►
From YouTube: Jitsi Community Call - May 7, 2018
Description
The Jitsi Community Call is held every other Monday at 10:30AM Central US time at https://meet.jit.si/TheCall
B
C
C
4
toolbars,
because
we've
got
to
pass
I
feel
step
on
the
toolbar,
always
in
top
toolbar
web
toolbar
mobile
toolbar,
so
we've
got
a
piano,
tries
to
unify
those
in
the
buttons
on
them
and
it
does
implement
some
basic
buttons
there
like
common
on,
although
in
all
those
four
cases
have
some
specifics
in
each
of
them,
pedestrian
have
abstract
basis,
dial.
Isn't
the
review
and
she'd
come
to
master
this
week,
then
there's
a
second
PR
building
on
top
of
that
which
works
more
buttons
and
there's
no
PR
currently
for
live
streaming
buttons.
C
So
I,
don't
know
when
that's
gonna
come
so
is
not
working
this
week,
so
I'm
just
hoping
to
get
this
ready
piercing
this
week,
I
know
so
leaves
always
here
started
working
on
those
indicators,
and
one
of
them
is
gonna,
be
the
wife
indicator
like
when
they're
serving
in
a
recording
guar.
So
obviously.
C
B
And
for
the
record
I'd
like
us
to
review
in
the
next
three
to
six
months,
the
way
we
handle
small
videos
on
mobile,
it
may
be
think
of
something
something
different,
because
we've
I
mean
now.
You
and
I
have
been
chatting
about
this
and
others
as
well.
Having
I
think
the
thumbnails
is
not
necessarily
the
most
optimal
optimal
thing
for
mobile.
You
don't
see
the
names
of
people
which
is
quite
annoying
and
you
rarely
actually
use
the
thumbnails
and
so
I've
been
I've,
been
thinking
that
we
could
try.
B
C
Okay,
go
ahead,
so
we
worked
on.
We
now
have
invite
a
P
I,
that's
in
master,
it's
like
two
pieces:
one
is
SDK,
consumers
can
handle
topping
bone
on,
invite
button
and
our
UI
and
it's
manifold
like
if
they
want
to
hook
up
their
own
like
he.
Why?
For
searching
for
people
to
be
added
to
the
confidence
and
stuff
and
then
we'll
invite
them,
because
we've
got
the
service
to
talk
directory
search
whatever
anything
like
certain
numbers
and
whatnot
there's
one
piece
and
the
other
piece
is
like
just
an
API
to
directly
go
and
call.
B
I'd
like
to
just
add
a
little
bit
of
context
to
the
whole
value,
I
think
and
I.
Don't
do
what
you
said
it,
but
because
it's
rather
important
I,
just
like
this
chess
and
the
whole
point
is
that
if
you
want
to
integrate
GT
in,
let's
say
a
chat
application
or
a
presence
based
application,
something
that
is
always
on
for
users.
It
is
very
likely
that
you
might
want
to
be
able
to
ring
people.
B
B
D
C
Those
coolest
things
were
stability.
Related
things
were
fixes.
There
was
a
crash
in
pictionary
Journal
enjoy
the
event
that
we
are
exposing
for
the
SDK
of
a
kind
of
incomplete,
didn't
always
fire
or
fire
like
inappropriate
time.
So
far
in
fixing
dogs,
Thank
You,
param
Zoli.
Do
you
wanna
add
something
about
the
things
that
you
didn't
want?
Those
first
weeks.
E
B
F
Sure
so
I've
been
working
with
Lenny
to
make
changes
to
Tootsie
meat
and
the
backend
such
that,
as
opposed
to
now
where
you
have
to
set
up.
If
you
want
to
do,
live
streaming
or
file,
recording
and
config
it
kind
of
deployment
time
for
making
change.
So
that's
both
of
those
buttons
would
be
available
in
the
UI.
So
you
can
configure
each
of
them
to
be
able
to
be
enabled
independently,
so
people
can
choose
if
they
want
to
do
by
putting
her
or
file
recording
her
live
streaming
in
the
causes
of
both
buttons.
F
So
we've
been
spending
some
time
on,
enabling
that
and
simplifying
that
message
flow
a
bit,
and
we
had
that
looking
pretty
good
at
the
end
of
last
week,
but
because
the
sip
gateways
that
uses
the
same
messages,
we
have
to
tweak
some
of
that
stuff
as
well.
