►
From YouTube: Jupyter Community Call - June 28, 2022
Description
Recording from the Jupyter Community Call in June 2022.
The notes from this call can be found here: https://github.com/jupyter/jupyter/issues/618
Read more about these calls on Discourse: https://discourse.jupyter.org/t/jupyter-community-calls/668
A
A
If
you
haven't
been
here
before
the
goal
of
these
calls
is
just
to
be
a
place
to
celebrate
anything
jupiter
related
that
people
are
making
that
people
are
seeing
that
people
are
just
getting
excited
about
and
also
help
us
build
cross
project
and
community
relationships,
because
we've
got
so
many
things
that
people
do
in
jupiter's
orbit
and
not
a
lot
of
ways
to
stay
connected.
A
So
this
is
the
fun
meeting
get
excited,
but
before
we
get
into
the
fun
there's
a
few
things
I
have
to
say
for
housekeeping
first
one,
it's
call
is
recorded.
You've
heard
that,
but
just
a
reminder
whatever.
That
means
to
you
two,
because
this
is
a
jupiter
event.
We
are
held
to
the
jupiter
code
of
conduct.
That
includes
me.
A
I
actually
have
to
get
you
all
a
new
url,
because
jupiter.org
slash
conduct,
I
think,
doesn't
exist
anymore,
so
be
patient
with
me
on
that
I
just
checked
today
and
three
if
I
mispronounce
anyone's
names,
please
let
me
know,
but
with
that,
what
we're
going
to
be
doing
is
following
the
agenda
I
linked
before
this,
and
so
our
first
stop
on
the
list
will
be
all
our
short
reports
celebrations.
Shout
outs,
so
these
are
just
exciting
things
that
don't
necessarily
need
discussion,
and
we
have
one
hooray
steve.
B
Oh,
my
okay,
so
I
just
wanted
to
give
a
little
bit
of
a
sneak
preview.
This
will
be
a
blog
post
once
everything's
ready
to
pull
the
trigger
on.
But
if
you've
ever
written
a
python
package,
you
probably
use
setup
tools
and
it's
a
it's
been
sort
of
a
pain
in
the
butt.
Over
the
over
the
years,
we've
been
like
basically
finding
ways
around
it.
B
We
had
jupiter
packaging
for
a
while
to
smooth
over
some
of
the
pain
points,
but
even
that
had
its
limitations
and
then
recently
there
was
a
packaging
back-end
added
to
the
python
packaging
authority
called
hatch
which
got
a
1.0
release,
and
so
I
started
to
try
and
experiment
with
that
and
see
what
we
could
do
as
a
hatch
plug-in,
because
it
offers
a
plug-in
base,
a
plug-in
system,
and
so
basically,
we
took
the
parts
that
were
in
jupiter
packaging
and
made
them
a
plug-in
for
hatch.
B
B
That's
either
instead
of
config
or
set
up
high
and
translating
that
into
stuff
that
either
hatch
or
hatch
jupiter
builder,
the
plugin
name
understands
it's
getting
pretty
close,
so
the
the
jupiter
lab
extension
cookie
cutter
ts
is
like
that
that
that
pr
is
open,
and
that
is
a
result
of
running
the
merge
script
on
the
thing
that
you
would
have
gotten
from
running.
The
cookie
cutter,
the
widget
cookie
cutter
ts
is
almost
done.
B
The
po
is
just
about
passing
and
I
gotta
do
some
updates
and
add
a
new
test
to
that
migration
script
and
the
last
one
I
wanted
to
get
was
the
the
javascript
extension
and
then
so.
We've
already
merged
this
into
jupiter
servers
using
it
jupyter
notebook
itself
is
using
this
back-end
and
then
there's
an
open
pr.
Jupyter
lab
that'll
use
it.
B
B
Hatch
also
supports
what
was
deprecated
instead
of
tools,
the
data
files
and
actually
it
does
it
better
and
it's
cleaner
to
use
and
it
it
works
in
edible
mode,
whereas
data
files
didn't
so
that
means
that
kernels,
you
can
also
supply
as
a
data
file,
the
kernel
json
file,
so
basically
all
of
our
types
of
extensions,
kernels,
notebook
extension,
server,
extensions
and
lab
extensions,
all
work
with
hatch
without
all
the
legacy
parts
of
setup
tools,
and
you
have
a
an
up-to-date
pep,
compliant
builder
and
hatch,
and
then
the
helper
function
and
hatch
jupiter
builder
to
get
the
rest
away
there.
B
So
I'm
pretty
excited
about
this.
I've
been
battling
subtools
for
like
15
years
now,
so
yeah
that
that's
it.
A
C
D
C
B
So
the
the
migration
script
hatch
itself
has
a
set
of
tools,
migration
script
and
then
we
add
another
one
that
like
mocks
out
the
jupiter
packaging
module
and
says:
oh,
you
were
trying
to
call
this.
I'm
going
to
intercept
that
and
write
out
pi
product
automal
data
instead
and
then
it's
some
other
tricks
they're
doing
there.
So
the
intent
is
99
of
it.
It's
just
going
to
be
migrated
for
you
and
then
it'll
for
things.
