►
From YouTube: JupyterLab Team Meeting - October 26, 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/152
A
B
Hey
just
wanted
to
thank
Jeremy
for
getting
the
3.5
release
out
over
the
weekend
after
the
releaser
did
not
work
for
me,
I
was
able
to
get
clarification
from
him
and
Steve
that
the
releaser
just
recently
updated
to
version
two,
so
that
requires
changes
on
the
Jupiter
lab
side
to
support
the
version
two
of
the
releaser.
So
until
those
changes
are
in,
you
can
just
use
the
1.x
branch
of
the
releaser
to
do
releases.
That's
how
Jeremy
got
the
release
out
for
people
who
are
dealing
with
that.
B
So
that's
my
update
since
I
originally
said
last
meeting
that
I
would
take
care
of
that.
So
Jeremy
saved
my
butt
after
two
days
of
thrashing
against
it.
C
Okay,
I
opened
a
proposal
to
what
fragment
identification
syntax
to
Jupiter.
The
idea
is
that
we
cannot
currently
points
to
a
cell
or
an
output
so
that
The
Notebook
will
scroll
upon
opening
or
turn
that
on
course
allowing
to
scroll
to
specific
fragments
of
The
Notebook,
and
there
is
a
standardized
method
of
describing
how
the
uncores
work,
which
is
called
fragment
identification
syntax.
C
This
is
the
part
of
the
URI
which
can
be
arbitrarily
interpreted
by
the
browser
and
it's
it
was
standardized
for
txt
files
for
CSV
files
for
PDFs
so
and
how
we
go
with.
It
is
a
little
bit
of
a
product-wide
decision,
but
in
the
meantime,
I
opened
a
proposal
on
adding
cell
ID
requirements
so
that
we
can
Target
itself
based
on
its
ID
and
created
a
pull
request,
providing
a
reference.
Implementation
for
Jupiter
lab
I
just
wanted
to
put
that
on
your
radar
in
case.
C
A
Yes,
the
Japs
are
one
of
the
main
things
that
the
new
software
steering
council
is
going
to
be
doing
is
talking
through
so
yeah.
That
is
one
that
is
probably
the
best
Avenue
if
you're
suggesting
something
that
is
Project
wide,
but
if
you're
doing
an
implementation,
that
is,
that
isn't
going
to
cause
Downstream
changes
to
other
clients,
and
it's
just
like
a
feature.
You
don't
have
to
do
a
job,
yeah
I,
don't
I,
don't
think
I
understand
fully
what
the
ramifications
are.
C
The
ramifications
are
that
if
we
decide
to
go
full
set
of
syntax
right
now,
it
will
be
difficult
to
change
that
later
and
other
clients
might
want
to
adapt
the
same
syntax
so
that
you
can
get
you
can
scroll
to
the
notebook
using
the
same
kind
of
URL,
whether
you
are
using
Jupiter
lab
notebook,
static,
HTML
exports
or
something
else.
But
yes,
that's
an
additive
change
which
I
think
that
this
pull
request
could
be
reviewed
and
and
potentially
merged
into
four
points:
zero
or
any
release.
C
Yeah.
Please
leave
a
comment
if
you
have
an
opinion
on
whether
this
is
the
right
thing
to
do,
and-
and
my
other
point
is-
that
I'm
working
on
creating
an
extension
for
benchmarking
of
user
interface
in
Jupiter
lab
and
that's
for
developers
only.
But
if
you
happen
to
have
like
a
user
who
stumbles
upon
a
performance
issue,
you
can,
you
would
be
able
to
say
them.
C
I
tell
them
that
you
can
install
this
extension
and
click
Start
and
some
benchmarks
will
run
and
they
can
report
back
to
you
with
a
Json
file
which
contains
information
on
impact
of
different
styles
and
potentially
in
future.
On
the
JavaScript
cell,
profiling
and
you
can
analyze
that,
so
that's
excluding
any
online
reporting
that
would
be
problematic
when
it
comes
to
privacy.
So
there
was
just
there
wouldn't
be
an
how
to
upload
it's
just
generating
a
static
Json
that
you
can
use.
E
Just
on
your
first
proposal
for
the
the
fragmented
identification,
do
you
do
you
want
to
move
forward
with
Jupiter
lab
before
waiting
for
or
what's
what's
your.
C
A
I
think
that's
a
really
good
idea
and
in
case
the
implication
was
missed
by
other
people
on
the
call.
This
is
something
that
goes
into
the
URL
right
and
URLs
or
public
apis,
and
we
have
thought
pretty
hard
about
what
things
we
put
in
URLs
in
general.
So
if
you
have
thoughts
in
that
sphere,
take
a
look
at
the
issue
because
our
the
The
Proposal,
rather
the
pr
because
it's
these
things
do
become
sticky
right
now,
because
we
have
some
time
before
four
and
we
have
some
Alpha
and
beta
Cycles.
F
Also
I
can
host
another
week
because
my
meeting
chaos
I'm
sorry,
you
had
to
start
it
I'm
hosting.
Where
are
we
on
the
agenda?.
A
F
Okay,
great
hi,
sorry
I've
been
running
around
a
bit
this
morning.
