►
From YouTube: WebAssembly Filters Meeting (Aug 16th, 2021)
Description
WebAssembly Filters Meeting - Aug 16th, 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,
let's
get
rolling,
though
we've
got
a
few
topics
and
a
few
people
here.
If
you
would
just
again
reminder
drop
drop
your
name
into
the
meeting
minutes,
if
you
would.
A
B
Yes,
yes,
you
do,
can
you
hold
on
a
second
piece
hold
on?
Yes,
I
have.
I
have
kubernetes
on
aws
and
azure
and
gcp.
B
A
A
It's
been
about
capturing
the
filter
as
a
binary
and
then
capturing
a
specific
configuration
that
or
I'm
sorry
configuration
that's
specific
to
the
filter
and
how
it
might
run,
which
is
good,
but
it's
only
part
of
the
story
that
more
of
the
story
has
to
deal
with
what
applications
to
apply
that
filter
to
what
proxies
to
insert
that
filter
into
what,
if
there's
a
chain
of
proxies,
where
does
the
filter
that
you're
working
with
land
in
that
chain?
A
Questions
like
that
things
like
that?
Does
that
is
that
filter
compatible
with
that
proxy
things
like
that,
and
so
there's
a
an
issue
out
there
that
identifies
the
need
for
to
capture
more
of
this.
Now,
as
we
look
at
more
of
as
we
look
at
supporting
envoy
filters
in
context
of
istio,
istio
has
a
number
of
well.
What's
the
right
word,
a
number
of
objects
metadata.
You
know.
A
Additional
constructs
additional
objects
to
track
metadata
about
those
to
track
for
an
envoy
filter,
and
so
the
they're
broken
down
here
and
there's
a
few
of
them,
and
so
in
each
of
those
there's
more
metadata
to
be
captured,
and
so
at
this
moment
I'm
not
sure
and
maybe
so
who's
on
the
call
or
gara
of.
Can
you
guys
tell
me
the
current
pattern,
components
that
we
have?
Are
they
capturing
this
level
of
of
data?
Do
we
have
do
we
have
placeholders
for
this
data.
D
Actually,
right
now
in
mesh
map
at
least
or
what
we
have
right
now
is
filter
is
treated
as
any
other
workload,
for
example,
html
any
virtual
service
or
any
other
code.
So
when
we
say
about
relationships
of
filters
with
either
an
envoy
filter
or
any
other
thing,
such
a
relationship
is
not
exactly
captured.
D
If
you
talk
about
these
fields,
I
guess
utkarsh
will
be
able
to
tell
you
more
about
these
fees
because,
whatever
is
being
returned
from
the
adapter,
all
of
these
fields
are
populated.
So
if
the
adapter
is
basically
returning
a
schema
which
has
these
fields,
then
we
are
capturing
these
fields.
A
Okay,
can
can
you
confirm
that,
for
me,
can
you
look
to
see
if
envoy
config
object
match
is
coming
from
that
workload
descriptor,
you
know
from
from
that
component.
A
Okay,
good
so
coming
coming
back
to
that
later,
so
the
the
next
topic
that
we
had
up
is
so
last
time
we
met.
We
talked
about
well
a
little
bit
of
the
design
in
mesherie's
ui
for
managing
filters
and
how
some
of
what
is
being
some
of
how
we
manage
filters
is
applicable
to
applications
and
patterns
as
well
and
so
gaurav
and
megana.
I
think
the
two
of
you
are
maybe
closest
to
this.
A
A
F
A
A
F
A
Nice,
that's
great
okay
for
context
I'll
try
to
pull
up
mesher's
ui
to
so.
Everyone
has
some
kind
of
context
as
to
what
we're
talking
about.
While
we're
talking
about
that
there's
a
couple
of
other
updates.
A
Like
the
the
thing
I
was
just
mentioning:
common
attributes
of
patterns,
applications
and
filters,
and
so
that
being
the
need
to
just
have
a
display
name
and
a
file
name
shown
so
gaurav.
You
didn't
work
on
that.
Megana
has
this
been
on.
I
don't
know
if
this
has
been
on
your.
A
Good,
so
let
me
give
context
to
this
to
make
sure
that
we're
everyone
understands
and
and
someone's
able
to
take
action
on
it.
So
within
life
cycle
management
of.
A
Of
applications,
filters
and
patterns,
all
of
those
a
user
is
capable
it
can.
A
user
can
upload
those
each
of
them
as
artifacts.
