►
From YouTube: Service APIs Bi-Weekly Office Hours for 20200624
Description
Service APIs Bi-Weekly Office Hours for 20200624
A
All
right,
we're
recording
this
is
office
hours
for
service
api's
on
June
24
this
year
is
flying
by
alright,
so
I
know
Harry
had
an
update
he
wanted
to
give
on
TLS,
but
maybe
we'll
bump
that
down
a
bit
just
because
there's
not
too
many
people
here
yet
and
we'll
move
PR
review
up
a
bit
and
leave
some
time
at
the
end
to
go
through.
If
that
does
that
sound
all
right,
Harry
yep.
B
A
Okay,
with
that
said,
I
know,
there's
a
a
a
bunch
of
PRS
that
could
use
some
attention
here.
So
I
wanted
to
spend
a
bit
of
time
on
this.
We
haven't
had
much
time
in
office
hours
recently
to
go
through
and
just
talk
about,
pr's
together,
just
looking
at
who's
here,
maybe
we
can
pick
PRS
from
people
that
are
on
this
call
right
now
to
start
with
Harry
it
looks
like
this
is
approved
and
just
needs
an
LG
TM
is
that
yes,.
B
B
A
B
A
Well,
awesome
well,
I'll,
be
selfish,
then,
and
go
through
my
own
I
think
this
one
has
already
been
lgt
and
I.
Just
I
have
had
bad
luck
with
rebasing
after
lgt
em,
and
you
know
they
whatever
so
I.
Think
I
just
need
to
go
back
through
and
rebase
one
more
time,
and
hopefully
we
can
catch
approval
at
just
the
right
point
to
make
this
work.
I
think
this
is
a
fairly
non-controversial
one,
so
yeah
I
I
will
leave
that
E
for
now.
Are
there
any
other
PRS
that
we're
aware
of
that?
A
A
A
Let's
see
yeah
I
think
this.
All
this
all
seems
find
me
and
that
there
was
this
big
discussion
about
unnamed
address
and
I
think
we
just
decided
it
was
out
of
scope,
but
it's
certainly
something
that
we
want
some
kind
of
DNS
entry
or
were
something
here
that
can
help
us
represent
the
AWS
load.
Balancers.
C
C
A
A
This
PR
looks
massive
because
it
hasn't
been
rebased
after
vendor
was
removed
and
dependencies
were
updated.
I
think
this
wound
up
being
significantly
smaller
DIF
once
this
is
rebased
on
those
with
that
says,
this
opens
up
a
broader
conversation
of
if
this
is
something
we
want
to
move
forward
with
yeah.
D
One
interesting
thing
that
John
brought
up
that
wasn't
apparent
to
me
was
that,
because
the
is
my
audio
okay,
yeah
you're,
great
okay,
the
because
the
dynamic
client
has
performance
issues,
because
the
cache
stores
like
unconverted
types
or
something
I
didn't
realize
that
that
was
the
case.
I,
don't
know
if
any
of
the
folks
on
the
call
have
run
into
this
at
all.
B
A
I
mean
so
there's
there's
an
upstream
issue
on
KU
builder
Jakub
builder,
moved
away
from
including
client
Jen
as
part
of
coop
builder,
and
so
there
was
some
discussion
on.
Why
and
part
of
that,
as
I
recall,
was
I,
don't
have
the
issue
right
now,
I'm,
not
sure
if
it's
referenced
here
but
yeah
that
there
was
discussion
on.
Why
and
part
of
that
was
well.
You
could
just
use
dynamic,
client
and
the
other
and
I
think
there
was
a
question
about.
A
If
you
know
if
we
could
just
generate
our
own
client
Jen
Lister
whatever
and
it
sounds
like
they
just
they
don't
mind,
there's
nothing
against
using
your
own
custom,
client
Jen,
climbing
your
own
custom
generated,
client,
client
and
Lister
Informer's.
