►
From YouTube: JupyterLab Team Meeting - March 16 2022
Description
A meeting to share and discuss features, ideas, issues, and pull requests in JupyterLab and other Jupyter frontends. This meeting is open to anyone and everyone.
Join future calls via the Jupyter community calendar: https://docs.jupyter.org/en/latest/community/content-community.html#jupyter-community-meetings
Notes for upcoming meetings can be found on the agenda: https://hackmd.io/Y7fBMQPSQ1C08SDGI-fwtg
Past notes can be found on the JupyterLab team compass: https://github.com/jupyterlab/team-compass/issues?q=is%3Aissue+label%3A%22Dev+Meeting+Minutes%22+
A
Okay,
welcome
everyone
to
the
march
16th
jupiter
lab
weekly
call.
Today
we
have
about
15
people
on
the
call
and
if
you
don't
like
that,
number
blame
daylight
savings.
A
B
Yeah
just
two
points
for
me:
the
I'll
go
for
the
second
one.
First,
I'm
not
sure
who's
active
on
the
jupiter
lab
get
project.
Currently,
I
know
frederick
is:
is
anybody
else
currently
active
on
there
or
like
doing
a
review
on
there
just
wondering?
Because
I'm
getting
some
asks
from
our
client
to
fix
some
bugs
and
issues,
and
I
want
to
know
who
to
ping
for
review
when
I
start
opening
those.
C
I
can
tell
you,
I'm
the
only
one,
because
I'm
also
mentoring,
two
students
from
from
a
american
project,
for
they
are
doing
some
some
bug,
fixes
and
future
implementation
right
now
or
so
for
for
client
of
that
that
institution,
so
I'm
just
as
as
a
mentor,
it's
not
officially
true
on
stock,
so
I'm
the
only
one
maintaining
it
lately
like
by
lately.
I
mean
for
two
years
something
like
that:
okay,
that's.
B
Good
to
know,
then,
and
then
my
other
bullet
point
is
a
fault.
Last
week
we
talked
about
adding
a
plus
button
to
the
tab
bar
to
open
the
launcher,
and
I
did
that
I
was
interested
in
knowing
if
anybody
wanted
to
see
a
demo.
Otherwise
the
pr
is
there
and
ready
for
people
to
try.
B
Interested
in
the
demo,
okay,
let
me
share
my
screen,
then
we'll
just
share
safari
just
to
be
safe.
D
B
Cool
and
everybody
can
see
that
there's
a
pretty
plus
button
here
now
so
yeah
when
you
click
it,
it
opens
a
launcher.
It
also
opens
launcher
specifically
next
to
the
current
launcher,
which
is
the
same
functionality
as
any
other
button.
You
can
also
make
multiple
sections
and
when
you
click
it,
it
will
open
in
the
area
it's
expected
to.
B
Fantastic
yeah:
this
is
great,
it
floats
and
it
fits
at
the
end.
So
I
have
to
say
thank
you.
I
forget
who
it
was,
but
somebody
implemented
this
in
lumino
last
may
so
this
was
much
easier
because
of
that.
How
does
this.
E
B
Got
it
awesome
plus
is
not
actually
in
the
list.
They
are
too.
The
whole
list
of
tabs
is
a
sister
dom
element,
okay,
which
is
why
getting
this
to
float
to
the
tab
side
was
actually
took
some
work
with
some
styling,
because
by
default
it
actually
just
always
sits
over
here
on
the
right
yeah.
E
B
So,
and
it
has
highlighting
so
just
like
any
other
tab
when
you
mouse
over
it,
it
highlights
and
it
looks
actually
in
my
opinion,
it
looks
even
better
in
dark
mode.
F
B
Super
cool
yeah,
essentially
any
tab
bar
that
is
created
in
the
dock
panel,
which
is
this
main
area
section,
will
have
that
button.
But
if
you
create
a
tab
bar
anywhere
else
in
code
so
like,
if
somebody
has
an
extension
and
they
create
a
tab
bar,
it
won't
be
included
by
default.
We,
the
launcher
extension,
specifically
enables
it
for
the
dock
panel.
E
Question
yeah:
will
the
plus
button
appear
if
someone
deploying
jupiter
lab
has
swapped
out.
B
F
B
This
plus
button
is
essentially
this
plus
button.
