►
From YouTube: JupyterLab Dev Meeting, August 26, 2016
Description
Meeting of the JupyterLab development team on August 26, 2016.
Meeting Notes: https://jupyter.hackpad.com/JupyterLab-Weekly-Meetings-UUJ3gIQ3iBS
A
B
You
sure
so
my
name
is
then:
we
are
from
company
called
type
fox
and
we've
been
starting
using
you
for
the
lab
just
recently
for
some
product
we
are
working
on.
That
is
like
more
like
a
cloud
product
for
our
developers
and
so
we'd
like
to
contribute
the
kind
of
basic
stuff
we
are
working
on
back,
like
biggest
thing
at
the
moment,
would
be
the
abstraction
for
the
editor
that
Antonia
has
been
working
on
and
and
also
the
maybe,
the
monocle
integration
like
using
monarch.
Win7
could
narrow.
A
A
Cool
excited
this
realize
everyone's
got
their
name
already.
Let's
see
everyone
else,
who's
been
here
and
is
that
meeting
before,
let's
move
on
to
discussion,
items
and
status,
update,
I,
think
for
me
the
biggest
discussion
item
and
status
update
would
be
third-party
extensions.
Are
there
any
other
topics
that
we
need
to
discuss
to
make
sure
we're
giving
ample
time
for
what
we
need
to
discuss?
Do
you
want
us
to
discuss
your
editor
work,
Anton
and
then
yeah.
A
E
F
All
right
so
I've
been
doing
some
research
on
how
other
editors-
and
this
is
sort
of
in
this
space
handle
third-party
extensions.
I
couldn't
I,
don't
know
of
one
that
is
fully
in
the
browser
and
allows
extensions,
I'm
still
researching
in
that
area,
but
vs
code
and
Adam
are
both
javascript-based
editors
at
lau
plugins.
F
They
are
both
desktop
applications,
so
they're
using
node
on
on
the
desktop,
as
opposed
to
being
in
the
browser
and
having
to
deal
with
the
browser
environment,
but
we
can
learn
from
some
of
their
packaging
methods
what
they
what
they
do,
is
they
provide
a
core
that
is
available
from
the
environment,
so
either
you
can
require
Adam
or
require
vs
code,
depending
on
your
which
one
you're
using
and
then
each
plug-in
is
a
fully
standalone,
a
chunk
of
code.
That's
self-contained!
F
You
can't
share
code
between
plugins,
so
they
they're
fully
self-contained
that
doesn't
lend
itself
well
to
a
browser
environment.
In
that
you,
don't
you
don't
want
to
have
to
fetch
all
of
these
files
over
and
over
again
over
a
remote
connection,
and
also
we
are,
we
don't
have
the
required
j/s,
loader
I'm,
sorry,
the
the
node
loader
require
loader.
We
are
forced
to
use
browser
loading
capabilities
which
right
now
is
as
either
script
tag
will
require
j/s
or
something
like
system
loader.
F
So
where
is
trying
to
rectify,
making
it
easy,
for
extension,
authors
to
produce
their
installable
things,
making
it
easy
for
vendors
such
as
anaconda,
or
are
going
to
to
provide
these
things
as
chunks
that
get
combined
and
balancing
the
shipping
of
code
versus
assembling
it
or
building
it
on
the
user's
computer,
preferably
not
having
multiple
build
chains
set
up
if
we
can
just
use
what
they
with
their
guarantee
to
have
with
Jupiter
in
that
a
working
Python
environment,
then
that
would
be
ideal.
But
there's
there's
a
lot
of
things
to
weigh
in
all
this.
A
A
Jupiter
have
already
has
an
NPM
dependency,
so
there's
NPM
already
going
to
be
on
the
server
in
that
case,
and
the
users
are
the
people
that
are
installing
the
extensions
we'll
have
hopefully
better
higher
technical
sophistication.
So
they
should
be
able
to
manage
a
an
NPM
based,
build
system
for
bundling
and
or
D
duping
the
any
third-party
extensions
that
are
installed.
So
to
me,
it
sounds
like
we
can
tackle
the
two
use
cases
if
they
have
different
requirements
and
different
different
capabilities.
A
H
A
D
A
G
I
think
that
if
you,
if
we
think
that
a
basic
data
scientist
with
that
people's
for
Wi-Fi
widgets
and
that
people
saw
trigger
some
NPM
beyond,
and
you
could
believe
that
that's
a
like
workable
scenario-
we
are
just
shooting
ourselves
in
the
foot.