They
just
didn't
want
to
maintain
it
I.
That's
my
recollection
of
that
issue.
I
think
I
know
at
least
Bowie
has
read
through
that
issue
and
commented
on
it
right.
D
Sally
I
forget
his
last
name
directly:
Sally
Ross
yeah,
one
of
the
people
who
works
on
that
and
basically
they
were
like.
Well,
you
could
do
this
and
but
it
won't
be
supported
in
queue
builder,
and
this
is
a
little
bit
finicky,
so
it
sounded
like
it
would
keep
working.
No
one
said
that
it
was
going
to
kind
of
stop
working.
B
B
E
D
Okay,
it
seems
like
I
can
take
a
look
and
see
if
this
is
just
merge,
able
I'll
ask
him
to
well,
it's
definitely
not
merge
able.
Yes,
they
like
me
do
to
commit
I'll,
follow
up
with
John.
A
Let
me
see
what
else
we
have
going
on
dependencies
update.
I
think
this
was
pretty
non-controversial
as
well.
It's
got
an
LG
TM
yeah
from
James
I.
Think
myself,
yeah,
okay,
yeah!
We
can
merge
that
one,
okay,
cool
and
okay
are
there
any
any
other
appears
we
want
to
cover
here
before
we
move
into
the
rest
of
the
content.
A
B
D
B
B
Essentially,
if
you
have
a
TCP
listener
on
the
gateway,
it
will
use
a
route
selector
to
select
a
TCP
route.
In
this
case,
it
should
only
select
one
TCB
route,
because
if
you
have
multiple
TCP
routes,
you
will
have
lesion,
so
so
that
seemed
pretty
straightforward,
so
that
applies
for
TCP
route
and
TLS.
Now
know
that
professor
specifically
does
not
introduce
those
types,
but
it's
more
about
how
defining
how
matching
behavior
works
and
then
the
second
thing
was
thing
Bui
brought.
B
This
up
was
very
interesting
in
a
very
important
point:
that
what
is
the
minimum
that
we
can
do
inside
route
11
for
TLS?
So
now
I
think
we
sorry
rubbish.
You
can
go
into
the
proposal
and
enter
a
table
of
content
scroll
up.
There
is
this
TLS
properties
from
the
bottom
quarter.
Yeah
I
did
fear
of
elaborated
on.
B
You
know
the
properties
that
are
being
introduced
as
part
of
this
proposal
in
the
Gateway
and
in
the
route
and
in
the
router
we
proposed
that
we
can
restrict
this
to
to
have
only
certificate
and
ebay
references
inside
that
out.
Everything
else
will
be
always
controlled
by
the
Gateway
level,
TLS
properties,
so
that
probably
achieves.
B
Something
like
not
making
routes
too
complicated
and
still
allowing
for
the
some
use
cases
or
the
worst
user
stories
that
we
have.
You
know
where
application
developers
would
like
to
specify
DNS
settings.
So
those
are
the
two
changes
at
the
high
level
that
I
have
made
in
this
one.
He
definitely
do
want
to
discuss
more.
If
you
know,
TLS
and
route
is
something
we
want
to
do
or
not.
He
right
now
don't
have
like
the
good
consensus,
I
think
around
it.
Some
people
wanted
some
people
are
opposed
to
it.
Yeah.
D
Thank
you
like
I,
guess
we
would.
It
would
be
nice
to
see
if
we
could
write
out
a
concise
way
to
validate
that.
The
combination
makes
sense
like
if
you
had,
you
can
delegate
it,
but
then
you
probably
can't
have
any
config
at
the
gateway
level
or
if
there's
like
only
a
wild-card
there,
it
would
be
nice
to
if
we
could
work
through
those
because
then
would
be
make
it
very
like
that
would
be.
My
concern
is
that
we
can
get
into
situations
that
are
kind
of
tricky.
