►
From YouTube: GMT 2018-05-15 API WG
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).
B
C
C
F
A
C
A
Right
so
for
the
matrix
format,
I
think
there
was
some
discussion
in
the
containerization
working
group
as
well
recently
and
in
the
our
frameworks
or
framework
matrix
channel
as
well.
Regarding
some
of
the
metrics
that
we
recently
added
and
we
have
had
for
the
container
successful
and
one
of
the
discussions
was
whether
we
should
figure
out
how
to
unify
some
of
the
metric
format
for
different
endpoints
for
the
containers
endpoint.
A
F
Know,
if
that's
how
I
would
characterize
it,
I
mean
I
think
what
I
understood
was
that
it
was
more
a
challenge
of
like
we
have
to
choose
those
tags
implicitly
and
burn
them
into
the
key,
but
they
may
not
correspond
to
the
tags
that
the
operator
wants
to
use
like
if
he
won
his
framework
name
and
we
decided,
we
were
gonna,
add
the
ID
in
there.
They
don't
know
what
the
ID
is.
Gonna,
be
it's
hard
for
them
to
monitor.
Based
on
that,
I
think.
A
A
C
H
F
C
A
A
I
A
I
C
A
A
F
C
I
think
another
question
to
ask
is
given,
and
maybe
we
need
to
do
some
more
investigation
and
scoping
first,
but
is
this
going
to
be?
Is
this
gonna
give
us
the
biggest
return
on
our
investment
of
time,
because
another
approach
is
to
encode
like
framework
idea
name
and
the
key
as
we
have
and
rely
on
some
kind
of
tool
and
in
between
basis,
metrics
and
whatever
metrics
consumer?
They
have
to
decode
those
so
like
MUX
you
get
on
to
the
consumer,
but
that
requires
very
little
work
on
our
part.
F
F
F
Unless
what
someone
tries
to
do
is
like
another
half-finished
thing,
we're
like
some
metrics
show
up
in
this
new
format,
but
all
the
existing
ones
in
the
old
format,
and
we
stay
in
that
world
for
a
long
time.
I
don't
see
if
they
can
like.
That's
the
only
thing
that
I
see
being
feasible
in
a
small
amount
of
time
and
that
doesn't
sound
too
great.
F
B
C
A
That
our
matrix
library
would
be
extended
to
actually
store
tags,
in
addition
to
names,
adjust
keys
and
values,
and
our
endpoint
will
have
different
except
types
and
they're.
Fine,
except
type,
says
I'm
going
to
understand
the
new
format.
The
civilization
would
be
in
the
new
format.
If
it's
not,
maybe
the
tags
will
not
be
shown
in
the
old
format,
then
they
can
stuff
them
and
it
fit
out.
Well.
F
F
F
B
F
A
A
So
yeah
I
guess
if
this
becomes
a
priority
for
us
at
least
on
the
mesosphere
side,
I
guess
we
will
have
to
put
people
on
it
depending
on
people.
If
here
we
figure
out,
if
that's
most
important
than
other
stuff,
like
improving
state
response
or
improving,
putting
them
in
different
actor
and
stuff
like
that,
I
guess
it
depends
on
which
becomes
a
higher
priority
for
us
in
terms
of
getting
the
biggest
win.
So
this
might
become
one
and
on
for
now,
I
think.
A
F
A
That's
one
question
that
I
had
starting
to
Gaston
the
other
day
was
so
whether
we
should
even
commit
the
perfect
metric,
as
is
or
should
we
wait
to
figure
this
out
because,
as
you
were
saying,
we're
going
to
introduce
a
new
kind
of
format
with
a
very
dynamic
key
I
mean
we
had
some
time
with
kids
before
as
well,
like
principle
based
metrics
and
some
of
the
dynamic
keys.
But
this
is
like
a
lot
more
dynamic
because
there's
an
ID
in
play
here
UID
should
be
like.