That
measure
then
manages
each
of
them
can
come
from
a
file.
Some
of
them
can
come
from
other
places,
so
they
have
common
needs
about
showing
the
name
of
the
pattern.
In
this
case
the
name
of
the
filter,
the
name
of
the
I'm
sorry,
I
think
I
confused
them,
but
the
name
of
the
pattern
name
of
the
application,
good,
so
th.
This
is
right.
A
Now
it
should
be
a
display
name
like
a
human,
friendly
name.
Someone
would
give
their
pattern
if
they
imported
it
through
a
file,
though,
that
the
location,
the
file
name
and
the
path
to
that
file
is
captured
and
it'd
be
nice
to
expose
that
as
a
common
like
as
another
column
and
do
so
across
the
other
three,
so
megan
does
that
make
sense.
Now
what
I'm
referring
to.
G
I
actually
missed
what
you
said:
what
should
be
the
column?
Could
you
just.
G
File
name,
you
know
you
mentioned
that
I
need
to
add
another
column.
A
G
Okay
and
what
will
be
the
application
name
or
the
patent
name?
Will
that
be
given
while
uploading
or.
A
It
I
don't
know
it
should
be,
I
mean
like.
If,
if
we're
not,
then
we
we
should
be
asking
for
one
optionally.
If
someone
doesn't
supply
it,
then
an
example
here
for
patterns
is
that
when
someone
is
creating
it
using
the
visual
designer
it'll
by
default,
the
file,
the
display
name,
will
be
until
design,
and
so
too
can
the
same
for
an
unnamed
filter
or
an
unnamed
pattern
is
that
we
I'm.
G
A
Okay,
let
me
give
you
an
example,
so
so
see
this
this,
this
google
document
that
I'm
using
at
the
top
it
says,
wasn't
filters
meeting
the
name
of
the
display
name.
So
this
document
is.
A
Awesome
that
it
can
be
whatever
I
want
it
to
be,
and
I
can
include
your
letters
and
characters
and
stuff
like
this,
so
I
could
do
something
awesome,
filter,
meeting
four
seven,
you
know
eight
slash,
16,
slash,
21.
and
normally,
if
you
try
to
throw
slashes
into
a
file
name,
that's
probably
not
going
to
fly
right
on
both
file
systems
so
that
that's
that's,
explain
it
and
then
I
bet
you
know
what
I'm
going
to
do
for
a
file
name.
A
So
there's
this
this
thing
called
location.
It's
really
like
a
it's
like
the
path,
but
the
attribute
the
metadata
about
a
file
or
a
pattern.
You
might,
it
might
also
you
might
see
it
called
location
and
so
that
an
example
there
might
be
like.
A
Now
to
your
point
about
like
why,
if
someone
goes
to
import
like
let's
say
I
go
to
upload
a
pattern,
I
bet
it
just
shows
me,
the
you
know
a
browser
to
a
file
system
browser,
so
I
would
choose,
choose
an
applicable
file
that
that
matches
the
you
know
that
I
can
upload
and
I
might
not
find
one
quickly,
but
when
I
do
well,
I
I
didn't.
I
wasn't
afforded
an
opportunity
to
also
give
it
a
display
name.
A
In
that
case,
we
either
need
to
make
sure
that
the
back
end
handler
the
api
that
is
being
called,
that
it
accounts
for
a
null
display,
name
and
auto,
generates
or
provide.
You
know,
auto
assigns
a
name
okay,
which
a
lot
of
them
do
and
that's
probably
the
best
place
to
handle.
That
is
to
just
put
that
requirement.
On
the
end
point.
H
So
we
already
do
provide
a
display.
I
mean
distance
name
is
already
independent
of
the
file
name
and
the
handler
already
handles
the
fact
that
it
will
actually
grab
the
display
name
from
the
patent
file
itself,
not
the
file
name,
but
rather,
if
you
would
open
pattern,
there
is
always
an
attribute
called
name
a
field
called
name
in
the
back.
It
will
basically
call
that
thing
as
actually
this
is.
H
I
guess
this
was
generated
using
meshmap,
so
it
bypassed
that
thing,
but
if
you
would
actually
upload
it
manually,
then
in
that
case
it
hits
a
different
endpoint
and,
and
basically
it
extracts
the
name
from
the
uploaded
pattern
file.
This
is
actually
a
originated
file.
That's
why
it
talks
differently.
A
Yeah,
some
of
these
discrepancies
aren't
ideal.
