►
From YouTube: Jupyter Community Call - July 27, 2021
Description
Recording from the Jupyter Community Call in July 2021.
The notes from this call can be found here: https://jupyter.readthedocs.io/en/latest/community/community-call-notes/2021-july.html
Read more about these calls on Discourse: https://discourse.jupyter.org/t/jupyter-community-calls/668
A
Here
you
go
wonderful,
hello,
everyone
and
welcome
to
our
july
2021
jupiter
community
call
I'm
isabella,
I'm
gonna
be
your
host
today
and
I'm
super
excited
to
see
everyone
here.
We
have
a
lot
of
different
people
today.
I
think
too,
so
also
super
excited
to
see
that
welcome.
If
you
haven't
been
here
before,
and
of
course
thank
you
for
returning.
If
you
have
been
here,
this
is
a
call
where
we
try
and
celebrate
across
jupiter
communities.
A
Try
and
unite
projects,
share
cool
things
just
generally
get
excited
and
remember
what
we're
here
for
and
of
course,
coordinate
larger
community
discussions.
Two
things
I
want
to
remind
you,
though
this
is
being
recorded
as
the
ominous
voice,
just
told
us
whatever
that
means
for
you.
These
will
end
up
on
youtube
at
some
point
and
two
we
are
part
of
a
jupiter
event
right
now,
so
that
means
we
are
held
to
the
jupiter
code
of
conduct.
A
That
includes
me,
and
with
that
I'm
ready
to
go,
I
think
we
have
anything
on
the
short
report
items.
First,
we're
just
gonna
follow
the
agenda,
which
I
will
link
one
more
time
actually
so,
starting
with
quick
reports.
Oh
we
do
actually
have
one
awesome.
Is
that
michael?
Would
you
like
to
speak
on
that.
B
Yes,
hello,
my
name
is
michael
milligan
of
the
university
of
minnesota,
I'm
one
of
the
developers
on
batch,
spawner
and
rap
spawner,
this
announcement's
a
couple
months
old,
but
I
saw
that
it
hadn't
been
made.
Yet
we
did
get
out
new
releases
for
both
of
those
packages.
Batch
spawner
is
specifically
for
cluster
and
scheduler
users.
Rap
spawner
is
not
specific
to
clusters
and
scheduling,
but
it
is
used
by
much
of
the
same
community.
A
C
I'm
away
from
keyboard,
but
we're
planning
on
3.1
of
jupiter
lab
today
this
afternoon,
which
will
be
our
first
release
of
real-time
collab,
so
we'll
make
an
announcement
on
the
discourse
channel
later
today.
I'll
try
it
out
at
a
time
it's
available
from
pre-installed,
but
it'll
be
available
as
a
final
release.
Solution
nice.
D
Well,
I'll
throw
one
in
because
mike
is
here.
I
had
a
pull
request
accepted
to
the
jupiter,
oh
authenticator,
enabling
globus
groups
for
group
management
inside
of
authenticator.
So
if
anyone
out
there
is
using
globus
auth
for
their
user
log
into
jupiter
hub,
that
feature
is
now
there
and
works.
I
think
it
works
well.
A
F
Quick
update
from
the
binder
side
of
things,
so
I
think
few
months
ago
we
started
blocking
launches
from
public
clouds
because
of
there's
a
click
to
mining
abuse.
There
was
a
new
pr
merged
a
few
weeks
ago.
That
means
there
should
be
a
new
work
around
so
hopefully,
if
we
can
launch
fine
launching
it
on
my
binder
and
you're
safe
from
amazon
or
whatever
it
should
hopefully
work
now.
E
That's
super
awesome,
simon.
Thank
you.
We
have
somebody,
we
have
a
tester
in
like
critical
need
of
this.
So
thank
you.
Thank
you.
Thank
you
yeah.
Well,
because
if
you
get
given
to
try
it
out,
and
it
doesn't
work,
give
us
some
feedback
cool
isabella,
you
think
you
can
ping
thomas
and
see
if
it
works.
I.
A
Was
going
to
yeah,
I
think
that
was
like
two
weeks
ago.
You
fixed
that.
Thank
you
so
much.
I
was
watching
yeah.
E
A
But
yeah
it's.
I
know
it
was
a
difficult
problem,
so
I
really
appreciate
people
spending
time
on
that
because
yeah,
especially
for
accessibility
work,
we
have
a
lot
of
people
sharing
software
via
that.
So
thank
you.
G
I
got
a
I
got
a
quick
one,
so
we've
been
looking
to
get
jupiter
video
chat,
which
is
the
one
that
lets
you
get
jitsi
inside
of
your
lab.
Looking
to
get
a
new
release
of
that
out,
got
it
mostly
working.
