►
From YouTube: 2022-10-26 meeting
Description
Open cncf-opentelemetry-meeting-3@cncf.io's Personal Meeting Room
A
A
C
A
D
This
is
convenient
because
we've
got
all
the
owners
of
the
ottl
on
the
call,
but
there's
a
ongoing
work
right
now
to
improve
how
the
transform
processor
and
any
user
of
the
open,
Telemetry
transformation
language
interacts
with
specific
contexts,
context
being
the
the
process
that
the
language
supports
in
order
to
like
retrieve
data
from
P
data.
D
Right
now,
the
transform
processor
is
pretty
inefficient
when
it
needs
to
like
update
things
like
resources
or
span
metrics.
It
has
to
Loop
through
every
single
span
versus
just
like
looping
through
the
resources,
and
so
the
changes
that
we're
working
on
right
now
will
help
improve
that,
and
also
is
limited
in
some
of
those
things
that
it
can
interact
with
so
right
now,
although
you
can
get
access
to
F4
spans,
for
example,
a
span
event,
you
can't
interact
with
individual
spans
events
and
you
can't
make
conditions
on
those
span.
D
Events
on
how
you
would
change
them,
so
the
proposal
is
to
create
a
Spanish
context,
and
this
could
work
as
well
for
other
low-level
objects
in
otlp,
such
as
spam
links
or
exemplars,
expose
them
as
contexts
so
that
users
who
want
to
interact
with
those
types
of
telemetry
and
make
individual
decisions
based
on
individual
span
events
or
exemplars
and
whatnot
would
be
able
to
use
in
a
context.
D
B
I
I
didn't
say:
I
don't
want
to
do.
This
I
said
that
I
want
to
limit
the
surface
that
we
expose
until
we
have
clear
used
cases
and
and
clear
requirements
to
have
this.
So
if,
if
people
believe
that
the
functionality
that
this
will
provide
like
dropping,
span
events,
it's
super
critical
and
we
have
more
than
one
user.
That
would
use
that
I'm
fine.
B
A
D
I
can
try
to
represent
some
use
cases
we
have.
We
have
some
customers
that
want
to
be
able
to
manipulate
their
span.
Events
dropping
is
the
is
the
key
one
that
they've
been
asking
us
for,
but
I
think
that
reduction
of
attributes
is
another
thing
that
some
of
our
customers
are
interested
in.
A
D
I
mean
I
can
okay
I.
B
C
Okay,
I'm
sorry,
I,
I,
I
I
was
trying
to
find
my
mute
button.
Sorry
about
that.
Obviously
Tyler
and
I
work
for
the
same
company
now
so
as
I
said,
I
mean,
as
we
said,
we've
seen.
People
calling
for
this
and
I
also
think
it's
like
we're
going
to
be
like
people
are
going
to
want
to
do
stuff
and
the
context
is
the
way
we
do
it
and
we're
gonna
need
context,
I
think
to
manipulate
all
of
the
things,
and
so
it
just
sort
of
feels
like
it's.
C
It's
a
necessary
part
of
the
surface
area
of
what
we're
doing
and
part
of
what
we're
trying
to
do
is
to
say,
let's
make
context
as
usable
in
general
as
possible,
so
that
this
language
can
support
whatever
people
want
to
do
with
it.
We're
trying
to
make
the
transform
processor
a
general
purpose
tool,
so
I
I
mean
yeah.
There's
there's
an
argument.
I
think
and
I
mean
it's
valid
to
say:
well,
let's
not
expose
things
before
we
need
them,
but
I
feel
like
there's.
C
D
There's
also
an
issue,
the
issue
that
I
Linked
In,
the
Sig
dock,
which
is
a
link
to
the
main
tracking
issue
around
efficiency
of
context.
One
of
the
related
issues
in
that
link
called
the
new
component
span
event.
Filter
processor,
which
was
open
quite
a
bit
ago,
is
related
to
this
work
and
in
that
issue
it
was
opened
by
another
honeycomb
employee,
but
in
that
issue
there
are
some
end
users
talking
about
there.
Okay,.
D
B
D
So
if
you
look
in
that
main
issue,
Kent
and
I
have
been
going
back
and
forth
on
how
to
change
the
transform
processor
to
actually
take
advantage
of
these
new
contexts
in
a
user-friendly
way.
I
started
working
on
that
as
well.
There's
some
PRS
out
there
for
that
work
and
like
there's
like
a
massive
PR
that
like
should
not
be
merged
and
reviewed,
because
it's
too
big,
but
it's
a
representation
of
like
how
it
could
look
and
then
there's
some
smaller
PRS
that
start
to
build
out
that
API.
