►
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
Selectors
are
a
long-standing
thing
that
we've
wanted
for
many
many
years.
There's
a
lot
of
conversations
gathered
over
the
years
as
to
what
we
want.
We've
tried
a
number
of
times
to
come
up
with
some
languages
and
its
indexes
that
would
kind
of
satisfy
a
lot
of
the
requirements
and
have
so
far
been
unsuccessful.
A
So
this
spec
begins
by
kind
of
articulating
what
what
selectors
are
and
links
to
a
bunch
of
prior
work
that
expands
a
lot
on
the
goals
of
selectors
installing
this
crisis,
a
few
kind
of
design
notes
for
what
we
would
want
out
of
time,
selectors
and
then
it
kind
of
narrows
on
the
problem.
By
saying
what.
A
A
That's
still
a
pretty
hard
problem,
there's
a
lot
of
requirements,
I
sorts
these
and
run
them
through
a
few
people
right
and
by
a
lot
of
people,
and
we
weren't
able
to
find
more
yet
there.
Many
other
brothers
collectors,
sorry
other
requirements
for
inspectors,
but
these
are
roughly
roughly
what
they
are
and
what
these
boil
down
to
is
that
you
can't
actually
come
up
with
one,
because
we
have
some
departments
around
a
great
ability
of
this
system
in
the
long
term,
and
we
want
to
be
able
to
learn
from
from
three
different
systems.
A
And
then
we
have
the
ability
to
have
different
types,
and
we
can
have
some
very
basic
selectors
that
we
should
buy
the
bulbs,
but
for
creative
process
by
which
other
groups
can
other
participants
and
so
on,
can
create
new
selector
types
and
contribute
them
to
the
to
the
system
and
then
over
time.
If
some
of
those
become
really
popular,
then
they
can
become
part
of
of
standard
libraries
and
so
on,
and
so
what
this
means
is
that
you
know
we
use
typing.
A
We
have
a
system
of
selector
types
that
allows
creative,
you
select
languages
and
synthesis
and
the
can
compose
them,
which
is
an
important
part
of
that.
We
create
an
easy
path
for
plugin
selector
text
into
a
Quixote
libraries
and
other
consumers,
and
then
we
need
a
light
process
for
testing
and
submitting
new
selector
times.
So
there's
a
this
outlines
a
whole
bunch
of
work
ahead
for
the
selector
effort,
which
means
hey.
We
need
a
well-defined
binary
human
human,
readable
cell
description
for
types.
That's
easy!
It's
just
a
multi
codec!
We
have
a.
B
A
Selector
interface
for
most
uses
of
selectors,
so
that
way
all
systems
that
need
to
use
the
lectures
things
like
grass,
bank
or
cluster
for
pinning
or
ID,
because
we're
pinning
providers
for
expressive
private
records-
or
you
know
things
like
Paco
and
when
they
need
to
synchronize
a
chain
or
some
other
things
like
that.
We
just
use
that
interface.
So
that's
a
various
obtain
a
description
of
what
is
lekker,
isn't
us,
and
then
we
need
a
senate
way
to
add
peace,
electric-type
implementation
to
ideally
faculty
library.
A
So
in
an
ideal
library
we
export
the
ethically
so
like
our
interface,
and
we
need
a
way
to
add
type
implementations
to
that
libraries.
So
that
just
means
like
a
simple
way
of
defining
the
modules
as
much
that
you
can
insert
new
types
over
time,
and
so
the
users
of
the
library
can
do
that
as
well
to
experiment
with
things.
A
Basically,
as
long
as
possible,
the
abstract
selector
type
and
we
don't
and
bad
some
like
specific
thing
all
over
the
place,
we
came
up
with
a
number
of
symbols.
Electrodes
that
cover
most
of
the
common
cases,
see
II's
paths.
Globs
I'll
show
you
a
few
more.
In
a
moment.
We
have
a
selector
type
that
allows
composing
selectors.
So
this
means
some
way
so
I
feel
that
handles
multiple
slackers.
The.
A
Selector,
that
might
change,
because
to
avoid
confusion
with
the
motor
formats
well
and
then
two-point
aims
are
that
one
we
want
to
aim
for
language,
independent
limitations
of
selectors,
so
think
of
writing,
parsers
and
execution,
and
so
on
in
a
language
agnostic
way,
so
I
work
in
a
portable
way.
Maybe
it's
the
best
better
way
to
describe
it,
and
that
would
be
no
long
term.
A
Things
like
blossom
but
in
the
meantime,
really
be
fine,
allowing
language
specific
implementation
of
stuff
selectors,
which
are
is
what's
gonna
happen
for
those
simple
ones
or
CDs
and
pads,
and
maybe
blobs
like
we're
just
gonna
right
at
them
chance
at
going
that's
much
easier
than
trying
to
like
shoehorn.
