►
From YouTube: 2023-01-17 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
B
C
B
For
sure
happy
New
Year.
How
are
you
I'm
all
right.
B
Good
yeah
hang
it
in
there
same
same
old
as
they
say:
In
The,
Biz,
foreign.
C
Ruby
for
good
mascot
excellent,
there
was
a
year
that
it
was
hosted
at
a
endangered
species.
Research
campus
and
one
of
the
projects
for
the
not
hackathon
was
I,
don't
know
it's
like
it.
C
Helping
researchers
track
things
about
the
endangered
species.
One
was
a
bunch
of
antelopes
like
it
was
like
a
dating
app
for
antelopes
because
they
were
so
endangered.
There
were
so
few
of
them.
You
had
to
mate
to
them
very
carefully
or
you'd,
get
weird.
C
D
The
joke
should
become
part
of
the
permanent
record
in
the
the
meeting
notes
Perfect
by
the
way.
B
E
E
Well,
I,
don't
know
how
long
it
would
have
taken
if
it
hadn't
been
over
Christmas
right
because
there
was
you
know,
holiday,
traveling
and
seeing
people
takes
a
lot
out
of
me,
but
I
had
like
a
week
and
a
half
on
either
end
of
that
so
and
I
didn't
do
anything.
I
didn't
go
I.
Just
like
slept
and
existed.
E
C
That's
all
I
have
that's
well.
That
was
that's
what
I
was
asking
for
and
I.
Thank
you
for
it.
Thank
you.
E
D
D
D
Yeah,
so
there
is
a
spec
1.17
release.
Today
there
are
a
couple
of
PR's
to
I,
guess:
expand,
expand
or
add
some
semantic
conventions.
One
is
definitely
an
expansion
around
the
database
pool
attribute.
D
Basically
there
there
already
exists
this
pull.
That
name
attribute,
but
in
the
case
where
there
isn't
a
pool
there,
suggesting
that
the
pool
name
should
be
host
Port
database
name
user
or
something
if
I'm
parsing,
that
properly.
C
D
There's
also
a
PR
for
capturing
the
bind
values
for
a
parameterized
SQL
statement
and
yeah
I.
Think
in
general,
people
were
not
opposed
to
this
surprise
that
doesn't
exist
already.
I
think
the
biggest
thing
is
just
making
sure
that
it
is
optional
and
opt-in
so
that
you
don't
accidentally
collect
pii.
D
I
think
a
lot
of
yeah
most
people
are
in
agreement
that
it
would
be
awesome
if
there
was
a
way
to
be
able
to
enforce
cardinality
cardinality
limits
on
your
metrics
by
the
meter,
meter,
provider
and
I.
Think
this
PR
suggested
the
limit
should
be
2000,
so
2000,
distinct
time
series
from
a
a
single
meter
provider
or
anything
more
than
that
and
it's
starting
to
look
like
you-
may
have
a
cardinality
explosion
and
something's
wrong
and
either
you
need
to
kind
of
fix
things.
D
Yeah
like
at
that
point,
there's
there's
something
wrong
in
your
code
and
you
need
to
go
fix
it.
You
either
need
to
like
Add,
A
View,
to
kind
of
like
filter
out
an
attribute
that
is
not
not
grouping
up
or
or
something
along
those
lines.
But
I
think
this
is
where
things
started
to
get
controversial.
It's
like
what
that
what
what's
to
do
when
you
hit
that
2000
limit
in
kind
of
like
a
running
process
and
I,
think
the
there
are
quite
a
few
suggestions.
D
I
think
Jim
MCD
was
at
least
was
initially
looking
to
just
just
drop
the
data.
It
seems
like
it's
an
invalid
stream
and
then
other
people
were
saying
you
should
just
report
up
to
2000,
but
that
seemed
like
it
was
a
really
bad
idea
actually
because
it
kind
of
makes
it
look
like
your
stream
is
okay,
because
you're
getting
data
but
you're
actually
dropping
a
lot
of
data,
and
it
might
not
even
be
rolling
up
at
all,
depending
on
what's
happening
with
with
your
attributes.
