►
From YouTube: Diagnostics WG meeting - May 15 2019
Description
A
A
Okay,
the
one
thing
I
would
announcer
or
share
is
that
Mike
Coffman
is
I.
Guess
he
couldn't
make
it
again
today
and
he
kind
of
reached
out
to
me
saying
he's
not
sure
how
often
he's
gonna
be
able
to
make
it.
So
we
were
just
wondering
if
there's
anybody
else
who
wants
to
volunteer
to
tour
and
you
don't
necessarily
have
to
volunteer
immediately.
But
if
you
can
think
about
that,
and
let
us
know
that
would
be
great
or
if.
A
Now,
that's
that's
good.
What
take
someone
William
Pierce?
What
does
it
mean?
What
I'll
do
is
I'll
make
sure
to
set
you
up
so
that
you
know
how
to
start
the
soon
meeting
start
the
streaming
and
then
basically
you
know
through
the
agenda,
and
if
you
can't
make
your
time,
then
you
know
it'd
be
up
to
you
to
find
somebody
else.
You
could
share
or
leave
the
meeting
for
you
is
everybody
the
issues
themselves
like
the
creation
of
and
I
guess,
the
other
thing
it's
related
is
making
sure
the
minutes
get
captured.
A
But
the
generation
of
the
issue
is
automated
from
the
from
the
calendar.
So
like
the
issue,
one
here
to
99
was
automatically
generated
three
days
on
things
that
are
tied
to
the
agenda.
I
reads
the
calendar
and
then
says:
okay
I'll
generate
it.
The
minutes
that
Google
Doc
is
also
created,
but
the
manual
it
is
a
manual
step
to
take
that
and
actually
commit
the
minutes
themselves
as
a
PR,
okay,
okay,
so
yeah
think
about
it.
Let
either
Mike
her.
The
other
Mike
know
somebody's
interested.
A
In
doing
this,
it's
not
too
much
work,
but
you're
gonna
be
at
the
meetings
in
LA.
You
know
it's
not
a
whole
lot
extra
okay
I'm
going
to
finance,
then
the
first
thing
on
the
agenda
is
number
24.
66
download,
I'm,
just
gonna
open
this
up
and
see
who
added
here
or
whoever
did
add
it
wants
to
jump
in
that's
good.
C
A
D
D
A
A
A
E
A
A
E
A
E
Would
that
would
probably
be
one
possibility?
I
mean
they're
like
I
guess
it
would
actually
be
a
good
thing
Institute
to
talk
about
it
at
the
summit
and
then
discuss
how
exactly
it
should
be
implemented,
because,
even
if
they
are
willing
to
do
it,
it's
probably
actually
how
it's
going
to
be
done
so
that
it's
not
going
to
cost
a
lot
of
performance.
A
Okay,
so
do
we
have
a
session
where
you
think
that
this
gun
can
fit
into
I'm
just
trying
to
make
sure
that,
like
the
the
summit
is
only
a
couple
weeks
away,
so
I
think
that's
a
great
idea
to
try
and
get
people
together
and
discuss
there,
but
how
we
make
sure
that
that
gets
on
to
one
of
the
agendas?
And
we
don't
forget
about
it.
I.
A
A
Okay,
so
if
we
can
all
try
and
remember
too,
if
we're
going
to
the
promises
session
to
try
and
bring
that
up
as
a
as
a
topic,
I
think
that
would
be
a
good
way
to
to
move
that
forward,
make
a
concrete,
like
then
capture
the
concrete
naps
next
steps.
In
terms
of
like
do,
we
need
to
ask
for
volunteer,
or-
and
at
least
if
we
could
document
you
know
at
a
top
level,
what
what
what
it
would
actually
do
and
how
it
would
work.
That
would
be
great.
