►
From YouTube: 📦 Package Managers WG Weekly Sync October 8, 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
B
A
little
bit
things
are
are
mostly
solid
and
gonna
be
paused
for
a
while,
while
I
move
on
to
other
things,
there's
some
pub/sub
stuff
getting
finished,
which
is
just
mostly
gonna
end
up
being
like
just
a
couple
of
rounds
of
like
just
back
in
back
and
forth.
The
remaining
non-trivial
thing,
I
guess,
is
efficiently
caching
pairs
that
we've
tried
to
discover
already.
B
Basically,
if
you
build
a
cache
and
the
cache
returns
multiple
pairs
and
then
this
thing
is
gonna
be
accessed
from
multiple
different
threads.
Then
there
needs
to
be
some
sort
of
state
in
there.
That
says,
like
I,
have
already
tried
this
and
maybe
some
sort
of
back
off
of
like
when
to
try
again
Steven
was
concerned
about
what
happens
when,
like
how
frequently
do
we
try
connecting
two
pairs
that
we've
already
tried
and
failed
to
connect
to
before?
A
A
B
A
pretty
good
start,
but
if
it's
gonna
need
a
bunch
of
time
and
I,
don't
think
this
corner
is
necessarily
gonna
have
that.
So
what
are
you
working
on
this
quarter
yeah?
So
this
quarter
is
between
now
and
whenever
the
testing
infrastructure
is
good
enough.
For
me
to
start
hammering
on
DHT
things,
it's
gonna
be
some
amount
of
like
miscellaneous
stuff,
plus
IP
FS,
add
performance.
B
There's
the
config
file
stuff,
which
needs
to
get
fixed
so
that
we
can
actually
you
know,
I
have
the
defaults
normally
like
have
actually
controllable
defaults
and
be
able
to
set
things
to
true
by
default,
would
be
nice
and
I.
Don't
know,
I'm
helping
daveed
figure
out
how
to
spin
up
like
give
us
instances
for
for
testing
things.
I
can
get
into
that
a
little
bit.
So
this
is
a
bunch
of
like
miscellaneous
stuff
and
then
start
looking
at
performance.
More
okay,.
A
A
Okay,
so
the
next
section
is
bit
swap
updates,
so
yesterday,
I
kind
of
finished
off
like
really
decent
version
of
the
that's
my
work,
I've
been
doing
and
collaboration
partner
ran.
It
I
found
that
it
was.
It
was
a
lot
better
and,
what's
out
there
master
sorry
so
that
this
week,
I
think
I'm
gonna
concentrate
on
like
really
making
a
bullet
for
fixing
up
all
the
tests
and
everything.
A
B
Yeah,
yes,
I
mean
we
don't
this
now
honoring
your
thoughts
on
I
as
I
mentioned
yesterday.
The
I
suspect
that
when
the
PAP
is
ready
to
put
more
time
into
pub/sub,
they're
gonna
have
to
think
about.
Like
hey
this,
randomly
running
into
peers,
isn't
working,
we
need
to
form
trees
and
they're
gonna
want
to
use
the
same
metrics.
In
any
event,
the
entire
system
is
just
like:
send
a
message,
get
a
message,
so
I'm
wondering
if
there's
a
way
to
like
it,
even
makes
sense.
A
B
Not
clear
I
mean
so
there's
some
stuff,
that's
like
already
there.
For
instance,
the
have
like
you
know,
I
have
I
want,
is
already
like
part
of
pub/sub,
like
they
already
went
in
like
in
a
sense
like
reimplemented
that
I
don't
know
when
they're
gonna
get
around
to
doing
the
tree
broadcasting
I
imagine
it
will
be
as
soon
as
somebody
starts
trying
to
use
this
in
production
and
realizes
that
randomly
running
into
peers
is
not
what
they
wanted.
B
B
The
things
are
not
ordered.
They're.
Just
like
this
is
the
number
we
hope
they're
incrementing,
but
there
was
someone
Stephen
and
some
other
folks
have
been
interested
in
like
having
pub/sub
but
using
like
cid
like
having
a
sort
of
reliable
in
a
sense
pub/sub,
where
I
just
have
like
a
chain
of
I
have
a
CID.
That's
just
like
a
growing
list
of
all
of
the
prior
messages
that
I
cared
about,
or
maybe
they
don't
have
to
care.
