►
From YouTube: WebAssembly Filters Meeting (Sept 20th, 2021)
Description
WebAssembly Filters Meeting - Sept 20th, 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
today
is
september
the
20th
and
we
have
a
couple
of
topics
to
discuss
from
last
week.
So
a
couple
of
action
items
from
last
week.
So
we
will
be
discussing
that
today
on
the
call
as
well
and
so.
A
To
like
all
our
meetings,
learfa
community
meetings,
this
meeting
is
also
recorded
and
this
video
will
be
published
on
youtube.
With
that
said,
let's
jump
into
topics
and
lee.
Would
you
like
to
talk
about
garrett's
question
yeah.
B
B
Okay,
so
so
thanks
for
providing
some
good
learning
content
for
us,
so
I
think
there's
a
couple
of
things
that
we'll
be
able
to
speak
to
with
standing
on
confident
ground
and
there's
some
other
things
where
we'll
probably
find
some
volunteers
on
the
call
to
do
a
bit
of
research,
because
in
part,
the
questions
that
you're
asking
are
just
like
good,
good
fodder
and
for
us
to
good
justification
to
go,
look,
go
off
and
learn
some
things.
So
so
on
there's
a
couple.
B
If
I,
if
I
may
try
to
break
down
the
question,
maybe
into
a
couple
different
a
couple
of
different
things.
One
is
that
at
the
very
end
of
this
there's
a
well
I
don't.
I
don't
know
that
the
the
last
portion
of
your
investigation,
you
know
the
last
portion
of
your
line
of
thought
is-
is
so
much
about
literally
the
proxy
or
the
filters
themselves.
B
But
if
it's
about
them,
caching,
the
the
requests
well,
the
like
caching,
that
particular
user's
token
or
caching
that
particular
sessions
token,
as
much
as
as
much
as
the
caching
here
is
more
about
stuffing
that
jot
stuffing
that
jwt
full
of
some
additional
info,
such
that
it
itself
behaves
as
something
of
a
cache
and
it
sort
of
carries
its
cache
around
as
each
microservice
were
to
interact
with
a
request
that
has
that
jot
associated
the
info
is
present
already
and
so
there's
sort
of
an
implicit
cache
happening
there.
It
like.
C
Yeah
yeah,
so
I
can.
I
can
I'm
apologizing
again,
but
I
I
can
read
from
from
the
customer's
kind
of
requirement
would
be
so
they
said.
The
enrichment
of
the
jwt
in
their
case
would
be
calling
an
internal
service
with
the
oauth
2
uuid
and
receiving
user
group
commission
data
back,
which
would
be
cash
for
the
lifetime
of
the
session
and
invalidated
on
updates.
C
So
I
think
it's
rather
than
then
going
back
to
this
internal
service.
Every
time
there's
potentially
some
caching
there
for
a
lifetime.
Does
that
make
sense.
B
Yeah
yeah,
which
actually
sounds
it
sounds
a
little
bit
easier
to
achieve
or
well
rather
something
that,
for
my
part,
I'm
a
bit
more
familiar
with
than
than
sort
of
generic
caching
of
http
requests,
and,
and
so
so
anyway.
So
the
good
news
here
is
yeah.
We
actually
there
was
another
and
there
still
is
another
user
in
the
community.
B
I
don't
know
if
samir
is
on
today,
if
he
is
we'll
pick
on
him,
if
not
it's
not
as
fun
to
pick
on
him,
but
anyway,
and
so
samir
had
come
with
real,
really
a
fairly
similar
request
in
this
case,
basically,
his
request
was
to
be
able
to
take
a
jaw
to
take
a
token
and
manipulate
it
like
based
on
some
info.
That's
inside
of
it
go
ahead
and
extract
some
info
out
and
then
shove
some
info
back
in
and
and
so
a
very
similar
pattern
in
this
case.
B
In
his
case,
it
wasn't
about
literally
shoving
in
like
a
role
along
with
specific
permissions
that
that
role
or
that
that
user
might
have,
but
in
the
end
it
ends
up
being,
but
that's
neither
here
nor
there
in
some
respects,
because
all
your
that
that
that
identity
system
or
that
authorization
system,
when
it
hands
back
that
info
yeah
or
rather
I'm
being
long-winded.
I
guess
the
first
part
of
this
is
yeah.
B
That
sounds
entirely
feasible,
like
that
sounds
like
some
a
bit
of
work
that
the
community
just
ended
up,
ended
up
doing
recently
and
tried
to
make
this
particular
filter.
Configurable
like
that.
That
was
part
of
the
point.
Was
it
like?
Hey
the
while
samir's
use
case
was,
while
we
were
trying
to
facilitate
that
and
achieve
it,
we're
trying
to
also
make
it
generic,
because,
because
it's
you
know
it's
not.
B
You
know
in
some
respects
you
could
sort
of
compare
the
concept
of
baggage
with
distributed
tracing
as
kind
of
a
similar
thing
that,
like
you,
want
to
prop.
You
want
to
take
a
bit
of
info
and
put
it
and
propagate
it
across
each
of
these
requests
and
there's
one
of
those
way
yeah
one
of
those
ways
you
get
that
token
sitting
right
there.
It's
capable
of
holding
things
like
great
as
a
matter
of
fact,
the
the
tool
that
we
talk
about
in
this
call.
B
B
It's
the
set
of
contributors
is
desirous
of
measuring
operating
in
something
of
a
similar
fashion,
so
that
it
can
be
somewhat
dynamic,
so
the
code
can
just
respond
to
who
it
is
that's,
making
the
request
and
they
see
the
appropriate
buttons
or
you
know
that
they
see
the
appropriate
things
to
interact
with
or
not
based
on
the
permissions
in
there.