We
really
need
to
have
a
form
of
extension
that
might
be
not
as
powerful
as
when
you
rebuild
the
entire
environment
within
p.m.
packages
that
doesn't
require
NPM.
Like
doing
a
bureau
on
the
user's
machine
is
like
bringing
all
the
complexity
to
the
user.
G
It's
going
to
be
a
huge
mess
and,
and
people
were
already
having
a
lot
of
difficulty
with
our
node
notebook
extension
and
server
extension
mechanism.
That
was
much
simpler
than
having
to
build
a
JavaScript
NPM
package
and
doing
a
bundle,
and
therefore
I
had
I
made
a
proposal
for
mechanism
of
extension
that
could
be
less
powerful
than
the
one
where
you
can
be
doop
everything
and
have
like
a
real
and
p.m.
extension
on
the
server
side
that
you
barely
do
to
the
lab,
like
the
ones
that
we've
done
last
time.
G
I
A
J
J
A
Runtime
dependencies
relationship,
runtime
dependencies
and
steve
had
a
good
idea
of
go
ahead
and
d
duping
at
package
build
time
stuff
that
we
knew
stuff
that
we
know
is
already
going
to
be
upstream.
So
you
know
your
upstream
packages
and
you
can
get
from
them.
What
dependencies
they're
already
pulling
in
for
runtime
dependencies,
and
you
can
d
do
partially
at
least
based
on
your
upstream
dependencies,
that
okay,
the
upstream
packages
so.
G
I
I
don't
agree
with
that.
Other
there's,
one
of
the
things
we're
talking
about
earlier
is
when
you
start
the
server,
they
could
compute
a
hash
of
the
things
that
you
need
to
run,
that
particular
launch
of
the
server
and
look
up
to
see
if
it's
built
the
bundle
for
that
hash
already
and
if
it
hasn't
go
ahead
and
one
packet
or
bundle
it
up
some
whatever
we
use
to
decide
how
we're
going
to
buckle
this
up.
So
I
disagree
that
you
need
to
be.
G
C
E
A
A
I
I
H
And
the
other,
the
other
aspect
that
that
we
can't
ignore
is
that
we're
using
all
these
web
pack
plugins
that
allow
us
to
load
other
types
of
assets,
images,
CSS
and
like
and
resolve
those
properly
with
the
imports
and
so
I
like.
So
there
has
to
be
something
in
that
place.
That
does
what
right
hack,
yep.
G
G
G
H
G
G
When
a
package
phone
clicking
attention,
then
the
contest
solver
find
the
version
of
weakness,
leading
circle
that
matches
both
requirements
and
then
the
J
is
wonderful.
We
just
index
and
we
must
have
a
compatible
version
so
we're
just
relying
on
the
counter
packaging
system
for
to
ensure
that
the
bundles
are
going
to
have
compatible
vs.
So
now
all
we
need
to
do
is
have
a
version
numbering
for
the
count.
A
package
that
any
time
the
JavaScript
keeps
a
every
time
issues
a
new
version
of
the
JavaScript
that
has
a
major
version,
number
change.
G
I
I
Really
quick,
there's
nothing
in
web
pack
that
allows
you
to
depend
on
external
dependencies
that
are
loaded
by
another
web
back
bottle
that
doesn't
exist,
so
we
have
to
hack
webpack
to
do
that
ourselves
or
write
some
shim
or
otherwise
customized
web
pack
into
something
that's
not
going
to
be
supported
by
the
main
line
packing.
But
the
second
thing
that
that
bothers
you
talking
I
get
back
to
this
okay.
I
Yes,
we
can
change
all
of
that,
transparently
on
the
server
side,
and
none
of
the
plugins
have
to
change
your
code.
They're
all
still
compatible
and
backwards
compatible.
If
we
choose
a
format
that
plugin
authors
have
to
pre-compiled
our
code
before
they
ship
it
into
a
particular
thing,
without
stuck
with
that,
and
we
can
never
change
the.
G
Other
thing
is:
if
we
use
requirejs
we're
stuck
with
where
we
are
stuck
with
with
reg
s,
for
why
at
least
on
it,
unless
we
make
a
big
breaking
change
the
future,
but
any
anyway
I,
don't
think
that
require
jess
is
giving
away
like
in
the
next
year,
and
we
don't
have
a
much
better
way
of
loading
dynamically
things
on
the
browser.
