of
this
filter
you're
like
that's
what
the
delete
button
is
for.
You
don't
want
it,
but
if
you
didn't
want
it,
here's
the
changes,
the
differences
and
you
might
also
say
well
yeah,
you
know
actually
there's
no
differences
between
them,
so
you
can
just
clear
out
your
system
delete
two
of
them
like
that's
a
minor
nuisance.
I
think
that's
like
that's
absolutely
fine,
certainly
for
now,
anyway.
D
The
second
issue
that
people
would
say
is
the
one
that
I
had
said
before,
which
is
like
well,
what's
what's
new
and
hot
about
that
new
version?
Oh,
it
runs
more
efficiently
boy,
I'd
like
to
gosh.
Why
didn't
you
just
give
me
that,
like
that,
of
course,
why?
Wouldn't
I
want
that
you're
like
well,
you
might
like
it
like
if
you
look
at
it.
D
If
you
take
the
example
of
a
pattern,
it's
like
well
best
practice
used
to
be
three
retries
for
every
service,
but
now
it's
two
and
we
don't
want
to
just
assume
that
you're
always
gonna
want
it
like
we're
gonna
and
that's
where
you'd
have
to
build
in
this
like
further
more
granular
life
cycle
management
around
that
and
it's
like
look,
we
can't
do
all
that
right
now.
We're
not
gonna.
Do
all
that.
D
D
Response
it's
one
that
can
be
improved
on,
but
but
it's
not
something
that
needs
to
be
solved
now
we
can
so
long
as
we're
avoiding
conflict
in
either
those
two
paths
we
can
come
back
to
enhance
with
questions
like.
Oh,
we
see
that
you
I
mean
like
one
of
the
things
you're
trying
to
do
right
now
and
if
you
can
achieve
it
quickly
and
easily
great,
that
is
you're
going
to
take
the
last
the
commit
sha
of
a
given
pattern.
D
D
So
if
you
can-
and
that's
that's
nice
because
then
that
eliminates
one
of
the
complaints
that
people
would
have
like.
Oh
I've
got
like
three
three
entries
of
the
same
filter,
just
the
different
versions
and
so
like.
If
they're,
if
you
do
that,
no
op
chat,
if
you
do
that
duplicity
check,
we
can
should
be.
That
should
be
relatively
straightforward.
We
can
reduce
the
number
of
like
this
thing
that
we're
talking
about
here
is
really
like
a
future
facing
problem.
Where,
like
that,
we
don't
really
have
right.
D
Now
we
can,
even
if
you're,
not
doing
that
get
shaw
check.
Fine,
like
let's
just
we'll
just
address
it
later.
I
think
when
I
mentioned
this
before
I
didn't
dish
out
this
comment
with
like
a
get
out
of
jail
free
card,
which
is
what
we
need
to
play
here
and
just
we'll
come
back.
You
know
we'll
come
back
to
just
a
good
note.
It's
just
a
good
like
entry
in
the
docs.
It
says
whatever
path
we
went.
E
Gotcha,
okay,
so
that
addresses
one
of
the
concerns
about
managing
versions.
The
the
other
is
about
like
the
current
implementation,
is
read
the
content
and
register
it
as
regular
patterns
and
filters,
so
so
that
it
can
be
delivered
with
the
general
api.
We
don't
have
to
make
changes
everywhere
and
rewrite
all
that
we
have.
We
already
have
patents
and
fitness.
So
for
that
we
are
just
using
the
same
tables
that
we
had
earlier,
but
that
comes
with
the
concern
that
admins
get
duplicate
entries
of
everything.
E
D
Remove
the
word
get
shot
and
just
say
unique
identifier
that
tells
you
if
the
ver,
if
the
thing
that
you're
looking
at
is
the
same
thing
as
the
other
thing
that
you're
looking
at
and
like
even
within
the
provider
implement.
You
know
like
in
this
case,
if
that's
a
hash,
which
is
what
the
shaw
is,
that
getcha
is
like
what
you're
able
to
do.