A
B
B
Are
we
created
this
document
which
breaks
down
to
the
most
common
usage
and
user
use
cases
and
recommended
tools
for
that
which
I'm
mentioning
some
gaps,
also
listing
the
tools,
so
first
of
all,
I
saw
in
the
Google
Docs
statistics
that
some
people
saw
the
document,
but
not
many.
So
we
would
need
to
basically
agree
on
some
points
around
the
document
and
finalizing
the
document,
and
there
are
some
open
questions
like
seeing
your
line
meant
around
the
course.
B
What
are
the
supported
use
cases,
as
so
on,
so
to
move
forward
on
that
I
proposed
a
set
on
the
collaborate
or
summit
and
I
just
checked
the
the
agenda
and,
it
seems
seems
to
be
it's
made
it
to
the
agenda,
and
it's
also
related
to
the
next
item
on
our
conversation
today.
The
best
practices
I'm
not
sure
that
was
of
any
progress
around
the
best
practices
guide,
but
basically
it
could
be
built
up
of
the
use
cases.
I
think.
A
B
We
don't
next
steps.
Is
that
yes
and
the
document
there
are
some
coins
which
I
believe
it
would
need
a
language
under
the
Emma
Street.
It
has
some
proposal
around
a
TS
million
enhanced
and
I
just
seen
the
agenda
that
from
the
question
that
you
are
leaving
the
model
ecosystem,
it's
at
the
same
time
when
this
session.
So
maybe
we
should
ask
the
organizers
who
I'm
sure
it.
A
It
that
would
be
great.
Okay.
Is
there
anything
that
we
should
talk
about
the
people
we
have
today
here,
like
you,
said,
there's
a
I'm.
A
Okay,
okay,
so
we'll
say
people
please
really
advance
the
summit's,
and
next
step
is
we'll
discuss
there
and
I
think
best
practices.
Unless
anybody
else
has
that
is
well.
You
know,
based
off
the
diagnostic
user
journeys,
we'll
talk
about
how
we
sort
of
build
off
those
into
starting
to
create
the
best
practices.
Okay,.
C
F
That
yeah
I
went
through
the
user
journeys
documentation
and
the
could
be
sections
where
there
is
a
clear
dependency
on
the
best
practices.
That
means
once
we
zero
down
onto
the
user
journeys,
then
the
information
from
the
often
proven
best
practices,
but
there
could
be
other
areas
in
the
best
practices
where
you
can
directly
go
ahead
and
act
on
it.
F
For
example,
the
crash
document,
which
I
am
performing
some
of
the
things,
for
example,
the
prerequisite
for
collecting
the
chorim's
prerequisite
for
the
machine
to
be
prepared
for
collecting
the
dams,
etcetera
can
actually
be
developed
in
as
well.
So
I
agree
that
we
should
have
a
together
discuss
this
one
in
the
upcoming
collaborator
summit.
In
fact,
I
was
planning
to
have
that
in
the
diagnostic
summit.
F
B
F
B
Basically,
my
idea
is
that
if
so
it's
a
one-hour
session
on
correctly
I
am
planning
to
spend
the
first
15
minutes
kind
of
giving
the
context,
probably
some
kind
of
short
presentation.
Then
we
have
some
kind
of
action
items
which
needs
alignment
and
then
then,
in
the
rest
of
the
time
we
can
see
more
about.
F
On
a
separate
note,
or
maybe
a
related
note,
as
part
of
the
best
practice
initiative,
there
is
a
PR
which
I
raised.
Number
285
is
basically
talking
about
the
basic
prerequisite
for
collecting
customs
or
preparing
a
machine
for
being
enabled
for
the
quorum.
It
doesn't
look
like
it
has
any
dependency
or
any
anything
that
needs
to
have
a
consensus
at
the
moment
other
than
people.
Having
a
look
at
that
and
talk,
you
know
just
provide
their
views
about
the
structure,
the
overall
layout,
how
the
information
is
presented,
etc.
