►
From YouTube: Meshery Development Meeting (Sept 30th, 2020)
Description
A demo of @smi_spec conformance improvements by @DhruvPatel0008.
A
All
right
welcome
everybody.
It's
four
minutes
in,
let's
start,
let
me
just
share
my
screen.
A
Oh
great
chris,
do
you
want
to
go
first?
I
think
you
the
way
to
leave
early.
A
B
C
By
the
way,
welcome
to
the
call
everybody
it's
the
mastery
development
call
and
so
kush.
If
do
you
want
to
give
a
little
bit
of
a
background
as
to
what
the
heck
nighthawk
is.
B
Okay,
so
nighthawk
is
in
their
seven
performance
benchmarking
tool,
which
is
maintained
by
envoy
community.
We
wanted
to
leverage
nighthawk
because
of
its
accuracy,
its
preciseness
in
the
test
results,
and
the
most
important
factor
is
that
it
soon
is
going
soon.
It
is
going
to
support,
distributed
load
generation
and
benchmarking.
B
B
B
So
we
have
got
a
pretty
high
talk
option
over
here
and
so,
let's
just
fill
in
the
fields
my
talk
demo
right
now
it
only
accepts
http
and
https
url
and
in
future
we
are
going
to
privilege
provision
a
grpc
server
for
nighthawk
so
that
we
can
get
more
variety
of
service
to
get
in
the
load,
testing
github.com.
B
B
B
D
A
Great,
does
anybody
have
any
questions.
C
I
have
critical
feedback,
you
know.
Does
anybody
have
can
any
so
as
we
go
through
this
I'll?
Have
some
questions
that,
hopefully
everyone
else
does
as
well,
something
that
I
would
encourage
everyone
on.
The
call
to
do
is
well
I'll
say
it
like
that
is
to
give
some
give
some
feedback
the
more
critical,
maybe
the
better.
Hopefully,
anyone
who's
demonstrating
is
encouraging
of
critical
feedback,
and
by
critical
I
don't
necessarily
mean
negative.
I
mean
thoughtful
and
critical
in
the
sense
that
you're.
C
Being
constructive
so
if
anyone's
participated
on
scrum
calls
in
the
packing
in
the
past,
where,
wherever
you
might
have
been,
or
maybe
even
more
than
a
scrum
call
more
like
a
an
iteration
demo
like
a
sprint
demo,
it's
kind
of
what
happens.
That's
kind
of
what's
happening
here
as
cush
and
as
others
go
to
demonstrate
the
functionality
that
they've
been
working
toward.
Is
that
it's
a
you
know
a
lot
of
times
some
of
the
times
these
are
like
half
baked
works
in
progress
some
of
the
times.
C
These
are
final
drafts,
irrespective
even
if
it's
a
totally
it's,
not
a
draft,
it's
fully
functioning
and
done
and
merged.
There's.
You
know
two
other
there's
a
couple
of
reasons
why
people
present
for
in
here
it's
one
to
help
and
help
is
one
to
request
feedback
two
to
show
off
the
good
works,
three
to
educate
the
rest
of
us,
keep
us
all
on
the
same
page.
C
C
I
encourage
you
to
pick
out
one
thing
here
that
you
think
either
can
be
improved
or
that
you
really
like
you
know
either
the
two
of
those
or
both
kush.
C
For
my
part,
I'm
curious
about,
if
I
don't
know
how
much
time
you've
been
able
to
spend
sort
of
looking
at
the
test
results
themselves,
but
you
know
understanding
that
each
of
these
three
load
generators
will
do
their
statistical
analysis,
look
slightly
differently,
there's
I'll
I'll,
say:
there's
a
dramatic
difference
between
the
way
that
the
analysis
is
charted
and
presented
between
fortio
and
wrk2.
C
It's
a
dramatic
difference.
You
can
see.
Wrk2
will
have
this
stepwise
cumulative
charting
of
the
latency
and
the
throughput.
C
B
C
B
C
To
expose
that
info,
how
disruptive
is
that
to
like
right
now,
the
chart
that's
produced?
