►
From YouTube: 📦Package Managers WG Weekly Sync April 30, 2019
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
B
Hello,
so
I
had
some
fun.
This
week,
I
wrote
my
first
bit
of
rust
and
contributed
to
the
weapon
package,
manager'
client,
to
be
able
to
add
to
install
a
particular
version
of
a
thing
which
is
pretty
cool
that
the
original
client,
as
it
was
published,
was
very
bare
and
is,
and
if
I
opened
a
few
issues
asking
for
some
features,
they've
been
very
responsive
and
then
we
had
a
call
with
them
on
Thursday
to
discuss
how
to
add,
ipfs
support
to
weapon
and
the
different
levels
of
integration.
B
That
could
happen,
which
I
then
kind
of
turned
into
an
issue
of
like
here's,
some
of
the
different
stages
or
the
ways
you
could
go
about
it,
which
I'm
planning
on
kind
of
then
delving
in
a
little
bit
further
on
stage,
one
to
literally
like
lay
out
his
some
of
the
commands
that
you
would
run
either
over
HTTP
API
or
a
if
you're
in
the
client
like
what
API
calls
to
make
to
help
them
get
tarballs,
hosted
on
ipfs
and
then
kind
of
1.5.
I
guess
is
thinking
about
how
the
client
can
integrate.
B
It
has
some
really
interesting
features
from
a
usability
perspective,
but
it
throws
out
loads
of
reproducibility
problems
out
there
kind
of
as
a
trade-off
that
it
feels
like
ipfs
can
I
can
help,
but
it
doesn't
feel
like
they
want
that
help
right
now,
so
is
a
different.
I
could
see.
Michaels
just
tend
his
video
because
he
has
opinions
or
incited
knowledge.
They'll
be
interesting.
B
C
I
mean
I,
know,
Ryan,
really
well
and
I've
known
him
through
creating
this
and
stuff,
and
he
yeah
he.
He
has
really
strong
opinions
about
package
managers
that
basically
boil
down
to
all
the
best
practices
that
we've
designed
over
the
last
ten
years.
He
doesn't
like
so
I
was
like
yeah,
I'm
necessarily
bet
on
him.
Agreeing
with
these
are
like
the
problems
that
need.
B
But
I
think
he's
got
a
note
to
talk
about
that
in
a
minute
and
the
other
thing
that
I
have
not
done
much
further
work
on
is
the
ipfs
concepts
mapping
document,
but
that
is
basically
where
I'm
drawing
all
of
the
weapon
related
kind
of
documentation
that
I'm
putting
together
for
them.
It's
just
focused
on
their
use
case
specifically,
which
is
a
database
backed
package
manager
to
kind
of
almost
try
it
out
to
see
how
much
I'm
missing
so
I'll
be
carrying
on
working
on
that
and
that's
about
it.
B
B
D
It
would
be
helpful,
so
thank
you
for
the
info
about
gosh
gosh
continuing
to
work
on
the
concepts
mapping
stuff,
because
eventually
I
would
like
to
take
a
little
bit
of
a
stab
at
integrating
that
into
some
of
the
work
that
I
would
like
to
be
doing
eventually.
D
If
that
ends
up
being
something
that
happens,
I
know
that
Molly
you've
got.
That
is
a
suggested
workshop
on
the
agenda
I'm.
So
looking
forward
to
seeing
how
that
shakes
out
and
then
the
stuff
I
want
to
be
working
on
in
the
next
couple
of
days
is
starting
to
work
on
user
segmentation
for
package
manager,
users
guide
tied
to
Andrew.
D
F
So
I
hosted
some
of
the
documentation
that
we
talked
about
last
week.
So
that's
an
issue
now
to
go.
Live
there
that
cladistic
tree
thing
I
still
feel
like
there's
so
many
different
ways
that
could
go
like
a
feather
on
the
end
of
it
like
it.
Just
got
up
to
the
point
where
we
discuss
what
decentralized
authorship
would
look
like.
I'm
gonna
stop
right
there,
so
I'd
love
to
see
somebody
take
that
and
run
with
it
someday.
F
So
all
these
concepts
like
a
lot
files
and
stuff
are
all
buried,
but
basically
comes
with
the
idea
that
you've
got
this
one.
Repo,
probably
is
on
github,
and
it's
gonna
produce
like
one
thing
and
that's
the
package
management
that
we're
talking
about
and
in
distress
just
resembles
something
totally
different
and,
like
other
interesting
concepts
that
those
people
in
user
stories
care
about
that
single
project
package
management,
just
like
doesn't
have-
or
vice
versa-
and
the
answer
is
like
yes,.