So
so,
yes
is
the
answer
to
your
first
question
like
yes
and
there's
a
sample
token
to
to
look
at
its
behavior
and
see
how
closely
it
fits
with
the
use
case
that
that's
being
described.
B
The
caching
portion
that
I
was
doing
a
bad
job
of
like
trying
to
clarify
was
that
was,
I
think
you
were
just
reaffirming
that
like
yeah,
hey
we're
not
talking
about
each
or
not
necessarily,
maybe
that
ends
up
being
the
case,
but
but
to
start
with
it
wasn't
like
the
customer
was
describing
each
individual
like
if
you
have
10
serp,
if
they
have
10
services
and
10
side
card
proxies
that
each
proxy
needed
to
literally
hold
a
cache
of
like
what
that
authorization
information
was
for
each
individual,
unique
user
request.
B
You
know,
because
that
same
user
would
have
the
same
permissions
across
the
it
wasn't
like
that
needed
to
be
cached
inside
of
the
filter
itself
inside
of
the
proxy
itself,
but
rather
it
sounded
like
that.
The
the
caching
in
this
case
was,
you
know
once
the
token
has
interfaced
with
that
sort
of
central
authorization
microservice,
it
was
like
hey.
Okay,
great
you've
been
here's
the
info
that
that
all
of
the
rest
of
the
the
the
upstream
services
that
you
might
interact
with
here's,
here's,
your
key!
Here's,
your
setup!
Here's
your
keychain!
B
Like
here's,
your
set
of
permissions!
You
just
take
that
with
you
and
then
that
way,
those
microservices
don't
have
to
come
back
and
hammer
on
what
will
become
a
bottleneck
in
the
system.
Eventually,
this
central
auth
service,
which
is
also
just
also
just
kind
of
inefficient
and
like
yeah
and
totally
like-
that's
like
no
doubt,
there's
some
there's
some
devil
in
the
details
of
some
of
this,
but
that's
and
anyone
else
on
the
call
so
that
was
working
on
the
jwt
9000.
B
The
transformer
9000
is
what
they
ended
up,
calling
it
does
anyone
want
to
point
out
where
I'm
overlooking
an
aspect
of
what
gareth
is
describing
and
that,
like
this
filter,
like
is
ill
positioned
to
handle
his
use
case.
B
Is
it
so
so
I
think
that
there's
there's
a
ton
of
power
in
the
intelligence
of
the
network
like
if,
if
you've
got
a
fine
grain
control,
this
layer,
7
control
over
requests
that
are
going
by
and
you're
able
to
dynamically
load
filters
like
the
one
that
we're
talking
about
that
will
do
a
special
function
for
you,
a
network
function,
so
to
speak.
B
That
great,
like
so
we're
trying
to
enhance
the
a
relatively
recent
cncf
project
meshery
as
it's
a
service
mesh
manager
when
some
of
the
folk,
the
contributors
on
the
call,
are
going
to
talk
about
how
mesherie
in
its
next
release
like
it,
will
go
to
this
repository,
which
has
this
filter
plus
a
couple
of
other
filters.
It
will.
B
You
can
tell
measuring
things
like
hey,
I'm
running
a
collection
of
20
microservices,
here's,
the
10
that
should
get
this
jot
transformer
filter,
here's
the
10
that
should
get
the
I
don't
get
the
the
other,
the
whatever,
or
get
no
filter
or
whatever
or
maybe
there's
a
chain
of
two
filters
like
it
needs
to
get
the
job
transformer
filter,
but
also
this
subsequent
one
that
it
should
pass
through
and
that
those
are
the
kinds
of
use
cases
that
mashery
is
chasing
after
and
so.
B
Yeah,
I
don't
know
that
I
mean
so
so
you
know
we're
all
still
collectively
learning,
so
some
of
us
are
hesitant
to
point
out
where
I'm,
where
I'm
wrong
on
a
call,
maybe,
but
but
to
my
knowledge
like
nope,
like
you
just
redescribed
the
like
the
same
use
case
that
we
were
trying
to
make
that
transformer
generically
capable
of.
B
I
think,
if
for
nothing
else,
my
sense
is
that
after
your
time
here,
you'll
you'll
step
up
you'll
step
back
from
me,
like
hey
yeah,
my
my
intuition
was
correct,
like
these
are
the
types
of
capabilities
that
are
also
you
know
well
aligned
for
these
are
also
service
mesh
use
cases,
and
so,
as
you
double
double
click
into
that
there
are
so
the
community
ended
up.
Writing
this
the
particular
field,
this
particular
jwt
manipulator.
B
For
for
that
use
case,
I
guess
what
I'm
saying
is
like
if
this
doesn't
work
for
the
use
case,
that,
as
you
collect
more
details
or
as
you
sort
of
explore
it
more
we're
really
happy
to
try
to
change
this
or
make
a
new
one.
Make
a
copy
of
this
one
make
a
new
one
to
try
to
meet,
to
try
to
get
that
to
work.
B
Cases
will
be
very
similar,
but
be
nuanced,
and
so
what
we
want
for
is
like
for
someone
to
be
able
to
go
into
measuring
and
then
say:
oh
actually,
here's
the
string
that
you
that
the
filter
needs
to
look
for
when
it
sees
that
that
string.
Maybe
it's
like
user
account
or
something
like
replace
it
with
something
else
from
a
different
portion
of
the
http
transaction
or
yes,
we
desire
for
it
to
be
configurable.