That's
how
it
goes
for
me
on
the
West
Coast
time
zones.
F
Usually
so,
yes,
my
update
is
just
last
week
we
talked
just
a
little
bit
about
Jupiter
lab
Zoom
stuff
for
anyone
that
wasn't
here
for
various
reasons,
accessibility,
maybe
presentation
mode
Jupiter
lab-
could
have
a
little
bit
more
intentional
behavior
for
what
happens
when
you
get
to
high
Zoom
percentages,
High
I
mean
like
200
400
percent,
500
percent,
which
is
supported
by
several
browsers
at
least
Chrome
and
Firefox
right
now
it
just
kind
of
all
squishes.
F
Together
we
have
various
issues
that
are
kind
of
documented
in
this,
but
yeah
I
was
hoping
to
discuss
this
a
little
more
I.
Don't
know
if
we're
already
at
the
additional
discussion
section,
though,
and
I
think
this
could
easily
be
more
than
a
five
minute
conversation
I
will
post
the
link,
though,
because
you
can
also
review
it
asynchronously
outside
this
meeting.
If
you
wish
I
believe
Mike
did
thank
you
for
that.
Unless
you
want
the
full
discussion
now,
but
I'll
be
quiet
because
I'm
caffeinated
I'm
just
talking
fast,
probably.
B
F
G
So
I
got
a
a
contact
from
Prashant
who's,
a
PM
in
my
organization.
He
pointed
out
there's
an
issue
which
I
linked
in
the
in
the
notes
Here
with
plotly
and
the
latest
version
of
Jupiter
lab.
G
It
says
that
figure
widget
is
not
working
with
ipy
widgets
8.0.0
RC
0,
which
I
was
trying
to
do
a
little
research
on
this
and
it
seems
like
Jupiter
lab,
should
work
with
ipy
widgets,
but
I
wanted
to
pull
the
group
because
he
contacted
me
just
yesterday
yesterday
afternoon
has
anyone
else
seen
issues
with
IPI
widgets
compatibility?
G
Is
there
a
newer,
iPod
widgets?
We
should
be
using
any
guidance
on
this.
A
8.0.0.0.2
right
I,
imagine
that
might
have
something
to
do
with
it,
but
Json
growls
on
the
call-
and
he
knows
I.
H
Can
see
it
in
his
eyes,
yeah
I
think
the
question
is:
does
plotly
support
yeah
I?
Probably
would
you
say
and
I
don't
know
the
answer
that
off
the
top
of
my
head,
but
maybe
it
doesn't
and
that's
the
that
would
be
the
most
obvious
thing
to
look
at
from
to
me.
E
G
E
This
there
is
an
open
PR
to
open
a
week
ago.
G
G
I
guess
I
missed
that
and
so
did
Prashant.
My
teammate
okay
looks
like
there's
still
some
active
discussion
on
this.
H
H
Okay,
there
is
an
iPad
widgets,
8
migration
guide
in
the
iPad
widgets,
a
change
log
documentation
that
might
be
helpful
here.
I
haven't
looked
at
this
issue.
G
H
I
I
just
want
to
advertise
that
python
311
had
this
really
cool
release
party
live
stream
on
Monday
and
I,
don't
know
just
pointing
it
out,
you
might
want
to
check
out
the
YouTube
recording
of
it,
but
basically
what
they
did
was
intersperse
the
release
manager
doing
key
steps
of
the
release
live
with
people
who
had
developed
key
new
features
in
Python
311
talking
about
those
new
features
so
that
the
talks
themselves
are
extremely
interesting
and
worth
looking
at
and
fairly
short
and
I'm,
mainly
mentioning
here,
because
people
here
are
interested
in
Python,
but
also
it's
actually
a
really
neat
seeing
you
know
somebody
do
a
live
stream
release
so
hint
into
whoever
does
the
next
or
any
Jupiter
project
releases?
J
So
fun
story
about
that
William.
We
actually
did
the
Jupiter
lab
1.0
release
that
way
cool
and
it
was
a
comedy
of
errors
like
we
were
having
issues
with
like
access
to
published
things,
and
we
well
like
we're
jokingly,
playing
The,
Final
Countdown
song
over
and
over
again.
Every
time
we
try
awesome,
but
it
was
a
good
time.
Yeah.
I
The
the
python
one
seemed
pretty
smooth
I
think
they're
pretty
practiced.
At
this
point
it
was
a
really
complic.
It
was
weird
because
there
are
questions
like
how
much
change
from
the
last
release
candidate
until
the
actual
released
version
so
interesting
discussion
during
the
live
stream
about
that,
and
it
was
well
over
a
hundred
commits
between
the
rc2
and
the
released
version
which
is
really
unusual
to
you
know.
I
H
I
F
A
H
Well,
I
would
just
like
to
say
thanks
to
Steve's
work.
Hopefully
it
would
go
much
smoother
this
next
time,
yeah.
F
Yeah
I
definitely
they're
very
practiced
at
this
process.
I,
don't
remember
how
many
release
streams,
they've
done,
but
I
think
I've
seen
at
least
two
other
ones,
so
they
definitely
know
what
they're
doing,
but
always
nice
to
hear
good
feedback
and
weird
idea,
but
well
I,
don't
know
if
they'll
ever
align
like
I
would
be
interested
too
if
anyone's
like.