B
Maybe
I,
maybe
I
decide
it's
not
reliable
and
I
just
send
you
see
IDs
for
each
message
right:
okay,
interesting
and
one
of
the
someone
tried
to
implement
this
in
Russell.
It
p2p
I,
don't
think
that
PR
got
merged,
but
they
wanted
to
do
that
as
an
option
and
I
think
that
it's
probably
worth
doing.
B
A
B
I
mean
the
bit
soft
case
is
obviously
like
different
in
that
you
only
have
one
layer
to
traverse
like
I'm,
only
directly
talking
to
my
peers.
They
might
happen
to
also
find
out
about
data
from
someone
else,
but
there's
no,
like
the
requests
aren't
being
propagated.
They're.
Just
it's
just
one
layer.
B
A
C
Well,
it's
a
good
question.
I
was
just
reading
back
on
the
issue
and
the
last
kind
of
note
goes
to
the
design
direction
was
that
there
was
a
sync
discussion
in
which
it
was
decided
to
do
it.
The
other
way
which
oh,
it's
just
you
know
just
pink
thing
on
there
and
just
said.
Has
it's
been
documented
anyway,
because
I
always
know
it's
all
in
Stephens
head.
B
D
C
B
D
A
Okay,
so
I'm
gonna
kill.
You
have
any
updates
on
the.
D
Amount
I
do
so
yesterday.
I
spent
some
time
basically
going
over
some
of
the
feedback
that
I've
gotten
on
it,
particularly
in
relation
to
like
how
memory
is
allocated
and
how
references
are
derived
and
stuff
like
that,
and
then
I
generalized
a
lot
of
the
traversal
logic.
So
everything
basically
just
implements
this
interface
and
it's
good
to
go
messing
with.
That
seems
quite
a
bit
better.
D
However,
I'm
getting
some
weird
issues
around
returning
or
operating
on
the
wrong
reference,
so
something's
opened
and
then
right
is
called
on
it
and
it
thinks
that
it's
not
open
because
something's
not
signed
properly
so
I'm
looking
into
that,
but
outside
of
outside
of
some
of
the
right
bugs
seems
to
be
going
well.
So
that's
around
that
on
that
all
right.
Any
questions
for
the
diamond.
A
B
B
Look
at
you
know,
maybe
like
one
of
these
is
you
know
about
we
paralyzed
rights,
but
we
sort
of
synchronously
flushed
everything
to
discs
that
sort
of
is
gonna
end
up
like
choking
it
or
you
know
a
few
other
like
miscellaneous
things
like
that
I
sort
of
just
need
to
get
my
like
set
up
for
testing
like
I
need
a
you
know,
set
up
some
scripts
that
are
gonna
like
actually,
let
me
go
test
this
and
see
how
it
goes
and,
and
then
we'll
we'll
see
what
helps
I.
B
A
I
created
a
PR
I,
don't
know
if
you
can
look
at
that
and
which
is
extremely
rough.
It's
just
like
you
know
me
thrown
together
ideas,
but
inside
that
PR
it
like
it'll,
actually
set
up
like
a
data
structure.
You
know
you
know
you
can
set
up
different
ones
at
different
sizes,
so
you
can
test
out
kind
of
different
scenarios
for
ad,
so
that
may
or
may
not
be
useful.
I
mean
you
might
just
want
to
use
I.
A
A
B
A
B
A
B
A
B
So,
yes,
maybe
a
question
for
you
der
is
like
so
I
talked
to
Steven
about
this
I
was
a
little
skeptical
of
working
out
before
missus
quarter.
Just
because
of
how
long
it
took
you
to
get
started
and
then
being
like.
So
if
I
spend
a
month
getting
into
this
and
then
I
haven't
pivot
to
the
DHT
thing
like
I've,
we
now
have
two
people
that
are
like
know
what's
going
on,
but
not
had
time
to
fix
it.
B
It's
the
best
use
of
time
like
and
Stevens
guess
was
that
there's
probably
like
fixing
it
TM
is
like
or
making
it
much
more
efficient
is,
could
be
a
while,
but
that
there
might
be
like
one
or
two
of
these
things
that
might
in
Vil
result
in
like
a
pretty
dramatic
speed-up.
A
Think
basically,
the
conclusion
we
came
to
is
we're.