B
Okay,
yeah
there's
was
what
other
aspects
of
what
gareth
said
are
worth
exploring,
so
gareth
you're
right
on
the
cusp
of
like,
like
you,
can
manually.
B
D
So
yeah
the
first
one
to
give
some
context
for
people
who
are
not
familiar
with
this.
The
first
issue
that
I'm
talking
about
is
a
fix,
fixing
the
url
import.
So
recently
we
added
the
feature
called
we
can
import
patterns
filters
and
even
applications
using
url.
So
we
can
get
the
data.
We
can
upload
the
data
using
that
url.
D
So
initial,
pr
that
that
was
added
to
add
this
feature
was
not
functioning
properly,
so
I
think
michael
corrected
it
for
applications
and
I'm
working
on
the
patterns
and
the
filters
part
of
it.
So.
D
Yeah
in
the
pr,
so
I
I
did
a
similar
approach
that
I
think
I'm
hoping
it's
michael.
So
I
I
did
the
same
thing
that
he
has
done,
but
the
thing
is
the
samples
that
I
have
for
patterns
and
filters
are
a
little
different.
So
if
you
can
scroll
up
a
bit
once.
D
So
for
the
pattern
sample
yeah,
it
is
just
it
just
opens
to
a
complete
data,
the
the
it's
the
data
is
shown
like
this,
but
for
the
filters
one.
What
happens
is
that
it
triggers
a
download,
so
I'm
not
sure
how
to
handle
it.
So
currently,
if
that's
filters
part
is
not
working.
D
So
I
added
for
the
patent
patterns.
It's
working
fine,
so
I
only
work
checked
it
with
the
two
samples
that
I
have.
That
is
one
thing
and
there
are
some
chicks
failing
there.
So
I
need
some
help
with
that
too.
So
that's
pretty
much
it
for
that
issue.
A
Oh
I
mean
like
I,
I
can
speak
for
the
filters
like
so
for
now.
I
think
you
can
try
the
all
the
correct
operations
with
a
pattern
file
ml
file
as
we
as
we
update
the
filters
ui
to
function
differently.
We
can
make
changes,
so
maybe
you
can
try,
try
the
filters
with
us.
The
same
pattern
fan
as
well.
That
should
work
now.
A
B
B
The
yeah,
because
it
looked
like
the
error
in
there,
was
saying
that
there's
a
particular
package
particular
module
that
needed
go
116,
and
the
thing
is:
is
that
that
environment
should
be
running,
go
116..
So
it's
it's
more
of
just
an
update
in
the
the
workflow
itself.
That
needs
to
make
sure
the
action
setup
go
using
the
right
version.
A
D
Yeah,
so
once
we
try
to
delete
items
or
anything,
for
example,
then
we
get
a
prompt
which
says
like
yes
and
no
thing
right
now.
Both
of
them
have
the
same
kind
of
buttons
and
they
suggested
that
it
will
be
better
for
the
us
if
we
make
the
s
in
the
background
color
and
know
as
the
one
like
transparent
background
which
won't
you
know,
it's
rarely
used,
so
we
keep
it
trans
transparent.
So
for
that
the
thing
is
the
component
we
are
currently
using
for
the
prompt
has
something
like
once
again.
D
I
didn't
go
through
all
of
them,
so
the
thing
is
if,
if,
if
we
want
one
as
the
most
primary
button
as
so
background
one,
what
we
can
do
is
like,
we
only
go
through
the
first
n
minus
one
once
and
then
for
the
last
one.
We
can
just
make
it
as
a
background
background
as
a
dark
one,
or
something
like
that.
That
is
one
thing
or
if
we
are
generally
only
having
two
options,
we're
just
using
two
options
for
all.
D
Then
we
can
just
remove
this
and
just
use
it
as
a
to
one.
That
is
the
thing
I
didn't
edit
anything
right
now
because
of
the
component,
how
it's
built
right
now,
so
I
just
want
some
inputs
on
that.
How
do
you?
Oh?
I
think
I
should
proceed
for
this,
so
here
yeah,
so
it's
taking
options
like
this
csm
no,
and
it
just
creates
buttons
for
each
of
the
options.
D
I'm
not
sure
actually
like
it's
used
in
different
places,
so
I'm
not
sure
how
it's
being
used
in
that
different
ones.
As
far.
E
Yes,
and
no
for
now,
like
everywhere,
we
have
used
that
component.
I
think
at
four
places
one
is
application
filters,
patterns
and
performance,
so
it
all
has
the
yes
and
no
option.
That's
it.
Yeah.
G
In
addition
to
that
yeah,
in
addition
to
that,
what
in
patterns
applications
and
filters,
we
are
using
that,
but
we
are
passing
the
array
values
as
of
objectives
in
filters
and
applications
everywhere,
and
we
are
configuring,
the
yes
and
no
option
by
conditions,
so
that
thing
can
be
changed
in
the
model
itself.
G
But
yeah,
I
said
that
it
is
using
a
single
component
for
a
button,
so
yeah
that.
H
G
Is
quite
tricky
to
have
a
change
in
primary
and
you
can
do
that.
D
So
so
just
to
break
it
again,
so
you
want
me
to
change
the
component
itself
to
accepting,
like
I
mean
it
has
as
if
it's
only
getting
two
inputs
and
options
right.
So,
let's
hope.
