►
From YouTube: SONiC DASH Workgroup Community Meeting Aug 16 2023
Description
Removal of ACL tags from P4
Kamil uplifting libsai, longterm goals to make SONiC instance that includes DASH behavioral model
Fix APIs and reconcile small shortcomings
Chris reviewing PRs, checking CI pipeline
If anyone has 'Stale' branches, please tidy them or bring to Kristina or Chris' attention
A
A
These
are
the
basically
the
notes
that
I
sent
out
from
last
time
so
up
here,
I'm
doing
a
new
thing
where
I'm
putting
basically
our
project
items
that
we
looked
at
last
time.
More
of
in
a
format
like
this,
where
you
can,
we
can
click
on
them.
Last
time
we
continued
the
conversation
about
the
Apple
tags
and
we
all
agreed
that
maybe
we
wouldn't
need
them.
So
we
talked
about.
B
A
There
and
I
I
ended
up
opening
an
issue
to
go
ahead
and
remove
those
from
P4,
and
then
we
Chris
mentioned
that
Camille's
doing
some
work
in
the
background
for
us
on
the
repo.
So
that's
Camille
good
Nick.
He
works
over
on
my
end
and
so
I
talked
to
him.
Yesterday
I
thought
maybe
he
could
come
explain
some
of
his
changes
or
the
work
he's
doing
he's
in
a
different
time
zones.
A
So
I
don't
know
if
he'll
show
up
today,
but
maybe
next
week,
and
so
we
could
look
at
his
closed
or
merged
items
there.
So
that's
what
happened.
Oh
and
then
of
course.
Last
week
we
had
a
couple
of
new
Arista
people
joined
the
team
and
Javon
and
Max.
Did
you
receive
the
meeting
notes
from
last
week
that
I
sent
out.
A
Cool
cool,
great
and
then
over
here
on
the
project
board.
I
noticed
that
this
week,
Vincent
mentioned
that
these
two
issues
are
related.
A
The
number
309
which
is
from
before
the
update,
scaling,
underlay
routing
service
tunnel,
private
link,
encoding
and
related
that
to
psi
attributes
for
Apple
rules
should
include
list
of
IP
prefixes,
because
Camille
was
working
on
this.
So
I
put
these
into
the
list
here,
Vincent
and
related
them
to
each
other,
and
so
we'll
see
if
we
can
get
work
done
on
those.
When
prints
comes
back,
yeah
no
problem
and.
C
C
Wanda
I'm
just
going
back
to
what
Camille's
working
on
he
did
follow
an
issue:
4
15,
where
we
could
use
Marion's
review.
Oh
okay
generation.
That's
been
trying
to
Ping
Mario
a
couple
times
on
that.
So
if.
D
C
A
15
yeah
since
since
I
just
explained
what
I
did
there
four
one
five
so
generated
in
lib
size,
so
I
won't
want
to
read
it
to
you,
but.
C
A
Okay,
cool
thanks
thanks
Chris
yeah
yeah,
so
I'll
go
ahead
and
put
this
into
our
list
and
I.
Think
I'll
start
filtering
out
all
this
work
that
we've
done
here
so
I'll
do
that
we
won't
have
an
update
today
from
rushman
unless
someone
else
on
the
call
from
Intel
might
be
able
to
tell
us
about
408
the
P4
dpdk,
possibly.
A
That
would
maybe
be
do
we.
A
Speeding,
slowly,
okay,
hey
Christina
I,
don't
have
the
answer,
but
I'll
flag
it
to
Russia
and
other
stakeholders.
E
A
Until
next
week
she
said,
thank
you,
chinos
yeah,
okay,
so
it's
kumaresh
on
the
coast,
I,
don't
see
him.
He
was
gonna.
Take
a
look
at
this
first
one
here.
I
can
check
with
him
later
too
so
he's
in
Azure
boot
camp
this
week.
So
I'll
take
a
look
at
this
one
offline,
the
meter
policy
scale-
and
we
also
have
a
this
one
here.
Did
this
Marion
did
this
one?
A
F
Haven't
done
anything
that
yeah.
A
Oh
okay,
cool
just
make
sure
I
don't
need
to
do
anything.
Sorry.
D
On
the
P4,
I
call
tag
I.
Suppose
Marion
is
a
PR
going
to
cover
the
update
of
the
documentation,
because
I
think
there
was
a
chunk
there
about
ACL
tags.
F
Yeah,
so
the
documentation
has
two
things.
First,
one
is
the
MDB
schema
which
will
stay,
as
is
because
tax
will
still
be
passed
into
site.
I
will
just
remove
the
part
that
pushes
it
to
site.