It's
now
ready
to
go,
but
if
there's
anybody
that
uses
that
I'd,
certainly
like
your
feedback
on
either
the
binder
off
develop
or
on
the
docks,
I'll
drop,
some
links
in
the
in
the
hack
md,
but
the
things
that
it
does
it
it
works.
G
The
o60
will
work
with
jupiter
light,
which
we
had
a
release.
Last
week.
I
don't
know
it
was
kind
of
kind
of
low
proof
and
we'll
also
work
with
retro
lab.
It's
makes
so
many
tabs,
but
that's
what
retro
lab
does
so,
if
anybody's
interested
in
that,
especially
anybody
that
self
hosts.
You
know
I'm
very
interested
in
some
feedback
on
that
piece.
D
A
A
I'm
gonna
watch
the
time
a
little
to
keep
it
close
to
15
minutes
with
discussion.
So
that's
just
kind
of
your
forewarning
and
the
first
person
we
have,
though,
is
matthias
who
welcome.
I
don't
actually
think
I've
seen
you
on
a
community
call
in
a
while.
So
thanks
for
being
here.
H
Thank
you
very
much.
Let
me
know
if
you
don't
hear
me
well
because,
as
I've
said
before,
I'm
tethering,
so
I
probably
don't
have
a
really
good
connection
and
I
will
start
with
two
two
short
demos.
So
then
I
can
stop
sharing
my
screen
and
then
we
can.
We
can
discuss
so
I'm
going
to
show
you
two
things:
I've,
oh
no!
I
can't
share
because
what's
happening,
I
need
to
open
system
preference
and
restart
the
zoom.
H
Oh
gosh,
I
hate
zoom.
So
I'm
going
to
try
to
do
that.
While
I'm
I'm
talking
so
two
two
specific
things
with
respect
to,
especially
when
you
work
with
with
with
schedulers
one
is
often
when
you
have
your
notebook
and
a
scheduler,
and
you
want
to
do
quick
iteration.
H
Okay,
so
let
me
try
to
move
it
around,
which
basically
starts
instead
of
the
ipython
kernel
is
used
as
a
proxy,
and
so
you
can
issue
a
magic
command
that
will
restart
the
kernel
on
the
remote
compute
node
without
having
to
to
go
through
all
the
scheduling
process,
because
from
the
server
point
of
view
the
kernel
has
not
restarted,
and
so
it
looks
like
that.
You
can
define
a
variable
check
that
it
exists
like
normal,
you
can.
H
You
can
change
change
the
value
and
if
ever
you
issue
the
command
restart,
the
proxy
will
intercept
this
magic,
not
send
it
to
the
kernel.
If
you
actually
kill
the
kernel
and
start
the
kernel,
and
now
I
have,
I
should
have
a
fresh
musician,
and
this
also
allows
you
to
do
a
couple
of
things
like
you
can
send
other
magics
to
the
specific
proxy
that
will,
for
example,
start
to
display
all
the
messages
it
is
going
through
on
the
output
channel.
H
So
you
can
really
see
what's
happening
if
you
ever
have
to
to
debug
the
connection
issue
or
something
like
that,
and
so,
for
example,
if
you
do
one
plus
one,
you
are
all
the
messages
that
are
exchanged
between
your
your
notebook
and
and
the
camera
right
now.
H
It
only
works
for
ipy
kernel,
because
I
have
a
hardcoded
it,
but
if
you
have
other
kernels,
there
is
no
reason
for
it
not
to
work
for
other
kernels,
because
it's
just
a
simple
proxy
and
you
can
deactivate
the
debugging
after
and
go
back
to
the
normal
normal
way
of
executing
and
what
I
am
going
to
show
you
another
things
we
are
working
on,
especially
for
for
people
who
want
to
parametrize
kernel
and
want
to,
for
example,
choose
choose
resources.
H
H
So,
for
example,
in
your
kernel,
you
can
say
that
you
have
json
schema
that
describes
the
amount
of
memory
you
have
the
number
of
cpu.
You
have
and
whether
or
not
you
have
a
gpu
and
we
have
these
small
extensions
still
working
on
it
and
that
allow
you
to
either
directly
edit
all
the
kernel
spec
scan
aspect,
values
or
expose
some
of
them
using
json
schema
as
something
like
relatively
relatively
easy,
like
a
drop
down,
and
if
you
want
a
gpu
gpu
or
not.
H
You
can
quickly,
as
a
user
is
not
have
to
be
really
aware
of
all
the
details
of
how
canal
spec
works,
and
so
that's
about
it.
For,
for
the
demo
online
series,
one
more
demo,
the
same
thing
for
parameterization
works
in
the
terminal.
H
So
if
I
do,
I
do
python
main.py.
