►
From YouTube: JupyterLab Team Meeting - 22 February 2023
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/170
A
A
We
have
an
agenda
which,
if
you
haven't
already
seen
I
will
drop
into
the
chat
here,
looks
like
we
got
a
pretty
light
agenda
so
far,
but
if
anyone
would
like
to
add
something
to
it,
please
feel
free
once
we're
done
with
the
agenda
and
I'll
stop
the
recording.
Before
this
we
can
have
some
additional
off
the
Record
discussions
and
today
we'll
take
a
look
at
issue
triage
for
Jupiter
lab
400
enhancements.
A
B
Hello,
everybody
sorry
I
have
a
bad
internet
connection
lately.
So
no
video
for
me
just
first
an
announcement
there's
been
the
opening
of
the
Google
season
of
ducks.
So
basically,
every
years
Google
is
providing
for
open
source
organization
Grant
to
to
finance
technical
writers
for
documentation.
I,
don't
think
we
have
applied
in
the
past,
but
I
think
it
would
be
good
if
we
can
try,
because
our
documentation
is
one
of
the
the
part
we
are
really
like.
A
user
is
looking
for,
are
looking
for
improvement
on
that
site.
Let's
say
so.
B
If
somebody
wants
to
to
to
submit
submit
something
or
maybe
we
can
create
an
issue
on
the
on
the
team
compass
and
try
to
synchronize
their
with
people
that
are
interested
in
that
subject.
So
basically,
what
we
need
to
provide
is
support
for
like
decide
what
we
want
as
a
as
a
target
for
the
technical
writer
and
also
provide
support
for
him
in
IMC,
and
so
that's
for
the
announcement,
and
the
second
thing
is
just
the
question
that
I
was
last
week
just
to
be
sure.
B
Is
it
Okay,
then,
if
I
do
a
release
candidate
of
lumino,
2
I,
don't
know
if
my
case
there
today
yeah
I,
see
Mike
because
I
think
you
you
were
trying
to
maybe
have
another
an
additional
thing:
okay,
so
okay,
so
Mike
said
on
the
jet:
go
for
it.
So
I'll
do
the
release
tomorrow
of
the
the
release
candidate
of
three
minutes
and
thanks
and
that's
why
I
set
up
an
issue
on
the
team,
Compass
I!
Think
for
the
Google
of
Doc
thing.
A
C
A
C
You
go
yep,
thank
you,
so
hey
everyone,
so
I've
been
having
and
I've
been
getting
a
error
message
in
extension,
manager
and
I
can
reproduce
it
pretty
reliably
and
it
overflows
so
I
was
looking
into
why
it
overflows,
and
this
error
message
and
a
lot
of
error
messages
over
the
Egyptian
lab
code
base,
our
formatted
with
pre-tax
bre
HTML
tags
like
the
pre-formatted
text
text,
which
the
default
behavior
for
the
stack
is
to
display
strings
and
white
spaces,
in
particular
in
the
same
manner
as
they're
displayed
in
HTML.
C
So
because
of
this,
you
can
display
the
ASCII
like
art
using
this
text,
but
you
know
text
would
overflow
if
you,
if
the
parent
node
doesn't
have
enough
bits
or
space
for
the
text.
So
you
can
go
to
this
link
to
see
more
details
for
this
discussion.
I
wanted
to
get
a
background.
Like
is
this
functionality?
Is
it
like
why
this
tag
is
used?
Is
my
assumption
that
it
is
used
for
a
sci
like
art
is
correct?
If
it
is
not,
light
is
used.
C
C
You
can
just
proceed
to
by
to
this
link
that
I
posted
in
chat,
and
that
is
also
in
the
agenda
and
participated
in
a
discussion.
I'm.
