►
From YouTube: SONiC DASH Workgroup Community Meeting Apr 26 2023
Description
PR289 - Still needs volunteer - Dockerfile.saithrift-client has a static tag in the FROM instruction
PR346 - Change SAI submodule URL to opencomputeproject/master - pending
Looks like there were a lot of changes into main/SAI that changed PTF stuff, need some more time to fix integration with latest SAI master.
Change broke automated framework for flexible # of ports - 1755 trying to get this fixed
Chris Sommers:
Annual P4 Workshop on Intel campus
Real world content presented, integration w/real systems & applications
DPU & Switch tech providers attended
SmartNIC IPU/DPU applications focus and P4 enabled devices
SmartNIC Summit June 13 - 15
Clark Lee - Alibaba
Alibaba would like to submit a NAT P4 code PR - send invite for Behavioral Model (HLD) - prep / process
Mukesh: change to bmv2 to check for the changes to the SAI APIs, it would make it easier for ppl to review against the generated APIs to review.
Any changes to SAI APIs to link to OpenCompute SAI APIs?
Breaking changes in the SAI header.
When we generate SAI, we also generate metadata, which verifies that changes are not breaking headers (in SAI repo).
Marian: not really a method to check for this yet
Marian is updating OCP SAI as we move bmv2 along (whenever there is a change, a PR is created) (not updating submodule pointer, but need 1755 fixed first)
We are supposed to have a submodule of SAI which is OCP SAI
A
Yes,
okay,
great,
so
what
I'm
showing
here
is.
These
are
the
meeting
minutes
from
last
week
that
I
sent
out,
and
so
you
know
you
know
we
talked
about
the
cluster
tag
or
the
tagging
again,
so
we
have
a
few
follow-up
items
there
and
then
we
also
held
a
behavioral
model
call
last
week
where
we
talked
about
private
link,
mapping
and
I'm
curious,
you
guys
just
a
housekeeping
note.
A
Do
you
guys,
like
the
format
of
these
notes,
or
do
you
prefer
the
format
where
I
don't
put
as
much
information
up
front
in
the
email
and
I
just
have
links
to
everything
that
you
can
click
and
go
read
so
I
guess
my
question
is:
do
you
like
it
right
in
front
of
your
face
in
the
email,
or
do
you
prefer
to
just
have
something
shorter
with
links?
You
can
go
click
if
you're
interested.
A
A
Know
that
I
don't
know
that
it
works,
though,
because
I
see
formats
of
different
meeting
notes
going
out
and
it's
like
well
maybe
that
works
better.
You
know,
I,
don't
know
so,
just
if
you
guys
want
something
different
or
want
me
to
do
it
differently.
Just
tell
me:
what
do
you
think
I'm
curious
Mukesh?
F
A
All
right,
all
right,
so
I
try
to
just
put
the
pr
what
the
title
is
summarize,
what
I
think
we
said
and
then
the
next
action
item
or
what
we're
gonna
do
and
then
move
on.
That's
what
I
try
to
do
so
and
maybe
a
little
interesting
screenshots
if
we
have
okay,
I'm
gonna,
keep
doing
it
this
way,
then
so!
A
Okay,
cool
because
I
have
seen
other
formats
where
you
just
go,
click
different
links
and
the
email
is
shorter,
so
I'm
open
to
either.
But
if
you
guys
like
it
we'll
do
it
this
way.
So,
okay,
so
we
talked
about
Uncle
tag.
We
talked
about
all
this
stuff
last
week,
and
so
we
did
have
a
couple
of
follow-ups
that
needed
to
happen.
We
needed
to
make
sure
and
Prince
isn't
here
today,
because
he
had
a
conflict,
but
I'm
gonna
follow
up
with
Prince
and
make
sure
that
he
can
add
notes.
A
You
know
in
in
the
hld
about
whether
we're
going
to
separate
into
two
different
rules,
we're
gonna
document
differences
as
part
of
the
work
agent
and
Marianne.
