►
From YouTube: 2021-09-29 meeting
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
Hello,
everyone,
so
it's
a
wonderful
wednesday
again
and
we
have
this
meeting
and
we
have
couple
of
items
in
the
agenda.
C
First
yeah:
I
need
to
find
the
links
to
this
description,
but
effectively
proto3
added
optional
field
presence,
and
it
was
an
experimental
feature
for
a
while
and
now
it
is
an
approved
feature
and
we're
looking
at
using
this
to
make
the
otlp
protocol
be
easier
to
evolve.
C
The
tldr
is
inside
of
the
proto
definition.
You
can
mark
a
particular
field
as
being
optional
and
then
the
language
library
is
required
to
remember
whether
or
not
that
field
was
present
in
the
binary
that
it
read
or
in
json.
You
know
whether
or
not
the
field
was
present
today.
Protos
like
if
you
don't
mark
something
as
optional,
you
get
a
default
value
effectively
and
if
you
actually
set
a
field
to
the
default
value,
the
client
might
not
send
the
field.
C
That's
that's!
That's
just
an
implementation
detail
of
the
client
library.
However,
there's
there's
a
lot
of
proto2
to
this
a
lot.
There's
a
lot
of
use
cases
where
relying
on
optional
make
sense.
C
We
actually
designed
otlp
metrics
around
the
notion
that
optional
was
allowed
and
it's
not
and
so
we'd
like
to
start
using
optional
field
presence
in
otlp.
So
the
discussion
here
is,
we
talked
about
this
in
the
metric
sig
or
the
specification
sig
bogtin,
and
I
there's
an
issue
where
gogo
proto
doesn't
support
optional
field
presence
at
all,
and
the
collector
proto
libraries
rely
on
optional
field
presence.
C
So
what
I
wanted
to
talk
about
was
kind
of
like
understanding,
whether
or
not
we're
comfortable
moving
with
optional
field
presence
in
the
collector
timelines
like
understanding
what
we
can
do
and
and
whether
or
not
external
help
would
be
useful
here
to
try
to
around
our
usage
of
proto
and
optional.
C
A
Yeah,
so
our
biggest
problem
is,
we
have
defined
our
p
data
that
the
library
that
contains
the
data
models
for
for
data
that
we
pass
through
the
pipelines
that
we
define
that
part
is
great
because
allows
us
to
to
change
the
internal
implementation
of
what
we
do
internally
and
right
now.
What
we
do
is
we
generate
the
the
protobuf.
A
Files
using
gogo
proto,
which
is
very
old,
hence
because
of
that,
I'm
not
sure
if
adding
optional
would
be
trivial
or
not,
it
may
just
simply
work
or
it
may
not
work.
So
that
being
said,
I
think
we
need
to
do
a
due
diligence
understand
exactly
if
we
can
have
an
optional
field
with
the
current
google
proto
version
that
we
are
using
also
keep
in
mind
that
google
proto
is
kind
of
deprecated
or
unmaintained
right
now,.
A
A
The
official
go
proto
library
or
do
what
we
had
in
mind
initially,
which
is
try
to
to
do
lazy
deserialization
or
to
do
to
write
our
own
serialization
and
this
realization
directly
to
proto,
which
is
much
faster
than
what
the
current
proto
does.
A
A
Then
what
does
that
look
like
then,
then?
Probably
a
couple
of
days
of
work,
we
need
to
add
in
our
p
data
we
need
to
generate
another
method
called
has
full
or
something
like
that
just
to
to
match,
with
the
the
concept
of
has
the
the
value
or
doesn't
have
the
the
value
will
still
be
returned
as
default.
If
you
query,
because
I
think
that's
the
the
default
behavior
for
for
for
proto,
so
yeah,
it
will
be
extremely
easy.
C
Okay,
that's
something
we
can
easily
find
because
we
have
it's
all
in
private
repos,
but
we
have
our
personal
repos,
but
we
have
a
a
set
of
performance
tests
in
otlp.
Right
that
that
is
go.
Go
pro,
go
go
proto
generated
to
check
collector
performance.
Every
time
we
make
a
proto
change,
so
I
can
actually
make
the
optional
proto
change
in
there
and
see
if
I
can
get
it
working.