C
Yeah,
but
problem
still
persists
if
we
have
larger
error
message
and
we
don't
have
apparently
wide
enough
which,
apparently,
if
if,
if
parent
div
displays
arbitrary
error
messages,
it
cannot
know
how
large
of
an
error
message
is
also
if
we
don't
have
limits
for
the
length
of
error
messages.
You
know
there
are
no
limits,
so
some
error
messages
will
overflow
with
how
they
are
set
up,
so
I
invite
everyone
to
participate
in
discussion.
If
you
have
anything
to
add
yep
in
a
written
form,
in
particular.
D
E
Just
a
brief
comment:
there
was
a
recent
post
showcasing
a
situation
where
an
overflow
in
y
y
overflow
was
causing
a
security
problem,
because
the
list
of
permissions
that
were
requested
by
an
extension,
it's
completely
not
related
to
Jupiter
lab,
was
suddenly
showing,
like
the
five
that
happens
in
like
Chrome
on
Mac
OS,
because
this
group
are
is
hidden
by
default.
So
we
should
probably
be
careful
with.
We
say
that
I
don't
know
if
people
seen
that
like
X
overflow
is
not
such
a
big
issue,
but
why
overflow?
D
Yeah
I'm
Curious,
so
we've
been
getting
some
feedback
on
navigating
notebooks
with
assistive
technology
and
we've
got
errors
and
we've
got
warnings
and
sometimes
those
things
like
blend
together.
I'm
wondering
does
Jupiter
lab,
have
like
a
consistent
concept
of
warning
and
error
so
that
yeah
all
the
applications
and
sort
of
error
messages,
kind
of
appear.
The
same.
C
So
if
anyone
could
can
expand
on
what
what
food
I
would
say,
please
add,
because
my
expertise
is
based
on
this
exploration
of
Jupiter
Labs,
that
I
did
within
trying
to
solve
the
issue.
I
was
just
talking
so
from
what
I
can
say.
No,
it's!
It's
pretty
much
just
a
strings
displayed
in
a
dialogues
or
wherever.
So,
yes,.
A
This
seems
like
this
would
be
a
good
enhancement
request
for
accessibility
reasons.
F
F
Yeah
everyone
is
welcome.
There
I,
don't
know,
there's
a
good
chance.
Tony
will
be
there
asking
questions,
but
we
just
talked
about,
but
you
never
know
what's
going
to
happen,
yeah
so
feel
free
to
drop
in.
If
you
have
time
the
other
meaning
I'm
going
to
announce
is
next
week
we
do
have
the
community
call.
I
know.
F
I
haven't
announced
that
in
these
calls
and
what
feels
like
a
while,
but
who
knows
so
yeah
if
people
have
demos
or
other
things,
they're
excited
for,
would
love
to
see
you
at
that
call
too
for
Pacific
time
people
we
managed
to
work
with
the
security
team,
so
big
thanks
to
them
and
Rollin
for
reaching
out
about
that.
So
we
are
now
at
like
8
A.M
Pacific
again.
If
that
helps
anyone.
F
So
sorry,
I
know
moving
later
is
probably
difficult
for
some
people
but
cycling
back
a
little
but
yeah
I
Linked
In
the
agenda,
the
discourse
post
with
all
the
details.
So
thanks.
E
A
A
G
Okay,
I'm
using
I
dialed
in
with
my
phone
as
well
so
yeah.
First,
a
quick,
a
quick
announcement
that
for
those
of
you
who
may
not
be
paying
attention
to
the
work
on
mist
and
I'm,
happy
to
expand
on
kind
of
what.
What
missed
is
for
those
who
may
not
be
aware.
But
there's
been
a
lot
of
progress.
G
I
left
some
some
links
in
the
agenda
about
to
to
the
core
the
core
tools
website
and
the
Jupiter
lab
plugin,
on
which
there's
been
a
ton
of
progress,
missed
for
as
a
quick
refresher
for
those
who
may
not
have
been
tracking
that
it
was
something
that
we
started
years
ago.