D
So
that
seemed
like
actually,
maybe
the
worst
idea
and
I
think.
Ultimately,
since
there's
even
a
summary
on
this,
the
the
things
that
were
a
little
bit
more
popular
was
trying
to
I
think
yeah.
The
biggest
thing
is
just
trying
to
like
get
this
information
to
the
user,
that
something
is
wrong
so
that
you
can
go
ahead
and
fix
it
and
I
think
they
were
going
to
see
if
this
could
degrade
to
some
street
or
yeah
degrade
to
a
single
stream
of
metrics.
D
That
indicated
that
there
isn't
there's
an
error,
so
I
think
it
makes
us
a
little
bit
more
concrete
if
you
like,
having
something
like
an
HTTP
request
account
or
like
a
throughput
metric.
This
is
one
metric.
That's
probably
going
to
be
vulnerable
to
to
this
issue
if
you
have,
depending
on
how
the
stream
is
being
named,
if
it's
being
named
on
some
parameter,
that
can
have
many
values
you
know
in
in
the
URL.
D
You
can
get
one
of
these
explosions
and
if
you
have
a
bunch
of
dashboards
set
up
where
you're
kind
of,
like
you
know,
have
a
filter
in
there
by
HTTP
URL
like
you're,
not
going
to
see
problem
right
away.
D
C
D
Don't
really
have
a
way
to
con.
You
don't
really
have
a
way,
I
think
to
continually
like
report
that
first
2000
or
something
and
whether
or
not
that
would
even
be
useful.
If
you
could
do
it
is.
E
This
is
relevant
because
it's
just
something
I
work
with
thinking
about
this
today,
but
think
of
something
like
Network
performance
monitoring,
where
you
may
be
tracking
a
metric
that
has
extraordinarily
High
cardinality,
because
it's
tracking
Source
IP
and
destination
IP,
and
maybe
that's
a
thing
you
want
it's
really
not
a
good
fit
for
tracing.
But
things
like
this
would
mean
that
you
can't
really
reliably
use
Hotel
meters
to
capture
that,
if
there's
going
to
be
a
limit
on
the
cardinality.
If.
E
Right
as
long
as
the
woman's
configurable
I
think,
I
I'll
take
some
time
and
try
to
read
through
this
today
to
make
sure
I
put
my
two
cents
in,
but
this
smells
to
me
like,
where
we're
gonna
run
into
some
issues.
There's
you
know
without
really
good
error.
Propagation
strategies
like
there's
no
way
that
people
would
necessarily
know
that.
There's
a
problem
they're
not
going
to
see
it
it's
going
to
affect
the
reliability
of
their
metrics
and
thus
they'll
think
that
otometrics
are
bunk.
E
You
know
if
we're
not
very
careful
about
how
it's
reported,
but
also
like
it's
a
limitation
of
the
Prometheus
back
end.
Where
we're
writing
the
specification
to
Prometheus
again
and
I
I,
really
don't
care
for
that.
I
know:
there's
a
strong
bias,
but,
as
Rob
is
all
too
well
aware
and
as
Eric
is,
there
are
other
back-ends
out
there
that
perhaps
are
a
little
more
flexible
with
how
much
cardinality
you
can
send.
C
Skimming
the
instead
of
the
comments,
the
text
of
The
Proposal
I
see
the
word
configurable
in
several
places.
So
that's
maybe
there's
maybe
we're
going
to
bike
shed
what
the
default
value
is,
but
the
we
need
to
read
it.
But
it
looks
like
it's
a
configurable
thing
which
I
think
that.
E
C
D
D
Cardinality
and
they
want
to
have
some
reasonable
ways
to
limit
and
detect
this
because
I
think
I,
don't
know,
I
think,
depending
on
where
you're
sending
your
data.
There
is
like
a
a
huge
variance
in
your
bill
that
you
will
get
depending
on
the
cardinality
of
your
stuff,
and
this
ends
up
being
an
important
thing
is
for
users
as
a
result,.
D
But
but
yeah
definitely
take
a
look,
definitely
make
sure
that
this
is
not
going
to
be
too
restrictive
for
folks
that
do
want
high
cardinality
and
ultimately
yeah.
