►
From YouTube: Node.js Tooling Group Meeting
Description
A
All
right
so
here
we
are
node.js,
tooling
group
meeting
this
Friday
March
6th
2020,
so
hi
everybody
and
there
was
a
late
add
to
the
agenda,
which
is
FS
hooks,
which
I
haven't
had
a
chance
to
look
at,
but
it's
the
first
thing
in
list
and
Brian
added
it
so
I'm
going
to
ask
Brian
to
talk
about
it.
Please
Brian.
B
B
B
Fine
I'm
just
teasin
so
yeah.
Basically
in
our
last
name,
we
discussed
the
the
notion
of
well
just
what
one
of
the
things
wrought
up
was
the
idea
of
FS
hooks
the
idea
of
being
able
to
not
necessarily
swap
out
the
entire
FS
implementation,
but
at
least
be
able
to
make
decisions
based
on
sort
of
user
input,
so
sort
of
what
I
mean
by
that
is
a
shim
layer.
B
Four
FS,
that's
built
in
as
opposed
to
the
sort
of
manual
shims
that
we
see
in
existing
tools,
basically
just
to
enable
those
tools
rather
than
work
against
them.
So
that's
that's
kind
of
long
and
short
of
it.
So
let
me
go
ahead
and
open
up
the
issue
here,
see
what
I
actually
said
in
it
right.
So,
like
sort
of
the
more
obvious
examples
we
have
today
of
things
that
might
find
this
sort
of
thing
useful
would
be
things
like
yarn
to
or
urine
PNP
or
tank.
So
basically
anything
that's.
B
Presenting
tarballs
or
various
other
structures,
as
if
there
are
real
files
for
both
require
and
for
regular
FS
access
on
the
flight,
transpiling
seems
like
a
pretty
appropriate
use
case.
Policy
enforcement
tools
might
take
advantage
of
this
sort
of
thing.
Apm's,
of
course,
I'd
be
remiss
if
I
didn't.
You
know
talk
about
the
fact
that
this
would
be
advantageous
for
my
employer,
and
then
you
know,
of
course,
things
like
ipfs.
That
are
this.
B
This
sort
of
notion
is
actually
pretty
pretty
well
geared
towards
as
well,
so
that
those
are
kind
of
the
use
cases
that
I
think
would
motivate
this
sort
of
work
and
then
I
guess
all
the
only
other
things
I
said
in
the
issue
is
that
you
know
today.
People
are
just
basically
shimming
FS
directly
and
manually,
which
is
prone
to
all
kinds
of
mistakes,
and
you
know
code,
that's
not
the
best
to
read
so
yeah.
That's
that's
sort
of
the
motivating
things
behind
this.
A
C
C
C
Haven't
really
had
a
chance
to
fully
investigate
but
I
believe
that's
Wes
made
a
proposal
for
import
map
support
like
native
to
know
jazz,
so
I
do
think
it
would
be
nice
if
nodejs
have
the
ability
to
basically
look
at
a
data
file.
That
would
say
you
know
this
module
isn't
in
node
modules
it's
over
here
in
this
tar
ball,
but
that's
kind
of
a
just
sidebar
and.
D
C
Yep
and
actually
just
to
say
about
that
I'm
not
maybe
wet
what
wes
is
proposing,
could
be
implemented
using
this
on
the
backend
I
guess.
My
point
is
that
I'd
like
nodejs,
to
support,
for
example,
reading
a
JSON
file
out
of
node
modules
and
using
that
to
look
up
kind
of
a
shared
cache
of
modules
rather
than
you
know,
having
a
hundred
copies
of
es
lint.
You
know
installed
to
my
NPM
clones.
F
It's
kind
of
it's
kind
of
interesting
to
means
to
so
to
your
point:
Brian
I
guess,
like
we've,
been
talking
about
for
like
months
well
over
a
year
now
about
how
we
would
facilitate
cooking
require
because,
yes,
M
modules,
you
know
if
you
don't
have
the
same
hooks
that
you
have
for
common
tasks
for
requiring
so
I
mean
with
this
idea.
We're
basically
saying
that.
Are
we
basically
saying
that
these
hooks
would
be
fired?
F
F
E
Well,
I
think
there's
some
subtle
things
that
this
enables
beyond
what
the
you
know
the
require
hook.
