►
From YouTube: Ethereum Core Devs Meeting #21 [7/28/17]
Description
Agenda: https://github.com/ethereum/pm/issues/19
Music: A Winged Victory for the Sullen - Steep hills of vicodin tears
A
A
B
There
is
not
a
lot
to
talk
about
today,
I
think
in
this
case,
but
I
left
some
comments
around
month
ago,
or
maybe
even
more
than
this,
and
so
I
have
I
have
like
different
implementation,
but
it
has
not
been
finished
yet
so
I
cannot
talk
about
this
this
time,
but
in
the
in
the
main,
pull
requests.
There
are
some
comments
and
one
so
it
serpent
rated
issue
about
signed
integers,
but
in
general
I
think
this
is
not
finished
case
in
terms.
D
B
Yeah,
okay,
it's
right
also,
there's
like
a
pre
requested
the
pre
quest
that
also
have
my
test
that
I
was
working
on
some
time
ago,
so
I
think
we'll
take
it
from
now.
It's
like
we
should
try
to
cooperate
after
the
meeting
next
week,
if
possible
on
this,
because
yeah
I
was
busy
with
other
stuff
recently
and
that's
a
bit
not
prepared
for
this
meeting,
but
I
hope
we
can
yeah
almost
finish
it
until
the
next
one.
A
D
A
C
There's
two
approaches:
one
of
them
is
for
the
first
fork
to
be
called
metropolis
in
the
and
to
be
called
something
else,
and
for
the
the
other
s
that
we
call
the
too
hard
for
Excel
to
typically
separate
things
and
amid
the
the
brands
metropolis
can
kind
of
abstractly
refer
to
the
whole
thing.
Okay,.
B
A
I,
what
I'm
gonna
do
probably
is
just
refer
to
it.
Abstractly
is
the
whole
thing,
but
yeah
doesn't
truly
matter
except
to
like
I
guess.
It
would
need
to
be
more
specific
for
things
like
technical
or
spec
papers
like
the
yellow
paper,
so
cool
all
right.
Well,
any
other
ideas
other
otherwise.
This
sounds
good
and
will
probably
float
it
and
a
few
more
channels
before
it's
like
official.
F
G
A
B
G
A
C
A
I
G
A
J
H
J
Measurements
from
a
while
back
I,
don't
really
know
how
to
how
we
go.
We're
gonna
go
about
to
decide
the
gas
costs.
We
should
compare
them
with
the
old
one
or
we
should
if
we
should
try
to
decide
on
what
kind
of
CPU
architecture
it's
a
target
one
and
then
try
to
reach
ten
minute
gasps
per
seconds.
So
basically,
what
I'm
trying
to
do
is
just
provide
some
numbers
and
you
can
take
it
collective
it
on
there.
I
don't
know.
C
Well,
we
realistic
would
just
be
in
a
network
slowdown,
so
if
it
was,
if
it's
we
terribly
miss
calculate,
then
the
network
look
there's
a
risk
of
dos
attacks
like
last
year.
If
we
slightly
miss
calculate
then
like
an
attacker
will
be
able
to
put
push
the
uncle
rate
up
to
like
25
percent
instead
of
15
percent,
for
otherwise
that
Ohama
long,
fine,
okay,.
J
E
Some
of
these
operations
use
code
that
is
heavily
tailored
towards,
or
at
least
on
some
architectures.
It's
it's
assembly
intensive
or
something
like
that
is
our
machines
like
like
arm
and
other
architectures
in
scope
here,
and
should
we
check
the
performance
there
too,
or
should
we
just
assume
that
it's
probably
fine.
C
E
C
I
B
K
B
Uranus
so
in
terms
of
C++,
we
don't
have
any
assembly
enabled
in
the
underlying
big
number
library
that
we
are
using.
But
the
reason
for
it
was
mostly
it
was
easier
to
integrate
this
way.
So
far,
so
the
library
called
MPI
R
has
done
like
different
paths
for
different
architectures,
even
for
different
revisions
of
Intel,
CPUs
and
so
on.
B
A
So
the
other
thing
I
wanted
to
mention
so
Tim
he
got
a
week
and
a
half
ago
was
talking
about
putting
parity
benchmarks
up
and
I
know
he
had
worked
on
them
before.