There
was
a
lot
of
discussion,
so
I
hope
I'm,
remembering
what
the
final
kind
of
detail
was,
but
I
think
ultimately
dropping
things
and
signaling
to
the
user.
D
Hopefully,
via
this
same
time,
series
or
by
a
Time
series
that
you
expect
to
see
was
was
the
approach,
but
there
were
all
there
were
many
other
suggestions
that
were
being
bossed
around
before
we
got
to
that
point.
D
D
Tristan
has
been
trying
to
add
a
suppressed
tracing
flag
to
to
the
early
implementation
and
has
looked
at
Ruby
untrace
as
well
as
JavaScript
has
a
context
based
flag
to
suppress
tracing
and,
ultimately
he
so
the
JavaScript
context
based
suppressed
tracing
flag
kind
of
works
on
its
own,
whereas
untrace
requires
a
parent-based
sampler.
So
he
kind
of
based
his
spec
update
on
the
JavaScript
implementation
and
was
asking
others
to
take
a
look
and
possibly
add
this
to
their
see.
D
If
they
would
add
this
to
their
languages
and
I
think
everybody
recognizes
there
is
a
need.
There
were
some
good
questions.
Good
questions
that
I
think
were
asked
were.
D
Do
these
sdks
have
more
than
just
tracing
and
does
it
make
sense
for
the
flag
to
be
more
General?
Should
it
be
like
a
suppress
everything
and
then
and
I
don't
know
I
at
least
chimed
in
with
my
two
cents
that
I
thought
that
people
are
going
to
want
to
be
probably
a
little
bit
more
they're,
probably
going
to
want
to
be
able
to
suppress
everything,
but
they're,
probably
going
to
want
to
be
granular
as
well.
D
So
but
maybe
a
suppress,
suppressed
Telemetry.
You
know,
method
with
various
parameters,
for
your
granularity
might
be
something
worth
considering,
but
at
any
rate
it
seems
like
what
we
learned
is
that,
in
addition
to
the
Ruby
and
JavaScript
implementations
of
this,
there
is
a
net
implementation
of
this
as
well
and
the
dot
net
implementation.
D
Somehow
it's
only
an
SDK
based
API.
So
it's
not
part
of
the
API
API,
it's
not
available
to
instrumentation
authors
and
others,
but
it
is
at
least
available
to
the
SDK,
I
believe
and
then
yeah
so
I
guess.
That
was
another
thing
and
then
there
is
also
an
Otep
that
was
referenced.
D
I,
don't
think
so.
I
think
this
was
an
attempt
for
Daniel
dyla
to
kind
of
spec
what
was
happening
in
JavaScript
in
the
past,
okay,
which.
D
All
right
giving
up
at
this
point
at
any
rate
there
there
is
an
Otep
about
suppressing,
while
suppressing
some
things
kind
of
by
layer
and
I
think
it
comes
from
the.net
world
as
well
and
I
guess.
This
is
something
that
the
Java
Sig,
the
Java
instrumentation
Sig
ended
up
implementing
and
they
said
it
has
been
useful
in
their
instrumentation
and
the
other
I
guess
interesting
thing
learned
during
this
conversation
was
that
JavaScript
has
a
Java
minus
the
script.
D
The
Java
instrumentation
Sync
has
a
ton
of
instrumentation
and
they
have
somehow
managed
to
get
by
without
needing
a
untraced
or
suppressed
tracing
flag,
which
is
a
little
bit
shocking,
but
I
think
worth
worth
noting
that
they
have
been
able
to
get
by
without
this.
D
So
that
was
one
kind
of
argument
that
maybe
we
need
to
think
twice
about
adding
this,
but
I
do
think
that
some
of
the
things
that
they
need
to
do
or
that
people
want
to
do
they
do
kind
of
send
them
down
the
custom,
sampler
route
or
like
a
rules-based
sampler
route.
So
it
might
not
be
quite
so
quite
so
straightforward,
so
I
think
that
ends
up
kind
of
being
that
ended
up
being
a
lot
of
the
the
conversation
is
that
a
lot
of
this
stuff
can
be
can
be
done.