B
Nice
megan
that's
great
by
the
way,
just
as
a
relevant
example
or
relevant
topic,
so
of
the
prompt
that
you
were
talking
about,
that
where
the
user
needs
to
choose
yes
or
no,
and
how
you're
looking
at
making
the
sort
of
the
I
don't
know
if
you'd
call
it
the
default,
but
the
most
the
primary
choice
like
if
they
chose
to
delete
something
from
the
ui,
then
they
probably
meant
to
do
that,
but
we're
prompting
them
just
to
make
sure
that
it
isn't
the
mistake
and
so
that
that's
good,
so
probably
probably
they're
going
to
answer
yes.
B
And
so
if
we
move
yes,
but
just
like
I'm
just
kind
of
repeating
some
of
the
stuff
you
just
said,
but
if
we
move
yes
to
the
right
hand,
side
sort
of
like
this
is
set
up
here,
that
there
would
be
a
no
and
a
yes
and
they
would
they
would
be
styled
in
a
similar.
You
know
similar
to
this.
I
was
bringing
this
up.
B
To
give
an
example
of,
I
don't
think
that
the
prompt
prompt
component,
prompt
confirmation
component
is
being
used
here,
but
I
just
bring
it
up
as
an
example
of
where
a
tertiary.
B
B
You
know
I,
by
the
way,
I
think,
just
to
help
with
consistency.
This
would
be
my
fault.
I'm
probably
been
fickle
on
this,
but
if
you
don't
mind
capitalizing,
the
yes
and
the
no,
I
guess
we
capitalize
in
the
other
areas,
so
that
that's
my
fault,
I
mean
grav.
Had
it.
D
F
D
A
H
Okay,
so
my
topic
is
okay,
so
we
had
already
done
the
registration
of
seed
content
for
patents
and
filters.
The
applications
is
left.
Rudraksha
is
an
open
pr
where
he
has
seated,
because
in
the
case
of
applications,
there
would
be
different
applications
for
a
different
service
meshes.
So
this
recent
pr
that
that
rash
has
made
he
has
we
have
we
have
these
book
info
http
bin
and
emoji
applications.
H
These
three
are
istio
specific
applications,
so
we
would
have
a
pattern
file
for
these
three
applications
and
it
would
be
seated
inside
the
easter
adapter
and
similarly,
each
adapter
would
have
its
own
application
and
what
we
are
hoping
is,
after
this
pr
gets
merged
and
after
some
changes
in
the
mesh
kit
at
the
runtime,
when
the
adapter
starts,
what
it
does
is
it
registers
its
capabilities
with
the
messy
server
and
while
it
does
that
it
will
also
register
these
applications
with
the
message
server
right
now.
H
There
is
no
end
point,
I
think,
in
the
messy
server
for
that.
If
I'm
not
wrong
or
we
can
use
the
existing
endpoint,
okay,
we
can
use
the
existing
endpoint
or,
if
you
want
to
avoid
all
that
token
thing,
then
we
can
create
a
new
endpoint
that
that
is
a
topic
of
discussion,
so
that
endpoint
would
just
take
these,
take
these
applications
and
depending,
and
it
would
figure
out
that
okay,
so
adapters
have
sent
these
applications
and
I
need
to
see
them
and
then
the
machines.
H
I
would
see
those
applications
inside
of
the
inside
of
the
database
and
then,
when
you
will
log
in
with
the
local
provider,
you
will
be
able
to
see
applications
as
well,
along
with
filters
and
along
with
filters
and
patterns,
so
the
so.
This
is
where
we
are
right
now,
with
the
with
this
thing.
B
Cool
does
that
make
sense
to
everybody
else
that
well,
that
the
adapters
are
vessels
they're,
they're
packaging
mechanisms
for
service
mesh,
specific
functionality
that
you
know
each
one
while
they're
similar,
and
they
share
a
lot
of
common
functions
that
each
one
packages
up
things
that
are
specific
to
that
mesh
and
that
they
would
continue.
And
so
today,
they
package
up
sample
applications.
B
Today,
if
you
take
the
linker
d,
adapter
it'll
probably
have
the
emoji
voto
app,
because
that's
the
linker
d,
you
know
sample
app.
It'll,
probably
have
the
istio
book
info
app
because
that's
another
prominent
sample
app
so
that
you
can
try
that
sample
app
on
linker
d
and
then
probably
have
http
bin.
Oh
yeah,
actually,
you've
got
the
three
listed
right
there,
so
good.
Now,
if
you
take
the
link
or
d,
I'm
sorry.
B
If
you
take
the
console
adapter,
it
probably
has
those
same
three
or
four
and
the
same
three
or
four,
and
so
actually
we
end
up
we're
being
repetitive
across
them
and
that's
you
know
intentional,
because
because
maybe
someone
is
just
using
one
adapter
so
and
if
that
adapter
doesn't
come
with
any
sample
apps,
then
the
user
doesn't
get
any
test
apps.
You
know
sample
apps,
okay,
well,
okay!
So
how
do
we?
How
do
we
give
the
users?
B
B
And
that
was
that
was
one
of
those
things
where
I
didn't
really
have
a.
It
wasn't
a
rhetorical
question
or
it
wasn't
like
okay,
so
a
couple
of
ways
to
achieve
better.
I
was
like
well
look.
It's
still
going
to
be
the
case
that
people
can
optionally
choose
to
load
up.
Just
you
know
one
adapter
or
two
adapters
whatever,
so
they
can.
You
know,
okay,
all
right.
So
if
we,
if
we
de-duplicate
the
apps
from
each
adapter,
then
if
someone
only
loads
up
one
adapter
they're
not
going
to
get
all
the
sample
apps.
Okay,
fine!
B
B
B
B
I
Yeah,
actually,
if
patents
and
filters
are
being
centrally
seated
and
server,
then
why
applications
would
like
so
either.