A
Okay
and
then
so
I'm
trying
I
I
think
this
one
here
the
clarifying
scale
in
route
so
I
feel
like
we
did.
Do
the
update
to
the
scale
tables
and
everything
a
few
weeks
ago,
but
I'm
trying
to
understand-
and
maybe
it's
me
I
was
trying
to.
Maybe
it's
me.
Why
is
it
important
to
know
whether
the
routes
are
inbound
or
outbound
and
knowing
the
scale
of
those,
maybe
if
you
guys
could
help
me
or
help
me
understand
that
maybe
I
could
take
it
back
to
the
team
internally.
A
A
A
A
A
A
C
C
They
were
supposed
to
look
at
it,
and
intel
was
going
to
look
at
it,
but
no,
no
activity
that
I'm
aware
of
okay.
A
A
B
A
B
C
Outside
set
of
eyes,
if
they
wish
just
put
I,
put
some
suggestions
there,
no
one's
obligated,
okay,
but
you
know
it
lingers
around
for
another
week
without
any
other
review.
Yeah,
maybe
we'll
want
to
advance
it.
It's.
C
B
A
Tab
here
we
actually,
this
is
mine
from
last
week,
so
we
have
nothing
new
from
last
week.
So
let
me
see
if
we
have
Camille
here
there
is
Camille
hi
Camille.
F
A
Hi
so
Camille
hi,
I'm
Christina.
This
is
everyone
in
the
dash
community.
So
I
have
everyone
here
from
broadcom
AMD,
Intel,
keysight,
peel,
Vision,
Nvidia,
gosh,
I'm,
sorry,
if
I'm
forgetting
anyone
dream,
big
semiconductor,
Arista,
Etc
and
so
a
Camille.
We
go
through
the
repo
once
a
week
and
the
community
call
here
and
go
over
pull
requests
and
issues
and
Q
a
and
things
like
that.
So
everybody,
this
is
Camille,
he's
a
principal
engineer
over
in
Prince
Camilla
I'll.
Let
you
introduce
yourself.
B
Hi
everyone,
I'm,
Camille
and
I
will
I
work
in
Microsoft
for
11
years
right
now
and
I,
don't
know
seven
or
eight,
maybe
in
in
Sonic
and
lately
I
was
given
a
task
to
to
make
some
changes
in
the
dash
repo.
So
our
goal
is
to
make
a
some
like
something
similar
that
we
had
in
the
side.
Ideas
like
the
wheelchair
Suite.
B
So
we
want
to
have
ability
to
connect
the
bmv2
with
a
P4
model
and
to
the
CD
and
be
able
to
you
know
to
make
make
use
of
the
dash
apis
from
The,
Arc
agent.
I.
Think
that
I
think
that's
our
goal.
There
is
also
another
possibility
that
we
that
we
are
deciding,
because
the
current
Dash
repo
is
just
supporting
only
the
dash
dash
part
of
the
PSI
API.
So
there
is
no
other
underlying
functionality
for
other
apis
and
although
I
saw
that
someone
already
posted
some.
B
Routing
apis,
but
I
think
our
goal
would
be
to
try
to
connect
the
virtual
switch
that
we
have
inside
of
this
report
that
have
all
the
abilities
for
the
regular
apis
and
with
the
DMV
tool.
Just
for
that
dash
part
and
have
that,
like
under
one
single
layer
of
the
PSI
apis.
A
Great
thanks,
Camille
thanks
yeah
yeah
appreciate
that
so
did
anyone
have
a
a
question
or
comment
for
Camille
or
Camille?
Did
you
want
to
go
through
the
I?
We
noticed
here
that
you've
been
doing
a
lot
of
issue
work
here
in
the
repo.
B
Yeah
so
I've
been
doing
last
week
or
last
two
weeks,
refactoring
of
the
site
in
the
dash
report,
so
I.
Basically
the
fact
of
everything
that
almost
everything
that
was
there
I
found
some
bugs
and
some
issues
which
I
already
opened
in
the
in
the
dash
issues.
I
saw
that
you
already
mentioned
one
of
my
one
of
my
bags
and
I
refactored.
The
codes
of
this
to
improve
the
code.
B
Quality
and
I
moved
that
I
call
the
all
the
stabs
API
to
site
API
to
its
respective
files
and
I
created
the
Class,
CC
plus
plus
class
for
the
entire
Dash
implementation
and
added
some
proper
logging
and
so
like
improve.
Overall
the
generic
code
of
the
of
the
sign
of
the
library
yeah.
A
Thank
you.
So
while
we
have
him,
did
anyone
have
a
question
or
comment
for
Camille
in
the
call
today.
A
Chris
or
anyone.