D
If
you
know,
if
you
know
enough
about
Samplers
or
if
you
are
well
versed
in
in
Samplers,
but
it's
not
something
that
is
easy
for
everybody,
especially
newcomers
to
the
project,
and
then
I
think
there
are
definitely
a
lot
of
like
wishlist
items
in
sampling
that
might
even
kind
of
unlock
this
stuff
and
make
it
a
lot
easier
than
it
already
is.
B
F
Is
it
possible
that
they
have
like
an
instrumentation
proxy
right
like
they
do
some
sort
of
like?
Maybe
it's
not
by
code
manipulation?
Maybe
they
have
Dynamic
proxies
that
are
emitted
from
like
a
factory
or
something
of
that
nature,
but
and
then
basically
like
their
internal
instrumentations
or
their
exporters,
don't
generate
the
proxy
objects
in
those
cases,
yeah.
B
You're
asking
good
questions,
I,
don't
know
yeah,
it's
just
odd
to
me
that
a
language
wouldn't
have
this,
even
if
it's
only
used
for
that
use
case
and
not
used
for
like
broader
consumption
by
users.
But
that's
what
people
you
know
it's
work
to
find
out
and
I'm
not
going
to
do
that
work
so.
C
C
So
the
the
mechanism
to
declare
all
right
from
this
point
until
I
tell
you
otherwise
don't
generate
spans,
means
that
the
last
Trace
spans
parent
continues
to
convey
until
a
point
where
you
say
something
is
traced
and
then
spans
being
actually
sent,
can
pick
back
up
and
declare
a
parent
that
was
sent
and
your
Trace
hangs
together.
B
Yeah,
no
for
sure
you
know
it's
just
it's
nice
to
see
some
work
going
on
here,
Upstream
in
general,
so
yeah.
D
I
don't
know
the
answer
to
that
question,
but
I
think
that's
a
good
question,
because
that's
kind
of
our
core
use
for
untraceed
and
yeah,
so
I
think
I
did
find
the
otab
by
the
way,
and
this
this
might
be.
This
might
be
a
way
that
Java
is
handling
this,
but
it's
a
proposal
to
create
Clarity
on
instrumentation
layers,
interaction,
how
to
suppress
duplicate
layers,
EG,
multiple
instrumented,
HTTP,
client
layers,
so
I
guess
you
can
kind
of
this.
Otep
will
tell
you
how
to
combine
like
nested
clients,
bands,
for
example,
into
one.
D
D
Figuring
out
what
what
the
best
way
forward
for
this
is
because
it
does
seem
to
be
something
that
people
ask
for
and
need,
and
it
is
super
useful,
so
coming
up
with
a
a
good
set
of
tools
for
for
this
area,
and
it
might
be
more
than
one
thing,
I
think
makes
sense,
and
also
just
as
it
kind
of
relates
to.
We
already
have
an
implementation
of
of
entrees
like
would
we
would
we
be
able
to
and
willing
to
kind
of
update
our
implementation
to
work
as.
B
Marriage
five
days
ago,
I,
don't
think
it's
released,
but
yeah
I
think
we
may
be
our
by
the
pal
by
the
divine
intervention.
Aka
Francis
I
think
we
might
be
on
spec
here.
If
this
option
thing
gets
merged,
I
I
haven't
reviewed
it
so,
and
it
looks
like
there's
certain
Deco
pilot,
so
I
don't
know,
but.
B
D
Like
this
makes
it
basically
the
same
as
what
JavaScript
is
doing
so.
A
D
D
B
D
All
right,
so
is
there
any
further
discussion,
I
guess
on
on
the
spec
related
stuff?
This
was
the
last
issue.
If
not,
we
can
move
on
to
talking
about
our
repo.
D
F
The
only
thing
that
might
be
interesting
is
that
you've
been
playing
along
at
home
this
issue,
that
kind
of
like
Richard
has
been
working
through
or
Ricky
I'm,
sorry
Ricky
McMillan,
which
is
a
kind
of
a
curious
issue.
Where
he's
seeing
that
some
Port
processes
that
he's
running
seem
to
hang
indefinitely.
