►
From YouTube: Meshery Development Meeting (Jan 20th, 2021)
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Let's
begin
welcome
everybody,
this
is
team
sridevka.
We
discuss
everything
related
to
mystery
on
this
particular
call
once
a
week
if
you're
new,
here
you
are
very
welcome
to
give
a
short,
intro
or
tell
us
a
little
bit
about
yourself
before
we
begin
pradav.
I
believe
you
are
due
on
this
call.
Do
you
want
to
go
first.
B
Yeah
hi,
so
I'm
currently
you
know
starting
my
bachelor's
in
india
and
I'm
a
third
year
student
and
I
just
joined
today,
and
I'm
quite
excited
to
be
a
part
of
this
community
and
start
contributing
I'd,
be
more
active
in
tomorrow's
call.
I
guess
the
newcomers
call,
so
this
is
just
me
sitting
here
and
observing
how
things
go
well.
Thank
you.
So
much
for
letting
me
talk.
A
A
A
All
right
bye,
we
wait
for
him.
B
Know
I'm
I'm
really
sorry
to
interrupt
you.
B
Yeah,
I
just
wanted
to
know
what
sort
of
conversations
do
you
generally
have
I,
I
know
you
said
everything
measuring,
but
maybe
a
little
more
in
detail.
Is
it
like
projects
that
are
ongoing
certain?
What
do
you
say
prs
that
you
have
to
review
and
accept
is
that
is
this
is
where
do
they
occur
here?.
A
Everything
that
you
just
said
actually
it
could
be
concerned
with
peers.
It
could
be
concerned
with
related
projects
or
directly
related
to
the
mystery
project.
Anything
that
we're
currently
working
on
and
that
which
needs
some
review
or
which
could
benefit
from
some
feedback
can
be
discussed
in
this
particular
call.
Does
that
answer
your
question
perfect?
A
All
right,
let's
move
on
did
lavender
finally
make
it
wander
the
call.
A
Guess
not
I'll
try
contacting
everyone's
partner
figuring
out.
What
exactly
is
going
wrong
is
another
one
called.
C
Oh
yeah
sure,
hey,
hey
guys,
hey.
C
I
have
a
headache
not
that
that
really
matters,
but
I'm
slightly
off
this
morning,
so
I
just
I
just
woke
up
you're
lucky
that
I
you
see
me
with
my
hair
brushed.
D
C
Okay,
fair
enough
good,
it
is
so
there's
a
lot
of
things
to
talk
about
today.
I'm
hoping
that
we
talk
about
a
bunch
of
stuff,
I'm
hoping
that
each
of
you
get
a
chance
to
say
something.
Actually
so
pranav
had
a
great
question.
I'm
gonna
try
to
type
out
the
answer,
because
a
lot
of
you
have
heard
me
give
that
answer
in
the
past,
which
means
we
should
probably
write
it
down
somewhere,
maybe
like
on
the
newcomers
page
on
the
layer,
5
ng.
C
If
anybody's
here
is
working
on
that
onirud
and
navendu,
and
a
little
bit
of
jay
batra,
I
was
hoping
some
sladen,
but
I
don't
see
him
on
is:
are
working
on
a
measuring
ctl
system
context
and
they're
working
on
bringing
context
to
measuring
ctl
and
anirude
isn't
on?
Will
somebody
poke
him?
Maybe
five
of
you
poke
him?
C
He
should
he
should
be
on
so
then
I
guess
that
means
that
I'm
I'm
up
on
this
topic,
which
is
we're
getting
ready
for
the
for
our
next
release.
It's
a
minor
release,
but
it's
got
it's
jam-packed
full
of
a
bunch
of
architectural
changes,
new
components
like
a
lot
of
new
components,
maybe
more
than
you
can
count
on
one
hand.
So
by
any
other
measure,
it
ain't
a
minor
release.
C
It's
a
major
release,
all
right,
it
qualifies
as
a
major
release,
we're
just
not
revving
the
1.0,
yet
we're
just
not
revving
the
major
release,
because
well,
because
informally,
1.0
signals,
production,
readiness
and
we're
not
we're
not
quite
there.
C
There
are
people
who
use
the
performance
management
aspects
of
measuring
for
production
things,
that's
fine!
That
area
works
well.
The
rest
of
it
is
on
its
way
and
it's
about
halfway
on
its
way,
maybe
if
it's
a
little
bit
hard
to
measure,
but
but
it's
about
right
anyway.
We
want
to
let
the
world
know
that
we're
about
halfway
there
and
there's
some
interesting
things
that
they
should
be
using
and
trying
out
and
giving
us
feedback
on
helping
inform
the
road
map
on
and
there's
any
any
number
of
you
on.
C
Now
is
a
chance
to
tell
everybody
about
it,
and
so
we
want
to
if
we
can
be
a
bit
organized
about
announcing
that
about
getting
the
well.
You
know
getting
the
word
out
about
writing
up
what
the
word
is,
and
so
it's
a
little
bit
of
a
task
onto
its
own.
Just
kind
of
organizing
excuse
me
organizing
a
release.
What
does
it
mean
to
make
a
release?