B
Yes,
yes,
so
there
are
some
of
at
least
a
few
of
those
situations,
and
we
need
to
be
explicitly.
You
know
explicitly
document
at
least
the
partner,
and
you
know
how
conflict
resolution
works
in
those
cases
so
I
can
I
can
take
those
things
where
you
know.
I
can
write
up
some
examples
of
how,
if
you
have
some
domains
that
are
being
controlled
by
the
Gateway
and
some
that
are
being
dedicated,
how
how
they
work
well
together
and
how
they
fail.
D
Yeah,
like
the
ideal,
would
be
that
for
the
tricky
ones
we
could
just.
We
could
basically
just
allow
them,
and
it
will
be
very
obvious
that
why
it
was
disallowed
from
my
user
perspective,
like
the
user,
can't
easily
end
up
generating
these
situations
and
then,
when
it
happens
like
they
can
understand
why.
It's
like
not
valid.
B
D
C
D
C
A
B
D
B
A
So
forth,
you
know
I
think
we
can't
go
too
deep
into
conflict
handling
and
without
much
structure,
but
maybe
it
would
be
worthwhile
to
just
try
to
list
off
the
different
kinds
of
conflict
we
might
need
to
handle,
not
necessarily
how
we
would
handle
it,
but
places
where
things
can.
Conflict
and
I
am
open
to
ideas
here.
I.
A
B
E
I
mean
we've
talked
before
about
config,
being
merge
Abul
on
other
calls,
and
so
I
think
that
I
mean
that
that
is
obviously
a
type
of
confident
illusion.
So
yeah
it's
conceivable.
That
might
be
an
answer
that
you
could
give,
but
yeah
I
think
the
the
the
obvious
ones
that
came
to
me
yeah,
the
stuff
that
the
stuff
that
the
current
ingress
kind
of
is
one
of
the
things
that
poops.
Sometimes
people
really
like
that
behavior,
where
you
can
specify
the
same
domain
multiple
times
and
you
currently
that
gets
merged
you.
E
If
you,
if
you
use
it
depends
on
the
controller.
Obviously,
but
you
know
our
old
friend
to
the
index
controller.
You
know
it
will
for
ingress
objects,
will
merge,
matching
domain
names
and
matching
paths.
So
if
you
have,
if
you
have
to
English
records
that
match
up
on
both
the
domain
name
and
the
path,
the
at
the
end
of
the
day
learned
upload
balanced
across
the
two
things.
E
D
A
D
E
I
mean
the
stuff
that
we've
had
happen,
the
way
that
we
handle
all
some
of
these
in
Congo,
because
we've
because
the
the
reason
the
conflict
resolution
is
important,
is
that
the
usual
ways
in
which
you
handle
a
conflict
like
this
are
not
straight
forward
in
Coverity.
So
if
you
wanted
to
do
first
right
or
win
semantics,
it's
actually
very
difficult
to
do.
You
know
and
ways
that
you
would
determine
which,
if
there
is
a
conflict
which
one
wins
is
not
straightforward,
and
that's
why
the
conflict
resolution
behavior
is
really
important.
Yeah.
B
B
E
Yeah
I
mean
I.
Think
one
of
the
one
of
the
things
that
we've
seen
is
that
you
can
very
easily
end
up.
You
know
one
of
the
reasons
that
we
built
has
to
be
proxy.
The
way
we
did
is
to
stop
the
problem
that
a
lot
of
people
saw
with
ingress
with
ingress,
where,
if
you
deploy
something
that
conflicts
with
someone
else
and
it's
broken,
then
you
can
break
a
significant
chunk
of
a
working
thing,
and
so
I
think
that
it
feels
to
me
like
one
of
the
guiding
principles
should
be
that
you
should.
A
E
D
E
Contour
we
specifically
chose
not
to
do
that
because
it
may
it
was
too
hard.
So
if
you
have
a
conflict
in
contour,
everything
that's
conflicted.