B
That
knows,
it
can't
do
itself
it'll
print
a
warning
at
the
end,
but
so
far
on
the
cookie
cutter,
the
only
one
that
was
was
having
the
static
version
field
in
version
up
high
instead
of
the
the
tuple
and
everything
else
was
just
migrated
for
you.
So
it's
just
run
the
migration
script
and
it's
a
compare
script
where
you
can
say:
it'll
build
a
wheel
on
like
a
clean
checkout
of
set
the
setup
tools
base
one
and
then
also
with
the
hatch
migration
and
compare
the
files.
B
So
do
that
on
the
wheel
and
the
s
disk
and
tell
you
if
you've
accidentally
missed
the
file
either
from
bad
metadata
or
you
you
didn't,
it
wasn't
picked
up
as
part
of
the
build
or
whatever.
So
it
gives
you
some
confidence.
B
Yeah,
so
I
mean
to
be
fair
to
set
up
tools
anderson,
I
can't
remember
their
last
name
has
been
modernizing
a
lot
of
it
and
and
bringing
it
up
to
date
with
modern
pep
standards.
At
the
same
time,
the
core
python
has
basically
dumped
this
details
on
them,
so
they're
having
a
struggle
to
jason
coombs.
I
think
his
name
is
basically
taking
the
brunt
of
that
work.
There's
just
a
lot
to
it.
B
A
lot
of
legacy
there,
so
even
even
as
they
modernize
and
migrate
you're
still
having
to
deal
with
a
lot
of
legacy
if
you're
using
setup
tools
and
another
data
point-
is
that
sci-fi
package
this
year
under
one
of
migration,
see
kind
of
seeing
the
writing
on
the
wall
of
they
were
going
to
have
to
rewrite
a
lot
of
stuff
anyway,
because
they
hadn't
they
were
using
numpy
at
this
details.
B
So
they
they
switched
to
the
mison
build
system
with
a
new
back
end
called
mizonpai
that
you
know,
instead
of
trying
to
wrap
just
details,
the
way
they're
doing
before
so,
and
then
there's
like
a
scikit
build
system
right
now
that
uses
setup
tools
like
murder,
setup
tools
and
cmake
they're,
also
looking
to
to
get
get
away
from
the
setup
tools.
B
Part
of
that
as
well
so
generally
people
are,
are
where
possible,
trying
to
get
away
from
set
tools
just
just
because
of
the
legacy
and
like
it's
there's
a
lot
of
conflicting,
confusing
documentation
out
there
about
it
again,
no
fault
of
the
the
maintainers,
just
the
the
the
nature
of
the
beast,
as
it
has
evolved
over
20
years.
So
yeah
there's
a
lot
of
things
going
on
in
the
packaging
world
right
now,.
A
B
Yeah
they've
been
they've,
been
doing
a
great
job.
Ralph
gomer
has
been
well
drawn
named
like
gomer's,
yes,
is
it
rough?
Okay
got
it
yeah
on
the
spot
has
been
leading
that
effort
so
involves
like
getting
things
into
cython
to
support
it
working
with
the
meson
folks,
so
they
still
have
a
couple
workarounds
for
limitations
on
both
of
those
projects,
but
they're
it
does
work
they're
able
to
build
the
new
version
of
scipy.
That's
an
rc
right
now.
B
I
think
it's
1.9
and
numpy
is
going
to
follow
suit
and
basically
replace
and
deprecate
numplata
dusty
tills,
probably
all
along
the
same
timeline
as
python
itself
is
so
yep
everything's,
moving
forward
to
the
new
standards
and
the
and
the
new
backends
that
are
coming
out.
D
Okay,
cool
yeah,
you
know
packaging
whoa
yeah,
I
mean
there's
just
you
know.
No
one
really
likes
someone
showing
up
with
I'm
gonna
fix
all
your
packaging
problem
pr's,
and
I
really
want
to
give
a
shout
out
to
steve
for
actually
doing
it,
because
it's
so
painful,
it's
so
painful
right.
D
People
have
such
scars
from
that
over
the
last,
as
you've
said,
like
literally
decades,
most
of
our
careers
have
been
during
this
dark
time
of
python
packaging
that,
like
we
want
to
reuse
software,
but
for
some
reason
python
wants
to
do
it
its
own
way,
but
its
own
ways
50
different
ways.
I
don't
know
so.
That's
awesome
man
that
you
actually
made
some
progress
there
and
and
crucially
gotten
people
upstream
to
carry
some
of
these
goals
that
we
have
right.
D
Like
you
know
this
ability
to
work
on
software
and
development
that
is
so
important.
You
know
I
don't
care
how
perfect
and
hermetic
your
build
is.
If
you
can't
work
on
the
software
while
you're
building
it.
So,
there's
all
these
weird
things
that
come
with
with
python
that
I'm
glad
they're
working
there.
I
guess
I'll
just
steal
the
quick
transition
there,
packaging
python
for
the
browser.
Oh
boy,
is
working
great
and
yeah.
It's
really
working
great.
D
You
know
the
folks
from
from
quant,
whatever
the
one
of
the
s
words
is
is
doing
some
very
cool
stuff,
it's
one
of
them.
I
don't
remember,
and
so
they've
got
the
better
part
of
the
condo
stack
working
in
there
for
the
zoo
stack.
So
with
that
you,
you
know
you
kind
of
know.
You
have
all
the
stuff
up
front.