C
C
Some
of
that
includes
you
know,
code
quality,
feature
functionality,
writing
up
of
things,
maybe
taking
some
videos
of
some
things,
taking
some
screenshots
or
some
things.
It
means
documentation,
it
means
blog
posts
being
social
with
the
updates.
It
means.
C
Yeah
code
cover
it,
it
just
means
it
means
a
lot
of
things.
It
means
then
sort
of
releasing
the
bits
at
this
point.
Hopefully
releasing
the
bits
is
one
of
the
easier
aspects
of
making
a
release.
C
So
much
so
so
much
effort
is
given
to
continuous
integration
that
we've
revived
a
immediate,
a
bi-weekly
meeting
on
focus
just
on
continuous
integration,
there's
a
lot
of
components
that
comprise
meshri
and
they
each
take
a
bit
of
ci.
C
So
it's
a
bit
of
a
long-winded
way
of
saying
of
asking.
Is
anyone
interested
in
taking
on
the
release
manager
hat
for
for
the
release,
so
I
sort
of
briefly
described
some
of
the
activities
within
there.
It's
it's
a
bit
of
program
management.
C
If
you
will
a
bit
of
sometimes
hard
to
distinguish
between
project
management
and
program
management,
but
it's
a
bit
of
herding
cats
corralling,
I
don't
know
corralling
animals,
it's
a
bit
of
a
going
around
making
sure
that
we've
got
docks
for
each
of
these
features
that
we've
got
a
release.
Announcement
created
that
we've
got.
C
So
we
want
to
make
a
couple
of
early
releases
in
a
beta
format
so,
and
we
have
a
particular
naming
scheme
for
the
way
that
we
version
the
software
created
here.
It
follows
semantic
versioning,
and
so,
in
accordance
with
that,
we'll
probably
have
a
you
know:
measuring
b050
beta
one
and
depending
upon
really
like,
depending
upon
how
many
bugs
we
find
we'll,
probably
determine
how
many
betas
we
have.
C
E
C
C
Is
everybody
everybody
can
hear
the
rooster
right
now.
C
C
Wow
dhruv,
I
think
you
gave
me
such
a
jolt.
I
accidentally
quit
the
other
app
that
I
was
in
in
their
browser.
Right.
C
Apologies.
Thank
you
for
that
drift.
Anybody
of
what
I've
described
this
far.
Anybody
is
anyone
raising
their
hand
on
taking
on
that
role,
the
release
manager,
role.
C
And
if
sometimes
we
do,
I
ask
questions
like
this,
and
people
are
intrigued
and
interested,
and
it
is
entirely
too,
and
they
have
some
questions
that
they'd
like
to
ask
in
private.
So
so
I
get
that
just
come,
ask
them
and
see
if
you
want
to
try
out
the
activities
I'll,
try
to
lay
them
out
a
little
bit
I'll,
write
them
down
and
but
to
finish
off.
What
I
was
describing
is
that
we'll
probably
have
a
release
candidate,
so
maybe
like
an
rc1
and
then
make
a
release.
C
Hopefully
we
would
just
do
sort
of
minimally.
These
two
things
get
a
beta
out
there
and
get
an
rc1
out
there
now
a
whole
separate
conversation.
This
is
something
for
us
to
talk
about
in
the
measuring
ci
call
which,
by
the
way,
I'm
going
to
fully
answer
pranav's
question,
not
verbally
non-verbally,
but
the
ci
call
is
a
great
place
to
finish
up
these
discussions
about
how
many
relationships
we
do,
what
the
stable,
how
the
stable
and
edge
channels
work
around
how
they
are
vehicles
for
these.
C
Cool
what
else
will
we
anybody
have
any
questions
on
that
anybody
tentatively
raising
their
hand
intrigued
by
a
signal,
if
so,.
C
F
Yeah,
so
the
issue
I
was
working
on
was
to
make
the
system
reset
command
contact
server,
so
the
earlier
behavior
was
that
the
reset
command
automatically
picks
the
latest
docker
compose
yaml
from
github,
but
actually
it
should
check
the
current
context
and
the
channel
and
the
version
attributes
and
pick
up
the
docker
compose
file
from
the
from
that
particular
release.
F
F
Yeah
so
I'll
just
show
the
code,
so
what
we
have
done
currently
is.
We
first
pick
up
the
current
context
and
we
will
then
pick
up
the
channel
of
that
particular
context
and
the
version
of
the
particular
contest.
F
And
if
the
channel
is
x,
then
we
will
just
pick
up
the
latest
docker
composer,
but
if
the
channel
is
stable
and
it
has
a
version,
we
pick
the
the
docker
compose
file
from
that
particular
version
and
github,
and
I
will
just
show
my.
F
F
C
F
C
Yeah
and
so
then
you're
already
presenting
on
the
call,
which
is
pretty
awesome
because
you
and
I
we
just
talked
about
that
command
like
a
couple
days
ago,
so
it
just
so
it's
pretty
fantastic,
because
you're
going
to
have
your
name
in
the
v050
release
like
you're
going
to
have
contributed
to
it.