It
does
exactly
the
same
functionality.
In
fact,
it
actually
calls
the
file
editor
wrapper
for
adding
a
launcher.
B
B
B
Sweet
thank
you
for
showing
this
off.
This
is
great.
How
do
you
get
it
to
stop
sharing
now
there?
It
is
okay,
yep,
so
quick
demo
for
everybody
that
pr
is
linked
in
the
notes.
If
people
want
to
go
track
for
themselves
and
review
it.
A
Okay,
frederick,
you
are
up
next.
C
A
G
Hey
everyone,
so
I
have
a
a
small
pr
about
a
new
concept
called
data
set
and
data
registry.
This
is
an
attempt
to
create
a
new
interface
called
data
set
and
a
data
registry
which
will
let
you
manage
data
sets,
so
I
didn't
know
exactly
where
to
open
it.
So
I
have
created
a
just
a
temporary
repo
in
in
my
alias
for
this.
It
has
two
files
only
right
now.
It
has
all
the
concepts
for
this.
G
The
interfaces
and
it
has
user
stories
for
the
data
set
and
data
registry
so
feel
free
to.
You
know,
comment
on
that.
Pr
I
can.
I
can
move
it
to.
You
know
a
new
repo,
if
there's
a
more
appropriate
place
for
this,
and
I
can
I
can
share
and
talk
about
it
a
little
bit
if
you,
if
you
want.
E
And
just
to
clarify
some,
some
of
what
we're
hoping
to
get
input
on
is
push
is
basically
continuing.
The
work
that
saul
shadowbrook
had
originally
done
on
the
existing
data
explorer
data
registry
for
jupiter
lab
at
this
point.
It's
more
or
less
a
complete
rewrite,
and
so
we're
not
clear
should
push
just
submit
a
pr
overriding
that
entire
repo.
Do
we
archive
the
old
one
and
pull
a
new
one
in
that
would
be
a
new
sub-project
or
what
we'd
like
guidance
on
these
questions?.
A
Is
that
repo
is
any
any
author
of
that
repo
currently
active?
It
seems
to
me,
like
maintaining
version
history
of
a
thing
that
is
a
natural
successor
is
preferable
to
clobbering
it
and
losing
that
history.
Unless
there's
no
one
there
to
merge
it
in
basically.
E
H
I
mean,
as
usual,
when
it
comes
down
to
things
that
are
very
advanced
use
cases
that
drastically
change
like.
Can
it
be
an
extension
first,
so
that
can
people
can
try
it
next
to
their
stuff
without
having
to
work
off
of
a
fork
or
whatever
right
like?
There
was
a
lot
of
interest.
I
you
know
I
briefed
right
after
you
out
at
nasa
about.
H
You
know
related
stuff
and
people
were
interested
in
the
data
registry,
but
they
couldn't
really
try
it
next
to
the
other
things
that
they
were
interested
in,
because
it
was
off
on
this
fork
and
you
had
to
go
through
a
rather
involved
thing
like.
Can
it
be
an
extension
first
that
people
can
try
out.
G
Yeah,
it
is
the
way
it's
the
implementation
right
now,
it's
it
is
an
extension,
but
we
we
just
wanted
to
lay
out
the
concepts
initially
to
get
some
feedback.
I
do
have
the
implementation.
I
can
post
that
as
well.
A
That's
my
intuition,
I'm
not
prescribing
unilaterally
this.
If
other
people
have
other
ideas,
you
know
we
should
be
open
to
it,
but
my
intuition
would
have
been
yeah
open,
a
pr
against
the
extension
that
already
exists,
maybe
even
bump
the
version
so
that
it's
clear
what's
happening
here
and
just
release
a
new
version
of
the
extension.
A
All
right,
so
I'm
looking
we're
talking
about
the
jupiter
lab
data
explorer
extension
right.
Yes,
yeah,
so
of
the
one
two
three
four
five,
six,
seven
people
I
see
in
the
committees
there's
saul
and
athen,
who
are
not
here,
there's
max.
Who
often
is
here
as
the
third
person,
so
maybe
he'd
have
input.
