►
From YouTube: Real-Time Working Group 2020-07-01
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
All
right,
so
this
is
the
real-time
working
group,
1st
of
July,
can't
believe
it
cool.
So
I
have
the
first
and
only
item
on
the
agenda,
but
this
is
why
I
wanted
to
actually
get
everyone
together.
A
The
way
I
see
this
like
it
would
be
the
same
as
if
we
delivered
any
feature
in
order
to
consider
it
shipped.
We
would
have
to
either
remove
the
feature
flag
and
find
a
different
way
to
to
conditionally
show
this
feature
or
retain
the
feature
flag
defaulted
to
on
either
way.
We'll
still
have
to
conditionally
show
the
feature.
That's
a
lot
easier.
If
we
support
only
the
standalone,
not
standalone,
only
embedded
oxygen
cable,
because
we
can
just
do
a
check
to
see
if
it's
enabled
it's
enabled
to
show
the
future.
A
If
it's
not,
we
don't
and
the
obvious
drawbacks.
I
guess
are
that
we
won't
be
able
to
talk
through
this
yeah
on
comm,
because
we
are
not
a
small
cluster
or
a
single
instance.
So
yeah,
that's
that's
the
fallback
but
I've
put
in
the
agenda
that
there
are
a
couple
of
ways
we
can
mitigate
that
and
plan
for
the
future.
But
the
point
is
we
don't
have
to
like.
A
We
can
defer
that
decision
till
later
and
we
could
even
exit
the
working
group
in
a
phase
one
and
then
in
phase
two
come
back
and
look
at
supporting
yeah
larger
deployments.
We
could
even
focus
on
creating
more
features
on
this
basis
and
we
could
revisit
the
idea
of
having
a
non
production
environment
deployment
of
of
this,
like
so
long
as
that
environment
isn't
managed
with
kubernetes,
I
guess
or
has
a
little
spare
resources,
and
the
only
downside
I
can
see
to
this
is
right.
A
B
Hey
so
I
wish
Henry
was
here
to
confirm
or
refute
that,
but
my
impression
was
that
in
the
next
implementation
he's
actually
using
that
approach,
so
that
so
I
want
to
say,
I
looked
at
an
imam
where
there
is
a
setting
now,
an
omnibus
which
is
disabled
by
default.
But
if
you
were
to
enable
it
it
would
run
so
it
would
allow
me
to
choose
you
between
embedded
or
a
dedicated
action
capable
deployment
and
the
default
any
was
is
to
run
embed
it.
So
setting
is
that
it
is
Kern
approach
already.
B
C
C
B
It's
just
a
question
then
comes
up,
you
know
what
does
this
need
a
full
like
redesign
of
the
feature,
because
because
it
was
just
not
build
without
common
mind,
so
that
would
be
my
only
concern
about
not
ruining
it.
That
said,
though,
I
mean
we
have
thought
about
it
quite
a
bit
and
we
kind
of
know
what
the
path
forward
would
be
for
calm.
So
so
I
guess,
like
I,
wouldn't
be
like
too
fussed
about
about
this
one.
A
A
Planning
to
to
ship
support
for
both
and
standalone
and
embed
it,
but
I
think
we
just
focus
on
embedded
for
now,
and
we
mentioned
that
in
the
in
the
documentation.
When
we
launched
this,
that
we
simply
say:
hey
look
it's
it's
only
for
singles
and
small
clusters
with
a
little
extra
memory,
overheads
and.
A
B
C
B
A
C
Not
really,
my
only
thing
has
been
mostly:
we
want
to
be
able
to
do
this
either
seamlessly
with
the
existing
fleet,
or
do
it
directly
in
kubernetes,
avoiding
creating
a
new
fleet
of
servers.
That's
basically,
the
only
goal
for
from
an
infrastructure
side
is
avoiding
avoiding
having
to
spin
up
new
chef
infrastructure
and
either
reuse.
The
existing
web
fleet
API
fleet,
whatever
or
spinning
up
new
instances
in
kubernetes.
That
makes
sense.
A
B
Yes,
exactly
it
was,
it
was
actually
my
favorite
option
to
begin
with,
because
would
have
because
it
allowed
us
to
iterate
very
quickly
because
it
does
not
affect
the
actual
application
logic
or
how
you
would
implement
the
actual
feature.
And
it's
exactly
the
overhead
is,
was
very
small.
It's
easy
to
test
that
way.
It's
just
like
early
on.