The
concern
could
have
been
that
what,
if
a
certain
certain
component
of
patent
doesn't
exist
and
how
would
we
know,
but
but
if
patents
are
being
centrally
seated
and
if
that
concern?
Basically,
if
we
are
not
dealing
with
dartmouth
concerned
with
patents
or
with
filters,
then
applications
are
sort
of
similar
construct.
I
So
definitely
if
we
see
that
centrally,
that
is,
if
it's
only
available
to
server,
there
is
a
concern
that
whatever
component
of
patent
is
a
confidence.
Application
is
absent
from
from
from
the
store
from
the
registry,
but
the
same
thing
actually
happens
with
patents,
and
if
we
do
this,
that
is,
we
actually
don't
think
stuff
unload
applications
centrally
they
are
in
machine
server.
Then
we
avoid
duplication
because
machine
service
anyhow,
definitely
it
could
be
a
bad
user
experience.
I
That
is
you
applied
on
applications
and
machine
service
telling
you
that
we
cannot
do
that
because
one
of
this
component
is
not
present,
but
it
won't
mess
up
with
the
infrastructure,
because
mr
would
take
care
of
identifying
that
this
particular
company
is
actually
absent.
So
so
now
we
get
back
to
the
same
concern.
We
don't
get
to
know
in
once.
We
list
them
out,
but
if,
if
the
user
has
all
the
confidence
it
works
or
else
the
mystery
server
will
let
them
know
that
no,
this
won't
work.
H
I
agree
with
you
there
and
the
only
concern
that
you
told
that
was
my
only
concern.
So
if
the
user
is,
it
has
been
shown
that
these
are
the
available
applications
and
he
clicks
on
them
and,
of
course,
if
the
node,
if
the
adapter
isn't
there
to
handle
that,
so
it
would
be
a
bad
user
experience.
But
again
now
I'm
good
quite
confused.
I
So,
actually,
even
I
think
we
can
deal
with
that
also
just
that
we
will
have
to
build
some
more
stuff
in
machine
server.
That
is
mrsa,
always
knows
what
is
present
and
what
is
not
it
if
it
has
the
list
of
the
application
it
can
quickly
pass
through.
Okay,
this
application
needs
all
these
components.
Let
me
look
at
the
registry.
I
Okay,
I
don't
have
this
particular
component
in
my
registry
that
belongs
to
this
application,
rejected,
don't
show
it
to
the
user,
so
we
can't
build
on
top
of
it
if
we
need
to
which
will
avoid
duplication
as
well
as,
and
it's
not
like,
we
won't
be,
we
won't
have
to
build
stuff
with
them
with
seating,
seating
applications
or
anything
into
that
office,
because
right
now,
I
don't
think
there
is
any
any
mechanism
through
which
adapters
can
convey
that
hey
see
these
items
into
your
registry
or
something
so
if
we
have
to
build
stuff,
then
we
can
probably
build
an
immersion
server
to
avoid
duplication
on
a
while,
while
getting
rid
of
the
duplication
that
we
have
at
least
right
now,.
H
I
We
can
actually
just
do
it
when
probably
the
user
loads
of
that
page.
So
yeah
are
we
going
to
process
it
every
time.
So
not
the
most
efficient
thing.
We
can
maybe
cache
it,
but
caching
comes
with
its
own
issues,
so
either
we
can't
do
it
every
time.
The
user
sends
the
page
that
okay
load,
these
things
up
load
my
seated
patterns.
I
B
Yeah,
the
thing
is:
is
measure
is
going
to
have
to
deal
with
both
with
applica
with
content
coming
from
both
places
anyway
or
you.
Even
the
content
was
never
never
came
from
an
adapter
as
we're
describing
it
would
still
have
to
deal
with
the
concern
where
a
user
brings
their
own
app
and
it
turns
out
their
app
is
incompatible
with
meschery's
capabilities,
whether
it's
mesh,
reservers
capabilities
or
the
adapter's
capabilities
like
meshri,
already
has
to
be
concerned
with
that
and
make
sure
that
the
user
has
a
good,
a
decent
experience.
B
We
all
we
were
also
you
know
like
and
again
like,
even
if
the
use
like,
if
you
take
a
different
use
case,
you
say
hey,
I,
I
wrote
a
jwt
transformer
filter
like
the
one
that
we
were
just
talking
about.
It's
that
version
one.
It's
doing
certain
things,
gareth
provided
some
feedback.
Let's
release
a
1.5
great.
Let
me
go
load
up
that
1.5
into
my
mastery
system.
That
already
has
my
1.0
wait.
B
What
should
mastery
do?
Should
it
it
like?
Should
it
overwrite?
Should
it
version?
Should
it
so
mastery
also
already
has
the
challenge
of
content
management,
content.
Versioning,
there's
a
lot
about
content
management
that
concerns
that
come
in
and
so,
whether
it's
an
adapter,
that's
registering
content
or
it's
a
user,
that's
doing
it
subsequently
or
if
meshre
server
as
part
of
its
initialization
routine
is
self-registering
some
things
like
it
doesn't
really
matter
what
vector
you
come
from.
It's
like
you've
got
that
concern
about
what
do
you?
What
do
you
want
to
do
about
duplication?
I
So
but
there
is
still
sort
of
a
difference
that
is
the
components
that
either
machine
registers
or
erupters.
They
are
actually
patent
components,
while
these
are
complete
pattern
files.
What
we
can
do
is,
I
mean,
I'm
not
sure
if
we
should
be
doing
this,
but
what
we
can
do
is
we
can
have
a
component
type
called
book
info,
not
the
best
solution
with
book
info
artistic
or
something-
and
this
is
now
a
confident
type.