A
See
what
happens?
Yeah
that's
an
option.
The
other
option
is
just
add
optional.
We
have
that
docker
image
that
we
are
actually
using.
So
it's
set
up
in
the
in
the
proto
repo.
There
is
a
setup
for
for
building
with
gogo
pro
to
all
things.
If
you
just
run
me,
go,
go
proto
and
buy.
The
generated
code
looks
reasonable
for
that.
A
C
C
Okay,
well,
I
mean
there's
like
when
we
were
doing
performance
benchmarking.
We
were.
We
had
like
eight
different
options,
we're
experimenting
with
to
see
how
they
all
worked
with
gogopro.
So,
if
that,
if
that
thing
has
the
default
set,
that
you're
using
beautiful,
am
I
able
to
change
that
if
we
need
to
for
optional.
A
C
Okay
and
is,
is
it
on
the
roadmap
eventually
to
move
away
from
gogo
proto?
I
only
asked
because,
as
you
mentioned,
it's
not,
it
doesn't
have
any
support
that
community
eroded
away.
It's
gone.
It's
we
need
to
decide
if
we're
going
to
own
it
ourselves
going
forward
or
not.
No.
A
We're
not
going
to
own
it
for
sure,
okay,
most
likely,
what
we
will
end
up
doing
is
because
so
that's
very
generic.
We
can
do
a
much
better
job
on
our
side,
so
most
likely
we'll
use
protoc
we'll
have
a
plug-in
for
generate
our
own
p
data
directly
from
there.
A
C
Around
it
just
so,
we
can
keep
abreast
of
that.
Okay,
basically,
if
we
try
gogo,
proto
and
optional
doesn't
work,
we
need
a
workaround
for
this,
like
we
have
so
we
have
ish.
We
need
to
make
a
decision
in
the
protocol
around
whether
or
not
we
can
rely
on
field
presence,
which
is
the
fastest
way
to
solve
an
existing
bug
with
metrics.
Today,
if
we
can't
rely
on
field
presence,
we
actually
have
to
re-implement
field
presence
via
on
the
wire
information
that
already
exists
as
part
of
proto.
No,
we
had.
A
C
Okay,
okay,
but
in
terms
of
timelines,
the
sdks
are
unwilling
to
implement
anything.
The
collector
doesn't
support
going
forward
sure
so
like
basically,
what
I
need
is
is
what
does
the
collector
need
outside
of
gogo
proto,
so
if
it
doesn't
work
in
gogo
proto?
What's
our
plan.
C
Okay,
okay,
and
for
for
reference,
the
java
community
switched
away
from
proto-c
to
a
thing
called
wire
to
do.
Proto-Generation.
A
Sure
that
java
changed
to
wire
versus
having
the
manually
written
serialization
that
I
wrote
six
months
ago
or
something
one
year
ago,
yeah
I
all
the
prs
went
through
this.
C
Month
so
so
effectively,
it's
not
just
that
they're
using
wire
they're
using
wire
as
a
hookable
proto-c
compiler,
and
they
have
their
own
set
of
templates
that
they're
using
to
generate
their
classes
and
a
custom
marshaller
right,
so
they're
directly
serializing
their.
You
know,
if
you
will
their
p
data
classes
into
proto,
but
they're
using
this
instead
of
proto-c,
because
it
was
easier
to
extend
for
them
yeah,
I
don't
and
again
all
I'm
just
all.
I'm
asking
is
like
what
are
the
requirements
here.
C
Okay,
cool
all
right,
so
I'll
do
the
investigation
on
the
gogo
proto
optional
thing:
I'm
not
hearing
major
concerns
from
people
with
using
optionality.
So
that's
wonderful
and
I'll
come
back
when
I
have
results.
A
A
Okay,
let's
move
to
the
next
one
here
I
think
we
have
travis
with
the
next
item.
E
Yes,
thank
you,
hello,
so
I'm
travis
with
vmware
and
we're
working
on
adding
a
exporter
for
the
tonsi
observability,
which
is
one
of
our
products,
and
we
submitted
a
pull
request
last
thursday
and
I
noticed
a
lot
of
the
workflows
aren't
running
in
it.