People
will
people
will
use
that
very
feature,
especially
anyone
who's
contributing
here.
That
features
were
super
super
convenient
to
being
able
to
well.
You
know
reset
to
factory
defaults,
so
to
speak,
like
I've
used
it.
C
I
don't
know
how
many
times
and
it's
been
a
pain
in
my
arse.
If
I
can
use
that
word
in
the
past,
because
it
will,
it
was
violating,
it
was
crossing
release
streams,
it
was
violating,
I
would
be
on
stable
channel.
I
would
go
to
reset
and
it
would
reset
to
the
edge
releases.
C
You
know
a
version
of
the
docker
compose,
so
so,
for
my
part,
I
thank
you.
This
is
gonna,
be
good
but
anyway,
so
so
I.
F
F
C
I'll
ask
who's
sana
hussain
who's
on
the
call
actually
she's
she's
made
of
well
in
the
best
of
ways.
She's
made
a
a
mess
of
things
inside
of
meshree
ctl,
so
she's
got
her
name
on
a
fair
bit
of
lines
of
code
in
there,
and
so
she
might
have
some
ideas.
C
One
of
the
ones
that
I
would
try
out
is
what
happens
if
you
try
to
like
you
know,
if
you
try
to
you,
subscribe
to
the
stable
channel
great,
that's
the
channel
that
you're
on,
but
you
know
what
happens
if
you
misspell
it
oops
and
then
or
what?
If,
if
you're
on
there
and
you're
subscribed
to
the
stable
channel
and
you're
trying
to
pin
to
the
version
v04
20
000,
420
201,
you
know
some
version
that
doesn't
exist.
C
Does
that
get
held,
and
so
you
know
just
fat-free
fat
fingering
things
like
that
is
always
you
also.
This
gets
a
little
opinionated
and
hopefully
we
approach
our
user
experience
in
a
consistent
way.
But
you
want
to
consider
what
happens
if
someone
types
in.
C
Well,
if,
in
this
case,
like
I
don't
know
you
you,
it
might
be
one
of
those
areas
where
we
would
just
first
pop
the
question
in
someone's
face
say:
yes,
no,
like
they
do
measure
ctl
system
reset.
Now
this
is
it's
a
factory.
It's
a
you
know,
it's
a
reset
to
factory
defaults
thing,
that's
exactly
what
this
behavior
here
is.
C
I
bet
you
more
often
than
not.
When
you
go
to
do
that
on
in
a
piece
of
software,
it
usually
says,
are
you
sure,
like
I'm,
about
to
wipe
out
your
configuration
like
just
to
confirm,
you're
sure
you
want
to
do
that
right,
so
we
might
do
well
to
present
people
with
just
a
quick.
Yes,
no,
like
this
will
reset
youtube
factory
defaults.
Are
you
sure
your
stuff
will
be
lost?
C
C
Both
our
clients,
the
ui
and
the
ctl,
are
both
clients,
so
so
who's
saying
that
I
guess
there's
a
call
out
to
you
to
try
to
make
try
to
make
nuvenue's
life
hard
see
if
see
what,
if
you
can
break
anything
in
there.
D
So
do
we
have
the
list
of
supported
releases
somewhere
that
we
can
have
it
as
a
list?
If
the
input
is
matching
to
that
list,
we
would
pick
it
up
from
there.
Otherwise
we
would
pick
it
up
from
a
stable
or
some
default
path
should
be
defined
and
also
it
would
be
good
to.
D
C
Nice
I'm
taking
meeting
minutes
since
no
one
else
is
writing
this
down.
So
that
was
to
confirm
ask
for
confirmation,
yeah.
F
D
Like
a
default,
docker
config
file.
D
F
Anything
like
currently
like
it
will
keep
it,
as
as
it
is.
F
I'll,
just
I'll
I'll,
just
yeah
yeah
should
we
go
for
like
a
default
docker
compose
file
yeah.
If
that
specific
version
doesn't
exist
or
like.
C
I
didn't
hear
that
I
wasn't
paying
attention
now.
I
rude
looks
like
he
isn't
either.
The
answer
is
that
it's
actually
over
time
the
docker
compose
file
that
we
have
so
I'll.
Explain
it
real
quick,
but
before
I
do,
let's
put
a
link
to
where
this
is
actually
written
down,
which
is
that
over
time
the
docker
compose
file,
it
begins
that
there's
a
template
config.yaml
that
gets
baked
into
mastery
ctl
itself
over
time,
messy
ctl
will
reference
less
and
less
of.
What's
in
the
compose
file,
and
more
and
more,
it
will
reference.
C
What's
in
config.yaml,
what
the
heck's
the
difference
between
the
compose
file
and
config.yml.
You
ask
great
question:
we'll
put
a
link
to
it's
in
the
this.
That's
such
a
great
question.
It's
inside
of
this
doc
on
your
road.
If
you
have
a
link
to
this
doc,
feel
free
to
point
people
to
it.
There's
a
number
of
faqs
in
there
that
that
speak
to
that
yeah.
There
is
a
def
default.
C
C
Yeah
hit
up
in
slack,
if
what
I
said
it
doesn't
make
a
lot
of
sense
and
if
what's
written
in
the
cli
and
documentation
doesn't
make
sense
either.