It's
the
same
chart
same
it's,
it's
the
same
chart
that's
produced
for
fort
fortio
and
wrk2
we're
storing
the
performance
test
results
in
the
same.
I
assume
in
this
are
in
the
same
database
format
or
we.
B
C
A
A
A
Audio
okay.
C
B
C
Yeah
or
yeah-
and
I
don't
that's
not
necessarily
to
suggest
that
we
have
to
it-
might
be
the
case
that
yeah
I
mean
it
would
be.
It
would
be
nice
at
some
point.
We're
gonna
have
to
diverge
from
the
total,
the
total
uniformity
and
consistency
that
we
have
right
now,
as
we
consider
that
we'll
have
multiple
instances,
nighthawk
running
it's
so.
B
What
I
thought
about
that
was,
since
we
are
giving
out
the
results
in
the
canonical
sm
smt
smps
format,
so
we
can
leverage
the
descriptiveness
which
nighthawk
gives
us
while
we
download
the
results
from
over
here
so
in
the
yaml
file,
which
is
downloaded
by
nighthawk,
so
we
can
basically
give
them
all
the
details
which
nighthawk
has
captured
so
like.
If
you
will
see
here,
we
have
only
recorded
file,
it
is
five
percentiles
which
are
listed
here
by
default,
so
we
can
make
changes
over
here.
C
I
guess
your
question
probably
answers
one
of
my
other
questions,
which
was
an
acknowledgement
that,
while
we're
not
visualizing
all
of
the
results
right
now
that
we
are,
it
sounds
like
we
are
persisting.
The
full
details
of
nighthawk
results.
C
B
B
B
So
I
don't
know,
what's
the
issue
over
here,
that
if
the
ui
is
unable
to
leverage
all
the
percentiles
over
here
or
is
it
something
with
the
golang
server
or
the
handler?
But
it
is
quite,
it
is
quite
irregular
like
here
you
can
see
only
five
and
that's
the
one.
The
the
following
discrepancy
is
only
in
the
case
of
night
talk
like
for
fort
io.
You
have
five
well-defined.
C
Percentiles
yeah,
it
does
seem
like
something
that
we
would
want
to
work
toward
that.
Certainly
the
spec
itself
yeah.
As
we
look
at
the
protos
there
it
you
know,
I
would
think
that
it
should
certainly
be
extensible
to
allow
people
to
define
different
percentiles
to
be
more
granular
if
they
so
choose
that
that
doesn't
that
that
doesn't
hurt
anything
in
in
so
much
as
the
implementation
that
we
have
inside
of
mesherie
yeah.
It
would
seem
like
in
time
we
would
want
to
work
toward
exposing
all
of
that
granularity.
C
You
think
about
making
database
queries
over
the
data
if
it's,
if
it's
in
the
same
format,
irrespective
of
which
load
generator
is
used,
but
one
of
the
result
sets
one
of
those
has
greater
depth
of
detail.
It's
the
same
descriptors,
the
same
json
descriptors,
but
one
of
them
has
a
few
more
percentiles
that
shouldn't
mess
up
database,
queries
and
things
like
that
so
yeah.
I
would
think
that
we
would
want
to
try
to
expose
the
additional
detail
when
it's
available.
C
What's
a
little
more
thought-provoking
is
and
and
for
everyone
on
the
call,
I
don't
know
if
you
realize
this,
but
part
of
the
reason
I
mean
one
there's
a
couple
of
reasons
why
nighthawk
support
is
being
added.
I
think
for
some
of
you.
If
I
was
sitting
there
watching
this,
I
would
be
thinking.
C
C
Aren't
there
isn't
there
more
important
things
to
work
on
at
least
that's
what
I
would
be
thinking,
and
so
the
the
point
of
the
biggest
point
of
why
support
for
nightmark
is
being
incorporated
is
to
go
after
the
goal
that
isn't
that
we
haven't
done
that
we
collectively
haven't
done
yet,
and
that
goal
is
to
not
just
run
an
instance
of
a
single
load
generator,
but
it's
to
run
multiple
instances
and
that
actually,
I
think,
as
we
engage
with
the
envoy
team,
there
could
be
a
lot
of
work
there
or
there
will
be
some
work
there
and
it
depends
on
whether
or
not
they're
they're
carrying
the
burden
of
coalescing
result.