F
F
F
B
A
quick
note
in
the
papers
that
I've
read
about
Paul
a
wild
all
he's
in
the
room
is
the
papers,
I've
read
about
package
round
room
or
almost
always
have
kind
of
a
distinct
angle,
very
rarely
cross
between,
say,
distro
package
managers
and
application
level
package
managers.
So
this
this
potential
into
ding
areas
of
kind
of
research,
of
the
crossover
or
lack
of
crossover,
as
as
Eric
mentioned
there
that
as
haven't
seen
but
I
may
do
another
troll
of
the
of
the
published
papers
to
see.
If
there's
anything
else
interesting,
there.
G
I
think
or
something
in
in
chatting
with
Stephen,
and
we
were
kind
of
brainstorm
and
all
these
different
who
ski
says
things
like
big
big
Linux,
builds
like
downloading
all
of
Ubuntu
and
stuff
like
a
different
characteristic
in
terms
of
size,
and
so
the
cost
of
say
downloading.
The
same.
Linux
builds
on
20
machines
in
a
computer
cluster
or
no
other
scenarios
we're
getting
this
sort
of
builds
into
an
environment
that
is
relatively
internet
or
bandwidth.
Constrained
starts
to
have
like
pretty
significant
impact
of
the
sort
of
deduplication
and
overhead
and
peer-to-peer
bandwidth
direction.
E
Yeah
in
scientific
high-performance
computing
with
Linux,
it's
typically
that's
a
core,
there's
a
usually
something
in
the
cluster
that
you
know
has
a
copy
of
all
the
repos
and
things,
and
so
that
that's
actually
a
place.
You
can
search
for
a
lot
of
prior
art
for
this
sort
of
thing,
because
it's
it
kills
you,
if
you
have
to
five
hundred
nine
cluster
you're,
not
going
to
download
the
same
packages
five
hundred
times
off
the
internet.
E
A
F
Gonna
talk,
I,
dropped.
Another
note
I've
got
a
similar
thing
in
in
our
shared
pad.
There's
the
project
from
uber
called
crackin,
that's
about
distributing
with
docker
images
inside
of
a
cluster
I
forget
if
we
can
mention
em
before,
but
it's
an
interesting
example
that
we
were
just
talking
about
like
it's,
not
exactly
that
they're
been
with
constraints
just
that
they
absolutely
know
they're
doing
the
same
thing
repeatedly,
and
so
it
makes
a
lot
of
sense
to
try
to
like
turn
this
into
a
small
operation.
G
E
Yeah
so
as
in
the
the
the
in
web
browsers
meeting
and
we're
talking
about
web
package
and
then
we're
all
talking
about
hey,
you
know
this
package
managers
is
the
focus,
but
it's
like
web
package
like
like
what
Google
is
doing
in
chrome,
is
they're
taking
websites
and
attorney
websites
into
packages.
So
it's
sort
of
funny
because
it's
like
a
crossover
with
what
we're
doing
here
so
with
the
with
the
HTTP
signed,
exchanged
things
we're
essentially
taking
a
bunch
of
what's
a
website
putting
into
essentially
the
exact
same
format
that
you
would
distribute.
E
E
B
You
there's
a
interesting
hook
back
in
there
to
Dino
as
well
Jim
that
if
you
just
treat
your
packages
as
URLs
that,
if
you
can,
if
you
can
get
them
to
work
in
a
similar
way
and
you're
acting
like
a
web
browser,
then
perhaps
you
can
be
like.
Oh,
you
can
fetch
this
from
anywhere
as
long
as
it
is
correctly
signed,
which
gets
pretty
close
to
ipfs
style
thing.
C
You
know
exists
in
this
weird
place
where
they're
automatically
translating
stuff
in
typescript,
and
so
it's
just
not
going
to
mesh
very
well
with,
like
all
of
these
newer
style,
important
module
techniques
that
need
only
a
browser
and
only
something
like
signed
exchanges,
because
there
there's
always
this
compile
step
in
between.
So
it's
like
I,
don't
want
spend
too
much
time
trying
to
do.
You
know
there's
like
not
really
anybody
using
it
yet
and
yeah.
C
C
A
G
That
kind
of
goes
back
to
kind
of
the
work
kind
of
a
cladistic
tree
and
and
the
like
the
different
layers
of
package
minute
and
stuff
like
what?
What
do
you
need
for
layer,
one
to
be
effective?
Well,
you
needed
to
be
fast
because
content
store
you're,
fetching
content
like
what
do
you
need
for
layer,
two,
alright,
now
you're
trying
to
like
mirror
registries
and
have
your
index,
and
then
you
know
layer,
three
yeah
decentralized,
publishing
some
amount
of
identity
flows
into
there.