Is
invalid.
Garlic
produces
the
exact
opposite
of
that
thing
where,
if
you
create
a
concluding
HTTP
proxy,
you
know
that
happens
to
conflict
on
your
main
website.
Didn't
guess
what
you
just
took
your
main
website
down:
yeah
I.
E
A
A
Think
you
know
I'm
just
thinking
about
other
forms
of
potential
conflict
and
this
I
don't
know
that
this
is
I,
don't
know
exactly
how
you
this
would
classify,
but
multiple
gateways
that
target
the
same
route.
I,
don't
think
that's
necessarily
an
issue,
but
it's
just.
It
seems
like
something
like
that.
We
need
to
at
least
you
need
to
yeah.
E
A
Yeah,
maybe
you
know
maybe
another
thing
worth
discussing
is
things
that
have
consistently
been
challenging
with,
let's
say
the
ingress,
API
or
other
API
is
in
this
space.
Where
conflict
resolution
has
been
painful
or
we're
not
straightforward,
so
I
know
like
for
ingress.
We
we
for
sure
know
that
it's
very
difficult-
or
this
is
kubernetes
in
general,
it's
very
difficult
to
understand,
which
was
the
last
updated
yeah.
B
A
B
It
really
like
having
TLS
only
on
gateway
minimizes,
that
to
some
extent
it
still
creates
an
issue
where
you
could
allow
for
a
TLS
certificate.
No
one
TLS
certificate
on
the
support
for
four
three
and
another
TLS
certificate
for
the
same
SNI
on
code,
four,
four,
eight
right,
and
so
so
that
does
create
those
issues
and
I
think
most
proxies
cannot
handle.
You
know
like
presenting
different
so
like
most
of
this
can
actually
handle
present
in
different
certificates
on
different
listeners.
B
The
real
issue
comes
up
when
you
delegate
DLS
on
tour
out,
and
this
is
what
pooi
was
also
referring
to.
You
know
if
we
can
sketch
out
those
situations,
because
that's
one
of
you
medicated
now,
anybody
can
introduce
a
TLS
certificate
for
foo.com
in
his
round
and
then
someone
else
goes
and
creates
the
same
similar
mouth
and
we
again
have
the
same
thing
as
and
get
a
3-1.
So
in
that
case
you
know
we
need
sort
of
a
guiding
principle
is
to
know
which
one
takes
precedence
and
how
to
highlight
it.
D
C
E
Sorry
I
just
I
just
realized,
remember
something
that
is
pretty
relevant.
We
actually
have
a
problem
in
contour
at
the
moment
that
we
discovered
where
the
it's.
E
Can
create
weird
problems
for
certificates,
the
if
you
are
doing
a
wild
card
cert
and
you've
got
your
doing
and
you've
got
like
individual
SSN.
Is
that
then
route
to
different
things,
browsers
aggressively
will
try
to
coalesce
the
connections
and
what
the
way
that
it
should
work
in
is
that
when
you
hit
the
proxy
asking,
for
you
know
food
calm
and
it's
the
same
IP
as
bar
calm,
your
browser
will
say:
are
the
dis
domain
name
resolves
to
this
IP
or
that
so
it's
they
come
it.
E
Sometimes
it'll,
try
and
the
proxy
that
the
the
Habeas
supposed
to
do
is
just
is
to
call
us
the
connections
and
send
them
to
the
same
place
with
the
same
and
with
the
same
s
and
I
I'm.
Even
it's
actually
headed
for
a
different
place,
and
so
you,
basically
what
happens,
is
you're
supposed
to
issue
the
proxy
is
supposed
to
eat
you.
A
41
I've
go
some
41
misdirected
requests
and
some
browsers,
Safari
I
don't
handle
that
and
they
will
just
show
you.
E
E
Yeah
because
it's
so
we
had
the
problem
where
people
were
like.