Some
wants
other
thing
into
either
of
these
systems
and
not
a
cover.
So
the
claim
is
that
CDN
path
might
cover.
A
A
A
A
Question
marks,
matching
exam
character
and
so
on
and
you'll
notice
here
that
that,
in
this
there's
kind
of
like
a
hint
at
what
the
binary
format
might
look
like
and
what
this
string
format
might
look
like
important
that
their
selectors
need
both
a
binary
and
a
string
representation,
because
it'll
be
a
lot
of
cases
where
selectors
get
dumped
out
to
the
user
and
the
user
needs
to
be
like
systems.
Building
on
selectors
need
to
just
count
on
the
fact
that
these
next
half
of
human
reader
works
with
each
other.
That's
not
gonna,
like
yeah.
A
There's
so
yeah
one
selector
that
just
it's
kind
of
like
a
nice
expression
that
applies
unions
in
effect
and
difference
over
selectors
and
then
one
part
of
the
spectate
respect
is
really
showing
what
these
interfaces
might
be
and
go.
So
this
is
meant
us
to
kind
of
like
a
general
thing,
but
it
became
like
more
just
real
quickly
becomes
like
close
to
any
permutation
yeah.
Your
prototype
is
your
implementation,
and
this
kind
of
goes
through
what
a
bag
is
and
kind
of
boils
down
to
a
simpler
interface.
A
There's
kind
of
a
proposed
traversal
special
powder
go
through
things.
This
is
kind
of
something
that
really
so
bunch
of
other
things.
We're
going
to
night
believing
you'll
to
crawl
through
that
similar
to
how
you
do
OS,
don't
walk
or
or
area
over
things,
and
then,
once
we
have
those
things
to
find
that
a
selector
is
pretty
simple.
It
just
checks
whether
it's
specific
path
is
part
of
that
selector
inside
of
the
bag,
and
that's
really
like
it
so
far.
It
has
these
other
got
kind
of
convenience.
Things
like
we
can
return.
A
The
route
of
the
selector
can
return.
The
type
can
return
it
its
bite
or
anything.
It's
v,
representation
straight
word
petition
and
there's
this
notion
of
whether
or
not
selective
absolute.
So
so.
The
important
thing
here
is
turns
out:
selectors
have
similar
to
paths
in
UNIX
file
systems.
They
could
be
absolute,
meaning
they're
like
routed
of
cid,
or
they
could
be
relative,
which
means
you
apply
them
on
top
of
any
word.
So
that
means
you
might
have
a
selector
that
you
apply.
A
On
what
you
like
business
yeah-
and
that's
it's
so
far
for
for
this
thing,
I
think
the
future
work
here
includes
defining
a
bit
more.
What
these
pieces
of
of
work
might
look
like
so
actually
going
through,
and
so,
instead
of
like
doing
it
just
yet
scoping
out
what
some
of
these
things
will
will
imply
and
kind
of
describing
requirements
for
things
like
that
that
we
libraries
you
have.
E
A
A
B
A
A
So,
for
example,
a
file
it
kind
of
is
like
one
object
right,
even
though
it
has
a
bunch
of
nodes
inside
of
it
or
I
started
directories
from
one
object,
so
you
would
want
to
be
able
to
traverse
things
like
this,
the
path
to
a
directory
that
happens
to
be
started
and
then
write
just
the
name
of
a
directory
and
then
move
forward,
and
then
that
might
actually
represent
a
whole
bunch
of
like
traversals
inside
of
a
larger
graph.
So,
ideally,
we
would
want
to
like
ours
over
over
those
higher
level
level
clusters
as
well.
A
They
can
likely
be
expressed
the
exact
same
way.
The
syntax
is
don't
change,
which
is
don't
change,
what
changes,
how
they
execute
and
how
they
apply
into
the
bags,
and
so
what
we
really
need
to
figure
out
to
be
able
to
make
that
possible
is
how
this
composite
object.
Data
structure
thing
how
that
behaves
as
an
actively
node.
If
that
whole
composite
object
can
satisfy
some
IP
LV
node
interface,
then
then
it'll,
then
then,
that
can
work
out
of
the
box,
meaning
I,
don't
think
as
a
selectors
problem.
That's
a!
A
E
A
A
A
I'm
saying
so,
I
should
probably
avoid
by
attending
here,
I'm
gonna
abort
and
say
that
I
that
is
definitely
concerned
and
I
think
you
can
need
to
be
addressed.