So
this
is
a
workload
definition.
I
Basically
when
and
it
registers
it,
we
have,
the
list
can
be
taken
care
by
the
the
same
workflow.
That
is,
he
said
I
need
some
book
info
dot,
sd
or
something,
and
then
it
goes
through
the
same
thing
and
mystery
server
can
quickly
check.
Okay.
This
thing
is
available
to
me
and
this
was
registered
by
this
tv
adapter.
We
can't
do
that,
but
the
thing
is
that
there
are
two
distinct
sort
of
registration.
I
The
only
registration
that
we
have
right
now
is
the
registration
of
components
that
come
from
that
can
come
from
any
direction.
While
this
is
a
pattern,
pattern
or
application,
because
this
only
this
sort
of
country
it
can
contains
only
applications
so
that
that
is
that's
the
reason.
I
was
listening,
that
if
we
can,
let's
centralize,
because
we
have
all
of
the
mechanism
in
mystery
server
to
check
stuff,
so
we
also
have.
We
also
have
now
sort
of
stages.
I
We
can't
even
do
verification
like
we
don't
have
to
deploy
the
entire
thing
just
to
check
if
it.
If
that
thing
is
bad
or
not
just
add,
two
stages,
identify
verify,
get
back
the
result
so
yeah.
B
Nice,
it's
probably
worth
while
for
as
much
as
I
would
like
for
us
to
make
a
decision
and
move
forward
briskly.
It's
probably
worthwhile
to
write
down
a
couple
of
the
considerations,
because
it
because
it's
because
it
gets
a
little
involved,
see
if,
if
you
supplant
the
use
case,
not
with
a
well-known
sample
app
but
with
some
users,
random
application
that
that
might
be
a.
B
Like
you
know
how
it
gets,
thank
you
megan
how
it
gets
verifi
validated
verified
if
we
don't
register,
so
I'm
not
arguing
one
way
or
the
next
in
terms
of
like
hey.
Where
should
these
be
sourced
from
like
when
you
look
at
the
web
assembly
filters,
you
there's
some
amount
of
sustaining
and
maintaining
that
goes
on
on
the
wasm
filters.
B
Repo
should
curate
those
and
make
sure
that
they
work
for
at
least
four
envoy
based
data
planes,
so
therein
they
actually
have
the
same
challenge
already
that
they've
got
to
or
that
as
they
get
registered,
they
should
only
be
applied
to
envoy
based
service
mesh
data
plans.
B
That
anyway,
my
point
being
is
that
as
the
sample
apps
are
sustained
over
time
and
as
they
potentially
version
based
on
enhancements
and
changes
to
their
respective
service
mesh.
It's
like
okay.
Are
we
dynamically
sourcing?
Those
are
we
centralizing
the
the
where
they
get
stored
somewhere
centrally
and
managing
them
there
and
then
or,
and
so,
but
anyway,
I'm
not
arguing
one
way
or
the
next.
All
I'm
saying
is.
B
We
should
go
write
these
down,
write
down
like
a
few
of
these
use
cases,
because
it'll
help
illuminate
the
challenges
that
we
face
and
in
some
respects
I
kind
of
consider
that
what
I
was
saying
before
continues
to
be
true,
the,
irrespective
of
whether
or
not
an
application
is
seeded
from
an
adapter
from
a
user
centrally
in
mastery
server.
It's
like
all
the
same
concerns
continue
to
be
there
about
de-duplicating
and
confirming
that
these
things
should
act
that
we
should
actually
let
a
user
try
to
provision.
One.
B
B
So
ashish,
perhaps
like
by
wednesday
or
before
wednesday,
if
you
can
articulate
it,
could
just
be
very
brief.
Just
bullet
bullets
of
like
what
the
use
cases
are
user
brings
their
own
adapter
brings
one
message:
server
brings
one
centrally
stored.
The
thing
gets
versioned
over
time
that
sample
app.
The
user
tries
to
load
up
multiple
of
the
same
one
you're
like
just
like,
as
you
bullet
those
out
it'll
become
a
little
more.
H
B
And
I
don't
know
that
there
is
a
wrong
answer
to
like
so
to
to
where
they
get
sourced
from
so.
B
Okay,
does
anyone
know?
Are
we
dynamically
sourcing?
The
sample
apps
today
are
those
manually
packaged
in
the
adapters
yeah.
So.
B
To
the
extent
that
they're
manually
packaged
part
of
what
utkarsh
is
saying
is
like
there's
some
mate:
there's
sustaining
energy
that
goes
into
manually,
curating
those
sample
pack,
those
sample
apps
in
each
adapter.
B
B
Cool:
it's
a
good
topic,
you're
up.
G
Yeah,
so
the
current
issue
is
about
having
multiple
import
options
for
patent
filters
and
apps.
So
previously
what
we
were
having.
F
D
G
Ui
is
an
upload
button
and
a
url
uploader
plus
now
we
will
be
having
our
github
upload
button,
plus
another
option
for
upload.
That
is
okay.
Let
me
just.
G
Items
in
a
single
model
so
that
we
can
have
an
user
usability
to
select
what
he
or
she
wants
to
upload
from
the
file.
So
currently,
these
configurations
are
not
working,
because
I
had
some
issues
with
the
material
gui
version
4
and
version
5..
So
I
will
be
configuring
this,
so
the
basic
idea
is
to
so.
The
basic
idea
here
is
that
user
can
have
multiple
options
in
a
single
model
to
rule
whether
he
wants
to
upload
locally
or
from
the
url.
