►
From YouTube: Config Working Group 5/2/2019
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
B
B
Requests
either
specific
config
types,
yeah
basically
have
some
way
for
pilot
to
not
just
sit
and
wait
for
a
first
push
of
every
single
config
type,
and
so
we
had
kind
of
two
ideas
around
that
one
was
to
kind
of
more
short
term
thinking
but
like
to
start
up
the
pilot
discovery
binary
with
some
kind
of
configuration
that
allows
you
to
say.
Okay,
when
pilot
starts
up,
it
only
needs
to
accept
these
five
types
of
config
or
all
the
config.
B
The
default
would
be
like
all
the
config,
and
maybe
it
could
have
configuration
to
accept
just
some
of
the
types
and
then
potentially
longer
term.
We
were.
We
wanted
to
bring
up
and
consider
this
idea
of
in
mCP
itself,
part
of
like
the
client
and
the
server
talking
to
each
other.
At
the
beginning,
there
would
be
some
kind
of
discoverability
function
where,
like
the
client,
maybe
asked
the
server
what
kinds
of
config
types
do
you
configure
and
then
only
expect
the
after
that
point?
Okay,.
A
So
I
need
to
go
back
and
check
one
thing,
but
I
think
we
can't
have
a
third
option
and
I'm
not
going
to
present
you
right
there.
This
is
gonna,
be
feasible
or
not,
but
we
should
differently
considers
and
that
would
be
then
pilot
request,
a
collection
that
doesn't
exist
Eames
if
he
can
automatically
say
no
right.
A
The
the
stack
itself
can
say
that
right,
this
kind
of
I
think
alleviates
some
of
the
concerns
that,
for
example,
custom
has
I
am
sad.
That
cost
is
going
from.
You
know
he
doesn't
want
to
like.
You
know,
create
this
like
swim
in
mobs
and
livers
in
this
system
and
cause
like
many
poor
things,
possibly
go
wrong
right,
so
we
want
to
have
like
a
very
robust
system
right.
That's
his
goal
and
I
kind
of
agree
with
the
goal
right
so
but
I
also
am
NOT
happy
with
the
idea
of
sending
yes
to
every
request.
A
Like
you
know,
hey
I
want
food.
Yes,
it's
an
intercept,
Hugo
right,
so
I
think
like
if
the
limits
you
want
to
have
the
discovery
server
so
that
somebody
can
actually
like
if
the
mCP
goes
more
advanced
and
we
have
more
use
cases,
somebody
can
do
a
method
programming
with
services
right.
You
know
you
can
get
to
discovery
and
then
decide
on
things,
but
otherwise
PI.
You
should
be.
A
Need
right
at
these
versions
and
I'm
gonna
go
ahead
and
ask
for
each
one
of
them
from
the
server
right
and
the
server
can
say:
okay,
here's
the
15
things
I
know,
and
for
these
five
things,
I
don't
have
them.
So
if
you,
if
you
know
it,
that's
why
I
think
things
should
still
be
robust.
The
pilot
side,
you
can
make
a
point
wise
decision,
saying
that,
okay,
you
know,
I
will
never
get
any
resources
for
this
thing.
B
A
A
Look
at
me
make
a
decision
at
that
point
in
pilot,
saying
that
okay,
we're
just
gonna
log
this
or
we're
just
gonna-
have
a
sign
there,
telemetry
so
that
it
can
be
observed,
but
otherwise
I
will
start
up
and
you
know
treat
that
collection
as
an
empty
internally
right,
I
kind
of
see
that
long,
as
is
a
good
compromise,
because
you
know
there's
much
less
amount
of
knobs
and
levers.
You
know
it's
kind
of
autocorrect
and
I
think
it
had
those
things
like
upgrade
scenarios
ready.
You
know,
I
have
an
older
gal.
A
C
Gonna
go
ahead,
I
worry
so
that
sounds
like
air
suppression
to
me,
and
I
worry
about
that,
because
it
could
hide
mismatches
in
version
numbers
or
like
bugs
between
the
two
of
them,
because
it's
like,
if
pilot
is
like,
if
pilot,
can
warm
up
just
fine
with
its
configurations
or
source
responding
to
nothing
and
it's
like
Oh
pilot's
running
fine,
but
the
configuration
source
is
saying:
oh
none
of
these
exist
I
can
provide
no
useful
configuration.
I
would
rather
have
an
error
in
that
scenario.
Yeah.
A
But
that's
the
thing
now
you
can
actually
make
that
decision
in
pilot
in
pilot.
You
can
make
mountain
of
your
configuration
becomes.
These
are
the
things
to
ask
for,
but
it's
not
gonna
be.
These
are
the
things
to
ask
for
what
to
do
when
you
figure
out
certain
collections
doesn't
exists.
That
you're
asking
for
say.
C
A
Will
all
it
has
already
I
have
an
enhanced
set
of
connections
that
it
was
looking
for
already
right,
you
just
keep
looking
for
those
and
then,
if
you
can
actually
now
create
infrastructure
or
make
a
policy
decision
between
pilots
in
the
case
where
it
doesn't
find
one
or
more
of
these
things.
You
know
your
policy
could
be
if
you
cannot
find
any
networking
collection
with
networking.
Related
connections
do
not
start
up,
but
otherwise
everything
else
is
should
be
fine,
for
example
right.
A
A
B
There
would
be
some
set
of
like
required
config,
so
there
would
be
logic
in
pilot
that
says.