It
began
as
a
private
Reaper,
because
I
was
kind
of
afraid
of
doing
this
in
a
public
repo
and
waiting
in
the
thick
middle
of
the
debate
about
academic
markdown
and
Publishing
markdown
and
whatnot
and
rmd.
G
But
since
the
common
market
discussions
never
kind
of
went
anywhere.
Eventually,
that
was
the
format
that
Chris
holgraph
picked
up
as
the
core
of
extended
format
for
Jupiter
lab.
It's
basically
markdown
extended
with
ideas
from
restructured
text,
but
within
the
syntactic
structure
of
markdown
and
the
tool
that
the
team
at
curvnote
has
been
led
by
Rowan
cockett
has
been
making
a
ton
of
progress,
there's
now
a
core
typescript
implementation
that
can
run
both
inside
Jupiter
lab
and
add
to
the
command
line
and
therefore
what
what
we're
thinking
is.
G
This
is
probably
the
future
of
the
executable
books
of
the
Jupiter
book.
Machinery
will
be
eventually
at
some
point
to
replace
the
Sphinx
Engine
with
this
typescript
engine
that
can
be
consistent
both
for
static
rendering
at
the
command
line
and
for
rendering
live
in
Jupiter
lab.
It's
really
really
cool.
G
There's
I'll
suggest
that
Rowan
maybe
come
to
the
Community
College
Isabella,
thanks
for
flying
that,
because
I'm
going
to
talk
to
them
and
see
if
either
him
or
someone
else
from
the
team
is
available
to
do
a
little
demo.
Next
week
this
weekend
he
was
texting
me
I'll
drop
some
links
on
the
chat,
some
of
the
cool
things
he's
done
with
like
inline
sparklines,
one
thing
that
will
be
relevant
I
know
many
people
have
asked
for
this.
G
In
fact,
I've
heard
a
number
of
people
who
have
said
they
can't
use
Jupiter
book
and
they're
using
Porto,
because
we
don't
have
inline
kernel
variable
rendering
in
the
markdown
in
in
by
default
in
Jupiter
and
quarto.
Does
that
and
our
markdown
does
that
and
this
weekend
Rowan
has
been
working
on
that
there's
an
issue.
That's
a
pretty
complicated
topic.
People
have
lots
of
opinions
both
on
the
syntax
and
caching
and
rendering.
G
G
Rowan
is
going
to
be
remotely
participating
in
the
notebook
format
Workshop,
but
Steve
perves
from
his
team
will
be
there
in
person
and-
and
my
hope
is
that
anyone
who's
interested
on
the
document
side
of
Jupiter,
the
markdown
side
of
things-
participates
in
these
discussions,
because
I
I
think
Mist
eventually
should
become
our
default.
Markdown
format
I
mean,
obviously
that
discussion
will
take
place
in
depth
at
the
workshop.
G
But
I
think
it
has.
It
meets
that
it
meets
a
lot
of
things
that
we've
needed
for
a
long
time
and
that
the
kind
of
execute
the
markdown
Community
never
quite
figured
out
a
solution
and
having
something
which
is
well
cleanly
woven
into
the
architecture
of
Jupiter
and
understands
the
rest
of
our
system
will
be
fantastic
I'm
using
it
I'm
testing
it
with
my
students.
It's
a
lot
of
fun.
G
It
does
things
that
we
wanted
for
years,
so
I
hope
people
will
provide
feedback
and
test
it
and
and
see
if
it
meets
their
needs
and
if
it
doesn't,
can
help
improve
it.
So
that
was
an
announcement
I'm
happy
to
answer
some
questions:
I'm,
not
the
one
writing
the
code,
but
I'm
working
pretty
closely
with
Rowan
and
the
team
on
on
that
and
I'm
happy
to
answer
questions
I
I'm
super
excited
about
it.
G
There's
a
binder
link
on
the
on
the
triple
lab,
plugin
demo
that
that
I
put
I
put
a
a
link
for,
and
you
can
play
you
can
play
with
it
on
binder.