You
know
my
like
some
some
I'm
using
a
wild-card
search
and
some
requests
are
just
not
working
yeah
as
like,
okay
and
then
so.
It
came
down
to
the
fact
that
yeah
browsers
will
do
this
aggressive
connection
coalescing.
We
actually
had
to
add,
because
we
use
envoys
that
back
in
proxy,
we
actually
had
to
had
a
lower
filter
to
the
connections,
to
like
check
the
S&I
on
the
host
header
matter.
E
Something
james
deal
well,
yeah,
so
yeah
it
was
a
bit
gross,
but
yeah
we
needed
to
do
that
so
because,
if
you
don't
do
that,
then
you
can
and
you've
announced
client
certificate
authentication.
Then
you
can
connect
to
the
proxy
with
with
any
valid
TLS
session
and
bypass
the
client
certificate,
authentication,
yeah.
B
This
is
super
painful
in
the
next
verse,
the
opposite,
where
it
does
the
same
thing.
What
the
browser
is
doing
to
the
upstream.
It
will
multiplex
multiple
connections
of
there.
They
ever
thought
if
they
are
going
to
the
same
IP
for
toppled-
and
you
know
it
will
put
all
the
requests
for
different
lesson
eyes
on
the
same
thing
and
that
my
bus
is
for
flying
authentication.
Yes,.
E
Yeah,
so
we
built
the
thing
a
while
ago
to
stop
that
happening
when
we
built
client
authentication.
But
then
we
found
that
yeah.
You
have
this
problem
that
causes
this
problem.
Yeah
they
come
with.
The
combination
of
client,
authentication
and
SNA
can
make
things
weird,
because
I
guess
the
the
headline
that
I
wanted
to
say,
though
yeah
that's.
B
We
don't
before
not
to
break
things
that
are
working
so
in
that
section
is
it
we
can.
There
are
some
times
issues
where
you
know
you
have
a
large
object
and
you're
like
five
ingress
rooms,
and
only
one
of
them
expired,
and
it's
pretty
hard
to
actually
signal
that
only
one
resources,
much
everything
is
classified,
and
so
we
probably
want
to
also
have
some
fine-grain
yeah.
B
E
The
that
comes
down
in
my
mind
to
the
to
the
problem
of
defining
the
status,
how
the
status
pipes
I
have
been
I,
have
been
pushing
some
more
on
trying
to
get
some
more
clarity
from
people
around
conditions
from
cigar
around
conditions,
and
you
know
what
we
should
know
the
parity.
What
the
expected
behavior
is
all
that
sort
of
stuff
still
no
agreement
on
the
polarity
stuff,
which
is
what
you
would
need
to
be
able
to
make
this
stuff
machine
possible.
E
But
one
of
the
things
that
did
come
up
out
of
that
is
something
that
you
would
talk.
You
were
talking
to
James
about
yesterday.
Rob
that
conditions.
It
seems
that
the
some
at
least
of
cigar
could
actually
feel
that
conditions
shouldn't
appear
and
disappear.
If
a
condition
is
valid
on
an
object,
it
should
always
be
there.
It
might
be
unknown
true
or
false,
but
it
should
always
be
there.
So
interesting,
yeah
yeah.
So
I
was
like.
Oh
that's,
not
how
I
thought
of
them
earlier
at
all
yeah
and
when
I
was.
E
That
that
was
how
some
people
thought
of
them.
Then
it's
like
whoa.
Okay,
that's
a
that's
a
reasonably
big
behavior
change.
But
yet,
if
you
look
at
a
lot
of
the
core
objects,
that's
how
the
conditions,
the
choreography
because
have
conditions
the
conditions
are
always
present
and
they
all
you
know,
have
different.
You
know
have
different.
Just
the
the
true/false
unknown
value
will
change.
Yeah.
A
E
Yeah
exactly
so
I
think
the
yeah
that
the
reason
this
is
relevant
is
because
you're
treating
something
that's
partially
valid.