We
were
going
to
ask
him.
Maybe
if
we
can
check
whether
we
can
express
all
in
ACL
mode
I,
don't
know
if
he's
oh,
he
is
on
the
call.
Did
you
have
a
chance
to
look
at
this
Marion.
A
Hey
Marion,
no,
not
yet
not
yet:
okay,
I'll
shoot
it
out.
Separate
I'll,
just
shoot
you
a
separate
email,
not
a
big
deal,
and
then
Prince
also
needs
to
add
this
requirement
for
the
Northbound
tag
and
the
translation
so
I
know
we
have
follow-ups
there
so
anyway,
last
question:
are
you
guys
receiving
these
notes?
It
sounds
like
you
are
like
they're,
hitting
your
inbox:
okay,
because
they're
not
hitting
mine,
somehow
I
got
left
off,
so
I've
started
copying.
A
E
Sure
hi
everyone
I
went
to
the
annual
P4
Workshop,
which
was
held
in
person
on
Intel
campus
in
Santa
Clara,
and
it
was
attended
by
I'm
saying
the
order
of
a
hundred
attendees.
It's
about
half
the
audience
of
the
last
in
person,
one
three
or
four
years
ago.
I'm,
not
sure.
Why
probably
there's
a
bit
of
a
downturn
in
terms
of
travel
budgets
this
time
as
we
all
know,
but
but
it
was
a
good
event.
E
There
was
a
lot
of
Engagement
a
lot
of
good
talks
on
a
variety
of
topics
and
compared
to
some
of
the
earlier
P4
workshops.
This
one
had
a
lot
more.
Let's
see
real
world
content
in
the
very
early
days.
There
was
a
lot
of
academic
talks
about
P4
and
all
kinds
of
Novel
algorithms.
You
could
do
for
congestion
management
things
like
that.
Now
it's
getting
a
lot
more
pragmatic,
there's
a
lot
more
things
about
integrating
with
real
systems
real
applications.
E
Dash
was
brought
up
in
a
number
of
different
contexts.
So
there's
there's
quite
a
bit
of
widespread
awareness
about
daesh
in
different
communities.
The
C4
Community
is
generally
aware
of
it.
The
the
open
programmable
infrastructure
or
OPI
Community
is
quite
aware
of
Dash.
So
there's
more
and
more,
you
know
General
awareness
there
were
dpu
vendors,
there
switch
vendors,
all
kinds
of
you
know:
Professor
types
and
PhD
student
types
presenting
interesting
works,
so
it
was
good.
E
There
was
a
fireside
chat
by
Nick
McEwen
who's
in
charge
of
the
Data
Center,
and
switching
and
Nick
division
of
Intel
I
might
have
gotten
the
specific
specifics
of
that
wrong,
but
he
gave
a
fireside
chat
and
talked
a
lot
about.
You
know
his
vision.
He
did
discuss
quite
candidly
about
the
Tofino
road
map
being
halted
and
and
why
mainly
economic
reasons,
there
wasn't
really
any
discussion
about
what
a
follow-up
made.
E
You
know
what
the
follow-up
position
is
for
people,
programming
and
P4,
but
I
would
say
that
the
discussion,
a
lot
of
the
discussions
focused
on
smart
Nick
or
DP
ipu
applications.
There's
a
lot
of
presentations
on
some
of
the
nitty-gritty
of
Plumbing
different
software.
Stacks
on
P4
enabled
devices
that
were
more
smart
Nick.
Like
so
I
would
say
you
can
definitely
see
a
trend.
Shifting
and
you
know,
Intel
did
quite
a
few
presentations
in
that
regard
so
yeah.
Obviously
it
was
a
good
conference.
A
So
just
so,
everyone
knows
that
smart
nuke
Summit
is
coming.
This
is
June
13th
through
15th
in
San
Jose
and
it
looks
like
I've
received
approval
to
go
so
I
will
be
there.