The
pie
died.
Side
is
getting
really
exciting
and
they're
adding
new
features
all
the
time,
so
it
really
is
becoming
pretty
viable
for
certain
kinds
of
stuff.
D
I
really
hope
people
don't
start
doing
like
you
know,
like
your
menu
bar
needs,
python
to
run
or
whatever,
but
you
know
from
being
able
to
control
your
data
and
compute
on
your
data
in
the
browser.
It's
actually
getting
really
exciting.
Our
little
jupiter
light
announcement
on
there
is
that
you
can
access
from
inside
of
both
zeus,
zeus,
jupiter,
light
and
pilot
the
one
that
is
still
bundled
with
it.
D
If
you
upload
a
file,
you
can
access
that
inside
your
kernel.
If
you
create
a
file,
it
immediately
shows
up
in
your
file
manager.
You
know
all
these.
These
weird
boundaries
that
we
had
to
create
for
ourselves
are
kind
of
slowly
dissolving,
oh
and
it
has
a
real
service
worker
so
that
it
does
the
heavy
duty
cache.
D
So
if
you're,
you
know,
if
you're
using
it
on
tri,
jupiter.org
and
stuff
like
that
it,
if
you
are
not
in
private
browsing
and
you
don't
clear
out
your
cache,
it's
actually
only
going
to
download
that
25
megabytes
of
numpy
once
right,
it's
not
going
to
keep
downloading
that
every
time
it's
not
going
to
download
pandas
every
time.
D
So
this
really,
you
know
we
really,
finally
are
making
some
headway
against
some
of
those
problems,
so
it
may
not
be
as
cool
as
pi
script,
but
the
fact
that
it
still
works
with
all
the
cool
stuff
that
people
are
building
in
the
community
still
makes
me
pretty
helpful
and
that's
about
it.
A
C
I
feel
like
I
feel
like
with
dick
and
steve's
like
last
story,
is
like
to
hear
them
now
versus
like
four
years
ago,
because
you
all
have
wanted
this
stuff
for,
like
so
long.
It's
almost
like
triumphant
to
hear
that,
like
there's
momentum
behind
things
that
you
know
you've
been
pushing
for
for
so
long.
That's
so
cool.
D
I
think
it's
getting
mainstream
right,
like
that's.
That
has
been
this.
This
dream
all
along
is
that
you
know
interactive
computing.
It
has
been
mainstream
for
a
while.
We
already
we
won
that
a
while
ago.
You
know
everybody
wants
to
put
it
in
their
front
end,
but
like
simple
packaging,
I
I
don't.
I
don't
it
it's
crazy,
it's
crazy
to
have
to
argue
for
that.
C
Packaging
really
has
gotten
easier
for
the
layman
like
in
the
past
I
have
to
I
used
to
have
to
be
like
nick.
Can
you
lift
me
up
so
I
can
see
what's
happening,
I'm
not
tall
enough
to
like
ride
the
ride
like,
but
now
it's
like
at
a
place
where,
like
yeah,
we
can
do
these
things,
but
for
a
while
it
was
just
like
the
layers
of
complexity
were
just
like
super
hard
to
handle.
D
That's
pretty
straightforward,
to
get
some
rust
in
your
python
these
days,
and
so
it's
really
cashing
in
on
that
that
long
held
promise
of
python
is
the
you
know
second
best
language
at
everything,
except
when
it's
the
first
best,
because
you
wouldn't
want
to
actually
do
the
interactive
computing
part
in
rust
or
you
wouldn't
want
to
do
it
in
fortran
or
you
won't,
wouldn't
want
to
do
it
in
c
plus,
but
you
need
the
power
that
those
languages
have.
D
A
B
Just
to
add
a
little
jupiter
flavor
to
what
nick
said,
the
win
win:
pi,
pty,
kernels
or
terminals.
So
it's
the
it's
it's
what
enables
terminato
on
windows.
Basically,
that's
it's!
Just
a
rust
wrapper
that
the
the
sparta
folks
are
maintaining.
So
it's
a
very
thin
wrapper
or
some
rust
code,
pretty
sweet
using
the
metron
build
back
in.
C
I
don't
know
if
we
use
pythantic
in
any
places,
but
I
know
that
they,
they
just
like,
changed
all
their
validators
to
rust
code
and
got
like
tons
of
speed.
Ups.
There
too,
it
yeah
that
matcher
and
project
seems
like
it's
just
doing
really
good
stuff.
D
E
Wi-Fi
package,
that
is
the
crdt
back-end
client,
is
actually
a
wrapper
around
wires,
which
is
spelled
y
rs,
which
is
the
rest
implementation
of
the
yjsc
or
dt
wire
protocol.
So
there's
pieces
of
core
jupiter
that
we're
going
to
be
shipping
that
are
python
wrappers
around
rust
libraries.
A
D
Yeah
and
then
and
then
to
to
you,
know
inception
all
the
way
fully
back
on
there
I
mean
that
I
still
think
that
webassembly
stuff's
going
to
show
up
at
some
point.
It's
it's
you
know.
2022
is
probably
not
the
year
when
that's
going
to
make
sense
to
have
that
be
an
option
everywhere
in
the
stack.
But
I
I
think
it's
getting
closer
and
there's
a
number
of
new
webassembly
runtimes
and
the
webassembly
component
standard
is
is
chugging
along.