I
know.
A
I
mean
we
can
ship,
should
we
ship
our
dependencies
I?
Think
that's
a
more
basic
question:
should
we
ship
dependencies
and
that's
something
we
can
walk
away
from?
We
can
say
your
distribution
you're.
The
plugin
distribution
should
include
the
node
modules
directory
for
runtime
dependencies
and
possibly
even
prune.
A
If
you
know
something
else
is
already
provided
by
upstream
one
of
your
dependencies,
and
so
if
you
just
ship
that,
without
transcribing
anything
you're,
just
shipping
common
jf
stuff
as
a
self-contained
package
or
as
a
self-contained
package,
assuming
that
up
dependencies
are
already
loaded,
then
we
can
take
care
of
whatever
bundling
needs
to
be
done
on
this
client
side.
If
there
is
bundling,
it
needs
to
be
done,
or
we
can
have
a
framework
like
sistas
like
steel
or
something
like
that
that
will
just
use
the
common
j
us
and
node
modules
directories.
First,.
G
That
defines
were
to
find
the
thing
that
needs
to
be
loaded
later
so
I.
Every
single
extension
defines
two
bundles.
One
is
an
extension
Jess
that
defines
all
the
symbols
that
is
going
to
be
required
by
the
other
one
and
the
other
one
is
what
actually
is
the
core
of
the
the
bundle
important
question,
so
you
can
look
at
for
export,
BQ,
platypi,
3gs
or
ipad.
If
I,
they
all
do
the
same
thing
they
have,
they
have
to
wait
back
buttocks.
G
I
A
We've
gone
down
the
path
with
Steve
of
trying
to
merge,
westpac
bundles
and
it's
pretty
complicated,
I'm
more
in
favor
of
shipping
dependencies
so
that
the
user
doesn't
need
internet
access
and
it
doesn't
need
NPM
to
pull
down
dependencies
and
server
time
shipping,
whatever
minimal,
set
or
dependent
on
time
dependencies.
That
plugin
will
need
to
be
self-contained
at
runtime.
Welcome
to.
I
A
I
A
If
you're
marrying
in
p.m.
right,
so
if
you're
marrying
in
p.m.
but
that
doesn't
so,
that
was
the
second
point.
The
other
point
was
that
it's
completely
self-contained
from
the
user
perspective.
They
don't
need
NPM
installed
as
a
runtime
dependencies
for
a
single
user
case
where
you
just
have
Python,
you
know,
that's
the
only
pendency
we've
had
so
far
it's
nice
to
have
something
at
self-contained.
It
doesn't
require
NPM
in
order
to
activate
it
an
extension.
What.
H
But
so
I
think
a
couple
things
are
important
here:
one
is
that
I
I
don't
think
we
should
design
assuming
no
network.
We
can
design
assuming
a
partitioned
network
that
would
require
someone
to
run
like
their
own
private
Conda
service
or
their
own
private
NPM
service.
The
other
is
that
we
need
the
capability,
whether
we
like
it
or
not.
We
need
the
capabilities
of
NPM
and
some
sort
of
bundler.
H
A
A
I
A
Point
here
being
that
we're
in
the
case
here
of
a
non
sophisticated
user,
single
user
localhost,
and
that's
why
I
brought
up
the
point.
I
think
we
have
to
use
cases
that
are
really
different
from
each
other
and
what
one
could
use
you
know
running
on
localhost
with
a
single
user
system.
I
think
is
different
than
a
multi-user
system
with
the
sophisticated
admin
so.
I
So
to
be
clear
when
I,
when
I
said
really
really
slow,
I
met
single
user.
Like
me
on
my
localhost
and
it's
still
painfully
slow
like
we're
talking
about
the
difference
between
running
loading,
Jupiter
lab
and
web
pack
taking
like
half
a
second
and
loading
it
with
steel,
j/s
taking
like
30
seconds
I
mean
we're
talking,
orders
of
magnitude
difference
because
I'm.
A
A
H
H
A
They
have
a
single,
they
have
a
single
bundle
that
they
can
install.
They
can
get
that
bundle
from
wherever
they
are
wherever
they
want,
and
they
you
know,
install
that
one
single
bundle,
things
work
things
might
be
slow
depending
on
how
we
implement
this.