So
we're
gonna
get
that
taken
care
of
and
play
up
some
PRS
up
this
week.
B
B
Well,
we're
actually
working
on
that
and
and
making
sure
that
it
would
be
more
easily
portable
in
a
production
environment.
So
there's
that
also
Boris
is
not
here,
but
he
spent
a
little
bit
of
time
trying
to
do
like
a
couple
of
days
trying
to
do
performance
testing
to
see
where
we
stand
and
and
have
we
had
any
regressions.
So
we're
not
done
with
that.
Yet
it
actually
turned
out
that
we
need
to
fix
a
few
things
before
we
were
able
to
do
that
in
our
environment.
B
I
assume
we're
I'm
hoping
we
would
eventually
have
something
for
next
time,
but
we
did
run
a
few
quick
tests
and
they
were
not
conclusive.
We
weren't
able
to
rule
out
or
confirm
issues
the
more
important
things
that
are
around
the
fact
that
we
started
monitoring
kind
of
general
per
megabit
CPU
usage
on
all
of
the
tens
of
bridges
that
we
have
running
around
the
world
and
there
doesn't
seem
to
be
they
actually.
We
have.
B
We
have
about
I
think
16
of
them
running
on
a
very
old
version
of
the
bridge
from
from
more
than
a
year
and
a
half
ago,
and
they
seem
to
be
giving
us
exactly
the
same
values
as
we
have
with
the
modern
bridges
running
in
today
in
production.
So
to
be
clear,
none
of
those
were
particularly
stressed
bill.
Are
they
all
were
same
kind
of
a
few
thousand
calls
a
day?
B
Okay
George,
would
you
like
to
talk
a
little
bit
about
the
works?
I,
don't
know
how
much
we've
talked
about
this
in
past
goes
because
I'm
short
memory,
apparently
but
George-
is
doing
this
general
initiative
on
trying
to
evaluate
our
bandwidth
estimation
trying
to
compare
the
different
sorts
of
balanced
estimations
in
chrome,
REM,
B
versus
feedback
since
I
transferred
white.
All
of
that
and
this
this
has
been
a
very
interesting
work.
This
is
step
one
step.
Two
would
be
to
see
what
what
are
the
things
that
we
can
actually
basically
step.
B
One
is
to
try
and
make
sure
that
we
are
performing
at
least
on
par
with,
with
with
a
chrome
reference
implementations
of
these
algorithms
and
step
two
would
be
to
see.
Are
there
any
ways
to
fine
tune
them,
so
they
for
the
bridge
so
that
they
perform
better
than
that
George?
Would
you
like
to
talk
a
little
bit
about
the
current
status
ago?
Yeah.
G
So
the
first
thing
that
maybe
it's
interesting
is
that
we
have
built
some
confidence
that
the
sensor
bound
estimations
in
the
jvb
scenario.
Like
you
have
a
jvb,
the
participants
are
connected
through
the
jvb.
This
specific
scenario
seems
to
behave
much
better
with
sense:
I
advanced
estimations,
as
opposed
to
receive,
say,
bounce
estimations.
Now
this
is
a
little
bit
hard
to
explain
because
in
both
cases
we
lifted
jvb
is
running
the
one
of
the
earliest
GCC
iterations
GCC
versions,
with
the
one
that
is
based
on
the
common
filter.
G
So
we
would
expect
a
similar
behavior,
but
it
turns
out
that
for
some
reason
that
we
haven't
really
understood,
it's
not
the
case.
The
sense
that
when
you
have
to
be
it's
inside
bound
estimations,
you
have
much
better
performance,
much
more
accurate
estimations.
The
bridge
is
making
much
more
accurate
estimations,
so
we
are
duplicate
if
I
could
add.
Oh.
G
G
Sometimes
it
achieves
50%,
better
performance,
still
better
performance
than
then
the
common
filter
based
GCC.
However,
even
with
the
linear
regression
when
compared
to
the
actual
capacity
of
the
channel,
linear
regression
doesn't
go
anywhere
near
the
actual
capacity
pass,
the
actual
capacity
of
the
channel,
as
measured
by
iperf
and
as
estimated
by
the
TCP
congestion
controller.
So
the
TCP
congestion
control
achieves
much
bigger
throughput
than
either
linear
regression
or
common
common
filter
based
GCC.