D
So
that's
going
to
leave
us
in
this
place
where
it's
it's
not
really,
because
the
first
implementation
happened
to
be
in
a
language
no
longer
needs
to
guide
necessarily
what
an
ecosystem
does
forever,
and
I
think
that's
an
important
thing
for,
for
you
know,
for
jupiter
like
again,
python
is
a
really
great
interactive
computing
language.
It's
not
a
really
great
systems,
programming
language,
unless
you
met
rocklin
so
like
it's,
you
know
there
there's
all
kinds
of
ways
that
we
can
grow
as
the
building
packaging.
A
A
Okay,
awesome.
No!
I
I
really
love
hearing
y'all's
updates
because
you
give
a
lot
better
context
than
sometimes
I
run
into
into
the
wild.
So
I'm
grateful.
It
is
agenda.
Item
time,
though,
currently
the
only
thing
I
have
on
the
agenda
is
tony
question.
Mark
question.
Mark
question
mark:
are
you
ready.
C
C
Okay,
so
the
seed
for
discussion
before
I
get
there.
I
really
like
having
new
notebooks
with
the
date
on
it.
I
don't
know
how
everybody
struggles
with
the
untitled
thing
now,
but
this
has
been
the
way
to
go
anyway.
It
makes
it
a
lot
easier
to
search
for
stuff
and
organize
things
anyway.
Now,
that's
on
the
record,
how
general
is
the
notebook
format
like
it's
hidden
in
plain
sight
to
a
lot
of
people,
and
it
keeps
the
millions
of
notebooks
on
the
web
that
we
have
honest.
C
C
But
you
know,
maybe
it's
more
of
a
general
document
format
we'll
come
to
that
at
the
end,
and
hopefully
you
know,
y'all
will
have
some
feelings
so
before
we
get
going
like
what
systems
use,
mb
format
like
we
know,
jupiter
lab
uses
it
and
like
mb
viewer
uses
it
and
vs
code
uses
it.
C
But,
like
you
know
how
big
is
like
the
ubiquity
of
the
notebooks
at
this
point,
like
I
mean
I
put
millions
on
here,
but
like
I
mean
when
are
we
at
the
hundreds
of
millions
of
notebooks
like
across
institutions
that
aren't
shared?
So
I
don't
know
if
anybody
wants
to
like
throw
some
things
in
the
chat
like
just
some
weird
stuff,
where
you're
like
oh
yeah,
it's
cool,
maybe
they're
using
it
over
here
or
if
anybody
wants
to
shout
into
the
mic
or
anything
where's
mb4
matt,
and
we
don't
even
realize
it.
D
Well,
I'll
take
a
quick
stab
at
that.
I
mean
I
remember
back
at
the
the
our
you
know,
second
or
third
jupiter
day.
You
know
it's
that
next,
the
next
billion
users,
because
then
you're
not
even
you're,
not
even
thinking
about
how
many
millions
of
notebooks,
because
you've
got
so
many
daring
users.
D
E
D
You
know:
that's
that's
millions
of
new
year's
users
every
year,
even
if
they
don't
stay
with
it
forever,
they
will
have
been
exposed
to
interactive
computing.
That
way,
so
I
think
it's
really.
You
know
once
you
once
you
get
some
of
those
some
of
those
big
corporate
accounts
as
it
were.
You
know
that
that
you
get
the
long
tail
of
usage
over
time.
C
Indeed,
and
when
they're
institutional
and
they
happen
at
younger
ages,
they
become
more
influential.
Anybody
else
want
to
throw
anything
out
there.
A
If
you're
looking
at
places,
we've
had
some
really
good
conversations
with
people
at
space
telescope.
A
C
A
lot
of
the
big
telescopes
do
it.
The
the
lsst
also
has
a
big
notebook
system
and
pipeline
machine
learning.
Pipelines
like
at
this
point.
It's
really
hard
to
like
even
like
number
like
we've
got
so
many
tools.
We've
got
interfaces,
we've
got
publishers
using
it
organizations
and
communities.
C
Julia
for
some
reason
wants
to
use
a
different
format-
that's
cool,
but
for
the
most
part
like,
I
think
we
have
like
critical
mass,
like
notebook
format
wise,
but
like
a
while
ago,
fernando
wrote,
a
blog
post
called
lyric
computing
and
computational
reproducibility.
There's
another
yeah
there's
another
yeah.
This
is
the
one.
C
Has
a
reproducible
yeah
this
one
has
a
reproducibility
case
in
it,
and
it's
just
got
this
like
really
great
story
about
how
when
we're
working
in
notebooks,
it
builds
off
of
donald
news
like
early
ideas
of
literate
programming,
but
when
we're
weaving
live
narrative
directly
into
the
computation,
it's
this
idea
of
literate
computing,
but
we're
interleaving
text
and
code
and
we're
really
telling
stories.
This
is
a
homage
that
I
often
do
to
donald
news
original
literate
programming
paper.
C
But
when
we
look
at
notebooks
as
a
document
or
as
a
thing,
their
literate
programs,
because
they
have
multiple
languages,
encoded
in
them,
notebooks
the
literature
and
a
lot
of
ways
they're
going
to
outlive
notebooks
the
software
and
when
we're
on
the
web,
we've
got
multiple
authors
and
underlying
all
of
this
is
still
nb
format
like
all
of
this
is,
if,
if
we
didn't,
if
there
wasn't
a
json
schema,
would
we
have
had
the
same
control
over
all
of
these
years?