B
We
were
thinking
with
gum
in
mind
and
we
wouldn't
want
to
deploy
it
hasn't
better
because
they
were
just
like
you
would
serve
two
entirely
different
workloads
from
the
same
notes
of
production,
and
it's
just
not
a
good
idea
for
a
lot
of
reasons.
So
that's
why
we
kind
of
maybe
even
prematurely
like
said.
We
don't
want
to
do
this
at
all,
and
then
we
kind
of
like
yeah
dialed
that
back
a
little
bit
right,
because.
B
The
other
approaches
we
looked
at
were
quite
yea,
difficult
or
involved.
If
you
wanted
to
get
memory
requirements
down,
while
still
being
able
to
run
dedicated
notes,
so
yeah
so
I
think.
Actually,
if
you
want
to
ship
something
quickly,
even
if
it's
just
a
self-managed,
this
sounds
like
a
sensible
approach.
A
A
D
D
A
C
D
Values
and
rereading
them
and
basically
said
saying
in
this
iteration
video,
if
you
had
dependencies
on
another
group,
you're
pretty
much
screwed.
You
know
because
of
this
reason,
and
so
like
in
my
what
we're
talking
about
workarounds
to
getting
because
we're
blocked,
and
we
want
to
move
forward
to
do
something
and
I
almost
would
argue
that
instead
of
shipping,
something
like
have
pie
and
worker
who
ever
just
working
on
this
become
less
dependent
on
delivery.
You
know
that's
smiling.
A
A
D
B
B
Was
never
never
done
and
then
we
realized
it
isn't
like
if
you
would
come
here,
I'm
less
concerned
about
this,
because
the
the
change
is
actually
quite
small,
actually
cable
already,
so
we
use
actually
cable
rightful
for
this,
and
it
already
has
support
for
running
either
a
dedicated
node
or
a
run
in
embedded
mode,
and
it
does
not
change
the
way
you
write.
The
application
I
mean
damn
I
okay,
so
there
might
be
some
subtleties.
That
and
I
have
to
be
honest,
because
I
haven't
worked
on
the
implementation.
There
was
mostly.
B
I,
don't
think
we
would
look
at
like
you
know
having
to
rewrite
half
of
it,
just
because
we
want
to
run
action,
cable
in
a
dedicated
server
mode,
so
I'm
not
too
concerned.
But
if
that's
what
you
meant
like
that,
we
in
terms
of
backtracking
like
having
to
redo
how
much
of
work.
A
B
A
Actually
agree
with
that,
like
I
saw,
this
is
kind
of
a
way
to
to
kind
of
fork
the
decision-making
off
at
a
certain
point
and
with
a
very
minimal
change,
just
ship,
something
the
real.
The
only
thing
is
really
just
feature
flag
support.
How
do
we
do
an
additional
check
for
supporting
action?
Cable
that
isn't
a
feature
flag,
so
in
other
words,
we
ship
it
to
a
customer
and
as
soon
as
they
enable
as
soon
as
they
enable
action
cable
on
their
servers.
A
They
get
the
feature
you
know
and
that's
that's
the
only
difference,
and
then
in
future
we
would
have
to
there's
no
way
I,
don't
think
the
detect
that
they're
running
a
standalone
action,
cable
server,
so
you
would
have
to
put
an
additional
configuration
value
in,
but
that's
not
changing
work.
It's
just
a
new
piece
of
work
and
it's
very
small.
We're.
D
Gonna
have
to
do
that
anyways
to
be
able
to
turn
action,
cable
on
them
off
or
throttle
it
or
whatever
for
the
smaller.
It's
just
like
raspberry,
pies
and
stuff
that
don't
have
enough
memory
to
run
it
so,
and
it
makes
sense
to
me
so
yeah,
that's
what
I
say
like
O's.
This
I
just
would
also
say:
like
a
can,
we
become
less
dependent
on
delivery,
because
I
have
a
feeling
migrating
where
we
migrated
were
migrated
from
sidekick
to
what.
A
B
C
A
And
I
believe
it
is
WebSockets
is
next
on
their
on
their
roadmap.
But
you
know:
I
asked
my
own
and
you
know
he
mentioned
13-3.
They
might
get
psychic
finished.
That
brings
us
nearly
into
September.
So
if
we
I
mean,
if
we
take
some
of
these
jobs
like
the
helm
charts-
and
we
can
make
some
progress,
we
can
maybe
speed
that
up
when
they
do
start
us.