Oh,
this
should
be
like
a
community
call
thing.
I,
like
the
idea
of
the
recording,
is
documentation
for
how
you
fix
some
of
those
things
too.
So
I
don't
know
weird
idea.
J
Oh
yeah
I
got
like
20
minutes
so
first
thing:
that's
not
on
the
agenda
because
of
security
related,
and
this
will
come
out
by
the
time
the
recording
is
published
but
we'll
be
releasing
a
cve
today
to
basically
stop
using
files
in
the
current
directory
as
part
of
config
python
files,
because
you
might
not
control
the
permissions
of
the
folder
you're
running
from
the
directory
running
from
so
someone
could
put
in
a
malicious
file,
and
that
would
give
you
a
bad
time.
J
So
IPython
had
already
done
this.
We
just
didn't
really
notice
and
coordinate,
but
now
Jupiter
is
following
the
suit,
so
this
will
be
in
Jupiter
core.
So
if
any
other
apps
are
like
overriding,
the
behavior
Jupiter
core
and
like
traitless
is
not
doing
that.
Jupiter
server
is
not
doing
that,
but
it's
worth
checking.
J
You,
if
you
own
a
cheaper
application
to
make
sure
you're
not
doing
that
and
inadvertently
re-exposing
the
cve
but
yeah,
so
that
should
be
published
today.
There's
already
a
published
fix
that
came
out
a
week
ago,
today,
Jupiter
core
and
so
practically
what
that
means
is
like.
We
had
some
binders
that
were
just
throwing
a
config
file
in
in
the
the
working
directory
and
and
just
assuming
that
would
work,
but
you
have
to
instead
patch
dash
dash
config.
J
Then
I
think
I
had
to
update
one
test
and
MB
convert
that.
K
Was
making
a
similar
assumption
and
pass
dash
dash
config
in
that
test?.
J
Could
you
do
to
file.
A
J
To
answer
Mike's
question:
let's
answer
your
question:
Darren,
it
is
a
file
name,
there
might
be
a
configur
come
in.
Oh
yeah,
yeah
Jason's
confirming
that.
But
you
know
it
so
Mike
to
answer
your
question:
it's
it's.
We
don't
look
at
the
current
working
directory
at
all,
so
that
includes
Json
files,
because
what
you're,
what
you're?
J
Putting
in
that
Json
file,
it
could
be
a
an
import
pass
to
us
to
a
malicious
path
of
disk
because,
like
some
of
these
configurables
are
are
class
paths
or
whatever.
So
it's
still
an
attack
vector
to
load
config
from
the
current
work
directory,
even
if
it's
Json,
because
it
gets
interpreters
python.
H
J
J
All
right,
so
next
is
Jupiter
releaser.
So,
as
Alex
mentioned,
we're
transitioning
from
V1
to
V2
and
I.
Think
Jupiter
lab
is
the
only
one
that's
a
bit
lagging.
Is
there
it's
doing
a
lot
as
part
of
the
release
process?
That
was
like
a
hard
case
for
early
sir.
So
there's
a
couple
minor
things
that
need
to
be
updated
in
Jupiter
lab
itself
to
accommodate
it,
but
once
that's
done,
it's
I
think
a
lot
better.
J
Even
if
you
don't
use
the
release
from
repo
workflow
that
that
V2
was
intended
for,
but
we
are
using
the
release
from
work
workflow
across
three
different
GitHub
orgs.
Now
as
test
cases
using
bot
accounts
for
email,
GitHub,
Pi,
Pi
and
npm
with
appropriately
scope
permission
tokens.
J
So
our
test
case
in
the
jupyter
lab
or
it
was
Jupiter
lab
server,
made
a
release
of
that
using
our
our
credentials
from
our
our
Jupiter
lab
bot
account
that
has
now
admin
access
to
that
repo,
but
checks
to
see
who's
running
the
the
workflow
to
make
sure
they
are
admin
as
well.
The
commit
comes
in
as
the
that
author's
email
address,
but
the
release
and
the
release
assets
are
made
using
that
Bots
credentials.
J
J
Go
so
there's
there's
a
set
of
bots
for
each
GitHub
work
right
now.
Three
sets
of
them
that
have
appropriately
two-factor
authorization.
K
Set
up
and
all
that
and
the
credentials
are
in
a
shared
for
the
Jupiter
admins
they're
in
a
shared
location.
For
that.
J
In
a
vault
so
yeah
this
is
still
sort
of
experimental.
So
if
anyone
has
any
feedback
on
that,
but
yeah
I
think
taking.
J
And
then,
hopefully,
once
Jupiter
lab
gets
those
couple
small
things
changed.
It
could
also
use
the
workflow
of
releasing
from
Jupiter
Labs.
So
you
don't
need
like
that.
As
long
as
you're,
an
admin
on
that
repo,
you
can
make
releases
now
and
you
don't
need
to
have
your
fork
of
Jupiter
releaser,
set
up
with
this
Pi
Pi
token
map
that
you
have
to
babysit.
So
the
the
that.