Stuff
does
so,
for
example,
when
you're
doing
something
like
the
pkg,
you
know
bundling
if
you
access
something
on
the
filesystem
relative
to
like
dur,
you
know
dur
name
or
whatever.
Now
that's
obviously
no
longer
valid
right,
so
this
would
enable
you
to
do
a
bunch
of
stuff
for
that
asset.
E
Loading
right
like
if
you
pull
in
a
data
file,
that's
not
through
require,
and
then
you,
you
know,
no
processing
that
these
type
of
hooks
would
enable
you
to
do
those
type
of
location
rewrites
as
well.
So
it
is
quite
a
bit
more
powerful
there.
So
it's
not
I,
see
where
you're
going
with
it
being
kind
of.
E
A
C
C
Just
wanted
to
comment
about
the
ESM
loader
hooks
at
this
point.
There
is
an
experimental
loaders
flag
that
you
can
point
to
a
ESM
file
that
supports
a
transform
source
hook
and
I
have
actually
used
that
in
one
project
to
essentially
essentially
I
use
babble
plug
in
Istanbul
to
get
coverage
of
esm
files.
You
know
loaded
with
native
nodejs
under
NYC,
so
that
that
is
supported
right
now
or
experimentally,
supported.
A
Go
so
I
had
a
I
mean
I
I,
don't
okay,
so
I
want
to
do
the
the
devil's
advocate
thing.
So
these
four
bullet
points
here,
so
the
the
first
one
and
any
issues
tool
why
this
would
be
some
use
cases
for
this
tools
to
build
up
a
fake
directory
structure
and
access,
packed
files
or
files
that
are
located
elsewhere.
A
G
C
Okay
from
what
I
understand,
the
way
it
works
is
that,
instead
of
unpacking
everything
into
node
modules,
there's
basically
a
script
that
I'm
not
sure
where
it
looks,
but
one
way
or
another
hooks
the
requires
to
wear
instead
of
loading
from
node
modules,
it
loads
from
say,
a
shared
cache
of
tarballs.
So
in
again,
in
the
case
of
let's
say
you
have
a
hundred
repositories,
cloned
I
believe.
E
So,
to
be
clear,
though,
these
are
two
cases
that
I
think
are
being
conflated,
that
don't
need
to
be
right,
you
can,
as
I
was
trying
to
show
in
the
import
Maps
do
this
without
filesystem
level
stuff
right.
It
doesn't
give
you
one
of
the
qualities
that
I
think
Darcy
was
referring
to,
which
is
file
based
stuff.
The
import
map
is
only
module
or
package
based
right,
so
it
would
still
download
and
install
the
entire
package,
even
if
you
weren't
using
half
the
files,
whereas
the
filesystem
shim,
which
is
what
I
understand
from
eyes.
E
One
conversation
with
Isaac's.
This
is
how
you
intend
to
do
tank
gives
you
specific
file
ability
to
do
that
on
specific
files,
as
opposed
to
a
package
entirely.
That
said,
though,
I
think
you
could
probably
do
the
same
thing,
maybe
not
with
import
Maps,
but
with
something
like
that,
without
necessarily
still
having
to
shim
the
filesystem
right.
E
The
the
example
that
I
did
with
the
import
Maps
it
just
overrides
a
resolved
module
or
whatever
it
is
my
result.
File
name
I
can
enter
the
name
of
the
method
anyway,
and
it
does
unfortunately
change
the
name
and
all
that
other
stuff.
That's
just
a
proof-of-concept,
though
I'm
sure
we
could,
you
know,
find
ways
to
work
around
that
with
commonjs.
E
C
A
Okay,
well,
I,
don't
wanna
yeah
I'm,
not
gonna,
go
off
on
a
tangent
about
the
relative
merits
of
this
sort
of
thing,
but
okay,
so
on-the-fly
transformation.
So
it
seems
to
me
that
there
used
to
be
a
thing
for
this,
which
is
now
deprecated,
which
is
like
require
extensions,
or
something
like
that
where
you
could
require
a
coffee
file
and
it
would
load
the
CoffeeScript
thing
and
CoffeeScript
running,
and
so
it
would
just
kind
of
do
it
behind
the
scenes.
A
B
Go
ahead!
Okay,
so
when
I
wrote
that
I
actually
had
a
slightly
different
use
case
in
mind
and
that
is
serving
up
JavaScript
assets,
you
know
with
you
know,
Express
static
or
something
like
that
and
then
having
those
files
being
transpired
before
they're
delivered
to
the
browser
like
transparently
right.