It's
nice
that
what
you
said
about
the
handler
the
patterns
handler
specifically,
you
know
intelligently
sourcing,
the
display
name
from
then.
You
know
from
the
name
inside
the
yaml
file,
that's
nice
and
provides
a
better
experience
than
just
untitled
untitled
untitled
untitled
entitled
we
do.
What
do
you
think?
Should
we
take
a
note
to
reconcile
the
two
like?
A
H
Actually,
I
I
think
the
behavior
is
different
right
now,
because
nishmap
chooses
it
to
be
different,
so
meshmap
has
the
feature
that
you
can
name
a
file
and
the
default
name.
There
is
untitled
design
or
something
so
that
is
why
it's
choosing
that
particular
name,
and
that
is
why
it
has
this
particular
particularly
and
but
if
you,
if
you
as
a
user,
are
uploading
from
this
particular
ui
from
any
using
visual
ct,
also
and
you're
providing
a
part
in
that
case,
you
are
not
providing
it
a
name,
so
it
will
automatically
grab
it.
H
So
the
thing
is
the
the
difference
in
behavior
is
because
mesh
map
or
different
clients
using
that
particular
route
or
if
they
won't,
if
they
don't,
provide
a
custom
name,
then
in
that
case,
actually
it
will
crack
the
name
from
the
yaml
itself.
Cool.
A
Yeah,
no,
I
understand
that
and
so
my
having
understood
that
my
question
is:
should
we
should
we
rectify?
Should
we
enforce
that
they
must
be
the
the
same
or
should
we
leave
it,
as
is
where
what
we're
essentially
doing
is
saying
as
a
convenience
operation.
If
the
user
doesn't
separately,
specify
a
display
name
for
that
client,
then,
instead
of
auto
generating
one
called
untitled,
the
you
know,
it'll
do
what
it's
doing
today,
which
is
sourcing,
perhaps
a
more
befitting
name
from
the
name
field.
A
It
seems
like
it
seems
like
in
this
case,
there's
probably
nothing
to
do
like
that.
It's
okay
and
it
is
valid
that
the
display
name
here
like
untitled
design,
doesn't
necessarily
you
know
it
doesn't
mandatorily
have
to
match
the
name
inside
the
pattern,
but
that
the
endpoint
that
you
were
describing
was
really
having
that
behavior
as
a
to
provide
good
ux,
where
there
wasn't
any
in
the
event
that
there
isn't
a
display
name
defined.
A
Okay,
all
right
cool,
all
right,
okay,
good,
so
by
the
way
so
off
the
top
of
your
head.
So
do
you
I
mean
crash?
Do
you
know
so?
Certainly
that's
the
case
for
the
pattern.
Endpoint
do
you
know
if
that's
the
same
for
the
other
two.
A
H
I'm
not
sure
about
the
behavior
of
filters.
I
think
I
think
the
code
is
very
similar,
so
I
assume
that.
F
A
A
A
G
So
I
didn't
do
the
special
special
case
for
the
github
file.
I
raised
a
pr
related
just
uploading,
but
I'm
not
sure
if
I
have
done
it
right,
so
I'm
just
going
through
the
reviews.
So
accordingly,
I
make
the
changes
and
lukash
has
been
helping
me
too
so
I'll
work
on
a
special
case.
Next,
so
that's
pretty
much.
It.
A
A
G
A
A
A
A
And
actually,
this
relates
back
to,
I
think
our
first,
the
first
question
that
I
had
had,
which
is-
I
guess,
we're
going
to
get
back
to
it
now.
So
so
you
were
going
to
check
on
whether
or
not
those
additional
level
of
details
for
envoy
filters
yeah.
We
do
capture
everything.
That's.
D
A
A
So
so
without
avi
here
that
this
probably
goes
to
you.
D
A
A
I
H
D
D
That's
that's
not
the
filter,
that's
just
a
dummy
component.
They
should
be
inside
istio
components
itself.
If
you
have
loaded
measuring
before
running
your
adapter,
maybe
you're
not
able
to
see
the
whole
list
of
workloads
because
you
have
started
adapters
before
starting
measuring
itself.
D
Okay,
yeah,
you
should
be
see,
you
should
be
like
able
to
see
all
of
the
workloads
over
restarting
mystery
and
then
starting
adapters.
After
that,
would
let
you
see
them.
A
A
Yeah,
it's
a
pretty
big
okay,
so
it
sounds
like
sounds
like
these
things
are
being
shown
in
there.