J
Forward
and
we'll
culminate
Williams
4.0
release
parties.
H
Because
that
live
stream
itself,
yeah
yeah
I,
would
love
to
use
this
for
iPay
widgets
and
you
know
obviously
there's
other
Jupiter
repos.
It's
been
a
bit
of
a
moving
Target
for
the
release,
Bots,
which
has
been
wonderful
because
there's
so
many
things
that
you've
been
adding
and
updating
and
and
improving
and
policing
do
you
have
an
idea
of
when
you
think
you
would
call
it
as
a
I?
Don't
know
1.0
release
like
you
would
say:
okay,
this
is
ready
for
other
Jupiter
projects
to
use.
J
Yeah,
so
it's
it's
publishing
the
actions
are
under
V2
tag
and
we're
the
the
python
package
is
so
called
Alpha
to
signify.
This
is
we're
not
finished
with
the
B2,
yet
I
think
once
Jupiter
lab
is
using
it
successfully.
I've
used
it
across
all
the
other
repos
that
I
release
on
so
it's
it's
done,
except
for
Jupiter,
lab
I,
would
say,
and
then
and
then
I
would
say
we
can
lock
it
down
and.
K
Stop
making
breaking
changes
for
version
two
I
move
forward.
B
Out
of
curiosity,
have
we
updated
the
cookie
cutter
because
the
cookie
cutter
instructs
the
user
once
they
open
the
cookie
cutter
to
use
the
releaser
to
release
cookie
cutted
extensions?
So.
J
J
Yeah
they're,
probably
still
using
the
B1
tag,
but
you
as
the
releaser
on
your
fork,
we're
using
the
V2
stuff.
Basically,
so
it
was
kind
of
in
a
mixed
state
that
that
was
hard
to
that's
another
another
way.
It
makes
more
sense
to
do
it
from
the
repos,
so
you're
kind
of
you're,
it's
more
stable
when
everything
is,
is
targeting
a
stable
version
tag,
as
opposed
to
having
a
fork
that
you're
updating
to
the
latest
commit
to.
J
J
That's
one
part
of
it.
Other
part
is
we're
using
Nest
async
IO,
which
is
completely
against
what
python
wanted
in
terms
of
having
re-entrant
event
Loops.
They
don't
support
it
explicitly.
Actually
Guido
had
mentioned
as
part
of
this
thread
recently
that
they
might
still
consider
going
forward.
It's
just
they.
They
thought
it
was
impossible,
but
he
has
an
idea.
It
might
be
possible
so
that
that
we
still
might
get
re-entrant
event
Loops
in
Python
eventually,
but
they
don't
exist.
J
So
the
the
break
and
change
for
us
was
that
get
event.
Loop
the
top
level
async
io.cad
event,
Loop
used
to
just
create
one
and
you're
ready
to
go.
If
there
wasn't
one
there
and
tornado
was
relying
on
that,
zmq
was
relying
on
and
therefore
we
were
in
a
lot
of
ways,
but
now
that
raises
deprecation
warning
because
they
intended
to
make
that
an
alias
for
get
running
Loop.
So
it
would
fail
if
there
was
not
an
actually
installed
and
running
Loop
event
Loop.
J
So
we
had
to
make
some
changes
at
the
starting
mid-made
centuries
of
cmq.
We've
got
some
changes
coming
in
Jupiter
client
8.0
to
accommodate
that
working
on
changes
in
Jupiter
server,
trying
to
work
that
up
the
stack
to
where
we're
in
in
alignment
with
what
or
python
is
doing,
and
also
trying
to
get
ourselves
off
of
necessary.
J
Sync
IO,
because
we've
seen
bugs
related
to
that
in
a
while,
where
like,
if
you
start
an
event,
Loop
and
then
try
to
Nest
things
to
go,
apply
it
after
the
fact
you
can
get
into
there's
like
a.
K
J
The
way
we're
handling
event
Loops
the
one
concession
that
they
might
make,
there's
an
open
PR
against
C
python.
3.12
right
now
is
that
they
might
make
get
event.
Loop
return
a
an
installed,
but
not
running
Loop.
So
it
would
not
be
an
alias
for
get
running
Loop,
but
it
would.
It
would
also
not
install
one
for
you,
so
it
would
fail
if
there's
not
one
there,
but
if
you
had
installed
a
loop,
you
set
event:
Loop
ran
it
to
completion
and
it
was
still
sitting
there
again
event.
J
Loop
would
return
that
one.
So
you
have
a
running
Loop,
that's
in
a
state
of
not
running
and
not
closed
that
you
could
run
again
on
that
would
clean
up
some
of
the
logic
we
currently
have
in
client
to
like
work
around
that
fact.
Where
we're
like
bookkeeping
the
the
loops
that
we've
created,
we
could
get
rid
of
that
logic,
but
that
would
be
a
nice
to
have.
F
Thanks
for
all
those
updates,
I
always
appreciate
that,
because
I
don't
know
how
I
keep
up
with
half
the
things
you
do
to.
F
Them
so
thank
you
so
much
Steve.
Anyone
else
have
things
to
say:
silence.