C
You
know
statistical
analysis
from
each
of
the
individual
nighthawk
instances
or
not
if
they
are
good,
there's
less
work
for
for
mashri
to
do,
but
either
way
we
think
about
it.
Kush.
Let's
say
that
there
were
three
nighthawks
running:
the
users
are
going
to
want
to
look
at
results,
statistical
analysis
from
four
perspectives
in
that
case,
look
at
it
from
a
perspective
of
each
of
the
individual
load.
Generators
that
we're
running
each
of
the
three
and
then
also
look
at
the
fourth
overall
sort
of
coalesced
results.
C
So
some
what
I
guess,
what
I'm
alluding
to
is
at
some
point
we're
going
to
have
to
break
into
the
charting
library
that
we're
using
here
anyway,
which
I
actually,
which
brings
up
a
good
point.
I
think
we're
using
two
charting
libraries
in
mashari
right
now.
One
is,
I
think,
the
one
that
we're
looking
at
here
is
chart.js
if
memory
serves
and
the
other
ones
that
we
use
for
graphing
out,
prometheus
and
grafana
charts
is
billboard.js.
C
B
Yep
so
like
basically,
the
ui
needs
to
be
a
little
more
descriptive
because
right
now,
if
you
will
see
that
wrk2
and
italk
are
provisioned
as
separate
containers
so
like
it
will
be
better.
If
we
have
a
nice
tooltip
over
here,
which
would
describe
all
the
load,
generators
and
the
way
to
run
them.
So
like
yep,
that's
a
pending
section.
C
All
right,
I'm
taking
notes
good
yeah,
keep
going
so
a
tool
to
an
explanatory
tool
tip,
maybe
that
talks
about
running
this
container
versus
running
as
a
library.
C
The
at
some
point
too.
I
think
a
common
question
that
people
are
going
to
have-
and
this
is
going
to
you
know
a
common
question
people
have
is.
Why
should
I
choose
one
versus
the
next
and,
moreover,
how
would
I
interpret
these
results
differently?
If
I
do
run
one
versus
the
next?
C
B
C
Nice
so
so
pratier
was
providing
some
additional
feedback,
suggesting
that
perhaps
perhaps
there's
some
ui
changes
that
can
be
done
here,
not
necessarily
specific
to
nighthawk,
but
in
general,
on
the
performance
test
page
and
pratt.
C
I'm
excited
to
think
that
you
might
focus
on
this
page
because
you're
right
there's
a
lot
of
thought
that
can
be
done.
A
All
right,
if
nobody
else
has
any
questions
for
kush,
we
can
move
on.
A
All
right,
one
of
the
articles
in
the
agenda
today
is
that
lay
five
is
participating
in
half
double
fest,
we'll
be
adding
a
program
card
by
today.
There's
a
pr
open
for
it.
If
anybody
wants
to
go,
have
a
look,
adding
suggestions
that
you
want
to.
A
For
those
of
you
who
don't
know
how
to
the
fest
is
a
get
to
know,
git
event
that
happens
throughout
the
month
of
october,
and
you
get
some
swag
for
completing
about
four
peers
in
any
active
and
live
organization
that
is
registered
for
the
event.
A
I'm
bringing
this
up,
because
if
anybody
has
any
ideas
for
possible
issues
that
can
be
created
across
the
mastery
or
the
s,
snp,
repos
or
any
of
the
adapter
repos
now
would
be
a
pretty
good
time
to
bring
it
up
because
it
starts
tomorrow
and
those
issues
can
be
created.
A
A
A
Please
do
suggest
any
issues
that
you
want
to.
This
is
a
pretty
good
way
of
getting
word
out.
A
It
hasn't
because
the
program
card
hasn't
been
merged,
yet
it
will
get
it.