A
I
don't
know
if
brian
henning
comes
to
these
calls
ever
and
then
there's
tim,
bryan
and
vidar,
so
it
seems
like
there's
at
least
some
people
pretty
active
still
who
know
something
about
that
and
would
be
capable
of
weighing
in
if
you
pin
them
on
a
pr
to
that
repo.
Okay.
So
that's
good
I'll.
Do
that.
G
J
Hey
there,
I
try
and
go
through
the
needs,
triage
tag,
as
has
been
requested
every
now
and
then
last
week.
That
led
me
to
a
question
about
how
to
cite
jupiter
lib
I'll
link
it
in
the
chat,
and
I
thought
we
had
some
kind
of
citation
guide
or
some
way
we
like
to
be
cited
for
jupiter
labs
specifically,
but
I
couldn't
find
one
and
I
wanted
to
check
if
that
was
me,
just
not
finding
it
this
time.
J
E
Yeah,
I
I
can
give
background.
This
is
one
of
those
things
that
is
on
all
of
our
very
long,
personal
and
and
corporate
your
community
to-do
lists.
Our
intent
was
always
to
write
a
paper.
E
I
forget.
What's
this
journal,
that's
focused
on
software
artifacts
joss
journal
of
open
source
software.
Lorena
barber
has
been
involved.
It's
super
simple.
It's
basically
like
a
one-page,
markdown
document
in
the
repo
and
then
a
submission
process,
quick
review,
you
get
a
doi,
so
it's
always
been
our
intention
to
do
that.
We've
never
done
it
there
and
there's
no
like
fernando
and
myself.
Any
of
the
other
original
people
involved
in
creating
jupiter
lab
have
not
written
a
paper,
and
so.
D
E
Yeah
I
mean,
I
would
say
the
the
current
answer
today.
Is
that
there's
not
an
official
way
of
citing
it?
If
we
want
to
take
an
action
item,
I
think
it
would
be
for
us
to
write
one
a
joss
paper,
which
is
a
pretty
simple
thing
it
we
did
this
for
altair
jake
van
der
pless,
and
I
did
it
for
altair
and
it
was
a
few
hours
worth
of
work.
It
was
straightforward.
A
A
I
don't
know
about
this,
but
one
potential
action
to
fix
that
issue
could
be
adding
a
citation.cff
file
to
the
jupiter
library
call
if
that
is
actually
appropriate
for
the
task
at
hand.
Again,
I
don't
know
this.
I
think
cff
stands
for
citation
file,
format,
I'll
I'll
link,
the
page
about
it
in
the
notes,
but
I
can't
I
can't
vouch
for
it.
This
is
just
like
literally
what
I
looked
up
while
we
were
talking
about
it.
Oh
thanks.
Gary
I've
never
seen
that
before
yeah
me.
A
J
Another
potential
intermediate
is,
I
noticed,
because
I
looked
around
what
other
projects
are
doing.
I
think
what
binderhub
does
is
they
cite
one
of
like
the
early
talks
that
they
gave.
I
don't
know
if
there's
any
talk,
someone
would
want
cited,
or
I
mean
I
know
there
is
the
there's
some
blog
posts.
I
don't
know,
if
that's
better
or
worse,
that's
why
I'm
bringing
it
up
yeah
the.
E
But
these
days
we
should
not
be
the
only
ones
being
cited
in
an
author
list
on
jupiter
lab.
That
would
be
a
complete
gross
misrepresentation
of
what's
happened,
and
so
I
I
I'll
look
up
the
document
that
jupiter
has
describing
the
process
of
how
we
would
build
an
author
list
for
a
paper.
That's
as
inclusive
and
and
representative
as
possible.
J
I
linked
the
binder
one
because
that's
what
I
thought
too,
when
I
checked
and
it
instead
it's
funny.
The
author
is
just
jupiter,
even
though
the
talk
was
definitely
given
by
specific
people.
I
think,
but
I
could
be
misunderstanding.
I
don't
have
the
same
academic
background
as
a
lot
of
you,
but
I
thought
it
was
an
interesting
approach.
J
A
E
I
just
pasted
the
link
to
this
document
in
the
jupiter
governance
about
writing
papers
related
to
jupiter.
A
Okay,
nick,
you
are
up
next.
H
Oh
no,
I
forgot
what
I
wrote.
I
think
I
had
a
couple
things
that
I
put
down.