It's
kind
of
it's
really
fun.
Yes,
if
you
install
the
plugin
by
default,
it
replaces
the
markdown
cell
render
and
that's
by
design
that
can
that
behavior
can
be
disabled,
so
it
can
be
disabled
with
a
flag
on
the
command
line.
G
If
you
don't
want
that
to
be
the
case
so
that
it
doesn't
replace
it
and
then
it
becomes
an
extra
option
in
open
with
and
you
can
open
the
notebook
either
with
the
classic
renderer
or
the
new
mist.
Renderer
should
be
a
new
cell
type,
though
Mike
question
I
honestly,
don't
think
so.
Yes,
Nick
emphatically
beat
me
to
it:
I
I.
G
We
we
had
more
cell
types
in
the
past
and
and
that
that
led
to
more
complexity
than
I
think
we
were
happy
with
so
I
think
I
mean
this
is
probably
a
more
detailed
conversation
to
have
in
the
in
the
in
the
notebook
format
discussion.
But
my
my
instinctual
reaction
is
also
that
we
should
try
to
not
keep
extending
cell
types.
G
We
also
had
heading
cell
types,
for
example
at
some
point,
and
we
realized
we
could
just
understand
heading
structure
from
from
within
existing
cells,
and
that
would
simplify
a
lot
of
things.
So
that's
kind
of
where
we've
landed
so
far,
but
obviously
a
more
common,
that's
a
topic
that
admits
a
good
discussion
and
and
and
that
probably
would
be
a
good
topic-
a
good
topic
to
talk
about
at
the
at
the
notebook
Workshop.
G
Unfortunately,
I
can't
make
it
to
that
Workshop
I'm
teaching
here
and
I
can't
go
any
other
questions
on
this
topic.
Let's
see
new
message:
oh
the
multiple
kernels.
In
one
notebook
conversation,
we
had
a
version
of
that
debate
like
in
depth
years
ago,
we
landed
pretty
squarely
that
our
architecture
should
be
one
kernel
at
a
time
because
and
that
it
would
be
possible
to
implement
multiplexing
kernels.
In
fact,
I
python
is
in
a
sense
multiplexing
with
Magics.
G
If
you
have
the
right
Magics,
if
you
have
Magics
like
the
siphon,
Julia
or
Magics,
those
effectively
give
you
a
kind
of
a
multi-carnal
behavior
and
you
could
add:
we've
never
added
a
data
bus
to
them
that
that
could
be
a
a
mechanism
for
sharing,
which
is
what
what
multi-language
multi-kernel
systems
tend
to
do
again.
That
could
be
Revisited,
but
we
definitely
had.
We
had
a
workshop,
a
one-day
all
hands.
G
I
mean
when,
when
all
hands
meant
a
lot
less
people,
but
years
and
years
ago
we
had
a
very
in-depth
discussion
about
that
and
we
all
eventually
seem
to
land
squarely
on
the.
Let's
have
a
single
kernel
in
the
architecture.
G
Now
that
was
years
ago,
new
ideas,
new
people,
new
thoughts,
certainly
welcome
topic
to
be
Revisited,
but
at
least
I'm
giving
you
a
historical
perspective
that
everybody,
after
going
around
pretty
in
detail
on
that
one,
we
seem
to
conclude
comfortably
conclude
that
we
should
not
try
to
extend
the
architecture
to
to
be
multi-cernal
that
the
complexity
cost
of
having
to
abstract
over
kernels
everywhere
in
the
design
did
not
offset
the
the
benefits
and
that
we
could
get
the
necessary
benefits
by
a
lot
of
kernels.
G
That
did
that
the
way,
for
example,
IPython
does
it
or
the
SOS
kernel?
When
is
where
can
I
find
a
notebook
format?
Meeting
it's
a
workshop.