It
should
pick
up
some
canon
spec
and
I
can
re-parameterize
them.
Oh
gosh,
that's
super
slow
and
so
same
thing.
It
uses.
Let's
say
I
want
to
have
1253
cpu
and
then
I
want
to
have
a
lot
of
memory
and
it
will
hopefully
rewrite
the
kernel
spec.
With
the
amount
of
memory
I
asked
the
number
of
cpu
I
asked,
and
all
events
are
gonna
expect
name
now
have
the
values
I
have
I
haven't
heard,
so
we
are
trying
to
bring
that
more
into
core
jupiter.
H
So
you're,
muted,
I
see
you
speaking
but
you're
just
and
nick
says
text.
User
interface
is
savage.
Yes,
it's
with
prompt
toolkit
from
toolkit
have
good
things
to
do.
To
do
to
do
to
do
things
and
I'm
way
more
frequent
in
front
of
you
than
I
am
in
in
react
and
json
schema,
and
I
am
struggling
with
typescript
and
I
think
silver
sound
is
on,
because
I
heard
a
small
amount
of
noise
when
you
plug
something.
I
Can
you
hear
me
now?
Yes,
wonderful,
okay,
sorry
about
that,
and
this
is
so
much
aligned
with
stuff
that
we
sort
of
had
in
the
backlog
for
so
long
that
I
I
may
have
like
tons
of
questions,
and
so
one
is
like.
Do
you
have
somewhere
in
the
kernel
spec
or
as
a
companion
file
adjacent
schema,
specifying
the
required
parameters.
H
So
right
now-
and
that
came
from
some
some
discussion
with
some
of
our
clients,
we
actually
put
the
templates
in
the
metadata
of
the
kernel
spec,
which
allow
you
to
send
a
notebook
to
someone.
And
if
you
send
the
notebook
to
someone,
then
they
have
the
templates
in
the
metadata
and
they
can
retemplate
like,
for
example,
if
they
don't
have
200
bigger
frame,
they
can
change
to
have
100
or
10,
or
something
like
that.
I
H
Right
now
they
can
have
spec
and
we
want
to
embed
that
into
the
letter
of
the
notebook
as
well,
so
that
the
users
can
re-parametrize
if
they
want
to.
I
mean
right
now:
it's
really
a
prototype,
especially
there
are.
There
are
some
questions
like
in
particular.
If
you
use
a
scheduler,
you
may
want
to
have
say
the
gui
that
says
and
third
gigabytes,
but
you
need
to
pass
the
amount
of
of
memory
as
kilobytes
to
slurm,
and
so
we
have
a
workaround
for
that.
H
Where
you
can
define
you
can't
really
execute
arbitrary
code
or
think
of
that
jumping
in
the
notebook.
So
for
some
things
we
allow
to
have
mappings
so
that
it's
slightly
outside
of
the
scope
of
json
schema.
I
can.
I
can
send
you
the
link
to
the
repository
and
the
various
user
stories
actually
return
your
stories,
even
if
I
don't
like
that,
to
show
what
is
possible
and
we're
we're
showing
that
I
mentioned
that's
here
today
to
get
feedback
and
see
what
what
would
be
the
requirement.
H
What
are
the
scope
in
which
you
want
to
you
want
to
do
that,
and
there
are
actually
two
sides
that
I
showed
here
is
only
the
the
ui,
and
I
want
you
to
point
to
kevin
works.
H
I
thought
kevin
is
here
downward
my
screen
about
kernel,
provisioning
that
was
merged
recently,
I
believe
by
steve
a
couple
of
hours
ago
or
hours
ago
that
basically
would
let
your
system,
as
far
as
I
understand,
tell
jupiter,
which
kind
of
resources
you
can
you
can
expose
and
then
the
ui
side
could
show
that
to
the
users.
So
there
are.
I
H
I
will
I
will
ping
you
ping
on
the
relevant
repo,
which
right
now
is
called
ksmm,
and
I
don't
remember
what
it
stands
for
with
it
at
some
point.
It's
called
aspect
manager
manager.
If
you
have
a
better.
I
H
I'm
happy
to
get
it.
I
H
And
that's
that's
it
for
me,
as
far
as
for
demos,
so
I'm
happy
to
discuss
more
about
that
later.
If
someone
has
more
questions.
H
So
this
is
running
both
the
server
extensions
to
be
able
to
manipulate
canal
spec,
because
there
is
no
good
api
endpoint
to
do
that
and
the
front-end
extension
for
jupyter
lab
the
front-end
extension
here
is
mostly
a
mock-up
like
some
of
it
works.
Some
of
it
doesn't
work
because
I
just
don't
understand
react.