G
The
thing
that
basically
we
want
to
implement
is
that
the
user
should
be
able
to
upload
it
from
the
github
repository
for
that
we
have
all
the
request
body
and
the
urls
over
here.
So
I
need
to
configure
that
accordingly,
that
thing
is
not
yet
configured.
B
B
G
Okay,
so
using
that
after
having
this
back,
option
should
be
there.
G
For
that
we
can
have
tabs
options
that
would
work.
B
B
It's
a
little
like
when
you
have
tabbed
a
tabbed
menu
people,
or
at
least
I
know
for
sure
I
feel
as
if
I
kind
of
need
to
go
exploring
first
and
look
across
the
tabs
to
understand
what
the
what
those
choices
are
about
as
opposed
to
here.
It's
like
hey,
look,
it's
atomic!
You
click
one
of
these
four
you
know
and
and
if
when
I
click
one
of
the
four,
it
just
expands
down
below
and
says:
okay!
Well,
what's
your
github
url
or
what's
your
import
url?
What's
you
know
like
that.
G
Yeah
makes
sense:
yeah
we
can
have
an
expand
down
model
for
that,
not
a
model
but
an
expand
on
option
and
just.
B
That's
cool
yeah
by
the
way
so
on
the
github
import.
This
would
be.
This
is
one
of
those
things
where.
B
Yeah
ooh
cars-
you
might
want
to
give
this
some
thought
this
is
this
is
I
think
this
is
that's,
I
think,
actually,
both
of
those
to
the
github
import
and
the
mesh
map
button,
the
cytoscape
json.
I
think
those
are
both
this
is.
I
Actually,
github
import
is
also
supported
by
none
provider,
while
cytoside
escape
is
also
supported,
but
so
every
for
all
all
of
them
is
supported
by
both
of
the
provider.
The
difference
is
that
in
github,
upload,
private
repositories
and
all
are
supported
only
or
organization
related
uploads
are
supported
only
by
the
mystery
cloud
and
not
by
the
name
provider.
Although,
yes,
both
of
them
can
also
be
queried
by
checking
out
checking
the
capabilities.
I
So
if
you
would
query
the
capabilities
you
would,
you
would
be
able
to
see
that
is
these
are
these
things
basically
is.
Is
these
kind
of
supported
by
a
particular
provider.
B
So
the
the
capabilities-
endpoint:
are
you
familiar
with
that
one.
B
Yeah
yeah
more
or
less
that
yeah,
you
know
we'll,
send
you
a
link
just
to
be
sure,
but
so
back
to
wood.
Crush's
comment
that
utkarsh
on
the
cytoscape
json
or
the
mesh
map
import
like
like.
While
it
may
be
technically
true
that
people
could
import
that
in
the
local
provider,
I
don't
know
that
there's
any
value
to
it,
because
they
wouldn't
wouldn't
persist.
They
wouldn't
be
able
to
okay.
So
there's
that
then,
on
the
other
github-based
one,
it's
like
is
the
github
import
behavior
for
public
repos.
B
B
B
A
We
have
one
more
topic
to
discuss
saying
like
do
you
have
an
update
on
updating
the
filters
ui
yeah
you.
F
J
So
I
was
actually
go
going
through
all
all
the
codes,
and
I
I
am.
I
have
identified
all
the
repeating
codes,
so
I
just
want
to
confirm,
confirm
some
things
and-
and
I
want
to
know
if
I'm
understanding
what
I'm
supposed
to
write
so.
Firstly,
okay
I'll
start
with
patterns.
J
So
patterns
has
this
workflow
when
you
click
a
configuration
button,
it
comes
into
another
space
and
we
have
the
form
for
the
patterns
where
we
can
input
fields.
I'm
so
sorry,
if
my
this
docker
desktop
running
kubernetes,
running
still
running
nws
and
all
on
my
windows,
so
it
might
crash
anytime.
J
So
I
don't
know
okay,
so
this
is
a
workflow
for
our
patterns
and,
as
doesn't
have
this
workflow
yet
so
I
think
I'm
supposed
to
be
implementing
that
to
this
as
well
so
right
now,
creators
brings
up
a
model
like
this
and
there's
just
an
option
for
direct
addict,
and
one
more
thing
I
notice
is
filters
has
only
a
one
delete
option
here.
We
don't
have
an
update
option
here
so.
J
Yeah,
so
applications
has
a
delete
button,
sorry
update
button,
but
filters
on
so
I
will
just
confirm
that
that
was
intentional
yeah.
So
this
is
what
the
patterns
workflow
to
filters
and
applications
and
centralize
the
whole
repeated
components.
J
B
We
wanted
yeah
it
ultimately,
like
long
long
term.
We
hope
I.
I
would
think
that
there's
hope
that
there
will
be
a
more
compelling
user
experience
but
short
term.
That's
very
much
needed
like
right.
Now,
people
don't
if
you
want
to
apply
a
specific
yeah.
I
mean
like
I'll.
Just
stop
there
and
just
say
it's
very
much
needed.
F
B
Okay,
yeah
I'd,
say
man,
it's
about
it's
about
time,
like
you
know,
come
on
just
kidding
just
kidding,
zayn,
that's
great
yeah.
It's
no
small
thing
running
all
those
systems
at
once,
nor
is
it
a
small
thing
to
come
up
to
fully
up
to
speed
on
measuring,
like
you
have.
So
that's
really
good.
It's
exciting
to
see
this,
I'm
I'm
going
to
borrow
the
ball
for
just
a
moment
since
we're
right
at
time.
I'm
going
to
say
that.