D
But
yes,
one
of
the
big
things
we're
going
to
have
to
work
on
is
in
order
to
get
the
efficiencies
of
like
I
just
want
to
change
the
attributes
on
my
resources
like
how
do
we
help
customers
know
that,
like
the
way
to
do
that
efficiently
is
through
the
resource
context
is
by
exposing
that
correctly
in
the
transform
processors
config.
So
we
started
working
on
that
as
well,
and
it's
documented
in
that
issue.
Yeah.
B
And
it
will
become
even
harder,
not
only
for
the
external
users
with
the
config,
but
also
for
the
for
the
developer
users.
That
will
try
to
use
this
package.
So
maybe
maybe
I
was
thinking,
maybe
evaluate
function,
maybe
we'll
get
the
whole
traces
data.
Traces,
for
example,
and
we'll
know,
based
on
the
rules,
for
example,
to
how
to
construct
these
different
contexts
and
yeah.
B
A
B
D
Yeah
we've
moved
towards
defining
a
group
of
statements
and
the
context
associated
with
that
statement
and
letting
it
be
flexible
like
exactly
like
the
from
traces
from
data
point
from
metric.
Like
that.
B
Let's,
let's
discuss
more
a
bit
about
that
that
part
where
we
users
have
to
Define
this
context,
so
so
I
I'm
worried
a
bit
that
you
will
be
complex
in
the
very
hard
for
users
to
get
it
right.
So,
okay,
so
we
will
move
forward
with
the
spanning
event
as
we
discussed,
but
we'll
have
to
revisit
very,
very
actively
how
the
API
surface
for
developer
and
for
the
end
user,
because
I
feel
like
we
are
overly
complicating
right
now.
B
D
B
Please
please
send
me
the
link
or
it's
linked.
D
B
A
E
Yeah,
hey
everyone
yeah,
my
name
is
Alex
Peach
I'm,
an
engineer
at
New
Relic
and
basically
the
context
of
this
particular
issue
is
that
I'm
trying
to
figure
out
a
way
to
actually
monitor
pipelines
in
an
automatic
automatic
way?
Just
the
goal
is
to
kind
of
monitor
the
structure
of
a
pipeline
that
collector's
running
or
multiple
pipelines
and
I
haven't
really
found
a
way
to
do
it
so
I'm
kind
of
proposing
either
kind
of
changing
the
interface
of
the
service
host
or
making
some
changes
to
Z
pages.
B
E
Ideally,
it
would
also
include
instances
but
I
guess
for
to
limit
the
scope
of
kind
of
what
I
was
doing.
I
was
just
trying
to
replicate
kind
of
the
information
that
was
in
D
Pages,
but
it's
I
don't
think
the
instance
information
is
like
available
even
in
the
Z
pages
extension.
B
They
may
not
be,
because
instance
is
a
goal.
Acting
ZIP
pages
is
a
text.
So
so
no
I
I
mean.
Are
you
do?
Are
you
looking
to
have
access
to
the
real
golang
object
instances
or
what
is
instance
for
you.
E
So
I
guess
the
so
like.
For
example,
let's
say
that
you're
running
you're
running
a
a
bunch
of
collectors
and
they
all
have
kind
of
the
same
Pipeline
and
you
want
to
visual.
You
want
to
visualize
which
pipelines
are
running
on
which
specific
instances
and
so
I.
B
Guess
that's
what
I
mean,
so
that's
that's.
The
instance
is
different
than
than
the
instance
that
I
mean
I
mean
object.
Instance
I
mean
an
instance
of
the
struct
batch
processor
like
golang
instances,
but
you
are
not.
You
don't
need
that.
You
need
other
instance,
which
is
the
instance
of
useless
ID
of
these
or
versions
of
this.
Or
what
are
you
talking.
B
Oh,
the
service
instance
idea,
but
okay,
this
is
available
in
the
restore,
but
we
can
put
them
in
the
zip
pages.
Okay.
Now,
let's
so
I
see
so
you
you
need
the
text
part
of
this,
but
in
a
in
a
in
a
computer
process
the
same
way
so
some
kind
of
Json
or
or
something
like
that,
you
need
access
to
this
informations.
So
I
am
more
than
happy
to
have
all
of
these
things
that
we
have
in
the
Z
pages
having
as
a
jsons
apis.