I
don't
understand
javascript
theory
in
general
or
why
but
yeah
you
could.
I
hope
to
move
that
to
a
more
to
a
more
more
core
things.
H
In
particular,
we
directly
manipulate
spec,
and
I'm
wondering
if
the
notion
that
can
expect
templates
by
itself
as
something
different
and
can
expect,
would
make
sense.
J
Just
as
a
follow-up,
I
believe
eric
was
planning
to
come
to
the
server
meeting
this
thursday.
We'll
talk
more
on
this
subject
as
well.
If
anyone
wants
to
join
that's
at
9am
pacific
time.
A
A
D
Yeah,
I'm
I
was
gonna,
do
slides,
but
I
think
I
can
keep
this
shorter
than
that
and
I
see
that
rolling
didn't
make
it
because
we're
also
on
pacific
time
and
8
a.m
is
not
super
early,
but
it's
a
little
early.
So
for
those,
so
I
know
at
least
one
of
you
matthias,
because
matthias
has
helped
me
present
and
invited
me
to
present
on
security.
D
Before
some
of
you,
I
used
to
work
at
san,
diego
super
computer
center
running
hpc
systems
and
help
users
run
stuff,
and
I
worked
for
globus
for
a
while,
helping
with
projects.
Globus
does
data
management
and
auth
stuff,
but
I've
been.
I
have
an
astrophysics
background,
and
so
I
use
jupiter
and
python
quite
a
bit
for
that
side.
D
But
what
I'm
talking
about
today
is
I'm
trying
we're
doing
that
classic
like
let
people
know
what
we're
up
to
and
then
get
your
input
so
roland
and
I,
after
the
first
round
of
community
workshops
we
submitted
for
a
second
one
on
security.
D
Ohio
supercomputing
center
has
offered
to
host
so
we're
likely
to
do
it
there
in
columbus,
ohio.
We
want
it
to
be
something
that
would
be
very
appealing
to
a
broad
range
of
participants,
and
so
we
didn't
didn't
have
to
be
the
classic.
Do
it
at
berkeley
or
nurse
the
outcomes
we're
looking
for
are
things
like
best
practices
guides
things
that
we
could
help
say,
show
users
like
if
you're
going
to
run
jupiter.
D
There's
already
some
really
important
things
like
the
security
group,
the
signed
key
and
stuff
like
that
that's
clearly
distributed,
but
as
we're
all
aware,
with
current
the
current
environment,
the
current
news
cycles,
security
is
becoming
more
of
a
concern
even
for
research
in
academia
and
so
we'd
like
to
say
hey.
What
can
jupiter
do,
even
though
we're
very
distributed,
diffuse
project
to
bake
some
stuff
in
and
there's
areas
in
particular
that
we
want
to
start
focusing
on
we've
done
interviews,
since
we
have
a
long
lead
up
to
this
workshop.
D
We
were
also
lucky
enough.
The
nsf
has
a
center
of
excellence
for
cyber
security
called
trusted
ci,
and
we
have
an
engagement
with
them
to
support
this
activity,
so
we're
getting
help
from
some
of
the
folks
that
are
folk
on
the
nsf
cyber
infrastructure
side
for
working
on
security,
and
what
we'd
like
to
know
from
this
group
is,
who
would
you
say,
is
appropriate
to
approach
talk
to
when
you
think
about
development
and
processes
and
awareness
for
security?
D
In
particular?
I'm
really
glad
that
I
just
heard
the
notebook
server
call
made,
because
that
seems
to
be
a
key
component
like
even
when
we're
talking
with
folks
like
the
broad
is
a
lot
of
folks
are
running
the
notebook
server
with
their
own
proxies
and
that
matches
what
we
do
when
we
work
on
our
laptops
or
we're
working
on
systems.
D
This
is
kind
of
our
first
step
into
letting
people
know
and
if
we're
welcome
on
the
thursday
call
at
8
a.m,
we'd
like
to
come
talk
there
too
and
see.
If
that's
an
area
where
we
can
have
a
more
a
closer
interaction
during
the
run
up
to
the
workshop.
So
please
any
feedback
or
we
can
just
roll
on
to
the
next
item.
A
I'm
just
I
can't
well
articulate
in
this
space
how
excited
I
am
to
hear
all
the
words
you
just
said
like
personally
yeah.
It
sounds
great.
D
I
see
you're
at
ucsd,
so
I'm
at
some
point
I
should
probably
say
hi
to
you.
E
F
A
D
You
look
in
the
participants
list
and
then
you
just
try
and
keep
track
of
who
raised
their
hand
first.
I
think
simon
had
his
hand
up
first.
F
You
all
right,
so
this
sounds
really
interesting.
