►
From YouTube: SONiC DASH Workgroup Community Meeting Aug 2 2023
Description
Merged 2 PRs PR403 and PR404, and opened a couple more for test case creation and P4 DPDK exploration. @Chris Sommers and @Kristina Moore performed a ‘stale branch’ cleanup of the Repo. Next week NVidia would like to revisit the ACL Tags conversation.
A
78Th
meeting
and
the
last
time
we
were
here
together,
I
hadn't
sent
the
notes.
Yet
we
talked
about
pr404
talking
about
the
or
might
have
been
402
talking
about
the
throwing
out
that
underlay
routing
functionality
here
in
this
information
here,
and
so
we
all
talked
about
that
and
left
some
comments
in
the
pr
and
left
it
open
for
people
to
talk
about,
and
then
we
also
closed
a
couple
of
PR's
and
we
welcomed
kumaresh
and
some
different
people
and
then
over
here.
A
This
is
the
the
board
that
we've
been
working
on
and
we
were.
We
changed
a
cup
oops.
We
changed
a
couple
of
these
items
to
closed,
so
these
ones
here,
Mark
just
done
so.
Marion
had
fixed
this
bmv2
pipeline
source
code
change,
and
then
we
had
a
couple
of
more
items
here
that
were
done,
and
then
we
have
these
items
that
are
still
to
do
this:
one's
the
oldest
one
number
289,
and
this
one
came
from
Prince.
This
generate
PSI
for
Apple
tags.
A
A
Okay,
great
thank
you
and
then
I
can
pop
it
in
here.
That's
awesome
thanks
for
letting
me
know,
and
then
we
did
open
like
an
actual
issue
for
this
p40bk
on
the
horizon,
because
we
talked
about
it
last
week
and
I
think
Rasmus
said
she
would
follow
up
on
it.
So
so
we
have
that
and
the
rest
of
those
are
done,
but
I
was
curious.
I
ended
up
joining
the
P4
call
this
week,
and
so
I
was
curious.
C
The
the
architecture
work
group,
the
Monday
yeah.
Oh,
we
continue
to
I'll
just
we
continue
to
emphasize
minutia
in
the
cross
vendors
and
their
P4
implementations
on
how
to
do
things
like
checksums
and
other
stuff,
so
yeah
looking
for
common
ways
to
that,
we
think
you
know
we
can
agree
on
so.
A
No
I
didn't
I,
didn't
hear
anything
Dash
specific,
but
it
was
a
really
good
call
so
yeah
and
then
so
this
week,
I
guess
I'll
put
it
out
there.
Did
anyone
want
to
bring
up
specific
PRS
or
issues
they
wanted
to
talk
about
this
week?.
D
A
A
The
one
where
we're
correcting
the
seconds
and
the
numbers
and
things
like
that
and
the
ones
where
I
did
three
of
you,
too
yeah.
E
D
A
totally
new
group
of
apis
for
next
hope,
ID
as
well
as
nexop
ID
group,
will
have
to
be
implemented
as
part
of
a
subsequent
PR.
So
current
PR
just
addresses
route
next
hope,
ID
apis
and
for
the
holy
CMP
you
will
have
to
be
a
separate
PR.
So
as
of
now,
we
are
adding
the
behavior
for
out
and
the
behavior
for
hair
pinning
we
are
leaving
it
unchanged
from
the
current
implementation
till
the
CMP
gets
implemented,
yeah
in
a
different
job.
D
Yeah,
so
the
fourth
we
Implement,
Port
ID
could
be
used
as
next
hop
ID
in
the
current
implementation.
Of
course,
future
implementation
can
now
support
for
interfaces
and
other
things,
but
to
have
support
for
interfaces.
First,
an
interface
needs
to
be
added
in
bmv2
and
so
on.
So
each
extra
apis
that
we
add
have
to
be
their
own
PR
and
add
additional
functionality.
B
D
Now
so,
if
you
want
a
route,
you
add
a
route.
If
you
don't
other
route,
it's
today's
Behavior
with
anyway
today
is
happening,
no
matter
what
okay.
B
A
So
does
that
answer
these
two
questions
here
we
were
asking
if
we
could
have
some
sort
of
global
capacity
to
bypass
underlay
routing
feature
and
if
you
do
nothing
with
the
next
step
or
Chris
did
answer
this.
Okay,
yeah.
A
There
will
be
additional
errors
or
what
did
you
say?
Next
Top
group.