Oh.
C
The
thread
in
cncf
slang.
F
Yeah
I,
don't
know
how
we
could
help
him
really.
I
was
trying
to
figure
out
a
way
that
it
we
could
be
helpful
and
the
only
thing
that
I
can
think
of
was
having
some
sort
of
like
little
Baseline
app
that
you
used
to
take
a
look
at
some
of
the
interactions
of
four
processes,
threads
and
fibers
and
kind
of
document.
F
What
we
see,
or
at
least
have
some
sort
of
way
to
like
hey,
run
this
and
take
a
look
at
the
results
on
our
bee
spy
and
the
flame
graph
or
or
something
of
that
nature,
so
that
at
least
we
have
some
sort
of
a
reference
point
that
we
can
share
of
like.
This
is
what
you
should
see
if
you
Strays
a
process
or
if
you
profile
a
process.
A
F
Or
at
least
for
us,
I
have
like
some
expected
predicted
Behavior
that
we
can
tell
people
like
this
is
what
you
should
see
at
least
I,
so
we're
kind
of
just
like
I,
don't
know
guessing.
F
What's
the
only
interesting
a
conversation
other
than
that,
it's
just
been
like
housekeeping
City,
where,
like
I,
wanna
and
updated
a
bunch
of
stuff
for
Robocop
and
got
rid
of
a
bunch
of
unused
parameters
and
whatnot,
and
you
know
I-
think
we're
ready
to
start
doing
the
work
to
switch
over
to
active
support
notifications.
I
I
got
that
one
PR
marginal
release
out
for
it.
E
F
It
works
well
question
mark,
I,
haven't
deployed
it
anywhere
because
I
don't
use
active
support,
notifications
anywhere
or.
A
D
A
F
Other
big
thing:
that's
really
I,
don't
know
now
I'm
turning
this
into
like
a
I
guess,
a
catharsis
meaning
so
anyways.
If
we
can
get
that
kicked
off
that'd
be
great
if
anybody
has
capacity
to
like
kind
of
dig
into
that
and
help
us
with
that
people
kind
of
just
been
reporting
these
errors,
or
they
have
these
gems-
that,
for
whatever
reason,
change
the
public
API
of
active
record
and
then
things
break
and
I'm
like
I'm.
Sorry,
that's
not
a
bug!
Don't
change
the
API
of
active
record
I!
Don't
you
know.
F
A
E
Have
bandwidth
to
start
working
on
it
now
that
I'm
fully
back
so
that.
E
F
Of
fun
so
I,
my
gripe
has
been
so
far
as
that
you,
we
don't
notice
these
things
as
much,
but
what
I
tried
to
do,
at
least
with
the
RoboCop
upgrade,
was
to
submit
cops
into
their
own
PRS,
so
that
people
can
look
at
cops
and
like
agree
or
disagree
on
them
like
for,
or
should
I
say
winter
rules
in
context,
just
like
as
a
small
batch
of
changes,
so
that
they're
not
overwhelmed
by
like
hundreds
of
files
changing
or
whatever
right.
F
But
what
that
revealed
to
me
was
a
constant
seg
fault
and
container
startup
issues.
So
that's
what
prompted
me
to
go
and
redo
the
build,
and
it's
kind
of
like
I,
want
to
isolate
all
the
Kafka
related
gems,
because
those
segful
all
the
time,
but
that's
not
because
of
our
code.
It's
just
because
it's
SEC
faults
because
there's
bugs
in
there.
E
So
I
mean
for
that
specific
one
if
it's
easier
to
just
split
Kafka
out
into
its
own,
build
somehow
and
copy
and
paste
more
just
to
make
it
a
little
more
manageable,
I
think
that
would
be
okay
if
we
want
to
be
more
proactive
and
better
about
cleaning
it
up
properly,
I'm,
also,
okay,
with
that
too
I
don't
need
that
feel
too
bogged
down
and
trying
to
make
actions
work
well
if
we
can
just
copy
and
paste
our
way
out
of
it.
Yeah.
F
Yeah
I
think
like,
like
the
other,
the
other
situations
were
like
whatever
the
container
failed
to
start,
because
redis
took
too
long.