A
Okay,
all
right
so
good
to
know:
okay,
okay,
so
great!
Thank
you
so
much
so
guys,
that's
really
all
I
had
for
today.
I
can
open
it
up
to
everyone
else.
Did
anyone
else
have
something
they
wanted
to
look
at,
discuss
or
issue
or
PR,
or
something
I
need
to
chase
down.
B
Maybe
I
have
just
one
question
here
sure,
because
what
I
noticed
is
that
those
Dash
feathers
in
the
industrial,
those
are
like
auto-generated
from
the
P4
model
and
those
are
actually.
There
are
some
changes
which
are
different
from
the
actual
Fighters
that
are
in
the
site
in
the
side
level.
So
I
think
when
I
spoke
with
Gohan,
he
told
me
that
we
should
probably
treat
the
daso
repo
as
a
grant
Truth
for
the
dash
headers.
B
So
but
I
also
noticed
that
there
are
some
there's
some
differences
and
probably
limitations
of
the
BMV
tool.
Simulator
I
think
one
of
the
limitations
is
that
it
doesn't
accept
like
the
least
of
the
attributes
yeah,
something
like
that.
A
Okay,
well,
I
think
this
is
a
question
for
Marion
for
sure
so
Martin.
Maybe
could
you
help
out
a
little
bit.
F
Yeah
there
is
this
limitation
for
now.
As
far
as
I
know,
there
was
a
work
done
to
just
partially
support
that
and
either
expand
a
rule
or
something
like
a
similar
to
that
or
support.
Only
one
value
out
of
the
list.
I,
don't
recall
who
was
doing
that,
but
it
was
someone
from
Microsoft,
so
I
can
I
can
check
that
and
point
to
how
this
is
being
overcome
in
the
BMV
too,
but
it
is
kind
of
like
partially
supported
yeah.
E
F
Oh
that's
what
our
agent
is
doing
is
expanding
tags
that
come
from
from
the
controller,
but
it
doesn't
expand
lists
so
or
Cajun
basically
still
generates
one
rule
with
a
list
of
prefixes
that
came
from
text,
but
expansion
of
a
list
is
done
in
the
style
layer.
I
can
look
for
the
plural
request
it.
It
should
have
been
emerged
long
ago
to
support
that
in
PSI
implementation
for
bmv2.
D
It's
kind
of
related
to
the
issue
that
I
brought
again
last
time.
It's
the
bmv2
is
doesn't
support
everything
that
needs
to
sorry.
We
generate
the
apis
from
the
bmv2,
except
that
there
are
a
few
caveats
on
that
right
now
and
that's
the
problem
we've
listed
in
list
of
prefix
and
all
that
jazz
that
we
find
in
ACL.
A
E
Both
all
software,
such
as
head,
the
Gap
that
they
do
not
support,
lists
directly
and
work
around,
for
that
is
the
same
for
all
of
them,
which
is
have
some
layer
of
Control
software
above
the
software
switch,
expand
the
lists
out
into
digital
career
prefix
entries,
that's
the
same
workaround
for
all
of
them.
I
would
recommend
a
reference
model.
D
D
D
E
E
D
D
F
D
I
know
it's
a
problem:
we
need
to
fix
it.
That's
all
I'm,
saying
and
you're
right,
I'm
actually
enabling
that
that
the
compile
flag
so
that
I
can
get
the
proper
API
generated.
But
Camille
is
for
the
oxygen
thing
D
and
all
that
stuff.
You
know
it's
you're
eating
product
ties.
F
Now,
as
I
was
saying,
as
I
was
fixing
that
I
don't
know
if
he
actually
got
to
that
issue
or
not,
but
we
know
that
we
have
two
versions
of
apis
generated
with
the
types
that
support
sorry
with
the
types
that
correctly
reflect
intended,
ACLS
and
with
the
types
that
can
run
on
the
mv2.
F
The
goal
was
to
marry
those
two
together
and
have
apis
that
are,
let's
say,
official
and
have
bmv2
on
running
on
on
the
types
that
it
can
support
and
just
do
a
simple
translation
with
some
limitations
so
that
we
will
have
some
simplified
version
of
ACLS
where
the
list
will
be
just
one
value
or
something
like
that.
D
D
A
I'll,
try
and
we'll
try
and
find
that
so
I
made
a
comment
into
into
the
issue
there.
Vincent.
A
Yeah
yeah
yeah
yeah,
because
I
I
remember
I
was
emailing
him
about
one,
a
sorry,
I
don't
want
to
belabor.
The
point
here:
do
a
quick,
quick
search
from
the.
A
Any
work
to
do
great
and
so
Camille
did
you
have
anything
else
before
we'll
have
to
let
you
go.