A
All
right
I'll
contact
under
the
gun
until
I
do
rule
for
abhishek,
do
you
want
to
take
the
next.
E
Is
it
me
or
no
I
didn't
get
it?
You
drew,
I
think,
oh
yeah,
so
currently
I'm
having
a
dependency
issue,
so
I
won't
be
able
to
show
what
I've
been
working
on
so
yeah,
probably
that
if
you
want
to
talk
about
that,
we
could
talk
about
it
after
everyone
else
is
completed.
C
G
G
All
right
is
my
screen
visible.
A
G
Great
all
right,
so
it
was
my
recent
pr
that
I
chipped
in
and
meshedy,
which
is,
which
is
to
introduce
an
extension
point
on
the
golang
backend
on
the
golang
machine
server.
That
missouri
runs,
which
can
load
up
plugins
during
runtime,
so
that
we
can
develop
different
plugins
to
implement
independent
functionalities.
G
G
So
first
I'll
show
how
it
works.
Actually,
so
I
had
designed
a
sample
plugin
that
that
prints
out
a
message
called
initialized
whenever
it's
been
loaded
up
and
I
have
my
machine
server
running
as
a
local
build
and
I'm
running
the
mystery
remote
provider
as
well
as
a
local
build.
So
when
you
boot
up
mystery.
G
Yeah,
so
it
looks
something
like
this
and,
if
so
to
in
order
to
initialize
the
plugin
in
here.
G
G
G
So
I
had
created
an
extension
point
which,
which
has
two
functions:
one
is
to
load
up
the
package
and
by
default
any
plugin,
I
can
show
you
the
sample
plugin
that
I
created
it's
in
here,
so
any
plugin
would
have
a
run
function
which
would
implement
an
input,
input,
parameter
of
extension
of
type,
in
extension,
input
and
output
of
type
extension
output,
and
these
are
the
definition
of
these
two
are
defined
in
measuring
itself.
G
Yeah,
so
right
now
the
extension
input
feeds
in
the
database
handler
that
measure
uses
it.
It
gives
basically,
it
shares
the
database
handler
to
the
plugin
so
that
the
plugin
can
do
operations
with
mesh3s
database
and
the
extension
output
here
gives
a
router
which
is
http
based
router,
and
this
can
be
used
to
extend
functionalities
of
the
measuring
api
itself.
G
So
so
this
is
basically
the
extension
or
definition,
and
so
this
particular
function
loads
up
the
whole
extension
package
in
here
and
then
the
extension
plug-in
in
here
and
the
handler
would
serve
the
plugin
as
and
if,
if
the
router
rewards
with
a
with
a
http
handler
to
extend
that
extend
its
capability,
this
handler
would
enable
that
and
and
yeah
that's
that's
pretty
much
how
this
has
been
implemented.
G
All
right,
yes,
so
lee
do
you
wanna,
want
me
to
discuss
about
graphql
extensions
as
well
like
towards
graphql
the
work
we
have
done.
G
Yeah
sure,
so
what
we
are
trying
to
do
is
to
run
a
graphql
server
alongside
meshrey's
api
server,
and
in
order
to
do
that,
we've
implemented
a
graphql
plugin.
Basically,
the
plugin
would
be
loaded
real
time
in
the
machine
server
and
it
would
serve
different
endpoints
in
here.
So
so
right
now,
if
you
look
at
this
code,
basically,
what
we're
doing
here
is
that
we,
we
wrote
some
sample
resolvers.
G
It
is
always
nothing
but
functions
that
that
perform
actions
for
every
api
call
or
our
query
in
case
of
crappy
and
so
for
every
query.
We
have
all
the
resolvers
in
place.
Now
to
now.
The
graphql
server
itself
is
implemented
over
h
over
a
tcp
socket,
so
it
returns
a
http
based
handler,
or
rather
it
implements
the
http
interface
of
the
the
native
http
package.
G
So
what
we're
doing
is
that
we're
getting
the
graphql
handler,
which
is
this
particular
portion,
and
we
are
passing-
we
are
giving
the
asset
again,
giving
that
as
an
output
of
output
to
the,
where
did
it
go
output
to
the
measuring
server
as
a
http
handler?
G
This
basically
graphql
handler
implements
this
interface,
so
we
are
just
passing
passing
the
graphql
handler
as
this
particular
value
and
then
to
extend
whatever
the
end
points
that
graphql
is
going
to
serve.
We
are
just
dealing
with
that
in
this
particular
function,
which
is
the
extension
handler
which
records
the
extended
endpoints.
According
to
the
loaded
plugins
inside.
G
All
right
yeah,
I
guess
that's,
that's
it
from
my
side.
C
Thank
you,
abhishek
good,
so,
by
the
way,
if
people
didn't
quite
get
what
abhishek
was
showing,
which
I
suspect,
since
people
didn't
ask
questions,
that's
totally
fine,
it
makes
sense
that
you
might
not
get
it.
Can
anyone
do
me
a
favor?
Does
anyone
who
think
that
they
understood
what
he
was
referring
to?