My
scale
took
too
long
to
start
on
the
particular
action
image
and
those
were
happening
over
and
over
and
I
think
we're
just
running.
So
you
know
eight
services
or
whatever
for
our
test
Suite
on
each
box
and
and
an
individual
test
run,
doesn't
run
all
of
the
gems
at
once
right.
F
E
Was
that
was
no
you're,
not
I,
know
you're,
not
exaggerating,
I
know
that
file.
We
can
definitely
split
that
up
more
like
that
was
a
very
like
initial
attempt
to
try
and
make
bring
some
sanity
to
our
builds
and
like
that,
may
not
be
the
right
way
to
do
it.
We
can
definitely
say
maybe
we
have
one
job
that
builds
everything
that
requires
redis
and
one
job
that
builds
everything
that
requires
Kaka,
and
that
way
we
don't
have
to
start
as
many.
That
would
be
a
totally
valid
Improvement
I
think,
like.
F
Yeah
that
that's
the
point,
that's
the
pr
that
I
set
up,
sent
up
and
I
put
the
linters
in
the
separate
build,
but
that
the
lenses
in
this
upper
build
is
a
waste
of
resources
and
time
so
I'm
gonna,
that's
the
thing
I'm
going
to
revert
and
then
just
keep
one
build
to
be
a
data
data
store,
specific
test,
job,
Matrix
or
whatever,
because
otherwise
we're
gonna
have
like
83
builds
again
and
I'm
like
I.
Don't
please
I
have
attic.
F
C
Ariel
you're
you're
grumpy
is
is
nicer
than
most
people's.
F
Apologize,
it's
okay,
man
just
jump
on
the
island
with
me.
I
just
feel
like
the
thing.
That's
really
bogging
me
down
and
the
thing
that's
making
me
really
feel
bad.
Is
this
threading
process
working
thing
which
is
like
when
situations
like
this
happen
and
it's
like
damn,
you
add
instrumentation,
and
it
makes
your
app
worse.
That
really
is
disheartening.
It's
not
a
good
look.
Yeah
yeah.
A
D
D
Like
I
think
you
were
kind
of
suggesting
us,
having
kind
of
like
a
threading,
threading,
async
playground
for
or
hotel
Ruby,
just
to
kind
of
demonstrate
what
we
expect
to
happen.
If
you,
you
know,
have.
D
You
know
have
are
tracing
things
across,
you
know
threads
or
fibers
and
then
are
also
adding
forking
into
the
mix
which,
as
I
say,
all
this.
These
are
all
recipes
for
disaster,
actually,
especially
the
spreading
plus
forking,
because,
usually
when
you,
if
you
have
multiple
threads
in
flight
and
you
Fork
only
the
main
thread
survives,
the
rest
of
them
die
and
that
is
usually
unexpected.
But
it's
the
way
things
work.
As
far
as
I
know,.
D
So
anyways,
so
two
options
would
be
for
us
to
kind
of
have
a
bit
of
a
playground
there,
where
we
try
to.
D
D
We
do
have
a
user
who's
having
these
issues,
while
it's
unlikely
that
they
can
kind
of
like
give
us
their
source
code
or
anything
to
at
that
level.
They
might
know
kind
of
enough
about
the
structure
of
their
program
to
kind
of
build
some
toy
version
of
it.
That
exhibits
the
problem,
and
that
would
be
something
that
we
could
actually
probably
try
to
make
some
progress
on
and
that
might
if,
if
we're
reaching
kind
of
like
a
dead
end-
and
we
don't
have
time
to
like
whip
up
our
own
playground
for
everything.
D
But
if
you
could
ask
ask
them
to
kind
of
Coach
them
through
kind
of
building
like
a
very
like
watered-down
version
of
what
they
actually
have
and
just
you
know,
focus
on
having
kind
of
the
same
constructs,
and
you
know
forking
at
the
same
point
in
time
as
their
original
app
does
and
just
seeing
seeing
if
they
can
get
like
a
minimal
viable
reproduction
that
they
can
toss
up
somewhere.