I
believe
Gerald
will
be
there
and
I
believe
Gerald
will
be
delivering
a
keynote
there,
and
so,
if
any
of
you
come
it'd
be
fun
to
see,
you
I
think
we.
G
A
Couple
more
people
yep
yep,
so
you
know
we
missed
it
last
year,
but
I
think
we
have.
What
is
it
like?
A
keynote
a
table,
an
exhibit
table
a
beer
and
pizza
event.
Something
like
that,
something
that
affects
so,
of
course,
now
I'll
probably
be
tasked
with
putting
it
together.
Maybe
I
can
get
leeway
to
come,
I
don't
know
so
we'll
see
right.
E
A
A
G
G
A
A
Oh
sure
or
I
can
send
a
ping
through
the
pr
it's
okay,
all
right
and
then
it
sounds
like
you
know:
I
didn't
have
a
ton
of
other
PR's.
They
just
look
like
the
normal
PR
list
here
and
which
is
fine.
I
can
give
time
back
this
week,
but
what
I
was
curious
about
so
so
Clark
is
here
from
Alibaba
and
he
had
a
question
about
how
to
prep,
for
he
wants
to
submit
a
PR
for
the
behavioral
model.
A
C
A
So
so
we
usually
like
to
have
some
sort
of
description
of
why
we're
doing
the
pr
I
guess
reshma
I.
Could
let
you
pick
up
on
that,
but
it's
super
helpful
to
have
a
description
of
why
we're
doing
the
pr
and
what
what
purpose
it
would
serve
and
then
Marian
is
there
anything
extra
you
like
to
see
in
behavioral
model
PRS.
B
E
I
I
would
I
would
make
a
recommendation
that
you
first
make
sure
that
the
in
your
local
Dev
branch
in
your
fork,
that
is
passing
all
the
build
and
CI
tests
in
your
GitHub
actions
and
that's
kind
of
a
sort
of
a
first
timer
mistake
that
keeps
happening
that
they
submit
a
PR,
that's
failing
where
it
could
be
that
it
can
be
tested
in
your
own
Fork
first,
and
what
you
want
to
do
is
in
your
local
Fork,
enable
CI
actions
in
your
GitHub
pipeline.
Then
you'll
see
if
they
build
and
pass
properly.
E
D
Yeah
I
would
also
like
to
just
add
to
what
the
it's,
not
just
only
enough
to
have
the
existing
test
cases
passing,
but
also,
if
you
know,
whatever
the
changes
that
you're
adding
make
sure
that
you
know.
Please
add
the
new
test
cases
tool
to
make
sure
that
it
is
accompanying
all
the
unit
test
cases
to
validate
a
verify,
and
you
know
test
the
additions
that
you
are
suggesting
right.
So
I
think
this.
E
C
C
E
G
F
F
D
So
no
I
was
saying
it's
a
great
idea
and
then
we
I'm
not
sure
in
our
GitHub
actions.
Are
we
running
besides
the
the
test
cases
to
ensure
that
okay,
you
know
our
pipeline
is
working
and
then
you
know
we
have
not
broken
anything
are.
Are
we
also
running
the
test
case
is
to
to
generate
the
API
and
ensuring
that
if
there
are
any
changes,
are
we
catching
them
or
or
flagging
them,
or
whatever
be
the
case
right
to
ensure
that
okay,
you
know
the
sanctity
is
preserved
right.
E
E
That's
actually
what
takes
the
most
time
in
the
make
file
is
not
regenerating
the
API
that's
pretty
fast,
but
rebuilding
the
sci
Thrift
server,
client
libraries
based
on
the
side
headers
du
jour,
so
it
is
rebuilt,
but
the
second
part
of
the
question:
do
we
look
for
changes
in
the
side,
header
or
breaking
changes?
We
don't
and
that's
a
real
good.
It's
a
real
good
point
and
I.
Don't
I,
don't
know
what
the
the
right
way
to
go
forward
is
on
that.
E
You
know,
it'd
be
probably
quite
a
bit
of
effort
for
someone
to
implement.