F
E
Yeah,
so
at
constack
we
have
been
working
art
of
bringing
a
better
RTC,
two
three,
the
three
branches
so
first
to
four
and
then
back
parting
as
much
as
possible
to
three
and
I
would
like
everybody
to
try,
because
we
have
a
binder
to
be
able
to
use
it.
So
I
will
it's
the
link,
that's
in
the
in
the
notes,
but
I
just
passed
the
link
into
the
the
to
the
chat
too.
E
So
just
open
it
and
write
things
and
edit
things
or
add
new
things,
and
just
like
break
things
just
to
see
it's
probably
gonna
break
like
it's
the
first
first
time
we
are
trying
it
with
so
many
people,
so
the
the
ID,
it's
it's
gonna
break
API
on
the
RTC
front
compared
to
3.5,
but
those
API
are
tagged
as
and
like
Alpha
features.
E
So
it's
fine
and
the
the
idea
is
also
to
still
keep
the
the
flag
so
that
will
the
the
RTC
stuff
will
be
used
only
if
the
flag
for
collaboration
will
be
set
up.
There
are
some
discussion
that
are
going
on
because
we
would
need
to
to
be
able
to
plug
also
the
optionally
to
put
the
the
new
file
ID
system
that
has
been
pushed
by
people,
a
colleague
of
Jason
mobile
on
this
Jupiter
server,
so
yeah.
E
Oh
I'm,
not
the
only
one.
There
is
lots
of
people
like
Carlos
David
that
are
also
connected.
They've
done
tremendous
work
on
that,
so
yeah,
yeah
and
I
think
yeah.
The
other
point
for
me
is
like
there
is
the
performance
meeting
just
after
this
one,
so
that's
been
in
20
minutes
for
people
that
are
interested
and
yeah
I
think
that's
it,
and
maybe
we
can
try
to
do
the
the
note
in
the
corporation
corroborative
Jupiter
lab.
Then
next
week.
F
F
E
F
E
D
E
E
D
E
Oh
then,
you
yeah
you,
you
eat
something:
okay,
but
yeah,
because
nobody
tried
to
rename
like
there
there's
gonna
be.
So
that's
why
I
was
speaking
about
the
fire
ID
service
that
we
want
to
be
able
to
plug
in
also
it's
it's
because
there
is
a
a
huge
constraint
in
case
of
somebody
rename
the
file,
especially
if
it's
done
externally
of
the
Jupiter
lab
environment,
but
so
it
seems
It's
not
related
to
that
this
time.
So,
okay.
E
Yeah,
your
right
is
not
saving
good
using
the
same
mechanism,
because
everything
is
done
automatically
in
the
back
end
by
time
to
time
is,
is
saving
on
the
on
the
art
on
the
file
on
system.
F
William
and
Vidar,
and
maybe
Mike,
can
I
copy
your
comments
into
the
notes.
So
we
have
them
for
later
saying.
Yes,
for
million.
E
F
Yeah,
no
thank
you
for
bringing
binder
links
like
that.
I
know.
For
me,
it
makes
it
much
easier
to
try
than
than
having
to
get
it
all
set
up
on
my
own.
So
thank
you
so
much
for
me
cool.
Where
are
we?
Where
are
we
and
performance
meaning
go
to
that?
The
performance
people
are
very
nice
Corey.
You
are
next
on.
The
agenda
am
I,
seeing
people
there
we
go.
L
Hello,
everybody,
hey
so
I'm
from
GitHub
I've,
been
working
with
Vidar,
we're
working
on
setting
up
MB
dime
as
a
rich
stiff
view
on
github.com,
but
always
been
lovely
to
work
with
and
very
helpful,
and
we're
going
to
start
preparing
to
do
like
a
beta
test
with
a
larger
group
of
people.
But
we
were
interested
to
see
if
anybody
here
wanted
to
try
it
out
and
provide
any
feedback.
L
Kick
the
tires
see
what
we
got
wrong,
because
there's
probably
things
and
we've
been
working
on
upstreaming
some
stuff
to
the
nbdyne
repo.
So
you
know,
hopefully
anything
that
we
find
we
can
can
Upstream
as
well
and
try
to
make
that
project
better,
but
I
kind
of
threw
together
a
really
I'm
a
programmer,
not
like
a
marketing
person
or
something
so.
H
L
Together,
like
a
really
ugly,
terrible,
Google
doc,
you
can
just
like
Drop
in
a
GitHub,
username
and
then
email
into
and
then,
if
you
do
that
here,
I'm
going
to
share
my
screen
real
quick,
let
me
see,
let
me
see
which
of
these
makes
sense
to
share.
L
L
F
L
Yeah
someday,
if
I
ever
meet
any
of
you,
we
can
talk
politics.
What's.
B
D
D
L
And
me
and
Vadar
have
been
you
know,
speaking
there's
some
there's
some
things
we'd
love
to
do
to
it.
But
again
it
all
comes
down
to
internal
politics.
We'll
just
see
what
happens.
F
Yeah,
oh
sorry,
go
on.
H
I
was
just
going
to
ask
about
performance
because
I
realize
you're
running
at
a
scale
that
you
know
boggles
everyone's
Minds.
Were
there
any
performance
issues
at
the
scales
we
were
looking
at
so.