You
know
I,
don't
want
to
write
a
bundler
and
Python,
that's
for
sure,
addy
duping,
maybe
but
not
a
fun
toy,
but
there
is
some
stuff
you
can
do
on
the
browser
size.
It
might
be
too
slow.
I,
don't
know
the
trade-offs.
Having.
K
H
But
we're
not
we're
not
go
saying
that
you
need
to
do
that.
I
mean
here's.
A
great
example
I
mean
so
the
question
and
I
think
it's
worth
saying.
So
when
we
talked
the
other
day
to
get
around
the
situation
where
users
may
not
have
an
MPN
installed,
we
could
actually
run
a
public
build
service
and
that
basically,
the
the
notebook
server
this
its
dependency
lists
it's
plugins
and
we
have
a
public
rest
endpoint.
That
can
bundle
based
on
that.
The
benefit
of
this
is
that
we
can
actually
cash
those
builds.
H
A
K
Honestly,
Ryan
you
want,
you
want
to
have
an
external
service
where
user,
post
data
and
your
return
date
I
to
them.
It's
I
mean
we
will
get
to
legal
trouble.
If
we
do
that,
we
already,
we
will
I
mean
yes,
I
mean
I,
like
you
will
stuff
to
get
user.
That
wondering
like
how
why
why
are
you
sending
some
of
the
data
which
is
on
mice
on
my
computer,
to
your
server?
K
H
K
C
K
I
C
A
K
A
A
F
H
B
So,
actually,
we
are
convinced
that
the
Monaco
editor
is
in
general,
much
better
than
code
mirror,
and
so
we
would
just
like
to
use
it
though
it
could
like,
for
the
time
being,
it
could
be
maybe
simpler
to
not
abstract
over
the
editor
with
you,
but
just
use
Monaco.
Everyone
agrees,
but
I
mean
it
might
change
in
two
years,
which
editor
widget
is
the
best
of
the
same.
So
that's
why
we
came
up
with
a
abstraction
ltd,
possibly
in
your
discussions
that
you
would
more
likely
houses,
kind
of
abstraction,
yeah.
A
D
H
I
Would
discourage
focus
have
focusing
too
much
on
worrying
about
how
notebooks
perform
with
one
particular
editor
and
lots
of
cells,
because
in
no
case
should
we
be
having
more
than
one
instantiated
editor
at
a
time
like
it
should
only
be
the
cell.
That's
currently
active
that
has
focus
should
be
a
rendered
editor
instance.
Everything
else
should
just
be
static,
HTML
those
things
in
there
yeah.
D
I
A
D
They
used
to
be
just
widget
code
because
mirror
widget
now
now
say
is
a
view.
It's
just
an
abstraction
over
content
editor
with
its
minako
or
code
mirror.
It
doesn't
know
anything
about
the
application
model,
like
cell
models
or
comment
document
context,
is
ntpc
epicenter,
who
actually
managed
a
model
and
can
change
the
view.
D
Yeah.
It's
like
MVP
is
a
pattern.
If
you
think
you
know
it.
Yes,
it's
a
nephew
tection,
so
it
simplifies
the
testing,
for
example,
of
your
logic,
because
you
can
you
don't
need
code
miura
for
testing.
You
just
mock
your
view
and
you
write
a
logic
for
presenter
which
work
with
the
model
and
what
one
example
can
be
computation
of
education.
D
So
it's
kind
of
like
a
separation
of
you,
I
logic
and
application
project
is
expected
to
set
a
separate
model.
It
is
the
biggest
change
and
yeah,
and
then
they
say
decorators
reviews
it's
in
the
presenter.
Doesn't
know
anything
about
witches,
but
sometimes
you
want
to
have
some
YouTube.
You
use
about
logic,
they
use
decorators.
So
basically
you
can
create
a
widget
which
implements
of
you
is
then
wrap
it
in
some,
a
decorator
which
pose
take.
It
decorate
a
widget
example.
It
adds
default
styles
or
you
compute,
a
caption
for
editor.
D
H
D
A
H
H
Immunity
all
right
what
I
was
saying
was
it
might
help
if
you
could
list
like
the
order
of
files
or
classes
that
we
should
review
in
in
a
complex
PR
like
this,
it's
often
difficult
to
know
even
where
to
start
with
all
the
interdependencies.
D
C
B
So
in
case
you
accept
this
PR
after
the
review
arm
I
got.
You
are
also
interested
in
to
the
general
Monaco
plugin
and
then
with
that
there
would
be
another
abstraction
we
would
need
to
talk
about.