G
I
think
question
mark
to
us
is
like
having
having
a
couple
of
flavors
of
each
of
these
would
be
really
interesting
to
like
compete
against
each
other
of
like
how
much
is
necessary
in
order
to
get
something
that
could
even
possibly
function
and
how
much
is
ideal
in
order
to
make
something.
That's
like
really
really,
you
know,
fits
everyone's
mental
model
in
the
package.
Manager'
worlds,
and
we
can
you
know
when
we
have
like
stages
within
each
of
the
stages.
G
A
F
I
feel
obliged
to
actually
push
back
on
that
ever
so
slightly
I
have
a
bunch
of
use
cases
and
reproducible
builds
where
I
don't
actually
give
a
about
passages
at
all
I
care
that
it's
reliable
and
I
care
that
it
gives
me
an
integrity
guarantee
without
pushing
further
work
up
to
higher
levels
of
my
stack.
It's
fast,
that's
nice,
but
it's
actually
I
can
wait
for
speed
optimizations
to
come
down
the
pipe
later
I.
F
So
late,
so
it's
interesting
because
I
can
concretely
imagine
an
application
where,
for
example,
the
IP
NS
resolved
speed
I
just
don't
care,
because
what
I
want
instead
is
the
snapshot
of
the
index
by
a
specific
cid
already
so
like
IP
NS,
just
don't
care
and
maybe
then
I
care
about
the
speed
of
the
diffuse
mount.
But
instead
I
might
have
some
of
these
case
where
I
can
like
I'm
doing,
reproducible
builds
like
literally
I'm,
going
to
build
the
same
thing
repeatedly
right.
F
So
I
can
just
copy
files
out
of
the
fuse
mount
onto
a
standard
amount
of
whatever
other
performance
characteristics.
My
underlying
disks
have
and
I
can
cache
that
by
the
CID,
which
I've
already
got
because
the
system
is
nice
and
so
I
don't
care
about
the
speed
of
that
either.
Like
I
can
immunize
myself
to
a
lot
of
this
stuff.
F
If,
at
a
super
high
level,
I
happen
to
have
the
the
cool
selection,
algorithms
figured
out,
cuz
I'm,
doing
a
distro
scale,
stuff
and
I
try
to
make
sure
that
I'm
using
the
minimum
spanning
selection
of
unnecessarily
different
versions
of
some
package,
then
that
just
increases
my
cache
a
way
for
all
these
other
things.
That
annoys
me
against
not
about
speed
of
the
transport.
So,
if
I
take
all
those
together,
yeah
I
get
rid
of
a
lot
of
the
speed
constraints
in
such
an
application.
F
C
C
So
for
taking
like
a
a
bulk
chain
set
to
to
a
large
piece
of
data
or
or
like
you
know,
if
we
want
to
take
the
last
10
seconds
of
MDM
activity
and
an
update,
like
you
know
the
whole
graph
for
all
of
them
game
or
whatever
it's
just
much
more
efficient
to
do
those
in
batch
operations
and
we're
going
to
write
api's
to
make
that
a
lot
better.
That
caches
that
they're
going
to
be
built
on
top
of
the
data
model
layer.
C
C
C
That's
how
like
a
lot
of
data
bases,
actually
work
when
they
have
like
big
block
states,
they're,
just
like
okay,
well,
while
I'm
waiting
for
things
to
update
I
just
batch,
all
the
rest
of
the
updates
and
then
I
can
do
the
next
batch
of
updates
quicker
once
that
returns
like
that
is.
That
is
probably
how
NFS
should
work
like
at
some
point
in
the
future.
Using
some
of
these
new
tools
that
we're
building
I.
G
G
Think
recap
from
the
call
like
I
think
it
went
super
well
they're
like
really
excited
they're
like
great
we're
gonna
like
start
integrating
ipfs
into
our
data,
like
our
content,
store
layer
like
in
a
week
now
like
a
week
from
now,
which
is
awesome
and
super
exciting,
and
so
definitely
like
work
on
our
plate
to
make
sure
they
have
like
a
super
great
experience
and
also
to
learn
from
this,
because
you
know
the
end
goal
is
not
a
single
package
manager.
Who
has
this
it's.
G
You
know,
understanding
the
pain
points
that
folks
run
into
and
solving
them,
not
just
for
package
managers,
but
for
all
these
other
use
cases
and
so
I
think
a
lot
of
a
lot
of
things
like
having
a
fast
channel
with
them
and
quickly
understanding
like
where
they're
heading,
both
from
a
design
space
from
like
a
technical
space
from
like
an
emotional
space
like
where
they
would
tend
to
go
and
what?
What
sort
of
information
is
really
valuable
to
them?