Evelators
of
data
structure
definition
selector
shouldn't.
We
shouldn't
have
multiple
types
of
like
core.
We
shouldn't
have
multiple
selector
levels
like
raw
I,
peel
these
selectors,
and
then
you
know
you
know
XFS
selectors
and,
like
a
whole,
different
family,
selector
daddy's,
going
to
get
massively
confusing
that'll,
be
like
having
different
types
of
different.
Like
packs
traversal
systems,
depending
on
which
file
system
you
have
HP.
A
A
F
B
G
A
There,
a
second
getting
balls
recording
a
new
sentence.
Are
there
more
questions
about
how
its
selectors
first,
so
you
before
you
Claus
I,
think
it
might
just
be
best
to
include
the
grasping
this
question
here,
because
a
lot
shorter
and
it's
really
dependent
on
all
this?
Okay,
so
having
it
separately.
If
people
will
be
like
what-
and
it
was
just
before
some
to
rewatch
with
all
this.
A
A
So
the
the
pulley
inception
of
graphs
and
close
to
getting
fit
swap
to
be
like
upgraded
with
selector
capable
stuff
and
downgraded
with,
like
all
of
the
market
dynamics,
meaning
they
stopped
originally
had
two
goals.
One
was
moving
around
just
simple
blocks
of
content
and
the
other
was
to
measure
there's
like
ratio
stuff.
A
A
So
this
right
now
is
a
much
earlier
piece,
but
the
whole
like
protocol
part,
has
been
figured
out,
and
so
this
is
taking
much
of
what
baseball
already
does
and
just
cleaning
up
a
certain
problem
around
requests
and
responses.
That
was
there
previously
and
adding
the
selector
piece,
and
so
this
is
work
that
I
did
with
Jeremy.
A
A
A
A
A
What
used
to
be
kind
of
been
a
very
bit
torrent,
inspired
thing
more
towards
HTTP,
API
is
on
our
pcs.
So
it's
a
slight
move.
It's
not
very
significant
good.
It's
still,
you
know
single
messages.
You
don't
need
to
track
the
notion
of
kind
of
requests
and
keep
them
open
in
the
same
way
as
HTTP
does
it's
still.
You
know
you
send
one
message
and
you
receive
a
message,
but
you
have
ways
of
communicating
about
those
those
requests
and
then
there's
like
some
some
of
the
kinds
of
pieces
responses
that
you
might
send
send.
A
So
this
I
went
through
this
with
the
people
making
sure
that
it
kind
of
covers
what
most
people
wanted
things
around,
just
information
like
being
able
to
send
additional
peers,
for
example.
This
is
about
a
common
device
where
you
know
that
somebody,
you
know,
other
participants
have
the
content.
So
it's
a
requester
passport
selects
the
responder
receives
electric
text
and
they
know
that
some
other
party
has
the
content
and
they
you
facilitate
that
distribution
without
having
to
resort
to
some
other
content.
Rhyming
system
like
with
UHD.
F
A
So
that
would
be
articulated
in
prose,
as
this
document
evolves.
That's
what
this
partial
and
terminal
mean
so
partial
means
that
I'm
sending
a
response
to
that
request
with
that
information
code,
but
it's
not
a
terminal,
so
it
might
send
you
as
many
of
these
as
I
want
until
I.
Send
you
one
of
these
terminal
ones
once
I
send
you
these
criminal
ones,
then
I
garbage
collect
all
notion
of
the
request
and.
A
So
this
probably
like
for
completeness
here,
this
spec
should
have
a
little
state
machine
of
how
these
these
work,
which
bits
off
already
implies.
It's
just
not
documented.
This
one
already
has
a
little
bit
of
a
state
machine
around
one
list
and
cancels
it's
just
not
described
anywhere,
and
so
this
spec
should
have
that.
That's
a
became,
so
it's
very
clear.
It's
like
you
know
you
sent
a
request.
The
the
requester
sends
the
request.
The
responder
receives
a
request
and
can
send
a
set
of
responses
until
they
send
a
terminal
response
and
that's
it.
A
We
also
gathered
a
set
of
example,
use
cases
of
grassing
some
of
things
like
syncing
watching
we
would
like
you
know,
give
me
give
me
this
hash
and
the
parent,
or
you
know
as
many
someone
back
or
give
you
the
knows
that
exists
in
some
hash,
but
not
attached
to
so.
This
is
like
the
selector
difference
right,
so
you
say:
I
want
everything
related
by
this
hash,
and
but
not
everything
at
this
other
hash
that
that
really
just
means
from
this
hash.