A
Mean
the
you
yeah
I
mean
it
seems
like
the
at
the
at
the
end
of
what
you
get
is
some
sort
of
representation
in
JavaScript
all
right,
some
sort
of
object,
or
what
have
you?
Maybe
it's
a
byte
array,
I,
don't
know,
but
it
seems
like
kind
of
the
same
thing
but
I,
don't
know
why
that
was
ever
like.
I
know
it's
deprecated
I
know
that,
but
why
was
it
deprecated
and
like
what?
What
are
you
supposed
to
do?
Instead?
Does
anybody
know.
F
Was
actually
gonna
volunteer
you
and
say
that
that's
what
Corey
was
talking
about
with
the
loader
hook,
so
it
was
just
people
wanted.
There
was
basically
contention
that
required
on
extensions
was
too
low-level
and
a
lot
of
people
to
create
thing.
As
an
example,
it
was
difficult
to
compose
multiple
require
hooks
together
and
it
was.
It
was
hard
to
know
in
the
node.js
runtime
one
had
actually
been
done
to
the
files
before
they
got
brought
in.
So
there
was
just
a
lot
of
thought
around
that
functionality
and
people
were
excited
to
remove
it.
F
With
the
perfect
new
world
of
esm
modules,
which
was
a
you
know,
we
could
start
from
scratch
and
everything
would
be
perfect.
There's
then,
a
lot
of
contention
around
how
you
still
would
support
tools
like
like
NYC,
because,
like
we
had
NYC
a
battle
over
Babel
over
fire,
you
know
typescript
require
yeah
well,
I
have
mocha
require
I'll,
give
you
all
these
things
that
they
were
relying
on
this
hook
and
we're
pretty
popular
things
in
the
community,
and
there
wasn't
really
a
an
API.
F
And
now
it
sounds
like
there
is
an
API
behind
an
experimental
flag,
which
is
what
Corey's
talking
is
a
my
concern
would
be
with
with
like
doing
the
lower-level
file
access
thing.
I
think
you
would
run
into
a
lot
of
pushback
from
the
same
folks
who
are
really
excited
to
deprecate,
require,
and-
and
you
know
what
I
do
understand
both
sides
of
the
argument
like
it
like.
It
was
really
hard
to
know
what
the
heck
people
were
doing
was
that
require
statement.
C
So,
just
to
correct
one
thing,
you
said
the
experimental
loaders
from
what
I
can
tell
that
will
not
help
us
move
toward
away
from
the
require
dot
extensions.
The
experimental
loaders
is
for
ESM.
Only
so
essentially
I
mean
we
still
have
to
support
CGS.
So
we're
going
to
continue
to
be
using
the
required
extensions
hook
and,
at
this
point,
I've
gotten
the
impression
that
it's
flagged
as
deprecated,
but
there
is
no
alternative,
so
I
don't
really
see
a
way
that
they
could
get
away
with
removing
it.
At
this
point,.
A
E
F
I
was
just
gonna
say.
One
really
interesting
use
case
to
me
is
it's
like.
If
you
look
at
how
Deano
approaches
things,
just
the
idea
that
you
could
be
requiring
like
I
guess:
Chris
didn't
like
that
yeah,
you
need
the
ideas
that
you
could
be
requiring
from
your
at
URI
directly,
like
you
might
be,
reaching
out
to
the
way
up
and
pulling
something
in
I.
B
A
B
A
B
So
I
mean
right
now
the
way
L,
the
way
most
node
APM's
work
is
they're,
going
to
basically
shim
everything.
They
want
to
trace
right
everything
they
want
to
measure
they're
going
to
ship
having
predefined
hooks,
makes
that
less
necessary
and
so,
for
example,
a
data
dog
in
other
languages
that
do
provide
hooks
for
various
things.
We
get
to
actually
the
legend
of
those
hooks
and
it
becomes
a
lot
easier
to
do
that
job.
A
So
I
mean
yeah.
If
we're
gonna
go
this
route,
then
I
think
we
need
to
be.
We
need
to
consider
what
all
what
are
all
the
alternatives
for
these
things.
You
know,
because
people
are
people
are
gonna,
be
like
no.
We
we
don't
need
to
do
that.
You
can
do
this
instead
right
because
that's
the
way
people
are
so
we
need
to
be
sure
that
you
know
this
is
really
going
to
be
something
that
we
can.