Could
you
try
to
explain
it
is
to
reinforce
your
own
understanding
and
then
help
everyone
else
understand.
At
the
same
time,.
C
So
so
the
topic
was
extensibility
and
it's
kind
of
a
big
old
deal
for
for
measuring.
C
Well-
or
I
should
say
it
like
this-
that
it's
a
big
old
deal
in
terms
of
like
how
much
effort
it
takes
to
try
to
make
the
project
extensible
it's
a
fantasy
of
maybe
mine,
and
maybe
some
others
here-
that
mescheri
becomes
a
well-used
project
and
and
part
of
what
will
help
make
this
project,
or
many
others
well
used,
is
to
give
people
some
freedom,
some
room
to
be
to
differentiate
and
by
people
I
mean
either
individuals
or
other
organizations
to
come
in
and
say,
hey.
You
know
this
is
a
useful
tool.
C
I'd
like
to
extend
it
I'd
like
to.
I
I,
like
the
logo
in
the
upper
right
of
our
left
hand
corner
it's
the
prettiest
logo,
I've
ever
seen,
but
I'd
rather
put
mine
there,
I'd
like
to
augment
and
change
the
way
that
this
thing
works
and
people
can
they
can
go
in
and
remove
the
mescheri
logo.
I
don't
know
why
they
ever
would
like
that
thing
is
just
but
and
put
in
put
in
something
else,
and
that's
that's
fine
like
that.
That
would
be
in
many
respects.
C
That
would
be
great,
like
that's
not
the
thing
that
we
would
encourage
the
most,
but
would
be
great
because
they're
finding
value
in
the
project
and
they're
using
it,
and
they
want
to
share
with
others
and
maybe
they're
a
consultancy,
and
they
want
to
I'm
just
giving
a
random
example
keeping
people
making
a
change
anyway.
Point
is
at
some
point,
maybe
we
should
say
hey
a
lot
of
people
kind
of
want
to
do
that.
C
Maybe
we
should
make
it
such
that
when
the
next
release
comes
out
the
new
logo,
the
messaging
logo,
doesn't
overwrite.
Their
change,
like
you
know,
be
kind
of
nice
that
way
they
don't
have
to
so
maybe
we
should
be
conscientious
about
where
people
can
extend
the
tool.
Maybe
that
and
it's
a
fine
balance
like
up
front.
It's
like
hey.
No
one
really
cares
about
extensibility.
If
the
tool
doesn't
do
interesting
things.
If
it's
not
like
you
know,
novel
and
really
helping
people
get
something
done,
then,
who
cares
about
extensibility
nobody's
using
the
tool?
C
So
so
hey
once
the
tool
is
really
interesting.
Let's
make
sure
that
we're
encouraging,
like
encouraging
other
organizations
and
individuals
to
invest
their
time,
to
augment
the
the
tool
to
do
things
they
want
to
do
people
right
now,
there's
a
number
of
areas
of
extensibility.
We
support
three
different
load
generators
by
the
way,
but
there
are
many
more
than
three
load
generators
in
the
world.
They
work
differently.
Someone
might
want
to
insert
a
different
load
generator
than
what
is
supported
directly
in
the
project.
Well,
good
we've.
C
There
have
been
considerations
made
for
that,
there's,
an
interface
behind
which
people
can
extend
and
insert
a
different
load
generator.
As
a
matter
of
fact,
it's
we
have
talked
about
this
a
little
bit
in
the
past,
rather
than
me,
blubbering
on
for
a
little
bit
more
I'll,
say
I'll,
try
to
wrap
up
my
description
and
leave
you
with
a
visual
diagram
of
the
fact
that
there
are
that
meshery
and
its
architecture
has
become
a
bit
littered
with
extension
points.
C
Extension
points
are
here
in
yellow
whoa
and
so
obviously
was
just
more
or
less
showing
that
that
there
are
in
the
in
this
upcoming
release
we
will
have
made.
There
have
been
extension
points
that
the
project
has
supported
in
the
past.
The
most
notable
and
sort
of
intuitive
extension
point
that
has
been
supported
to
date
is
adapters.
C
There's
a
lot
of
adapters
out
there
and
that's
a
point
of
extensibility.
Someone
can
write
another
adapter.
As
a
matter
of
fact,
I'm
looking
at
some
individuals
on
this
call,
who've
written
new
adapters
and
people
on
the
call
who've
totally
rewrote
the
entire
framework
of
the
way
that
adapters
work.
C
Good,
that's
nice,
because
you
know
hey,
guess
what
there's
going
to
be
another
service
mesh
announced
soon.
So
it's
coming
and
I'll
tell
you
come
hit
me
up.
If
you
want
to
know
what
it
is,
what
am
I
trying
to
do
extension
point
so
he's
just
showing
an
extension
point,
which
is
that
you
know
with
the
measuring
supports
graph,
will
will
support.
Graphql
has
an
interface
through
which
you
can
query
the
database.
C
C
Go
ahead,
go
ahead
and
when
you
do,
when
you
upgrade
to
v06,
your
queries
won't
be
overwritten
because
you'll
put
them
inside
of
the
safe
zone
inside
of
the
extension
point,
and
so
in
some
another
way
of
articulating.