A
A
All
right
moving
on
through,
we
can
go
ahead
with
yours.
G
Yeah,
so
this
is
a
very
small
change,
but
I'm
sure
he's
gonna
take
this
time
to
tell
what
he
wants
confidence
to
be
the
basically.
I
shifted
the
smi
test
results
into
a
new
navigation
library
and
currently
the
conference
navigation
bar
and
the
smi.
The
results
bar
lead
to
the
same
sms.
We
have
not
decided
what
we
want
to
have
in
the
conformity
space
for
now
at
least.
C
Oh
nice,
it
drew
questions
and
feedback.
Then
do
we.
I
wonder
if
we
might
be
able
to
be
rid
of
that.
First
faucet,
that
first
air.
C
G
G
C
C
C
Speaking
of
the
table
that
we
have
in
the
smi
results,
it
has
a
shadow
to
it.
The
the
second
table.
I
wonder,
yeah
what
it
would
look
like.
G
C
Gotcha,
when
you
go
to
the
performance
test
results-
and
we
do
a
quick
comparison,
so
you
might
have
to
run
a
quick
performance
test.
Oh
wait.
One
second.
G
C
Oh
sorry,
okay,
so
yeah.
So
here
it
is,
there's
no
there's,
no
shadow.
It's
just
a
an
additional
set
of
data.
C
Some
like
the
challenge
that
we're
having
the
reason
why
a
drop
shadow
was
used
on
the
smi
results
was
because
it
sounded
like
what
you're
saying
is
without
it.
There
we
have
it's
like
really
hard
to
offset
the
table
and
table
effect.
G
C
Yeah,
if
it
does
hopefully
I
I
could
see
that
maybe
it
wouldn't
work
or
something
but
but
like.
Hopefully
we
would
try
to
make
it
the
same
and
then,
but
if
it's
not
working,
I
understand
if
we
need
to
deviate.
C
So
this
one
right
now,
this
one
has
the
ellipsis,
on
the
right
hand,
side
to
get
a
get
a
modal
with
like
to
so
so
that
here
that
we're
getting
a
visualization
and
if
the
user
wants
to
get
the
tabular
format
for
that
visualization,
they
yeah
they
click
on
the
ellipse
and
which
I
think
is
a
an
issue
that
continues
to
plague
us
in
some
areas,
but
they
they
normally
get
the
modal.
That
kush
was
demonstrating
earlier,
like
one
like
this.
C
C
Gotcha
but
yeah
anyway,
I
guess
that's
sort
of
just
maybe
the
biggest
general
feedback
there
is
to
the
extent.
Today
we
can
reuse
and
provide
a
consistent
experience.
We
would
want
to
start
from
there
and
deviate,
if
necessary,
hopefully
not
necessary,
but
deviate
if
needed.
G
Sure
what
I
can
do
is
I
can
show
you
some
different
formats.
If
you
like
of
that
table
itself,
I
guess
the
detail
thing
can
be
implemented
should
be
implemented.
I
will
definitely
do
that,
I'm
not
sure
about.
If
you
want
to
change
the
ui
which
I
decided
for
xmi
but
yeah,
I
can
show
you
some
more
instances.
C
Nice
josh
and
fonzie
had
a
couple
of
comments
as
well.
A
Yeah,
if
you
want
to
speak
up,
please
go
ahead
fongie.
I
don't
think
your
audio
situation
is
that
good.
G
A
I
think
he
means
the
right
most
one
in
the
top
number.
Oh
I'm
so
sorry
my
accent
is
setting
set
up.
E
G
The
the
mui
table,
which
we
are
using
of
certain
things,
which
you
have
disabled
right
now,
like
downloading
the
data
printing,
the
data
searching
for
specific
things
and
like
choosing
which
columns
which
you
want
to
have
right.
We
have
just
enabled
searching
in
columns
for
results
and
for
sms.
I
also
added
filters.