We
can
champion.
A
Also
Jordan
commented
below
that
there
should
be
a
way
to
disable
this
right.
So
if
you're
running
code,
where
you
don't
want
some
other
tool
to
add
some
weird
hook,
you
should
be
able
to
say
so
and
I
don't
know
if
that
that
starts
to
go
into
the
policy
area.
Where
there's
all
these
concerns
and
and
there's
like
you
can
only
be
so,
you
can
only
lock
it
down
so
much
without
other
changes.
You
know
what
I
mean
and
so
I
don't
know,
but
I
mean.
B
E
G
I
mean
that's
no
different
than
any
other
kind
of
like
plug-in,
transform
like
api's,
that
you
see
in
the
wild
for
other,
like
ecosystems
like
I'm,
not
very
slight
little
loose
yeah,
but
you
could
potentially
bucket
them.
So
we
could
decide
like
if
there's
an
access,
specific
transform.
First,
resolving
like
resolver,
transforms
or
something
like
that
or
workers
or
whatever.
You
would
like.
One
of
that
to
look
like
and
then
how
you
load
them
or
specify
the
ones
that
you're
loading
I,
don't
know
what
that
API
would
look
like.
But
my.
A
And
so
you
started
to
getting
this
this
this
this
place,
where
you're,
adding
all
these
hooks
for
various
things,
and
you
don't
really
know
what
the
hell's
going
on,
and
it's
really
hard
to
reason
about
like
what
your
program
is
doing.
You
know
I
think
that
yes,
I
could
see
the
potential
for
something
like
this
to
be
abused.
A
F
My
point
of
view
right
now
about
this
idea.
It
mainly
is
just
like
I'd
love
to
see
some
things.
Land
like
I'd
love
us
to
actually
have
a
good
story
about
how
we're
transpiling,
ESM
and
I'd
love
us
to
at
least
get
this
a
terminal
point
on
like
import
maps
or
export
maps
or
whatever
how
many
maps
there.
It
feels
like
there's
a
lot
of
car
parts
all
over
the
garage
right
now,
I'm
gonna
be
good
to
clean
them
up
a
little
bit.
That's
just
my
personal
opinion.
Right
now,.
G
G
A
G
E
One
one
thing
I'd
also
like
to
add
is
the
future
of
some
theoretical
API.
Like
tank.
We
do
have
a
I
think,
there's
a
lot
to
work
out
there
and
whether
that
will
be
able
to
land
as
people's
normal
JavaScript
workflow
I
have
my
doubts,
but
I
think
this
particular
proposal
serves
much
more
than
just
that
use
case.
E
E
C
So
could
I
just
comment
on
number
three:
the
policy
enforcement
I
am
a
little
concerned
depending
on
how
that
was
documented.
I'd
be
concerned
that
we
want
to
make
sure
that
we're
clear
that
it's
not
really
a
security
feature
since
it
can
be
circumvented
theoretically,
pretty
I
assume
it
could
be
circumvented
easily
and
maybe
even
recommend
that
you
know.
If
you
need
real
security,
then
you
need
to
look
at.
You
know
to
the
operating
system,
whether
it's
SELinux
or
I,
don't
know
other
platforms,
so
whatever
the
other
platforms
would
provide
for
process
level.
B
C
B
C
Okay,
thank
you.
A
A
A
F
I
do
think
it
dovetails
off
the
earlier
conversation
know
that
we
need
to
a
weekly
I,
don't
know
if
we
surface
to
the
maybe
talk
this
one
on
the
TSC
and
be
like
listen
like
in
the
tooling
meeting.
We've
noticed,
there's
a
lot
of
things
that
overlap
in
terms
of
how
we
handle
file
system
access,
like
just
white
circles,
saying
like
it
would
be
good
to
lender
voice.
Maybe.
F
C
One
is
kind
of
on
me,
but
I
haven't
had
any
time
to
work
on
it.
To
be
honest
at
this
stage,
I
I
think
we
could
remove
this
from
the
agenda
for
now,
because
it
might
be
a
few.
It's
probably
going
to
be
a
few
weeks
before
I
have
a
chance
to
work
on
that,
and
actually
the
yeah
I'm
not
sure
the
same
may
apply
to
the
next
item
as
well,
better
way
to
hook
process
exiting,
although
for
that,
one
I'm,
not
sure
that
I'll
necessarily
be
able
to
work
on
that
at
all.