Obviously,
I'm
coming
from
the
jupiter
point
of
view
and
I
think
one
of
the
big
challenges
with
jupiter's.
No,
it's
not
just
distributed.
It's
also
that
we've
got
very
wide
community
participation
and,
for
instance,
one
of
the
difficulties
I
have
in
you.
When
I
look
at
maintaining
jupiter
projects
is
jupiter.
D
Yeah,
I
we
very
much
acknowledge
that
this
is
something
where
we
know
that
this
is
there's
additional
effort
required
and
it
may
take
new
contributors
before
new
contributors
or
some.
D
G
Nice
yeah,
so
the
best
practices
badge
is
a
great
process.
It's
a
great
checklist.
It's
self-certified!
It's
not
like
you
have
to
go.
Ask
somebody
for
it.
So
that's
very
I've
found
that
very
effective
on
you
know.
If
you
go
to
a
security
conscious
organization,
they
just
want
to
know
if
you
have
done
a
checklist
and
their
checklist
might
be
different
from
your
checklist,
but
some
of
those
map
to
you
know
some
of
them
have
have
different
ones
between
them.
G
So
that's
a
really
great
step
forward
and
it's
also
another
place
to
get
google
juice.
Second
point:
so
we've
been
doing
all
this
space
around
webassembly-based
kernels.
The
enorx
project,
which
I
believe
is
in
my
copy-paste
buffer
is,
is
a
place.
I'd
like
to
see
that
go
at
some
point.
You
know
right
now,
you
know
systemd
containers
are
good,
but
they
only
work
on
certain
flavors
of
debian
with
all
the
stuff.
You
know
hotted
up
all
the
way.
G
If
web
assembly
became
something
that
we
could
rally
around
as
something
that
works
in
all
the
run
times
that
we
care
about,
that
might
be
an
interesting
place
to
handle
the
you
know
the
elephant
in
the
room,
which
is
you
know,
remote
code
execution
as
a
service,
and
so
there's
there's
certainly
some
stuff
moving
towards
that
way,
but
that
that
whole
that
whole
workflow
is
nascent.
G
But
you
know
having
provable,
not
formally
verified,
but
but
you
know,
well-verified
runtimes,
that
you
then
can
load
your
your
user
code
into
and
really
understand,
irrespective
of
where
you're
deploying
it.
What
that
baseline
is
that
you're
you're
going
to
be
better
off
because
it's
better
designed
than
say
the
the
docker
oci
whatever
runtime
for
that.
So
I
would
be
very
interested
in
seeing
some
stuff
happen
in
that
direction.
D
Yeah
thanks
yeah,
the
I'm
glad
that
you
have
a
positive
reaction
to
the
best
practices
badge,
because
I
felt
it
was
similar.
It's
like
you
know
it's
not
a
nist
standard
for
security,
but
it's
like
it
is
one
of
the
few
ways.
An
open
source
project
can
sort
of
say,
like
hey,
we've
looked
internally
and
we're
trying
to
do
something
positive,
and
you
know
it
it.
You
know,
has
some
adoption
and
it's
backed
by
linux
and
stuff
like
that,
so
I
think
it's
worth
trying
and
see
how
it
does
the
point.
D
I
also
like
your
point
like
there's
feels
like
two
sides.
One
is
what
is
process
and
you
know,
governance
and
other
stuff
around
security,
and
then
what
are
specific
areas
in
terms
of
actual
development
that
are
going
to
benefit
our
users
and
the
code
and
be
enabling
for
usage
and
deployment
and
enarch
sounds
like
one
of
those.
H
Yeah,
I
agree
a
lot
of
what
was
said
and
I
think
one
of
the
difficulty
with
everything
being
distributed,
that
no
one
is
responsible
and
things
fall
to
the
crack.
So
we
recently
received
two
button:
child
vulnerability,
disclosure
and
like
on
one
of
the
security
mailing
lists,
and
only
a
few
people
responded
and
all
the
people
who
responded
and
and
and
tried
to
get
engaged
or
to
do
the
fixes.
H
We
don't
have
much
time
and-
and
I
think
I
think
we
should
probably
have
a
known
group
which
is
really
responsible
for
responding
in
a
timely
manner
to
everything
which
is
security
related
and
have
a
process
to
really
make
sure
that
things
get
followed
through,
because
I
believe
there
is
a
third
potential
thing
that
is
not
too
critical
but
was
overlooked
and
and
has
not
been
fixed
yet
so
so
yes,
I
I
don't
know
that
nixon.
Let's
take
it
takes
money.
H
Yes,
it
takes
money,
but
we
need
a
way
to
finance
this
money.
You
need
people
who
are
always
watching
and
and
and
really
like
respond
quickly.
I
also
want
to
have
a
way
for
the
community
to
organize
timed
releases.