B
So
essentially,
if
you
put-
let's
assume
we
do
something
like
if
you
put
slash
Services
question
mark,
whatever
format
equals
Json
or
something
like
that
to
return
Json,
otherwise
to
return
HTML
of
the
same
page,
I'm
more
than
happy
to
to
start
thinking
this
way.
Is
that
reasonable,
for
you
is,
do
you
think
it's
is
gonna
work
for
you
and
I
think
it
will
allow
programmatically
for
others
to
do
this
now.
The
problem
is
how
much
guarantees
do
we
have
for
these
jsons.
E
Yeah
then
I'm
not
sure
I
mean
there's
some
discussion
on
that
issue.
I
think
another
engineer
had
brought
up
this
solution,
basically
doing
like
content,
negotiation,
I
think
that's
a
good
approach
and
I'm
I'm
kind
of
like
at
the
point,
where
I'm
kind
of
willing
to
put
in
like
a
PR
for
this
particular
feature
but
I
think.
Ideally,
what
I'd
also
like
is
to
be
able
to
monitor
this
pipeline
endpoint
this
Json
endpoint
and
actually
like
ship
it
to
like
a
receiver.
So
then
it
can
eventually
be
exported.
E
So
that
becomes
like
an
issue
where
now
we
have
in
order
to
use
this
sort
of
Z
pages,
you
need
like
either
like
another
extension
or
like
another
receiver.
That's
dependent
on
this
extension,
which
it's
kind
of
similar
to
how
I've
seen
like
the
Prometheus
endpoints
work
is
that
you
know
if
you
want
to
monitor
The
Collector,
but.
B
Yeah,
but
why
why
why
do
you
need
to
ship
it
via
collector
to
with
what
protocol.
E
A
E
A
B
Because
that
if
we
give
you
the
final
config
Gamble,
because
that
that's
another
issue
that
asking
is
asking
us
to
to
have
a
z
page
is
to
print
the
config
gamble
the
resolve,
config
gamble,
then
then
you
will
be
able
to
parse
it.
It's
a
yaml.
You
will
be
able
to
parse
it,
and
you
see
all
the
structures
and
everything
that
we
have
there.
E
Okay,
I
guess
like
how
does
that
work?
In
the
case
of
like,
let's
say
that
this
collector
implements
like
op,
amp
or
something,
and
then
the
pipelines
are
actually
Dynamic,
so
the
actual
config
like?
Would
that
allow
would
that
option
allow
to
introspect
the
config?
Let's
say
that
it
changes
it's
Reloaded.
B
E
Okay,
is
there
a
way
to
get
that
config
in
currently
in,
like
a
receiver
or
because
I
wasn't
sure
how
to
actually
do
that
and.
B
I
wasn't
sure
there
is
no
way,
but
people
are
asking
us
to
provide
that
endpoint
in
the
zip
Pages
as
a
config.
Is
that
okay,
so
I'm
I'm,
trying
to
to
brainstorm
with
you?
Is
that
something
that
will
fix
your
problem
as
well,
because
then
we
I
don't
have
to
implement
two
things:
I'm,
just
gonna
Implement
that
and
fixes
people
need
to
access
the
config,
but
also
your
problem
is
that
reasonable?
Otherwise
we
can
discuss
other
solution,
but
I'm
I'm,
pretty
stormy.
Now.
E
F
So
Alex,
how
would
you
get
the
collector
sorry,
Duane,
I'm,
Erica
I,
also
work
and
you
were
like
how
would
you
get
the
config
in
either
mode,
whether
it's
from
Z
pages
or,
however,
you're
spitting
it
out
to
the
back
end?
Is
it
also
otlp
logs,
like
you
said,.
E
B
Yeah,
oh
so
you
have
a
problem
because
you
you
don't
you
cannot
script
that
endpoint,
so
you
have
to
the
end
point
to
send
it
to
you.
So
that's
that's
what
you
are
trying
to
solve
right.
Do
you
have
your
own,
your
own
distribution
of
The,
Collector
or
you
want
to
work
for
for
any
collector
I.
A
Would
would
the
with
the
traces
that
that
Pages
currently
produce
be
helpful
here
like
if
you
could
export
the
traces?
You
would
essentially
see
what
path
through
a
collector
the
requests
are
taking.
Yes,
sir.
E
Go
ahead!
Sorry,
so,
right
now,
basically
I
have
a
hack
that
basically
it's
an
extension
that
parses
the
HTML
that's
presented
by
Z
pages
pipelines,
but
it's
it's
a
hack
because
it's
non-semantic,
HTML
and
I
don't
want
to
contribute
that
back
up
just
because
it
is
a
hack.
So
that's
what
I'm
currently
doing.