There's
a
couple
of
things.
That's
nice
about
that
approach,
as
opposed
to
like
oh,
if
that
comes
from
the
sanctified
official
repository,
then
we'll
do
something.
D
It's
like
wait
in
the
database.
How
many
times
do
you
want
to
represent
a
unique
object.
D
So
before
we
even
and
so
that
so
that
there
needs
to
be
indirection
between
the
content,
the
con,
the
object
model
that
you
haven't
got
just
gotten
to
yet,
which
will
help
clarify
this.
For
you,
I
think,
which
is
if
your
content,
then
you
have.
You
need
indirection
between
the
end
user,
which
is
this
end
user
belongs
to
or
has
a
role,
which
is
part
of
their
part
of
a
group.
That
group
has
assigned
a
role
which
either
the
role
allows
access
to
this.
D
E
D
But
but
I
guess
my
my
point
is
like:
never
never
will
you
have
or
should
you
have,
this
piece
of
content
is
always
belongs
to
this
one
user,
and
so,
if
another
user
wants
that
same
piece
of
content,
they
need
another
copy.
No,
no,
they
don't.
Why
do
they
need
a
copy
like
they
can
just
have
a
copy?
They
can
have
their
own
through
access
through.
So
it's
like
god,
what
is
it
in
memory
counters?
D
What
is
it
yeah,
if
you're,
using
a
language
that
doesn't
have
garbage
garbage
collection
on
memory
or
what
requires
a
memory
encounter.
D
E
A
All
right
any
other
comments
here
before
we
move
on.
A
Thank
you
draksha,
so
like
what
should
we
do
next?
What?
What
is
the
next
action
item
here
like?
Should
we
review
these
updated
spec
and.
A
All
right,
thank
you.
Thank
you,
daksha.
So
next
stop
we
have
seen
sane.
Were
you
able
to
check
on
this
centralization
of
configuration
so.
H
Yeah
from
our
previous
discussion,
we
me
and
we
had
discussion
about
what
needs
to
be
changed
in
the
pr
so
one
of
which
were
code
improvements.
So
there
were
some
code
improvements
suggested
by
abby,
and
I
have
implemented
that
one
issue
right
now
is
there's
a
functionality
issue.
The
application
is
not
actually
being
loaded
on
the
yaml
editor.
I
haven't
yet
been
able
to
figure
out
what
the
issue
is.
So
I
am
working
on
that
and
might
be.
I
might
need
help
as
well.
D
Nice
yeah
wait
remind
me:
what's
this
one
about
the
configuration
form?
Oh,
I
see
to
stop
repeating
ourselves
across
the
three
yeah.
D
A
All
right
next
up,
we
have
a
demo
from
gaurav
on
multiple
import
options:
github
options
for
patterns
filters
and
applications.
G
F
So
yeah
the
this
draft
pr
is
about
the
github
import
issue,
so
it
is
not
particularly
about
the
multiple
import
option
which
I
demoed
earlier.
So
this
thing
I
have
two
before
asking
the
queries.
I
will
just
show
the
basic
import
thing
that
we
have
further
in
the
ui,
but
I
have
two
questions
that
will
help
me
to
move
forward
with
the
issues
is
first
one.
Is
that
what
all
basic
components
or
fields
we
require
for
the
github
import
other
than
this?
F
Getting
the
url
and
directly
importing
it
from
the
remote
repository
and
second
one
is
that,
based
on
this
info
for
in
the
measures
and
service
mesh
patterns
about
when
the
user
is
logged
in
with
provider
or
none,
we
have
the
following
paths
and
urls
to
follow
and
using
the
method
post
and
all
so.
I
have
set
the
conditions
and
made
falling
changes
for
the
filters
only,
but
I
am
not
sure
would
that
be
a
valid
approach
to
follow.
G
F
I
mean
yeah
yeah.
My
first
query
is
about
what
all
fields
I
mean
the
components
like
we
have
import
and
all
required
in
this
model
for
the
github
import
plus.
