►
From YouTube: Policies and Telemetry WG 2018-06-20
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
So
we
have
I
mean
all
the
code
that
was
necessary
to
necessary
for
GOP
C
adapters
for
n
for
no
session
based
adapters
has
been
checked
in
I'm.
I
have
a
walkthrough
guide
that
I'm
writing
right
now,
I'm
almost
done
so
I
send
that
send
it
out
and
and
then
hopefully
write
some
difference
talks
tomorrow,
but
I
think
it's
it's
mainly
mainly
going
to
be
referencing.
B
A
lot
of
documentation
is
already
in
the
Prados,
so
it's
just
going
to
be
edition
of
whatever
we
had
and
the
existing
add
up
a
developer
guide,
so
yeah
that's
the
status
and
then
one
dollar.
When
he
comes
back,
he
will
resume
the
things
that
are
left.
Is
we
don't
have
what
support
right
now
and
no
session
that
present?
Not
you
know
it
in
there.
So
those
two
things
when
it
comes
back
is
going
to
be
certainly
starting
to
work
on
it.
B
Yeah,
it
is
checked
in
community
in
the
walkthrough
I'm
going
to
refer
to
the
two
checked
in
adapters.
They
have
the
test
right
next
to
it,
so
on
its
own
and
even
the
proper
files
and
everything
make
files
are
in
there.
So
if
anybody
wants
should
want
to
try
out
full
place,
they
should
be
able
to
do
it.
B
A
Just
wish
everything
over
to
the
now
I
guess:
it's
like
static
and
dynamic,
or
the
distinctions
versus
legacy
and
whatever
was
there
before
so
there's
some
changes.
There
was
some
testing
and
then
I
can
start
when
that
then
I'll
start
going
back
and
converting
all
the
old
guarantees
stripping
out
the
config.
So.
B
B
So
I
was
thinking.
Do
we
can
we
somehow
ditch
that
completely?
And
if
user
just
writes
at
once,
and
if
there
exists
both
a
compiled
in
and
an
on
compile,
then
then
we
can.
We
can
in
memory
duplicate
those
things,
because
everything
else
is
just
the
same.
The
the
params
answer
is
the
same.
So
instead
of
users
I
think
it's
going
to
be
syntactic
I
mean
it's.
Our
current
code
remains
the
same.
B
A
A
B
Compiled
in
templates
at
least
for
now,
we,
if
you're
having
the
combines
in
adapters,
we
actually
don't
need
if
you
are
ready
to
pay
the
cost
of
transforming
and
the
even
the
anything
yeah.
So
we
would
need
to
put
a
shame
for
the.
If
we
continue
to
have
to
compile
in
adapters
which
are
working
off
the
go
interface,
then
you
will
need
a
shim,
so
we
would
need
two
things.
B
C
C
B
C
B
B
C
B
A
C
A
B
C
A
A
A
And
this
will
impact,
poetry
and
or
so
like
as
soon
as
I,
think.
The
big
thing
here
is
as
soon
as
it's
working.
We
need
to
make
sure
we
update
all
of
our
metrics
and
dashboards
and
logs
and
everything
to
reflect
the
context
reporter
attribute,
so
that
we're
not
double
counting
and
we're
not
lying
about
where
the
data
is
coming
from,
and
so
we
can
tease
out
client
server
side
latency,
that
kind
of
stuff.
A
So
this
came
up
in
tracing
because
often
now
we're
getting
client
side
spans
as
well,
and
we
weren't
separating
them
that's
the
current
status.
This
is
sort
of
my
high
level
summary
another
way
to
the
statute.
Changes
in
pending
metrics
changes,
I
just
want
to
mention
there
now
that
we've
switched
over
to
source
workload
or
workload
based,
sort
of
orientation
and
the
attributes,
especially
reserves
resolves
the
source
we
want
like.
A
We
need
to
start
thinking
about
a
way
that
we
want
to
reflect
something
that
crosses
deployments
and
we
need
to
update
all
the
metrics
and
logs
and
everything
to
reflect
the
workload.
So
I
think
I
want
to
be
starting
that
work
in
the
next
couple
of
days,
so
that
we
can
hopefully
have
something
that
we
can
run
through
testing
next
week,
because
I
think
the
proposed
target
for
cut
for
whanau
is
the
14th
of
July,
which,
with
the
fourth
of
July
holiday,
really
doesn't
leave
that
much
time.
A
A
As
anyone
they're
looking
at
health
checks
and
how
and
mixers
roll
and
health
checks,
has
anyone
been
following
that
I
know,
there
was
a
proposal
that
endorsed
turnout
and
I
know,
there's
some
PRS
out
right
now
and
the
repo
related
to
that
I.
Just
wondering
I'm
trying
to
come
up
to
speed
on
that
quickly,
but
I
don't
have
the
full
contact
with
anyone.
Anyone
been
looking
at
that
or
I.
C
A
A
Feel
like
this
is
I,
don't
know
this
is
I,
don't
know.
Maybe
you
have
a
insight
here.
I,
don't
know
this
is
something
we
should.
We
should
escalate
and
make
sure
it
gets
resolved
before
100
I
mean
I
feel
like
this
is
something
that's
relatively
important
to
have.
Someone
look
at
I,
just
don't
know
where
the
band
would
come.
A
C
D
A
A
A
C
A
A
C
Can't
do
it
at
the
top,
because,
ultimately,
you
need
to
call
one
log
statement.
One
log
make
one
log
call
for
every
log
entry,
so
I
could
make
at
all
upstairs
and
then
just
call
log
log,
log,
log
log
log
for
everyone,
but
that'll
still
end
up
with
synchronous
I/o
for
every
single
one
of
them.
I
think
you
need
to
be
underneath
the
log
thing
and
just
accumulate
the
data
and
then
write
to
disk
right
to
the
iostream.
Just
once.
C
C
I
emailed
Wayne
yesterday
he
said
he
added
performance
counters,
but
we've
never
gotten
numbers
out
of
them.
There's
a
task
force
or
SWAT
team
that
started
4-perf,
and
some
of
these
issues
are
around
that
around
measuring.
What's
going
on
and
the
mixer
client
nobody's
got
as
far
as
I
saw
the
the
documents
from
the
working
week,
there's
no
concrete
ideas
on
how
to
improve
performance.
C
Well,
that's
true
yeah
in
general,
as
we're
talking
about
the
template
stuff
in
the
beginning,
I
think
it's
fine
for
us
to
burn
a
few
cycles
and
mixer
for
the
sake
of
keeping
things
simple
and
and
more
generic,
because
that's
not
where
our
performance
is
going.
The
performance
is
all
in
the
I/o
and
in
the
client
side
it's
well
it's
better
to
optimize
for
simplicity
and
consistency.