L
This
is
interesting
I.
Actually,
just
this
year,
I
rebuilt
the
GitHub
notebook
server,
because
it
failed
about
60.
Of
the
time
was
our
internal
stats
and
to
make
NB
convert
actually
work
at
scale.
We
had
to
change
some
things
about
like
how
validation,
Works
and
we're
looking
at
some
of
the
things
around
syntax,
highlighting
because
syntax
highlighting
an
NB
converters
are
really
slow,
so
that
was
some
of
the
performance
things
that
we
looked
at
over
there.
L
So
we
we
actually
pushed
up
a
change
to
NB
convert
to
do
what
we
call
like
optimistic
validation,
because
NB
convert
validates
after
every
single
filter
runs,
which
can
add
up
to
70
of
the
total
load
time
to
that
tool.
So
we
just
ran
it
at
the
end
because,
like
why
validate
after
every
filter,
we
don't
need
that
and
that
really
helped
and
then
another
thing
we
have
to
deal
a
lot
with.
L
Is
people
who
push
up
invalid
notebooks
like
you
could
have
an
invalid
head
or
an
invalid
base,
and
and
there's
lots
of
different
states
of
invalid
that
we
get
so.
L
We've
had
to
do
a
lot
of
work
around
telling
users
like
your
notebook
is
invalid,
which
notebook
is
invalid
and
what's
invalid
about
it,
which
we've
been
leading
on
MB
format
for
that
tooling
and
then
performance
stuff
on
mbdime,
The,
Brute,
Force
algorithm
isn't
the
fastest,
but
you
know
there's
other
things
in
the
system
that
are
that
are
slower
on
just
in
getting
things
off
of
git
and
stuff
like
that.
L
One
thing
that
we
did
push
out
PR
for
last
week
that
that
I
built
was
that
the
current
version
of
mbdime
will
render
every
single
cell
in
the
notebook,
even
if
it
didn't
change
so
we
we
changed
it
to
do
a
lazy
display
and
that
made
the
front
end
faster,
yes,
cell
ID
support
would
be
great,
like
Vidar
is
saying
another
thing
that
would
be
I.
L
Think
nice
in
in
like
looking
at
how
the
diff
algorithm
works
is
if,
if
the
cells
actually
had
a
time
stamp
of
when
they
were
changed,
then
you
could
just
diff
the
ones
that
had
changed
since
the
last
version,
which
would
make
that
easier.
L
L
We
just
weren't
given
the
allotted
time
to
build
it,
but
it
would
be
easier
to
to
do
that
if,
if
there
were
cell
IDs
or
some
sort
of
an
ID,
it's
a
lot
of
commenting
at
the
cell
level.
Otherwise
we're
going
to
have
to
do
some
sort
of
Json
Source
mapping
solution.
L
E
Yeah,
it's
just
a
not
to
put
pressure
or
just
an
interesting
question.
Is
there
any
plan
to
move
to
code
mirror
six
instead
of
core
mirrors,
5
or
maybe
you're
already
using
another
idea?
My.
D
D
F
Okay
yeah:
this
is
super,
exciting
and
also
just
I
hope,
you're
patting
yourself
on
the
back,
we're
getting
to
share
it
with
us
all.
In
my
experience,
that's
really
hard
work
to
get
this
out
here
publicly.
So
thank
you
and
congrats
Darian.
A
I
think
this
is
awesome
and
I
think
it'll
help
a
lot
of
people.
I
was
curious.
Do
you
have
when
you
deploy
features
like
this?
Do
you
have
accessibility
requirements
for
shipping
a
feature
and
will
this
meet
your
accessibility
requirements
and
if
it
doesn't,
can
you
help
us
help?
It
meet
your
accessibility
requirements,
so
it
can
help
meet
our
accessibility
requirements
too.
Yeah.
L
So
that's
a
real
problem:
it
does
not
come
anywhere
close
to
meeting
our
accessibility
requirements.
We
took
it
to
our
accessibility
team
and
they
were
just
like.
This
is
terrible.
L
The
only
thing
that
you
had
going
for
you
as
I
was
able
to
take
them
to
the
accessibility
page
and
say:
look
the
entire
ecosystem
is
inaccessible
and
they
said
well.
If
the
entire
ecosystem
is
inaccessible,
we
can't
ask
you
to
make
this
one
thing
inaccessible,
because
no
one
could
use
it
anyways.
L
A
Will
help
you
a
little
bit
right
code,
mirror
6
will
at
least
make
the
cells
not
not
black
boxes,
but
that's
only
a
little
bit
but
yeah
I
mean
if
this
is
something
that
you
plan
on
doing
more
work
on
than
you
know.
Please
Loop
Us
in
on
it,
because
we
want
to
make
we
want
to
make
the
exact
same
components
work
better
for
everyone.
L
Yeah,
no,
that
would
be
nice
I.
Don't
think
that
I,
don't
I
have
no
idea.
I
have
no
idea.
What's
on
the
horizon,
for
my
team,
so
I
don't
know.
I
do
I
do
know
that
that
was
like.