How
can
I
add
the
fields
this
thing
this
url
or
do
we
need
certain
endpoints
or
I
can
directly
use
the
url
validator
that
we
are
using
for
the
machine
url?
C
Okay,
if
you
would
go
to
the
google
docs
that
you
were
sharing
yeah,
so
here
you
can
see
there
is
this
url
path
and
save
save
is
obviously
something
that
is
required
so
and
that
that
would
be
implicit
to
the
ui
like
if
someone
is
importing,
they
probably
do
want
to
save
it.
The
second
thing
is
the
path
that
is
also
required,
and
what
this
path
is.
C
This
path
is
basically,
if
they
don't
want
to
import
everything
from
entire
repository,
they
can
just
specify
a
path
that
okay
go
inside,
o
and
slash
door,
two
and
then
properly
get
everything,
or
only
only
just
reverse
this,
but
so
this,
in
short,
this
path
is
a
required
field.
What
would
be
the
best
ux
to
show
them?
That's
not
something
I'm
sure
about,
but
this
would
be
a
required
field,
or
this
could
be
implicit.
That
is,
we
don't
expose
this.
C
Probably
we
should,
but
we
don't
expose
this
and
we
say:
okay,
we
are
going
to
just
recursively
traverse
the
entire
repository.
So
in
short,
yes,
path
is
a
required
field,
and
in
the
url
url
looks
a
bit
different.
You
can
see
so
https
code
in
slash
flash
github.com
repository.
That
is
the
common
like.
That
is
how
github
urls
are
the
slash
branch
that
is
in
the
end.
That
is
something
you
can't
have
a
field
up,
for
that
is
what
branch
do
you
want
us
to
traverse?
C
The
default
is
mean
for
now,
but
if
so
yeah,
so
this
is
something
that
you
can
probably
have
a
feel
for,
but
you
can
say
that
okay
main
is
the
branch
name
that
we
are
going
to
consider.
If
you
want
to
override
that
specify
that,
and
you
can
append
that
in
the
end,
whatever
branch
thing
you
would
give
that
would
be
respected
unless
until
you
don't
give
any.
In
that
case,
we
will
traverse
me
so
yeah.
I
think
two
fields
are
three
fields
are
required.
G
F
Yeah,
okay,
so
this
endpoint
will
be
changed
according
to
the
data,
endpoint
would
remain
same,
the
post
body
would
yeah.
Okay,
I
got
it
the
post
body
and
for
the
ux
exp
ux
thing.
F
C
Actually,
I
don't
think
we
should
expose
that
this
is
the
way
we
internally
consume.
Url.
You
can
probably
just
have
https
github.com
something
and
internally,
when
you
actually
send
the
body
you
can
augment
it
to
to
how
fee
how
the
api
internally
expects
it.
So
yeah,
that's
problematic.
In
that
case
we
can
reuse
the
previous
url
validators.
F
C
A
So,
let's
move
on
to
the
next
topic,
so
I
guess
meghna
is
not
on
the
call
and
in
the
last
meeting
we
were
looking
to
find
some
area
for
meghana
to
get
involved
in
in
in
this.
So
I
I
don't
know
if
we
have
any
updates
on
that.
So
lee
like
do
you
have
any
updates.
D
No
like,
like
there's
a
no
gaurav,
opened
up
like
a
good
first
issue
for
calculating
the
number
of
filters
or
any
piece
of
content.
That's
shown
in
the
table,
gar
of
that
that's
fantastic.
If
no
one
signs
up
in
a
day
and
a
half
from
now,
then
please
just
take
that
issue
and
and
implement.
But
it's.
D
Like
pranav
who's
on
the
call,
potentially
kk
who's
on
the
call
either
the
two
of
them
might
want
to
pick
that
up
and
give
it
a
shot,
and
if
they
do
they're
looking
for
a
starter
issue
in
this
area,
they
should
go,
see
you
and
see
the
one
that
you
just
opened
for
megana.
I
don't
think
we've
expressed
like
given
her
necessarily
clear
guidance.