E
I,
don't
think
any
of
us
can
top
the
antelopes,
cantaloupe
joke
that
was
earlier.
That
was,
that
was
top
tier
other
updates,
like
I,
said
I'm
back
I'm
going
to
do
working
on
things,
I'm
gonna
I'm,
getting
my
bearings
again
should
be
more
active
I'm
going
to
work
on
the
active
support
notifications
things
a
bit.
There
was
a
question
that
was
brought
up
about
it.
That
was,
will
I
be
able
to
access
the
current.
F
E
Within
you
know,
if
we
convert
to
notification
based
instrumentation,
the
answer
is,
it
depends,
but
probably
so
I'll
verify
all
that
depends
on
how
the
notification
is
actually
emitted.
Upstream.
The
other
thing
is
metric
stuff
I'm,
going
to
try
and
get
the
lay
of
the
land
again.
I
know
we
had
a
person
from
thoughtpot
who
was
interested
in
helping
drive
that
to
completion
so
I'm
going
to
start
working
Earnest
on
that.
E
My
my
main
focus
for
the
next
few
weeks
is
going
to
be
how
tall
Ruby
for
my
day
job
unless
unless
management
informs
me,
otherwise
that
next
time
I
speak
to
them,
but
so
I
plan
on
being
around
and
yeah
do
that?
E
E
D
F
I
already
merged
those
Matt
where
you
been
I'm.
Sorry
man,
all
right,
although
as
soon
as
I
merge
everything
they
like
released.
Another
point:
you
know
a
minor
version
of
Robocop,
so
I'm
probably
gonna,
throw
that
up
there
pretty
soon
too,
but
no
I
think
the
only
outstanding
pairs
I
mean
you
can
click
through
them.
I
think
the
only
ones
that
are
left
are
kind
of
like
breaking
the
build
apart
and
dropping
real
rails
five
support,
which
will
be
a
big
change.
F
E
Yeah
I
mean
it's
not
super
important
I
was
I
was
mainly
curious
because
the
changes
that
were
being
made,
I
was
sort
of
like
I,
wonder
how
much
that
actually
helps
but
I'm
not
super
worried
about
it
either
way
like
if
you
they
they're,
harmless
changes,
I
think
so
one
way
or
the
other
should
be
fine.
E
Regarding
the
removing
rails,
five
support
I'm
generally
on
board
since
end
of
life,
I'm
wondering,
though,
how
many
users
that
would
actually
affect
and
I
know
there
is
some
awesome.
Hot
income
data
set
about
rubygems
downloads
and
I
can
never
find
it,
and
I
was
wondering
if
Rob
could
send
the
link
again.
Yep.
F
C
F
E
F
D
So
yeah
I
I
did
have
a
similar
question
when
I
saw
talk
of
of
removing
rails
five,
just
I,
don't
know
from
my
my
vendor
days
where
there
were
lots
of
Ruby
users.
It
was
always
surprising
to
see
how
old
of
stuff
people
would
run
forever.
A
F
So
one
thing
I
will
say:
is
we
haven't,
dropped
rail
support?
We
won't
be
sorry.
Let
me
take
a
couple
of
steps
back
here.
Our
instrumentations
will
still
work
with
the
rails
5.2.
You
just
have
to
be
pinned
to
the
version
that
supports
it.
F
Sure
we
just
will
not
make
any
other
changes
to
the
instrumentations
to
support
end
of
life.
Gems,
that's
reasonable
right
and
it's
kind
of
like
this
is
the
last
version
that
supports
it.
Stick
with
this,
you
you
have
that
constraint
with
every
other
Gem
and
Plug-In
or
whatever
in
the
world.
It's
like
you're,
stuck
at
a
yeah.
C
He
heard
you
that's
what
he
did
no
dance,
though,
sadly,
well.
E
Done
well,
I'm
on
board
with
dropping
rails
five
support.
Honestly
I
was
curious
about
how
many
users
that
might
affect,
but
I'm
only
mildly,
curious,
and
that's
it
not
I'm,
not
going
to
think
about
it
past.
This
meeting
I
think
I
think
you're
right.