G
C
Yeah
or
I
think
what
funji
was
saying,
maybe
we
wouldn't
have
the
collapse
all
show
hide
all
if
there's
only
if
there
are
zero
or
one
results,
but
only
when
there's
like
two
or
more.
G
G
D
C
The
yeah,
actually
the
smi
test
results
table
has
or
at
one
point
had
had
some
pretty
cool
options.
I
think
some
options
that
you
wouldn't
find
on
the
performance
test
table
yeah,
I
disabled
them.
C
G
Yeah
sure,
but
the
problem
over
there
would
be
right
now,
if
you
click
on
that,
it
will
just
give
you
the
results
which
is
shown
over
here.
It
won't
give
you
the
details
results
to
which
we
are
looking
to
get
basically
the
this
results.
It
won't
give
you
that
gun
out
of
the
box.
We
need
to
customize
it
to
give
those.
It
will
right
now
only
give
the
results
which
are
viewable
by
the.
C
Yeah,
there's
a
there's
a
few
improv
I
mean
yeah.
Actually,
as
we
look
at
the
performance
test,
results
table,
there's
probably
a
few
improvements.
Some
one
thing,
that's
not
obvious
to
everyone,
including
myself,
is
it
takes
the
user
check,
marking
the
box
for
a
given
result
in
order
for
two
things
to
show
up,
so
one
thing
that
will
show
up
is
yeah
in
the
upper
right
hand
corner.
This
is
a
really
neat
feature
the
ability
to
compare
two
tests
together
so
yeah,
if
you
click
it
it'll
give
well.
C
This
is
maybe
not
the
best
example.
I
guess,
but
but
it'll
it'll
overlay,
two
charts
on
top
of
each
other,
so
you
can
visually
difference
between
on
the
result.
Sets
which
is,
I
think,
is
really
neat.
It's
like.
Oh,
if
you
made
a
change
in
your
environment
now
you
can
see
what
the
performance
difference
is,
but
so
that's
good.
It's
not
always
the
most
intuitive
thing
like
people.
It
takes
people
check
marking
the
other
thing
that
also
how.
C
C
G
Export
hello,
I
can't
hear
you
for
some
reason.
I
can't
either
actually.
C
There's
some
things
for
everyone,
maybe
to
think
about
here
about
whether
or
not
so,
if
you
just
check
mark
one
of
them
yeah,
then
you
get
that
download
yeah,
and
I
don't
know
that
that
feels
kind
of
buried
kind
of
you
know
hidden
it
to
me.
Anyway.
C
G
The
thing
which
we
had
before
in
the
conference,
so
I
just
reserved
a
special
icon
right
over
here
to
download
stuff,
but
even
in
that
case,
if
nothing
is
selected,
you
will
download
all
of
the
things
and
if
some
something
is
selected,
it
is
anyways
giving
us
the
option.
Okay,
it
is
only
giving
us
option
for
one
of
the
downloads.
C
I
think
one
way
to
help
with
discovery
for
users
to
understand
that
there
is
functionality
here
is
to
potentially
show
the
download
icon
just
in
a
disabled
state
up
until
the
point
that
someone
check
marks
a
row
and
then
enable
that
state
could
be
the
same
thing
for
the
comparison
as
well,
which
I,
which
again,
like
I
I'll,
say
it
again
like.
I
think
that
that's
a
really
cool
feature,
but
isn't
always
people
don't
always
know
it's
there.
G
C
So,
for
everyone's
context,
the
when
you're
interpreting
these
results,
what
the
user
has
done
has
gone.
They've
gone
over
to
a
specific
service,
mesh
they've
told
mesri
to
run
a
set
of
tests.
There's
there
will
probably
be
25
35
tests,
individual
tests.
That
will
be
run
where
the
a
given
mesh
will
be.
C
C
It
might
make
sense
to
do
that.
We
squash
that
a
little
bit
or
sort
of
make
it
yeah.
G
C
And
yeah.