That's.
A
That's
by
design
github
change,
something
if
you
are
not
a
member
of
the
organization,
the
workflows
will
not
run
automatically
for
you,
and
somebody
has
to
press
a
button
to
to
run
the
workflow.
Okay.
Okay,
it's
for
security,
because
some
of
the
workflow
may
have
access
to
credentials
and
stuff,
and
we
don't
want
to
leak
the
credentials
by
you,
changing
the
workflows
and
and
other
things.
So
anyway,
it's
I
know
it's
annoying,
but
it
is
what
it
is.
A
Also
one
thing
in
the
contrib
repo:
we
have
this
model
where
every
component
has
to
have
one
or
two
owners
co-donors
who
who
is
the
co-donor
for
for
exporter,
tons
of
observability.
E
I
am,
are
you?
Are
you
the
only
one?
No,
no,
the
they're
not
open
guard
is,
is
the
technique
so
he's
also
known.
A
Yeah
so
then
you
should
ask
the
other
one
to
first
review
the
pr
before
asking
an
approval
to
to
look
at
the
pr.
E
Okay,
that
makes
sense.
Okay,
so
we'll
we'll
do
that
and
then
how,
when
once
he's
looked
at
my
code
and
has
checked
off
on
it,
is
there
anything
that
we
need
to
do
on
our
side
to
let
you
guys
know
that
it's
time
to
look
at
it.
A
Usually
not,
but
you
can
just
probably
mention
it
in
the
slack
if
nobody
looks
at
it
for
24
48
hours
or
something
like
that,
just
ping
us
on
the
slack
channel
to
make
sure
somebody
is
looking
at
it.
E
F
Yeah,
actually
I
was,
I
was
just
hey
everyone.
I
was
just
adding
in
particular
the
status
information,
I'm
particularly
new
to
the
community
as
well.
I
just
had
a
question
related
to,
since
we
did
basically
release,
I
think
close
to
the
20th
or
22nd.
We
had
a
release
on
for
having
that,
basically
relay
to
the
rest
of
the
community,
about
the
collector
itself
of
going
ga
stably.
Do
we
need
to
really
update
the
status
on
the
particular
link
that
I've
updated
on
the.
A
So
on
the
22nd,
when
we
release,
we
did
not
say
it's
ga.
We
said
that
we
guarantee
stability
for
the
tracing
pipelines
and
tracing
data.
I
think
the
status
that
is
there,
I
think.
A
Give
me
a
second
experimental,
stable.
A
G
Yes,
hi.
I
just
wanted
to
bring
up
a
question
about
this
pr
here,
I'll
link
it
in
the
google
docs
as
well.
G
Yes,
so
it's
regarding
removing
the
default
components
from
the
core
and
there
was
a
discussion
to
move
it
to
move
the
default
components
function
to
a
different
package
than
the
example
testing
components.
G
Yes,
yes,
and
just
wait
to
bring
up
to
see
if
there
may
be
a
better
sort
of
place
to
move
this
function
too.
D
I
think
this
this
ties
into
the
other
pr.
That's
that's
also
outstanding
about
removing
hotel
call
maine
entirely,
which
probably
should
land
before
this,
and
then
that
won't
be
an
issue
for
this
pr,
but
I
think
that's
held
up
because
we're
we're
needing
to
know
do
we
actually
still
need
that
or
not.
It
looks
like
the
that
pr
is
held
up
because
you
want
something
to
make
that
will
still
build.
A
So
I
was
thinking
to
use
the
builder
to
still
build
a
binary
so
that
people
can
test
it
locally.
So,
by
removing
the
hotel
call
right
now
anthony
we
give
zero
chances
for
users
to
test
locally
anything
in
core.
D
Right
yeah,
I
mean
I've
been
making
use
of
that
ability
with
the
the
feature
gate
capability
that
I've
been
working
on,
it's
nice
to
be
able
to
build
a
simple,
collector
and
and
use
it.
So
I
just
want
to
make
sure
that
we're
understanding
what
the
the
sequencing
of
these
should
be
then
right.
So
if,
if
you
want
to
remove
default
components,
it
seems
like
that
has
to
wait
until
after
hotel
call
main
has
been
removed.