The
one
loophole
we
were
able
to
get
this
in
under
is
to
like
leave
the
Json
version,
so
that
someone
who
has
an
accessibility
need
can
can
use
the
version
of
the
Json,
because
our
Json
is
more
accessible
than
actually
using
jupyter
notebooks.
F
Yeah,
thank
you
for
asking
Darian
and
thanks
for
the
honest
feedback,
yeah.
G
F
F
Yeah
good
safe
answer
cool
since
we're
close
to
time.
This
is
really
awesome
and
I'm
sure
this
could
sustain
discussion,
but
I
want
to
try
and
get
to
the
rest
of
the
agenda.
So
please
follow
up.
I
know
Corey
provided
opportunity.
Thank
you
so
much
for
showing
this
Chris.
Are
you
ready
to
talk?
Yes,.
M
I
am
hi
everyone,
I'm
Chris,
a
product
manager
also
from
GitHub,
wanted
to.
Let
everyone
know
that
you
know
based
on
a
bunch
of
user
feedback,
we've.
H
M
Hearing
we
rolled
out
or
we
are
planning
to
roll
out
Jupiter
lab
support
in
code
spaces,
so
I
know
we're
short
on
time.
I'll
show
a
super
super
quick
demo,
keep
it
as
brief
as
possible.
Can
you
see
my
screen.
M
This
is
just
a
stable
diffusion
repository
that
I've
set
up.
I
have
a
code
space
in
here,
that's
already
running
so
now.
I
can
just
click
on
that
and
it'll
bring
me
into
Jupiter
lab
in
the
code.
Space
running
in
the
cloud
and
I'll
be
able
to,
you
know,
modify
and
do
all
of
the
get
operations.
So
in
this
case,
yeah
I
can
just
select
the
last
couple
of.
M
Run
those
selected
cells-
and
none
of
this
is
happening
on
my
machine-
it's
all
happening
in
the
cloud,
but
it's
a
also
accessed
directly
to
the
GitHub
repository.
M
Excited
about
this
and
are
planning
to
roll
it
out,
at
least
in
beta
shortly,
and
wanted
to
give
everyone
a
heads
up
that
that
was
on
the
way
in
terms
of
the
initial
beta
functionality.
I
showed
you
pretty
much
most
of
it.
You
can
access
Jupiter
lab
via
the
web
application
or
via
the
CLI.
We
do
also
we're
planning
to
enable
gpus
in
code
spaces,
we're
still
working
on
making
sure
that
we
have
the
appropriate
inventory
to
do
that
as.
M
But
right
now,
if
you
have
specific
needs,
you
can
reach
out
to
me
and
request
GPU
access
and
we
can
give
that
to
you
and
then.
Finally,
the
the
last
piece
of
this
that
we're
not
going
to
be
rolling
out
with
in
beta
but
we'll
be
adding,
hopefully
soon
after
that
is
remote,
kernel
execution,
so
being
able
to
do
that
directly
from
the
GitHub
website
and
from
Jupiter
notebooks
in
there
as.
M
Yeah,
we
are
super
excited
about
this
and
please
please.
We
really
want
feedback.
I
really
want
feedback
and
I
would
love
to
make
sure
that
we're
representing
it
properly
when
we
start
getting
to
you
know
the
blog
posts,
social
things
like
that,
so
please,
reach
out
to
me.
I,
would
love
to
coordinate
on
that
messaging
and
also
would
love
to
give
any
of
you
who
are
interested
access
to
start
playing
around
with
it.
F
F
Yeah
I
know
I'm
really
interested
in
this
one,
so
you'll
probably
be
hearing
from
me,
awesome,
yeah,
and
also
just
so
that
you're
you're
both
getting
it.
Thank
you
also
for
sharing
this
like
Jason,
said,
but
also
thank
you
for
whatever
I'm
sure
quite
invisible
work
that
it
took
to
get
it
to
share
with
us.
So
that's
a
lot.
Thank
you.
D
Is
this
something
we
would
have
to
up
to
ask
for
access
to,
or
was
this
available
for
everyone.
M
Yeah
great
question
repos
so
right
now
you
have
to
ask
for
access
to
it
when
we
roll
it
out
in
the
beta
which
we're
planning
to
do
in
the
coming
weeks.
I
know
it's
recorded,
so
it
can
be
just
in
an
update,
been
burned
on
that
before
in
the
in
the
coming
weeks
to
month,
we'll
be
planning
to
roll
out
the
initial
beta
of
this,
at
which
point
it
will
be
publicly
available
to
everybody.
M
F
Awesome
good
question:
if
I
can
I'm
going
to
transition
us
in
our
last
minute
to
Jason
grout,
you
have
some
stuff
here
too.
H
H
Perfect
excellent,
so
the
idea
here
is
something
that
Pete
brought
up
a
few
weeks
ago
from
a
demo.
He
did
a
few
years
ago.
We
pursued
this
last
week
in
the
widgets
Workshop
that
was
in
London
and
the
idea
is
we
want
to
make
it
possible
for
people
to
provide
Rich
outputs
without
installing
extensions
and
just
having
sort
of
the
feeling
from
notebook
6
of
being
able
to
just
work
on
the
browser
and
push
stuff
to
the
browser
straight
from
the
kernel.