F
A
F
2
6
to
54,
basically
on
the
on
the
meeting
we
had,
we
agreed
that
we
identify
with
top
symptoms
and
the
tools
those
will
help
to
address
the
symptom
types
and
then
generate
the
guide
Dena.
We
identify
the
developer,
who
is
called
all
of
the
kite
etc.
But
at
the
moment,
based
on
the
discussion
with
Peter
I
guess
we
should
take
that
and
then
we
have
a
discussion
in
the
collaborator
Summit
I'm
happy
to
have
that
action
until
then
move
on
till
then.
F
Because
I
mean
the
team
who
actually
proposed
and
came
up
with
this
set
of
priorities
or
the
set
of
tools
was
actually
less
than
ten
I
play.
I,
don't
remember
the
number,
but
if
you
are
having
a
wider
discussion
in
the
collaborator
summit
and
if
you
are
coming
up
with
a
more
collective
decision
on
that,
probably
it's
that
the
priorities
can
change.
I
mean
right
now,
it's
a
memory
that
is
the
top
one.
It
would
be
possible
that
these
things
can
slightly
differ.
C
So
we're
on
it,
we've
been
talking
to
the
federal
team
there
for
past
two
weeks
and
they're
working
on
a
client
library,
that's
more
territories
and
they
would
use
case.
So
it's
gonna
ball
a
lot
of
questions
not
sure
if
anyone
here
is
good,
like
look,
I
prefer
to
I
think
that,
but
it's
probably
gonna
post
some
changes
to
use
their
new
client
library
or
be
let's
open.
It
note
to
introduce
it
to.
G
Yes,
they're
also
working
on
a
another
way
of
adding
Chris
points
that
involves
a
simple
post
lambda.
So
basically
you,
you
basically
add
a
trace
point,
that's
easier
to
use
and
much
nicer
looking
than
then
just
a
plain
macro,
and
that
also
allows
customizability
that
in
the
lambda
you
can
construct
below
that,
you
can
attach
so
should
be
able
to
be
a
really
nice
API.
G
A
A
A
A
D
One
and
I
need
to
use
some
rapper
script
in
the
middle
that
actually
tried
to
capture
a
CD
error
of
old
processes
and
try
to
parse
them
to
get
a
WebSocket
URL
that
I
need
for
the
backing
it.
Sometimes
it
works,
but
sometimes
it
doesn't
work.
The
problem
is
that
sometimes
the
process
can
spawn
child
process
and
by
default
propagate
environment
as
default,
be
here,
but
can
totally
ignore
STD
error
or
totally
ignore
any
STD
out,
and
in
this
case,
I
propagated,
inspect,
barricade
flock.
D
This
process
will
be
paused
until
any
connection
actually
appear,
but
since
I
ignored
the
air
I
cannot
actually
get
web.
Second
back,
so
I
am
in
state
when
I
cannot
do
anything
at
this
point,
and
the
root
of
this
problem
is
that
we
actually
use
two
different.
Let's
say,
transfers
to
actually
pass
inspector
wave
pass
flag
to
an
inspector
and
get
port
back,
and
is
there
some
minor
issues
with
this
approach
that
you
need
to
parse
STD
error?
D
D
Don't
control
and
now
the
case
I
can
easily
imagine
some
clouds
case
when
you
run
just
some
node
processes
from
time
to
time,
using
Chrome,
for
example,
and
you
don't
want
to
have
some
rappers
that
you
need
to
write,
and
this
factor
should
be-
can
be
very
complicated
in
case
when
it
should
work,
for
example,
support
of
work
in
process
when
node
for
process.
It
expect
that
all
this
own
message-
listeners
I
installed
before
anything
and
in
this
case,
is
hard
to
write
this
far
after
in
ways
that
support
all
use
cases.
D
So
it's
possible,
but
it's
hard
proposal.
I
just
go
right
to
the
proposal
and
then
I
will
share
some
other
alternatives.
It
I
can
see
here,
I'd
like
to
propose
new
flag
denote
this
flag.