Really.
The
only
way
that
you
can
do
anything
about
that
is
for
the
control
is
for
there
to
be
a
schema
on
the
status
that
the
controller
can
update.
To
say,
you
know,
hey
your
conflict
as
most
is
mostly
good
yeah.
Overall,
your
thing
is
ready,
you
know
because,
but
there
are
problems.
A
Yeah
that
makes
sense
I
mean.
Is
this
leading
to
a
place
where
we
kind
of
define
a
a
minimum
segment
of
a
resource
that
needs
to
be
valid
for
it
to
be
at
least
partially
valid?
So
as
an
example,
one
listener
may
be
valid
on
a
gateway
and
another
one
may
not,
but
you
still,
you
still
make
that
unique,
valid
listener
work.
Yet.
E
C
E
E
A
E
The
guidance
from
the
current
API
conventions
to
show
you
that
conditions
should
be
abnormal
true
polarity,
ie.
If
something
is
wrong,
they
are
true
that
the
folks
from
K
native
did
a
whole
bunch
of
UX
research
and
they
found
that
when
you
actually
show
people
abnormal
true,
could
it's
very
hard
to
parse
them
initially,
because
it's
kind
of
its
kind
of
like
the
good
state
is
a
double
negative,
yeah,
easier,
false
like
ever-present
yeah
and
then
like
yeah.
So
it.
E
E
That's
why
yeah
and
the
current
rule
that
what
their
name
was
sorry
I
was
reading
it
before
this
meeting,
but
still
really
for
me,
so
Daniel
I
think
yeah
Dana
Smith
has
replied
on
one
of
the
four
five
to
one
in
the
community.
Repo
is
the
is
Evan
from
Katie's
PR.
That
was,
let's
change.
The
I'll
put
it
in
the
chat.
E
E
Well,
it
feels
like
there's
a
bit
of
a
split
in
you
got
to
chew
anyway,
but
yes,
so
the
latest
thing
is
that
Daniel
C
there's
there's
a
lot
a
lot
of
backers
and
force
here
if
this
is
yeah
so
polarities
for
people,
so
the
state,
the
condition
in
a
natural
way
for
humans
to
understand,
yeah
and
then
absent
means
you
don't
know
the
value
I
mean
you
also
don't
know.
If
the
setting
is
running
centers
of
conditioned
shoots,
set
the
condition
for
the
objects
that
are
attempting
to
monitor
straight
away.
A
Yeah,
this
is
helpful
yeah.
Thank
you
for
bringing
this
in
this
out.
You
know
I
feel
like
the
question
I
asked
in
slack
the
other
day
was
already
kind
of
weighted
in
in
a
bad
way,
because
I
was
asking
about
a
condition
that
didn't
make
any
sense
in
a
false
state.
The
condition
no
such
gateway
class
equal
false,
is
very
confusing,
like,
like
you
were
dealing
with
K
native,
like
that.
How
do
you
understand
that,
as
someone
like.
A
E
Yeah,
that's
why
I
sort
of
said
that
the
approach
and
I'm,
probably
gonna,
actually
try
and
write
a
response
is
to
say
if
conditions
should
already
exist,
so
it
always
exists.
It
actually
makes
more
sense
for
them
to
be
positive,
polarity,
it's
easier
to
read
if
they
are
positive,
because
then
then
it's
like
everything
is
good.
True
is
what
you
want
to
say.
E
You
just
want
to
say
true,
true
true,
true
true
true,
true
yeah,
you
know
and
then
it's
easier
to
pause
them,
it's
easier
to
understand
what
they're
doing
as
a
person,
yeah,
I'm
beer
so
and
then
I
mean
that's
super
relevant
to
the
thing
we
were
originally
talking
about,
because
because
the
only
way
to
communicate
partial
validity
is
through
the
use
of
something
like
conditions.
Even
if
you
were
to
build
something
separate,
it
would
use
a
similar
mechanism.