F
I
have
any
cycles
I
am
interested
like
I
would
love
for
the
stuff
that
NYC
does
and
the
stuff
that
c8
does
to
just
be
something
any
user.
Land
tool
can
easily
write
and
that's
a
combination
of
detecting
exiting,
which
is
important
for
writing
reports
at
the
end
of
your
program
and
also
being
able
to
pass
the
same
environment
variable
to
each
sub
process,
which
I
think
Cori
figured
out
some
better
ways.
Well,
there's
no
way
to
make
them
sticky
so
I'd
like
that
to
be
a
feature
too
so.
F
I,
don't
know
a
guy
and
they
start
leaving
this
stuff
on
Jordan
LJ
harb
who's
in
the
node
tooling
slack
was
just
asking
this
week
of
how
to
knock
knock
something
in
a
subprocess
and
I
was
like
man.
If
we
had
all
this
stuff
that
we
used
for
NYC
and
very
janky
ways
just
available
to
everyone
in
user
lands,
then
you
could
do
stuff
like
that.
You
could
be
like
I
wanna
in
every
sub
process.
Mach
the
HTTP
connection,
super
useful,
I.
Think
I.
C
A
Okay,
so
looks
like
a
source
map
support
so
then
put
some
stuff
in
here,
and
one
of
those
things
is
that
a
blog
post
was
written.
All
I
see
is
pardon
the
interruption.
Okay,
so
yeah.
So
there's
a
nice
article.
I
haven't
actually
read
your
article
sorry
ban.
Well,
it's.
F
Funny
I'd
actually
noticed
like
it
got
really
good
pickup
like
it
was
published
by
the
node
blog
and
it
got
tweeted
by
got
J's
weekly
and
everything.
But
I
noticed,
like
my
readership,
compared
to
the
one
we
wrote
on
coverage
was
way
down
so
I'm
kind
of
tending
to
agree
that
maybe
it's
time
to
move
away
from
medium
as
a
blogging
platform
like
that,
what
was
the
one
you
suggested:
dev
dot,
t--.
F
Medium
so
I
feel
bad,
but
I
can
understand
the
pay.
Walls
may
be
driving
people
away
or
something
so
yeah,
but
give
it
a
read.
If
you're
curious,
it
was
mostly
just
to
get
it
on
people's
radar,
so
they
actually
know
that
feature
exists.
But
the
the
other
agenda
item
was
I've
been
talking
to,
is
it
Simon
and
it
works
unjust,
I
think.
F
Hope,
I'm
not
camp
blogging
here,
yes,
so
I've
been
walking
to
what
they're
talking
to
one
of
the
maintainer
suggests
and,
and
they
ingest
do
a
bunch
of
source
map
juggling.
They
were.
They
like
are
inserting
source
maps
on
the
fly
and
doing
a
bunch
of
work
with
them,
and
he
was
asking
whether
we
could
extend
the
API
to
let
you
just
kind
of
insert
a
source
map
right
now.
F
The
API
allows
you
to
get
out
a
source
map,
so
you
can
see
what
source
maps
have
been
used,
but
we
don't
actually
let
you
insert
a
source
map
currently,
so
I
was
thinking
if
it
could
get
just
more
excited
about
using
some
of
these
features,
I
was
thinking
of.
Maybe
we
should
add.
The
ability
to
add
a
source
map
as
well
would
become
a
neat
and
kind
of
the
cool
thing.
There
is,
if
you
like,
maybe
you're
generating
code
without
an
import
or
acquire
statement.
B
F
Your
you
know,
I,
don't
know,
I,
don't
exactly
know
what
you're
doing,
but
you
could
still
potentially
put
some
things
in
the
source
map
cash
and
still
have
a
stack
trace.
Printout
a
nice
stack
trace,
which
I
thought
was
kind
of
cool,
no
I'm
trying
to
figure
out
more
use
cases
for
the
source
map
stuff
so
that
people
are
more
excited
about
it.
I
tried
to
estimate
it
and
patch
them
open
how
to
turn
down.
F
A
F
We've
just
been
starting
to
like
in
our
mocha
config
we've
started,
putting
in
just
enabled
source
map
support
as
one
of
the
mocha
flags
and
then
because
you're
smart
and
pass
it
along
to
no
DJ
s.