That
is
a
code
completion
which
it
you've
developed
I
think
it
would
be
better
idea
to
just
use
the
malacca
one.
B
It's
richer
and
like
what's
nice,
more
nicer
with
in
Morocco,
though
we
would
propose
another
PR
where
we
abstract
/
The
Witcher,
then
you
could
still
kind
of
use
that
that
widget,
you
have
the
completion
widget,
but
currently
it's
kind
of
tightly
bundled,
and
so
we
would.
We
couldn't
kind
of
easily
exchange
it,
but
I'm
pretty
optimistic
that
you
will
like
the
Monaco
proposal
better.
H
One
question
about
you:
if
we
start
to
have
only
one
active
rendered,
editor
/
notebook
and
basically
only
have
static
HTML
for
all
the
other
cells
that
are
not
being
edited.
One
thing
nice
about
code
mayor
is
that
you
can
use
it
to
do
static.
Syntax
highlighting
for
those
in
active
cells
does
monocle
have
that
type
of
capability.
H
B
I
Okay,
great
great
yeah,
I
think
the
key
here
is
that
the
reason
it
becomes
expensive
with
multiple
code
mirrors
on
the
page
is
because
it
sometimes
it
does
pulling
in
the
background
and
depending
on
what
resize
events
happen.
All
those
editors
are
triggering
Dom
measurements,
which
causes
a
bunch
of
reclose
on
every
resizable
book,
and
we
like
to
avoid
that,
and
so
once
you
can
decide
that
hey.
This
is
what
I'm
rendering
is
the
document
that
needs
to
be
there,
and
you
can
disable
all
that
right
now.
D
D
B
B
But
I've
like
I
mean
the
internet
says
that
the
villa's
code,
editor
is
very
fast.
So
there
is,
there
are
some
blocks
about
it
and
it's
the
same
widget
so
I'm,
pretty
optimistic
that
it's
kind
of
ok,
not
sure
how
it
compares
to
code,
0,
I'm
sure
it
compares
like
compared
to
a
sham.
It's
much
better
but
I
mean
maybe
that's
not
so
hard
yeah.
I
This
code,
you
only
have
a
maximum
three
editors
visible
at
a
time
and
in
a
notebook
we're
talking
about
an
editor
Purcell,
so
you
can
have
definitely
more
than
three,
and
so
it's
just
it's
just
something.
To
look
at
like
a
Monaco
solves
the
problem.
We
don't
have
to
do
static,
rendering
then
then
great
it's
a
solved
problem.
Definitely
with
codemirror.
We
can
only
afford
a
blown
off
the
page
at
a
time
or.
A
D
A
H
I'll
just
call
mine,
so
I've
been
going
through
doing
more
cleaning
out
the
flaws
less
and
less
and
I'm
gonna
going
to
basically
continue
that
and
try
to
take
a
pass
at
all
of
our
different
plugins
I
also
changed
a
little
bit
of
how
we're
doing
triage
I
created.
So
I
went
through
all
the
issues
that
have
no
milestone
and
assigned
milestones.
The
only
issues
that
don't
have
milestones
are
ones
where
we're
still
waiting
for
information,
and
we
can't
decide
the
other
thing.
H
J
So
but
I
change
the
I
ever
PR
0
and
I
tag
you
in
it
the
less
compilation
mode,
less
has
a
thing
called
strict,
math
and
I've
turned
that
on,
because
otherwise
it's
breaking
any
time
we
use
calc
in
our
CSS.
It
does
this
weird
thing
where
we
say:
calc,
one
hundred
percent
minus
20
pixels.
Somehow
that
equals
eighty
percent,
which
is
absurd,
so
I
put
that
in
I.
Didn't
we
didn't
want
to
merge
it
in
without
you
seeing
it,
so
you
might
want
to
take
a
look
at
it
great.
H
G
Sign
that
I
started
diving
into
the
notebook
with
a
legacy,
notebook
repo
and
the
goal
is
to
separate
a
base
server
out
vo
Jupiter
server,
which
is
not
linked
to
the
JavaScript
of
the
legacy
code
book.
But
eventually
we
do
another
big
spread
out,
be
tributo
server
and
the
car
notebook
has
an
intentional
eventually,
but
already
just
in
doing
that
separation
of
classes
I
found
some
issues.