H
Yeah
I'm
on
it
I'm
on
it.
I
found
it
okay,
so
we
are
using
a
file
loader
and
url
loader
and
some
other
loaders
that
are
not
actually
compatible
with
webpack,
five
anymore,
apparently
or
they're
not
compatible
with
the
newer
style
or
I
don't
know
whatever.
But
I
ran
into
a
bunch
of
trouble.
Replacing
those
with
the
new
asset
stuff
does
two
cool
things:
one.
You
don't
need
to
do
as
much
weird
loader
stuff
two.
It
doesn't
even
parse
those
files,
so
it
actually
makes
the
build
faster.
H
So
if
we
were
not
web
packing,
you
know
if
it
if
it
never
even
opened
the
json
files,
if
it
never
even
opens
whatever
files
that
might
help
the
build
performance
so
not
on
three,
because
people
rely
on
those
versions
of
those
loaders
and
how
they're
configured
and
everything
like
that,
but
on
four
we
should
definitely
try
and
lean
as
heavily
as
possible
on
the
asset
stuff.
So
I
just
figured
this
out
last
night,
so
I
haven't
made
an
issue
or
anything
like
that.
H
But
jeremy
said
that
we
should
talk
about
it
upstream
and
I
know
everybody
loves
webpack
problems
because
they're
the
best,
the
other
one
there's
an
issue
open
on
jupiter
lab
server
about
what
to
do
about
the
lsp
handlers.
H
Should
we
move
start
thinking
about
how
to
bring
those
into
the
house,
and
I
got
lots
of
opinions
because
I
don't
think
there's
a
right
answer,
but
I
just
wanted
to
point
that
out
as
a
place
for
people
to
weigh
in
if
they
have
users
that
are
using
lsp,
we
have
a
lot
of
people
using
the
jupiter
llc
stuff
and
really
not
that
many
people
actually
like
contributing
back
other
than
help
me
with
my
language
server
problems
right.
H
So
it
would
be
super
helpful
if
we
got
some
feedback
on
how
we
might
eventually
bring
some
of
that
stuff
into
core
or
if
we
never
want
to
touch
it,
because
we
don't
trust
microsoft,
which
is
also
totally
valid,
and
we
don't
need
that
inside
of
our
inside
of
our
code
base.
That's
that's
also
fine.
Those
are
just
two
issues.
A
So
what
are
these
opinions
that
you
have?
What,
with
being
informed,
it's
helpful
to
hear
what
you
think
should
happen.
H
Well,
I
I
mean,
like
I
originally,
for
example,
when
the
when
the
new
yjs
thing
came
out,
it's
just
a
brand
new
websocket
hanging
out
there
and
it
only
lives
on
it
only
lives
on
the
jupiter
lab
server
right
or
it's
in
jupiter
lab.
It's
not
a
google
app
server.
Maybe
that's
part
of
the
problem
right.
So
when
we
introduce
stuff
on
a
leaf
package
that
ties
it
forever
to
that
implementation
of
that
server
and
it
lives
in
the
open
api
space
of
that
server.
H
Much
to
our
discussion
on
the
previous
hour,
it
makes
it
harder
for
other
jupiter
clients
to
make
use
of
that
without
also
upstreaming
the
server.
If,
instead,
we
can
move
stuff
into
the
specification
space
in
within
jupiter,
so,
for
example,
you
know
hey.
I
got
this
websocket
over
here,
it's
called
comms.
I
can
run
it
next
to
the
code
of
interest
down
the
pipe
or
whatever.
Then,
all
of
a
sudden,
the
implementations
can
use
existing
stuff
that
they
already
have,
and
they
don't
have
to
implement
things.
H
So
I
would
love
it
if
it
was
possible
to
run
language
servers
inside
the
kernel.
I
would
love
it
if
kernels
were
able
to
emit
language
server
protocol
messages
so
like
there
was
a
discussion
last
week
about
completion,
documentation
or
something
lsp
has
already
destroyed.
H
That
message
space
I
mean
it's
got
like
20
different
things
that
you
can
say
about
completions
way
beyond
what
we
have,
and
it
doesn't
say,
experimental,
jupiter
lab
completion,
icon
or
whatever
you
know
like
there's
croft
inside
of
our
our
our
messages
right,
I
would
rather,
we
could
just
embrace
and
extend
those
lsp
messages
rather
than
having
to
iterate
on
the
jupiter
servers.