C
Recently,
we've
seen
a
push
of
notebooks
towards
like
files
where
it's
like,
hey.
Let's
take
this
notebook,
let's
turn
into
markdown,
let's
turn
into
python.
Let's
turn
it
into
a
file
system.
Let's
turn
it
into
some
cached
code,
all
of
this
kind
of
stuff,
but
for
the
most
part
like
we're
focused
on
file
systems,
but
when
we
get
rid
of
that
notebook
format,
when
we
get
rid
of
that
json
structure,
we
lose
a
lot
of
things
that
we
know
so
with
all
the
work
that
nick's
talking
about
with
pi
script,
pilot
jupiter
light.
C
You
know
this
idea
of
posix
is
like
a
little
bit
different
when
we're
working
at
the
web.
It's
weird
to
do
file
stuff
on
the
web,
but
you
know
json
becomes
the
thing
that
becomes
most
important
right.
So
you
know
this
notebook
format
as
an
interface
for
applications
at
mass
scale
for
documents
might
have
something
to
it
so
like
when
you
have
json.
Is
it
valid
json?
C
Is
it
a
valid
mb
format?
How
big
is
it
how's
this
going
to
affect
performance,
but
I'm
not
going
to
hit
that
point,
so
this
is
where
it's
got
me
too
thinking
these
days,
at
least
in
my
work
in
my
past
experiences.
Writing
notebooks
and
teaching
people
to
write
notebooks.
C
The
notebook
format
is
a
hypermedia
collage.
It
lets
you
put
tons
of
document
types
in
a
data
structure
and
they
have
natural
places
to
go.
Markdown
and
python
naturally
go
in
their
cells,
mime
types
they
go
in
the
display.
All
of
the
data
that
we
want
can
live
in
the
metadata.
C
So
you
know
at
a
really
general
level.
The
notebook
format
is
a
co
hype.
It's
just
a
collection
of
a
bunch
of
different
hyper
media
right.
So
some
of
the
questions
for
a
discussion
and
like
the
future
we're
working
off
of
draft
four
of
the
notebook
newer
schema
actually
are
more
composable.
We
can
say
stuff
like
there's
a
vocabulary
that
references
the
old
schema.
C
This
is
the
one
we're
going
to
use.
So
you
know
there
are
things
from
the
past
that
we
can
use
when
experimenting
with
the
future,
but,
like
one
of
the
interesting
questions
becomes
like
what
happens
when
we
start
thinking
about
cells,
what
happens
when
our
notebooks
get
too
large
or
our
revision
histories
get
too
large?
How
do
we
get
to
parts
of
a
notebook?
How
do
we
start
searching?
Can
we
go
and
use
these
composable
ideas
of
json
schema
and
say:
hey?
C
We
just
really
have
a
list
of
references
to
cells,
and
maybe
the
cell
is
the
more
critical
abstraction
and
in
some
way
or
another,
the
cells
aspect
is
hindering
us.
So
the
last
question
that
I
like
to
put
down
here
is:
is
the
cell
type
real
and
the
reason
I
say
that
is
because
this
whole
presentation
is
written
in
pull
in
pidgey.
C
So
when
we
go
back
here,
what
we
start
to
see
is
expand.
All
the
code
view
expand
all
code,
all
right
great.
So
when
we
start
to
go
back
here,
all
of
these
cells
are
written
in
markdown.
C
They've
got
ginger
templates
in
here
that
allow
me
to
do
some
formatting,
I'm
really
mixing
languages,
mixing
media
mixing
forms
here
to
be
able
to
tell
like
stories
that
I
don't
think
I
would
have
been
able
to
tell
with,
like
other
implements
right.
I've
just
got
some
python
here.
This
is
a.
C
This
is
a
markdown
cell
alone
right,
so
when
I
go
and
switch
between
markdown
and
code
cells
back
and
forth,
like
I'm,
not
sure
if
markdown
and
code
really
mean
anything
as
much
as
you
know,
text
x,
python
and
text-
slash
markdown
because
like
to
me,
this
is
almost
on
and
off,
at
least
in
my
experience.
Writing
this
very
weird
language,
but
really
thinking
that,
like
in
the
future.
C
Folks
are
gonna,
bring
language
to
code,
they're
gonna
come
with
a
lot
of
markdown
skills
and
the
need
to
communicate
and
code's
going
to
be
kind
of
just
another
device,
along
with
data
and
so
on
and
so
forth.
So
as
we
get
back
up
to
the
top
here-
and
we
start
to
look
at,
you
know
kind
of
a
very
weird
approach
to
writing.
Notebooks,
you
know
how
general
is
the
mb
format
and
like
how
general
will
it
be,
is
kind
of
my
question
so
yeah
there
you
go.
That's
what
I
did
today.
C
That's
what
I
did,
because
there
wasn't
an
agenda.
So
this
is
why
we
do
the
agenda
in
the
future.
Anyway,
I'm
done.
E
E
And
I'm
not
sure
it's
a
fit
for
purpose
document
for
that,
but
it'd
be
kind
of
cool.