B
Yeah,
but
even
that
extension,
so
so
there
are
two
problems
here:
one
exposing
these
and
second
pushing
this
to
a
pipeline,
so
but
I
think
I.
Think
the
we
have
a
yaml
I,
think
the
the
current
TCP
or
whatever
file
exporter
or
whatever
is
called
it's
capable
of
scraping
an
HTTP
endpoint
and
purchasing
it,
even
if
it's
yaml
so
that
that
may
be
you.
F
B
I
I
would
say
it's
a
log
like
you,
you
put
everything
I
would
either
not
even
transform
the
ammo
I
would
just
put
a
holy
ammo
in
the
body
of
the
log
and
Mark
it
as
with
a
specific
name
that
you
get
and
then
do
the
track.
The
the
parsing
and
everything
on
your
server
side,
so
I
would
just
be
like
give
me
the
whole
endpoint
as
a
yaml
file.
Put
it
in
the
body.
B
B
F
B
But
Opera
not
yet
Define
how
they
want
to
get
access
to
that
config
so,
but
I
think
they
went
to
a
model
where
they
they
have
a
supervisor.
So
their
problem
is
not
this,
so
what
they
will
do
they
have
the
supervisor
scraping
this
endpoint.
F
B
Have
you
talked
to
to
TIG
run
about
this
and
mention
that
hey?
We
would
like
to
have
the
same
way
as
you
have
access
to
the
entire
config.
We
will
do
have
the
same
if
we
have
the
same
mechanism,
this
will
solve
our
problem.
Just
mention
to
him
that
we,
you
are
okay
with
the
same
ways
as
we're
gonna
present
this
to
to
them.
He's
he's
leading
that
effort
with
the
world
Palm
star,
so
talk
to
him
about
this,
and
maybe
maybe
he
he
has
better
ideas
but
I.
B
B
Yeah
Secrets
secret
credentials.
One
thing
we
can
do
is
now
that
we
have
a
marshall
function,
is
to
call
on
marginal
under
config
and
then
call
it
Marshall
after
that,
and
then
make
make
some
of
these
strings
blurry
themselves.
Like
I,
don't
know
how
is
the
term
whatever,
but
just
put
stars
there
mask
masks,
yeah.
F
B
Pablo
had
this
this
issue,
maybe
you
look
into
that,
so
what
I'm
hearing
so
far
Alex
and
Erica
is
we
should
investigate
the
following.
We
should
investigate
if
exposing
the
config
solves
your
problem
and
talk
to
tigran
about
how
he
plans
to
expose
that
to
a
pump,
and
maybe
we
we
rely
on
the
same
mechanism.
If
not
we
back
to
the
same
original
thing
with
the
formats
and
make
the
Z
pages
return
systems
and
stuff
like
that.
B
So
these
are
the
two
options
that
I
think
I
would
prefer
to
do
the
first
one
since
it
was
asked
by
other
users
as
well
to
have
this
config
list,
like
active
configuration
exposed
to
the
to
the
an
endpoint
and
then
from
there
you
you
get
access
to
that
as
well.
So
so
that
would
be
the
the
best
because
we
I
I
hit
multiple
requests
in
the
same
VR.
B
And
I'm
I'm,
fortunately
I'm.
Sorry,
because
tigrant
is
the
more
experienced
in
this
Oppam
stuff.
So
he's
not
here
he
is,
it
could
become,
but
yeah,
maybe
maybe
put
another
item
next-
seek
to
make
sure
we
don't
drop
this
and
because
tigrant
will
be
back
and
I
want
to
make
sure
we
give
yourself
a
proper
solution.
D
Sorry
I
just
had
a
quick
question.
I
was
looking
through
the
Telemetry
that
the
collector.
D
B
Answer
is
yes,
but
nobody
built
a
configuration
for
the
exporter.
So
if
you
are
having,
if
you
are
proposing
this
shoot
me
a
proposal,
don't
don't
put
the
entire
PR.
This
should
be
the
config
proposal,
because
that's
where
we
will
debate
the
most
and
then
we
can
build
the
entire
API.
But
let's
start
with
a
proposal
that
proposal
and
take
it
from
there
make
sense.
D
B
D
Yep
I
may
work
on
it
at
some
point
in
the
future,
because
I'm
interested
in
it
but
I'm
definitely
not
planning
to
work
on
it
anytime.
In
the
next
seven
days,.
D
B
B
A
B
But
I
have
to
do
mine,
I
will
do
first
and
then
you
will
do
another
one.
Okay,
good
see
you
guys.