You
know
on
the
jupiter
kernel
messaging
space
forever.
H
If
there
is
a
path
forward
where
other
clients
can
make
use
of
of
that,
you
know
technique
and
then
the
other
one
is
if
we
wanted
to
actually
ship
anything
like
a
useful
ide
experience
that
had
jump
to
reference
and
refactoring
and
and
look
up
definition
and
all
that
that
means
we
would
have
to
then
grandfather
in
something
you
know:
we'd
have
to
pick
a
language
server
that
worked
with
something
we
had
and
I'm
really
not
comfortable
with
that
right
now,
because
it's
again,
the
one
that
seems
to
be
any
good
is
not
open
source
right,
the
pylance,
whatever
from
whatever
from
redmond.
H
H
Even
if
you
only
care
about
settings
you
care
about
json
you
care
about
json
schema.
So
if
we
could
ship
in
browser
language
servers
that
jupyter
lab
didn't
know
that
it
was
in
the
browser,
that
would
be
great,
so
we
could
ship
the
json
language
server.
We
could
ship
the
yaml
language
server,
because
those
are
jupiter
things
that
we
have
to
deal
with,
but
at
the
end
of
the
game,
what
will
it
mean
when
you
are
working
with
a
polyglot
document?
H
And
there
are
many
sources,
not
just
one
language
server,
but
there
are
many
sources
of
of
annotations
that
are
interested
in
the
layers
of
what
you
have
there.
So
it
might
have
a
a
link.
Checker
for
your
markdown
is
one
kind
of
thing,
and
your
spell
checker
can
work
on
your
markdown,
but
it
can
also
work
on
comments
inside
your
code,
etc,
like
there
isn't
really
a
model
like
that
today,
because
most
of
the
lsp
clients
work
with
a
line
and
column
based
source
document.
H
There
isn't
a
lot
of
examples
out
there
of
I'm
looking
at
a
polyglot,
hypermedia
document
that
is
also
computable
and
has
outputs
that
might
be
relevant.
So
you
know,
we've
got,
we've
got
many
directions
that
we
could
go.
I
think
we
need
to
go
in
all
of
them,
but
getting
something
in
first
is
probably
better.
So
those
are
my
opinions-
and
I
have
laid
them
out
on
on
that
issue
there,
but
I'm
really
interested
to
hear
people
are
actually
pushing
lsp
on
any
of
their
their
clients
and
what
their
experiences
have
been.
A
So
what
do
other
people
think-
and
just
just
I
can't
imagine
nick
accidentally
said
embrace
an
extent
when
talking
about
microsoft
protocol,
but
anyway,
what
do
other
people
think
about
this?
This
is
a
this
is
a
really
attractive
feature
set.
I
I'd
like
to
see
how
we
can
get
it
out
to
more
people
and
I'm
curious
what
other
people's
thoughts
are.
So.
K
Yeah,
I
feel
bad
because
I
just
hop
back
on
in
the
end
of
nick's
explanation,
because
I
was
dealing
with
some
emergency
here,
but
we've
had
some
discussions
earlier
with
nick
about.
K
Where
do
we
elect
the
lsp
protocol
like
whether
we
make
it
try
to
make
it
part
of
you
know
the
jupiter
kernel
protocol
or
like
as
a
separate
thing,
and
so
just
wanted
to
add
one
thing
so
rather
than
you
know,
contrasting
or
whatever,
because
I
haven't
heard
everything.
K
K
So
we
definitely
want
to
sort
of
have
like
core
support
for
this
in
jupiter
lab.
So
I
just
wanted
to
second
that,
from
from
nick.
A
It
sounds
like
the
same
way
we
had
a
for
adding
the
debug
adapter
protocol
to
the
kernel,
that's
probably
a
good
model
for
adding
lspd,
and
that
way
we
don't
have
to
deal
with
the
problem
of
okay,
but
what
language
server?
Do
we
ship
we're
just
talking
about
a
messaging
spec,
which
is,
as
I
understand
it?
One
of
the
approaches
you
proposed
and
the
one
you
advocated
for
nick
is,
is
that
correct.