A
To
this
other
cache
things
like
downloading
factors,
appendices
loading
something
from
deep
within
a
giant
dataset,
so
this
kind
of
like
stuff,
and
so
all
of
these
in
a
sense,
are
expressing
requirements
or
selectors.
This
is
why
selector
for
so
relevant
to
graph
sync,
because
what
what
a
request
for
graphics,
sync
so
far,
kept
anyone
being
able
to
express
these
kinds
of
requests
down
at
the
lowest
layer
of
this
well,
I
supposed
to
doing
is,
like
you
know,
kind
of
back
and
forth.
A
A
E
A
E
Currently,
whenever
we
run
GC,
we
walk
through
every
single
pin
in
general,
which
is
not
going
to
work
with
the
large
dataset
section,
well
forever,
something
the
data
store.
Thinking
about
like
telling
we
have
them
Eliza
lost
information,
so
we
only
have
like
do
this
once
when
we
add
move
them
yeah,
so
serious
traveling.
Ideally
you.
E
When
I
go
to
my
Center
fish
form,
Exeter
fisheye
effect
to
make
means
their
effect.
We
wanna
be
able
to
do
is
run
a
GC
without
looking
at
all
the
data
on
the
look
at
the
daily
amateur
move,
but
the
selectors
like
it
makes
it
difficult
unless
you
can
actually
compare
them
because
I
got
otherwise
it.
This
is
gonna
mope.
This
one.
E
And
unpinning
selector
now,
which
of
I
of
the
softer
supplies,
if
I
can
order
them.
Instead
of
this
bucket
commute
to
the
selector,
it's
really
easy,
then
X
base.
We
find
at
once
doctor
that
it's
like
contains
all
the
others
and
then
I
pick
that
one
otherwise
I
heckle
up
is
to
run
every
single
auger
to
okay
now,
which
is
the
new
one
that
applies,
if
you
guys
and.
A
A
One
thing
Elam
note
is
in
the
requirements:
there
is
a
requirement
selectors
that
they
say
it
can
select
with
seated
sitter
enemas,
and
this
might
seem
completely
weird
and
arbitrary.
The
reason
that's
there
is
that
issue
be
possible
for
one
party
to
send
effectively
the
same
selector
to
several
parties,
but
give
them
some
mod
modules
that
they
have
to
so
easy.
I
want
to
go
to
requests.
Give
me
this
graph
mod
7,
because
I'm
asking
is
having
people
so
just
giving
every
7th
note.
A
A
This
might
not
solve
it
solve
all
the
cases,
but
it's
like
a
thing
that
can
be
used,
so
there
might
be
other
things
like
that
with
selector
degree,
so
maybe
not
random
in
this
example,
but
some
other
kind
of
example
of
like
saying
give
me
this
part
this,
like
left
side
of
the
tree,
and
you
guys
want
a
party
to
give
you
the
right
side
of
the
tree.
This
it
gets
easier
with
the
bit
fields
because
with
midfields
you
could
actually
directly
ask
like
I
want
to.
You
know
this
chunk
of
notes
from
you.
A
This
other
chunk
of
notes
from
the
other
person
there's
a
chunk
of
notes
from
the
other
party,
and
you
can
keep
changing
these.
This
is
what
the
Torah
dust,
but
during
doesn't
very
quickly
and
effectively,
because
everything
is
around
bit
fields
and
you
can
modulate
which
it
feels
you
sent
to
what
party
and
and
you
can
govern.
You
can
tell
others
what
what
to
send
you
to
avoid
sending
a
lot
of
duplicates
up.
H
A
So
it's
basically
saying
like
take
ticket.
Suppose
there's
a
one
part
of
this
record:
let's
place
the
sub
back
now,
I
want
to
you
to
roll
a
die,
and
if
that
guy
is
like
you
know,
a
one
then
give
me
that
back
by
the
dogs
determined
by
you
can
make
it
country
different
by
the
content.
Wait
it's
not!
If
you
make
it
only
determined
on
that
node.
So,
basically,
is
like
the
diet.
The
module
is
check
is
on
the
hash
will
be
particular
node.
E
H
A
So
yeah
this
does
not
cover
the
questions
around.
If
I
have
a
whole
swarm
of
participants
who
don't
ask
for
what
and
how
do
I,
you
know,
decrease
the
duplication
or
optimize.
You
know
the
likelihood
of
me
getting
content
quickly
from
the
parties
that
I
have
and
so
on.
That's
kind
of
left
at
a
different
layer,
so
the
goal
of
grasping
was
to
really
really
separate.
These
concerns
with
one
concern
is
just
the
synchronizing
of
grass
with
collectors.
The
other
part
is
who
do
you
ask
what
is
likely
to
be
successful?.