Gonna
need
to
mess
around
with
the
way
it
works,
so
I
don't
think
it's
gonna
happen
fast,
honestly,
I
think
at
least
if
the
last
thing
we
were
discussing,
that
issue
was
like
changing
fundamentally
the
way
that
it
it
sort
of
ordered
stuff.
So
you
know
in
order
to
properly
test
that
I
think
it's
probably
gonna,
take
a
while.
A
B
A
B
C
B
With
with
that
talk,
I
don't
think
I
mean-
maybe
it's
just
me
I
couldn't
do
it
I
feel,
like
you,
gotta
look
at
some
of
the
slides
also
and
be
like
okay.
This
is
what
they're
doing
I,
don't
think
it's
actually
necessarily
so
dissimilar
from
what
we
want
to
do
anyway.
I
think
it's
just
a
particular
like
maybe
I
misinterpreted
it,
but
I
think
it's
just
a
particular
flavor
on
it,
which
is
like
okay,
like
that's,
why
I
was
trying
to
compare
it
to
like
the
different
chunker's.
B
It's
like
yeah
I
mean
you
can
have
the
same
data
structure
and
then
glue
a
bunt.
If
they
have
links,
you
can
glue
the
links
together
and
like
arbitrary
ways
right,
but
that
that
kind
of
structure
might
be
efficient,
I,
don't
know
it's
and
they
were
they
were.
It
was
interested
in
like
more
sequential
things
right.
How
do
you
I
want
to
ask
for
a
bite
range
right,
which
I
think
UNIX
must
be
to
you
want
to
do
in
general
is
like,
because
all
of
your
file
system
commands
operate
on
a
bite
range
right.
D
B
Some
percentage
of
the
data
will
become
inaccessible
to
a
new
version
of
the
same,
a
new
formatted
version
of
the
same
file,
which
is
like
really
annoying,
and
so
that's
why
we
don't
want
to
like
iterate
on
this
quite
as
rapidly
as
maybe
we
do
other
stuff,
but
which
is
why
I
think
1.5
is
yeah.
Okay,
like
everyone
knows,
this
isn't
gonna
last
very
long.
Let's
do
it,
but
like
I,
don't
know,
I,
don't
know
how
long
it's
gonna
last
right.
D
It's
definitely
an
issue.
I
remember,
like
kind
of
tangentially
related
someone
was
mentioning
about
schema
changes
between
like
different
schema
versions
and
things
like
that
and
migrating
from
or
having
interrupts
between
newer
versions
of
applications
with
newer
versions
of
the
schemas
still
talking
to
the
older
ones
like
that
and
while
I
don't
know
what
much
came
of
that
conversation.
I
know
that
Eric
Meyer
was
excited
to
talk
about
that
with
this
person,
so
hopefully
things
of
that
nature
are
being
considered
so
that
we
don't
end
up
in
situations
like
that.
D
D
Yeah
I
imagine
the
geeks
people
and
all
that
are
probably
very
focused
on
similar
things
to
of
like
when
it
comes
to
reproducibility
right,
like
I,
want
to
just
always
construct
the
same
exact
object,
regardless
of
what
I
am
capable
of
I
guess.
If
I
have
a
new
schema,
I
might
still
only
use
the
old
one
and
all
that
I.
B
A
Minutes
left
so
I'm
going
to
just
things
we
did
before
the
ends.
So
one
is
last
week
I
think
Steven
was
asking
about
whether
whether
we
feel
like
this
meeting
is
like
useful.
If
you
want
to
keep
doing
every
week,
is
there
an
to
have
a
strong
opinion
about
that
at
the
moment,
I.
B
C
A
C
A
D
A
A
B
Wanted
to
mention
Alex
that
if
you
are
interested
in
getting
all
in
the
coat
and
then
go
code
buy
stuff,
it
can
be
a
little
rough
and
it's
like
yeah.
It
can
be
a
little
rough
so
like
recommend
reaching
out
to
a
person
cuz
the
examples
like
the
Lib
p2p
examples
have
gotten
better.
The
ipfs
examples
are
still
not
great,
because
there's
like
designed
for
people
who
want
to
use
like
the
daemon
or
used
like
the
HTTP
API
to
talk
to
the
daemon
and
you're
like
okay
I,
couldn't
just
done
this
from
JavaScript
I.