They
can
tend
to
an
old
version
of
the
instrumentation
if
they
want
and
that's
probably
not
a
bad
thing.
I
feel
like
if
you're
running
an
end
of
life
rails
version,
you
should
be
used
to
that.
So.
C
That's
that's
the
coping
strategy
we've
come
up
with
for
our
honeycomb's
proprietary
beelines
at
a
certain
point,
we're
like
use
Hotel,
because
they're
tracking,
the
current
stuff,
but
if
you're
using
old
rails,
you're
probably
used
to
pinning
things
at
this
point
so
add
another
one
to
your
PIN
list.
C
C
C
A
C
Do
what's
that?
How
do
I
expand
time
upper
right
if
you
scroll
up
last
30
minutes
at
the
top?
There.
A
C
If
you,
if
you
want,
you
want
to
look
for
five,
if
you
scroll
up
to
the
top,
since
version
numbers
are
string.
Sadly,
if
you
scroll
up
to
the
query
boxes
up
top
and
then
the
where
box,
if
you
click
on
downloaded
gem
name
rails
up
top
that'll
turn
into
an
edit
box.
E
Looks
like
over
the
past
last
month,
I
mean
there's
several
hundred
thousand
downloads
of
5.2.8.1.
D
Work
with
them
to
make
something
work
for
them,
I
think
that's
that
will
be
a
a
good
test
of
the
process,
I
think
and
as
long
as
as
long
as
we
come
up
with
something
that
will
not
alienate
any
current
rails,
five
users
I
think
that's
that's
fine.
A
C
Which
is
something
that
we
could
consider
a
thing
I've
done
with
gems
in
the
past
was
declare
that
past
this
version.
We
don't
test
past
this
version
of
something
to
like
rails.
We
don't
test
past.
We
start
our
testing
at
version
six,
we
didn't
say
that
minimum
version
is
six
in
the
gym
metadata
because
it
might
work
it's
just.
We
don't
guarantee
it'll
work,
but
we
didn't
prevent
you
from
installing
the
current
version.
It's
just
it's
on
you
to
figure
out
if
it
works.
D
So
we
have
in
instrumentation
base
the
compatible
check.
D
This
might
be
an
inappropriate
place
to
actually
check
out
your
rails
version,
and
if
it
is,
you
know,
after
we've
dropped
rails
five
if
they
are
on
Rails,
five
have
a
nice
log
message
with.
C
A
F
F
F
F
Whatever
information
we
can
give
them
so
that
they
don't
get
caught
by
surprise,
attack.
C
F
C
It
we
can
also
put
a
in
in
trying
to
communicate
and
prepare
people
for
this.
We
could
open
a
discussion
in
the
repo
saying
we
intend
to
drop
support
for
rails
five,
as
of
this
date,
post
that
post,
a
link
to
that
discussion
into
the
cncf
slack
and
the
people
who
are
engaged
will
see
it
and
we'll
have.
However,
much
time
we
give
to
comment
on
it
and
before
we
start
making
any
code
changes.
Maybe
we
just
make
the
announcement
of
intent
and
then.
A
C
A
discussion
would
be
a
way
to
tempt
them
to
report.
E
D
Yeah
I
think
as
long
as
we're
not
alienating
users
and
I
think
this
will
kind
of
be
a
trial
run
for
how
we
are
going
to
handle
this
stuff
in
the
future.
Because
one
day,
rail
6
will
also
be
EOL
and
we
want
to
just
have
a
process.
So
we
can
select
that
out
of
the
out
of
the
Matrix
as
well
and
you're,
not
not
alienate
any
users
who
who
may
be
depending
on.
C
I
could
just
for
our
query:
interest
I
could
put
a
derived
column
on
this
Ruby
Ruby,
together,
gem
set
or
a
data
set
gem
set,
my
Ruby
and
my
honeycomb
have
collided,
and,
and
we
can
trim
I
can
have
a
a
Virtual
Field
here
that
we
could
query
on
or
at
least
see
the
results
for.
That's
just
major
minor
of
gem
version.
C
C
I'll
I'll
add
that
and
drop
a
link
in
the
slack
for
your
curiosity.