You
can't
you
just
update
your
mocha
config
to
have
source
Maps,
and
then
it
starts
working
right
down
in
your
applications,
because
you
look
a
little
bit
better.
F
F
F
No,
okay,
cool,
so
is
it
but
a
long
story
short?
We
were
actually
in
agreement
on
the
three
of
us.
Obviously
we're
I'm
sure
everyone
will
have
different
opinions
as
we
start
to
share
more
widely,
but
but
we
were
pretty
pretty
much
an
agreement
and
a
lot
of
the
minimal
features
like
we
were
like
okay,
we
managed
to
basically
redline
tons
of
things
that
are
in
most
argument,
parsers
and
still
felt
we
had
something
relatively
useful.
F
There
were
a
couple
interesting
points
of
contention.
One
of
them
was
like
single
flags
like
if
you
go
ABC.
Some
argument
parsers
take
that
as
a
true
B,
true
C,
true-
and
we
were
talking
about
whether
it
like
we
said
that
was
like
the
one
that
was
kind
of
the
one
feature
we
did
not
agree
on
the
parsing
up
and
our
like
our
conversation,
I,
don't
know
it's
kind
of
happy.
It
got
me
excited
guys.
F
I
think
that
there
was
a
minimal
feature
set
from
my
point
of
view
as
the
maintainer
of
yards,
that
I
could
pretty
much
throw
yards
parser
and
build
a
little
bit
on
top
of
this
minimal
feature
set
and
be
able
to
just
use
the
thing.
That's
kind
of
the
agreed-upon
baseline,
so
I
felt
like
we
had
something
close
to
what
I
could
build
on
and
again
like
with
a
fraction
of
the
features
that
are
in
New,
York's,
parser
or
minimis
or
Commander,
or
any
of
that
stuff.
F
I
thought
that
was
neat
and
I
also
felt
like
what
was
cool
is
that
it
was
enough
of
a
foundation
that
you
could
then
like
have
much
nicer
code.
Samples,
which
is
a
big
thing
we
do
it
at
my
job,
is
writing
code
samples
for
a
lot
of
different
languages,
and
it
would
mean
that
the
code
sample
just
read
really
elegantly
right,
like
it
would
still
look
a
lot
like
a
nice
full-featured
argument.
Parser
who's,
like
you,
get
a
nice
object
bag
at
the
other
end,
so
we
weren't
having
a
ton
of
contention
I
think.
F
Maybe
once
we
went
to
a
wider
audience,
so
you
might
get
more
contention,
I,
don't
know,
I!
Think
it's
a
neat
idea,
though,
because
I
think
what
you
did
in
that
new.
It's
like
someone
could
build
something
like
commander
airy
args
on
top
of
it
as
a
baseline
and
then
give
you
like
the
pre
command
line
output,
giving
you
all
that
stuff
I
think
if
we
had
the
rate
minimal,
I,
don't
know,
I!
Think
it's
a
good
idea.
A
A
A
To
be
clear,
we're
trying
to
focus
on
the
use
case
of
the
the
the
author
of
code
samples,
not
necessarily
the
the
library
author
who's
building,
something
like
yarks
and
who
wants
to
build
on
top
of
it
because
those
you
know
it
serves
two
different,
two
different
sets
of
people
and
we've
seen
that
people
can
already
build
yards.
So
what
I
think
we're
lacking
is
serving
the
people
who
just
want
to
parse
like
a
couple
command
line
flags
and
and
not
have
to
pull
down.
F
My
point
of
view
like
I
can
be
like
I
can
definitely
be
convinced
to
go
in
that
direction.
If
we
have
another
meeting
and
like
that's
really
where
we
slice
things
down
to
I
was
getting
kind
of
excited.
That
I
was
like
oh
shoot.
This
actually
couldn't
work
as
a
foundation
as
yards,
because
it
turns
out.
We
were
agreeing
a
lot
on
a
lot
of
the
minimal
features
which
was
cool,
so
yeah
I
feel
like
another
hour
session
and
then
maybe
share
it
with
the
whole
group
would
be
good
yeah.
F
A
D
D
F
Is
in
the
details,
it's
like
okay
cool.
What
can
let
people
do
that
in
user
land,
but
now
we
need
to
distinguish
between.
Has
someone
provided
a
double
dash
or
a
single
dash,
and
do
we
track
the
double
dash
or
do
we
track
the
single
dash,
and
is
it
like
the
you
know,
I
mean
it
just
starts
that
you
start
to
realize
why
your
ex
has
3000
unit
tests.