H
So
here's
the
basic
idea,
we're
doing
a
stock
display
call
with
a
mind
bundle.
This
mine
bundle
has
two
mime
types.
One
mime
type
is
our
new
mime
type
that
we're
introducing,
and
this
new
mime
type
is
an
es6
module
and
the
CS6
module
has
exports.
One
function
called
render
and
the
render
function
takes
an
output
message
I.E.
H
This
message
that
you're
looking
at
right
here,
a
Dom
element
and
a
context,
argument
that
contains
a
bunch
of
apis,
that's
to
be
determined
and-
and
this
mind
bundle
also
has
one
other
mine
type,
just
some
data,
and
so
what
happens
is
when
I
execute
this?
H
It
goes
to
the
front
end,
it
takes
this
es6
module
and
it
runs
and
it
and
it
runs
the
ASX
module
Imports
this
es6
module
and
calls
the
render
function
with
this
message,
including
this
data
down
here
and
lets
this
es6
module,
actually
format
the
message
and
format
the
message
as
a
rich
display
on
the
front
end,
and
so
what
you
end
up
here,
it's
just
creating
a
HTML
table
in
case
you're
wondering
the
encode
here.
All
it's
doing
is
taking
this
string.
That
represents
an
es6
module
and
encoding
it
as
a
data
URL.
H
So
the
idea
here
is
we'd
like
to
write
a
renderer
for
this
mime
type
right
here.
That
provides
an
API
that
you
know.
If
you
export
a
render
function,
we
will
call
it
with
this
message
with
these
particular
you
know
elements
from
the
front
end.
That's
doing
this
render
and
again.
The
idea
here
is
any
package
would
be
able
to
include
with
itself
the
JavaScript
to
render
Rich
items
out
of
that
package,
and
it
could
work
across
many
front
ends.
Right
now.
You
know
a
rich
renderer,
it's
an
extension.
H
You
know
front-ends
or-
and
this
allows
a
minimal
API-
that's
based
on
the
browser
apis
in
order
to
do
rich,
outputs
and
many
different
front-ends
and,
and
it
avoids
the
problem
of
trying
to
keep
an
extension
up
to
date
with
what's
in
the
kernel,
which
is
an
issue
that
we've
always
had
in
iPad
widgets
by
just
having
everything
sort
of
in
the
kernel,
and
it
allows
you
to
just
really
easily
quickly
explore
without
having
to
write
an
extension
ways
of
rendering
your
my
data
in
a
rich
way.
H
So
the
at
actually
the
code
for
this
minimal
demo
is
super
short.
It's
using
the
modern
browser
apis.
Essentially,
this
is
all
the
magic
right
here.
We
just
import
that
string
as
a
data
URL
and
then
call
the
render
function
with
this
things
that
we've
said
we
would
call
it
with
so.
Our
big
question
is
Pete
and
I
would
like
to
work
on
this.
H
D
I
do
not
have
any
objections,
but
I
have
a
addendum
or
extra
information
that
that
context.
The
API
there
is
a
the
goal
of
to
be
that
is
to
be
super
stable
across
major
versions
and
to
be
kind
of
once.
The
symbol
has
been
defined
and
added
to
that,
you
can
never
change
the
signature
API,
because
it
would
need
to
be
supported
across
all
front
ends
and
all
of
them
will
be
optionally
present.
So
if
it's
not
there,
we
should
be
able
to
fall
back.
H
D
Where
most
of
the
interesting.
D
Will
happen
I
assume,
but
the
kind
of
minimal
version
Jason
presented
here,
I
I
think
should
be
pretty.
H
Yeah,
we
imagine
that
context
dictionary
might
have
facilities
for
creating
comms
to
a
kernel
to
be
able
to
interact
with
the
kernel,
maybe
a
function
that
would
tap
into
whatever
the
run
the
the
current
front-end
rendering
system
is
so
you
could
render
inside
of
yourself
a
mind,
bundle
and
there's
a
few
other
ideas.
We
have
for
sort
of
some
initial
minimal
things
that
are
would
be
provided
in
that
context.
Optionally,
okay,
so
I
see
a
raised
hand
and
a
couple
of
thumbs
up
for
moving
the
repo
into
Jupiter
lab
Mike.
H
Excellent
question
I
think
that's
a
question
for
how
the
front
end
renders
this
particular
mime
type.
You
know
clearly
it's
it's
better
if
it's
in
an
iframe
security
wise
at
the
same
time,
we
already
render
JavaScript
without
an
iframe,
so
my
guess
is
any
of
the
Enterprise
front-ends
like
collab
or
databricks
Etc
will
make
sure
that
they're
rendering
it
inside
of
a
knife
frame,
but
really
that's
up
to
the
front
end
and
right
now
we
we
adopted
the
same
model
as
Jupiter
lab
has,
which
was
hey.
We
just
run
JavaScript
in
the
current
context.
H
Okay,
I'll
move
the
repo
into
Jupiter
lab.
That's
a
sort
of
a
low
friction
thing
to
move
forward.
Pete
and
I
are
planning
on
working
on
this
and
as
it
as
it,
matures
we'll
take
the
question
of
whether
we
should
pull
it
into
core
thanks.