It's
like,
okay,
let's,
let's
start
to
identify
how
to
break
down
that
user
experience,
and
that's
that's
what
this
is
alluding
to
is
like.
Is
you
know
what
like
an
initial
suggestion,
is
something
like
take
each
of
the
sections
that
are
here
if
I
can
call
them
that,
like
the
dot
apply
to
dot
patch
context
and
put
them
as.
A
Right
kind
of,
except
I'm
thinking
more,
super
tiny
like
because
you're
gonna
have
one
you
have
this
many
of
them.
A
D
D
H
Probably
issued
after
is
this
the
adapter
when
tries
to
reach
out
to
measure
server,
it's
not
finding
it
because
then,
basically,
the
service
name
is
different
from
what
it
is
expecting
it
to
be,
and
that
is,
I
think,
the
reason.
A
H
Hunt,
okay,
I
think
I
know
the
issue.
The
issue
is
that
it
is,
it
must
be
trying.
I
know,
because
I
wrote
the
word:
it
must
be
trying
to
reach
to
mystery.
Basically,
the
dna's
name
mystery,
but
the
container
name
is
an
underscore
one.
So
probably
it
tries
to
reach
it
fails.
It
tries
again.
Basically
it
will
try
for
10
minutes
and
it
will
finally
fail
because
it
could
not
reach
to
mystery
and
just
go
on.
H
Oh
basically,
it
couldn't
reach
the
mystery,
but
then
because
the
name
of
actual
the
name
of
the
container
is
actually
machine,
let's
go
one.
I
think
that's
the
issue.
A
Okay
I'll
send
that
over
okay
but
anyway,
the
point
is:
there's
some
refinement
of
that.
What
that
user
experience
needs
to
be
and
and
to
the
extent
that
it's
done
in
those
json
json
schema,
react
components
that
that
your
efforts
in
mesh
map
might
but
then
benefit
measure
ui
in
terms
of
getting
that
laid
out,
refined.
F
Yep,
I
am
so
so
for
the
dogs
part
like
what's
going
on
right
now.
Is
that
I'm
unable
to
like
build
the
docs
locally,
like
it's
giving
me
a
broken
output
and
pews
on
slack,
said
that
samson
was
already
working
on
a
like
pr
to
fix
it.
So
I'm
kind
of
stuck
without
the
talks
being
generated
properly.
I
I
Also
actually,
the
swagger
workflow,
which
was
there
previously,
was
not
working
as
it
should
be
like
now
it
is,
it
will
generate
the
it
will
update
the
swagger.tml
in
both
the
place
like
in
the
docs
and
in
the
hilbert's
directory,
but
the
problem
is
that
whenever
it's
going
to
check
the
pr
it's
failing,
so
I
think
I'm
just
going
to
fix
that
like
it
will
be
done
by
this
friday
or
someday.
F
Know
other
than
the
workflow
like
when
we
are
building
it
with
the
make
swagger
docs
command
when
we
are
building
the
swagger
yaml
from
the
annotations.
The
output
is
kind
of
broken
like
some
of
the
responses
which
are
already
there
like.
They
are
kind
of
missing.
I
Yeah,
actually
I
figured
it
out
like
we
have
to.
I
just
change
the
make
file
and
have
to
write
the
end,
and
I
can
connect
with
you
in
slack
and
tell
you.
F
A
Makes
sense
good,
okay,
you
guys
are
you
guys,
will
be
collaborating
and
that's
great
moving
to
the
next
topic.
So
there
was
discussion
of
a
potential
design
spec
for
oh
how
sample
filters
are
well.
Maybe
I
shouldn't
use
the
word
sample,
but
how
certain
filters
are
pre-bundled
and
shipped
as
part
of
meshri,
and
so
that's
an
open
need
something
that
you
know
so
anyway.
Just
we'll
track
it
for
potentially
for
next
week
to
take
a
look
at
some
thoughts
on
paper.
A
A
Bootcase,
you
were
testing
to
see
if
the
behavior
of
the
filter
endpoints
in
the
application
endpoint
is
the
same
with
respect
to.
H
Yeah
there
is
a
minor
difference,
I'm
not
sure
if
it's
minus,
so
the
entire
behavior
is
same
except
that
in
filters
it
now
expects.
So
what
you
are
uploading,
it
expects
it
to
be
binary
and
not
yaml.
You
can't
upload
yaml,
because
it's
just
a
string
of
bite,
it's
no
longer
being
treated
as
yaml.