Did
he
ever
put
those
up
or
that
I
guess
follow-up
questions?
Should
there
be
a
centralized
place
for
all
this,
or
is
this
something
that
can
just
stay
in
the
core
dev
channel
and
we
keep
up
with.
H
A
I
missed
that
link
I
might
try
to
find
it
okay
cool.
So
let's
we
should
get
some
people
on
that,
like
cuz
I
feel
like
this
has
been
going
on
for
a
while.
So
we
probably
just
need
a
group
of
people
and
I
know
Tim
volunteered,
so
I
might
try
to
get
in
contact
with
him
and
then
yeah.
So
you
were
asking
Martin.
Should
this
be
on
one
I,
guess
type
of
just
be
targeting
one
type
of
architecture
you
were
saying,
but.
J
A
J
A
A
Yeah
so
I'm
gonna
think
more
on
this
and
maybe
Martin.
If
we
could
collaborate
on
getting
this
figured
out,
but
I
think
we
can
just
figure
out
who-
and
the
chat
has
talked
about
running
these
and
then
if
we
need
to
get
more
numbers,
that's
fine
and
then,
if
not
start,
throwing
out
recommendations
for
people
to
comment
on.
F
A
J
A
A
A
Okay,
great
next
one
review
time
estimate
for
testing
and
release.
It's
gonna
be
August
in
a
few
days
and
I
believe
we
originally
said
something
like
August
to
September
for
testing
and
release
would
be
ideal.
I
know
that,
last
time
let
me
go
to
the
last
agenda.
There
was
a
couple
of
like
very
important
to
do's
and
I
wanted
to
go
over
each
one
of
those
and
see
where
we
were
at.
A
C
So
for
96
we
are
continuing
on
Nick,
is
working
on
the
tests
and
will
continue
to
go
back
and
forth
on
that
and
will
report
back
to
the
group
when
we
have
something
that
looks
like
it's
passing.
Everything
for
98
98
is
the
intermediate
state
removal
rate,
so
I.
Remember
that
the
thing
that
we
had
wanted
to
do
was
basically
replace
the
30
starting
from
metropolis
block
replaced
of
32
bytes
with
the
intermediate
state
route,
with
a
single
byte
that
he
is
a
1.
C
C
So
that's
modular
exponentiation,
but
the
main
thing
that's
still
needed
to
be
figured
out
was
the
multiple
karatsuba
multiplication
in
complexity,
formula
and
I
had
suggested
one
a
few
days
ago,
and
it
seemed
that
and
it's
well,
it's
basically
a
piecewise
quadratic
quadratic
that
tries
to
approximate
a
n
to
the
power
of
one
point.
Six
and
people
seem
is
relatively
fine
with
it.
A
There
was
a
couple
other
to
do
so
let
me
go
through
the
rest
of
them,
I
think
only
two
more
because
we
already
talked
about
gas
benchmarks
and
opcode
gas
data,
so
the
EIP
for
minor
reward
reduction
was
gonna,
be
combined
potentially
with
669
and
I.
Actually
haven't
checked
on
that
did
anyone
happened
to
put
in
the
minor
reward
reduction.
A
C
G
A
A
Nick
also.
Is
it
on
that
conversation?
This
is
a
long
thread.
Okay,
we'll
look
at
this
a
little
bit
more,
but
this
isn't
something
that
has
to
be
done
like
today.
It
can
be
figured
out
by
the
next
core
dev
meeting,
but
yeah
I
think
that
we
came
to
consensus
last
quartet
meeting
that
we
are
doing
the
issuance
reduction
unless
something
major
comes
up
and
I
do
see
a
few
people
from
the
community
who
were
against
it.
So
we
might
look
at
it
at
least
one
more
time.
C
D
J
J
D
J
G
D
A
C
D
J
H
A
C
They
kind
of
more
broader
in
general
point
of
procedure
I've
been
noticing
that
there's
that,
because
we
have
this
mechanism
where
on
the
EIP,
is
an
issue
first
and
then
becomes
a
pull
request,
it
has
been
causing
some
annoyances
because
basically,
the
community
gets
used
to
debating
in
the
IP,
while,
what's
still
in
the
issue
forum
and
then
when
any
go
heads
to
the
implementation
phase,
because
there's
some
way
to
be
different
numbers
for
the
same
thing.