A
F
A
F
I,
don't
know
if
it
was
tag
related,
I
think
it
was
more
just
that
that's
what
they
use
I
would
have
to
give
it.
They
have
to
give
it
data
and
it
needed
to
understand
the
type
of
the
metric,
whether
it
was
a
counter
or
a
gauge.
So
James
was
looking
at
basically
exposing
that
for
E
and
X
was
looking
at
exposing
that
through.
F
We
ran
into
some
some
wrinkles
around
how
we
do
histograms
compared
to
how
they
expect
them
to
be
structured,
so
I,
don't
think
those
ever
landed
and
I'm
not
sure
what
they're
doing
okay
I
think
they
were
abandoned.
Oh
I'm,
not
sure
why
I
think
it
just
maybe
their
priorities
shifted
or
something
and.
A
F
F
The
constructs
they
have
from
the
exporting
you
look
at
the
code
'hail,
which
is
now
drop
wizard,
those
metrics
they
have
exporters.
Different
exporters
can
export
to
stats,
D,
JSON
thought,
JSON
or
other
other
systems
that
they're
not
tied
to
one.
We
don't
really
have
that
at
this
point,
and
we
also
don't
have
some
of
the
constructs
that
people
would
expect.
I
also
don't
know
what
what's
gonna
happen.
If
we
add
constructs
like
tags
and
say
something
like
stats,
D
doesn't
support
them.
F
A
H
F
A
Do
we
talk
to
Apple
and
uber
folks
about,
for
example,
seeing
what
their
use
are
and
maybe
send
any
much
the
community
mailing
list?
If
anyone
is
interested
in
this
stuff
or
not,
is
it
a
high
priority?
I
mean
I?
Guess
the
hope
is
if
it
was
I
party,
they
would
have
come
to
this
meaning
and
asked
us
but
start
that
well-attended.
So
we
don't
know,
that's
a
good.
What.
D
A
C
F
C
A
A
What
will
you
do
with
tagging
of
matrix,
but
since
we're
adding
now
partly
about
matrix
that
you've
probably
seen
in
the
doc-
and
you
commented
we're
wondering
if
you
have
any
comment
sessions
on
our
matrix
library-
words,
how
you
guys
are
using
them?
Would
you
like
to
see
some
kind
of
improvements
in
terms
of
tagging?
There
are
any
comments
on
that.
J
I,
don't
really
have
that
much
to
say
we
I
heard
a
rumor
that
we
have
gravano
as
of
like
this
week.
Maybe
so
our
metrics
are
all
aspirational
other
than
something
we're
actually
using
in
in
production,
I.
Think
in
general,
the
original
Prometheus
patches
were
just
a
one-to-one.
Mapping
from
the
JSON
name
is
to
Prometheus
names,
I
think
conventionally
Promethea,
when
you're
modeling
things
in
Prometheus.
You
prefer
to
do
things
with
tags.
J
F
F
We
just
for
some
context.
We
got
some
complaints
recently
about
especially
around
their
approach.
We
took
for
framework
metrics,
because
the
expectation
was
that
the
key
would
be
consistent
across
all
the
frameworks
and
that
the
framework
metadata
would
be
tags
like
the
name,
the
ID
and
so
on,
and
so
we
were
trying
to
figure
out
if
this
is
like
something
we
should
look
into
now
or
like
before.
We
commit
the
purveyor,
metrics
or
but.
J
J
Another
approach
which
I
was
thinking
about
the
other
day
is
is
I
know
there
was
a
discussion
with
G
about
how
to
get
Prometheus,
metrics
and
more
into
committee.
Stop
matrix
more
into
the
core
of
my
sauce,
but
a
slightly
different
approach
might
be
to
fold
the
measles
exporter
into
the
larger
me
sauce
project.
J
F
Yeah,
it
kind
of
makes
sense
to
me
to
long
term
pull
in
different
exporters,
but
I
didn't
want
to
be
able
to
have
hit
an
endpoint
and
get
stat
Steve
format
or
some
other
format.