H
The
second
thing
is
now
like,
unlike
pattern,
what
pattern
does
is
if
you
hit
the
pattern
upload
endpoint,
it
will
grab
the
grab
the
name
from
the
yamaha
and
basically
call
that
particular
file
with
the
name
that
you
actually
called
it.
Basically,
whatever
name
was
inside
the
yarn,
but
because
now
filter
is
treating
it
as
binary.
It
doesn't
do
that
it.
It
basically
assigns
it
the
name,
a
test,
filter
yeah.
Basically,
these
are
the
only
two
differences.
H
So
the
thing
is
that,
unlike
earlier,
it's
no
longer
expecting
it
to
be
yama,
it
expects
it
to
be
binary,
and
the
second
thing
is
that
the
name
is
not
dynamically
generated.
Rather,
it's
it's
a
same
thing
test
filter
every
time
for
whatever
you
actually
upload
application
pattern
is
actually
same
exactly
same,
so
application
is
just
pattern.
It's
just,
and
there
is
actually
no
other
enforcements
like
that
it
has.
The
type
has
to
be
application.
H
It's
just
you
can't
upload,
basically
any
part
in
there
and
the
name
is
generated
exactly
how
buttons
name
is
generated.
A
Good
deal,
so
you
know
just
to
circle
back
then
confirm
for
you,
the
application
and
pattern
endpoints
those
behave
the
same
as
previously
described
for
the
filter
endpoint
it.
We
might
need
to
open
up
a
back
end
change
to
allow
a
client
to
specify
a
display
name.
A
Nice,
so
is
there?
Is
there
anyone
on
the
call
that
wants
to
to
take
that
on
the
an
update
to
the
filter.
H
Actually,
I
think
I
think
the
support
is
partially
there.
So
if
you,
if
the
ui
will
choose
to
provide
a
name,
then
backend
is
going
to
respect
that
it
it.
It
will
not
assign
test
filter
name
automatically.
It
does
so
only
if
it
sees
that
you
didn't
provide
anything.
So
I
think
what
you
can
do
is
from
the
ui
side.
Only
we
can
handle
it.
H
That
is,
you
can
ask
the
user,
for
the
name
provide
that
to
the
end
point
of
basically
in
the
request
you
can
provide
that
and
the
backend
is
going
to
respect
that
for
patterns
and
applications.
I
think
like
no,
actually
I'm
talking
about
filter
so
patterns.
Behavior
is
same,
but
patents
is
more
dynamic.
Filter
is
not
that
annoying,
but
still
it's
going
to
respect
that.
If
you
provided
a
name.
H
A
Okay,
I
think
you
said
yes,
so
good.
All
right
last
on
the
agenda
is
probably
just
talking
to
kk,
maybe
pranav
on
other
items,
so
I
think
kk.
I
think
we
we've
yet
to
get
your
meat
hooks
synced
into
something
I
think
at
least
at
least
on
this
call.
C
Yeah,
I
would
be
glad
to
learn
something
yeah
if,
if
yeah
I'm
pretty
newbie
in
this
but
yeah,
if
I
get
some
help,
I
would
be
pretty
sure
I
can
do
something
here
and.
A
Yeah
there's
there
well,
I
think
we
might
have
said
this
before
like
there's
a
number
of
things
to
do,
they
haven't
been
designed
yet
so
it's
not
they're,
not
not.
All
of
them.
Are
that
easy
to
pick
up.
We
just
covered
some
of
the
easier
ones
that
didn't
require
a
design
other
than
just
a
quick
example.
A
For
example,
if
you
go
into
life
cycle
well,
so
actually
here's
one
before
we
even
get
into
anything
complex
is
if
you're
using
the
menu,
you
can't
really
tell
what's
a
top
level
menu
and
what's
a
sub
menu,
so
kk
like
super
like
it's
a
very
easy
first
go
is
to
is
to
put
in
bold
these
top
level
menu
items
just
persistently
bold,
whether
they're
selected
or
not.
That
will
help
offset
them
a
little
bit.
C
Yeah
I
I'll
drive
my
level
like
I'll,
try
to
make
sure
like
the
things
which
are
made
menus
to
be
bold
and
the
sub
topics
to
be
in
the
normal
text
so
that
it
pops
out,
if
I'm
not
wrong.
That
is
what
you
meant
right:
yeah
yeah,
that's.