Mm-Hmm
personally.
D
A
A
L
A
B
B
Yeah
boss
doesn't
matter
but
and
then
like
from
administration,
put
as
you
like
ability
to
to
mute.
The
comment
Accord
is
able
commenting
and
particular
cases
like
to
enforce
actually
moving
from
one
one
issue:
to
do
another
pull
request,
but
I'm
just
asking.
If
this
is
this
technical
option,
I
think
admins
should
have
this.
A
D
A
I'm
on
the
wrong
repository
nevermind,
that's
right,
ya
know
so,
right
now
we
have
Martin
BZ
metallic
Geoff,
Gavin
Fabien
me
Casey,
Nick
and
and
for
the
most
part,
Vitalik
Martin,
Jeff
and
gab
Fabien
to
an
extent
haven't
really
been
as
active
as
far
as
editing.
So
we
probably
should
get
some
new
people
in
there.
A
A
C
A
A
Oh
sorry,
all
right
great
but
ya
know
we'll
talk
about
it
later
to
see
if
like
because
we
might
even
change
editor
responsibilities
to
make
it
a
little
bit
more
relaxed.
So
we
don't
have
to
like
have
a
certain
amount
of
work
done
per
editor
or
something
like
that,
and
it
could
just
be
kind
of
more
chill.
The
next
item
just
comments:
yeah.
A
A
A
D
A
D
A
D
A
D
F
F
A
A
K
So
I'm
not
entirely
familiar
I'm,
looking
at
6:58
request,
I'm
not
entirely
familiar
with
the
with
this
intermediates
receipt,
but
my
question
is:
how
related
is
that
to
the
transaction
receipt
RPC
call
now
stole,
so
it
wouldn't
make
any
sense
to
include
the
status
for
revert,
wouldn't
make
any
difference,
including
DNA
in
this
change
only
have
to
bear
use,
success
or
failure.
You
don't
get
a.
K
Sorry,
what
are
you
suggesting
whether
it
would
make
sense
or
any
difference
to
have
three
codes?
What
is
it
so
if
you
look
at
the
revert
PR
back
in
the
day,
there
was
a
proposal
that
at
some
point,
call
could
be
adjusted
or
a
new
call
could
be
made,
which
returns
three
values,
success,
failure
or
reverse.
L
K
A
B
K
C
J
J
D
Sort
of
assuming
that,
if
it
to
take
to
reverse,
then
that's
because
it's
expecting
what
that
contract
supports
it
and
it's
gonna
go
look.
It's
gonna
go
trace
it
to
determine
what
was
it.
Data
was
like
the
exemption
does
know
where,
as
a
failure
that
isn't
you
to
revert,
there's
no
point
in
doing
that,
because
you
know
there
won't
be
any
yeah,
so
I'm
still
not
sure
I'm
not
convinced
either
way
on
the
right.
J
J
D
K
K
D
D
K
A
J
D
A
G
So
if,
if
you
have
already
implemented
it
so
hard
walks
instead
of
one,
you
have
to
start
these
other
books
at
the
same
block,
because
the
test
case
is
only
one
book
and
there
are
certain
undetermined
gas
costs.
So
you
have
to
set
this
undetermined
the
gas
costs
manually.
Maybe
you
have
to
look
how
the
ship
implementation
to
figure
out
the
gas
of
the
formula
tentatively
use,
but
why
not
now
I
mean
okay,
I,
remember
I,
do
remember.
Some
time
ago,
Vitalik
sent
us
some
bug
reports
on
the
tests
when
he
tried
the
pi.
A
C
A
C
L
A
C
I
think
it
would
be
a
good
idea
to
first
set
like
our
goals
for
two
and
four
weeks
from
now.
So
like
one
example
for
a
goal
would
be
in
two
weeks
from
now.
