Binder interface for wpa_supplicant?

[Christopher Wiley]

On Tue, Jan 19, 2016 at 6:10 AM Johannes Berg <johannes at sipsolutions.net> wrote:
>
> On Fri, 2016-01-15 at 23:33 +0200, Jouni Malinen wrote:
> >
> > One more challenge with the D-Bus interface has been that it is still
> > far from the capabilities of the ctrl_iface implementation. The
> > latter happens to be the one that I use as the first target for
> > everything new and in practice, it will be superset of all control
> > interface in what kind of functionality gets exposed. Having active
> > contributors to help maintain any new interface would obviously be
> > important to avoid issues with an interface not being sufficient for
> > something.
>
> This is kinda why I've always thought it'd be good to generate all the
> APIs from a single base specification. That would also help
> tremendously for consistency, and even implementation-wise it'd be far
> less "code".
>
> I don't really have a good feeling for how feasible this is yet, but it
> seems that the biggest issue would likely be around the existing
> ctrl_iface rather than any new additions.
>
> Christopher, are you or anyone else involved going to be at netdev next
> month, perhaps we could discuss this there?
>
> johannes

Agreed that for any non-trivial amount of logic, sharing it among as many
interfaces as possible becomes desirable.  I'm less sure how much generation
should be going on.  Our internal prototypes of this interface (in
Binder) generate
the serialization boilerplate if that's what you have in mind.
Unfortunately, differences
in type systems and semantics between DBus and Binder make a common interface
specification kind of its own project.

On the other hand, it would be nice if, past the IPC interface, we
were able to share
the operations against common supplicant objects.  I guess in my mind
that means that DBus
and Binder would both notice a new request is available on their
control FD, deserialize
it as appropriate, and then both call into some common backend,
sending back responses
as appropriate.

Our tentative plan is to prototype this a little internally, then send
up some RFCs to get
your thoughts.  Unfortunately, I can't be at the netdev conference,
although it would be
nice to travel.

wiley