Let's
go
inspect
store
and
this
flag
will
take
a
folder
as
an
argument,
and
some
external
tool
should
create
this
folder
and
responsible
to
make
it
writable
to
what
should
prepare
this
folder
and
to
clean
up
this
folder
is
the
end
of
tool
work.
D
For
example,
my
node
debugger
will
create
at
ahead
of
time
and
this
right
after
and
when
this
flag
is
fast,
not
process.
When
inspector
started,
we
will
put
in
this
folder
some
files
that
describe
WebSocket
URL
this
guy
some
another
additional
information
buzz
is
noth
process
if
needed,
and
then
note
will
make
best
effort
to
remove
this
file
at
the
end.
But
at
the
same
time
we
cannot
provide
strong
guarantee
here,
just
because
not
can
be
just
killed.
D
Actually,
the
proposal
was
born
after
Burfict
I
mean
I
created
the
proposal
after
numerous
of
attempt
to
implement
something
else.
For
no
debugger
and
alternative,
she
was
alternatives
that
unmentioned
in
the
back.
That
actually
is
required
to
just
inject
some
JavaScript
in
note,
and
the
JavaScript
can
use
inspector
module
to
get
WebSocket
and
be
as
flexible
as
possible
to
store
this
data
to
pass
this
data
to
another
place
and
do
whatever
actually
possible
to
do
at
the
same
time.
D
This
magic
script,
first
of
all
is
definitely
will
add
a
lot
to
not
process
myself
and
if
I
measure
startup
startup
time
of
my
node
process,
I
don't
want
something
to
be
injected.
Otherwise,
all
my
numbers
are
not
related
with
real
production
and,
for
example,
it
will
add
some
something
to
heap
and
erica's
its
first
problem.
Second
problem:
if
I'd
like
to
implement
something
like
wait
for
connection,
I
need
to
introduce
a
function
in
user
land
that
can
block
everything
until
some
condition.
D
D
Third
alternative
is
provide
some
way
for
inspector
to
allocate
the
support
in
predictable
way.
It
only
will
work
with
WebSocket
and
we
can
say
okay
inspector.
We
always
will
try
next
for,
for
example,
in
this
case,
we
need
to
have
some
demons
that
will
ping
all
ports
all
the
time
and
it
sounds
like
just
waste
the
CPU
cycles.
To
do
it
and
there's
no
strong
signals
that
some
no
textual
started
last
one
reversed
connection
instead
of
passing
flags
at
first
node.
D
It's
this
approach
definitely
will
work,
but
it
requires
some
demon
to
running
can
background
as
well
all
the
time
just
because
if
we
for
some
reasons,
if
our
demon
is
restarting
because
it's
crashed
and
we
need
to
restart
it,
it
happens,
then
we
can
miss
this
little
connection
in
the
middle
and
in
some
race
conditions
it
might
happen
and
when
it
happen
it
definitely
will
be
super
hard
to
do.
Backward
textual
is
wrong
there.
So
the
approach.
G
That
many
production
debuggers
use
rates
of
there's
a
reverse
connection,
where
the
the
application
connects
to
a
debug
server,
I.
Think
and
Google
cloud
would
have
like
a
online
debug
service
production.
That
runs
like
this
so
like
what
are
the
reasons
for
the
for
the
demon
to
actually
be
start,
like
I?
Think,
if
you're
expecting
a
tool
that,
like
like
NDB
here
under
UI,
to
be
offering
to
debug
a
client
application,
I
would
expect
that
it's
reasonable
to
expect
that
that
is
functional
and
working.
Well,
the
user
wants
to
debug
the
gurus.
D
D
I,
definitely
current
demos,
it's
just
always
there,
but
at
the
same
time
I
can
just
use
cron
and
schedule
it
and
run
it
from
our,
for
example,
and
it
will
record
some
information,
I
mean
for
five
seconds
or
something
like
this
and
then
I
can
implement
it.