But
those
things
need
information
about
the
underlying
data.
Right
like
don't.
They
need
to
know
the
framework
metadata
in
order
to
be
able
to
put
in
the
right
tags
for
the
different
systems.
J
D
F
J
A
J
A
Workflow
for
something
like
that,
and
what
does
it
mean
umbrella
project
or
is
just
directly
in
the
repo
we?
We
don't
really
have
any
other
projects,
the
ecosystem,
that
we
maintain,
I,
guess
just
Apache,
mesas
and
I
guess
as
a
meso
site
project,
it's
in
a
different
repo.
It's
all
automated,
no
human
is
involved.
So
that's
less
of
an
issue
alone.
How
people
and
you
know
about
it-
and
you
even
exist-
update
society
but
yeah.
If
you
want
to
pull
in
export
as
a
different
project,
then
we
need
to
figure
out.
J
Seem
plus,
plus
and
I
think,
if
you're
looking,
if
we're
looking
for
someone
to
maintain
it,
maintaining
it
as
a
current
project
in
guy
is
way
easier
and
it's
way
easy
to
find
people
then
to
maintain
it
as
a
abstraction
in
in
the
process.
If
anyone
how
much
make
ongoing
maintenance
is
required,
obviously,.
C
J
J
The
metrics
store
that
we
have
in
production
would
basically
fall
in
a
heap
if
we
all
of
the
new
framing
metrics,
all
the
time,
we're
moving
to
kind
of
cost
us
coped,
Prometheus
stores,
and
that
seems
much
more
promising
for
this
kind
of
stuff.
I
think
that
our
initial
approach
would
be
to
add
support
in
the
me
sauce
exporter
to
this,
in
which
case
we'd
do
the
tax.
J
But
you
know
that's
a
little
bit
begging
the
question,
because
we
would
use
the
tags
because
we're
using
Prometheus
wait
to
stock
my
troops
into
into
Prometheus
if
you're
using
something
that's
not
prometheus,
then
I,
don't
really
know
what
the
meaning
of
tags
is.
So
there's
no
tax
in
some
kind
of
raw
JSON
format.
It's
really
depends
a
lot
on
on
what
your
system
is.
I.
C
J
I
mean
some
metric
systems
have
like
I
said
I
like
that
I'm
familiar
with
performance.
Copilot
has
the
notion
of
an
instance
domain
and
you
you
would
model
you.
You
could
model
these
things.
As
you
know,
an
instant
had
met
which
has
instances
for
a
set
of
frameworks,
but
how
to
take
that,
how
to
make
it
general
Jason
representation.
That
would
seem
there's
kind
of
tricky.
I
C
A
A
F
We
might
follow
the
capability
approach
that
has
been
followed
in
the
past.
Where
there's
a
capability,
it's
not
there.
By
default,
you
can
specify
the
capability
yourself,
which
would
enable
it,
but
at
some
point
that
capability
becomes
a
default
commode
and
there's
no
need
to
explicitly
enable
it,
but
enabling
it
doesn't
do
anything.
Special
are.
A
F
A
F
A
F
I
I
A
C
A
A
A
C
A
C
C
That
G
Chow
brought
up
recently,
it
seems
like
it
makes
more
sense
not
to,
rather
than
requiring
the
operator
to
specify
a
capability
explicitly.
Why
don't,
we
just
add
the
set
of
required
capabilities
to
whatever
the
operator
provides
and
not
require
them
to
specify
it.
We
have
a
union.
Give
me
a
union
of
the.
F
F
I
D
F
F
C
A
B
C
A
That
it's
gonna
happen
anyway,
I
think
because
internally
and
mesosphere
going
to
use
missiles
exporter,
so
that's
going
to
get
updated.
You
understand
performance
metrics.
So
that's
going
to
happen
next,
so
that's
gonna
happen,
but
it's
going
to
continuously
keep
happening.
Dad's
I,
don't
know
that
we
need
to
figure
out
how
to
make
that
part
of
the
process.