Like
one
service,
for
example,
if
we
had
to
do
an
emergency
fix,
we
can't
just
make
a
normal
lease
or
merge
and
do
the
release.
A
week
later,
we
have
to
coordinate
with
vendors
and
say:
hey
we're
going
to
release
at
that
specific
time.
D
Well,
matias,
I
did
mean
to
send
you
an
email
and
I
apologize
for
not
doing
it
sooner
because
we,
you
know,
you've
been
engaged
in
this
and
helped
us
get
started
and
like
this
example
is
one
where,
for
you
know,
I
know
that
jupiter
spread
out,
but
there
are
some
folks
at
the
top.
We
really
try
to
say
what
are
key
things
we
have
to
focus
on
and
when
it
comes
to
security,
having
some
policies
like
you
know,
we
will
prioritize
security
work
such
that
you
know.
D
If
a
vulnerability
comes
in,
you
know
it
gets
a
rating
within
a
few
days
say,
for
example,
and
then
you
know
we
will
work
on
it
and
do
a
coordinated
release
like
it's.
You
know
it's
it's,
how
it's
putting
a
little
bit
of
structure
that
fits
the
community
slowly
over
time.
So
I
would,
I
think,
it's
one
of
the
potential
outcomes
and
it
could
be
a
recommendation
from
the
workshop
or
the
process
that
we're
doing
so
because
that's
yeah.
These
are
real
real
needs
for
the.
A
D
Yeah,
so
we
have
a
discourse
post.
We
actually
let
me
once
I'm
done
chattering
I'll,
find
the
link
and
so
roland,
and
I
aren't
on
jitter
and
we
do,
but
we
do
try
to
follow
that
discourse,
post
and
so
you're
welcome
to
reach
out
to
us
there
and
of
course,
there's
our
emails
and
stuff
like
that.
But
I
think
for
this
the
discourse
would
be
the
most
appropriate.
A
D
Well,
yeah,
and
but
this
is
what
we're
we're
kind
of
considering
it
is,
even
though
the
workshop
will
be
next
year,
the
more
we
get
done
along
the
way,
the
more
focused
the
workshop
can
be
on
stuff,
where
we
feel
like.
We
really
need
people
in
a
room
discussing
things
so
the
more
we
can
work
out
at
a
reasonable
pace
ahead
of
time.
D
H
Do
you
have
any
insight?
I
know
that
google
and
some
other
company
are
sometimes
giving
money
to
to
open
your
software
to
focus
on
security.
I've
posted
the
link
on
the
chat.
Do
you
believe
jupiter
would
have
any
any
chance
to
get
accepted
into
something
like
that.
D
I
think
in
the
near
term,
I'll
have
to
look
at
what
nick
just
posted,
but
I
think
this
is
really
ripe
for
say.
An
eager
award
from
the
nsf
sweet
eager
awards
are
something
that
are
limited
funding.
D
They
tend
to
be
like
100
to
150
000,
but
if
we
could
come
up
with
a
defined
activity
to
say
like
we
really
want
to
establish
a
security
program
for
jupiter,
you
know
and
and
sometimes
they're
used
to
host
workshops,
but
fortunately
we
already
have
the
num
focused
money,
so
I
think
there's
the
near
term
for
it
and
then
I
think,
there's
the
question
of
longer
term.
D
E
D
I
appreciate
the
feedback
and
I'll
put
roland
and
rick
and
rollin
and
tiffany,
if
she's
up
for
showing
up
on
the
agenda
for
thursday
morning
to
ask
about
what
we
might
be
able
to
do
there.
Thank
you.
A
A
K
K
K
So
you
can
wait
here
at
the
top
level,
which
is
a
nice
feature,
but
unfortunately
it
still
blocks
the
kernel.
So
if
I
launch
basically
the
same
code
in
the
next
cell,
it's
going
to
be
run
after
the
first
one,
which
is
not
really
what
you
expect
when
you
have
async
io.
Usually
you
want
things
to
be
run
in
parallel,
so
it's
nice
because
here
you
can
run
a
function
which
is
depending
on
asynchro
features,
but
still
you
would
expect
to
have
a
concurrent
behavior
and
that's
what
a
canal
is
about.
K
K
And
what
happens
with
this
widget
is
that
it's
fetching
some
information
from
the
front
end
and
this
code,
which
should
work,
actually
blocks
it's
trying
to
get
the
layers
from
the
front
the
front
end,
but
it
cannot
receive
the
the
message
because
it's
blocking
executing
some
code,
it's
kind
of
a
deadlock,
and
that's
that's
something.
That
is
that's
an
issue
we
have
in
ipi
widgets,
but
so
it
could
be
solved
at
the
ipad
widget
level,
but
it
can
also
be
solved
by
a
channel
at
the
kernel
level.