B
That
there's
another
there's
another
couple
of
things
that
are
needed
with
respect
to
filters,
applications
and
patterns.
So
one
of
those-
and
these
are
great
kind
of
first
first
items
actually
for
people
or
for
well-seasoned
contributors.
These
are
still
really
good.
Items
is
oh.
B
If
I
had
named
a
filter
when
I
first
uploaded
it,
I
might
want
to
change
the
name,
so
I
just
want
like
rename
the
display
name
I
can
go
in
and
I
can
edit
the
details
of
the
filter.
This
is
great
now,
if
I
edit,
but
but
I
don't
seem
to-
I
probably
can't
what
happened
to
the
save
button.
Oh
probably
missing
a
save
button,
but
also
I
don't
know
that
I
can
change
the
name
of
the
filter
there
apps.
I
got
this
book
info
pattern.
B
Oh
okay,
okay
and
he
may
okay,
so
this
actually
does
let
you
change
the
so
yeah
going
back
to
that.
Zagara
was
saying
this
is
intentional
yeah
right
he's
right
that
what
was
uploaded
here
is
a
binary,
and
so
there's
a
little.
So
so
actually
this
is
more
to
the
point
of
what
you're
bringing
then
like.
A
Yeah,
let
me
just
I
add
something
here
like
the
by
default,
like
every
filter,
we
upload
gets
filter
because,
like
we,
we
we
built
this
using
how
our
patterns
were
built.
So
in
patterns
it
fetches
the
name
from
inside
the
pattern
file,
but
in
case
of
awesome
filters,
there
is
no
option
for
that.
So
we
have
to,
like
probably
add
a
new
or,
like
fetch,
the
name
from
the
the
actual
filter
binary.
A
F
B
Nice
anybody
who's
interested,
just
you
know,
dropped
your
name
in
for
either
one
of
those
two
but
other
than
that.
The
venue
anything
else
that
we
want
to
cover
today.
H
I
had
a
minor
update,
so
so
the
last
thing
I
worked
on
was
dynamically.
Registering
the
components
there
was
a
little.
I
don't
know
whether,
whether
to
call
it
a
blender
or
a
mistake.
There
was
a
mistake
with
the
filters,
so
what
was
happening
is
all
of
the
om.
Schemas
were
same
for
all
of
the
band
components.
There
was
this
problem
with
the
filter
and
it
went
unnoticed
because
you
know
nobody
reads:
what's
the
omrf
schema
and
I
didn't
do
so.
I
have
made
the
change
on
my
local
system.
H
I
will
create
the
pr
soon,
but
yeah
I'll
have
to
change
the
filters,
and
before
doing
that,
I
will
also
have
to
make
a
lot
of
changes
in
the
mesh
kit,
because
first
thing
that
in
the
in
the
in
this
key,
basically
in
the
when
we
are
generating
components
for
the
kubernetes,
the
queue
open
api
json
schema
does
not
support.
That
referencing
thing
question
talk
more
about
it,
so
I
have
to
write
a
function
and
go
to
actually
handle
it.
H
So
first
that
needs
to
be
added
another
thing
that
I
missed
and
somehow
it
it
went,
nobody
saw
it
in
the
pr
review.
Also,
we,
I
have
hardcoded
a
lot
of
things
inside
the
inside
the
mesh
kits
function,
which
needed
to
be
there
passed
from
the
config
as
as
the
cfg.filter.something
filter.
So
two
filters.
I
was
not
even
using
them
and
it
worked
because
all
of
the
adapters
have
the
same
exact
template,
so
it
was
working,
but
it
just
went
overlooked.
H
So
all
of
these
changes
I
have
made
on
my
local
system-
and
I
will
be
making
the
pr
soon
and
in
the
in
that
pr
most
probably
I'll,
also
handle
the
k8
the
thing
for
kids
components
as
well,
so
there
would
be
a
pr
in
mesh
kit
and
meshery
and
all
of
the
adapters
changing
that
particular
filter,
so
that
everything
can
so
so
that
we
can
have
a
different
om
schemas
for
a
particular
component.
So
that's
that's
what
my
update
was.
B
Makes
sense
makes
sense
by
the
way
as
we
go
to
as
we
go
to
close
here
I'll
just
ask
if
roulin,
mr
tang,
if,
if
you
can
hear
us
or
if
or
if
you
might
be
able
to
say,
hi
it's
nice
to
have
you
on
the
call.
B
But
but
I
just
wanted
to
call
out
roulen
in
case
casey's
still
with
us,
so
it's
good
to
have
him
on
the
call.
Okay,
all
right,
nice
great
call
today.
So
one
of
the
action
items
for
next
week
is
outside
of
the
ones
that
are
being
lined
up
here.
I
think
there's
one
more
on
well,
so
the
and
like
following
up
on.
B
Well,
like
it
would
be
good
if
someone
explained
the
phantom
token,
it
might
be
nice
for
everyone
to
understand
what
that
is
also
on
like
when,
as
we're
looking
at
configuring
filters
and
applying
filters.
Well,
this
is
a
great
use
case.
It's
like
when
you
go
to
apply
the
jwt
manipulator
9000,
like
that
specific
filter,
it's
configurable!
B
Do
we
have
so?
Basically,
we
need
to
generate
probably
generate
a
component
right
for
that
can
follow.
Along
with
this,
is
that
right.
A
Oh
one
more
thing,
and
that
we
may
need
to
do
is
improve
the
docs
of
the
dot.
The
jot
manipulator
filter.
A
F
A
A
I
think
we
covered
every
topic
so
let's
meet
up
next
week
and
follow
up
on
these
items
thanks
everyone
for
joining.