We
should
have
literally
all
tests
implemented
in
C++,
including
block
tests
for
the
op
codes,
that
only
that
our
block
related
so
things
like
EAP,
9,
96
and
98,
and
for
an
extremely
optimistically
two
weeks
but
realistically
for
weeks.
Somebody
have
all
the
tests
passing.
A
A
G
D
G
And
I
guess:
that's
the
best
way
to
resolve
differences,
resolve
uncertainties,
so
the
I
mean.
Obviously
they
are
the
implementers
themselves
far
more
better
than
I
am
in
fixing
this
or
determining
the
cause
of
these
problems.
So
that's
the
best
way
to
help
I
guess
so
because,
yes,
I'm
solving
existing
Pro
existing
errors
is
bringing
hype.
A
Okay,
so
basically,
if
you're
a
client-
and
you
have
an
error-
please
go
to
the
Gator
channel
that
I'm
pasting
in
here.
It
also
has
a
Skype
bridge,
so
if
you'd
rather
be
in
Skype,
but
I
I
mean
it'd
be
ideal
to
have
one
person
from
each
client
team.
Who
is
basically
the
person
who
can
you
know,
talk
about
and
work
with,
the
testing
team
on
errors
or
just
fix
them
themselves
and
then
report
back.
G
A
C
A
C
A
F
A
F
A
Problem
cool
alright,
so
that
is
where
we
are
there
any
other
comments
on
any
of
this.
Basically,
the
verdict
is
we'll
try
to
get
all
of
the
clients
passing
the
test
and
two
weeks,
preferably
so
that
we
can
then
reassess
and
try
to
set
a
hard
date
block
number
for
the
next
two
months,
because
we'll
need
some
time
for
testing.
A
Cool
all
right
next
item
make
a
decision
on
the
hard
fork,
maeda
VIPs,
that
was
Alex
and
basically
he's
just
talking
about
combining
so
there's
like
a
few
e
IPS
that
are
meta
e
IPS.
That
just
include
information
like
what
block
number
it
started
on
and
other
like
what
a
IPS
were
included
in
different
hard
forks
like
Homestead.
F
K
Yes,
so
there
is
the
two
two
three
three
which
just
explains
the
process,
and
it
would
be
nice
to
actually
merge
it.
It's
just
an
informal
what
type
meta
or
is
it
informal
anyway
that
just
describes
that,
for
every
heart
focus
should
be
such
a
meta
VIP
created,
which
list
code
names
of
the
fork,
the
block
numbers
they
were
activated
and
all
the
apes
which
were
included.
And
now,
on
top
of
that,
we
have
three
hard
Forks
hard
for
me
to
it's
written
up,
706,
708,
709,
m
and
KZ
kept
them
really
updated.
K
I
think
they
should
be
done
at
this
stage
and
it
would
be
nice
to
merge
them
because
they
unlikely
to
change
since
they
are
for
past
hard,
Forks
and
I.
Think
609
is
the
one
for
metropolis,
which
is
not
final
yet
so
it
there's
no
need
to
merchant
and
it
would
be
nice
to
get
just
a
generic
idea,
whether
people
like
this
or
and
if
anyone
is
against
it.
If
not,
then
I
would
hope
to
merge
it.
Yep.
A
I'm
totally
for
it,
the
only
thing
would
be
I've
been
talking
to
some
people
about
the
naming
convention
for
them.
So
do
you
have
a
preference
on
the
naming
convention
like
only
because
if
we
put
these
in
there,
it
might
be
scattered
in
a
chart
if
it's
just
called
like
233
and
240
whatever
or
607
608
like
so
maybe
we
could
just
like
name
it
meta,
homestead
or
meta,
blah,
blah
blah
and
like
a
different
chart.
Oh,
they.
K
A
A
K
And
yeah,
so
the
motivation
is
right
now
swap
in
dope
are
limited
to
the
top
16
items
and
there's
the
only
way
to
reach
any
items
below
is
to
pop
items.
First,
and
this
EEP
introduces
a
two
new
instructions,
one
for
swap
one
for
two
which
take
an
immediate
value
and
can
reach
up
to
the
full
depth
of
1024
items.