K
E
That's
rad,
so
does
this
work
with
stuff
like
can
I
use
like,
but
they
yeah
can.
I
use
like
trio
with
this.
K
So
now
it's
using
a
pure
async
io,
no
no
trio
yet
or
I
know
also
any
io-
would
be
a
a
cool
library
to
support
because
it's
basically
an
abstraction
layer
on
top
of
async
io,
which
gives
some
trio
like
features,
but
you
don't
care
basically
if
the
backend
is
trio
or
asynchro
so
yeah.
This
is
this
is
a
nice
direction
for
development,
but
right
now
it's
only
pure
snk,
so
it's
yeah
yeah.
Obviously
it
raises
a
lot
of
questions
because,
like
now,
all
cells
are
run
concurrently
and
is.
K
Is
it
really
what
we
want?
I
mean
you,
you
don't
want
all
your
code
to
run
in
parallel,
so
there
are
semantics
to
be
to
be
invented.
For
example,
if
I
share
my
screen
again
so
we
saw
that
these
two
cells
are
run
concurrently.
K
F
K
H
Switching
and
having
the
synchronous
by
default,
to
not
change
current
notebook
and
by
keying
being
a
async
like
in
the
await
task.
You
would
do
something
like
as
excel.
K
H
K
Actually
it
doesn't
if
your
cells
are
not
asynchronous,
if
they
don't
have
a
in
a
weight
word
in
them,
it
falls
back
to
to
the
default
behavior,
because
this
cell
is
blocking
it's
going
to
pause
the
execution
of
this
one.
So
let's
try.
K
So
if
you
don't
use
the
weight
at
the
top
level,
you
have
the
same
behavior
as
a
ipi
kernel.
Basically,.
E
So
I
think
you're
actually,
so
I
love
what
you're
doing
here
david,
because
it
raises
some
difficult
teaching
challenges,
but,
like
is
jupiter
the
right
place
to
learn
about
concurrency.
I
have
a
lot.
I
have
a
hard
time
with
some
of
the
education
on
concurrency
like
I
can
see
this
being
like
a
really
cool
substrate
to
learn
about
that
programming
method.
The
other
thing
that
I'm
seeing
at
least
when
I'm
teaching
folks
to
write
notebooks
is
that
restart
and
run
all
or
it
didn't
happen.
E
K
I
also
think
that
it
can
really
show
its
potential
when
designing
dashboards
and
yeah,
of
course,
using
voila
and
so
on,
because
when
you
are
using
ipi,
widgets
and
widgets
in
general,
you
quickly
fall
in
the
kind
of
callback
hell,
because
you
have
to
observe
some
some
traits
in
order
to
to
do
some
action,
and
basically
you,
if
you
you
use
callbacks
everywhere
and
it's
become
it's
becoming
quite
complex,
and
I
think
this
this
kind
of
concurrent
cell
can
solve
this
problem
because
you're
always
still
in
the
notebook
you're,
not
somewhere
else.
K
You
executing
some
some
function
that
cannot
have
an
output
in
the
notebook,
for
example,
but
yeah
as
far
as
as
using
that
to
teach
asynchronous,
maybe
yeah.
It's
not
the
right
thing
tool
to
do.
H
H
If
you
want
to
analyze,
something
which
is
dynamic
like
like
a
live
event
stream,
or
something
like
that,.
K
K
Notebook
so
yeah,
but
anyway
it's
just
wanted
to
share
that
and
yeah
some
people
in
at
unfolded
are
interested
in
that
because,
basically
they
cannot
solve
their
widget
issue
right
now,
so
yeah,
let's
see
how
it
goes.
K
The
thing
is,
they
are
pushing
towards
this
direction
like
to
have
at
least
a
synchronous
handling
of
com
messages
in
ipad
widgets,
but
it's
kind
of
dangerous
because
it
breaks
some
up
assumptions
very
low
level
assumptions
that
we
really
don't
know
about
all
the
consequences.
So
it
could
be
an
opt-in
behavior
that
the
widget
designer
chooses
to
to
take
at
its
his
own
risk,
but
yeah.
We
are
not
sure
to
to
merge
that,
but
it's
one
of
the
potential
solutions.
A
E
I
just
know
the
fact
that
we
have
this
top
level
weight
statement.
I've
been,
I
made
a
little
hack
for
ipython.
That
gives
you
top
level
return
statements,
so
you
get,
you
can
use
return
and
get
displays
in
different
places,
but
I
guess
at
some
point
this
awaits
stuff
is
going
to
be
python
proper
right,
like
the
top
level
weight
statement,
is
a
pep
or
something
is
that
going
to
change
how
a
weight
stuff
gets
implemented
in
the
kernel.