This
is
that
oh,
my
gosh
another
way
of
articulating
this
is
that
you
well
we've
got
like
a
bunch
of
more
contracts
happening
a
lot
of
times.
You'll.
Think
of
an
api,
as
being
you
know,
is
something
of
a
contract
with
users.
This
is
how
you
can
programmatically
interface
with
mesh
rear
with
the
softwares
through
an
api.
C
Sometimes
that's
you
know,
that's
a
rest
based
thing.
Sometimes
it's
a
sql
based
thing.
Sometimes
it's
a
anything
can
be
an
api.
It's
the!
What
what
makes
an
apis
is
anyway,
that's
a
discussion
for
another
time,
but
one
of
those
characteristics
is
an
implicit
contract.
So
in
some
respects
we
just
littered
our
project
with
a
bunch
of
contracts.
Oh
no.
That
means
we're
gonna
have
to
move
more
slowly
and
adhere
to
all
of
these
contracts
like
well.
C
I
don't
think
so,
because
I
say
we
spent
a
long
darn
time
on
a
number
of
these
extension
points,
they're
pretty
neat.
They
work
pretty
well
and
they're
dynamic.
These
are
different
than
you're
going
to
find
in
any
other
project.
In
okay,
not
not
every
single
project.
Kubernetes
is
maybe
the
only
one
that
does
some
similar
things
like
this.
C
A
I'll
interrupt
here,
another
is
on
the
call
he
could
probably
pick
up
his
topic.
C
C
And
so
back
to
what
I
was
saying
is
that
yep,
that's
pretty
neat,
so
so
that's
what
honor
abhishek
was
talking
about
with
resolve
resolvers
and
the
ability
to
dynamically
inject
go
code
on
the
fly.
C
Cool
drew's
quickly
or
actually
said
truth
boot
karish.
I
think
you're
up
any
other
questions
for
abhishek
before
kash
goes.
C
H
Okay,
so
we've
been
actually
trying
to
create
another
construct
in
mystery
by
construct,
I
mean
basically
now
you
can't
give
a
message:
you'd
be
able
to
give
me
a
a
particular
file,
a
yamaha
file,
which
we
called
pardon
this.
This
is
nothing
but
kind
of
you're,
mentioning
that
these
are
the
services
that
you
want
to
deploy.
This
looks
very
much
similar
to
docker
compose,
so
you
are
mentioning
in
here
that
the
basically,
what
I'm
call
what
this
should
be
called
is
sdsm,
it's
random.
H
You
can't
call
it
anything,
and
the
services
that
you
want
to
be
deployed
are
seo,
and
that
is
a
certain
type
is
service
mesh
and
you
want.
You
also
want
some
add-ons
to
that
particular
service
mesh.
That
is,
the
type
of
service
which
add-on
variant
is.
You
also
mentioned
depends
on
so
this
is.
H
This
is
the
like
the
configuration
of
your
entire
deployment
that
you
want
to,
and
you
want
to
have
what
what
machine
would
do
behind
the
scenes
is
that
it
will
take
this
thing
and
it
will
convert
this
pattern
into
multiple
oem
components.
Ryan's
sound
was
here
on
the
call
he
explained
really
well.
What
oem
is
so?
H
I
won't
go
in
depth
what
oem
is
all
about,
but
what,
if
what
machisa
would
internally
do,
is
it
will
convert
this
pattern
into
om
competence,
oem
application
component
and
om
application
configuration,
and
it
will
also
apply
certain
other
kind
of
policies
onto
it.
For
example,
here
I'm
mentioning
that
hey
this
grafana
grafana
service,
this
depends
on
istio.
H
Would
implement
some
best
practices
onto
it?
That
is,
it
will
say.
Okay,
you
are
saying
that
in
rafana
depends
on
histo,
but
I
know
that
it
should
also.
I
mean
it
or
not
that
you,
you
would
be
deploying
rafana
before
prometheus,
but
a
measure
would
enforce
that.
So
it
will
add
another
dependency
to
your
to
your
deployment,
because
machine
is
smart,
I
mean
it
will
figure
it
out
so
so
it
will
make
sure
that
rafana
it
gets
deployed
only
after
prometheus
gets
support.
H
So
right
now
this
I
mean
because
it's
still
in
testing
seats,
so
there
is
no
ui
for
it.
The
end
result
would
be
that
you
would
be
able
to
access
this
thing
from
mischief
ctl
as
well
as
from
measure
e1.
Let
me
show
you,
I
would
obviously
have
a
ui.
I
mean
all
the
ui
extensibility
to
support
those
things
and
measuring
measure
ctrl,
basically
support
to
basically
run
a
command
like
mystery,
mystery
ctl,
deploy
hyphen
this
particular
file
right
now.
What
I
have
on
my
system
is
this
endpoint.
H
This
endpoint
is
actually
what
is
responsible
for
which
will
be
like
handling
this
particular
file
behind
the
scenes.
So.