D
Yes,
ndcmp
and
I
looked
at
those
apis,
they're
already
specified
in
swss
light,
and
it
has
their
ecmp
type
and
that
will
address
I
would
believe.
The
questions
which
are
asked
in
the
last
meeting.
A
A
Look
at
us
getting
work
done,
okay
and
thanks
so
much
for
those
comments.
That's
great
anyone
else
have
something
they
want
to
talk
about
here.
E
I'd
like
to
make
a
comment:
if
you
look
in
issues-
okay,
it's
just
a
general
comment
to
keep
people
informed
if
they've
been
paying
attention,
but
fellow
named
Camille,
kudne
he's
a
Microsoft
employee,
very
active
inside
I
think
he's
a
maintainer
inside.
It
looks
like
he's
going
through
and
doing
a
big
scrubbing
operation
on
okay
on
the
dash
PSI
implementation,
which
is
a
good
thing
he's
trying
to
make
a
compatible
to
compile
it
into
sync
D,
for
possibly
a
virtual
switch
have
an
emperor.
E
We
talked
about
probably
a
year
and
a
half
or
two
years
ago.
Could
this
ever
be
done?
So
it
looks
like
I'm
trying
to
align
it
with
a
standard,
PSI
implementations.
So
it's
going
to
require
a
bit
of
work
on
various
people's
part
he's
doing
the
most
of
it,
but
he's
gonna
be
submitting
issues
and
poll
requests
to
to
align
the
website
and
get
it
out
of
its
little
sandbox,
which
it's
in
and
make
it
a
little
more
conformant.
E
E
E
This
one's
a
hard
one
I,
don't
know
how
to
answer
it,
but
he's
got
some
other
ones
which
are
a
little
easier.
F
Previous
time
when
we
had
generated
PSI
apis
right,
Dash
apis
from
the
bmv2
and
we
had
found
some
changes
and
all
those
things
were
fixed,
we
are
not
regenerating
at
this
time.
Our
is
coming
regenerating.
E
I
think
he's
trying
to
take
the
regenerated,
PSI
headers
from
latest
commits
and
manually
copy
them
into
the
psy
repo
Dash
version
to
bring
it
up.
So
that's
just
one
of
the
things
he's
working
on
and
it
looks
like
zigon
is,
is
Angel
based
on
that,
although
Intel
might
want
to
keep
an
eye
on
what's
going
on
too,
because
you
you
guys
and
Nvidia
have
been
the
most
active
in
you
know,
curating.
E
This
aspect,
I
won't
have
to
say
on
on
that,
but
when
it
comes
to
things
like
the
implementation
of
the
bnb2
libs
I
I'm,
trying
to
assist
where
I
can.
A
Yeah,
so
it
looks
like
you
asked
Marion
to
help
out
a
bit
on
this,
so
I'm,
just
going
on
my
call
in
Russia,
okay,
great
okay
can.
E
E
Camille
might
want
he
might
Merit
collaborator
access
to
the
dash,
so
he
can
ask
for
reviewers.
He
did.
He
hasn't
been
able
to
do
that.
Yet
I,
don't
think
I
think
he
reached
out
to
Gohan
and
said.
Could
you
add
me
but
I
think
it'd
be
easier?
If
you
just
did
it.
A
G
Chris,
you
mentioned
you
say
something
like
it's
a
copy
to.
Is
it
somewhat
officializing
the
generated
experiment,
experimental
headers,
to
push
them
up.
E
G
Okay,
that's
that's
fine.
I'm
gonna
go
back
to
an
issue
that
we
have,
which
I
think
is
related
to
the
HCL
rule
definition,
which
is
incorrect
out
of
the
box.
In
my
opinion,
when
bmv2
is
compiled
because
it
doesn't
use
prefixes
in
the
ECR
rules,
and
you
should.
E
Yeah
and
and
I
I
just
need
to
correct
myself.
It
looks
like
this
okay
right
there
yeah
it's
actually
merging
into
open
compute
Master,
oh
so
if
I
was
wrong
about
it.
Being
that
he's
going
from
the
dash
Branch
into
master
I,
don't
know,
I
think
that's
a
Psy
maintainer
kind
of
decision
which
I
have
a
lot
to
say
about,
but
it
makes
your
comment
more
timely.
G
E
E
Think
I
I,
don't
disagree
with
you.
I
don't
have
a
strong
opinion
on
that,
but
it
does
seem
like
it
needs
to
be
addressed
and
I
don't
recall
it.
The
reason
it's
not
addressed
is
because
there's
a
technical
disagreement
or
it's
just
work.
Someone
has
to
do
to
to
correct
it.