I
know
matias
had
to
fight
pretty
significantly
to
get
this
top
level
weight
statement.
H
I
I
would
love
to
have
something
like
that
before
I
python.
My
python
has
just
such
a
complexity
of
what
it
supports
that
it's
relatively
hard
to
hard
to
add
right
now
until
we
know
the
end
semantics
right
now,
I'm
not
even
sure,
but
ipython
is
completely
compatible
by
2010.
A
A
G
I
just
want
to
give
a
shout
out
to
the
team
working
on
lab
the
started,
rolling
rc2
on
a
couple
things
and
if
you
follow
all
the
rules-
and
you
write
a
document
viewer,
it's
just
going
to
work
with
the
collaborative
editing
stuff,
it's
kind
of
insane
that
my
jaw
literally
dropped
the
first
time
that
it
worked
for
ipad
raw.
So
it's
going
to
be
very
exciting.
All
new
proper
problems.
A
I'm
probably
gonna
cry
after
this
call,
that's
kind
of
amazing.
Let
me
find
a
link
for
ipidroio
in
case
anyone
needs
it.
E
What
does
observable
do
a
weight
wise
is
observable,
no
are
observable,
notebooks
concurrent
async
or
are
they
blocking?
Does
anybody
know
it
rewrites
all
your
code.
G
L
Explicit,
I
have
a
quick
question
about
rtc
in
in
jupiter
hub,
so
the
server
side
is
there
somehow
it's
it's
really
cool
that
it's
landing
in
jupiter
lab,
but
is
there
anything
what
where
to
read
up
on
what's
happening
on
the
hub
site.
A
G
Yeah
yeah,
so
I
mean
you
know
we're
kind
of
stuck
with
the
ygs
architecture.
Now,
with
all
of
its
you
know,
joys
and
the
what
we've
done
to
make
it
work
on
one
of
the
things
called
jupiter
light.
G
We
use
the
webrtc
based
one
so
it
you
know,
your
client
calls
out
to
a
signaling
server
that
you'd
be
able
to
stand
up
on
your
hub
as
a
managed
service
right
now,
there's
a
node.js1,
but
you
know
whatever
we
can
re-implement
that
and
whatever
makes
sense
at
scale
for
the
deployment
purposes,
and
you
would
specify
that
so,
instead
of
using
the
one
on
your
individual
server,
you
preload
all
your
environments
with
use
this
other
lab
extension
swap
out
that
back
end
put
this
other
one
in
and
be
able
to
use
it.
That
way.
A
I
Thank
you,
yeah
regarding
the
integration
with
hub.
We
don't
have
a
good
solution
right
now,
but
it's
like
the
main
focus
of
the
team
of
like
having
a
proper
oath
and
meaning
authentication
story
and,
like
presence,
and
you
know,
sort
of
ui
and
the
the
the
main
challenge
is
that
there
are
so
many
develop.
You
know
deployment
scenarios
and
you
know
authentication
stories
around
jupiter
that
it's
a
really
you
know
range
of
use
cases
to
address
that
things,
and
I
I
don't
think
we
can
take
all
this
in
the
two
minutes.
G
Well,
it's
just
you
know,
documenting
it,
making
the
yjs
stuff
like
there
are
so
many
magic
numbers,
it's
just
like
whatever.
Let's
just
make
a
random
number,
you
know
establishing
some
of
those
types
so
that
it's
more
predictable
for
what
the
client
is
going
to
expect,
I
think,
is
the
thing
forward
and
making
more
parts
of
it.
G
L
A
Hope
to
see
you
on
some
of
the
discussions
in
the
future,
but
we'll
see
well
that
that
is,
we've
hit
the
hour.
Everyone,
if,
first
of
all,
if
you
have
to
run
now
to
another,
call
feel
free.
All
I'm
gonna
do
is
wrap
up
and
tell
you
when
the
next
call
is
so.
Thank
you
all
for
being
here
today.
I
really
appreciate
your
time.
A
I
hope
you
had
a
great
time
as
well,
because
I
love
those
are
amazing,
shares
today
and
a
lot
of
topics
I
hadn't
heard
here
before
the
next
call
is
going
to
be
on
august
31st.
A
You
can
also
give
feedback
on
this
call.
Let
me
get
you
the
link,
it's
basically
just
a
few
questions
that
say:
hey
was
the
audio
broken
and
did
you
have
a
nice
time
and
any
other
feedback,
and
you
can
already
sign
up
for
the
next
call
if
you've
been
so
wildly
inspired
by
these
great
shares,
that's
the
august
31st
agenda
and
I
hope
to
see
you
all
in
a
month.
Thank
you
so
much.