If
there's
like
a
notebook
file,
that's
an
actual
notebook
file.
You
can
render
it
as
a
notebook.
Look
at
it
modify
things,
but
you
can
string
together
a
real
spreadsheet,
whereas
certain
cells
are
code
cells,
but
the
rest
of
them
are
sort
of
filled
in
or
something
that
was
one.
E
The
other
thought
was.
I
think
we
specify
an
md
format
that
notebook
must
have
more
than
zero
cells,
and
that
seems
like
an
oversight,
and
I
like
how
flexible
and
out
of
having
sort
of
just
the
standard
notebook.
C
Thanks
man
yeah,
I
like
so
one
of
the
tools
I
worked
on
a
while
ago
was
basically
I
was
using
desk
to
ingest
the
jupiter
con
file
contents
manager
right
and
then
you
dump
the
notebooks.
You
dump
the
all
that
stuff
to
data
frames
and
then
you
know,
search
gets
really
good,
but
there's
a
and
like
with
the
mime
types
and
all
of
those
things
you
can
actually
build
a
really
strong
query
language
across
notebooks.
C
So
you
know
at
a
large
organizational
scale
like
redoing
work
that
somebody
else
has
already
done
and
rewriting
code
is
expensive,
especially
when
you
have
a
lot
more
people
that
could
potentially
be
doing
it
and
figuring
out
how
to
access
parts
of
things
and
search
parts
of
things
at
like
big
scale.
C
A
Another
thing
this
made
me
think
of
was
a
couple
weeks
ago.
We
ended
up
talking
about
mv
format
in
a
jupiter
lab
call.
I
think,
because
the
I
think
it
was
for
the
jupiter
lab
benchmarks.
Repo
federic
was
trying
to
put
together
markdown
centric
tests,
and
we
started
talking
about
the
the
markdown
schema
for
mv
format
too.
I
don't
think
that
ended
up
anywhere
in
particular,
but
less
related
to
your.
Where
is
it
going
to
be
used
and
more
related
to
people
are
using
it
and
constantly
running
into
limits
with
that.
C
Oh
markdown's,
a
great
gateway
drug
to
computing,
like
I
think
we
need
just
like
the
most
inclusive
markdown
world
right,
because,
like
people
love
corto
over
in
r,
you
know,
they're
heavy
are
markdown,
users,
mist
is
gaining
allegiance
and
you
know
pidgey
has
one
user
he
just
presented,
but
two.
D
D
Yeah,
I
mean,
I
think,
I
think,
a
bunch
of
the
tools
that
strip
that,
if
the,
if
the
output
of
computation
never
appears
in
along
with
the
document
that
created
it,
if
it
only
ever
goes
into
a
throwaway
place,
if
it
only
ever,
you
know
shows
up
in
some
pop-up,
you
know
chart
window
or
something
like
that,
we'll
all
you
know
we'll
be
able
to
do
language
server,
protocol
stuff,
we'll
be
able
to
do
source
graph
kind
of
stuff,
but
the
tool
that
actually
figured
out
how
to
do
meaningful,
search
that
tied
inputs
and
outputs.
D
Together
I
mean
that's
knowledge
management
right,
like
if
you
know
all
the
stuff
that
your
data
scientists
have
been
working
on,
that
they
have
chosen
to
share
not
like
creepy
panopticon
kind
of
stuff,
but
you
know
hey.
I
put
this
into
our
chat
room,
and
everyone
understands
that
the
chat
room
is
is
is
a
place
where
knowledge
goes
not
not
to
die,
but
once
it
hits
chat
now
it's
really
part
of
the
index.
You
know
now
it's
really
part
of
the
proceedings.
You
know
every
day's
chat
is
a
published
every
day's
check.
D
C
So
one
of
the
things
that
isabella
and
I
make
jokes
about
when
we're
when
we're
talking
about
things-
and
this
is
no
disrespect
to
you
all
because
for
me
for
a
really
long
time,
tumblr
was
my
favorite
place.
I
really
enjoyed
tumblr
but
y'all
seriously,
how's
tumblr,
not
a
notebook
like
I
mean
it's
basically,
just
like
a
notebook
with
the
outputs
on
the
top
and
the
input
on
the
bottom.
But
like
the
end
of
the
day
I
just
like
I
I
just
I
have
one
notebook,
that's
basically
my
tumblr
for
the
day.
C
D
You
know
and
the
threaded
comments
that
that
helped
in
their
position
and
their
content
helped.
You
know
who
was
interested
in
what
you
were
doing
and
which
things
you
were
doing
were
most
interesting,
which
are
really
hard
problems
like
if
you're
out
there
just
throwing
stuff
into
the
ether
and
making
a
blog
and
making
a
gist
or
whatever
it's
hard
to
assess
whether
those
things
are
interesting
and
the
more
technical
that
your
stuff
gets,
the
even
harder
it
is
for
someone
to
imagine
coming
in
and
talking
about
it
and
so
yeah.
D
If,
if
all
you
can
ever
talk
about,
are
140
character,
tweets
or
whatever,
that's
small
enough
that
you
can
actually
take
it
in.
But
I
think
a
lot
of
people
have
thoughts
that
are
bigger.
You
know
and
they
need
more
space
and
the
notebook
is
a
great
way
to
put
stuff
in
it.