H
Well,
yeah,
I
mean
the
the
specific
one
the
the
mvp
that
I
built
was.
There
is
a
kernel
that
takes
over
the
role
of
bespoke
websockets,
so
the
way
that
lsp
works
right
now,
it's
a
server
extension
and
it
adds
some
new
routes
and
they're
named
on
the
server
or
whatever,
and
it
spins
them
up
when
they
need
them
so
that
we
didn't
have
to
you
know
we
we
didn't
have
a.
It
only
has
a
restful
api.
H
H
You
just
have
one
attack
session,
one
markdown
session
and
there's
a
lot
of
limitations
with
that,
where
you're
expecting
the
code
to
be
co-located
with
the
server
and
all
this
stuff
breaks
down
entirely
when
you
get
to
jupiter
kernel,
gateway
and
stuff
like
that,
so
I
want
there
to
be
some
way
that
you
can
run
kernels
next
to
language
servers
on
the
same
file
system
and
it's
hard
as
well
as
running
local
servers.
H
It's
got
to
be
really
easy
to
get
it
stood
up,
so
I
I'm
I'm
just
I'm
not
sure
about
adding
a
whole
other
set
of
websockets
to
core
that
already
require
some
other
extremely
complicated
piece
of
software
to
run
next
to
them.
Nobody
has
success
running
these
language
servers
out
of
the
out
of
the
box.
You
know
so,
whereas
we're
heavily
invested
making
kernels
work.
So
I
I
don't
know
I
feel
like
if
ipython
started
started,
you
know
or
ipi
kernel
started,
putting
out
language
server
messages
that
would
start
getting
people
into
a
cooler.
K
K
Kernel
is
about
execution
and,
most
of
the
time
when
we
leaked
the
information
to
the
kernel
about
what
the
client,
what
client
we're
using
whether
it
was
a
notebook
or
console,
we
considered
that
it
was
an
abstraction
leak.
So
we
tried
to
make
the
kernel
completely
unaware
of
any
notion
of
document,
while
the
lsp
is
really
about
the
documents
and
about
the
entire
workspace.
K
It's
it's
a
global
thing.
Where
you
may
have
many
python
kernels
running
in
a
given
environment.
You
must,
typically,
you
will
have
only
one
language
server
for
that
environment,
so
they,
you
know
the
abstractions
don't
really
align
to
me,
and
so,
in
my
opinion,
the
debug
protocol
really
worked
well
with
the
kernel
protocol,
but
not
lsp.
K
So
what
I'd
like
to
see-
and
I
think
having
a
new
web
socket
is
fine.
I
mean
we
have
a
websocket
endpoint
for
yjs.
You
know
for
real-time
collaboration
and
we
have
a
websocket
endpoint
for
kernels,
and
now
we
may
have
another
one.
For
you
know
language
servers
right,
probably
if
we
make
a
simple
one
available,
I
don't
know
for
json
or
some
other
simple
language.
A
Okay,
this
is
this
is
something
that
I
guess
a
good
takeaway
from
the
conversation
would
be.
Do
we
need
a
to
move
forward
on
this,
or
is
there
a
way
of
adding
these
endpoints
just
to
the
next
version
of
server,
because
it's
a
major
version
change
and
that's
it
like?
If
we're
not,
if
we're
not
imposing
anything
on
the
kernel
spec,
if
we're
simply
adding
end
points,
it
seems
like
a
natural
thing
to
do
in
the
next
version
of
server,
without
going
through
a
formal
like
ratification
process,
do
you
have
an
intuition
on
that.
I
H
No,
I
know
it's
a
hard
problem
too.
I
would
really
like
to
see
something
that
worked
in
browser.
I
think
that
json
fellow
running
inside
the
browser
that
could
already
make,
even
though
we
just
got
our
brand
new
awesome
settings
editor,
but
making
schema,
constrain,
json
editing
better.
Would
it
would
already
move
things
forward
right?
You
know
that.
Would
that
would
start
getting
us
thinking
about
more
of
these.
H
Right
like
to
your
point
yeah,
we
shouldn't
even
try
and
pick
a
working
python
one
and
we
shouldn't
write
one
either
good
lord,
so
I
I
feel
like
there's,
you
know
some
opportunities
for
getting
some
of
the
features
and
using
the
features
for
simple
non-executable
documents
to
kind
of
get
our
feet
wet
before
going
to
all
of
it.