K
This
is
a
problem,
a
compiler
issue
which
can
be
workaround
that
in
many
cases,
but
I
wonder
if
there
are
in
any
optimizations
and
death
or
parity
which
take
this
16
stack
limit
into
account
and
put
the
way
the
rest
of
the
stack
in
a
less
accessible
place,
where
there's
any
reason
not
to
be
able
to
access
the
entire
stack.
If
there
is
no
such
reason,
I
think
it
would
be
useful
to
make
this
possible.
D
So
I
have
a
concern
that
this
effectively
encourages
misuse
of
the
stack
like
it
seems
to
me
that
it
would
be
far
more
sensible
to
store
local
variables
in
memory
rather
than
stay
whole
time,
because
that
creates
this
dependency
on
what
occurred.
If
mistake,
I'm
also
concerned
that
adding
a
new
of
code
that
has
an
immediate
argument
complicates
I
simplify
code.
Where
is
currently
only
one
when
opcode
has
an
immediate
argument,
which
is
to
push
on
codes.
E
C
E
K
D
I
very
much
like
the
way
the
avian
currently
only
has
one
immediate
thing
and
I
argued
against
Briggs
introduction
of
newer
medians
for
the
Cindy
instructions
as
well.
I
think
that
if
we
gonna
change,
he
was
a
more
something
that
would
be
a
good
time
to
completely
restructure
things.
But
in
the
meantime
we
should
try
and
keep
the
simplicity
of
existing
opcode
structure.
D
K
D
What
about
the
other
half
of
my
comment
about
using
memory
instead
because
imagine
if
CVM
it
on
the
head,
like
soft
1,
&
2
per
one
from
the
very
start,
writing
language,
which
expected
local
variable
mistake
would
never
have
been
an
option.
How
would
you
have
architecture
till
then,
and
why
not
do
it
like
that?
Now.
K
D
Yes,
but
I,
don't
think
that
the
overhead
of
storing,
locals
and
memory
instead
of
stack
is
overly
high
and
in
something
the
optimizer
can
easily
pull
out
onto
the
stack
if
it
determines
as
possible
and
cheaper,
and
it
seems
like
a
more
sensible
general-purpose
way
to
do
it.
It's
also
the
way
that
traditional
proto
compilers
and
processors
work.
K
Well,
just
a
single
comment
to
that:
the
gas
cost
can
be
higher
to
not
encourage
I
mean
that
was
in
there
as
well.
Well,
actually,
what's
the
IP
the
way
it
is
written
up,
it
had
the
push
required
up
front
and
instead
of
the
immediate,
and
because
of
that
it
was
obviously
more
expensive
than
the
current
doop
instructions.
A
J
A
K
K
F
K
A
lot
of
other
uses
in
functions
for
shifting
very
useful
and
things
like
any
any
quite
building
methods.
What
this
introduces
is
cheaper
ways
to
shift
it's
a
tenfold
decrease
compared
to
you
so
like
currently
implementing
it
with
a
division
of
multiplication
and
exponentiation.
It
usually
cost
around
35
guests,
and
if
we
take
the
shift
instructions
natively
at
Regas,
that's
a
ten
ten
fold
reduction
in
gas
cost
regarding
their.
K
A
Yeah
I
see
Greg's
commented
on
it
and,
let's
see,
and
then
the
add,
Nix
Nix
good
on
it
from
April.
So
this
would
be
okay,
cool
in
the
EVM
yeah
that
this
seems
like
everyone's
pretty
much
agreeing
on
it.
I'd
say
it's
good
to
go.
Is
there
anyone
in
here
who
kind
of
objects
to
and
the
people
who've
been
following
it
and
commenting
on
it
seem
okay,
I
think.
D
A
K
A
K
A
A
Okay,
great
I
think
we're
good.
Then
everybody
have
a
good
weekend
and
I
will
I've
written
some
notes
that
I'll
put
in
the
agenda
comment
and
hopefully
start
doling
out
to
people
every
after
every
Court
have
meeting
I
get
busy
and
then
like
get
Lacs
on
on
sending
out
reminders
to
people,
but
I'm
gonna
try
to
not
do
that.
This
time,
all
right,
cool,
great
meeting
everybody
and
talk
to
you
in
two
weeks,
bye.