It's
one
of
the
community
workshops.
I
can
Google
it
Jupiter
Community
hold
on
notebook
4.
here,
I
just
found
the
link
here.
Let
me
drop
it
on
the
chat.
Andre
there
you
go
so
and
Frederica
will
be
will
be
hosting
is
wrote,
the
blog
post,
I'm
sure
it
knows
more
Nick
yeah.
That
was
a
discussion
I
said
years
ago.
G
G
This
call
was
that
it
was
a
long
time
ago,
so
it
could
certainly
be
Revisited
but
I'm,
giving
you
the
historical
kind
of
perspective
of
kind
of
where
we
are
and
yes
syntax
highlighting
is
trash
in
Magics,
I
I
believe
that's
always
been
a
sore
point
of
magics
I
think
that
is
something
that
we
could
fix.
It
separately
from
the
question
of
multiple
kernels
I.
Think
that's
it's
a
valid
UI
question
for
Jupiter
lab
to
improve
on,
but
but
I
think
I
believe.
G
G
G
Jason
Wilde
any
thoughts
on
like
running
the
meeting.
Do
you
want
to
dig
a
little
bit
deeper
into
like
the
mist
and
kernels
topic,
or
should
we
should
we
talk
about
the
RTC
problem.
A
G
Okay,
oh
thank
thank
you
William
for
that
link,
I
appreciate
it
and
I'll
I'm
going
to
leave
that
leave
that
open,
so
I
can
take
a
look
at
it
later.
Thank
you,
yeah.
So
the
RTC
problem
I
left
the
link
I
left
links
on
the
on
the
notes.
I
left
several
links
to
the
key
issues.
I
opened
a
new
one,
14
of
31
there's
also
another
one
and
the
previous
one
that
we
had
a
year
ago.
It
was
open
for
context,
I'm
being
pretty
aggressive
in
dog
fooding,
the
RTC
Machinery
I.
G
A
year
ago,
I
turned
it
on
in
my
class
I
in
the
spring,
I
typically
teach
a
class
with
about
70
students.
These
are
fairly
kind
of
competent,
experienced
Jupiter
users.
Typically
in
statistics,
computer
science,
economics
physics,
a
bit
of
a
mix
data
science
I
have
about
70
of
them.
It's
big
enough
to
get
very
real
world
usage,
it's
small
enough
that
that
I
can
be
somewhat
experimental.
G
A
year
ago,
we
found
data
corruption
that
led
to
a
ton
of
work
from
the
team
which
I
deeply
appreciate
and
right,
as
the
year
was
beginning
right,
around
Christmas
3.6
was
coming
out
and
and
I
was,
it
seemed
like
36
might
address
those
issues
after
a
ton
of
work
that
was
done
on
what
was
causing
data
corruption
and
so
I
give
it
a
try.
G
I
put
3.6
as
soon
as
the
semester
started
and
came
out,
I
added
it
to
our
system,
I'm
able
to
iterate
on
that
system
pretty
easily,
because
I
can
I
control
how
the
image
is
built.
It's
cloud-based
deployment
of
Jupiter
Hub
with
Jupiter
lab
and,
and
we
turned
RTC
on
and
immediately
we
started
seeing
data
corruption
again
in
this
case,
the
documents
do
not
become
as
badly
corrupted
as
we
were
seeing
them
last
year.
G
What
happens
is
they
can
duplicate
content
and
this
happens
in
notebooks
and
it
happens
in
text
files
as
well,
so
pretty
much
anything
that
you
have
open
with
more
than
one
tab
visiting
and
we
have
seen
it
happen
not
only
with
multiple
actual
multiple
people.
Different
people
collaborating,
but
simply
the
same
user
logging
in
from
two
browsers
or
having
like
my
ta,
opened
his
laptop
in
class,
and
it
was
teaching
and
then
close.
His
laptop
got
to
his
office
left.