You
know,
first
of
all,
decide
what
to
do.
You
know.
Do
we
just
report
the
changes
as
a
report,
or
do
we
somehow
look
for
breaking
changes,
I
think
that
might
be
done
in
the
ocp
PSI
repo
I'm,
not
quite
sure,
but
but
isn't
it
true?
I
think.
E
E
B
D
But
before
before,
before
I,
you
know
just
took
me
one
one
point:
Chris,
isn't
it
true
that
if
we
are
doing
this
thing
as
part
of
GitHub
actions
and
we
generate
and
I,
you
know
the
apis
and
we
build
the
PSI
and
we
test
it
out?
If
there
was
you
know
if
if
this
I
apis
had
broken
or
something
else
had
happened,
our
test
would
fail.
Wouldn't
it.
E
D
E
We
sort
of
keep
everything
working,
but
some
other
person
who's
spent
time
on
their
own
Fork,
even
private
Fork,
creating
their
vendor
specific
platform
is
using
a
previous
header
definition
that
our
tests
wouldn't
check
that
we
broke
something
for
somebody
else
right
and
that
that's
where
we
need
maybe
a
committed
version
of
the
headers
in
this
in
the
ocp
sci
repo.
You
know
that
that's
the
kind
of
situation
that
might
arise
someday.
B
B
I
was
thinking
purely
about
site
apis.
You
can
rename
something,
but
that
won't
be
backwards
compatible.
If
someone
has
implemented
them,
that's,
but
we
do
not
include
vendor
implementations
of
Dash
site
and
to
our
into
that
project
anyway.
So
I
I,
don't
know
how
to
solve
for
that
and
to
check
for
backward
compatibility.
E
Yeah
I
guess
we're
we're
going
to
be
getting
to
a
point
where
we're
going
to
need
to
commit
our
Dash
apis.
You
know,
update
a
Psy
repo
version
of
the
dash
apis
right
and
that
will
sort
of
be
a
stake
in
the
ground,
but
yeah.
The
good
news
from
all
this
is
at
least
we're
always
self
checking
this
repo
against
all
the
test
cases.
Every
time
with
fresh
builds.
B
F
Options
my
question
was
more
towards,
like
the
manual
verification
part.
So
currently,
if
you
make
a
change
in
the
BMV
too,
it's
very
hard
to
know
what
change
it
cost
to
the
site
API
unless
you
actually
build
build
the
build
the
code
and
look
at
this
IPS
for
yourself,
so
every
reviewer
would
have
to
do
that.
F
I
was
wondering
if
we
can
put
this
I
if
we
generated
once
also
in
the
report
it
shows
up
in
the
day,
then
it
would
make
it
much
easier
for
people
to
review
it
and
stuff
like
that.
I
yeah
I'm
not
sure
how
to
do
that,
how
to
get
the
generated
apis
into
the
repo
and
make
sure
that
it
also
gets
updated
every
time.
The
pr.
E
G
F
E
F
E
F
They
might
submit
it
without
updating
the
generated
code
so
which
then
it
might
still
not
show
the
dip
so
but
yeah
yeah,
it's
not
automatic,
is
what
I'm
going
to
say,
but
yeah
at
that
point
you
could
say:
okay,
it
looks
like
there
is
a
change
in
this
IAP
and
you
have
not
updated
the
generated
code
and
ask
them
to
push
again
over,
but
yeah
I
thought
it
might
be
easier
to
review.
If.
G
G
G
B
B
B
D
D
B
Why
I
think
that
committing
Auto
generated
changes
to
our
Fork
is
not
a
good
idea.
What
we
can
do
for
visibility
is
to
generate
a
div
and
show
it
and
automatically
in
the
pull
request,
but
committing
them
will
lead
to
a
different,
commit,
IDs
and
I.
Don't
know
how
we
can
I,
don't
think
there
is
a
way
to
solve
that
every
time,
because
we
will
need
to
sync
it
with
OCD
side
and
it
will
happen
for
every
such
Dash
comment.