I
think
maybe
it's
time
for
her
to
shift
her
focus
to
the
new
ui.
D
So
this
is
a
healthy
reminder
to
navendu,
as
you
outlined,
the
mastery
development
meetings
agenda
to
include
just
like
you're,
including
a
standing
section
for
measuring
ctl.
The
new
ui
will
be
an
ongoing
theme
for
a
while,
so
it'll
be
there's
a
tracking
spreadsheet
that
this
should
be.
You
know
he
should
have
a
section
where
he's
covering
that
aditya.
If
we
can
get,
is
it
on
no
okay?
D
D
On
the
next
question,
I
I
just
had
this
general
question
and
it
was
you
know
we
we
would
cars.
We
have
this
ability
to
deploy
these
awesome
filters
to
envoy
based
data
planes,
and
do
we
still
have
like
refresh
my
memory?
What
what
are
we
at
it?
Where
are
we
at
in
terms
of
implementation
compatible
with
all?
C
Right
now,
the
support
is
very,
very
minimal
right
now
we
are
only
supporting
istio.
We
don't
have
a
high
level
construct,
which
yeah
we
don't
have
high
level
construct
which
can
operate
on
all
of
the
service
meshes,
that's
one
of
the
things.
The
second
thing
is
that
even
the
because
we
don't
have
a
high
level
construct,
so
even
the
pattern
which
is
capable
of
deploying
filter.
That
is
also
very
like
it's
using
it's
using
all
of
the
constructs
of
s
two
in
order
to
so.
C
Basically,
if
you
would
read
the
pattern,
you
would
be
able
to
see
that
we
are
actually
directly
leveraging
its
use
construct
and
there
is
no
upper
a
higher
level
construct
for
configuring.
Any
of
that
so
yeah,
that's
the
state
at
the
moment,
and
that
is
why
we
are
actually
right
now.
Only
supporting
istio
service.
D
D
Did
you
guys
get
those
same
at
t
commercials
like
non-stop
reach
out
and
touch
someone
like
yeah,
I
think
before
cellular
was
the
thing
anyway.
So
we
did
we're
touching
two
intel
engineers
whose
names
I'm
gonna
have
a
hard
time
pronouncing,
but
but
they
we
just
got
connected
with
them,
and
so
we're
gonna
try
to
get
them
scheduled,
hopefully
for
next
week.
Hopefully
this
time
works
for
them.
D
The
top
the
project
they're
working
on
is
the
webassembly
system
interface
for
neural
networks,
which
is
potentially
well,
which
in
concept
is
highly
related
to
one
of
the
abilities
that
we
desire
for
measuring
to
have,
and
that
is
to
calculate,
to
characterize
the
performance
of
your
service
mesh
from
within
each
and
every
filter
from
within
each
and
every
sidecar.
D
D
I
don't
know
it's
a
little
bit
of
a
long
shot.
I
think,
where
it
becomes
potentially
really
interesting
is
when
you
take
into
account
use
cases
like
locality-based
routing.
So
if
you're
trying
to
facilitate
failover
for
certain
services
that
might
be
failed
or
capable,
while
others
might
not
be
or
when
you
go
to
incorporate
services
that
span
not
only
kubernetes
but
edge,
the
latency
involved
there
and
the
routing
that
goes
on
with
something
like
that.
D
It
brings
in
a
lot
of
brings
in
room
for
this
type
of
these
types
of
performance
characterizations
to
like
actually
take
action
on
them
and
have
the
action
be
meaningful
so
good,
so
we
did
yeah.
We
got
in
contact
with
two
new
intel
engineers,
so
we'll
see
we
seem
to
have
a
lot
of
intel
folks
coming
to
the
calls
which
is
fantastic.
D
D
So
serverless
web
assembly
for
compute
intensive
workloads
at
four
yeah
that
one
oh
wait
hold
on
hold
on.
Oh
that's
funny!
I
didn't
realize
if
you
go
back
one
I
guess
in
the
venue.