G
His
laptop
in
his
backpack
opened
his
desktop
logged
in
again
to
the
hub,
kept
working
and
then
all
of
a
sudden,
the
notebook
like
quadrupled
all
of
his
cells,
like
it
was
a
notebook
with
only
two
cells,
and
then
it
appeared
four
times
so
you
end
up
with
four
copies
of
the
content:
I
I
Was,
preparing
a
demo
actually
for
mist
and
then
all
of
a
sudden,
I
committed
I
had
a
tech
file
in
that
in
a
report,
I
was
making
a
lot
of
repo
for
binder
I
put
it
in
and
then
later,
when
I
went
and
opened
the
binder.
G
The
tech
file
had
the
entire
content
of
the
latex
file.
Twice
I
hadn't
even
noticed
because
I
just
had
it
open
in
more
than
one
tab.
So
it's
pretty
bad.
It
seems
like
it
happens
very
easily
and
unfortunately,
it's
kind
of
a
bummer
because
I'm
willing
to
let
my
students
be
a
little
bit.
Guinea
pigs
in
the
sense
that
I
told
them
we
are
testing
this
it's
in
the
spirit
of
the
class.
G
The
course
is
about
open
source
and
open
science,
and
they
know
that
we're
using
new
things,
but
at
the
same
time
they're
my
priority
and
my
responsibility
to
them
as
University
students
is
they're
learning
not
they're,
not
they're,
not
a
Testing
Lab
they're,
not
q,
a
interns
and
so
data
loss
is
just
unacceptable
and
so
I
wanted
to
ask
the
team.
If
people
had
looked
into
this
there's
been
some
discussion.
G
If
you
think
there
is
any
chance
that
we
might
get
an
actual
fix
for
this,
very,
very,
very
fast
I
might
keep
it
open.
Otherwise,
I'm
gonna
have
to
turn
the
turn.
This
attempt
off
again
and
unfortunately
that
means
we
lose
the
opportunity
to
get
feedback
on
the
ux.
If
we
have
no
data
loss,
I'm
happy
to
experiment
with
the
system
we
were
already
I
had
already
opened
an
issue
with
our
team
to
begin
creating
collaborative
accounts
on
the
system.
G
We've
taken
some
inspiration
from
what
Coco
does
we're
collab
we're
in
contact
with
what
Min
and
others
are
enrolling.
Thomas
and
others
are
doing
it
nurse
because
there's
a
deployment
that
nurse
nurse
is
one
of
the
Department
of
energy
reading
centers,
where
we
have
a
deployment
with
RTC
we've
added
it
with
their
security.
So
we
have
an
opportunity
to
get
feedback
on
on
the
ux
and
what
this
collaboration
look
like.
G
What
are
the
missing
features,
but
obviously
we
can't
do
any
of
that
if
it's
actually
corrupting
user
data,
like
that's
just
a
no-go
so
I
wanted
to
put
this
on
the
table.
Rtc
has
been
for
a
long
time
very,
very
badly
desired
features,
something
that
we've
I
know.
We've
worked
a
lot
on.
G
I
I
think
that,
obviously
it's
unfortunate
that
this
is
happening,
but
I
wanted
to
get
some
feedback
from
the
team
on
what
you
think
might
be
fixed
in
very
soon
and
if
not,
unfortunately,
I'll
have
to
like
turn
off
turn
it
off
in
our
systems.
Go
back
to
the
teaching
the
course
normally,
but
then
I
won't
be
able
to
provide
any
feedback
on
on
this,
because
I'll
have
to
turn
it
off.
G
H
Yeah
I
think
the
most
useful
for
us
would
be
to
have
access
to
this
y
style
file,
because
this
is
really
how
we
can
reproduce
the
the
issue
on
our
side,
if
it's
possible
to
for
you
to
have
access
to
it.
I
know
that
you
had
issues
with
that,
because
it's
it's
erased
after
a
session.
I
think.