A
But
realistically,
are
we
gonna
get
all
that
done
anyway
by
September
and
do
we
really
want
to
wait
to
give
customer
feedback
on
features
when
we
could
unblock
other
people
who
might
want
to?
You
know
even
just
expand
simple
features
like
parts
of
other
parts
of
the
sidebar
you
know
and
to
be
able
to
release
those
to
customers
and
get
feedback.
B
Another
thought
I
had
on
this
was
that
we
should
involve
definitely
someone
from
QA
and
in
making
sure
that
this
stuff
is
tested
before
like
feature
tested.
You
know
before
before
we
even
consider
releasing
that,
because
I
mean
yeah
the
problem
with
talk
footing.
Obviously
it's
also
when
develop
I
mean
there's
like
a
bunch
of
examples
where
things
work
a
little
bit
differently
in
omnibus
and
developers
usually
do
not
test
using
that
new
bus
but
sort
of
a
couple.
Handful
manual
examples,
maybe-
and
that's.
D
B
Running
your
seed
in
this
container
is
fairly
easy.
I
found
it
like
the
way
to
obtain
and
yeah
okay,
so
it
kind
of
depends
like
if
you,
if
you
work
on
a
new
feature
of
good
lab,
and
you
want
to
test
this
using
an
omnibus
build
yeah.
You
would
kind
of
like
have
to
go
to
the
CI
infrastructure
and
like
download,
like
whatever.
B
The
that's
probably
much
request
you
would
have
to
have
to
have
a
job,
build
your
custom,
omnibus
image
and
then
yeah
pull
that
container
a
little
bit
image
and
run
a
container
to
that's
my
workflow
anyway.
It's
like
you,
know,
super
straightforward,
but
it's
much
more
straightforward
for
anything,
that's
already
on
the
master,
so
it's
like
already
kind
of
done.
B
D
A
A
That's
right!
You
can
simply
build
the
image
with
your
branch
right
and
it
will
use
that
instead
of
the
upstream
image.
So
I
don't
know
if
it's
something
wrong
yet
involved
in.
But
if
you
did
this
work
anywhere,
if
you're
curious
about
native
good
lab
anyway,
it
might
be,
and
it's
a
cool
way
to
do.
I'm.
B
C
D
A
Think
grant
did
some
work,
creating
reference
architectures
to
just
test
the
resource
usage,
but
again
yeah.
He
I
think
he
mentioned
at
the
time
that
it
was
gonna,
be
extremely
difficult
to
do
like
full
stack
testing
on
this,
like
maybe
it's
easier
actually
now
with
the
embedded
server,
but
you
would
still
need
it
yeah
if
you
wanted
to
a
full-stack
test,
you'd
still
need
to
spin
everything
up
today,
yeah.
B
I
think
it's
due
to
the
nature.
Sorry
like
maybe
maybe
that's
the
rail
this
but
like
just
maybe
in
the
sentence.
I
think
the
tricky
bit
is
is
is
here
that,
just
by
virtue
of
it
being
pushed
based
you
you
need
to
trigger
some
kind
of
domain
events
to
observe
the
behavior
right.
So
something
like
someone
else
changed
the
assignee
of
this
issue
so
that
web
page
that
I'm
not
even
interacting
with
updates
itself,
so
that
yeah
I
don't
know
if
we
do
any
kind
of
test
like
this
already
elsewhere,
but
I.
B
A
Why
most
people
are
testing
it
manually,
but
there
is
an
additional
way
and
that's
to
assign
somebody
using
a
quick
action,
so
it
could
be
just
the
case
of
browser
loads.
The
page
sends
a
request,
a
post
request
to
create
a
comment
with
a
quick
action
assigned
to
the
user
and
it
could
be
tested
that
way,
instead
of
minting
in
two
separate
things.
B
D
A
Cool
we're
out
of
items
I'll
go
back
and
annotate
some
of
the
agenda,
because
I
know
I
missed
some
stuff,
but
yeah
I
can't
do
both
I
can't
converse
and
type
just
I've,
given
up
trying
thanks
everyone
for
your
time.
I'm
gonna
create
that.
Mr,
if
you
wouldn't
mind,
give
any
feedback
and
I'll
leave
it
for
a
couple
of
days
before
Marg
in
it
like
so
in
case.
Anything
comes
to
mind
by
why
we
shouldn't
do
this.
Yeah
and
that'll
be
the
next
action
from
here
cool
thanks.
Thank.