I
didn't
realize
that
those
two
intel
engineers
they
did
speak
at
the
conference
just
right
there
at
3,
35.,
okay,
cool,
so
they'll
have
fresh
content
and
that's
great.
D
D
D
But
yeah
cool.
It's
it's
really
interesting
to
see
web
assembly
programs,
you
know
be
being
a
a
right
once
run
many
a
warm
aurora
run
anywhere.
I
guess
aurora
right
once
run
anywhere,
but
but
to
have
people
using
examples
like
in
serverless
environments.
That's
really!
That's
heartening!.
D
So
other
than
the
venue
has
anybody
studied
a
little
bit
of
neural
networking.
D
A
D
All
right
anything
else
today,
oh
hey,
I've
got
something
I
forgot.
I
just
saw
azure
on
that
page.
This
is
something
that
I'll
announce
in
the
community
meeting,
but
but
azure
is
kind
enough
to
give
us
ten
thousand
dollars
in
credits
on
their
platform,
so
I've
got
I've
got
at
least
like
half
a
bitcoin
that
I'm
looking
to
so,
which
is
good
because
there's
some
aka
there's
a
bunch
of
compatibility
testing
that
needs
to
be
done.
Aks
in
specific,
and
so
we'll
talk
more
about
that.
D
Oh
and
for
open
service
mesh
right
for
smi,
sorry
for
s,
mic
and
performance
testing
on
open
service
mesh
which,
like
there's
no
doubt
or
there's
very
little,
doubt
that
microsoft
will
eventually
offer
an
open
service
mesh,
offering
right,
whether
that's
its
own
standalone
service
on
azure
or
a
checkbox
on
your
aks
config,
or
something
like
that.
That's
I'll!
Take
that
bet!
A
All
right
does
anyone
else
have
anything
else
to
discuss
before
we
close
this
meeting.
B
So
actually,
I
think
utrecht
just
told
that
these
two
were
missing
due
to
being.
History
was
not
installed
in
my
system,
but
in
the
morning.
What
I
did
was
that
I
took
this
same
two
commands
like
I
have
my
I
have
already
done
a
desktop
and
over
here
I
think,
I'm
being
using
mini
cube.
So
I
have
just
copy
pasted
code
and
it's
just
installed
my
message
and
then
I
used
this
command
like
this
exact,
same
command,
message,
detail
pattern
apply
in
this
command
and
you
can
see.
B
Like
yeah
this
one,
this
image
this
I
have
uploaded
from
there
only,
but
it
didn't
have
any
error
over
there
so
like.
What's
the
difference
that
this
from
actually
sequel,
it's
not
creating
any
error,
but
that
workflow
it's
creating
an
error.
So
what
might
be
the
difference
between
these
two.
C
It
could
be
that
you
either
have
istio
installed
or
you
had
sd
installed
and
because,
right
now
we
don't
clear
out
the
custom
resource
definitions
crds
because
we
are
using
10
base
installations.
So
your
cube
api
server,
probably
like
basically,
your
community
still
has
those
crds.
So
if
you
can't
just
do
cubecut
it'll
get
customer
resource
definitions
and
if
you
see
those
custom
customers
in
your
environment.
Probably
that
is
the
reason
you
didn't
see.
The
error
in
your
local
environment.
A
Lee
about
the
questions
like,
I
will
just
do
some
quick
research
to
make
sure
that
they
are
not
that
trivial,
but
it
surrounds
things
related
to
like
I
I
bet
like
we
will
be
using.
We
will
be
exploring
tensorflow,
so
that
is
the
like
best
thing
to
use
in
production.
So
comparing
the
performance
running
natively
on
tensorflow
and
awesome
and
like
maybe
if
they
have
some
experience
with
that.
But
I
will
share
those
on
slack.
A
All
right,
so
I
guess
that's
it
for
today's
meeting
and
we'll
follow
up
on
some
of
the
stuff
that
we
discussed
next
week,
and
maybe
some
of
these
will
be
discussed
on
the
machinery
development
call
later
this
week.
So
I'll
see
you
guys
there.
Thank
you
for
joining
and
bye.