E
I
would
say
it's.
The
definition
is
wrong.
Folks,
like
Prince,
would
be
very
interested
in
that
right.
G
G
Don't
say:
I'm,
not
nothing.
There
are
changes
that
Prince
is
going
to
propose
for
the
new
ACL
tag
stuff,
so
the
ACL
tire
is
going
to
move
still
a
little
bit
on
the
ACL
tag.
Rule
this
hill,
sorry
DSL
rule
hasn't
changed
and
we
definitely
need
to
rewire
that
so
I'll
add
that
to
the
pr
so.
B
Vincent,
like
I,
have
posted
the
actual
tag
here
in
the
chat.
You
can
check
that.
A
A
Aside
for
Apple
takes,
someone
was
going
to
post
that
for
me
over
in
the
other
repo.
A
There
it
is
okay,
all
right
all
right.
Thank
you,
so
I'll
make
sure
and
add
that,
okay,
so
back
to
what
we
were
talking
thanks
Vincent
for
that
that's
awesome.
So
basically
it
looks
like
Camille's
doing
this
work
and
I
messaged
him.
While
we
were
just
talking
now
and
I,
let
him
know
that
I
added
him
to
the
repo,
so
it
would
be
easier
for
him
to
work
with
us.
E
Good
high
power
resource
to
help
us
out
yeah.
E
G
So
another
goal
actually
has
to
say
without
Chris
would
be
to
run
test
cases
like
since
we're
using
the
side
Challenger
the
side,
changer.
Okay,
either
interface
with
the
sci3
server,
which
is
the
current
model
or
the
Cindy
which
I've
played
with
and
it
works
well
yeah,
so
maybe
I
don't
have
a
goal
is
to
get
the
thing.
The
integration
tested
with
the
side
Challenger
and
some
test
cases.
E
E
A
A
Okay,
awesome,
okay
and
it
looks
like
one
of
these
okay,
so
so
FYI
to
us.
These
are
being
worked
on.
Okay,
thanks
Chris,
that's
awesome.
Was
there
anyone
else
who
had
something
to
bring
up
today?
A
H
Hi
Chris
yeah
nice
meeting
you
I'm
Naveed
I,
go
by
Navid
from
Arista
networks.
We
met
at
the
dpu
conference
this
year,
yeah,
okay,.
A
Nice
to
see
you
nice
to
see
you,
so
this
is
kind
of
the
Cadence
of
what
we
do
we
get
together.
We
socialize
we
talk
about
PR's
issues,
forward,
movement
in
the
repo
and
and
the
repo
here
Naveen
is.
We
actually
have
different
sections
with
documentation
and
probably
what's
more
interesting
to
the
developers
is
the
pipeline
area
and
that's
where
the
bmv2
and
the
side
pieces
are
so
just
FYI,
but
there
there's
a
lot
of
documentation.
A
That's
the
high
level
design
here
and
it
describes
a
lot
more
things,
and
then
we
have
this
one's
more
authored
by
me
and
Prince
and
whoever's
helping
and-
and
these
are
the
older
documents
that
are
actually
still
pretty
useful,
with
a
lot
of
pictures
and
things
like
that
with
packet,
Transformations
and
pipelines,
and
Enis
and
vnis
and
encapsulation,
and
examples,
and
things
like
that.
So
that's
what
that
is.
So,
okay,
thanks.
Thanks
for
coming
on
camera,
okay,
yeah.
H
A
You
yeah
anyone
else
have
anything
today.
E
F
F
E
Far
so
far,
I
just
answered
I
just
provided
him
like
a
synopsis
I
just
wondering
if
there's
any
progress,
something
we're
still
going
to
work
on
I'm
just
going
to
review
with
the
team.
Well,
we'll
see.
E
Okay,
yeah
there's
no
urgency.
It's
it's
I'm
kind
of
trying
to
help
people
who
have
like
vendor
interest.
It
works
fine
in
the
dash
repo,
as
is
but
I,
can
see
how
this
isn't
a
worthwhile
thing.
I
wouldn't
be
surprised
if
Camille
stumbles
into
it
too.
He's
turned
it
into
sync
D
and
I'm,
not
sure.
A
All
right,
well,
I'm
gonna,
stop
the
recording,
then,
unless
someone
else
has
anything
and.
A
I
already
did
it
I
think
in
the
call?
Oh
well,
we
were
sitting
there
talking
I,
believe
I
did
Chris
ping
it
to
me
and
I'll
make
sure
I
did
but
I
think
I
did.