A
A
Yeah
and
here
as
well,
it's
good,
but
it's
it's
so
subtle
that
it's
like
you
know,
so
the
the
bolding
will
help
offset
the
other
thing
that
then
subsequent
to
this
would
be
that
there's
one
two
three
submenu
items
under
here,
for
example,
and
there's
like
eight
or
so
under
here.
Well,
you
see
that
the
highlight
that
you
get
when
you
hover
it's
that's
great,
but
it
would
actually
be
a
lot
more.
A
Obvious
to
to
users,
if
each
of
these
sub-menu
items
each
of
these
three
actually
had
a
lighter
background
to
start
with
like
when
this
expands,
they
all
have
a
lighter
background.
It
would
become
really
obvious,
then
the
difference
between
top
level
and
sub
menu
items
we
still
desire
to
have
an
on
hover
effect.
A
So
you
have
to
you
know,
pick
a
shade.
That's
you
know
not
not
the
same.
J
J
You
know
to
begin
with,
so
that
if
there
is
a
sub
menu
within
the
sub
menu
item,
then
you
can
just
keep
setting
it
at
an
incremental
level,
and
that
would
solve
your
highlighting
issue
as
well.
I
guess
so
just
a
smidge
lighter
would
be
helpful.
J
And
yeah,
so
I
guess,
can
I
just
go
ahead
and
ask
a
question.
So,
in
three
weeks
I've
gone
like
there's
been
a
metric
load
of
changes
and
I'm
seeing
this
and
it's
completely
new.
It's
like
it's
like
I'm
starting
over
again,
so
I
just
wanted
to
sort
of
get
to
know
what
is
functional.
What
has
was
like
the
what
it
works
and
what
the
where
the
main
work
is
lying.
So
just
a
quick
recap
as
to
where
the
pro
what
the
state
of
the
project
is.
A
A
That's
actually
part
and
parcel
to
the
same
conversation.
We
were
just
having
about
trying
to
get
that
same
user
experience
in
the
and
the
mesh
for
ui,
which
is
when
someone
goes
into
the
table
today
and
they
want
to
go
to
apply
to
filter,
deploy
the
filter.
They
would
need
to
answer
those
same
questions.
E
A
Yep,
it
kind
of
know
like
in
the
simplest
way.
I
think
we
could
deploy
the
filter
it,
but
without
so
right
now
are
we
would
crash.
Have
we
ever
tried
capturing
a
bunch
of
other
configuration
for
the
other
attributes.
H
Yeah
we
so
in
terms
of
patterns.
So,
if
you
use
patterns
for
deploying
you
can't
apply
any
filter,
provided
you
are
running
distilled
version
1.9.7
in
later.
H
A
A
One
of
them
is
well
if,
when
measuring
is
shipped
today,
it
doesn't
have
any
filters
pre-loaded
and
available.
It's
like
well
there's
the
wason
filters,
repo
that
has
you
know,
you
know,
has
the
collection
that
you're
familiar
with
those
are
some
of
those
are
just
generally.
You
know
helpful
in
general.
A
That's
a
like
this,
that's
a
that's!
Something
that
needs
to
be
considered
is
like.
J
So
just
a
thought
about
that
would
be
that.
So
in
your
filters
page,
you
do
have
like
a
drop
down
where
you
can
create
sort
of
like
an
overall
packaging,
the
configuration
along
with
the
binary
so
over
there
could.
We
just
give
like
a
drop
down
which
basically
acts
as
a
link
to
one
of
those
urls.
J
So
then
you
know,
if
you
want
to
pack
it,
you
won't
have
to
necessarily
you
know
package
it
along
with
the
application
itself,
but
rather
than
that,
we
just
have
that
url
pop
up
in
the
drop
down,
and
we
can
then
update
it
as
when
it
will
be
some
sort
of
an
api
call.
It
will
make
saying
okay,
what
are
the
available
packages
that
are
being
shipped
and
it'll,
just
pull
it
and
populate
the
drop
down,
just
as
a
suggestion.
J
Right
so
you
have
the
like.
You
said
right:
you
have
a
place
where
you
can
enter
the
url
to
get
the
filter
or
package.
So
why
not
just
take
that
and
then
have
like
a
sort
of
default
drop
down
and
that
way
it'll
just
have
to
make
an
api
call,
and
it
can
we
don't
even
have
to
rely
on
packaging
it.
So
that
could
work
out.
A
Yeah
yeah
and
like
I
think
I
think
that
unto
its
own,
like
it
stands
or
that
that
will
work
okay
and
then
that's
that's
a
good.