G
Yeah
at
this
point,
I
I,
don't
know
I
mean,
unfortunately,
the
y-score
files
of
when
this
happened
are
law
are
gone
because
they
were,
they
were
already
deleted
right.
We
went
and
they
were
in
sessions
that
happened
a
few
days
ago,
and
so
they
don't
exist
anymore.
We
put
the
Y
store
on
a
temporary
file
system.
That's
on
the
Node,
because
putting
the
white
store
sqlite
database
on
the
NFS
home
directory
was
just
killing
performance
and
Brian
suggested
that
we
move
it
to
a
temporary
temporary
SSD
based
local
local
disk.
G
So
we
did
that
which
means
those
files
were
gone.
We
could,
of
course,
add
a
I
look.
I
talked
to
Ryan
Lovitt
about
adding
shutdown,
a
script.existent
shutdown
that
copies
it
back
to
the
home
directory
and
timestamps
it
so
that
we
keep
them.
But
obviously,
at
this
point
those
would
be
new
files.
The
problem
micro
might
not
happen,
and
the
issue
is
I.
G
G
It's
documents
for
breakfast
as
soon
as
you
turn
it
on
and
so,
and
so
I
could
try
to
do
that
and
engineer
like
the
copying
the
white
store
file
back,
but
it
won't
be
any
different
from
what
anyone
else
who
actually
tests
this
could
could
do,
because
it
just
it
seems
to
happen
pretty
quickly.
H
The
other
thing
is
that
we
don't
see
that
on
our
side,
it
works
pretty
well,
we
actually
have
tests
also
in
in
the
CI
of
Jupiter
lab
foreign.
H
I
G
Yeah
I,
don't
remember
what
the
what
the
idle
calling
parameters
are
doesn't
shut
down
immediately,
but
it
does
shut
down
eventually,
but
it's
not
immediate
I,
don't
remember!
We
can
try
to
find
those
parameters.
I
can
give
you
I
can
show
you
where
the
configuration
of
that
of
that
is
and
give
me
a
second
I'll.
I'll
tell
you
because
this
is
all
done
with
public
infrastructure.
G
So
this
is
the
folder
that
has
all
of
the
kubernetes
files,
the
environment
files
that
build
the
image
everything
and
in
one
of
those
yaml
files,
those
parameters
for
starting
to
prohibber
somewhere
in
there.
But
I,
don't
know
I,
don't
remember
what
the
what
the
idle
shutdown
timers
are.
It's
probably
on
the
order
of
half
an
hour,
maybe
an
hour,
it's
my
guess,
but
but
I
haven't
looked
I
can
ask
I
can
find
out
if
that's
important.
I
Yeah,
because
so
the
thing
is,
there
is
an
issue
that
we
know
about.
Basically,
if
the
user
keeps
jupyter
lab
open
and
the
document
open
in
in
in
its
browser,
it
closes
the
leap
of
the
laptop
and
go
somewhere
else.
If
the
in
the
room
for
that
document
in
the
back
end
closes
if
somehow
the
wire
store
gets
removed.
I
That
means
next
time
that
the
user
reconnects
it
will
have
the
the
the
data
loaded
into
the
browser
in
memory.
But
we
don't
have
that
data
loaded
in
in
memory
in
the
bucket,
and
that
creates
a
duplication
of
the
data
from
that,
because
we
don't
discard
that
data
from
the
front
end.
That's
why
we
were
discussing
about
adding
document
sessions,
and
that
could
happen.
For
example,
if
the
server
session
shuts
down
you,
you
were
saying
that
you
remove
as
soon
as
they
said
that
shuts
down
you,
remove
the
wire
store.
G
I
G
Because
the
Y
store
right
now
is
in
slash
TMP,
it's
a
temporary
file,
which
means
it's
on
the
file
system
of
the
node
of
the
docker
image,
and
so
when
the
docker
image
goes
down,
that
file
disappears
with
it
and
then,
when
the.