A
A
What
if
I
give
it
three
dash
V's
like
how
do
we
handle
these
things,
and
so
we
just
we
need
to
figure
out
what
what's
reasonable
for
for
most
people
and
and
I
mean
you
know-
maybe
maybe
one
of
the
things
we
need
to
do
is
sit
down
with
somebody
who
who
hasn't
written
a
command-line
tool
in
node
and
be
like
all
right,
so
you
need
to
parse
these
options.
What
do
you
expect
to
happen
here?
You
know
I
mean
so
like
is
it?
Does
this
make
sense
just
kind
of
you,
whatever
you
call
it.
F
I
could
be
like
just
even
considering
that
either
like
that's
a
great
idea,
I
think
because
it
would
really
help
the
case,
therefore,
eventually
can
open
up
a
PR,
a
node
and
create
a
giant
argument
with
the
community
about
whether
we
should
do
this
or
not.
It
might
make
sense
to
already
have
some
of
that.
Like
listen,
we've
talked
to
some
early
node
users
and
we
have
this
feedback
and
they
found
this
confusing
and
here's
what
we've
come
up
with
just
rather
than
it
all
being
opinion.
C
F
F
C
Like
I
wonder
if
we
could
get
a
proof
of
concept
in
userspace
like
I
I'm,
not
really
sure
how
anything
in
the
argument
parser
would
not
work
would
only
work
if
it's
in
the
node
core.
So
if
there
was
some
kind
of
module
where
I
could,
basically,
you
know,
require
and
play
with
then
kind
of
see
how
it
works.
Then
it
might
be
easier
to
critique.
F
F
E
A
So
what
I'm
working
on
right
now
I
have
nothing
to
show
because
it's
kind
of
broken
but
I'm
working
on
getting
mocha
to
run
tests
in
parallel.
So
that's
interesting
it
once
upon
a
time
long
before
I
came
on
with
the
project,
it
did
have
this
functionality,
but
it
was
torn
out
for
reasons
and
I.
Don't
know
what
those
reasons
are.
A
A
E
A
Is
anybody
looking
after
that?
One
is
TJ
not
actually
looking
after
that
one
I
don't
know
anyhow,
so
what
I'm
doing
with
the
the
parallel
runs
is
right
now,
I'm
experimenting
with
workers
and
so
basically
taking
a
file
or
basically
so
I,
create
a
worker
pool
and
then
like
get
all
the
gather,
all
the
files
up
and
then
hand
out
files
to
these
workers
and
they
run
it,
and
then
they
like
send
the
results
back
to
the
main
process
and
then
I
can
aggregate
them
together.
A
And
so
you
know
my
initial
experiments
with
this
is
like
mocha
zone
unit
tests.
You
know
they
in
cereal.
They
take
took
three
seconds
to
run
and
with
eight
cores
on
my
on
my
Mac,
it's
one
and
a
half
seconds
so
yeah.
But
you
know
workers
do
have
some
limitations.
You
can't
do
certain
things
and
I
don't
know
it
may
be
that
you
know
the
benefits
in
terms
of
speed
are
going
to
not
be
worth
it
and
maybe
I
just
want
to
do
child
processes.
But
you
know.
F
I
remember:
Isaac
was
going
through
this
refactor
for
tap
about
a
year
ago
and
tap
was
it
was
again
like
trying
to
parallelize
things
which
you
you
might
talk
to
Isaac
my
point,
I
guess,
because
I
think
he
was
even
experimenting
with
workers
versus
sub
processes,
because
sub
process
is
made
tap
really
slow.
So
he
was
trying
to
figure
out
a
way
to
make
that
just
an
optional
thing.
Well,.
A
D
A
Why
does
it
do
that
because
they
never
think
it
can
run
in
parallel?
Hey
well,
if
your
tests
clean
up
after
themselves,
why
couldn't
die?
That's
not
necessarily
a
given.
Well,
that's
the
thing.
I
use
this
one
look
as
one
process.
Well,
it
has
two
processes,
but
there's
one
process
where
it
runs
all
the
tests,
so
you
screw
those
up.
Well,
you
screw
them
all
up,
because
there's
only
one
process.
So
that
means
that
suggest
me
well.
If
you
keep
your
tests
clean,
then
you
should
be
able
to
to
reuse.