H
If
I
send
a
request,
I'm
sending
this
particular
file
to
message
server
and
here
are
the
logs,
so
the
I
mean
I
have
enabled
logs
not
that
we've
been
getting
all
this
stuff,
so
you
can
see
that
it
figured
out
okay.
If
there
does
not
have
any
dependency,
so
it
will
deploy
spokes
then
it
will.
This
is
nothing
but
topological
sort.
So
then
it
will.
You
can
see
here
grafana
the
dependency
for
grafana
number.
Two,
I
mean
the
number
of
dependency
dependencies,
although
you
mentioned
that
dependency
for
graphana
is
one
but
mystery.
H
Mystery
basically
figured
out
and
said:
okay,
although
you
mentioned
only
one
dependency,
I'm
going
to
inject
another
dependency,
and-
and
that
is
why
that
is
the
reason
that
grafana
would
be
deferred.
The
last
I
think
this
is
this-
is
the
previous
state
of
my
mesh
of
my
tv
release
twister,
but
now
you
will
see.
I
hope
everything
is
up.
So
everything
is
up
in
your
system.
Namespace
grafana
is
theo.
I
mean
the
same
resistance,
all
of
the
other
components.
H
It
internally
also
measure
server,
but
internally
also
figure
out
that
what
our
machine
adapter
should
be
used,
because
we
have
a
lot
of
products,
it
will
figure
out,
based
on
the
variance
that
that
we
are
giving
giving
you
know
so
yeah,
that's
it
I
mean,
although
the
basic
basic
concept
is
that
you'll
be
giving
it
a
patent
file,
but
the
power
that
you'd
be
getting
is
that
machine
would
be
figuring
out
a
lot
of
smart
things
for
you
right
now.
H
What
like,
without
what
I
demoed,
was
I'll
figure
out
automatically
that
grafana
should
be
deployed
after
prometheus,
but
there
would
be
a
lot
of
other
best
practices
that
mystery
server
would
be
enforcing
on
your
patents
file.
C
C
G
G
H
If,
if
the,
if
the
I
mean,
if
the
dependency
does
not
exist,
then
the
deployment
will
fail,
I
mean
it
will
not,
because
we
have
explicitly
marked
the
dependency
that
this
particular
service
will
have
a
dependency
honesty
and
if,
if
any
of
the
dependency
would
fail,
then
then
the
deployment
for
that
particular
service
will
also
fail
and
and
right
now,
of
what
I
have
like
in
the
code.
H
We
can
toggle
this
behavior
together
and
since
that
we
can't
even
give
not
that
we
have
that
query
parameter
in
here
right
now,
but
we
can
talk
about
that
behavior
right
now.
What
would
happen
if
any
of
the
deployment
of
basically
any
services
if
any
services
deployment
would
fail?
Entire
deployment
will
basically
fail
so
yeah
yeah.
H
If
the,
if
the
dependencies
like,
if
the
requirements
doesn't
fulfill,
then
it
will,
then
the
deployment
for
that
particular
service
will
escape
and
because
the
behavior
right
now
is
that
if
one
deployment
fails,
then
all
of
them
will
fail.
So
basically,
what
I'm
trying
to
say
is
then,
all
of
like
all
that
you
mentioned
this
particular
pattern,
unlike
pattern
or
all
of
the
deployment
will
kind
of
fail
just
because
one
of
the
dependencies
was
not
being
resolved.
C
A
lot
more
and
so
the
concept
here
is
that
what
folks
like,
like
michael
gefeller,
who's
on
the
call
and
rudkarsh,
and
any
of
you
that
want
to
participate
here,
that
you
can.
But
there's
been
a
few
folks,
who've
been
thinking
about
this
area,
and
this
is
the
area
is
about
patterns
and
about
about
allowing
people
to
use,
measuring,
to
exercise
a
pattern
and
by
pattern.
C
In
this
case,
what
we
mean
is
it's
sort
of
like
a
a
feature,
that's
commonly
used
by
that
you
use
a
service
mesh
to
perform
like
a
thing
that
you
do
with
a
service
mesh
and
there's
a
lot
of
things
that
you
do
with
the
service
mission.
Sometimes
it's
just
looking
at
metrics
how
much
traffic
is
going
across
my
services.
C
There's
like
there's
a
pattern
in
there
about
how
people
use
those
metrics
you'll
hear
people
espouse.
You
know
principles
of
methodologies
around
monitoring
the
p99
without
getting
into
specifics
the
ways
in
which
people
will
look
at
metrics
there's
a
pattern
by
which
there's
like
a
best
practice,
basically
a
pattern,
best
practice
kind
of
synonymous,
an
architectural
pattern.
If
you
build
an
architectural
customer
and
so
there's
a
bunch
of
things
that
you
can
do
with
the
service
mesh,
you
can
redirect
traffic
in
a
coordinated
way
to
facilitate
the
release
of
a
new
service.
C
For
example,
you
might
do
a
canary,
so
you
might
do
a
blue
green
test
or
a
red
blue
test
or
a
purple
neon
orange
test,
which
is
to
say
that
you
might
split.
You
might
have
two
versions
of
the
same
service.