There
was
an
incredible
notebook
just
the
other
week,
a
guy
designed
a
a
brain.
D
Processor
in
python
that
he
compiled
in
one
notebook
so
like
like
base,
you
know,
did
the
pass
through
the
vhdl
and
then
actually
emulated
it
inside
of
the
kernel.
You
could
not
do
that
that
I
think
I
think
the
notebook
might
be
the
only
platform.
You
could
actually
convince
yourself
that
an
eso
lang
like
that
is
worthy
of
having
a
dedicated
cpu
and
I'm
going
to
write
it
right
here.
You
know,
and
with
with
the
ability
to
do
text
layout
good
enough
to
handle
a
white
space.
Only
language.
D
Front
page
of
hacker
news:
it
was
when
you,
when
you
look
on
there,
it
says
the
it
has
the
the
the
domain.
So
I
always
my
eyes
always
perk
up
when
it's
envyviewer.org
right
so
share
your.
C
C
B
Yeah,
let
me
do
a
screen
share.
Is
this
gonna
work
yeah
firefox,
so
I
saw
jeremy
start
start
this
repo.
It
was
like
shut
up.
My
github
feed
and
I
was
like
wait
a
minute.
I
think
I
took
this
class.
This
is
digital
communications
at
university
of
florida.
I
took
this
class
in
like
2008
and
it
was
in
matlab
and
now
it's
online
as
a
jupiter
light
instance.
This
absolutely
blew
my
mind
like
what
the
heck
so
cool,
so
cool.
C
B
D
Yeah,
I
I
think
well,
the
its
impact.
The
the
complexity
of
hosting
compute
in
an
organizational
setting
cannot
be
overestimated.
How
how
hard
that
is,
and
if
your
job
is
to
offer
a
a
complex
topic
like
whatever
you
just
sent,
which
seemed
totally
non-trivial
right.
You
know
the
amount
of
gear
that
you
would
need
to
do
that
justice
is,
is
kind
of
overwhelming
for
a
lot
of
instructors,
they're
really
good
at
what
they
do.
D
But
they're,
not
you
know,
kubernetes
administrators
or
certain
companies
come
in
and
they
you
know
they
get
your
your
your
I.t
to
start
moving
all
their
stuff
over
their
cloud
services.
And
then
you
have
no
way
to
offer
anything
bespoke
that
isn't
like
fully
buying
into
that
ecosystem.
And
then
your
content
lives
there.
So
yeah.
I
think
we're
going
to
see
a
lot
of
jupiter
light
and
related
technologies
and
to
to
that
point
earlier
of
of
pi
script.
D
Being
cool,
like
the
shortcoming
that
it
has
is
that
they
did
not
give
you
a
set
of
tools
for
building
for
building
editors
out
of
the
box
right
like
that's.
They
they,
if
they
had
deconstructed
notebook
classic
or
if
they
deconstructed
ipython.
Frankly,
you
know
terminal
based
lines.
You
know
line
by
line
base
and
giving
you
an
easy
way
to
build
that
that
might
have
if
it
was
easier
to
build
a
bespoke
whole
ui,
but
it's
still
used
to
lift
a
lot
of
water.
D
C
D
Oh,
oh
yeah.
I
bet
that
would
almost
work
today
with
the
zeus
back
end
one.
So
you
know
we
can
like
that's.
The
thing
we
consider
we
can
consider
is
is
trying
out
what
pidgey's
like
on
top
of
that,
but
it
doesn't
have
in
development
software.
Yet
it
has
to
be
well
published.
D
You
know
not
saying
pidgey's
poorly
published,
but
like
the
feedback
loop
on
building,
something
that
only
runs
in
the
browser
is
is
very
critical.
So
having
that
read
the
docs
thing
or
whatever
I
mean,
read
the
docs,
so
that's
that's
a
thing
here.
That's
actually
a
there's.
A
topic
up
on
the
community
call
right
now
that
got
started
by
by
who's
who
started
it.
Who's
who's
up
there,
chris
doctor,
hair
doctor
hold
a
graph
and,
and
you
be
chimed
in
a
bunch
of
other
people,
found
it.
D
You
want
you
like
all
right,
you
don't
have
to
run
webpack,
but
you
do
have
to
go.
Get
a
bunch
of
paper
condo
packages
to
extend
out
your
ui
because
there
is
not
there
is
today.
There
is
not
a
globally
trustworthy,
namespaced,
efficient,
cdn,
that's
not
going
to
break
on
you
tomorrow
or
that
we're
like
you,
can't
trust
unpackage.
You
can't
trust.
Js
deliver
all
all
that
stuff.
D
B
It's
a
fork
of
the
jupiter
light
demo,
repo,
okay.
D
D
You
know,
but
the
the
ui
and
the
ux
that
you
get
out
of
out
of
publishing
stuff
on
on
read
the
docs
is
really
good
and
you
know
they've
got
they've,
got
nice
features
and
stuff,
so
yeah
somewhere
in
there.
I
don't
know
if
it's
going
to
be
self-publishing
that
you
somehow
can
do
through
your
browser.
D
Well,
it
used
webdav
man.
I
wish
that
we
had
never
built
jupiter
contents
and
it
had
only
been
web
dev
since
the
beginning,
like
web
dab,
is
so
awesome.