That's
a
healthy
step
forward,
it's
good
for
the
user.
They
get
value
out
of
the
box.
You
know
that
and
subsequent
to
that,
you
might,
you
might
expand
it
say:
okay!
Well!
Also,
if
you
have
a
repo,
you
know
an
additional
one
that
you'd
like
for
measuring
to
reach
out
to
and
pull
from.
A
Maybe
to
do
that
and
it'd
be.
You
know
the
same
functionality,
just
a
different
repo
we've
discussed
in
the
past
that
maybe
those
are
private
repos,
and
so,
if
someone
has
offed
and
provided
measuring
access,
also
with
off
with
their
github
id,
then
meshwhere
could
go
retrieve
those
so
megana.
As
we
talk
about
like
the
upload
by
url
and
how
github.com
is,
it
can
be
a
special
case,
that's
kind
of
what
what
is
meant
by
that
is.
A
If
is,
if
someone
has
authenticated
using
their
github
profile,
that's
when
things
can
start
to
change
like
if
they
haven't
mesher,
you
can
still
pull
from
github.com.
It's
just.
It
won't
be
using
any
special
credentials,
it'll
just
be
another
api
or
another
http
endpoint.
G
A
Yeah
so
pranav,
I
think
the
yeah
long
term.
I
think
we
intend
to
have,
or
hopefully
even
more
than
what
you'd
said
what
you'd
said
is
directionally
aligned
like
it's
and
that's
a
good
first
step
it
it's
a
healthy
step
to
getting
to
like
hey.
What
does
this
thing
do
for
me,
I'm
just
seeing
a
bunch
of
empty
tables.
It
doesn't
do
anything.
A
Well,
no,
no
here
here,
here's
a
bunch
of
stuff
out
of
the
box.
Here's
patterns
out
of
the
box,
so
there's
a
patterns
repo
as
well
for
surface
mesh
patterns.
So
to
your
to
the
same
point
is
like
well
what
okay?
What
patterns
do
I
have?
Well,
not
you
know,
depending
if
you
first
signed
in
it's
like
nothing,
great
where's,
the
where's,
the
beef,
where's,
the
meat
and
so
good.
A
Now
what
happens
if
and
I
think
that's
fine
like
we
could-
we
can
take
that
approach
and
and
it
works,
and
then
it's
good
for
certain
deployments.
People
will
say
well
hold
on.
Actually,
this
is
an
air
gap
to
deployment,
or
this
is
highly
or
we
don't
let
people
willy-nilly
reach
out
to
grab
from
repos,
and
so
those
users
might
like
to
sanctify
what
patterns
or
what
filters
are
loaded,
so,
okay,
well,
one
of
the
next
steps
would
be
okay.
Well,
what?
A
If
the
meshri
that
they
download
and
install
has
the
same
collection
of
filters
in
the
same
collection
of
patterns
packaged
you
know,
bundled
inside
of
the
mesh
reservoir
image.
A
Or
maybe
the
mesh
or
maybe
a
specific
adapter,
if
it's,
if
the
pattern
or
the
filters
is
service
mesh
specific,
I
think
in
both
cases
for
patterns
filters
and
actually
for
applications
as
well.
These
three,
the
goal
is
that
they're,
inevitably
they
won't
work
across
all
meshes,
but
the
goal
is,
I
mean
for
filters
anyway.
The
goal
is
that
they'll
work
across
the
same
filter
will
work
across
many
many.
So.
J
So
all
these
entries
are
basically,
I
guess
inside
the
database
right,
that's
running
yeah,
so
then,
okay,
I
guess
one
flag
or
something
like
that-
would
make
sense
in
that
case,
because
if
it's
a
database,
you
won't
need
to
pull
it
from
a
separate.
What
do
you
say,
source
and
then
integrate
it?
So
I
guess
that's
a
little
more
long
term
the
providing
that
air
gap
functionality.
E
E
J
So
that
actually
kind
of
brings
up
another
thing.
So
what
about
like
security
and
messaging?
So
is
that,
like
a
priority
as
now,
I
guess
it's
kind
of
reaching
the
stage
where
we
need
to
start
considering
that
as
well,
not
necessarily
for
just
filters
but
like
in
general,
with
the
way
things
are
accessed
so
yeah.
A
A
Hopefully
I
think
we,
you
know,
we've
got
most
of
these
things.
I
yeah.