Okay,
these
don't
show
up.
There
will
be
an
error,
but
otherwise
it
it
can
start
up.
Just
fine.
Does
that
mean
that
the
mCP
server
has
to
indicate
some
way
to
pilot
that
it's
not
going
to
send
certain
types
of
configuration,
yeah.
A
I
kind
of
see
that,
as
like
think
of
the
HTTP
server
case
right,
when
are
you
connect
and
you
request
a
resource
and
you
get
200
and
we
get
the
resource
or
it
gets
404
right.
It's
kind
of
the
same
idea
you
connect,
then
you
ask
for
something
and
it's
not
there.
So
you
get
a
404
now
you
need
to
make
a
decision
right.
A
A
You
know
we
actually
have
a
post
decision
in
Pinal
and
the
post
can
be
you
know,
hey
the
collection
doesn't
exist,
oh
well,
you
know
we're
just
gonna
move
on
right,
which
is
which
aligns
with
the
robustness
expectation,
I,
believe
that
customers
right
things
can
just
start
up
or
you
can
actually
not
additional
class.
If
you
care
about
the
scenario
saying
that,
like
you
know,
that's
important,
we
want
to
be
able
to
see
when
collections
doesn't
exist.
If
you
find
something
that
doesn't
exist,
exits
right.
B
Mm-Hmm
so
so,
there's
like
two
putts
of
logic,
one
where
we
say
pilot
when
you
start
up
like
please
actually
error,
if
you
don't
get
all
of
these
configs
and
then
by
default,
there's
a
couple
of
like
required
things
that
pilot
needs
to
come
up
anyway.
Make
sure
that
those
are
there
and
error,
if
not,
but
otherwise
go
ahead.
A
So
I
don't
want
to
design
the
behavior
of
pilots
here
right
here
right
now,
all
I
am
saying
yes,
okay,
that
becomes
an
abstract
bubble
of
policy
decision,
and
now
we
can
have
a
further
discussion
with
the
networking
foulest
and
everything
about
I
cannot
okay.
How
do
we
handle
this
right
so,
but
the
default
behavior
would
be
like
enough
for
pilots.
If
you
supply
nothing
else
starts
up
asks
for
collections.
One
of
the
collections
is
not
found.
It'll
just
keep
functioning.
A
C
Okay,
and
so
that's
that's-
that
was
written,
that
was
that
code
was
introduced
in
this
most
recent
version
and
we're
considering
changing
that
because
it
didn't
used
to
like.
Obviously
it
didn't
used
to
complain
if
it
didn't
get
all
collection
sources
like
it
sounds
like
we're.
Reverting
back
to
previous
behavior
and
so
I'm
wondering
if
there
was
a
conversation
involved
in
like
making
pilot
refused
to
configure
any
on
voice
until
it's
completely
warmed
up
or
if
that
was
just
like
something
that
you
know
got
merged.
It.
A
A
We
actually
are
saying,
although
this
collection
doesn't
exist,
okay,
so
the
first
like
click,
stop
zero
of
this
implementation
would
be
once
you
get
it,
you
treat
it
as
a
okay.
We
are
synced
right.
We
are
not
going
to
get
anything
else
from
from
this
collection
anymore
right,
but
we
don't
drop
the
connection.
You
know
we
don't
treat
this
as
a
like
a
major
failure
case.
We
are
just
gonna
start
up
right
because
you
will
never
get
any
color
any
data
from
that
collection
from
this
server
right
now,
click
stop
one.
The
next
step
is.
A
C
C
A
Think
needs
to
be
called
that
one
of
them
is
I
need
to
go
and
double
check.
If
this
is
possible
in
this
streaming
case,
it
should
be
right.
So
we
may
need
to
change
the
protocol
for
this,
but
it
should
be
possible
to
add
this
one
to
the
protocol.
The
second
thing
is
on
the
cid
side,
there's
also
an
additional
concern
which
I
don't
think
you
guys
are
interested
in
but
I.
Nevertheless,
so
we
have
cases
very
unlike
communities.
A
A
So
then,
when
Pilate
asks
for
a
gateway,
even
though
the
CID
doesn't
exist,
gali
should
still
say.
Okay,
you
know,
here's
like
a
gateway
connection,
and
it
just
says
empty
stuff,
because
over
time
the
CID
can
be
introduced,
and
then
the
Gateway
resource
can
appear
at
some
point.
So
but
that's
essentially
a
galley
problem
more
than
a
pile
of
mCP
problem.
B
A
Unsupported
collection,
support
to
MCP,
changing
the
behaviors
so
that
we
don't
drop
the
connection
and
the
collections
that
found
and
on
pilot
side
treat
that
as
okay.
This
is
fully
Singh
right,
so
which
would
be
separate
from
at
the
discovery
discussion
which
I
think
mCP
should
still
have
at
the
limit.
I
think
that's
still
a
good
idea,
just
that
I,
don't
think
it
is
needed
to
solve
this
particular
problem.
A
B
A
First,
create
the
propose
a
lot
of
this,
so
that
are
like
I
mean
this
is
essentially
me
kind
of
proposing
this
idea
at
this
point
right
so
I
think
we
can
start
a
proposal
and
get
every
more
with
this
I'd.
Rather
do
that
and
avoid,
like
you
know
having
that
conversation
in
APR,
where
you
don't
think
SIA
implemented
and
might
change
in
a
major
way
right.
That.