Working
really
well,
yeah,
they're
doing
a
good
job
on
python
lsp,
but
man.
That
was
a
really
painful
experience
over
the
last
couple
years.
H
Right,
like
I
know,
they're
doing
a
good
job,
and
I
know
a
lot
of
people
are
invested
in
lsp,
but
individual
servers
are
really
hard
to
pick.
We,
we
grandfather
in
ipay,
kernel
on
on
jupiter
lab
and
that's
what
most
people
expect
to
get.
I
don't
think
we.
I
don't
think
there
is
a
clear
front
runner
on
which
one
we
say
like
if
you
download
jupiter
lab
you're,
also
going
to
pull
in
this
right.
H
Like
you
know,
it's
it's
it's
hard
to
pick
at
this
point,
whereas
there's
really
only
one
or
two
json
servers,
because
there's
only
so
much
you
can
really
do
with
it.
H
H
We
got
some
more
opinions
here.
Yeah
pyrite's,
lockdown,
difficult
to
customize,
yeah.
F
H
I
mean
you
know:
people
talk
about
lsb
being
safe
and
it's
like
not
really.
Only
some
of
the
language
servers
actually
work
without
code
execution,
I
mean
they
need
access
to
a
lot
of
stuff,
but
more
of
its
dependency.
Bloat
right.
That's
that's
the
thing
that
gives
me
the
most
nightmares.
Is
you
know
you
you,
you
kinda
install
whatever
jupiter
lab
today
and
you
are
pulling
stuff
from
all
over
the
universe.
H
Right
like
it
is
very
large
and
so
I'm
very
hesitant
to
add
any
of
those,
because
they're
usually
very
tightly
tied
to
the
language
grammar.
You
know
all
that
kind
of
stuff,
whereas
json
we
kind
of
understand
what
that
is.
We
already
have
ajv
in
the
browser.
That's
like
the
primary
dependency
of
the
the
language
server
for
json.
H
Hodor
there
we
go
so
anyhow.
A
H
I
don't
know,
there's
a
I
I
would.
I
would
like
us
to
get
the
ui
of
the
language
server
stuff
squared
away
before
trying
to
make
it
work
like
we
have
to
do
insane
things
just
like
that
does
to
get
notebooks
working
inside
of
a
random
language
server,
and
that
is
very
painful
and
it
doesn't
understand
things
like
I
don't
know
it
doesn't
like.
What
is
how
would
you
auto
complete
on
what
what,
if
you're,
in
a
markdown
document?
H
How
would
you
auto
complete
on
what
attachments
are
available
right
like
a
language
server,
had
no
idea
but
have
no
idea
how
to
do
that
kind
of
stuff
so
yeah?
I
don't
know
that
more
more
insights
there
than
just
me
and
sylvan
shouting
at
each
other
and
any
stories
that
people
have
about
how
lsp
is
helping
or
frustrating
their
users
is,
is
kind
of
what
we're
looking
for
there
yeah.
A
Look
in
the
chat
there's
a
lot
of
really
good
stuff
in
there,
but
also,
I
think
we
should
take
up
this
conversation
and
tomorrow's
jupiter
server
call
as
well,
because
I
think
actually,
a
first
move
toward
bringing
these
features
to
core
is
the
end
points
and
then
a
second
move
is
writing
the
wrappers
for
those
endpoints
and
jupiter
lab
services.
So
why
don't
we
continue
the
conversation
there
unless
there
is
something
that
someone
wants
to
add
right
now,.
H
Okay,
cool
and
maybe
it's
relevant
to
the
previous
mb7
thing.
You
know
really
jupiter
server,
proxy
and
lsp
server
have
many
of
the
same
requirements,
and
originally
the
lsp
stuff
was
built
on
top
of
jupiter
server
proxy,
but
we
needed
more
control
and
stuff
like
that.
So
maybe
jupiter's
server
needs
to
be
a
general
purpose
process
manager
with
streaming
logs
and
start
and
stop-
and
you
know
environment
variables
and
all
that
stuff.
I
don't
know.
I.
A
I'm
gonna
say
that's
outside
the
scope
for
the
jupiter
lab
call.