So
these
are
basically
the
conclusions
that
we
have
drawn
so
far,
and
these
are
all
questions
open
questions.
G
We
have
to
understand
why
this
happens,
especially
why
the
Jamie
B
behaves
the
way
it
behaves.
Even
it
runs
the
same
algorithm
when
you
move
it
from
the
sensor
from
the
receive
side
to
the
send
side.
You
have
better
performance,
more
accuracy
to
the
bandwidth
estimation.
So
that's
something
we
need
to
understand
and
we're.
At
that
point
we
are
adding
some
logs
in
Chrome.
This
is
microscopic
analysis
now
we're
going
into
the
details
and
we
want
to
see
how
the
engine
actually
behaves
and
why
it
gives
the
results
of
the
use.
B
Thank
you
very
much.
George
I
think
we
are
mostly
good
with
the
update.
Does
anyone
else
want
to
add
something
yeah.
H
So
from
my
side,
the
two
merrily
visible
thing
that
resulted
in
the
two
weeks
was
two
weeks
other
than
the
ones
that
you
mentioned
was
that
the
filmstrip
is
now
separated
from
the
non-local
thumbnail
is
separated
from
the
filmstrip.
It's
it's
a
beginning
of
a
larger
change
on
mobile,
and
we
now
have
a
domain
list
support
that
also
beginning
of
a
larger
change.
If
you
come
we're.
C
H
A
One
more
thing:
we
given
push
the
announcement
yet,
but
we
can
start
here
so
last
week
we
pushed
new
stable
release,
yeah
yeah,
so
we
push
the
new
stable
release
of
legitimate.
This
is
the
version
that
is
currently
running
immediately
and
with
that
we
added
automatically
when
pushing
new
stable
releases
from
now.
A
There
will
be
a
new
tag
in
the
source
code
of
shitty
media
for
in
the
video
Bridge
attack,
which
is
stable
subversion
and
you
can
easily
spot
it
when
going
to
the
release
type
of
the
project,
and
there
is
also
a
change
lock
attached
with
the
text
file
and
there
is
still
a
manual
process
where
we
sum
up
some
add
some
highlights
of
those
change
locks,
but
it's
still
it's
something.
We
have
a
stable
tag
now
with
a
change,
looks
when
we
release
so
that's
it.
I
won't
I
also
wanted
to
add
some
stuff.
A
So
it's
about
some
small
fixes
that
may
be
interested
for
somebody.
So
there
was
this
change
on
legitimate
that
preserves
the
written
EGS
namespace
when
loaded
from
a
script
tag
I'm,
not
sure.
If
somebody
has
issues
with
that,
but
yeah.
This
should
work.
Now
also
there's
another
fix
from
today
for
iOS
and
the
mobile
web
version.
A
A
B
F
I
Okay,
great
well
I
have
a
huge
appreciation
for
all
the
jitsi
developers.
It's
it's
so
amazing
to
use
jitsi
nowadays
and
I
have
a
question
about
how
many
people
would
you
expect
a
call
to
be
able
to
support.
B
B
And
we
have
our
codes
limited
to
50
people
now
50
people
is
probably
pushing
it
a
little
bit
and
in
order
for
that
to
work
a
little
bit
better
I
mean
pushing
it
not
because
the
bridge
wouldn't
handle
it.
It's
because
you're
you
I,
wouldn't
really
be
adopted
for
that.
For
that
use
case
in
order
for
that
to
work
better
and
sorry
was
it
Liz?
Yes,.
I
I
I'm
calling
from
the
public
laboratory
community-
and
we
are
not
yet
hosting
our
own
instance,
so
I
am
talking
about
using
me,
touch
it
see
and
I'm
happy
to
hear.
The
number
you
just
started
with
is
higher
than
what
I
expect
I
expect
to
have
not
more
than
25
on
our
call
tomorrow
and
I
expect
people
to
be
mostly
connecting
laptops,
but
we
have
quite
a
geographic,
it's
gonna
be
quite
a
global
call.
B
You
so
yeah.
Let
me
let
me
just
finish
up
what
I
was
saying
in
that
case,
the
whole
thing
around
last
death
is
not
relevant
to
your
place,
then,
because
you
cannot
change
it
one
day.
If
you
have
your
own
instance,
you
can.
We
have
this
property
called
last
time,
which
means
how
many
video
streams
will
you
you
will
be
shown
at
any
point,
reducing
that
basically
lets.