It
handles
so
many
problems,
but
it
would
have
made
us
lazy
because
it
can
list
like
there's
all
these
weird
things
that
that
come
out
of
the
web
of
that
actually
makes
sense
when
you
get
to
a
system
that
doesn't
have
those
limitations,
but
you
have
to
pay
for
it
or
like
every
time
you
list
your
s3
buckets.
D
C
A
one-liner
right,
what's
the
name
of
that
comedian
isabella,
who
did
the
three
months?
C
I
don't
remember
neil
neil
brenner,
something
like
that.
He
has
this
good
special.
It's
like
he's,
got
three
mics
one's
for
short,
jo
one-liners,
medium-sized,
jokes
and
then
like
long
ones.
So
you
know
what
you're
getting.
This
is
a
one-liner
here.
So
okay,
so
there
are
tons
of
schema
online
yeah,
I'm
obsessed
with
schema,
but
they
it
serves
a
purpose
here.
So
this
I
wrote
a
library
recently
it
passes
the
entire
json
schema
test
suite
it.
C
I
rewrote
all
of
validators
because
well,
why
wouldn't
you
so
anyway,
when
I
run
this?
What
just
happened
so
I
pulled
in
a
thing
that
has
schema
from
the
web,
but
what
it
actually
is
is
a
fully
fledged
python
type.
So
basically
it
went
and
built
you
a
full
python
type.
It
does
if,
if
then
else
logic
all
of
logic,
it
does
all
the
properties.
Things
does
all
the
references,
dynamic
references
so
on
so
forth.
C
C
But
the
reason
I've
been
thinking
about
this
is
because
y'all
have
gone
and
put
python
and
javascript
and
rust
and
prologue
and
sql
and
everything
all
in
one
place
now
and
everything
has
its
own
type
system
but
like
how
are
we
like
what
is
life
like
when
we're
managing
like
all
of
these
types
in
all
of
these
places-
and
I
don't
know-
I
think,
there's
a
lot
to
think
about
like
when
you're
in
jupiter
light
land
and
we're
doing
json
stuff
all
the
time
and
it's
like,
can
I
just
have
my
type
system
as
something
I
import,
but
yeah
I've
been
using
this
because
it
lets
me
turn
things
into
traitlets
widgets,
clis,
ui
ton
tons
of
stuff,
but
that
was
my
short
little
guy.
D
Yeah,
that's
the
the
the
the
types
thing
are
one
of
the
underlying
requirements
for
that
webassembly
component
system.
D
You
know,
city
of
webassembly,
that
lives
in
your
browser
you'd
be
able
to
talk
over
these,
these
well-defined
types
and
yeah
at
the
bottom
of
it.
It's
like
you
know,
unsigned
int
and
boolean
and
and
string
right
like
utf-8
string.
It
doesn't
take
that
many
things.
You
know
data
shaped
kind
of
kind
of
stuff.
It
doesn't
take
that
many
things
to
start
at
the
bottom,
and
I
just
don't
know
how,
if
it's
going
to
get
up
to
where
it's
like,
what
is
a
markdown
type
right?
C
Right
yeah,
it's
interesting,
like
open
api,
has
really
pushed
where
json
schema
has
gone
like
there's
been
a
lot
of
like
company
and
community
investment,
especially
around
open
api,
like
that
system
got
so
popular
that
people
really
wanted
it
to
be
extensible
and
stuff
like
that.
So
when
you
bring
in
like
the
standard,
json
schema
and
open
api
into
this
mix,
you
can
define
pretty
much
most
of
well
and
like
a
json
table
schema
or
something
like
that.
You
can
pretty
much
define
most
of
the
at
least
type
annotations
or
type
systems.
C
I
don't
know
how
to
deal
with
the
multiple
language
thing
I
feel,
like
it's
tech,
slash,
markdown,
plus
some
variant.
Some
vendor
thing
right
because,
like
quarto
actually
maintains
a
fork
of
a
common
mark,
spec
miss
just
has
their
own,
like
ht
json
blob
that
they
validate
against.
I
think
it's
almost
like
up
to
the
variant
with
that
one,
that's
a
hard
problem.
D
Yeah
angus
and
I
still
haven't,
figured
out
anything
good
for
that
over
on
jupiter
markup
he's
almost
done
with
his
his
dissertation,
though,
which
means
you
know
what
that
means
great
time
to
write
some
scientific
software.
Oh
yeah
right
as
we
all
know,
that
is
when
the
best
software
gets
written.
D
No,
I'm
just
saying
that
the
old
story,
that's
why
we
have
ipython
in
the
first
place,
because
fernando
wrote
it
while
procrastinating.
I.
A
D
E
D
A
It
yeah
it's
a
mobile
of
jupiter
and
all
the
moons
right,
yeah,
perfect
yeah.
Well,
on
that
note,
thank
you
all
for
being
here.
This
is.
I
had
a
lot
of
fun.
This
call.
I
really
appreciate
the
open
discussion
too,
so
yeah
in
that
case,
we'll
have
a
call
in
a
month
that
will
be
in
july
time
to
be
decided
so
we're
all
on
this
adventure
together.
Thank
you
so
much
for
your
time
today
and
I
hope
you
have
a
great
rest
of
your
week.