We
actually
don't
have
this,
but
but
we
have
most
of
these,
we
have
basic
crud
on
these
things,
which
is
good.
We
need
you
know
we're
gonna
need
more,
but
yeah
so
yeah.
It
was
like
a
couple
of
releases
for
now
to
get
to
authorization.
A
A
That
providers
can
use
to
identify
whether
or
not
a
given
user
should
be
allowed
to
import
a
new
filter
or
connect
to
a
repo,
or
just
this
litany
of
things
so
within
mesh
reservoir
within
every
api
endpoint
each
of
those
each
of
those
operations,
each
of
those
resources
and
the
operations
that
you
can
perform
on
them
need
to
have
a
a
key,
a
unique
identifier
assigned,
so
that
then
you
can
form
a
key
chain
and
so
that
that
forms
like
a
that's
like
a
permission
or
a
basically
it's
it's
a
key
that
says:
if
you
have
this
key,
and
that
says
that's
that's
specifically
assigned
to
import
a
filter
that
key
fits
that
you
know
operation.
A
You
can
do
it
in
general,
if
maybe
you're,
if
maybe
you're,
just
you're
a
user,
you
can
you
have
a
key
for
importing
a
filter,
a
key
for
deploying
it.
You
have
a
key
for
the
these.
You
know
importing
a
pattern
importing
it.
Maybe
you
don't
have
the
key
to
delete
them,
and
so
anyway,
the
key
chain
that
you're
handed,
which
is
a
collection
of
keys
like
the
keychain
that
you're
handed
as
a
user.
A
When
you
sign
in
might
be
for
a
certain
role,
like
the
user
role
versus
the
admin
role
or
and
so
yeah,
it
gets
into
a
really
long,
a
fairly
long
conversation
about
the
design
by
which
some
of
it's
fairly
straightforward,
like
laying
out
what
roles
and
what
keys
there
are.
Basically
any
anywhere.
You
see
an
input
field
or
you
have
a
button
to
click.
Ideally,
every
single
one
of
those
has
a
key,
so
you
can
be
really
granular
about
whether
or
not
an
individual
is
allowed
to
do
something
on
that
entity.
J
So
do
you
think
a
framework
for
handling
security
like
an
external
framework
would
be
better
because
I
guess
those
exist
like
there
are.
I
there's
like
something
like-
I
guess,
the
equivalent
of
keystone,
that
you
know
you
can
separately
deploy,
which
is
going
to
be
sitting
there,
handling
all
your
stuff
and
that
sort
of
provides
a
more
universal
security
management
thing
right.
So
instead.
J
A
Yeah
we
talked
about
that.
We
cut
out
some
specific
ones
and
we're
kind
of
waiting
to
go
research.
Those
go
so
there
is
one
that's
written
in
rust,
that
is
of
interest,
which
so
happens
to
fall
right
toward
your
question
and
kind
of
in
your
lap,
in
terms
of
it
being
in
rust,
I'll
point
it
out
to
you:
it's
a
good
thing
to
research.
A
It's
a
little
bit
of
an
well.
The
project
is
getting
older,
it's
still
young,
but
a
key
point
to
what
we
want
to
try
to
achieve.
Yeah
is
definitely
we
want
to
use
a
different
framework.
A
pre-existing
framework,
if
we
can.
A
One
thing
we
want
to
achieve
is
the
stuff
that
I
just
described
good,
but
also
for
it
to
be
extensible,
which
is
to
say
that
authorization
you
can.
You
can
have
your.
A
We
want
that,
but
also
when
largely
to
date,
anyway,
identity,
authentication,
authorization-
these
are
all
of
the
in
the
domain
of
the
provider,
meaning
that
the
provider
should
the
authorization
should
be
extensible
such
that
the
provider
can
bring
their
own
predefined
roles
in
groups
that
they
have
to
do
so
in
accordance
with
the
framework
that
mesherie
is
using,
which
which
we
don't
have
yet.
But
so
when
we
look
at
that
framework,
when
you
evaluate
it
is
want
to
do
so
in
context
of
allowing
providing
an
out-of-the-box
experience.
That
is,
you
know.
E
A
Know
fairly
well
designed
has
default
roles
and
things
like
that,
but
but
then
it's
extensible
as
well
good,
I'm
glad
you
brought
that
up,
because
that
actually
gives
us
something
to
put
in
front
of
you.
A
Nice:
okay,
thanks
all
for
coming
nice
to
see
everybody
see,
you
see
some
of
you
tomorrow.