You
add
more
people
to
a
call,
because
there's
less
traffic
per
call,
there's
no
last
end
limitation
on
on
me
duty.
B
There
is,
however,
another
setting
which
automatically
video
mutes
and
audio
mutes
everyone
who
joins
after
the
tenth
participant
and
they
have
to
manually
unmute
if
they
want
to
speak.
So
the
caveat
that
you
need
to
keep
in
mind
for
your
car
tomorrow
is
that
if
you
start
having
more
people,
the
me
UI
is
not
currently
really
fine-tuned
for
larger
conferences.
B
So
you
would,
you
should
make
sure
if
you
start
seeing
things
going
out
of
control
a
little
bit
and
I
would
manifest
by
some
people
getting
more
traffic
that
I
can
handle,
for
example,
or
something
like
that
ask
people
to
video
mute
by
default
unless
they're
speaking
to
just
mute
also
unless
they're
speaking
and
people
in
general
people
generally
tend
to
do
that.
As
you
can
see
on
this
call,
for
example,
also,
you
would
be
able
to
mute
people
actually
without
asking
them.
B
However,
you
won't
be
able
to
unmute
them
for
privacy
reasons,
so
that's
another
tool
that
you
have
at
your
disposal
now
we're
currently
working
on
something
that
would
that
would
have
multiple
replications
down
the
line.
That's
something
that
Boris
is
particularly
involved
in.
It's
the
our
octo
project,
which
is
about
getting
that
number
at
least
250
participants
per
call
and
being
able
to
have
the
same
conference
hosted
over
multiple
geo
simultaneously
right
now.
B
The
way
things
would
work
tomorrow
for
you
and
in
your
next
call
is
that
your
call
would
be
hosted
in
a
geography
in
the
geography
of
the
first
participant
to
join.
So
if
you
want
to
prove
that
make
sure
that
you
join
early
from
the
location
where
you
would
like,
where
you
believe
most
of
your
participants
would
be
joining
from.
Basically,
if
you're
joined
from
North
America,
that's
where
the
bridge
will
be.
If
you
join
from
Europe,
your
bridge
is
going
to
be
in
either
Frankfurt
or
ireland's
so
in
in
the
future.
B
That
will
no
longer
be
the
case.
There
would
be
a
bridge
in
every
geo
where
there
are
participants
in
those
bridges
would
be
cascaded.
We
expect
to
ship
that
somewhere
around
the
end
of
the
end
of
summer,
and
by
that
point
we
also
expect
to
have
at
least
some
work
that
makes
the
user
interface
a
little
bit
better
fine-tuned
for
for
larger
conferences,
potentially
Oh
to
cutting
off
some
some
some
automatic
configuration
of
last
end.
Some
additional
Auto
muting
features
some
additional
work
on
reordering
thumbnails
sort
of
that.
B
E
E
E
We're
working
with
a
couple
of
people
who
are
geographically
and
from
a
network
perspective
very
far
away
with
a
very
poor
connection
and
they're
the
the
coal
quality
setting
by
default
is
on
the
highest
level.
So
we
would
like
to
set
up
a
instance
where
that
setting
by
default
is
on
either
for
you
only
or
the
lowest
part
section.
Does
that
make
sense?
Because
so.
D
B
B
But
if
you
go
down
that
direction,
actually,
I
would
probably
suggest
that
you
drop
the
number
at
which
you
start
out
a
video
muna
now
OH
volume
unit,
because
it
would
be
less
confusing
for
people
to
start
doing
video
again,
if
you
just
put
everyone
in
audio
only
we
have
found
that
that
option
is
not
always
entirely
clear.
They
would
be
trying
to
be.
This
is
would
be
trying
to
hit
their
camera
button
and
it
will
be
wondering
why
the
am
I
not
getting
any
video
yeah.
B
B
G
E
Been
asked
to
see
if
this
is
possible,
people
who
are
joining
from
remote
regions
in
Africa
and
talking
to
people
in
the
States
and
Canada
I've
not
experienced
it
myself,
but
I'm
told
that
these
people
experience
it
and
the
first
thing
that
they
do
all
the
time
is
tweak
that
and
video
quality
study
they
they
just
want
to
be
able
to
not
have
to
do
that
manually
every
time
they
join
in
a
conference.
So.