On
top
of
daemon,
I
can
wrap
and
some
light
demons
that
will
store
to
the
file
system
actually,
but
at
the
same
time
it
that
complexity
is
that
I.
Don't
know
how
and
I.
A
A
A
My
first
thought
was:
should
this
be
more
generic
to
like?
If
we
did
this,
should
it
be
something
like
no
discovery
as
an
option,
as
opposed
to
just
strictly
tied
to
inspector?
So,
like
you
know,
one
thing
that
might
be
written
to
the
to
the
file
is
the
inspector
port,
but
similarly
you
might
want
the
bid
so
that
you
could
like
find
all
the
note
instances
if
you
wanted
to
like
kill
them
all
or
send
them
all
signals,
or
something
like
that.
I
D
I
I
So
those
files
will
accumulate
because
I'm
able
to
die,
and
if
you
have
a
cluster,
the
just
spawns
workers
for
a
long
time,
lawn,
ryan'
job
to
just
pond
worker,
and
sometimes
they
work
at
die
before
right
in
the
file,
and
it
was
just
making
a
mess
in
that
folder
and
at
some
point
you
run
out
of
disk
space
and
do
go
after
instance.
Is.
D
Down
you
have
solved
many
note
instances
that
are
killed
there.
Firstly
killed.
You
have
some
other
problems,
but
it's
a
nice
point.
I
like
this
point,
I,
don't
have
a
good
solution
for
this
and
my
main
point
at
the
same
time
that
I'd
like
to
avoid
the
user
leant
since
inspector
protocol
actually
ready
to
be
used
in
production.
D
A
D
No,
if
you
you
first
of
all,
yes,
we
need
some
special
kind
of
requires
that
have
some
girl
ranches,
as
this
requires
just
said
by
the
way,
is
three
key
just
because
someone
else
will
use
the
same
require
and
who
should
be
trust
in
this
case
and
but
if
we
have
some
way
to
say
around
the
script
cost,
then
it
might
work.
What
do
you
will
miss
bootstrap
if
you'd
like
to
the
back
node,
bootstrap
and
profile
node
bootstrap
that
might
be
significant
for
some
simple
node
apps
you?
F
Believe
the
file
based
solution-
it's
not
probably
very
much
scalable
and
RCN,
because
you
know
it
requires
a
lock
when
multiple
node
process
want
to
write
into
that
simultaneously
and
at
times,
if
one
of
the
VM
crashes
abruptly
the
file
will
be
in
an
inconsistent
state
and
every
time
you
will
need
to
sync
all
the
running
know
and
the
content
of
the
file.
There
could
be
various
complexities
well.
A
Logan
I
think
the
cleanup
side
of
things
can
I
think
what
I've
seen
done
is
if
you
write
the
pin
to
the
file
and
then
you
have
a
tool
which
makes
those
files
it
can
basically
like
when
you
start
up
the
tool,
say
well,
there's
no
kid
for
this
file.
I'm
just
gonna
delete
it
because,
honestly,
it's
no
longer
exists
right.
Yes,.
D
D
F
So
the
actual
requirement
is
to
figure
out
the
inspector
port
that
is
used
by
a
process
which
was
started
earlier
or
crosses
started
by
some
other
agent,
not
in
your
control.
Basically,
an
arbitrary
node
process
is
running
and
you
want
to
know
which
port
is
being
used
and
then
one
of
the
more
cleaner
solution
would
be
used.
The
diagnostic
report
as
an
option,
because
that
gnostic
report
anyway
Curry's
the
whole
internals
of
the
node.js
vm
components.
D
D
B
A
D
A
Landed
trees.
Do
you
remember
we're
in
the
bootstrap
sequence
it
gets
generated.
Well,
I
guess
it
doesn't
know
it's
generated
later
on.