Maybe,
though,
bring
it
up
tomorrow,
I'm
sure
it'll
be
real,
lively,
okay,
so,
let's
see
jeremy,
you
are
up
next.
L
Oh
yeah,
a
quick
one.
We
published
a
blog
post
yesterday
about
jupiter
lights
and
a
blog
post
is
called
jupiter
everywhere
and
the
whole
point
behind
this
was
just
to
show
people
how
they
can
embed
their
own
python
console
on
their
website.
L
And
if
you
want
to
give
it
a
try,
check
out
the
breakfast,
because
we
try
to
make
it
as
easy
as
possible
for
you
to
or
for
anyone
to
use
it.
That's
it.
A
So
this
is
super
cool.
I
don't
know
if
you've
seen
it,
but
now
people
are
embedding
jupiter
lab
in
all
sorts
of
places
that
you
wouldn't
have
thought
and
yeah
follow
those
links.
They
are
delightful.
A
Okay,
it
looks
as
though
we
don't
have
any
other
people
who
wanted
to
talk
about
anything
and
no
one
added
any
additional
discussion.
Can
I
just
give
you
a
minute
in
case
someone
has
thought
of
something
they
would
like
to
talk
about,
since
we
have
some
time.
A
H
I
know
right,
I
dropped
it
over
in
the
chat
in
addition
to
in
browser
language
servers.
I
think
we
might
want
in
browser
kernels.
I
think,
there's
value
there.
I
don't
know
I
don't
know
so.
The
light
experience
like
we've
already
had
someone
ask
for
that
and
they're
trying
to
do
really
complicated
like
iframe
stuff-
and
I
don't
know
whatever,
but
there
is
the
the
webassembly
is
not
going
anywhere.
H
We
will
be
seeing
more
things
available
for
it
and
if
jupiter
is
the
best
way
to
do
that,
a
lot
of
the
other
in-browser
stuff
is
not
as
cool
as
as
jupiter
lab
is
right,
like
we
can
do
cool
stuff,
we've
got
the
rich
display
system.
We've
got.
You
know,
we've
already
been
down
a
bunch
of
those
roads.
I
think
if
we
were
a
part
of
that,
that
would
be
really
good
and
then
we
wouldn't
need
the
jupiter
light
project
anymore.
It
would
just
be
a
different
flavor
of
jupiter
lab.
L
Yeah,
I
think
the
what
the
tripty
right
project
brings
is
different
than
what
lab
brings.
It's
really
just
this
in-browser
server
that
we
kind
of
try
to
unlock
so
that
we
can
start
kernels
in
a
browser,
but
we
could
imagine
lab
having
some
changes
to
allow
these
kinds
of
things
to
be
blocked
plugged
in
more
easily.
L
Maybe
it
requires
like
a
deep
restructuring
of
like
how
the
service
manager
works,
for
example,
but
I
mean
it
would
be
really
cool.
Indeed,
if
we
could
have
some
kind
of
hybrid
mode
where
people
could
even
choose
whether
they
just
want
the
normal
gpu
live
experience,
jupiter
plus
light
or
just
a
light
static.
That's
it
like.
It
is
right
now,
so
I
think,
having.
L
We
will
still
need
to
be
able
to
have
this
flexibility
and
not
require
again
this
python
server,
but
I
mean
if
we
can
yeah,
I
guess
you
make
things
simpler
right
in
the
long
term,.
H
Yeah
drop
a
link
there
I
mean
to
to
do
the
the
kernel
side
of
the
house
of
having
in
browser
kernels
or
just
whatever
kernels,
if
the
service
manager
offered
ways
to
add
new
sources
of
kernel
specs.
So,
instead
of
just
talking
to
services,
if
there
was
something
that
could
look
for
kernel
specs
and
then
when
you
fire
up
that
new,
when
you
try
and
create
a
new
session
with
that
kernel,
you
need
to
intercept
that
those
are
the
only
two
things
that
are
needed
for
that
side
of
it.
H
L
But
light
also
has
this
plugin
system
for
the
server
right
we're
using
the
definite
extension.
I
guess
we
will
also
want
that
to
work
if
we
have
this
hybrid
system
that
that's
the
way
people
bring
their
own
kernels
to
the
page
yeah
another
thing
to
consider.