Let's
say
that
gmail
has
there's
like
gmail
next
generation
and
they're
changing
their
interface,
and
they
want
to
test
out
that
interface
on
some
beta
users.
People
signed
up
well,
they
have
a
list
of
beta
users.
You
sign
in
you
authenticate
you
get
a
job,
a
jwt.
C
You
get
a
token
when
you
authenticate
you
send,
in
your
request
from
your
web
browser
to
pull
up
gmail.
It
says,
oh
that
person
that
token.
That's
being
presented,
they
they
are
on
the
service.
Mesh
will
look
at
that
and
say:
oh
that
person
is
on
the
beta
list.
We
should
send
them
redirect
their
traffic
to
the
beta
interface
great.
How
many
people
should
we
sign
up
and
how?
How
many
times
should
we
send
them
over
to
the
beta
interface?
C
Maybe
depends
upon
things
like
how
how
long
I'm
sorry,
how
how
many
errors
they're
receiving
there's
a
pattern
here,
there's
the
best
practice
for
what
to
do
and
to
distill
all
of
that
best
practice.
All
of
that
pattern
into
yaml
is
in
many
respects
what
karsh
is
sort
of
exploring
here
and
trying
to
articulate
and
do
well,
if
you're
going
to
redirect
traffic,
you
might
want
to
have
metrics
from
prometheus,
like
on
line
22
about
that
traffic.
Well.
Well,
what
metrics,
maybe
there's
a
particular
prometheus
configuration
that
you
need.
C
C
We've
had
ryan
zhang
of
the
open
application
model
project
oam
I'm
on
this
call
early
late
last
year,
presenting
on
an
application
model
and
we've
looked
at
it
a
little
bit
and
so
ludcrush
is
just
doing
some
early
prototypes
of
leveraging
part
of
what
that
model
is
trying
to
apply
it
to
the
domain
space
of
service
meshes.
C
Ultimately,
we
would
like
to
be
able
to
espouse
to
the
you
know:
tell
the
world
that
there
are
here's
60
practice
patterns,
70
80,
I'm
not
sure
how
many
there
will
be.
There
will
be
variance
they'll,
be
more
likely
they'll
just
be
you
know.
C
All
right,
hey,
there's,
there's
a
couple
things
we
didn't
get
to
cover
today
entirely
so
so
there's
a
lot
of
work
going
on
in
in
measuring
ctl
system
context.
Anirude
has
been
sort
of
headlong
focused
on
in
this
area,
so
he's
the
great
individual
to
go,
see
about
the
work
that's
going
on
there.
He
there's
it's
there's
a
lot.
C
It's
very
exciting!
It's
funny
because
in
one
hand
it's
so
simple,
yet
it's
so
confusing.
It's
a
very
interesting
place
to
get
into
so
do
see
any
road.
This
area
of
of
patterns
clearly
like
this,
was
sort
of
the
tip
of
the
iceberg
of
like
a
bunch
of
discussions
and
thinking
and
design
that
needs
to
happen,
and
so
we'll
be
doing
that
we'll
we'll
get
some
utkarsh
if
there's
a
design,
spec
or
anything.
C
Nice
good
deal-
I
know
I
know
who
has
been
thinking
about
this
and
he's
he's.
Gonna
want
to
get
his
hands
dirty
and
then
the
extent
anyway,
that
I
can't
wait
to
get
this.
The
it's
going
to
take
us
a
little
while
to
get
the
release
out
what
an
amazing
thing
it
will
be
hard
to
concisely,
articulate
how
much
is
in
there
and
yet,
what's
really
sort
of
frustrating
for
me
personally
is
in
the
end
measuring
other
than
mesh.
C
Sync
measuring
won't
have
done
be
doing
that
from
an
end
user
perspective
won't
be
doing
that
much
differently
or
much
more
for
the
end
user
kind
of
frustrating,
because
it's
so
much
better.
What
a
major
release
this
is,
but
we've
yet
to
take
full
advantage
of
those
architectural
components
just
yet
so
the
work
on
the
patterns
that
will
be
predicated
upon
that
there's
a
really
good
framework
for
adapters.
Now
that
can
be
used
to
deliver
something
like
a
pattern:
a
visual
topology
mesh
map.
C
Boy
yeah
istio
con
cfp
ended
yesterday.
We
tossed
in
tossed
in
something.
C
I
want
you
to
know
that
anyone
anyone
that
hangs
out
here
long
enough
you're
in
danger
of
being
put
on
stage,
so
it
just
comes
with
it.
This
community
works
a
little
bit
differently.
You
do
get
called
out
on
these
calls.
That's
because
it's
because,
like
me,
you're
here
to
grow
and
learn
things
and
the
way
that
you
do
it
is
by
being
uncomfortable.
C
That's
no,
no
pain,
no
gain.
I
think
that's
the
anyway
all
right!
I
can
ramble
on
forever
very
nice
to
talk
to
you
all
good
to
see
you
on
on
the
webcam.
The
guy's
got
more
wisdom
flowing
out
of
his
chin
than
I
can
okay
see
you
guys
tomorrow
on
the
newcomers,
call
see
you
for
the
community
calling
friday
bye,
y'all.