F
To
be
honest,
I
see
the
bootstrap
sequence,
keep
changing
quite
frequently
already
over
the
last
couple
of
months.
So
I'm
not
sure
where
the
report
is
getting
loaded
in
the
whole
series.
F
D
D
A
D
A
Okay,
thanks
that
was
a
good
discussion.
I
look
forward
to
seeing
how,
if
that
works.
So
if
using
the
node
report
works
or
not,
so
let
us
know
for
sure,
yeah
I,
don't
care.
Okay,
that's
the
end
of
our
agenda.
We're
also
only
have
one
minute
left,
but
if
anybody
has
any
other
things,
they
want
to
add
to
the
agendas
now's.
The
time.
D
I'd
like
to
say
that
during
class
a
couple
days,
I
read
a
lot
of
stuff
about
inspector
protocol
on
the
Twitter
and
I'd
like
to
find
the
channels
that
I
can
use
to
communicate.
That
inspector
protocol
is
ready
for
production
just
because
there
are
a
lot
of
miscommunication
on
this
part
from
our
side
from
my
side
maybe,
and
we
ate
in
the
flow
site
and
people
first
question
from
people
with
always.
Is
it
ready
for
production?
Can
we
use
it
in
production
and
a
lot
of
answers
like
no?
And
it's
not
right.
I
D
C
C
G
This
lights,
the
the
one
concern
I've
had
is
the
the
permission
granularity.
The
last
time,
I
checked,
is
to
course
right.
So
either
you
have
the
inspector
or
you
don't.
So
so,
let's
say
if
you
only
want
to
enable
like
expose
the
CPU
profiler,
like
you
have
a
port
on
which
the
inspector
is
listening,
then
you
expect
a
profile
to
connect
to,
but
let's
say
something
bad
happens
now.
Wait
a
second
second
I
have
full
control
of
my
production
machine,
where
somebody
cannot
run
the
debugger
into
arbitrary
code.
Injection
I.
D
Have
an
addition
for
this:
first
of
all,
we
can
introduce
inspect
pipe
and
you
don't
need
to
expose
WebSocket
at
all.
It's
a
local
one
and
fight
for
something
secure
believe
and
at
the
same
time
we
definitely
can
build
something
to
know
that
will
enable
and
a
couple
domains.
But
its
usual
knows
this
better
yeah.
So.
G
I
think
one
thing
I
would
really
like
is
like
at
startup
like
if
I
could
listen,
the
domains
I
wanted
to
enable
like
it
should
be
in
the
control
of
the
user.
So
even
on
a
pipe
like
a
pipe
maker
connect
to
something
else,
then
may
inherently
in
Edward
idli,
throwing
too
many
permissions
right.
So
the
security
needs
to
be
locked
down
by
default
and
opened
up
as
as
the
need
is
for
the
user.
So
if
there's
something
from
feature
like
that
with
when
the
spectra
quite
useful,
so.
I
D
I
G
A
You
know
kind
of
writing
up
a
blog
post
that
says
hey.
We
think
this
is
ready
for
production
use,
because
you
know
here's
the
data
we've
collected
that
convinces
us
you
know
like
the
overhead
is
not
going
to
be
an
issue
because
you
can
turn
on
these
things.
Security
is
okay,
you
know
to
address
what
Ali
is
and
then
discussing
that
in
the
workgroup
here
to
say
like
are
we
all
comfortable
with
this?
Is
the
message
we're
sending
about
the
inspector
protocol
from
the
working
group?
A
Once
we
have
that
consensus,
then
we
can
get
the
the
foundation
communications
people
to
help
us
promote
it,
like
I,
think
we'd
want
to
land
something
a
document
in
our
repo
saying
you
know
this
is
this
is
our
consensus.
We
recommend
that
it's
okay
or
not,
and
then
we
can,
you
know
all
help
publicize
that
through
the
foundation
and
other
resources
to
say,
okay,
this
is
this
is
the
message
we
want
to
get
in.