G
One
of
em
is
that
it
seems
that
the
order
in
which
we
compared
to
an
80
handlers
is
like
super
sensitive
and
when
you
do
inheritance,
you
have
to
you
know.
Actually
you
cannot
call
up
the
base
class
method
then
register
on
you
happy
because
you
have
to
insert
the
Henry
the
right
order
today
are
I
things
like
this.
I'm
not
quite
sure
understand
the
logic
I
might
have
to
ask
virtues
to
brian
or
my
tias
when,
when
they
are
variable,
okay,.
G
Notebook
out
on
the
patio
I
can
do
it
right
now.
If
you
want
it's,
so
it's
not
book
people
and
the
idea
is
to
separate
a
base.
Server
app
from
notebook
app
and
the
server
app
would
have.
None
of
the
different
agility
is
like
pylab
and
stuff
like
that
and
unexposed
services.
But,
for
example,
of
odd
me,
we
have
a
mathjax
property
me
that
seems
to
be
required
for
counseling
service
I.
G
F
Working
off
notebook
master
to
do
that.
It's
a
ham
because
there's
the
open
question
min
has
an
issue
about
as
to
whether
we
just
need
to
pull
webpack
out
of
that
entirely
to
be
able
to
support
the
old
workflow
of
requiring
using
an
AMD
require
an
individual
module
of
note
of
the
notebook
which
breaks
using
webpack.
F
H
Colonel,
but
also
so
then,
please,
please
make
sure
you
coordinate
that
with
Matthias
and
grant
Nestor,
because
obviously
you're
gonna
have
to
man.
It
figure
out
how
to
manage
the
fact
that
the
notebook
repose
itself
moving
forward,
and
so
there
may
be
some
coordination
that
has
to
happen
there.
Yeah.
K
H
K
Some
discussion
recently
and
a
lot
of
few
people
complain
that
they're
still
on
version
2
and
3
of
ipython,
not
Jupiter,
and
they
basically
I,
have
the
chance
to
update
every
couple
of
years
and
I
it's
starting
to
be
really
painful
for
them.
So
I
would
like
to
be
really
careful
about
changing
so
notebooks
and
having
a
smooth
path
for
a
while.
So
I
agree
that
there
is
a
lot
of
things
that
we
want
to
to
ditch.
I
think
that
having
a
may
be
true
really
is
where
we
literally
you
really
carefully.
K
C
A
To
me,
it
seems
like
one
big
question
and
difference
of
opinion
is:
is
it
too
much
to
require
all
users
to
have
NPM
on
the
on
the
server
side
and,
and
there
might
be
some
other
things
that
I'm
trying
to
get
down
to
like?
What
really
is
the
fundamental
difference
in
opinion,
and
you
know
kind
of
middle
of
things
that
were
that
we're
discussing
yeah.
I
If
you
know
so,
there's
there's
two
steps
that
need
to
happen
when
installing
a
new
plugin
there's
one
that's
installing
the
plug-in
and
its
dependencies
that
can
be
done
with
NPM
or
any
other
package
manager
that
can
install
directories
and
stuff
and
there's
the
other
act
of
actually
running
the
server,
which
requires
all
those
dependencies
to
be
bundled
up
in
something.
That's
served
for
the
client.
The
first
that
first
step
of
installing
packages
is
the
job
of
a
package
manager,
whether
that's
in
p.m.
I
whether
it's
Conda,
whether
it's
Wi-Fi,
that
can
be
done
by
whatever
means
necessary,
and
the
is
the
way
that
we
actually
serve
that
code
to
the
users,
and
so
there
are
really
two
problems
we
need
to
solve.
One
of
those
absolutely
I
think,
is
going
to
require
a
note
and
that's
you've
got
files
on
your
system
and
we
need
to
serve
those
to
the
user
and
a
deduplicated
more
than
likely,
bundled
way
and
I
think
that's
going
to
require
a
node.
The
first
first
one
might
not
require
know,
depending
on
what
package
battery
using.
A
I
I
thought
it
was
the
other
way
around.
I
didn't
want
to
run
an
NPN
jfcom
clone
or
force
a
package
to
be
able
to
package
up
all
your
dependencies.
But
let's
keep
the
conversation
on
the
PR
thanks
for
splitting
up.
I
totally
agree
thanks
everyone
until
next
week,
thanks
especially
Senate
Anton
I
hope
you
guys
can
join
us
more
often
and
thanks
for
your
contribution
from
species,
if
I.