C
Of
the
leftmost
column
yeah,
if
the
id
is
shortened
up-
or
you
know,
truncated
or
curtailed,
yeah
the-
I
guess
I
don't
know
when
people
come
and
search-
it's
probably
the
date
that
they're
most
that's
like
the
primary
key
for
them,
so
to
speak,
and
second
to
that
is
probably
the
type
of
service
mesh.
So
all
of
these
fields,
all
these
columns
are
sortable
or
no.
G
G
C
Do
you
know
is
the
filter
that
we're
using
I'm
sorry
the
sort
that
we're
using
on
the
back
end?
Is
that
just
an
ascending
date,
or
is
that
by
date,.
C
But
what
do
you
want
it
to
be?
Yeah,
probably
the
default
would
just
be
yeah
like
latest
at
the
top
last
last
round,.
C
Interesting
yeah,
if
so
this
is,
we
were
talking
a
little
bit
earlier
about
the
notion
that
there's
we
can
begin
to
bring
in
some
in-line
documentation
on
the
demonstration
that
kush
was
giving.
C
This
is
probably
another
example
where
there's
a
little
more
info
that
people
are
going
to
want
that
are
generally
going
to
want
to
understand,
and
it's
like
it's
one
of
the
questions
that
people
are
going
to
have
is
when
they
see
a
test
result,
fail
they're
going
to
yeah
they're
going
to
want
to
know
the
reason
for
sure
and
the
reason
could
be
right.
Now
it's
not
found
it's
a
very
short
reason.
C
This
crd,
this
crd,
is
not
found
and
it
could
be
longer
so
we'll
want
to
have
a
plan
for
how
to
control
that
length
and
and
yet
still
give
people
full
details.
C
That
makes
total
sense
for
the
question
of
like
what
is
this
test
and
what
is
being
tested.
That
makes
a
lot
of
sense.
It
might
be
again
like
if
we
do
it
in
line,
it
might
be
a
modal
where
that
documentation
is
relatively
static
and
we
can,
but
for
the
second
question
which
is
like
okay,
I
understand
what
this
test
is,
but
why
did
it
fail
in
this
case?
D
C
G
Just
to
the
extent
immediate
thing
would
be
a
proper
break
to
give
on
more
details,
but
I
guess
in
long
term
a
multi-line
box
would
like
me
force
it
out.
I
don't
know.
C
Yeah,
it
could
be
like
that.
You
know
that
ellipse
on
the
other,
it
could
be
a
way
of
when
someone
a
tool
tip
would
be
kind
of
a
poor
man's
solution.
A
multi-line
row
is
probably
hard
to
read
after
a
while,
because
it
could
go
on
for
many
many
rows
and
it's
kind
of,
but
that
ellipse
or
something
for
additional
details
with
a
modal
that
gives
kind
of
all.
You
know
everything
else
might
be
helpful.
G
C
Probably
a
pop-up
like
you
could
there
could
be
a
tool
tip
that
gives
the
whole
string
and
that
might
suffice,
but
then
yeah
the
a
modal
like
a
pop-up,
and
it
could
be
that
if
that
modal
could
do
two
things,
it
could
do
what
you
were
saying
about
answering
the
question
of
like
what
is
this
test
and
what's
being
tested
or.
F
C
And
yeah,
and
by
the
way,
speaking
of
nomenclature,
does
everybody
else
call
they?
You
know
kind
of
refer
to
the
menu
as
a
hamburger
that
the
three
the
three
stacked
anyway,
as
we're
all
getting
onto
the
same
anyway
pratik.
It
also
had
some
other
another.
He
was
just
saying:
hey,
you
know
an
faq
on
the
left.
Menu
for
tests
that
are
failing
a
pop-up
might
block
the
result
below
the
current
results.
When
we're
comparing
two
results.
C
A
A
I'm
looking
forward
to
thank
you,
so
much
is
afk
for
the
moment,
he'll
be
giving
us
his
tutorial
next
week
or
hopefully
in
the
community
meeting.
That's
it
for
today.
Folks,
thank
you
for
joining
us.
Thank
you
for
coming
on.
Thank
you
for
everybody,
for
the
tutorials
kush
and
drove
see
you
next
week.