If
the
user,
reconnects
they're
gonna
get
a
new
server,
but
they
get
a
brand
new
session
when
they
get
a
new
server.
So
the
whole
thing
restarts
all
over
again
right.
H
Because
the
the
issues
that
the
documents
are
not
going
to
match,
if
there
is
not
a
corresponding
y
star,
a
white
star
corresponding
to
to
the
document
that
is
present
in
the
memory
of
the
front
end,
it's
a
bit
technical
but
yeah.
The
real
solution,
as
we
said,
is
just
to
implement
document
sessions.
I
So,
basically,
if
once
the
the
new
server
start,
if
the
user
needs
to
reload
the
web
page
Jupiter
lab
does
not
the
the
issue
is
not
because
of
the
document
sessions.
If
the
user
doesn't
need
to
reload
the
Jupiter
lab
that
it
sees
the
document
that
it
just
reconnects
to
the
new
server,
then
it's
because
of
document
sessions.
I
And
then
it
probably
is
not
because
of
document
sessions.
We
will
need
to
look
into
that.
G
G
Right
so
it
seems
like
this
is
more
complicated
than
than
unfortunately,
then
I
can
like
afford
to
handle
with
live,
live
class
of
students,
so
I'm
going
to
turn
it
off
in
our
courses
and
I
want
to
recommend
that
we
turn
it
off
here
at
Berkeley
and,
unfortunately,
I
won't
be
able
to
provide
much
feedback,
because
I
I
could
provide
feedback
kind
of
within
the
bounds
of
actually
teaching
the
course,
but
if
I
have
to
but
I
don't
have
the
bandwidth
and
I
and
I
can't
have
the
students
be
tested.
G
So
I'll
do
my
best
to
provide
feedback
on
and
test
on
my
own,
but
but
I
will
turn
it
off
because
it
doesn't
look
like
this
is
an
easy
one
and
I
really
appreciate
people's
time
and
input
in
this
I
know
this.
This
is
hard.
So
thanks.
Thanks
for
the
feedback.
G
Was
actually
going
to
suggest
that
I
was
a
good
good
call,
William
I
remember
when
I
logged
into
hack,
MD
I
thought
we're.
Basically
at
the
point
of
having
that
and
and
one
of
my
concerns
is
that
I
appreciate
that
the
team
is
testing
a
lot
of
a
lot
of
this
and
that
you
guys
are
not
seeing
these
problems
but
I
wonder
how
many
of
you
use
Jupiter
lab
as
you're
like
working
system
rather
than
developing
it?
G
Are
you
all
day
logged
in
creating
notebooks
and
writing
that,
probably
not
because
your
jobs
are
different
right,
I'm,
not
I'm,
not
blaming
anyone
here,
I'm,
simply
saying
that
your
jobs
are
different
and
probably
are
more
focused
on
doing
development
than
using
the
system.
I
see
these
problems
when
I'm
using
it
not
when
I'm
trying
to
like
add
features
to
it,
which,
obviously,
at
this
point
I
I,
don't
I.
G
Unfortunately,
don't
have
the
bandwidth
for
so,
but
using
it
for
the
team,
as
the
RTC
engine
I
mean
the
hack,
and
the
experience
is
actually
at
this
point
pretty
much.
Basically,
Works
and
I
mean
you
can
have
exactly
kind
of
the
split
pane
render
experience
kind
of
works,
and
so,
if
we,
if
we
hosted
a
an
instance,
that's
persistent
that
would
be
that
would
be
nice
and
something
where
we
could
iterate
quickly
test
it
with
patches
test.
G
Even
like
have
a
meeting
with
a
copy,
with
a
version
of
a
specific
version
that
might
be.
That
might
be
a
bit
a.
A
J
It
but
I
would
say
we
should
use
it,
because
it's
good
to
dog
food,
that
sort
of
thing.