E
A
Yeah
I'll
have
to
spend
some
time
thinking
about
that,
but
I
know
our
conditions
right
now
do
not
currently,
or
at
least
the
conditions.
I've
added
do
not
currently
follow
this.
This
guidance
but
I.
E
A
Yeah,
hopefully
we
can
get
to
view
an
alpha
one
and
then
I'm
some
kind
of
my
backwards
compared
to
me,
but
at
least
be
a
little
slower
to
to
break
things,
but
right
now
all
we
can
break
things
be
great
to
get
status
right
so
yeah
it
was
there
anything
else
we
want
to
I
guess,
maybe
that
the
right
question
is
one
of
the
next
steps
for
this.
This
conflict
handling
semantics
discussion.
They
like
there
there's
a
lot
of
different
things
we
can
cover
here
Minh.
A
We
could
dig
through
what
happens
in
every
single
one
of
these
potential
conflicts
or
what's
happened
like
maybe
we
need
to
discuss
each
of
these
individually
and
how
we
would
avoid
these
potential
problems
or
handle
these
issues,
these
conflicts,
or
maybe
we
need
to
first
define
in
some
way,
whether
it's
in
documentation
or
what
what
what
our
guiding
principles
are
like
the
prefer
not
to
break
things.
Things
that
are
partially
valid
should
be
parsed
and
used,
however,
possible
yeah.
B
I
want
that
and
I
think
so,
since
we're
all
writing
different
controllers
and
then
all
of
our
controllers
are
different
behaviors.
What
we
could
do
is
we
can
take
a
couple
of
you
know,
common
examples
that
are
present
in
most
mendacious,
and
you
know
we
all
have
some
experience
around
it
and
then
we
share.
You
know
how
we
think
the
the
config
resolution
should
work.
B
So
with
that
we
can
arrive
at
sort
of
a
shared
understanding
and
then
derive
like
sort
of
guiding
principles
based
on
those,
because
we
can
have
high-level
guiding
principles,
but
then
our
understanding
still
could
be
very
different
and
then,
when
we
go
and
actually
drive,
okay
answer
questions:
okay,
how
should
conflict
the
resolution
happen
in
this
case.
They're
out
is
clashing
between
gateways
or
something
else,
and
then
we
come
with
different
answers.
A
A
D
A
A
A
This
should
be
parsed
and
should
be
handled
as
long
as
there's
sufficient
information
in
them
to
get
some
validity.
So
maybe
even
just
starting
that
discussion
and
making
sure
that
there's
enough
support
for
that
approach
is
good
and
then
and
then
we
can
go
from
there
and
talk
about
different.
How
how
different
implementations
are
currently
handling
conflict
resolution
and
maybe
try
and
find
some
common
ground.
A
A
Cool
I
will
take
that
one
on
and
then
the
last
thing
on
our
agenda
today
was
just
if
we
could
schedule
some
time
to
read
through
the
API
spec.
This
was
an
idea
that
James
had
and
maybe
I'll,
maybe
I'll
wait
till
I
get
a
chance
to
talk
with
him
again
because
I
know
I.
He
would
want
to
be
involved,
but
it
was
a
good
idea.
It
was
basically
at
some
point
in
the
not-too-distant
future.
A
We
should
just
spend
some
time
together
just
reading
through
the
current
tight
stock
go
and
make
sure
we
haven't
missed
something
like
really
obvious,
and
maybe
maybe
that
takes
place
and
then
upcoming
office
hours,
and
we
just
announced
it
schedule
and
have
been
it
a
bit
in
advance
because
it
would
be
a
bit
different
than
your
usual
office
hours,
but
yeah
I.
Think
I
think
this
would
be
would
be
helpful
for
everyone
as
work
approaching
that
V
1
alpha
1
release
or
you
know,
I
hate
to
say
final,
but
some
form
of
API
stability.